Pennsylvania

Economic Association

Annual Conference

May 30-June 1, 2013

The University of Scranton

Scranton, PA

Proceedings of the Pennsylvania Economic Association

PENNSYLVANIA

ECONOMIC

ASSOCIATION

ANNUAL CONFERENCE

May 30-June 1, 2013

The University of Scranton

Scranton, Pennsylvania

Visit the Pennsylvania Economic Association Home Page at

http://aux.edinboro.edu/pea/index.html Acknowledgements: Thanks to the Federal Reserve Bank of New York Cleveland and Dr.

Michael Mensah, Dean, Kania School of Management, The University of Scranton, and Fr.

Kevin P. Quinn, S.J., President, The University of Scranton for their support of this

conference.

Proceedings of the Pennsylvania Economic Association

Pennsylvania Economic Association: 2012-2013

Executive Board President: Tracy Miller, Grove City College

President-Designate: William Bellinger, Dickinson College

Vice President, Program and Proceedings: Sandra Trejos, Clarion University of Pennsylvania

Vice President, Publicity: Deborah Gougeon, The University of Scranton

Vice President, Membership: Vacant

Secretary: Stephanie Brewer, Indiana University of Pennsylvania

Treasurer: Steven Andelin, Penn State - Schuylkill

Co-Editors, Pennsylvania Economic Review: Thomas Tolin & Orhan Kara, West Chester University

Webmaster: Michael Hannan, Edinboro University of Pennsylvania

Immediate Past President: Orhan Kara, West Chester University of Pennsylvania

Board of Directors Ron Baker, Millersville University

Charles Bennett, Gannon University

Kosin Isariyawongse, Edinboro University

John McCollough, Penn State - Lehigh Valley

Brian O'Roark, Robert Morris University

Mark Schweitzer, Federal Reserve Bank of Cleveland

Yaya Sissoko, Indiana University of Pennsylvania

Luke Tilley, Federal Reserve Bank of Philadelphia

Ralph Ancil, Geneva College

Soma Ghosh, Albright College

Jolien Helsel, Youngstown State University

Ex-Officio Directors Abdul Pathan, Pennsylvania College of Technology

Andrew Economopoulos, Ursinus College

Andrew Hill, Federal Reserve Bank of Philadelphia

Bijou Yang-Lester, Drexel University

Brian O’Roark, Robert Morris University

Brian Sloboda, US Postal Service

Daniel Y. Lee, Shippensburg University

David Culp, Slippery Rock University

David Yerger, Indiana University of Pennsylvania

Donald Dale, Muhlenberg College

Donna Kish-Goodling, Muhlenberg College

Elizabeth Hill, Penn State-Mont Alto

Gayle Morris, Edinboro University

Gerald Baumgardner, Pennsylvania College of

Technology

Heather O'Neill, Ursinus College

Ioannis N. Kallianiotis, University of Scranton

Jacquelynne McLellan, Frostburg State University

James Dunn, Edinboro University

James J. Jozefowicz, Indiana University of

Pennsylvania

John A. Sinisi, Penn State University-Schuylkill

Johnnie B. Linn III, Concord University

Joseph Eisenhauer, University of Detroit Mercy

Kenneth Smith, Millersville University of

Pennsylvania

Lawrence Moore, Potomac State College of West

Virginia University

Lynn Smith, Clarion University

Margarita M. Rose, King's College

Mark Eschenfelder, Robert Morris University

Mehdi Hojjat, Neumann College

Natalie D. Reaves, Rowan University

Osman Suliman, Millersville University

Patrick Litzinger, Robert Morris College

Paul Woodburne, Clarion University

Robert D'Intino, Rowan University

Robert Liebler, King's College

Rocky Jui-Chi Huang, Pennsylvania State University -

Berks Campus

Roger White, Whittier College

Stanley G. Long, University of Pittsburgh/Johnstown

Tahereh Hojjat, DeSales University

Thomas O. Armstrong, Pennsylvania Department of

Community & Economic Development

William F. Railing, Gettysburg College

William Sanders, Clarion University

Editor’s Introduction and Acknowledgments

Sandra Trejos, Ph. D.

Editor of Proceedings

2013 Annual Conference of the Pennsylvania Economic Association

This volume contains papers presented at the 2013 Annual Conference of the Pennsylvania Economic

Association (PEA) held at the University of Scranton from May 31 to June 2, 2013. Only the papers

and discussions submitted according to manuscript guidelines are included in these proceedings. Not

every paper listed in this program was submitted for inclusion though. I thank Lori Klepfer, secretary

for the Department of Economics, and Yang Yang, my graduate assistant, in the College of Business

Administration at Clarion University of Pennsylvania for their excellent assistance in preparing these

proceedings.

University faculty, research professionals, graduate students, and undergraduate students gathered in

Scranton, Pennsylvania, to present their scholarly work from different corners of our country and the

world. Sessions allowed not only the presentation of fine work, but also the constructive criticism and

discussion present in any productive and meaningful professional meeting. The different concurrent

sessions were properly complemented by our Luncheon Speaker Mr. Peter Danchak, President of

PNC Bank, and the Federal Reserve Bank of New York lecture featuring Mr. Michael F. Silva, Senior

Vice President in the Financial Institutions Supervision Group. The annual conference ended with a

general meeting in which different aspects of our association were discussed with all members in

attendance.

The PEA would like to take this opportunity to thank Deborah Gougeon for hosting the 2013

meetings in such an organized and smooth manner. Her planning, energy, and organization skills

were crucial for another year of success for our conference. Thanks are also extended to the Federal

Reserve Bank of New York and Dr. Michael Mensah, Dean of the Kania School of Management, The

University of Scranton, and Fr. Kevin P. Quinn, S.J., President of The University of Scranton for their

support of this conference. Thanks, finally, to the PEA board and directors for their continued

commitment to and support of our organization, and all participants for their enthusiasm and

dedication to knowledge shown in their work and lively discussions. You make of this conference a

valuable event and we would like to invite you to continue being part of such a great group of

scholars. See you next year!

Table of Contents

Conference Agenda……………………………………………………………………………………………… page 1

The Effects of Power, Prestige, and Performance on Salary in the National Hockey League

Ashley M. Alt and Daniel R. Oberkofler, Indiana University of Pennsylvania …………………… …… page 16

Pennsylvania Tax Simplification: Nuisance Tax Credit, Obsolete Taxation and Administration Provision

Repeals, Including Proper Placement within the Tax Reform Code

Thomas O. Armstrong, Commonwealth of Pennsylvania, Department of Revenue……………… …… page 28

Business and Real-Estate Price Cycles across the Us: Evidence from a Markov-Switching Regression Exercise

Aram Balagyozyan, University of Scranton, Christos Giannikos, Baruch College, and Kyoko Mona,

Manhattanville College………………………………………………………………………………………… page 34

A Model of Relative Consumption

Chong Hyun C. Byun, Wabash College……………………………………………………………………… page 41

Peer Mentor Development as Secondary Leaders at the University Level

Dana D’Angelo and Susan Epstein, Drexel University…………………………………………………… page 48

Climate Neutrality in the Higher Education Sector: Making the Commitment

By Soma Ghosh, Discussion by Solomon T. Tesfu, Mount St. Mary’s University……………………… page 52

The U.S. Dollar as an International Currency Reserve and Its Continuous Depreciation

Ioannis N. Kallianiotis, University of Scranton……………………………………………………………… page 54

Trade Flows between S. Korea and the U.S.A.

Orhan Kara, West Chester University………………………………………………………………………… page 67

A Simple Model of Baseball Desegregation

Timothy F. Kearney & David Gargone, Misericordia University………………………………………… page 78

Impact of Migrant Remittances on Education Outcomes

Tanu Kohli, University of Illinois……………………………………………………………………………… page 87

Manufacturing Productivity in Pennsylvania

James A. Kurre, Penn State Erie, The Behrend College…………………………………………………… page 97

Force-Using Firms in the Competitive Equilibrium When the Only Publicly Provided Good is Envy

Gratification

Johnnie B. Linn III, Concord University…………………………………………………………………… page 109

The Effects of Increases in Cigarette Prices on Smoking Behaviors: Estimates Using MSA as a Natural

Experiment

Zhen Ma, Misericordia University…………………………………………………………………………… page 114

Mortgage Cash Flow Analysis and Pricing Using CAS (Collateral Analysis System)

Stephen M. Mansour and Riaz Hussain, University of Scranton………………………………………… page 123

Discussant Comments: The Geographic Concentration of Economic Activity

Stephen M. Mansour, University of Scranton…………………………………………………………………page 129

Credence Goods and State Mandated Vehicle Safety Inspections: How Non-Profit Inspection Services Can

Correct For Market Failure

John McCollough, Lamar University………………………………………………………………………… page 130

Some Possible Reasons for the Irrational Choice of Gift Cards

David Nugent, Robert Morris University……………………………………………………………………… page 137

Service Quality in the U.S. Airline Industry: Factors Affecting Customer Satisfaction

Logyn Pezak and Rose Sebastianelli, University of Scranton ……………………………………………… page 142

What Affects New Zealand Wine Prices? Estimation of the Effects of Sensorial, Reputational, Objective, and

Quality Factors in the Hedonic Price Model

Angela M. Rowland, Indiana University of Pennsylvania ………………………………………………… page 150

Author Index………………………………………………………………………………………………………page 162

Proceedings of the Pennsylvania Economic Association 1

Pennsylvania Economic Association

2013 CONFERENCE AGENDA

THURSDAY, May 30

4:00 pm - 9:00 pm Registration (Brennan Hall, 1st Floor Lobby)

5:00 pm - 8:00 pm Board of Director's Dinner/Meeting (Brennan Hall 509)

6:00 pm - 10:00 pm Reception (Brennan Hall 5th Floor Lobby)

FRIDAY, May 31

8:00 am - 12:00 pm, 2:00pm - 4:00pm Registration (Brennan Hall, 1st Floor Lobby)

7:30 am - 10:30am Continental Breakfast (Brennan Hall, 2nd Floor Lobby)

9:00 am - 10:15 am Concurrent Sessions (Brennan Hall)

10:15 am - 10:30 am Break (Brennan Hall, 2nd Floor Lobby)

10:30 am - 11:45 pm Concurrent Sessions (Brennan Hall)

12:00 pm - 1:45 pm Luncheon and Speaker - Mr. Peter Danchak, President of PNC

Bank, NE PA Region (Brennan Hall 509)

2:15 pm - 3:30 pm Concurrent Sessions (Brennan Hall)

3:30 pm - 3:45 pm Break (Brennan Hall, 2nd Floor Lobby)

3:45 pm - 4:45 pm Fed Lecture Featuring Mr. Michael F. Silva, Senior Vice President in

the Financial Institutions Supervision Group, Federal Reserve Bank of New York

(Pearn Auditorium, Brennan Hall 228)

5:00 pm - 8:00 pm Fed Sponsored Reception (Brennan Hall, 5th Floor Lobby)

SATURDAY, June 1

7:30 am - 10:30 am Registration (Brennan Hall, 1st Floor Lobby)

7:30 am - 9:00 am Continental Breakfast (Brennan Hall, 2nd Floor Lobby)

9:00 am - 10:15 am Concurrent Sessions (Brennan Hall)

10:30 am - 11:00 am General Membership Meeting (Pearn Auditorium, Brennan Hall

228)

11:15 am - Closing

Proceedings of the Pennsylvania Economic Association 2

FRIDAY, May 31, 2013

Conference Registration 8:00 a.m. – noon, 2:00-4:00 p.m., Brennan Hall Lobby, first

floor

7:30 a.m. – 10:30 a.m. Continental Breakfast (Brennan Hall, second floor)

Sessions F1: Friday, May 31, 2013, 9:00 a.m. – 10:15 a.m.

Session F1A: Housing Market

Location: Brennan Hall 102 Chair: Shuang Feng – Edinboro University

Mortgage Cash Flow Analysis and Pricing Using CAS

Stephen Mansour Rutgers University

The Effects of Housing Tenure on Labor Market Outcomes: Evidence from the USA

Md. Alauddin Majumder Middle Tennessee State University

Business and Real-Estate Price Cycles across the US: Evidence from a Markov-Switching Exercise Aram

Balagyozyan University of Scranton

Discussants:

Michael Hannan – Edinboro University

Patrick Litzinger – Robert Morris University

Steven Breslawski – College at Brockport, State University of New York

Session F1B: Economic Development I

Location: Brennan Hall 103 Chair: Yaya Sissoko – Indiana University of PA

Private Returns to Investment in Education in Cameroon; A Quintile Regression Analysis

Yaya Sissoko Indiana University of Pennsylvania

Wilfred Awung University of Buea, Cameroon

Antecedents and Consequences of the Aging Developed World: Implications for Business Systems

Abhijit Roy University of Scranton

Health Gap between Countries: Does Globalization Matter?

Khaled Elmawazini Gulf University for Science and Technology

Discussants:

Minh Tam Schlosky – West Virginia University

Tanu Kohli – Rutgers University

David Latzko – Penn State York

Proceedings of the Pennsylvania Economic Association 3

Session F1C: Women and Gender Economics

Location: Brennan Hall 105 Chair: Alberto Posso - RMIT University

The Latin American Mystique: Discrimination against Women in Ecuador's Labor Market Alberto Posso

RMIT University

An analysis of Gender Wage Gap over time in China’s urban labor market

Zhonghui Liu State University New York at Binghamton

Measuring Links between Labor Monopsony and the Gender Pay Gap in Brazil

Brandon Vick Indiana University of Pennsylvania

Discussants:

Charles Telly – SUNY Fredonia

Debarshi Indra – SUNY Buffalo

Zhonghui Liu – State University New York at Binghamton

Session F1D: Pricing and Market Reaction

Location: Brennan Hall 202 Chair: Travis Yates – Penn State Erie

The Effects of Increases in Cigarette Prices on Smoking Behaviors: Estimates Using MSA as a Natural Experiment

Zhen Ma Misericordia University

Psychological impacts on markets' response to earnings

Qian Hao and Anthony Liuzzo Wilkes University

A Pricing Course with Client-Based Experiential Learning

Robert Schindler Rutgers University - Camden

Alternatives for Highway Financing: Impact on Equity and Efficiency

Tracy Miller Grove City College

Discussants:

Ruttana Ruttanajarounsub – Edinboro University

Kosin Isariyawongse – Edinboro University

Travis Yates – Penn State Erie

Zhen Ma – Misericordia University

Proceedings of the Pennsylvania Economic Association 4

Session F1E: Student Session I

Location: Brennan Hall 203 Chair: Abdul Pathan – Pennsylvania College of Technology

Effects of Power, Prestige and Performance on Salary in the National Hockey League

Daniel Oberkofler and Ashley Alt Indiana University of Pennsylvania

Honor Killings: The Economic Parasite of Pakistan

Lauren Tassone Clarion University of PA

Challenges and Opportunities of Mass Customization: A Chinese Market Perspective

Huilan Zhang University of Toledo

Environmental Degradation and the Brazilian Economy

Erin Krotoszynski Clarion University of Pennsylvania

What Affects New Zealand Wine Prices? Estimation of the Effects of Sensorial, Reputational, and Quality Factors in the

Hedonic Price Model

Angela Rowland Indiana University of Pennsylvania

Discussants:

Deborah Gougeon – University of Scranton

Johnnie Linn, Concord University Margarita Rose – King’s College

Abdul Pathan – Pennsylvania College of Technology

John Kallianiotis – University of Scranton

Sessions F2: Friday, May 31, 2013 10:30 a.m. – 11:45 a.m.

Session F2A: Economic Development II

Location: Brennan Hall 102 Chair: Tracy Miller – Grove City College

Trade Flows between Korea and the U.S.

Orhan Kara West Chester University

Is There Moral Hazard in the HIPC Initiative Debt Relief Process?

Minh Tam Schlosky West Virginia University

Impact of Migrant Remittances on Fertility and Education Outcomes

Tanu Kohli Rutgers University

Bank Lending Channel in MENA Countries: Evidence from Dynamic Panel Model

Aram Belhadj University of Orléans

Proceedings of the Pennsylvania Economic Association 5

Discussants:

Jeremy Schwartz – Loyala University of Maryland

Sabri Yilmaz – Lycoming College

Orhan Kara – West Chester University

Tracy Miller – Grove City College

Session F2B: Taxation

Location: Brennan Hall 103 Chair: David Buehler – Penn State Harrisburg

Economic Impact of the Earned Income Tax Credit (EITC) for Erie County, Pennsylvania

Kosin Isariyawongse Edinboro University

Title Cost Function Estimations: Regularity Conditions and Cost Minimization

Kwami Adanu Gimpa Business School, Ghana

Pennsylvania Tax Simplification: Nuisance Tax Credit, Obsolete Taxation and Administration Provision Repeals

Thomas Armstrong PA Department of Revenue

Discussants:

James Kurre – Penn State Erie, The Behrend College

Aram Balagyozyan – University of Scranton

Elsy Kizhakethalackal – Bowling Green State University

Session F2C: Financial Economics

Location: Brennan Hall 105 Chair: Abhijit Roy -- University of Scranton

Informed Trading at Capitol Hill: Evidence from Congressional Trading over the 2004-2010 Period

Serkan Karadas West Virginia University

A Note on an Equation of Modigliani and Miller

Steven Andelin Pennsylvania State University

Help Your Students Realize Their Retirement Dreams by Quantifying the Cost of Procrastination

Jonathan Kramer and John Walker Kutztown University

Discussants:

Lee Siegel – PA Dept of Labor & Industry

Steven Breslawski – College at Brockport, State University of New York

Abhijit Roy – University of Scranton

Proceedings of the Pennsylvania Economic Association 6

Session F2D: Regional Economic Development

Location: Brennan Hall 202 Chair: Stephen Mansour - University of Scranton

Determinants of Economic Growth in ECOWAS Countries

Brian Sloboda US Department of Labor

Yaya Sissoko Indiana University of PA

The Geographic Concentration of Economic Activity across the Eastern United States, 1820-2010

David Latzko Penn State York

Credence Goods and State Mandated Vehicle Safety Inspections; How Non-Profit Inspection services can correct for market

failure

John McCollough Lamar University

Discussants:

John McCollough – Lamar University

Stephen Mansour – University of Scranton

Yaya Sissoko – Indiana University of PA

Session F2E: Economics of Education

Location: Brennan Hall 203 Chair: Sandra McPherson - Millersville University

Unders and Overs: Using a Dice Game to Illustrate Basic Probability

Sandra McPherson Millersville University

Students Perceptions of Hybrid Course Learning: A Case Study in Principles of

Microeconomics

Jui-Chi Huang Penn State Berks

Climate Neutrality in the Higher Education Sector: Making the Commitment

Soma Ghosh Albright College

Discussants:

Tufan Tiglioglu – Alvernia University

Huilan Zhang – University of Toledo

Solomon Tesfu – Mount St. Mary's University

Proceedings of the Pennsylvania Economic Association 7

Lunch Speaker

Friday, May 31

12:00pm Brennan Hall 509

Mr. Peter Danchak

President of PNC Bank, NE PA Region

Mr. Danchak joined PNC Bank in 1984 and has held various positions of responsibility in corporate

banking. He was named regional president of the Northeast PA market of PNC Bank in January 2001.

Mr. Danchak currently serves as co-chair of the Pennsylvania Early Learning Investment Commission and

is a member of the Executive Leadership Council of Pre-K Counts in Pennsylvania. He sits on the board

of trustees at Marywood University and Scranton Prep. He is a member of the board of directors of the

Pennsylvania Bankers Association, Blue Cross of Northeastern Pennsylvania, the ARC of Northeastern

Pennsylvania Foundation, the Luzerne Foundation, the Northeast Regional Cancer Institute, the Regional

Chamber Partnership and Scranton Lackawanna Industrial Building Company. He also serves on the

President’s Advisory Council for Keystone College and the Kania School of Management at the

University of Scranton.

Mr. Danchak previously served as a member of the board of directors of the F.M. Kirby Center for the

Performing Arts, King’s College, Keystone College, Johnson College, Junior Achievement of

Northeastern Pennsylvania, the Greater Pittston Chamber of Commerce, the Greater Scranton Chamber of

Commerce and the Greater Wilkes-Barre Chamber of Business and Industry.

Mr. Danchak received a Bachelor of Science degree in accounting from the University of Scranton.

Proceedings of the Pennsylvania Economic Association 8

Sessions F3: Friday, May 31, 2013 2:15 p.m. – 3:30 p.m.

Session F3A: Education Performance

Location: Brennan Hall 102 Chair: Zhonghui Liu - State University New York at Binghamton

Preliminary Results of Factors Contributing to Grade Changes

Robert Balough Clarion University of Pennsylvania

Rod Raehsler Clarion University of Pennsylvania

Teenage Socializing Behavior and Schooling Outcomes for American Youth

Solomon Tesfu Mount St. Mary's University

Analysis of Differences in Learning Outcomes: Online versus In-class Course

Sunita Mondal and Soma Ghosh University of Pittsburgh at Greensburg

Retention of Microeconomics Knowledge by Content and Cognitive Constructs

Steven Breslawski College at Brockport, State University of New York

Discussants:

Zhonghui Liu – State University New York at Binghamton

Brandon Vick – Indiana University of Pennsylvania

Steven Breslawski – College at Brockport, State University of New York

Serkan Karadas – West Virginia University

Session F3B: Health and Public Economics

Location: Brennan Hall 103 Chair: Khaled Elmawazini - Gulf University for Science and Technology

Force-Using Firms in the Competitive Equilibrium when the Only Publicly Provided Good is Envy Gratification

Johnnie Linn Concord University

Impact of Disaggregated Health-aid on Child Vaccinations: A Quantile Regression Analysis

Elsy Kizhakethalackal Bowling Green State University

Length of life, Individual Benefit Streams, and the Social Discount Factor for Long Lived Public Policies

William Bellinger Dickinson College

Discussants:

Khaled Elmawazini - Gulf University for Science and Technology William Bellinger -- Dickinson College

John Walker -- Kutztown University

Proceedings of the Pennsylvania Economic Association 9

Session F3C: Regional Economics

Location: Brennan Hall 105 Chair: Robert Schindler – Rutgers University

Identifying and Quantifying a Local Economy’s Exports

Travis Yates Penn State Behrend

Employment and Population Growth in Florida’s Counties

Sooriyakumar Krishnapillai American University of Nigeria

Manufacturing Productivity in Pennsylvania

James Kurre Penn State Erie, The Behrend College

Discussants:

Aram Belhadj -- University of Orléans

Dana D’Angelo -- Drexel University LeBow College of Business

Xuebing Yang -- Penn State Altoona

Session F3D: Economic Utility

Location: Brennan Hall 202 Chair: Margarita Rose – King’s College

A Model of Relative Consumption

Chong Hyun Byun Wabash College

Some Possible Reasons For The Irrational Choice of Gift Cards

David Nugent Robert Morris University

Service Quality in the U.S. Airline Industry: Factors Affecting Customer Satisfaction

Logyn E. Pezak and Rose Sebastianelli University of Scranton

Discussants:

Margarita Rose – King’s College

Abhradeep Maiti – Middle Tennessee State University

Sabri Yilmaz – Lycoming College

Session F3E: Labor and Demographic Economics

Location: Brennan Hall, 203 Chair: Zhen Ma -- Misericordia University

A Simple Model for Baseball Desegration

Timothy Kearney Misericordia University

David Gargone Misericordia University

Trials and Tribulations Of Analyzing Ferreted Pennsylvania Data: A Case Study

Lee Siegel Center for Workforce Information & Analysis

Recidivism of Juvenile Offenders

Proceedings of the Pennsylvania Economic Association 10

David Kalist Shippensburg University

Discussants:

Zhen Ma -- Misericordia University Steven Andelin -- Pennsylvania State University

Brian Sloboda -- US Department of Labor

Session F3F: Pedagogical Panel

Location: Brennan Hall 205 Chair: Abdul Pathan

Making the Principles of Economics Class Interesting to your Students

A group of faculty members will share methods they use to engage students in the classroom

Federal Reserve Lecture

Friday, May 31

3:45pm Pearn Auditorium, Brennan Hall 228

Mr. Michael F. Silva

Senior Vice President in the Financial Institutions Supervision Group, Federal Reserve

Bank of New York

Mr. Silva joined the bank in August 1992 as a law clerk in the Legal Group. He held positions with increasing

responsibilities in the legal group and in September 1995 was appointed an officer of the bank with the title of

Counsel. In December 1998, Mr. Silva was promoted to assistant vice president and became the lead counsel for

the Bank’s international account relationships and currency distribution. He was promoted to vice president in

December 1999 with those same responsibilities. In June 2006, Mr. Silva was promoted to senior vice president and

Proceedings of the Pennsylvania Economic Association 11

moved from the Legal Group to the Executive Group, where he served as chief of staff for president Geithner and

subsequently president Dudley. Mr. Silva’s tenure as Chief of Staff included all of the 2008 - 2009 financial crisis.

Mr. Silva concluded his assignment as chief of staff in September 2010 and moved to the Financial Institutions

Supervision Group, where he currently serves as the Senior Supervisory Officer for The Goldman Sachs Group. As

a collateral duty, he served as assistant corporate secretary of the bank from December 1995 to December 1999.

Mr. Silva holds a B.S. degree from the United States Naval Academy and a J.D. from Columbia Law School. He is

also a graduate of the Harvard Business School’s Advanced Management Program.

In 2004, Mr. Silva received the Department of Defense Joint Civilian Service Medal and the Secretary of the

Treasury’s Honor Award, both for his service in Iraq as a coalition advisor to the Central Bank of Iraq. Mr. Silva is

also the author of “A Central Banker in Iraq”, Journal of International Business & Law, Spring 2004.

Prior to attending law school and joining the bank, Mr. Silva served as an officer in the United States Navy from

May 1983 to September 1989. During that period, he was assigned to Fighter Squadron 142 as an F-14A Tomcat

Radar Intercept Officer and also to the On-Site Inspection Agency, a joint military and civilian agency in

Washington DC responsible for conducting the first ever arms control verification inspections in the former Soviet

Union. Mr. Silva is a 1986 graduate of the Navy Fighter Weapons School (a.k.a “Topgun”).

5 –8 P.M. Brennan Hall Lobby, fifth floor Reception hosted by the Federal Reserve Bank

Proceedings of the Pennsylvania Economic Association 12

SATURDAY, June 1, 2013 ~ 7:30 – 10:30 A.M. Conference Registration-Brennan Hall Lobby, first floor

Continental Breakfast-Brennan Hall Lobby, second floor ____________________________________________

Sessions S1: Saturday, June 1, 2013 9:00 a.m. – 10:15 a.m.

Session S1A: Minimum Wage and Labor Market

Location: Brennan Hall 102 Chair: Michael Hannan, Edinboro University of PA

Optimal Unemployment Insurance: When Search Takes Effort and Money

Jeremy Schwartz Loyala University of Maryland

Labor Demand Elasticity in the United States

Abhradeep Maiti Middle Tennessee State University

A Wage You Can Live On Minimum Wage Debate: Public Perceptions vs. Academic Findings

Jui-Chi Huang Penn State Berks

Discussants:

Robert Balough, Clarion University of PA

Michael Hannan, Edinboro University of PA

Sooriyakumar Krishnapillai, American University of Nigeria

Session S1B: International Economics

Location: Brennan Hall 103 Chair: Jonathan Kramer, Kutztown University

Patterns of Product Level Trade

Xuebing Yang Penn State Altoona

Trade Creation, Diversion, and Displacement: A Shift-Share Analysis Of EU Enlargement

David Buehler Penn State Harrisburg

The U.S. Dollar as an International Currency Reserve and its Continuous Depreciation

John Kallianiotis University of Scranton

Discussants:

Tufan Tiglioglu, Alvernia University

David Nugent, Robert Morris University

Jonathan Kramer, Kutztown University

Proceedings of the Pennsylvania Economic Association 13

Session S1C: General Economics

Location: Brennan Hall 105 Chair: William Bellinger, Dickinson College

The Mythological Hero and Its Development from Socrates to Dante, to Mandeville, to Adam

Smith, to Hayek and the Modern Manager

Charles Telly SUNY Fredonia

Developing University Peer Mentors as Second Chair Leaders

Dana D'Angelo Drexel University LeBow College of Business

Additive Manufacturing and Costs of Production

Patrick Litzinger Robert Morris University

Discussants:

Tracy Miller – Grove City College

David Latzko – Penn State York

David Kalist – Shippensburg University

10:30 a.m.-11:00 General Membership Meeting

Brennan Hall 228

This Annual Business Meeting of the General Membership of the Pennsylvania Economic Association is open to

the entire membership of the PEA, including all registrants at the conference. Door prizes will be awarded.

SATURDAY, June 1, 2013 CLOSING

Proceedings of the Pennsylvania Economic Association 14

Program Author & Participant Index

First Name Last Name Email Session

Kwami Adanu kadanu@gimpa.edu.gh F2B

Ashley Alt bbpq@iup.edu F1E

Steven Andelin sla7@psu.edu F2C, F3E

Thomas Armstrong thoarmstro@pa.gov F2B

Aram Balagyozyan aram.balagyozyan@scranton.edu F1A, F2B

Robert Balough balough@clarion.edu F3A, S1A

Aram Belhadj aram.belhadj@etu.univ-orleans.fr F2A, F3C

William Bellinger bellinge@dickinson.edu F3B, S1C

Steven Breslawski sbreslaw@brockport.edu F3A, F1A, F2C

David Buehler dlb74@psu.edu F2B, S1B

Chong Hyun Byun byunc@wabash.edu F3D

Dana D'Angelo dangeldc@drexel.edu S1C, F3C

Khaled Elmawazini k.elmawazini@alumni.uottawa.ca F1B, F3B

Shuang Feng sfeng@edinboro.edu F1A,

David Gargone dgargone@misericordia.edu F3E

Soma Ghosh sghosh@alb.edu F2E, F3A

Deborah Gougeon gougeond1@scranton.edu F1E,

Michael Hannan hannan@edinboro.edu F1A, S1A

Qian Hao qian.hao@wilkes.edu F1D

Jui-Chi Huang jxh74@psu.edu F2E, S1A

Debrashi Indra dindra@buffalo.edu F1C

Kosin Isariyawongse kisariyawongse@edinboro.edu F2B, F1D

David Kalist dekali@ship.edu F3E, S1C

John Kallianiotis ioannis.kallianiotis@scranton.edu S1B, F1E

Orhan Kara dearorhankara@gmail.com F2A

Serkan Karadas Serkan.Karadas@mail.wvu.edu F2C, F3A

Timothy Kearney tkearney@misericordia.edu F3E

Elsy Kizhakethalackal ekizhak@bgsu.edu F3B, F2B

Tanu Kohli tkohli@rutgers.edu F2A, F1B

Jonathan Kramer jkramer@kutztown.edu F2C, S1B

Sooriyakumar Krishnapillai sooriyakumar.krishna@aun.edu.ng F3C, S1A

Erin Krotoszynski E.Krotoszynski@eagle.clarion.edu F1E

James Kurre k12@psu.edu F3C, F2B

David Latzko dlatzko@psu.edu F2D, S1C, F1B

Johnnie Linn linnj@concord.edu F3B, F1E

Patrick Litzinger litzinger@rmu.edu S1C, F1A

Proceedings of the Pennsylvania Economic Association 15

Zhonghui Liu zliu8@binghamton.edu F1C, F3A

Zhen Ma zma@misericordia.edu F1D, F3E

Abhradeep Maiti am4m@mtmail.mtsu.edu S1A, F3D

Md. Alauddin Majumder mm4v@mtmail.mtsu.edu F1A

Stephen Mansour stephen.mansour@scranton.edu F1A, F2D

John McCollough jmccollough@lamar.edu F2D

Sandra McPherson smcpherson@millersville.edu F2E

Tracy Miller tcmiller@gcc.edu F1D, F2A, S1C

Sunita Mondal sum46@pitt.edu F3A

David Nugent davidanugent@hotmail.com F3D, S1B

Daniel Oberkofler danoberkofler22@gmail.com F1E

Abdul Pathan apathan@pct.edu F1E, F3F

Logyn E. Pezak pezakl2@gmail.com F3D

Alberto Posso alberto.posso@rmit.edu.au F1C

Margarita Rose MargaritaRose@kings.edu F1E,F3D

Angela Rowland a.m.rowland91@hotmail.com F1E

Abhijit Roy roya2@scranton.edu F1B, F2C

Ruttana Ruttanajarounsub ruttana@edinboro.edu F1D

Robert Schindler rschindl@camden.rutgers.edu F1D, F3C

Minh Tam Schlosky tammyschlosky@gmail.com F1B, F2A

Jeremy Schwartz schwarjf@msn.com F2A, S1A

Rose Sebastianelli sebastianer1@scranton.edu F3D

Lee Siegel LSIEGEL@pa.gov F3E, F2C

Yaya Sissoko ysissoko@iup.edu F1B, F2D

Brian Sloboda bsloboda@email.phoenix.edu F2D, F3E

Lauren Tassone L.Tassone@eagle.clarion.edu F1E

Charles Telly tabak@fredonia.edu S1C, F1C,

Solomon Tesfu tesfu@msmary.edu F3A, F2E

Tufan Tiglioglu tufan.tiglioglu@alvernia.edu F2E, S1B

Bradon Vick vick@fordham.edu F1C, F3A

John Walker walker@kutztown.edu F2C, F3B

Xuebing Yang xyang@psu.edu S1B, F3C

Travis Yates tmy5026@gmail.com F3C, F1D

Sabri Yilmaz yilmaz@lycoming.edu F2A, F3D

Huilan Zhang hzhang11@rockets.utoledo.edu F1E, F2E

Proceedings of the Pennsylvania Economic Association 16

THE EFFECTS OF POWER, PRESTIGE, AND PERFORMANCE

ON SALARY IN THE NATIONAL HOCKEY LEAGUE

Ashley M. Alt

Indiana University of Pennsylvania

2315 Franklin Street

Johnstown, PA 15905

Daniel R. Oberkofler

Indiana University of Pennsylvania

587 Saxony Lane

Yardley, PA 19067

ABSTRACT

The impact of individual player statistics on National Hockey

League (NHL) players’ salaries is analyzed empirically using

career statistics. A double-log regression equation is

estimated using ordinary least squares (OLS). Dependent

variables in this study are the 2011-2012 salaries.

Independent variables are the natural log of goals, assists and

shots on goal, points per game, plus/minus, hits, penalty

minutes, height, and weight, save percentage, wins, losses,

the natural log of games played, natural log of shutouts.

Dummy variables are also assigned. Shots on goal, points

per game, goals, and star, assists, hits, save percentage and

the natural log of games played are significant determinants

for respective groups.

JEL Codes: L83, M2, J30

Keywords: NHL Salary, hockey statistics, salary

determinants

INTRODUCTION

Team Profit Maximization

Basic economic theory asserts that the goal of every firm is to

maximize profit. Firms maximize profit; in part, through the

use of the wages they pay employees. For efficiency, the

harder an employee works, the more compensation they will

receive in return for their work. All rational, profit-

maximizing firms operate under this principle, and the 30

National Hockey League (NHL) franchises are no different.

Under the profit maximization principle, an NHL team will

strive to increase total revenues and decrease total costs.

Increasing total revenue includes creating an environment in

which fans will want to attend games, purchase merchandise,

and support their respective teams. Fostering a winning

culture within a franchise leads to higher revenues.

A fundamental additive in maximizing profit consists of

delving into how the General Manager (GM) of each team

decides on the salary each player receives. However, this is

not always an easy task for a GM because of the wide range

of skills amongst players. Player performance during the

season is used as a benchmark for determining a starting

point for salary negotiations. Since successful teams earn

higher revenues, and every team relies on its players to

perform at a certain level in order to be successful, the GM

must spend money in order to create that success. Yet,

spending more money does not necessarily create success.

Therefore, the goal of the GM is to create a competitive

franchise without overpaying the players. In other words,

maximizing total revenue and minimizing total cost. It is

important to note that while all firms operate under the profit

maximization principle, the same goes for all sports leagues.

Major League Baseball (MLB), the National Football League

(NFL), and the National Basketball Association (NBA),

among others, all function according to the same criterion.

That is, they all want to keep salary as low as possible while

still championing a successful team.

NHL Labor Market

To begin, we can draw an abundance of similarities between

the NHL and the conventional job market. For instance, just

as workers are represented by unions in large industries,

players in the NHL are represented by the NHL Players'

Association (NHLPA). The NHLPA serves as the go-

between for players on matters similar to unions; working

conditions, contractual rights, and the NHLPA also serves as

the collective bargaining agent for players. Both traditional

labor unions and the NHLPA help protect employees from

unfair working conditions.

Moreover, there is a set of restrictions built into the wage

system within professional sports leagues. All athletes must

earn at least the league approved minimum salary, and

though these differ across professional sports leagues, they

are very similar to minimum wage laws. By the same token,

rookie athletes generally earn less than league veterans

because they lack the experience and additional skills one

gains through multiple years of play. At the same time, new

employees at any firm often receive lower wages to start and

must climb their way up the wage ladder by gaining more

experience and expanding their skill set. It also makes

Proceedings of the Pennsylvania Economic Association 17

economic sense that those employees who contribute more to

the firm will be rewarded in kind, thus we can confidently

expect that the more a player contributes to a team, the more

they will be rewarded financially.

Furthermore, professional sports leagues are similar to the

traditional job market in that professional sports have the

concept of free agency. Free agency gives a player the right

to sign with another team if he feels dissatisfied with his

current team. An obvious parallel to this is the time it takes

for one person to find the right job. Workers are not forced

to keep the same job for the rest of their lives, and many

people change jobs or even careers several times in their

lifetimes. Workers may do this in search of higher wages,

better working conditions, or a better location. The same

goes for professional athletes, many of whom switch teams

several times as well. Clearly, because of the similarities one

can draw between the traditional job market and professional

sports leagues, it is apparent why sports franchises should be

studied using economic theory and econometrics.

At the same time, professional sports franchises can be easier

to study than conventional firms because of the massive

amount of statistics every player accumulates each year.

Another bonus to the accumulation of these statistics is that it

becomes easier to find which ones have the most significant

impact on salary determination. On the other hand,

employees in many firms do not amass large numbers of

individual statistics, making it difficult to show how their

input affects firm output. By analyzing the determination of

salary for NHL players, we can better understand how an

individual's impact on firm output is rewarded with salary.

This study examines the factors that are significant in

predicting NHL salary. In other words, how do performance,

prestige, and power factor into determining National Hockey

League players' salaries? This paper examines previous

literature on NHL salary determinants and identifies the

independent variables used in this regression. Based on the

results, we can draw conclusions about salary and which

predictors are significant. Additionally, we used many of the

same independent variables as Vincent and Eastman (2009);

career goals, assists, points per game, plus/minus rating, draft

round, penalties in minutes, height and weight. We also

included hits as part of our defensive model, something that

had previously not been done. Thus, our study contributes

something new to the pre-existing literature. Furthermore,

for goalies, shutouts were not included in any reviewed

literature. Moreover, the way the star variable is constructed

is completely original. Finally, our study examines data for

the most recent season available.

Outline of Paper

The second section of the paper reviews previous

studies of salary determinants in the NHL. Explanatory

variables and data are discussed in the third section. The

econometric model for this study is explained in section four.

Section five examines the regression equations and results.

The last section of this paper concludes by discussing the

findings and their implications.

LITERATURE REVIEW

Numerous empirical studies have been conducted on the

determinants of salary for professional athletes. There are;

however, only a smattering focused on the National Hockey

League. Zimbalist (2010) stressed the importance of

collecting the correct data in order to correctly define

compensation for players. In order to conduct research

efficiently, it is imperative to understand the National

Hockey League’s definition of salary, revenue, and how

those definitions are created. This study accentuates the fact

that different leagues define salary, and salary as a function

of total revenue, differently. Moreover, in Marchand,

Smeeding and Torrey (2006), the effects of overall team

performance and teamwork were analyzed. They used 2,866

player-season observations of 1,001 players over four NHL

seasons: 2000-01 through 2003-04. Specifically, they

discussed whether or not there is a "star effect" versus a

"journeyman effect," or, whether having a few star players

who earn above median salary leads to better performance.

In fact, their results showed that the presence of a player

deemed a "star" increased the spread of assists while the

goals scored spread remained relatively the same. In other

words, other lower paid players were jumping into the play

and contributing on the ice more.

Vincent and Eastman (2009) used a quantile regression with

data from the pre 2004-2005 lockout period to see how

changes in their explanatory variables would be reflected

within different quantiles; the 10th

, 25th

, 50th

, 75th, and 90

th.

In the end, they found that variables have different effects

across quantiles. For example, their results implied that one

more game played is much more valuable to a player in the

75th

or 90th

quantiles of games played than it is for a player in

the 10th

or 25th

quantiles. Furthermore, the study discussed

the difficulty in truly analyzing player performance because

of the combined offensive and defensive aspects of the game.

Lambrinos and Ashman (2007) explored the effects of

arbitration on salary for the 2001-02 season. They found that

often arbitrated salaries do not differ from negotiated salaries

for both forwards and defensemen. However, they did

acknowledge a possible sample size problem since only 17

forwards and 20 defensemen went through arbitration in

2001-02.

Chan, Cho and Novati (2012) discussed the contribution of

different player types to team performance. Their analysis

was based on Vincent and Eastman (2009), but they

expanded by putting additional emphasis on defensive

Proceedings of the Pennsylvania Economic Association 18

categories for forwards and defensemen. These categories

included: first line, second line, defensive, and physical

players for forwards, and offensive, defensive, average, and

physical for defensemen. They also included elite, average,

and bottom categories for goalies. They concluded that

goalies contribute the most to team performance, followed by

forwards, and then defensemen. Furthermore, they found

that player-types previously considered less attractive

become investments with a comparable return to more

offensive-oriented forwards. Another study done by Vincent

and Eastman (2012) analyzed player mobility. Their analysis

found that players who frequently change teams incur a

negative cumulative effect on their salaries.

Berri and Brook (2010) analyzed the most important position

in professional sports; the goaltender. The focus of their

study was mainly on whether general managers efficiently

ascertain a goaltender's worth in salary through the Vezina

Trophy (best goaltender), previous salary data, and a list of

independent variables, of which they found save percentage

to be most significant. Their results suggested that the

difference between a goalie considered "one of the greatest

ever" and "just average" is negligible.

DATA

The dependent variable in this study is the natural log of the

individual player salary, for the 2011-2012 NHL season.

Vincent and Eastman (2009) also used the natural log of

player salary in their analysis. Data is obtained from USA

Today, similar to Marchand et al. (2006). We separated the

dependent variable amongst the forwards, defensemen and

goalie groups. By using a large sample size of 375 NHL

forwards, 200 NHL defensemen and 48 goalies, we can

determine which statistics are the most influential in

determining players’ salaries. Since there is such a large

sample of NHL players, we do not have to be concerned

about potential degrees of freedom issues. Like Vincent and

Eastman (2009), forwards and defensemen are split because

each utilizes a separate set of skills in order to accomplish the

objectives of their position. Goalies also use a separate

regression, for obvious reasons. Therefore, a different set of

independent variables will be used for each in order to

generate a model that explains salary differences.

Salaries observed are for players who played at least 20

regular season games during the given season for the purpose

of consistency and the elimination of potential outliers.

Since the majority of teams are located in the United States

of America, with the exception of seven Canadian teams,

U.S. dollars are used to calculate salary. The player salaries

for the seven teams who use Canadian dollars are converted

into U.S. dollars.

Additionally, we used many of the same independent

variables as Vincent and Eastman (2009); career goals,

assists, points per game, plus/minus rating, draft round,

penalties in minutes, height and weight. The player

performance data was collected from the individual team

websites hosted by NHL.com. All of the data collected for

the study includes career as well as season totals. For the

purpose of consistency, statistics for the players’ entire

careers are used instead of just the 2011-2012 season. Career

statistics account for all years, including years of exceptional

performance as well as years of poor performance. Outliers

are a potential issue due to the variation in human

performance variables and the length of players’ NHL

careers. Players have varying degrees of experience, with

some having only a few years, and others like Jaromir Jagr,

have been playing for over two decades. Salaries, as well as,

human performance variables have a tendency to be right-

skewed. The natural logs of assists, goals, shots on goal,

shutouts and games played are taken in order to reduce this

potential outlier problem.

Expected Signs

Each season NHL players amass several statistics. Table 1 in

the appendix displays the variables used in this study and

their expected signs. Easily, the most important of these

statistics is goals scored (G_ln). We used the natural log in

order to address outlier issues. A goal is awarded to the last

player to touch the puck before it crosses the goal line. Goals

are the deciding factor in winning or losing. Typically,

forwards will have more goals than defensemen. A player

who scores more goals should theoretically earn a higher

salary. We expect that goals will have a positive coefficient

for both forwards and defensemen. There is an abundance of

studies that confirm our expected positive sign. For example,

Vincent and Eastman (2009) conclude points, which are

defined as the summation of goals and assists, have a

significant positive impact on salary for forwards and

defensemen.

Another useful statistic is assists (A_ln). Again, assists are

typically right-skewed, thus the natural log is taken. An

assist is attributed to up to two players of the scoring team.

They must have shot, deflected or passed the puck to the goal

scorer, or touched it before another player earned the goal.

For every goal, there typically is at least one assister, so we

can expect assists to have a positive coefficient for both

forwards and defensemen like goals scored. Vincent and

Eastman (2009), and Marchand et al. (2006) incorporate

assists into their models.

Shots on goal (SG_ln) can be a very influential aspect of

hockey. To correct for right-skew, the natural log is taken. It

is not possible for a player to earn a goal unless they are

actively shooting on the net. Shots supplement goals and

assists, and help to further explain variations in salary.

Typically, a player who contributes more shots on goal helps

create offensive chances, thereby increasing the likelihood of

Proceedings of the Pennsylvania Economic Association 19

his team achieving a win. Therefore, we expect the

coefficient for shots on goal to be positively related to salary.

It is also likely that a forward will have many more shots on

goal than a defensemen, for this reason shots on goal is

included only for forwards. Moreover, no existing literature

that was examined for this study included shots on goal,

which is another way this study contributes to the current

literature.

Another key statistic in determining salary for NHL

forwards, in particular, is plus/minus (PLUSMINUS).

Vincent and Eastman (2009) and Lambrinos and Ashman

(2007) also use plus/minus as an explanatory variable. A

player earns a plus one for being on the ice when a teammate,

or the player himself, scores a goal. Conversely, a player will

earn a minus one for being on the ice when a player of the

opposing team scores a goal. Naturally, a player with a

higher, positive plus/minus total will earn a higher salary.

Thus, it is logical to assume a positive sign for this

coefficient.

Points per game (PPG) statistics give insight into how well a

player is performing on a game-by-game basis. Points per

game are calculated by summing the number of career goals

and career assists, and then taking that sum and dividing it by

the number of career games that were played. We expect

points per game to have a positive sign because there is a

positive relationship between points per game and

goals/assists. Additionally, we expect that points per game

have a much more important impact on salary for a forward

than it does for a defenseman. Using this reasoning, points

per game is included in only the forwards model. Vincent and

Eastman (2012) also used points per game as an explanatory

variable.

Crucial statistics for defensemen, in particular, are penalty

minutes (PIMS). A penalty occurs when one team is awarded

an extra attacker due to a rule infraction from a player on the

opposing team. Defensemen often take more penalties

because the nature of the position requires more aggressive

play. Coaches generally approve because a defenseman’s

primary objective is to keep the puck from getting to the net.

Moreover, few other events rile up the crowd like a big hit or

a fight. On the other hand, a penalty always hurts a team

because it is put on the defensive, moreover missing one

man. Therefore, taking penalties can either carry a positive or

negative sign. Vincent and Eastman (2012) used penalties in

minutes as a covariate; however, they predicted a positive

relationship with salary.

Height (HEIGHT) and weight (WEIGHT) are used in several

other studies, such as Lambrinos and Ashman (2007) and

Vincent and Eastman (2009), as significant determinants of

salary for defensemen. This makes sense because bigger

players have the ability to hit harder, thus keeping the net

safer from the opposing team. For the purpose of determining

salary, we believe that the expected sign of both height and

weight will be positive.

When determining a player’s salary, for defensemen

specifically, hits are a vital statistic. Hits (HITS) are typically

recorded when a player body checks a player of the opposite

team causing him to lose control of the puck. As a

defenseman, hits are essential because it is necessary to play

rough in order to get the puck away from the net and out of

the defensive zone. Therefore, hits should have a positive

sign. Furthermore, we did not identify an existing study that

included hits in their model.

Dummy variables are created for the round a player was

drafted (DRAFT), and whether a player was either invited to

or played in the 2010-2011 All-Star Game, or had won the

Stanley Cup in the 2009-10 and/or 2010-11 seasons (STAR).

For the round drafted, a value of one was given to the players

that were drafted in either the first or second round of the

draft, and all others are represented by zero. For star power,

players are given a one if they either were invited to play, or

played in the All-Star Game, or won the Stanley Cup in the

aforementioned seasons, and those who did not are given a

zero. A first or second round draft pick should carry a

positive expected sign because the best players are selected

first, and therefore should be projected to have a higher

salary. Similarly, only the best players are invited to play in

the All-Star Game so we should expect those players to earn

a higher salary.

On the other hand, a prominent statistic for goalies is save

percentage (SAVE_%). Save percentage is the number of

shots a goalie saves divided by the number of shots faced. A

higher save percentage generally garners confidence from

both the front office and the players on the ice in their

goaltender because they trust the goalie to make the save.

Therefore, it is logical to expect that goalies with a higher

save percentage earn higher salary. Berri and Brook (2010)

used save percentage as an explanatory variable.

Another useful statistic for goalies is wins (WINS). A goalie

earns a win in a game for playing the entire game and

winning, or when he becomes the goalie of record by being

on the ice for the game-winning goal, regardless of the team

that scores. Naturally, wins are expected to carry a positive

sign. Conversely, losses (LOSSES) are also used. Goalies

earn a loss when the game-winning goal is scored on them.

We expect losses to have a negative sign. While Berri and

Brook (2010) included wins in their analysis, losses are not

used in any study we examined.

Goalies are also evaluated on their experience, thus the

natural log of games played is taken (LN_GP). The natural

log is taken because there are a plethora of goalies with very

few games played and only a few with a high number of

games played. The natural log helps control for this skew in

Proceedings of the Pennsylvania Economic Association 20

the data. Generally, with more experience there is a higher

salary, so we expect the natural log of games played to carry

a positive sign. Berri and Brook (2010) used minutes as a

measure of experience, instead of games played.

Also, for goalies, the natural log of shutouts is included

(LN_SHUTOUTS). A shutout is recorded when the

opposing team does not score. Again, this data is right-

skewed, so the natural log is taken to correct the skew.

Shutout victories are typically accredited wholly to the goalie

for their exceptional play. Again, we expect shutouts to carry

a positive sign. Shutouts were not used as an explanatory

variable in any existing study.

Finally, a goalie star dummy variable is created (STARG)

which indicates the 2009-2010 and 2010-2011 All-Star

goalies, the Stanley Cup winners, as well as the Vezina (best

overall goaltender) and Jennings (fewest goals against)

trophy winners from those years. We expect the goalie star

variable to carry a positive sign.

Descriptive Statistics

The mean salary for all NHL forwards is $2,500,508. Table

2.1 displays the descriptive statistics for forwards. The

maximum salary earner for forwards is Brad Richards,

bringing in a whopping $12,000,000, while 15 players earned

the league minimum $525,000. The mean number of career

goals is 103; similarly, the average number of career assists is

143. Jaromir Jagr has both the most career goals and assists

with 665 and 988, respectively. Averaging 1.40 points per

game, Sydney Crosby leads the pack in points per game. The

mean number of points per game is 0.48. Again, Jaromir

Jagr owns the highest plus/minus statistic at +280. The

league average is +8. The league leader in shots on goal is

Jaromir Jagr with 4,766. The mean number of shots on goal

for the league is 914.

Table 2.2 displays the descriptive statistics for defensemen.

Christian Ehrhoff leads NHL defensemen earning

$10,000,000. The mean salary for NHL defensemen is

$2,551,294. Nicklas Lidstrom leads defensemen in goals

with 264 and assists with 878. The average number of goals

for defensemen is 33 and 109 for assists. Zdeno Chara is a

monster on the ice at a towering 81 inches, the shortest player

comes in at 68, and the mean is 74 inches. The mean weight

for defensemen is 210 pounds. The player with the most hits

is Luke Schenn with 270. The league average is 88.

Table 2.3 displays the descriptive statistics for goalies. The

league leader is Ilya Bryzgalov earning $10,000,000. On the

other hand, the mean salary for goalies is $2,820,563.

Tuukka Rask commands the best save percentage with 0.926,

meanwhile the league mean is 0.912. The mean number of

wins is 157.75 and the average number of losses is 188.6.

ECONOMETRIC MODEL

This study utilizes ordinary least squares (OLS) regression

analysis to measure the importance of performance, power,

and prestige on a player’s salary. The dependent variable for

all equations is the natural log of salary for the 2011-2012

NHL season. For all equations, the X matrix consists of

performance variables and the Z matrix consists of power

variables, and STAR/STARG is the relevant prestige

variable. The empirical specification for defensemen is:

LnSali = 1*Xi + Zi3*STARi + i (1)

where X is a matrix of control variables, in relation to the

defensive position, consisting of G_ln, A_ln, HITS, PIM, and

DRAFT. Z is a matrix of control variables including

HEIGHT and WEIGHT.

The empirical specification for the offensive

position is:

LnSali = 1*Xi + Zi3*STARi + i (2)

where X is a matrix of control variables, consisting of G_ln,

A_ln, and SG_ln. Z is a matrix of control variables including

PPG, PLUSMINUS, and DRAFT.

The empirical specification for the goalie position

is:

LnSali = 1*Xi + i3*STARGi + i

where X is a matrix of control variables, consisting of

SAVE_%, LN_SHUTOUT, and WEIGHT. Z is a matrix of

control variables including WINS, LOSSES, and LN_GP.

Econometric Issues

The analysis is cross-sectional and therefore

heteroskedasticity may be present. Thus, the White test was

run on all models. The defensive, offensive, and goalie

equations were found to be heteroskedastic, and the models

were corrected for heteroskedasticity using White

heteroskedasticity-consistent standard errors. Final models

discussed in the following section reflect adjustments to the

specifications required to address multicollinearity issues.

Outliers are a potential issue due to the variation in human

performance variables and the length of players’ NHL

careers. Players have varying degrees of experience, with

some having only a few years, and others like Jaromir Jagr,

have been playing for over two decades. Salaries, as well as,

human performance variables have a tendency to be right-

skewed. The natural logs of assists, goals, shots on goal,

shutouts and games played are taken in order to reduce this

potential outlier problem.

Proceedings of the Pennsylvania Economic Association 21

RESULTS

Table 3.1 in the appendix displays results for the defensive

models. For all equations, assists are found to be positive

and significant at the 1% level. This corroborates past

studies which have also noted the significance of assists for

defensemen, including Marchand et al. (2006), among others.

Furthermore, hits are found to be positive and significant at

the 1% level for all models. As stated earlier, no literature

reviewed for this study included hits, hence a significant

factor in determining salary for defensemen has been

discovered by this study. The star variable also proved to be

positive and significant at the 1% level. Though we

constructed the star variable differently, previous studies like

Marchand et al. (2006) and Vincent and Eastman (2009) also

concluded that the star variable is positive and significant.

Goals is shown to be positive as expected but not significant,

perhaps because coaches do not necessarily expect

defensemen to score.

On the other hand, draft round selected and height show

negative and insignificant signs. One reason for the

unexpected sign on height could be the correlation it shares

with weight. Penalties in minutes is found to have an

insignificant negative coefficient, owing to the fact that when

a team is penalized they lose a player, thus making it more

difficult to defend the net. Likewise, since the 2005

Collective Bargaining Agreement, and especially since the

injuring of super-star Sydney Crosby, the league has been

stepping up its player safety efforts. Our adjusted R-squared

values, ranging from 0.557 to 0.572, are higher than Vincent

and Eastman (2009), whose values range from 0.299 to

0.533.

The regressions run for forwards can be seen in Table 3.2 in

the appendix. For all models, points per game is shown to be

positive and significant at the 1% level. Similarly, Vincent

and Eastman (2009) found points per game to be positive and

significant at the 5% level. Shots on goal is found to be

positive and significant. Goals are found to be significant at

the 5% level in Model 2, and significant at the 1% level in

Models 3 and 4. Moreover, the star variable is found to be

positive and significant at the 5% level. Previous literature,

such as Marchand et al. (2006), and Vincent and Eastman

(2009), all found their versions of the star variable to be

positive and significant. Although the draft dummy variable

and plus/minus variable did not turn out to have any

significance, they are left in the models in accordance with

Vincent and Eastman (2009). Moreover, the signs are

consistent both with literature and expectations. Vincent and

Eastman (2009) obtained adjusted R-squared values ranging

from 0.324 to 0.758. The range of our adjusted R-squared

values are more concentrated, ranging from only 0.681 to

0.692.

The equations for goalies are displayed in Table 3.3. For all

but one model, the natural log of games played is significant

at the 1% level and the circumstance in which it is not, it is

significant at the 5% level. Our findings corroborate Berri

and Brook (2010) who also found their experience proxy,

minutes, to be significant at the 1% level. Furthermore, save

percentage is found to be significant at the 5% level for all

models. Again, this corroborates Berri and Brook (2010),

who found save percentage to be significant at the 1% level.

Weight and the natural log of shutouts were found to be

insignificant. Wins and losses were not found to be

statistically significant; however, they are left in the models

in accordance with previous literature. The star variable

failed to garner significance in any model. Berri and Brook

(2010) obtained an adjusted R-squared range of 0.27 to 0.48.

In comparison, our adjusted R-squared ranges from 0.362 to

0.376.

CONCLUSION

Selected based on past literature, power, prestige and

performance variables are utilized to estimate player salary

for 200 defensemen, 375 forwards, and 48 goalies in the

NHL during the 2011-2012 hockey season. Several important

results are evident and show the importance of separating

defensive, offensive, and goalie equations. To begin, results

show the importance of lagging the star variable back to the

2009-10 and 2010-11 seasons to indicate that success from

past seasons has an effect on future salary. From our results

we are able to conclude that players, who are considered to

be of star quality, do garner a higher salary. As mentioned

earlier, other studies, such as Vincent and Eastman (2009),

also found a positive and significant relationship between star

and salary. Ultimately, our hypothesis was that there is a

positive and significant relationship between star and salary,

and the results support our presumptions clearly.

When comparing descriptive statistics it is important to note

that defensemen do earn a slightly higher salary, on average,

than forwards. We attribute this difference in salary to the

fact that defensemen play the harder position. Defensive

players must make sure that their skills are top notch since

overall defense is very difficult, and a player is required to

keep extra sharp, mentally and physically. We conclude that

defensemen should put added emphasis on hits and assists

while on the ice because these factors contribute the most to a

higher salary. Therefore, an aspiring defenseman should be

sure to practice both his offensive game, specifically passing

and setting up shots, and defensive skills like hitting and

using their body weight to keep opposing players away from

the net.

Goalies earn the highest salaries of all positions. We believe

this is because, as previously mentioned, goalies are often

described as playing the most important position in

professional sports. Goalies must constantly adapt to the

Proceedings of the Pennsylvania Economic Association 22

ever-changing nature of the game, adding an additional layer

of difficulty to the position. Furthermore, a majority of

hockey analysts believe that an outstanding goaltender is

paramount to a Stanley Cup Championship. We conclude

that goalies should focus their attention on stopping the puck

because this has implications on the save percentage variable,

which is found to be significant. Moreover, goalies should

continuously work to perfect their skills, in order to gain the

trust of the coaching staff because a higher number of games

played leads to a higher salary.

The findings of this study indicate that NHL franchise

owners seek to maximize profit by signing defensemen who

contribute on the offensive side of the ice, as well as, the

defensive side. Likewise, owners maximize profit by signing

forwards who put a large number of shots on goal, and who

accrue a large number of points per game. Finally, owners

maximize profits through the signing of goalies with a high

save percentage, and a large number of games played. The

results from this study can help predict how much a GM is

willing to spend on a specific player since the ultimate goal is

to create a winning culture without overpaying players.

These kinds of econometric techniques can be applied in

order to help GMs arrive at decisions.

Proceedings of the Pennsylvania Economic Association 23

Table 1: Variables and Expected Signs

Dependent Variable Definition

Salary_ln Ln of salary from 2011-2012 (Defensive, Offensive & Goalies)

Variables Definition Expected Sign

G_ln Ln of Career Goals (All Positions) +

A_ln Ln of Career Assists (All Positions) +

PIMS Penalties in minutes (Defensive) ?

HITS Career hits (Defensive) +

HEIGHT Height of player in inches (Defensive) +

WEIGHT Weight of player in pounds (Defensive & Goalies) +

PPG Career points per game (Offensive) +

SG_LN Ln of Career Shots on goal (Offensive) +

PLUSMINUS Career plus/minus (Offensive) +

DRAFT Dummy variables were used for round in which player was drafted. Value of 1

given to players drafted in first two rounds. All other rounds were given value of

0. (Defensive & Offensive)

+

STAR Dummy variable to indicate whether a player was asked to play in the All-Star

game for the 2010-11 seasons, as well as if they won the Stanley cup. If invited

and/or won, player was given value of 1. If not invited and/or did not win,

player was given a value of 0. (Offensive & Defensive)

+

SAVE_% Career save percentage (Goalies) +

WINS Career wins (Goalies) +

LOSSES Career losses (Goalies) -

LN_GP Ln of career games played (Goalies) +

STARG Dummy variable to indicate whether a player was asked to play in the All-Star

game in 2010-11, if that goalie won the Stanley cup, and if that goalie won the

Vezina and/or Jennings awards. If invited, won, and/or got an award, the player

was given a value of 1. If not invited, did not win, and/or the goalie did not win

an award, the player was given a value of 0. (Goalies)

+

Proceedings of the Pennsylvania Economic Association 24

Table 2.1: Descriptive Statistics – Offensive Players

Variables Mean Standard

Deviation

Maximum Minimum

Offensive Salary 2,500,508 213,840 12,000,000 525,000

Goals 103.3680 104.3502 665.00 1.00

Assists 142.7893 150.7759 988.00 0.00

Shots on Goal 914.4400 824.9344 4766.00 23.00

Points Per Game 0.476410 0.246206 1.403226 0.035714

Plus / Minus 7.970667 46.14261 280.00 -112.00

Observations

375

375

375

375

Table 2.2: Descriptive Statistics – Defensive Players

Variable Mean Standard

Deviation

Maximum Minimu

m

Defense Salary 2,551,294 1,888,990 10,000,000 525,000

Goals 32.62500 37.05319 264.00 0.00

Assists 108.885 109.2558 878.00 3.00

Hits 88.26500 55.14244 270.00 6.00

Penalties in

Minutes

292.2800 314.2796 1809.00 2.00

Height 73.8100 2.113346 81.00 68.00

Weight 210.2450 15.94794 270.00 180.00

Observations

200

200

200

200

Table 2.3: Descriptive Statistics - Goalies

Variables Mean Standard

Deviation

Maximum Minimum

Goalie Salary 2,820,563 2,189,261 10,000,000 570,000

Save Percentage .912 .0057 .926 .900

Wins 157.75 119 664 17

Losses 118.60 86.77 373 16

Games Played 331.90 235.34 1204 44

Weight 198.15 14.745 220 166

Shutouts 23.13 20.2 120 0

Observations

48

48

48

48

Proceedings of the Pennsylvania Economic Association 25

Table 3.1: Defensive Models

Independent

Variables

Model 1 Model 2 Model 3 Model 4

Constant 12.5473

(7.448485)

11.0072

(16.55968)

11.17213

(17.37732) 11.40563

(21.41883)

G_ln 0.1415

(1.284556)

0.1479

(1.347288)

A_ln 0.4173***

(3.404970)

0.4161***

(3.376466)

0.5191***

(10.88916)

0.5010***

(13.85968)

HITS 0.0021***

(3.265746)

0.0022***

(3.619744)

0.00195***

(3.139579)

0.00197***

(3.156443)

PIM -0.0002

(-1.083743)

-0.0002

(-1.107060)

-0.0001

(-0.652943)

HEIGHT -0.0284

(-1.007627)

WEIGHT 0.00795**

(2.077327)

0.0052*

(1.930346)

0.0044

(1.639127)

0.0035

(1.475723)

DRAFT -0.0052

(-0.063005)

STAR 0.3911***

(2.895259)

0.3764***

(2.823645)

0.4209***

(3.320626)

0.4294***

(3.410982)

Adjusted R2

Observations

0.570636

200

0.572124

200

0.557485

200

0.558237

200

Significance Level: ***1% **5% *10%

Parentheses contain t-statistics which are based on White heteroskedasticity-consistent standard errors.

Table 3.2: Offensive Models

Independent

Variables

Model 1 Model 2 Model 3 Model 4 Model 5

Constant 11.2995

(34.45103)

12.4239

(126.1133)

12.4725

(142.0312)

11.4943

(64.83570)

12.4609

(141.3301)

G_ln -0.0735

(-0.8537)

0.1795**

(2.5975)

0.285187***

(7.897713)

0.2844***

(7.867554)

A_ln 0.003

(0.0389)

0.1168

(1.601749)

SG_ln 0.3960***

(3.6114)

0.325294***

(8.899114)

PPG 1.6464***

(7.752388)

1.3784***

(6.539435)

1.4430***

(7.051820)

1.5842***

(8.838350)

1.4814***

(7.728525)

PLUSMINUS 0.0004

(0.561448)

0.0004

(0.533011)

0.0004

(0.5086)

0.0004

(0.555996)

DRAFT 0.0574

(1.086464)

0.0696

(1.260)

0.0697

(1.272083)

0.058739

(1.110243)

0.0677

(1.240681)

STAR 0.1703**

(2.220124)

0.1642**

(2.1005)

0.1619**

(2.051448)

0.16943**

(2.213890)

0.1651**

(2.136421)

Adjusted R2

Observations

0.690912

375

0.681463

375

0.680389

375

0.692272

375

0.680941

375

Significance Level: ***1% **5% *10%.

Parentheses contain t-statistics which are based on White heteroskedasticity-consistent standard errors.

Proceedings of the Pennsylvania Economic Association 26

Table 3.3: Goalie Models

Independent

Variables

Model 1 Model 2 Model 3 Model 4

Constant -27.05846

(-1.912256)

-27.40919

(-1.905393)

-24.75199

(-1.749214) -25.46925

(-1.575011)

SAVE_% 40.75025**

(2.524084)

40.385571**

(2.516581)

38.00340**

(2.365291)

39.26572**

(2.200933)

WINS 0.002028

(1.042287)

0.001983

(0.983937)

0.001601

(0.672149)

0.001970

(1.000411)

LOSSES -0.005684

(-1.500329)

-0.005627

(-1.446286)

-0.005656

(-1.486731)

-0.005550

(-1.438604)

LN_GP 0.861590***

(3.558112)

0.860657***

(3.487055)

0.905895***

(3.920409)

0.788406**

(2.119856)

WEIGHT 0.001313

(0.254666)

STARG

0.165882

(0.468582)

LN_SHUTOUT 0.059181

(0.289365)

Adjusted R2

Observations

0.376399

48

0.362184

48

0.364360

48

0.362201

48

Significance Level: ***1% **5% *10%

Parentheses contain t-statistics which are based on White heteroskedasticity-consistent standard errors.

Proceedings of the Pennsylvania Economic Association 27

REFERENCES

Berri, D.J., & Brook, S.L. (2010). On the evaluation of the

“most important” position in professional sports. Journal of

Sports Economics, 11(2), 157-171.

Chan, T. C.Y., Cho, J.A., & Novati, D.C. (2012).

Quantifying the contribution of NHL player types to team

performance. Interfaces, 42(2), 131-145.

Deutscher, C. (2009). The payoff to leadership in teams.

Journal of Sports Economics, 10(4), 429-38.

Jones, J.C.H., & Walsh, W.D. (1988). Salary determination

in the National Hockey League: The effects of skills,

franchise characteristics, and discrimination. Industrial and

Labor Relations Review, 41(4), 592-604.

Lambrinos, J. & Asham, T. (2007). Salary determination in

the National Hockey League. Is arbitration efficient? Journal

of Sports Economics, 8(2), 192-201.

Marchand, J.T., Smeeding, T.M., & Torrey, B.B. (2006).

Salary distribution and performance evidence from the

National Hockey League. Retrieved from

http://scholar.googleusercontent.com/scholar?q=cache:NPIU

uLCZKJcJ:scholar.google.com/+salary+distribution+and+per

formance+evidence+from+the+national+hockey+league&hl=

en&as_sdt=0,39.

Vincent, C. & Eastman, B. (2009). Determinants of pay in

NHL: A quantile regression approach. Journal of Sports

Economics,10(3), 256-277.

Vincent, C. & Eastman, B. (2012). Does player mobility lead

to higher earnings? Evidence from the NHL. The American

Economist, 57(1), 50-64.

Zimbalist, A. (2010). Reflections on salary shares and salary

caps. Journal of Sports Economics, 11(1), 17-28.

Proceedings of the Pennsylvania Economic Association 28

PENNSYLVANIA TAX SIMPLIFICATION:

NUISANCE TAX CREDIT, OBSOLETE TAXATION AND ADMINISTRATION PROVISION REPEALS,

INCLUDING PROPER PLACEMENT WITHIN THE TAX REFORM CODE

Thomas O. Armstrong*

Commonwealth of Pennsylvania,

Department of Revenue, Harrisburg, PA 17128

ABSTRACT The optimal taxation literature argues that successful tax

reform reduces economic losses to society where resources

can be transferred to higher valued uses. Taxing jurisdictions

can reduce inefficiencies by enacting tax simplification

processes resulting in administrators and taxpayers spending

fewer resources to administer taxes. Tax simplification gains

can be achieved by repealing nuisance tax credits, obsolete

taxation provisions, and obsolete administrative provisions as

well as placing relevant law within the tax code. The Corbett

Administration and the General Assembly in recognizing the

benefits of tax simplification have enacted for Fiscal Year

2013-14 tax reform measures including tax simplification

repeals.

I. INTRODUCTION

1

One purpose of taxation is to raise revenue to finance

government expenditures. Once a set of taxes are in place,

economic inefficiency results (Weimer and Vining, 1992).

From an optimal taxation perspective, minimizing

inefficiencies or deadweight losses of taxes relative to

benefits received will result in greater availability of

resources to enhance economic growth and development

(Slemrod, 1990; Scully, 1992).2

The optimal taxation literature places the administrative costs

of taxation into its research agenda when considering optimal

taxation possibilities (Slemrod, 1990). Tax simplification is

aligned with the classical objective of proper taxation policy

due, in part, to its efficiency objectives. With fewer

resources expended to administer taxes, greater resources are

readily available for more efficient market activities.

Successful tax simplification is when taxes are easy to

understand with relatively low compliance and administrative

costs, which promote a high degree of voluntary compliance

among taxpayers. Relatively low administration and

compliance costs with a simplified tax system occurs when

the system has relatively few taxes; simple tax laws; a limited

number of rates for each tax; limited exemptions, credits,

rebates or deductions; a broad base; and limited use of the tax

system to achieve not too many social goals (Handbook for

Tax Simplification, 2009). A higher degree of voluntary

compliance results in more returns processed, lower

administrative costs, and a higher degree of economies of

scale.

Gale and Holtzblatt (April 2002) consider complexity of a tax

system as the sum of compliance costs, incurred directly by

individuals and businesses, and administrative costs, incurred

by government. Compliance costs can include the time and

expenditures taxpayers spend preparing and filing tax forms;

researching the law to determine the application and if the tax

code is still in effect; maintaining records; taxes prepared by

third party preparers; respond to audits; and resources to

avoid or evade taxes.

Administrative costs include the resources allocated to the

tax agency, and resources required of other agencies to help

administer tax programs. It should be noted while

administrative costs are incurred by the government; it is

ultimately borne by taxpaying individuals along with the

compliance costs. Effective tax simplification results in less

compliance and administrative resources for collection of

certain revenue.3

The Corbett Administration and the General Assembly have

proposed and enacted a number of tax simplification

measures for Pennsylvania Fiscal Year (FY) 2013-14: Act 52

of 2013 (HB 465, PN 2211, Omnibus Tax Reform Code) and

Act 71 of 2013 (SB 591, PN 1328; Omnibus Fiscal Code).

The expectation is resources currently used to comply with a

complex state tax code will now be used in more efficient

economic activities as the result of a more simplified tax

code.4

II. TAX SIMPLIFICATION

Tax complexity increases the overall cost of taxation as well

as increases the likelihood that taxpayers make inadvertent

mistakes in calculating their tax liabilities (Kopczuk, 2006).

Other “…commonly recognized and negative effects of

complexity are (1) decreased levels of voluntary compliance

stemming from taxpayer confusion; (2) increased costs of

compliance for taxpayers; (3) reduced perceptions of fairness

in the …tax system; and (4) increased difficulties in the

administration of tax laws,” (Joint Committee on Taxation,

p.101, April 2001).5 Tax simplification refers to reducing

impediments or complexity of tax processes.

Proceedings of the Pennsylvania Economic Association 29

The goal for tax simplification is to increase the ease of

compliance, reflected in the costs and time saved for

taxpayers and tax administrators (Handbook for Tax

Simplification, 2009). The positive result is more resources

are available for other economic activities, thereby reducing

economic inefficiency (Final Report of the Pennsylvania Tax

Commission, March 1981).

The Joint Committee of Taxation Report (April 2001)

identified various sources of tax complexity relevant to

Pennsylvania’s FY 2013-14 tax changes that will reduce

complexity, thereby enhancing simplification. Three

provisions that have added to complexity within the

Pennsylvania Tax Reform Code of 19716 and the Fiscal

Code7 are 1) transactional complexity, 2) obsolete provisions

complexity, and drafting comlexity.8

“Transaction complexity” refers to the extent that tax laws

complicate the planning and execution of transactions by

taxpayers. State nuisance tax credits that increase the

complication of transactions and administrative costs relative

to few or no beneficiary recipients would fall into this

category of complexity.

“Obsolete provisions complexity” is the numerous

superseded or invalidated tax provisions within

Pennsylvania’s Tax Reform Code and Fiscal Code. While

most taxpayers are unaffected by the obsolete provisions,

these provisions require time and resources to determine

whether a particular provision has continuing applicability.

The tax and fiscal code bills repeal many obsolete taxation

and administration provisions contributing to better resource

usage as the result of tax simplification.

“Drafting complexity” occurs when a tax law is misplaced

within Pennsylvania’s statutes, where the time to locate and

associate the law can take an unnecessarily excessive time

period. The omnibus tax reform and fiscal code bills, Act 52

and Act 71 of 2013 remove tax code language from the fiscal

code and places the language within the Pennsylvania’s Tax

Reform code of 1971.

Tax simplification initiatives from a tax reform/budget focus

were last enacted in 2001 (Armstrong and Brehouse, 2001).

Governor Corbett proposed and the General Assembly passed

into law tax simplification initiatives of repealing state

nuisance tax credits and obsolete taxation and administration

provisions as part of Governor’s budget and tax reform

initiatives announced February 5, 2013.9,10

The tax

simplification measures are projected to have a minimal

fiscal impact of $500,000 out of a General Fund enacted

budget of $28.375 billion. The efficiency gains to the

Commonwealth are expected to be greater than the nominal

revenue loss.

III. STATE NUISANCE TAX CREDIT REPEALS

State nuisance tax credits are an accumulation of many prior

legislative session enactments, where the original intent of

certain tax credits may no longer be valid, no longer

consistent with current policies, or have very few recipients

where the benefits are less than the economic costs of the tax

credits. Upon reviewing the Commonwealth of

Pennsylvania’s taxation statutes, the determination was made

that the following four taxation provisions caused

unnecessary taxpayer and administrative burdens relative to

the little or no economic benefits by these state nuisance tax

credits where the effective repeal date for all the tax credits

listed below is July 1, 2013:11

Coal Waste Removal and Ultraclean Fuels Tax

Credit. This Credit was enacted in 1999 in the Tax

Reform Code (Article XVIIII-A) and available for

certain capital expenditures for companies that

produce synthetic fuels from coal, culm, or silt.

This credit is capped at $18 million per year. The

Credit can be used against sales and use tax,

corporate net income tax, and capital stock/foreign

franchise tax. From the Legislative Budget and

Finance Committee report (June 2010) and

Department of Revenue, no eligible developer has

applied for, or claimed, this tax credit. While

nominal, the Department of Revenue incurs the cost

of maintaining the administration of this credit.

Call Center Tax Credit (CCTC) Program. The

CCTC program was set up to entice call centers to

relocate into the state. $30M was set aside for this

program. A tax credit is available to call centers for

the sales and use tax paid on incoming and outgoing

interstate telecommunications. Beginning in on or

after January 1, 2004, the credit is equal to the gross

receipts tax paid by a telephone company on the

receipts derived from the incoming and outgoing

interstate telecommunications. This program has

been underutilized since the inception and the

number of applicants has decreased over the last

several years. While nominal, the Department of

Revenue incurs the cost of administering this

underutilized credit. Between 2005 and 2009, $6.4

million in tax credits (out of $30 million available

capped annually) were approved and awarded to 51

call centers applying for the credit. For 2010, 12

call centers applied and were approved for refunds

totaling $379,910 and for 2011, 9 call centers

applied and were approved for refunds totaling

$547,903. Currently, only 9 call centers applied for

2012 totaling $655,315 of the available $30M.

Proceedings of the Pennsylvania Economic Association 30

IV. OBSOLETE TAXATION AND ADMINISTRATIVE

PROVISIONS INCLUDING PROPER PLACEMENT

WITHIN THE TAX REFORM CODE

A review of the Commonwealth of Pennsylvania statutes has

indicated that a number of obsolete provisions exist that

should be repealed. It is expected that the repeal of these

obsolete provisions as well as proper placement of statute

within the Tax Reform Code will reduce the time and

resources expended by taxpayers to become familiar with the

Commonwealth’s applicable taxation laws. The obsolete

provisions can be classified into two categories: 1) Obsolete

Taxation Provisions, and 2) Obsolete Administrative

Provisions.

Obsolete Taxation Provisions Repeals—These taxes have

been superseded by other legislation or have been declared

unconstitutional by a court, yet the legislation remains in

Pennsylvania statutes. The effective repeal date for all the

obsolete taxation provisions is the date of enactment. There

is no fiscal impact for repealing the obsolete taxation

provisions.

Treasurer Report of Municipal Loans (1929 0

P.L.343 A176 §709). This law in the fiscal code

requires the treasurer to submit to the Department of

Revenue an annual report of municipal loans and

pay appropriate taxes. This was enacted April 4,

1929. This law is no longer actively imposed or

applied and is repealed. This repeal is part of the

Department of Revenue's recommendation to the

Local Government Commission concerning state

mandates on local governments.

Registers to File Monthly Inheritance Tax

Statements (1929 0 P.L.343 A176 §724). This law

in the fiscal code requires the Register of Wills to

file monthly statements of inheritance taxes paid.

This was enacted April 4, 1929. This law is no

longer actively imposed or applied and is repealed.

This repeal is part of the Department of Revenue's

recommendation to the Local Government

Commission concerning state mandates on local

governments.

County Treasurers Receive Use Tax (1971 0 P.L.6

A2 §226). This provision in the Tax Reform Code

designated county treasurer as receiver of use taxes

from any person other than a licensee and sets

procedure for transmittal to the Department of

Revenue. This was enacted March 4, 1971. This

law is no longer actively imposed or applied and is

repealed.

Inheritance Tax Poverty Exemption (section 2112).

This provision in the Tax Reform Code provides an

exemption from Inheritance Tax for transfers of

property for certain classes of persons, spouses.

This law is no longer actively imposed since June

30, 1995, and is repealed, due to being superseded

by an exemption for transfers to spouses.

Obsolete Administrative Provisions Repeals—These

provisions are non-functioning tax related administration

laws that remain in Pennsylvania’s statutes and should be

repealed. Repealing these provisions will reduce time and

resources expended by taxpayers to determine whether the

non-functioning tax related administrative laws are no longer

applicable. The proposed effective repeal date for all the

obsolete administrative provisions is the date of enactment.

There is no fiscal impact by repealing the obsolete taxation

provisions.

Motor License and Vehicle Operators’ License Fees

(72 P.S. §1206). This is an administrative provision

directing the Department of Revenue to collect fees

for the registration and titling of vehicles, and to

perform licensing and titling functions. The

Department of Revenue’s powers related to the

registration and titling of motor vehicles were

transferred to the Department of Transportation,

effective July 1, 1970 (Act of July, 1970 (P.L. 356,

No. 120)). Despite this transfer of responsibility,

this provision was never repealed and thus remains

in the Fiscal Code. Because the Department of

Revenue is no longer responsible for these

functions, this provision does not serve a function.

Collection of Amounts Payable to State Institutions

(72 P.S. §1209). These sections require the

Department of Revenue to place its agents in every

state institution for the purpose of collecting moneys

due the institution for any expenses accrued on

account of patients, pupils or inmates. This

procedure is not operative. The second paragraph of

section 210 of the Fiscal Code contains a similar

provision (72 P.S. §210).

Proper Placement Within the Tax Reform Code—The proper

placement of tax legislation will reduce the time and

resources expended by taxpayers to become familiar with the

Commonwealth of Pennsylvania’s applicable taxation laws.

There is no fiscal impact of transferring exactly existing

taxation language into the Tax Reform Code.

Neighborhood Improvement Zones, NIZ (Act 26 of

2011 and Act 50 of 2009, omnibus fiscal code

legislation). The Acts authorized a city of the third

class to designate a NIZ for the purpose of

improvement and development of a Zone and to

construct a facility or facility complex within the

Zone. Certain state and local taxes attributed within

Proceedings of the Pennsylvania Economic Association 31

the NIZ is allocated to the NIZ authority for

development.

Keystone Special Development Zones, KSDZ (Act

26 of 2011, omnibus fiscal code legislation).

Creates a new program for the designation of a

KSDZ for parcels of real property certified as

Special Industrial Areas by the Department of

Environmental Protection pursuant to the Land

Recycling and Environmental Remediation

Standards Act, and which as of July 1, 2011 had no

permanent vertical structures affixed to it. The Act

provides a tax credit for employers within a KSDZ

for new full time jobs created in the zone.

V. CONCLUSION

Tax complexity increases the cost in the form of time or

money imposed on taxpayers and tax administrators to

comply with the tax law. Tax complexity can enhance tax

evasion and avoidance that reduce economic efficiency. The

preferred method in reducing tax complexity is to correct the

tax code. The Corbett Administration and the General

Assembly has introduced and enacted tax simplification

measures into law that will improve the administration of

taxes for taxpayers and the Commonwealth. As a result,

efficiency gains can be expected.

ENDNOTES

* The author would like to thank Sally Fishel and discussant

for their assistance and comments. The conclusions do not

necessarily reflect the positions of the Pennsylvania

Department of Revenue. All possible errors are the author’s.

1. Information for the next two sections are generally from

Armstrong and Brehouse (2001), the Report (April 2001),

and Gale and Holtzbatt (April 2002). The Report and Gale

and Holtzbatt suggest that there is no general consensus on

the appropriate method of measuring the effects of

complexity; while recognizing that there is general agreement

that complexity has adverse economic effects.

2. It is recognized that there are other taxation policy

objectives. Some of the classic objectives are adequacy,

neutrality, equity, accountability, and ease of administration

(Armstrong, 2002).

3. Kopczuk (2006) argues that simplification can reduce tax

evasion and avoidance more efficiently than traditional

enforcement measures of costly detection and penalty

compliance administration.T ax avoidance is a reduction by

legal means to reduce the amount of tax that is payable whilst

making a full disclosure of the material information to the tax

authorities. Examples of tax avoidance involve using tax

deductions or changing one's business structure through. By

contrast, tax evasion is a reduction by illegal means to reduce

or evade the payment of taxes. Tax evasion usually entails

taxpayers deliberately misrepresenting or concealing the true

state of their affairs to the tax authorities to reduce their tax

liability, and includes dishonest tax reporting such as under-

declaring income, profits or gains; or overstating deductions.

4. Kopczuk (2006) argues that by reducing tax complexity,

taxpayers will be more responsive to changes in taxation.

5. It should be noted that it is not the overall level of

complexity within a tax system but the costs and benefits of

taxes including the degree of complexity relative to the

advancing of policy goals. The Handbook for Tax

Simplification, (2009) provides a list and rationale for

reasons that tax complexity arises.

6. The Pennsylvania Tax Reform Code of 1971 (Act of Mar.

4, 1971, P.L. 6, No. 2 Cl. 72), codifies and enumerates

certain subjects of taxation and imposing taxes thereon;

providing procedures for the payment, collection,

administration and enforcement thereof; providing for tax

credits in certain cases; conferring powers and imposing

duties upon the Department of Revenue, certain employers,

fiduciaries, individuals, persons, corporations and other

entities; prescribing crimes, offenses and penalties.

7. The Pennsylvania Fiscal Code (Act of Apr. 9, 1929, P.L.

343, No. 176 Cl. 72) includes relating to the finances of the

State government; providing for the settlement, assessment,

collection, and lien of taxes, bonus, and all other accounts

due the Commonwealth, the collection and recovery of fees

and other money or property due or belonging to the

Commonwealth, or any agency thereof, including escheated

property and the proceeds of its sale, the custody and

disbursement or other disposition of funds and securities

belonging to or in the possession of the Commonwealth, and

the settlement of claims against the Commonwealth, the

resettlement of accounts and appeals to the courts, refunds of

moneys erroneously paid to the Commonwealth, auditing the

accounts of the Commonwealth and all agencies thereof, of

all public officers collecting moneys payable to the

Commonwealth, or any agency thereof, and all receipts of

appropriations from the Commonwealth, authorizing the

Commonwealth to issue tax anticipation notes to defray

current expenses, implementing the provisions of section 7(a)

of Article VIII of the Constitution of Pennsylvania

authorizing and restricting the incurring of certain debt and

imposing penalties; affecting every department, board,

commission, and officer of the State government, every

political subdivision of the State, and certain officers of such

subdivisions, every person, association, and corporation

required to pay, assess, or collect taxes, or to make returns or

reports under the laws imposing taxes for State purposes, or

to pay license fees or other moneys to the Commonwealth, or

any agency thereof, every State depository and every debtor

Proceedings of the Pennsylvania Economic Association 32

or creditor of the Commonwealth (title amended June 21,

1984, P.L.407, No.83).

8. The Report (April 2001) identifies other sources of

complexity including a general level of complexity:

computational complexity. This refers to the complex

calculations to determine tax liability. The greater the

complex calculations, the greater resources required such as

hiring tax professionals or purchase software to assist in

preparation of liability.

9. Act 72 of 2013 repeals Corporate Loans Tax (CLT)

effective January 1, 2014. The treasurer of the corporation

assesses and withholds the CLT from interest paid at a rate of

4 mills on the nominal value of scripts, bonds, certificates,

and other evidence of indebtedness. Banks pay it as well as

other corporations, and LLCs are included. Technically, the

issuer of a bond is withholding this 4 mill tax for resident

individuals who earn the interest on the bond, but individuals

never have to pay it directly. Non-PA corporations only pay

it if they have their treasurer located in PA. They only have

to withhold the tax on interest paid to PA residents. The CLT

adds a lot of complexity to tax administration for a small

amount of revenue. A full fiscal year impact ranges between

about $11 to $14 million impacting about 7,000 taxpayers.

By repealing the CLT, tax simplification for taxpayers and

tax administrators is increased.

In addition, Act 52 of 2013 enacts Pass Through Business

Compliance legislation, effective January 1, 2014, where one

of the provisions is to authorize the assessment of tax at the

entity level for pass-through entities such as partnerships,

LLCs and S corporations. The change will simplify tax

administration and tax compliance. It is more cost effective

to issue one assessment to a large partnership than to

separately bill hundreds of partners for small amounts. One

assessment will create one appeal as opposed to hundreds of

appeals on the same issue. The provision will only apply if

the partnership as a whole understates income by at least $1

million, and has at least 11 individual partners. It does not

make any partner liable for another partner’s tax.

10. The omnibus tax reform code legislation, Act 52 of 2013,

includes the extension of the Capital Stock and Foreign

Franchise Tax (CSFT) phase-out. The CSFT is currently at

0.89 mills. The CSFT will be phased-out at 0.67 mills on

January 1, 2014; 0.45 mills on January 1, 2015; and repealed

on January 1, 2016. The CSFT is calculated as an average of

net income and net worth of a firm. From a tax

simplification perspective, compliance and administration

costs will be eliminated for 101,000 taxpayers and the state

government administrators after repeal (Armstrong and

Brehouse, 2000).

11. Governor Corbett proposed in his budget address on

February 5, 2013, for repealing the following additional state

nuisance tax credits enacted as free standing acts separate

from the tax reform and fiscal codes:

Alternative Energy Production Tax Credit (73 P.S. §

1649.701 et. seq.). Act 1 of 2008, Special Session Number 1,

created the Alternative Energy Production Tax Credit.

Taxpayers that develop or construct energy production

projects located within the Commonwealth, which have a

useful life of at least four years, may apply to the Department

of Environmental Protection for a tax credit beginning

September 2009. The total amount of the tax credit that can

be awarded is from $2 million to $10 million per fiscal year.

Act 2009-48 reduced the annual credits available to $0 for

Fiscal year (FY) 2009-10 and FY 2010-11. Afterwards, the

credit reverts back to the previous total amounts. Beginning

for FY 2011 and onwards, no taxpayers have received this

credit. This credit is generally not used. The costs to

administer this credit are borne by the Departments of

Environmental Protection and Revenue. As present, HB

1171 to repeal this Credit was voted out of the House Finance

Committee.

Organ and Bone Marrow Donor Tax Credit (35 P.S. § 6120.1

et, seq.). The freestanding Organ and Bone Marrow Donor

Tax Credit, Act 65 of 2006, provides for a tax credit for

expenses incurred when a business firm grants to any of its

employees a paid leave of absence for the purpose of

donating an organ or bone marrow. The Tax Credit expired

after tax year 2010. The economic rationale for repealing

this tax credit is the savings in time and resources expended

by taxpayers to determine the tax credit is no longer

applicable. Currently, no bill exits to repeal this credit.

REFERENCES

Armstrong, Thomas. 2002. “State Taxation Reform Proposals

for Pennsylvania,” Pennsylvania Economic Association

Proceedings, 1-10.

Ibid and Jason R. Brehouse. 2001. “State Tax Simplification:

State Nuisance and Obsolete Provision Tax Repeals,

Including Proper Placement within the Tax Reform Code,”

Pennsylvania Economic Association Proceedings, 171-175.

Ibid. 2000. “Capital Stock and Franchise Tax Phase-Out

Initiative for Pennsylvania,” Pennsylvania Economic

Association Proceedings, 18-28.

Final Report of the Pennsylvania Tax Commission. March

1981. Pennsylvania Tax Commission.

Gale, William G. and Janet Holtzblatt. April 2002. The Role

of Administrative Issues in Tax Reform: Simplicity,

Compliance, and Administration. In United States Tax

Reform in the Twenty-First Century, George R. Zodrow and

Peter Mieszkowski (eds.) Cambridge University Press.

Proceedings of the Pennsylvania Economic Association 33

Handbook for Tax Simplification. 2009. Washington, DC:

International Finance Corporation.

Kopczuk, Wojciech. 2006. In Max Sawicky (ed.), Bridging

the Tax Gap. Addressing the Crises in Tax Administration,

Washington, DC: Economic Policy Institute, 111-143.

Joint Committee on Taxation. April 2001. Study of the

Overall State of the Federal Tax System and

Recommendations for Simplification, Pursuant to Section

8022(3)(B) on the Internal Revenue Code of 1986.

Washington: U.S. Government Printing Office.

Pennsylvania’s Tax Credit Programs. June 2010. Legislative

Budget and Finance Committee. Harrisburg, PA.

Scully, Gerald, W. 1992. Constitutional Environments and

Economic Growth. Princeton, N.J.: Princeton University

Press.

Slemrod, Joel. Winter 1990. “Optimal Taxation and Optimal

Tax Systems.” Journal of Economic Perspectives. 4:1, 157-

178.

Weimer, David L. and Aidan R. Vining. 1992. Policy

Analysis: Concepts and Practice. Englewood Cliffs, N.J.:

Prentice-Hall, Inc.

Proceedings of the Pennsylvania Economic Association 34

BUSINESS AND REAL-ESTATE PRICE CYCLES ACROSS THE US:

EVIDENCE FROM A MARKOV-SWITCHING REGRESSION EXERCISE

Aram Balagyozyan

The University of Scranton,

Scranton, PA 18510

Christos Giannikos

Baruch College,

One Bernard Baruch Way, New York, NY 10010

Kyoko Mona

Manhattanville College,

2900 Purchase St., Purchase, NY 10577

ABSTRACT

This study examines whether house price cycles lead or lag

business cycles in state-level US data from 1979 to 2012.

Using a Markov-switching model, we test various lead/lag

scenarios in every US state as well as the aggregated US. For

the majority of states, we reject the hypothesis that house

prices did not lead the economy. Between 2002 and 2011,

house prices led the economy nationally as well as in twenty-

two US states. However, the co-evolution of the two series is

driven by factors that are specific to geography and time.

INTRODUCTION

Business and real-estate downturns can be devastating to the

welfare of households and investors. The US recession of

2007 was a bitter reminder of the strong ties between real

estate markets and the broader economy. The idea that these

two sectors are closely related is not new; scholars have been

actively investigating the nature of this relationship since the

eighteenth century.

It has generally been found that housing cycles lead

economic cycles at the city, state, and national levels (Case et

al. (2000), Iacoviello(2005), Leamer (2007), Ghent and

Owyang (2010), Strauss (2013)). Broadly speaking, the

literature offers four explanations for this observation. First,

the housing market may be a proxy for another, more

important variable that leads economic cycles. For example,

Strauss (2013) suggests that improvements in consumers’

expectations of future income may forecast improvements in

both housing and the general economy. If housing reacts to

these expectations faster than the economy, then it would

predict the economy. The second explanation casually

connects housing to the broader economy through residential

construction. When the housing market picks up, so does

employment in the construction industry. Rising incomes in

the construction industry are transferred with a multiplier to

the GDP and employment. The third explanation is also

casual, but operates through the credit channel. When the

housing market weakens, so does the strength of the lenders’

balance sheets. Foreclosures and mortgage defaults

undermine the financial strength of the banking sector,

causing the banks to tighten their lending standards, which in

turn weakens the consumer demand for goods and services

that are financed by credit. We suspect that the systemic

impact of this channel became stronger with the

securitization of the credit market. Finally, there is the wealth

effect. As real estate appreciates, homeowners become

wealthier. According to the life-cycle theory of consumption,

they spend more and thus increase equilibrium output and

employment.

While the wealth and credit effects of housing on aggregate

expenditure are the two most widely quoted channels, there is

no clear consensus about their validity and magnitude. Case

et al.(2005) studied a panel of US states and found that the

housing wealth effect has an important effect upon

consumption. They also found that the wealth effect of

housing exceeds that of the stock market. Muellbauer (2007)

studied a cross-section of countries and found that the wealth

and credit effects are strong everywhere but stronger in the

US and UK. Moreover, the wealth effect in the US became

much stronger after liberalization of the housing market.

However, despite such evidence there are good reasons to

argue against this interpretation of data. If the wealth and

credit effects are truly valid and significant, then they must

work through changes in real-estate prices. But real-estate

prices are often considered rigid in the downward direction,

and for this reason, many authors reject their ability to lead

the broader economy. Leamer (2007), for example, shows

that while national housing indicators such as residential

investment and volumes are powerful predictors of business

cycles, house prices are not. Ghent and Owyang (2010) and

Strauss (2013) reach similar conclusions for US cities and

states respectively. Hence, the results of these authors

indirectly imply that the wealth and credit effects are weak

and possibly invalid

In this paper we use a Markov-switching vector

autoregressive model to investigate the coevolution of

business cycles and house price cycles in the cross-section of

all US states as well as nationally. Our findings suggest that

at the national level, house prices sometimes lead and

Proceedings of the Pennsylvania Economic Association 35

sometimes lag business cycles. Moreover, in line with the

conclusions of Leamer (2007), Ghent and Owyang (2010),

and Strauss (2013) we do not find any consistent relationship

between state business and house price cycles. Yet, for the

overwhelming majority of US states as well as the aggregated

US, we fail to reject the hypothesis that house prices do not

lead the economy. We also find that the deterioration of each

state’s business conditions in late 2007 and early 2008 was

preceded by a decline in house prices in nearly half of US

states. Although the national recession of 2007 started

sometime after a downturn in real-estate prices, the housing

market recovery lagged the economic recovery by at least

two years. These results imply that even though we fail to

observe any systematic patterns in which house prices lead

state or national economies, neither can we prove that they do

not. Hence, the credit and wealth effects of housing on

consumption cannot be written off as weak or even non-

existent; they remain legitimate subjects for further

investigation.

The remainder of the paper proceeds as follows: the

following section describes the Markov-switching model and

hypothesis testing methodology. We describe our data in the

Data section. The Results section explains our findings, and

we offer some concluding remarks in the Conclusion section.

MODEL AND METHODOLOGY

In order to investigate the joint dynamics and turning points

of house prices and the economy, we rely on the

methodology employed by Hamilton and Lin (1996) and

Smith et al. (2000). The US business cycle turning points are

traditionally dated by the National Bureau of Economic

Research. On the other hand, a vast body of literature

investigates regime changes in time series using the Markov-

switching autoregressive model of Hamilton (1989). Our

model belongs to the latter category. Specifically, we assume

that the economy and housing prices in a given region evolve

according to the following vector Markov-switching model

with no autoregressive dynamics:

[

] [

] [

] (1)

The variables and represent the growth rates of the

economy and housing prices respectively. The innovation

terms and are assumed to be jointly normally

distributed, with zero means and time-invariant correlation

coefficient . Their variance-covariance matrix therefore has

the following form:

[

] (2)

The coefficients and are the mean growth rates of the

economy and house prices respectively. These coefficients

are assumed to be time varying and subject to discrete

switches between two regimes: low ( ) and high ( ). The

superscripts refer to the regime . It follows that

the growth rates of the economy and house prices in the

region jointly assume four regimes:

(3)

Hence, the vector of means,

, may also assume four

regimes. The individual and joint regimes are unobservable,

but we assume that they follow a Markov process.

Since forecasting is not our primary objective, we do not

include any autoregressive or moving average components in

the model. We are only interested in the mean coefficients

and . Since the asymptotic estimates of these coefficients

remain unbiased and are identical to those estimated under

more sophisticated linear specifications, we use the most

parsimonious model.

Shifts between different regimes are governed by a

transition probability matrix, which we estimate along with

the other parameters of the model (1)-(3). The matrix is:

[

] (4)

where , is the transition

probability that the regime was in the previous period and in the subsequent period. We assume that the matrix is time-

invariant. Each column must add to unity. Hence, when

estimating the model only twelve probabilities are left as free

parameters. A single element in each column is equal to one

minus the sum of the remaining elements in the column.

Using the procedures described in [8], [10], and [11] we can

obtain maximum likelihood estimates of the following

regression parameters:

(5)

as well as the transition probabilities in (4). Thus, the

unrestricted model (1) has a total of 19 free parameters.

When estimating the model, we enforce the constraints

and .

In principle, the transition probability matrix (4) should

describe all aspects of the joint dynamics of the housing

market and economy. A useful way of thinking of these

Proceedings of the Pennsylvania Economic Association 36

relationships is through the classification of shocks that may

affect both the economy and the housing market. The

transition probabilities reflect five types of shocks. First,

there may be no shocks. In this case inertia dominates both

sectors, so the regimes do not change. If regime-changing

shocks are rare in the sample, we expect to estimate

significant values of the probabilities on the principal

diagonal of the transition probability matrix, , , ,

and . Second, sterile shocks, that affect one sector but not

the other. If these type of shocks is dominant we expect to

estimate significant values of , , , and . Third,

systemic shocks affecting both sectors simultaneously,

although not necessarily in the same direction. If this type of

shocks is dominant we expect to estimate significant values

of probabilities along the minor diagonal of the transition

probability matrix, , , , and . Fourth, shocks

affecting the economy first, and housing prices after some

delay. If this type of shocks are dominant, then real-estate

prices would tend to follow the economy. In this case, we

would expect to estimate significant values of and .

Finally, shocks affecting housing first, and the general

economy with some delay. If this type of shocks are

dominant, then the economy would tend to follow real-estate

prices and we would expect to estimate significant

probabilities and .

The above interpretations imply that if housing prices in a

region tend to lead the economy then we should be able to

reject hypothesis : . Rejection of this

hypothesis implies that we are unable to reject the wealth and

credit effects of housing. Similarly, the rejection of

hypothesis , implies that the economy

in a given region may lead housing prices. Although the

rejection of hypothesis does not lead to any conclusions

about the wealth and credit effects, it would shed light on

how consistently house prices in a region lead the economy.

If for a given region both hypotheses ( and ) are rejected,

then we can conclude that both leading relationships exist in

the data, although not at the same time. On the other hand, if

we fail to reject either hypothesis, then neither sector leads

the other. For example, this may happen when housing and

business cycles coincide. Based the outcomes of tests and

, we can place each state into one of the four categories

. These four categories are summarized in Table 1.

Note that the number of states for which we reject (or not)

any of the two hypothesis can be obtained by combining the

states fallen in two categories along the rows or columns of

Table 1. If, for example, we are interested in obtaining the

states for which only hypothesis was rejected, we can

combine states which fall in categories 1 or 2.

We use the likelihood ratio test to determine the significance

of transition probabilities and reject (or not) hypotheses

and for each individual US state as well as the US as a

whole. Since each test imposes two restrictions on the

transition probability matrix, the likelihood ratio statistics has

a distribution with two degrees of freedom. We test these

hypotheses at significance level.

DATA

In order to estimate equation (1), we rely on two monthly

indicators from January 1979 to September 2012. First, is the

monthly coincident index of economic activity, which

proxies for the level of economic activity in a each US state

and the US as a whole. This seasonally adjusted index is

compiled by the Federal Reserve Bank of Philadelphia and

combines four state-level economic indicators: non-

agricultural payroll employment, unemployment rate,

average hours worked in manufacturing industries, and real

wage and salary disbursement.

Our second data set is the Freddy Mac housing price index,

which reflects the average level of real-estate prices in each

state. The Freddie Mac index includes valuation and location

data, and is based on the combined portfolio of loans that

were purchased by either Freddie Mac or Fannie Mae since

January 1979. The portfolio of loans covers every state,

although it reflects Freddie Mac and Fannie Mae’s collective

market coverage and thus is not random across states.

Furthermore, the loans are limited to one-family detached

and town-home properties financed by first-lien conventional

and conforming loans. We seasonally adjust the housing

price index using the Eviews (7th version) seasonal

adjustment utility which utilizes the U.S. Census Bureau’s

X12 seasonal adjustment program.

RESULTS

For each US state as well as the aggregated US, we estimate

the Markov-switching vector autoregressive model (1).

Parameter estimates for the unrestricted model are presented

in Table 2. Following this exercise we test hypotheses and

, and based on the outcome of the tests, place each region in

one of the four categories as described in the Model and

Methodology section. The results of this exercise are

presented in the first column of Table 2. When the entire

sample between 1979 and 2012 is used, we reject both

hypotheses and

for 40 out of 50 states. This implies

that for 80% of all US states, there is statistical evidence that

housing price cycles in the region may both lead and lag

business cycles at different times. Hence, consistent with

Leamer (2007), Ghent and Owyang (2010), and Strauss

(2013), we cannot establish that house prices are reliable

leading indicators of state business cycles. On the other hand,

since there are no states for which both hypotheses are

sustained, we can conclude that economic and real-estate

price cycles also do not have a consistent tendency of

evolving independently. North Dakota and Hawaii are the

only states where house prices tend to lead business cycles.

Proceedings of the Pennsylvania Economic Association 37

ND is largely an agricultural state, and in this case the test

decision seems to be dominated by a single housing recovery

that pre-dated an economic recovery in the early 80s. In eight

other largely agricultural and oil-producing states, real-estate

prices follow business cycles but not the other way around.

The aggregated US falls in Category 1. Thus, national real-

estate price cycles and business cycles each have some

probability of leading the other as suggested by the cross-

correlation analysis. Interestingly, for the US, the -value of

hypothesis is 0.008. Therefore at the 1% significance level

we would reject hypothesis only marginally. This implies

that the data are fairly close to placing the US as a whole

(along with Hawaii and North Dakota) in the 3 category,

meaning that real-estate prices lead the economy but not the

other way around.

Strauss (2013) suggests that during the US recession of 2007,

the worsening economy in most US states was preceded by a

decline in building permits. We are interested in whether the

pre-recession period was also accompanied by a decline in

house prices. If house prices are downward rigid, as for

example Gao et al. (2009) suggest, then the housing decline

prior to the recession of 2007 must have impacted the

economy through channels other than the wealth and credit

effects. This would obviously go against conventional

wisdom. We re-estimate the model and test hypotheses and

for the sub-sample starting in January 2002, right after the

end of the national recession in 2001, and ending in

September 2012. The results of the test for this sub-sample

are presented in the second column of Table 2. Although

Category 1, in which housing prices and business cycles may

each lead the other, remains the most populous (19 states),

the number of states in which house prices lead business

cycles (Category 3) increased from 2 to 10. The US as a

whole remained in Category 1, but it is again a borderline

case, close to the Category 3 region in which real-estate

prices lead the economy but not the other way around.

Category 2, in which the economy leads house prices but not

vice versa, now consists of 17 mostly agricultural and oil-

producing mid western and southern states

Being intrigued by the small number of states in Category 3

(house prices lead economies but not vice versa), we note

that our sample may be influenced by the recent signs of

housing recovery that started appearing in early 2012, more

than two years after the national recovery. Hence we

conjecture that some states ended up in Category 1 solely on

the basis of these recent data, but should be in Category 3 if

we focus the analysis on years prior to and during the

recession. That is, if the housing downturn in a region started

before an economic downturn and ended after an economic

recovery, we are likely to (properly) conclude that house

prices both lead and lag the economy, and label this region as

Category 1. However, if we what to answer the specific

question of whether house prices led the economy before the

recession of 2007, we should exclude the post-recession

housing recovery from our sample. To restrict our viewpoint

to the years prior to and during the recession of 2007, we

consider a second, shorter sub-sample running from January

2002 to September 2011. The test results based on this sub-

sample are presented in the third column of Table 2. Between

January 2002 and September 2011, house prices led state

economies but not the other way around in 21 out of 50 US

states, as well as nationally. During this period, state

economies led house prices in 15 mostly agricultural states.

This confirms our suspicion that in many US states, declining

house prices led the economic downturn. However, it is hard

to draw a single conclusion about the interplay between

house prices and the economy that holds for all US states and

the aggregated US. This result is a confirmation of one of the

main conclusions reached by Hamilton and Owyang (2011),

namely that there is no single pattern that characterizes the

behavior of business cycles across different US states.

CONCLUSIONS.

In this paper we investigated whether real-estate price cycles

can be written off as precursors of business cycles. If so, then

the validity of the wealth and credit channels as transmission

mechanisms from housing to the economy should also be

dismissed. Using a Markov-switching vector autoregressive

model, we test the order of precedence of state-level house

price and economic cycles in the US over the sample period

between 1979 and 2012. We fail to reject the hypothesis that

real-estate prices did not lead business cycles in 42 out of 50

US states as well as nationally. In a sub-sample preceding the

recession of 2007 and the recent housing market recovery

(2002 to 2011), house price cycles preceded regional

economies but not vice versa in 21 US states and the

aggregated US. This result implies that real-estate prices and

hence the wealth and credit effects of housing should not be

omitted in theoretical and empirical research

A word of caution is due. Although our results indicate that

house prices should not be dismissed as important precursors

of business cycles, they do not imply that house prices are

reliable predictors of state and national recessions and

expansions. If anything, the opposite is true. Consistent with

the findings of Ghent and Owyang (2010) and Strauss

(2013), our results indicate that house prices may lead, lag, or

coincide with regional economies, and that the dominant

patterns depend upon place and time. Thus, the interplay

between real-estate prices and economic activity requires

further theoretical and empirical research.

Proceedings of the Pennsylvania Economic Association 38

Table 1: Possible decisions based on the outcomes of hypotheses A and B

Test A:

Reject Do not reject

Test B

:

Reject

Category1

Business cycles

follow housing

price cycles

&Housing price

cycles follow

business cycles

Category 3

Business cycles

follow housing

price cycles &

Housing price

cycles do not

follow business

cycles

Do not

reject

Category 2

Business cycles do

not follow housing

price cycles &

Housing price

cycles follow

business cycles

Category 4

Business cycles do

not follow housing

price cycles

& Housing price

cycles do not

follow business

cycles

Proceedings of the Pennsylvania Economic Association 39

Table 2: Test decisions based on the outcomes of hypotheses A and B.

Jan.1979-

Sep.2012

Jan.2002-

Sep.2012

Jan.2002-

Sep.2011

Category 1

Business cycles follow

housing price cycles

&

Housing price cycles

follow business cycles

(A: Reject, B: Reject)

AK,AL,AR,

AZ,CA,CO,

CT,DE,FL,

GA,IA,ID,

IL,IN,KY,

LA,MA,MD,

ME,MI,MN,

MO,MS,NC,

NE,NH,NJ,

NM,NV,NY,

OH,OK,

OR,PA,RI,

SD,US,VT,

WA,WI,WV

(41)

CA,CO,CT,

IL,IN,MD,

ME,MN,NH,

NY,OH,OR,

RI,US,VA,

WA,WI,WV,

WY

(19)

CO,CT,IN,

KS,ME,NY,

OH,OR

(8)

Category 2

Business cycles do not

follow housing price

cycles

&

Housing price cycles

follow business cycles

(A: Do not Reject;

B: Reject)

KS,MT,SC,

TN,TX,UT,

VA,WY

(8)

AL,AR,AZ,

GA,IA,KS,

KY,MO,MT,

ND,NE,OK,

SC,SD,TN,

TX,UT

(17)

AL,AZ,GA,

IA,ID,KY,

MN,MO,ND,

NE,OK,SD,

TN,UT,WV

(15)

Category 3

Business cycles follow

housing price cycles

&

Housing price cycles

do not follow business

cycles

(A: Reject;

B: Do not Reject)

HI, ND (2)

AK,DE,HI,

LA,MA,MS,

NC,NJ,NV,

PA

(10)

AR,CA,DE,

HI,IL,LA,

MA,MD,MI,

MS,NH,NJ,

NM,NV,PA,

RI,US,VA,

VT,WA,WI,

WY(22)

Category 4

Business cycles do not

follow housing price

cycles

&

Housing price cycles

do not follow business

cycles

(A: Do not Reject;

B: Do not Reject)

(0)

FL,ID,MI,

NM,VT

(5)

AK,FL,MT,

NC,SC,TX

(6)

Proceedings of the Pennsylvania Economic Association 40

REFERENCES

Case, K., Glaeser, E., and Parker, J. 2000. Real estate and the

macroeconomy. Brookings Papers on Economic Activity,

2000(2):119-162.

Case, K. E., Quigley, J. M., and Shiller, R. J. 2005.

Comparing wealth e¤ects: The stock market versus the

housing market. Advances in Macroeconomics, 5:1-32.

Gao, A., Lin, Z., and Na, C. F. 2009. Housing market

dynamics: Evidence of mean reversion and downward

rigidity. Journal of Housing Economics, 18: 256-266.

Ghent, A. and Owyang, M. 2010. Is housing the business

cycle? evidence from US cities. Journal of Urban Economics,

67:3: 336-351.

Hamilton, J. D. 1989. A new approach to the economic

analysis of nonstationary time series and the business cycle.

Econometrica, 57(2): 357-384.

Hamilton, J. D. 1994. Time Series Analysis. Princeton

University Press, Princeton, NJ.

Hamilton, J. D. and Lin, G. 1996. Stock market volatility and

the business cycle. Journal of Applied Econometrics, 11:

573-593.

Hamilton, J. D. and Owyang, M. T. 2011. The propagation of

regional recessions. NBER Working Paper Series, w16657.

Iacoviello, M. 2005. House prices, borrowing constraints,

and monetary policy in the business cycle. The American

Economic Review, 95(3):739-764.

Kim, C.-J. and Nelson, C. R. 1999. State-Space Models with

Regime Switching: Classical and Gibbs-Sampling

Approaches with Applications.The MIT Press.

Krolzig, H.-M. 1997. Markov-Switching Vector

Autoregressions: Modelling, Statistical Inference, and

Application to Business Cycle Analysis. Springer.

Leamer, E. E. September 2007. Housing IS the business

cycle. NBERWorking Paper, No. 13428.

Muellbauer, J. 2007. Housing, credit and consumer

expenditure. Proceedings of the Jackson Hole Symposium on

Housing, Housing Finance, and Monetary Policy.

Smith, R., Sola, M., and Spagnolo, F. 2000. The prisoner’s

dilemma and regime-switching in the greek-turkish arms

race. Journal of Peace Research, 37(6):737-750.

Strauss, J. 2013. Does housing drive state-level job growth?

building permits and consumer expectations forecast a state’s

economic activity. Journal of Urban Economics, 73:77-93.

Proceedings of the Pennsylvania Economic Association 41

A MODEL OF RELATIVE CONSUMPTION

Chong Hyun C. Byun

Department of Economics

Wabash College

301 W. Wabash Ave.

Crawfordsville, IN 47933

ABSTRACT Habit formation and Keeping Up with the Joneses are two

specifications of relative consumption in which a household’s

utility depends, not only on its own current consumption, but

also on previous consumption levels, or on the consumption

levels of their neighbors. In such an economy, externalities

are generated by such actions. These externalities are not

accounted for when a household chooses its current

consumption level. Therefore, the competitive solution is not

the same as the social planner’s. Consequently, there is room

for improvement in the form of an optimal taxation policy.

I. INTRODUCTION

Relative consumption, in the form of either keeping up or

catching up with the Joneses, or habit formation, has now

been established as a relevant analytical tool in modifying

standard preference specifications to more adequately

represent consumer behavior. A household’s utility can

depend not only on its own current consumption level, but

also on its previous consumption levels, as well as on the

consumption levels of their neighbors. However, in such an

economy, externalities are generated by such comparisons.

Households do not realize that in choosing their

consumption level today, they will be influencing what their

neighbors do both today and tomorrow. And if they are

strongly influenced by habits, then their choice today will

also affect what their choices will be tomorrow. These

externalities are not accounted for when a household

chooses what to consume today. As a result, households

will consume too much, compared to the case where a

social planner chooses optimal consumption levels for all

households. Consequently, there is room for improvement

in the form of an optimal taxation policy that will induce

households to choose consumption levels that are identical

to the social planner’s outcome.

In this paper I will analyze a model that includes both habit

formation and catching up with the Joneses in the

household preference specification. Compared to the social

planner’s solution, where the planner takes the externalities

into account, the competitive solution is not the same as the

efficient one. Government intervention is required to ensure

the efficient solution, and this comes in the form of taxes on

capital and labor. Taxation on either capital or labor would

change the relative price of consumption either today or in

the future, and would encourage agents to shift

consumption accordingly.

Ljungqvist and Uhlig (2000) consider a model with

catching up with the Joneses preferences and find an

optimal tax policy for this economy. For an economy where

individual welfare depends on relative consumption and

income levels, Boskin and Sheshinski (1978) derive a

taxation system that will optimally redistribute wealth and

consumption. Alonso-Carerra et al. (2004) formulate a

model that includes both habit formation and both keeping

up and catching up with the Joneses preferences. Based on

these consumption spillovers, they derive an optimal tax

rate that will restore the efficient outcome to the economy.

My work builds upon that of both Ljungqvist and Uhlig and

Alonso-Carerra et al. In Ljungqvist and Uhlig’s model, the

utility function exhibits catching up with the Joneses

preferences. Based on this specification, they develop

optimal tax policies in an economy with productivity

shocks. Then they calculate welfare gains due to taxation by

introducing a stochastic productivity shock into three types

of economies (laissez-faire without taxes, the social planner

outcome with optimal taxes, and an economy where the tax

is kept constant at the steady state value). These welfare

gains are compared to those of two nonstochastic

economies: the laissez-faire and the social planner

outcomes.

However, Ljungqvist and Uhlig do not incorporate capital

accumulation in their model. In Alonso-Carrera et al,

preferences are modified to include habit formation and

both keeping up and catching up with the Joneses

preferences. Capital accumulation is included in the model,

and they analyze the optimal taxation rates on both

consumption and capital. But they do not include labor

supply, nor do they consider the welfare gains from the

optimal taxation schemes. My focus is to build on these two

prior works and extend them by including both capital

accumulation and labor supply in a model that includes both

habit formation and catching up with the Joneses in the

specification for household preferences. The next objective

is to compare the competitive equilibrium outcome with the

social planner’s solution for this model. Based on these

different outcomes, an optimal tax rate on labor and capital

can be determined.

Proceedings of the Pennsylvania Economic Association 42

The remainder of the paper is organized as follows. Section

II outlines the model and the household’s choice problem

when preferences are modified to include both habit

formation and catching up with the Joneses. Then the social

planner’s problem is outlined and I compare the results

from the competitive solution versus the social planner’s

solution. Since the preference specification generates

consumption externalities, these two outcomes will be

different. Consequently, there is room for improvement and

welfare gains if the government steps in and imposes an

optimal tax on consumption. This tax induces households

to choose a consumption level that is in accordance to the

socially optimal level. The derivation of the optimal tax is

described in Section III. In Section IV, the equilibrium

efficiency outcome is discussed. Finally, Section V

concludes.

II. THE MODEL

2.1 The Household’s Problem

Consider an economy with infinitely many identical

households, each with the same utility function, facing an

infinite lifetime. Here, a household’s preferences depend

not only on the current value of their consumption, but also

on the lagged value of their consumption, and also on the

lagged value of average consumption over all households in

the economy. The utility function is the standard constant

relative risk aversion format.

(1)

The variables are defined as follows: c t is consumption at

period t for the agent, ct-1 is lagged own consumption at t-1

(habit formation), and is the consumption in period t-1

averaged over all households, that is, catching up with the

Joneses. The parameters are defined as follows: > 0 is the

coefficient of relative risk aversion, measures

how important own lagged consumption is to an agent (i.e.,

the strength of habit formation), and measures

the importance to the household of lagged average

consumption. If 0 < < 1, this represents catching up with

the Joneses, that is, marginal utility of lagged average

consumption is negative. Note that if –1 < < 0 is true, this

represents lagging behind the Joneses, in that marginal

utility of lagged average consumption is positive.

The household faces the budget constraint:

(2)

where kt is capital, wt is wage rate, lt is labor, rt is rental

rate, kt is capital, is depreciation of capital, and

represent taxes on labor and capital respectively, and S t is

the lump sum transfer payment from the government.

The government faces a budget constraint of

(3)

where it receives income from taxes collected on both

capital and labor income. The production technology for the

economy is a standard Cobb-Douglas specification:

(4)

where Y is the output, K is the capital input, and L is the

labor input. Equation (4) can be rewritten as

(5)

where kt is the capital to labor ratio. In a competitive

equilibrium, the factor prices will be equal to their

respective marginal productivities, which are given by

(6)

where MPK stands for the marginal productivity of capital.

The rental rate on capital is equal to its marginal productivity.

The same is true for labor, and the wage rate is equal to the

marginal productivity of labor, given by

(7)

where MPL stands for the marginal productivity of labor. I

will define the marginal utility of current consumption as

(8)

and the marginal utility of lagged consumption as

(9)

Then for the household’s maximization problem, the first

order conditions with respect to consumption, capital, and

labor are

0t

θ1

t

σ1

1t1ttt

θ1

lA

σ1

)cαγc(cβu(t)

1tc

(0,1)γ

1,1)(α

t

k

ttt

l

tttt1ttS)τ(1kr)τ(1lwδ)k(1kc

k

t

l

t τand τ

tt

l

ttt

k

ttlwτkrτS

1η0 ,LKY η1

t

η

tt

η

tttk)f(ky

)(kf'ηkL

KηMPKr

t

η1

t

η1

t

t

t

tttt)k(kf')f(kMPLw

t

1tt1tt

1c

)c,c,c,u(c(t)u

1t

1tt1tt

2c

)c,c,c,(c(t)u

Proceedings of the Pennsylvania Economic Association 43

respectively, where t is the Lagrange multiplier for the

budget constraint. Equation (10) is the first order condition

with respect to current consumption, and represents the

marginal utility of consumption in period t. Equation (11) is

the standard consumption Euler equation for the household,

based on the first order condition with respect to capital.

Finally, Equation (12) shows that the marginal rate of

substitution between consumption and leisure are equal to

the after tax wage rate. An intuitive explanation of this last

equation is that it shows the rate at which a household is

willing to substitute consumption for leisure. The price of

leisure is the foregone income that could have been earned

by working, and ultimately represents the foregone amount

of consumption.

The competitive equilibrium for this economy is defined by

the positive paths of ct, kt, lt, and t satisfying (10), (11) and

(12) along with (2) and (3), and the following transversality

conditions.

(13)

(14)

In a symmetric equilibrium, is true. Define the

gross rate of growth of marginal utility as

(15)

Then from the first order conditions (10) and (11), I derive

the Euler equation

(16)

Defining the gross rate of growth of consumption as

(17)

then Equation (15) becomes

(18)

Combining the individual’s and government’s budget

constraints, I derive the resource constraint of the economy:

(19)

The first order difference equations (16), (17), (18), and

(19) with the transversality conditions and the initial

conditions on k0 and c-1 fully describe the equilibrium path

of the variables xt, t, ct, kt.

Assume now the government follows a stationary fiscal

policy so that taxes are constant throughout time, i.e.

. Then at the steady state, Equations

(16), (17), (18) and (19) respectively become:

(20)

(21)

(22)

To characterize the steady state, let . An

interior steady state exists if and only if the following

condition holds:

(23)

2.2 The Social Planner’s Problem

The social planner now considers the problem of

maximizing utility for the entire economy. In order to do so,

the planner must account for the consumption externality

generated by the household’s preference specification. The

planner internalizes the consumption externality such that

, i.e. that each household’s past consumption is

the same as the average consumption for the entire

economy. Here, catching up with the Joneses no longer

matters, and the utility function the social planner now

faces is the following:

(24)

where are the variables for consumption in

period t, consumption in period t-1, and labor in period t for

(12) 0)τ(1)k(kf')f(kλAl

(11) 0δ)(1)τ)(1(kf'βλλ

(10) 0γ)()cαγcβ(cλ)cαγc(c

l

ttttt

θ

t

k

1t1t1tt

σ

tt1tt

σ

1t1tt

0kλlim1tt

t

0(t)cuβlimt1

t

t

1t1tcc

(t)u

1)(tu

1

1

t

δ)](1)τ)(1(kβ[f'

1

βγ1

βγ1k

1t1t

1t

1t

t

tc

cx

α)(γ

x

11α)(γx

t

σ

1

t1t

tcδ)k(1)f(kk

tt1t

t τ τand ττ ll

t

kk

t

1x

)τβ(1

δ)β(11(k)f'

k

δkf(k)c

(0,1))τβ(1 k

δ)β(11

11 tt cc

θ1

lA

σ1

α)(γccβ(t)u

θ1

t

σ1

1tt

0t

t

tttlcc ˆ and ,ˆ ,ˆ

1

Proceedings of the Pennsylvania Economic Association 44

the social planner’s problem. The resource constraint the

social planner is

(25)

I define the following notation for the partial derivatives of

the utility function with respect to the two variables as

(26)

(27)

The first order conditions with respect to consumption,

capital, and labor are

(28)

(29)

(30)

respectively, where is the Lagrange multiplier for the

resource constraint. The solution to the social planner’s

problem is defined by the positive paths of ,

satisfying (28), (29), (30), the resource constraint, the initial

conditions , and the following transversality

conditions.

(31)

(32)

Define the gross rate of growth of marginal utility in the

social planner’s economy as follows.

(33)

Using first order conditions (28) and (29), and (33), we get

the Euler equation:

(34)

Define the gross rate of growth of consumption under the

social planner’s economy as follows.

(35)

Thus, from (33), we can derive the following.

(36)

The set of first order difference equations (34), (35), (36),

and (25), with the transversality conditions, and the initial

conditions completely describe the dynamics of the

variables At the steady state:

(37)

(38)

(39)

These three steady state conditions are exactly equivalent to

Equations (20), (21), and (22) from the competitive

equilibrium. The only difference between the competitive

equilibrium solution and the social planner’s solution is

between Equations (16) and (34). When k = 0 we get

When tax on capital is zero, the steady

state of the competitive solution coincides with that of the

efficient solution.

III. OPTIMAL TAXATION POLICIES

Inefficiencies may arise from the competitive equilibrium

outcome since agents are not taking into account the

spillover effects of their own consumption. The

externalities arising from an agent’s own lagged

consumption (the habit formation) will influence marginal

utility today. Agents are also influenced by past

consumption averaged over all the agents (the catching up

effect). When agents choose consumption today, they will

affect marginal utility of consumption tomorrow for all

agents in the economy.

Since the competitive equilibrium outcome can be

inefficient, there is a role for the government to step in with

a tax policy. An optimal tax will force consumers to shift

their consumption between periods such that the

competitive equilibrium outcome will be the same as the

social planner’s solution. We consider in this section what

the optimal taxation rates should be on labor and capital.

ttt1tckδ)(1)kf(k

t

1tt

1c

)c,c(u(t)u

1t

1tt

2c

)c,c(u(t)u

0λα)(-γα)(γccβα)(γcct

σ

t1t

σ

1tt

0)]k(f'δ)[(1λβλ1t1tt

0]k)k(f')k[f(λlAtttt

θ

t

tλ

ttttλ and ,l ,k ,c

10c and k

0kλlim1tt

t

0c(t)uβlimtt

t

t

(t)u

1)(tuˆ

1

1

t

δ)](1)k(β[f'

1ˆ

ˆα)β(γ1

ˆα)β(γ1

1t

t

t

1t

1ˆ

ˆˆ

t

t

tc

cx

α)(γ

x

11ˆα)(γx

t

σ

1

t1t

10c ,k

.k,c ,x ,ˆtttt

1ˆx

β

δ)β(11)kf(

kδ)kf(c

k.k c,c ,ˆ

Proceedings of the Pennsylvania Economic Association 45

To characterize the optimal tax on labor, I combined

Equations (12) and (30) to derive an expression for tax on

labor.

(40)

Using the first order condition for the competitive

equilibrium, Equation (10) and the first order condition for

the social planner’s problem, Equation (28), the tax on labor

can be expressed as the following equation.

(41)

Based on this equation, I can characterize the optimal tax on

labor, according to it.

Proposition 1: Optimal Tax on Labor

The optimal tax on labor is defined by

represents the rate of time preference

represents catching up with the Joneses

represents habit formation.

Consequently, the sign of βα determines the direction of the

tax, either positive or negative.

If α > 0, then βα > 0, which implies a positive tax on labor

income. Since the household has catching up with the

Joneses preferences, this positive tax is required to

discourage excess consumption. This is because the

marginal utility of one additional unit of consumption is

greater under the decentralized case compared to that in the

social planner’s economy. Households have to be

discouraged from consuming too much in the decentralized

economy, so this positive tax is necessary.

On the other hand, if α < 0 then βα < 0, which implies a

negative tax on labor. For α < 0, this implies that

households have a preference structure such that they

receive positive marginal utility with increases in average

lagged consumption. This is termed “lagging behind the

Joneses”, in that the household’s utility increases when their

neighbors’ consumption in the past increases. Thus the

households are not consuming enough; they are neither

keeping up nor catching up, but instead falling behind. In

this case, a negative tax (that is, a subsidy) is required on

labor income. The households have to be encouraged to

consume more. Note also that for α < 0, this means that the

marginal utility of one additional unit of consumption is

greater under the social planner’s case than in the

decentralized case. Once again, households have to be

encouraged to increase their consumption in the

decentralized economy, so a negative tax is necessary.

The size of the tax on labor is influenced by the parameter

γ, which represents the strength of habit formation. This

parameter γ is restricted to be between zero and one. For

bigger values of γ, habits are stronger, and as a

consequence, will be larger. On the other hand, if γ is

smaller, then habits matter less to the household, and the tax

on labor income will be smaller. Finally, it is worth noting

that if α = 0 or both α = 0 and γ = 0, then . No taxes

would be necessary in this case since catching up with the

Joneses preferences and habit formation do not exist.

To characterize the optimal tax on capital, we combine (11)

and (29) and evaluate them along the efficient path, to get

the optimal taxation condition for capital.

(42)

where the superscripts denote d=decentralized economy and

p=planned economy.

Proposition 2: Characterizing the optimal capital tax

rate

(a) For MRSd> MRS

p, we have

(b) For MRSd<MRS

p, we have

(c) For MRSd=MRS

p, we have

Part (a) means that consumers in the decentralized economy

want to shift their consumption to period t+1. To discourage

them from doing so, it is necessary to put a tax on capital

income. Part (b) means that consumers in the decentralized

economy are less willing to shift their consumption to the

future. In this case, the government must subsidize (i.e. use

a negative tax) on capital to encourage people to consume

less now and more in the future. Part (c) means that when

MRS for both the planned and the decentralized economies

are the same, so no tax is necessary.

IV. EQUILIBRIUM EFFICIENCY

Compare the equilibriums for an individual, when

. Comparing (6) and (34), we see that

(43)

t

tl

tλ

λ1τ

βγ1

βα1τ

βγ1

βατ

1β0

1,1)(α

(0,1)γ

0

)τ)(1k(f'δ)(1

)k(f'δ)(1

)c,c,c,c(MRS

)c,c,c,c(MRS

k

1t1t

1t

1tt1t2t

p

1tt1t2t

d

.0k

t

.0k

t

.0k

t

0ττ lk

1)(tuβ(t)u

2)(tuβ1)(tu

1)(tβu(t)u

2)(tβu1)(tu

21

21

21

21

Proceedings of the Pennsylvania Economic Association 46

So the competitive equilibrium is efficient if and only if

(44)

for all t and some along the competitive equilibrium path

of consumption.

Since the externalities enter the utility function

subtractively, are linearly dependent for all t.

In fact, it holds that along the competitive

equilibrium path of consumption. So by using (15), now

(41) becomes the efficiency condition

(45)

From this we have the following proposition.

Proposition 3: Let

(a) The competitive equilibrium is Pareto Optimal at the

steady state if and only if = 0.

(b) The competitive equilibrium is Pareto Optimal off the

steady state if and only if = = 0.

Since along the competitive equilibrium path

of consumption, the tax on labor at the steady state becomes

(46)

For = 0, we get . If = = 0, then utility will only

depend on current consumption, and there will be no

externalities, so the competitive equilibrium will be

efficient off the steady state as well.

V. CONCLUSION

A utility function exhibiting habit formation and catching

up with the Joneses will result in inefficiencies in the

competitive equilibrium outcome. Households do not

internalize the externalities their consumption choices may

cause. Thus, their consumption decisions today will

influence their own and all other households’ marginal

utilities today and tomorrow. Due to the distortions

introduced by this externality, the competitive equilibrium

will not be identical to the social planner’s solution.

Therefore, there is a role for government intervention in the

form of taxes on labor and capital. Optimal taxes will

induce agents to choose consumption levels that are

identical to the social planner’s solution.

Based on the work of Ljungqvist and Uhlig (2000) and

Alonso-Carerra et al. (2004), I extend their models to

include capital accumulation and labor supply in the

specification. For this new model, I derived the outcomes

under the competitive equilibrium and the social planner’s

problem. From these outcomes, I further derived optimal

taxes on capital and labor that will induce households to

choose consumption levels that are in accordance with the

social planner’s choice. It turns out that the optimal tax on

labor depends on the parameters that represent both the

catching up with the Joneses preferences and habit

formation. Depending on the direction (positive or negative)

and size of the “catching up” parameter, the tax on labor

will be either positive to discourage consumption, or

negative to encourage consumption. The habit formation

parameter plays a role in determining the magnitude of this

tax. If habits are stronger (i.e. the habit formation parameter

is larger), then taxes will be larger in either direction.

For my future research, I would like to consider the welfare

gains associated with these optimal taxation policies.

Ljungqvist and Uhlig determined the welfare gains from the

optimal taxation in their model, using the following three

stochastic economies: laissez-faire without taxes, the social

planner outcome with optimal taxes, and the social planner

outcome with taxes held constant at the steady state value.

These three are compared to two economies with no

productivity shock: the laissez faire and the social planner

to see the possible gains under these types of economies.

Based on the optimal taxes I derived for my model, I plan to

examine the welfare gains for the same economies as above.

This future work will help establish the importance of

relative consumption models for welfare gains compared to

the competitive equilibrium outcome.

REFERENCES

Abel, Andrew B., “Asset Prices Under Habit Formation and

Catching Up with the Joneses,” American Economic

Review, May 1990, vol. 80, no. 2, pp. 38-42.

Alessie, Rob and Lusardi, Annamaria, “Consumption,

Savings, and Habit Formation,” Economics Letters,

February 1997, 55, pp. 103-108.

Alonso-Carerra, Jaime and Caballé, Jordi, “Income

Taxation with Habit Formation and Consumption

Externalities,” Working Paper, July 2001.

Alonso-Carerra, Jaime, Caballé, Jordi, and Raurich, Xavier,

“Consumption Externalities, Habit Formation, and

Equilibrium Efficiency,” Scandinavian Journal of

Economics, Vol. 106, no. 2, 231-251, 2004.

Boskin, Michael J. and Sheshinski, Eytan, “Optimal

Redistributive Taxation when Individual Welfare Depends

1)(tβu(t)uψ1)(tuβ(t)u2121

(t)u and (t)u11

(t)u(t)u11

ψ1ψ)]γ(1β[α t

.0 k

t

l

t

)()(ˆ11

tutu

βγ1

α)β(γ11 τ l

0τ l

Proceedings of the Pennsylvania Economic Association 47

Upon Relative Income,” Quarterly Journal of Economics,

November 1978.

Campbell, John Y. and Deaton, Angus S., “Why is

Consumption so Smooth?” Review of Economic Studies,

July 1989, 56(3), pp. 357-73.

Carroll, Christopher D., “Solving Consumption Models

with Multiplicative Habits,” Economics Letters, 2000, vol.

68, pp. 67-77.

Carroll, Christopher D.; Overland, Jody and Weil, David

N., “Savings and Growth with Habit Formation,” American

Economic Review, June 2000, vol. 90, no. 3, pp. 341-355.

Constantinides, George M., “Habit Formation: A

Resolution of the Equity Premium Puzzle,” Journal of

Political Economy, June 1990, 98(3), pp. 519-543.

Duesenberry, James S., “Income, Saving, and the Theory of

Consumer Behavior,” Harvard University Press, 1952.

Dynan, Karen, “Habit Formation in Consumer Preferences:

Evidence from Panel Data,” American Economic Review,

June 2000, vol. 90, no. 3, pp. 391-406.

Faria, Joao R., “Growth and Labor Supply in the Presence

of Habit Formation in Consumption,” Economics Bulletin,

2002, vol. 4, no. 5, pp. 1-5

Ferson, Wayne, and Constantinides, George, “Habit

Persistence and Durability in Aggregate Consumption,”

Journal of Financial Economics, 1991, vol. 29, pp. 199-

240.

Fisher, Walter H. and Hof, Franz X., “Relative

Consumption, Economic Growth, and Taxation,” Journal of

Economics, March 2000, vol. 72, no. 3, pp. 241-262.

Frank, Robert, “The Demand for Unobservables and Other

Nonpositional Goods,” American Economic Review, March

1985, vol. 75, no. 1, pp. 101-116.

Fuhrer, Jeffrey C., “Habit Formation in Consumption and

its Implications for Monetary-Policy Models,” American

Economic Review, June 2000, vol. 90, no. 3, pp. 367-390.

Gali, Jordi, “Keeping Up with the Joneses: Consumption

Externalities, Portfolio Choice, and Asset Prices,” Journal

of Money, Credit, and Banking, February 1994, vol. 26, no.

1.

Grinblatt, Mark, Keloharju, Matti, and Ikaheimo, Seppo,

“Interpersonal Effects in Consumption: Evidence from the

Automobile Purchases of Neighbors,” Working Paper,

January 2004.

Gurdgiev, Constantin T., “Life-Cycle Model of Leisure

Demand with Habit Formation,” Working Paper, June 2001.

Hall, Robert E., “Stochastic Implications of the Life-

Cycle/Permanent Income Hypothesis: Theory and

Evidence,” Journal of Political Economy, December 1978,

86(6), pp. 971-87.

Harbaugh, Richmond, “Falling Behind the Joneses:

Relative Consumption and the Growth-Savings Paradox,”

Economics Letters, September 1996, 53, pp. 297-304.

Hayashi, Fumio, “The Permanent Income Hypothesis and

Consumption Durability: Analysis Based on Japanese

Panel Data,” Quarterly Journal of Economics, November

1985, 100(4), pp. 1083-113.

Lettau, Martin and Uhlig, Harald, “Can Habit Formation Be

Reconciled with Business Cycle Facts?”, Review of

Economic Dynamics, 2000, vol. 3, pp. 79-99.

Ljungqvist, Lars and Uhlig, Harald, “Tax Policy and

Aggregate Demand Management Under Catching Up with

the Joneses,” American Economic Review, June 2000, vol.

90, no. 3, pp. 356-366.

Mehra, Rajnish and Prescott, Edward C., “The Equity

Premium: A Puzzle,” Journal of Monetary Economics,

March 1985, vol. 15, pp. 145-161.

Modigliani, Franco, “Life Cycle, Individual Thrift, and the

Wealth of Nations,” American Economic Review, June

1986, 3(76), pp. 297-313.

Muellbauer, John, “Habits, Rationality, and Myopia in the

Life Cycle Consumption Function,” Annales d’Economie et

de Statistique, January-March 1988, (9), pp. 47-70.

Rhee, Wooheon, “Habit Formation and Precautionary

Savings: Evidence from Korean Panel Data Studies,”

Journal of Economic Development, v. 29, n. 2, 2004

Seckin, Aylin, “Consumption-Leisure Choice with Habit

Formation,” Economics Letters, 2001, 70, pp. 115-120.

Skinner, Jonathan, “A Superior Measure of Consumption

from the Panel Study of Income Dynamics,” Economics

Letters, 1987, 23(2), pp. 213-16.

Toche, Patrick, “Savings and Growth with Habit Formation:

A Comment,” Working Paper, May 2004.

Wendner, Ronald, “Capital Accumulation and Habit

Formation,” Economics Bulletin, vol. 4, no. 7, pp. 1-10,

2002.

Proceedings of the Pennsylvania Economic Association 48

PEER MENTOR DEVELOPMENT AS SECONDARY LEADERS AT THE UNIVERSITY LEVEL

Dana D’Angelo, Clinical Professor

Susan Epstein, Associate Clinical Professor

LeBow College of Business (Drexel University)

ABSTRACT

There has much discussion and use of the term "peer

mentor". Another way to consider what a peer mentor

represents is to recognize that peer mentors are, in fact,

"secondary leaders". The Peer Mentoring Program in the

LeBow College of Business at Drexel University was

developed over 15 years ago to provide leadership growth

opportunities for its undergraduate, upper-class students, as

well as mentoring benefits to new students, and support

to their partnered faculty. The program has seen tremendous

success over time and has been modeled by other programs

within the university, and at other

institutions. Undergraduate Teaching Assistants (UTAs) are

recruited based on highly selective criteria. Thus, on an

annual basis, up to 40 high achieving students participate in

the credit earning program. The peer mentoring program

combines both theory and practice, which supports the

university's experiential learning foundation. It is primarily

based upon academic research including that of Daniel

Goleman and Steven Covey. The focus on Emotional

Intelligence and Leadership Styles and Roles provides a basis

for the tasks, behaviors, and assignments that lead to personal

growth for the UTA and the mentoring for new students, As a

part of the evaluation process, a questionnaire is used to

compare perceptions among the peer mentor, the new

students, and faculty, to assess the effectiveness of the

application of specific leadership roles, skills, and

behaviors. The goal of an analysis of the results is to

recognize that peer mentoring is fundamentally secondary

leadership development.

PEER MENTORING IN ACADEMIA

Peer mentoring is a widely recognized approach for a

positive way to develop less experienced performers in

business and industry. Higher level academia has accepted

this mentoring relationship as one that can also create a

motivational environment and thus promote success among

new students. Traditionally, the benefits of the use of peer

mentors as peer leaders in the classroom has focused on the

outcomes achieved by the mentee, the more inexperienced or

new student. Peer mentoring provides these students the

opportunity to receive guidance, support and instruction from

someone other than the faculty member but who is also a part

of the classroom and course. The relationship pairs

collegiate upperclassmen with underclassmen, allowing the

underclassmen to have a role model from an experienced

peer who has met the challenges of the academic

environment. Peer mentors provide another foundation for

what new students need to be successful in the college and

university settings - general support and understanding,

positive role modeling and encouragement, and guidance and

instruction for a different station of life. Active peer mentors

are able to motivate and drive mentees to reach the next level

of success, while offering advice on how to get there.

Based on three mentoring projects, in the United Kingdom,

Korea and New Zealand, retention gains of up to 20% with a

return on investment of the order of magnitude of several

hundred per cent may be possible when peer mentoring is

used. (Boylea, Kwong, and Simpsond, 2010). Feedback

measures student responses to motivation, study skills, study

goals, and belongingness. The outcomes are as follows.

Motivation - Mentors helped motivate 92% of the

students to keep going with their studies

Study skills - Many students (90%) agreed that

mentors helped them with their study skills, and the

majority gave this a rating of 4–5 in terms of

importance to their learning

Study goals - The majority of students (86%)

considered that mentors helped them to achieve their

study goals and rated this help as 4–5 in importance

and 87% stated that the mentoring supporters had

helped them with strategies to manage workload

Belonging - Almost all (98%) of the students who

were surveyed stated strongly that belonging to a

learning group

Abraham Maslow identified Safety and Social needs in his

Hierarchy of Needs, aimed primarily at increasing

motivation. The sense of belongingness provided by the

relationship between peer mentors and their students clearly

contributes to creating affective commitment to the

university, thereby motivating students to academic success.

In addition, academic peer mentoring is often credited with

increased retention rates among mentees. The fostering of

mentoring relationships may assist organizations in

simultaneously promoting effective knowledge transfer and

the affective commitment that assists in the retention of key

knowledge workers. (Fleig-Palmer, 2009). In addition to the

mentee benefits, the positive outcomes to the mentor in an

academic environment are significant. The widely accepted

outcomes of increased communication skills and personal

Proceedings of the Pennsylvania Economic Association 49

leadership roles are clear. In addition, satisfaction of helping

a student reach her academic and professional goals,

recognition at work for participation in a job-related activity,

an expanded network of professional colleagues, recognition

for service to the community, increased self-esteem, self-

confidence and affirmation of professional competence are

all seen as mentee benefits. (Brainard and Ailes-Sengers,

1994).

THE PEER MENTORING PROGRAM IN LEBOW

COLLEGE OF BUSINESS AT DREXEL UNIVERSITY

Drexel University’s LeBow College of Business developed

its undergraduate peer mentor program over 15 years ago,

building on the university’s foundation of experiential

learning and leadership. The program provides the

mentoring students with a sequencing of growth over the

course of two mentoring relationship opportunities, focused

on creating and developing leaders. In LeBow’s peer

mentor program, second year/sophomore students are given

peer mentoring roles in an introduction course for all

incoming freshmen business students, and upper-classmen /

junior-senior student, having previously developed basic peer

mentoring skills in that course, are given the role of

Undergraduate Teaching Assistant (UTA) in the Foundations

of Business I and II classes, also taken by all incoming

freshmen business students. Applicants for the peer mentor

program must first be recommended by a faculty member

from LeBow College of Business. Grades, interviews and

peer recommendations are considered in the evaluation

process. Selected students are then “trained” in a classroom

setting, taking a course based primarily on the understanding

and development of various leadership styles. Beyond this

coursework, peer mentors are guided by their faculty partners

in developing self-confidence and assertiveness in the

classroom.

In his Leadership Pyramid, Steven Covey showed the

progression from Modeling to Mentoring to Teaching. He

described the role of modeling to be building trust and

credibility, the role of mentoring to be developing

relationships, and the role of teaching to be the pinnacle of

the leadership pyramid; it requires a foundation in modeling

and mentoring. (Covey, 1992). Emotional Intelligence has

also been identified as a key component of any leadership

development; and empathy is considered a platform for

human understanding, communication, and relationships as

well as being the foundation for effective leadership, both

personally and professionally. (Goleman, 2005). Based on

Goleman’s Five Components of Emotional Intelligence, peer

mentor learning outcomes within LeBow are designed to

include empathic listening, self-awareness, self-regulation,

motivation and social skills. (Goleman and McKee, 2008).

LeBow peer mentors are challenged with strengthening and

using leadership skills even beyond the basic expectations

normally associated with peer mentoring at the collegiate

level. The Big Five personality traits are explored, helping

the peer mentors identify their comfort zones, and the

students are challenged to demonstrate flexibility in these

areas and application with their mentee groups. At the

conclusion of the second peer mentoring relationship in the

six month UTA course, a 360 degree evaluation is used. The

questionnaire is given to and collected from the peer mentor,

the faculty partner, and the mentee student.

PEER MENTORS AS SECONDARY LEADERS

The sequencing of the mentoring programs at LeBow leads to

the development of secondary leaders, who are trained

beyond the relationship role to become leaders whose goal is

to promote independence among their mentees. A second

chair leader is a person in a subordinate role whose influence

with others adds value throughout the organization. (Bonem

and Patterson, 2005). A second chair leader fundamentally

works under a primary leader as a subordinate; in the case of

our peer mentors, that individual is a faculty member. But

together, these two leaders make up a team, one that

significantly impacts the overall program and the freshmen

students in it. The resulting value added by the second chair

makes the organization much better than it otherwise would

be. The relationships that exist between the primary and

secondary leader will vary, and as such outlooks of goals and

ideas may as well. Additionally, the perceptions of the led

group are also an important part of the dynamic and results

for both leaders.

As part of the evaluation process of the peer mentor

in LeBow’s Foundations of Business I and II course

sequence, a questionnaire was given to the students in the

peer mentor’s class, as well as to the peer mentor him or

herself, and to the supervising faculty member. The results

were primarily used to provide feedback for discussion with

the peer mentor regarding development and future goals, and

well as for overall performance evaluation purposes. Two

faculty members then observed that additional insight may be

provided by comparing results among the participants of the

questionnaire, in particular the primary leader (faculty), the

Proceedings of the Pennsylvania Economic Association 50

secondary leader (peer mentor) and the constituents

(freshman students). The questionnaire (Table 2) was

designed to elicit feedback based on the four roles of a leader

identified by Steven Covey: Modeling, Aligning, Pathfinding

and Empowering. The roles are an extension of principle-

centered leadership, where leaders are viewed as individuals

who do the right thing. The concept was also in direct

strategic placement with the objective of ethical awareness

and development of LeBow.

Four three-item statements were used to gain feedback on the

use of and effectiveness in these major roles of a leader, and

each was rated on a scale of 1 to 5 (1-Rarely, 2-Occasionally,

3-Sometimes, 4-Quite Often, 5-Almost Always). Table 1

shows the averages for the four roles for each of the

participants.

Table 1

Model Align Pathfind Empower

Student

Mentees

4.39 4.21 4.37 4.44

UTA/Peer

Mentor

4.14 3.92 4.25 4.53

Faculty 3.93 3.67 3.97 4.07

For three of the four roles, the mentees perceived stronger

behavioral roles in the peer mentors than either the peer

mentors did themselves or the faculty members did. The

student mentees also rated all four roles well above a 4.0,

indicating they experienced the behavior from the peer

mentor quite often to almost always. In all four roles, the

peer mentor perceived their behavior stronger than the

faculty member did. All three participant groups rated

aligning as the least practiced role and empowering as the

most practiced. In all roles and from all participants, the

averages were well above 3.0, suggested that the behaviors

were indeed perceived on a frequent and observable and

impactful basis.

The mentees clearly see the benefits of the inclusion of peer

mentors in their curriculum, and the peer mentors themselves

have positive outlooks on their roles and their achievements

as secondary leaders. Although the faculty perceptions are

not as strong as the other two groups, they are still positive

overall, and the information provided by the evaluation

questionnaire as a whole can assist faculty in identifying and

working on specific areas with the peer mentors for their

continued growth as secondary leaders in the classroom and

program.

The results of the questionnaire will have direct impact

on LeBow’s Peer Mentor Program going forward. Based on

the feedback, specific tasks and assignments within the peer

mentoring course can be developed to address roles and

behaviors that require more attention, including readings,

group discussions and hands-on activities. The responses to

the questionnaire also are able to give the faculty and peer

mentors collaborative information to use in setting goals for

themselves and in overall course designs for both the peer

mentors and the student mentees. Finally, the evaluation

process can be enhanced to draw more connections to these

goals, activities and results within the program.

Peer mentoring has proven itself to be a valuable aspect in

academia. Most of the benefits traditionally focus on the

mentees’ growth. In looking at the peer mentor as also a

secondary leader, additional benefits can be observed for and

by the mentees. More importantly, however, the

consideration of the peer mentor as the secondary leader as

well allows attention and focus to also be on the benefits to

the peer mentors themselves. Whatever the goals of the

sponsoring program or of the leadership development may

be, whether in roles skills, or behaviors, peer mentors are

indeed secondary leaders, and should be both acknowledged

and developed as such.

Table 2

The UTA

1. Set a personal example for expectations of

students in the class

2. Challenged students to innovate and think

creativity

3. Initiated ideas and activities that supported class

goals

4. Was enthusiastic and upbeat about the class and

LeBow

Proceedings of the Pennsylvania Economic Association 51

5. Conveyed a positive message about future

opportunities as a business student

6. Followed through on commitments regarding

the class, its students and instructors

7. Communicated open-mindedness and

encouraged success

8. Displayed effort toward ensuring students met

guidelines for the class

9. Created an interest and understanding about the

role of business within varying aspects in the

world

10. Connected learning in the class to outside

happenings

11. Showed respect and support for students and

faculty

12. Provided feedback and guidance to students

ACKNOWLEDGEMENT

The authors would like to thank Jordan Kenney,

Undergraduate Student, LeBow College of Business, for his

work in the collection and analysis of the questionnaire data.

REFERENCES

Allen, T. D., Russell, J. E., & Maetzke, S. B. (1997). Factors

related to protégés‘ satisfaction and willingness to mentor

others. Group & Organization Studies, 22(4), 488-507.

Black Issues. Mentoring: The Forgotten Retention Tool

(2002). Diverse Issues in Higher Education

Bonem, M. & Patterson, R. (2005). Leading from the Second

Chair. San Francisco: Jossey-Bass.

Boylea, F, Kwong, J, Rossc, C.,& Simpsond, O. (2010).

Student-Student Mentoring for Retention and Engagement.

Open Learning, 25(2) 115-130.

Brainard, S.G. &Ailes-Sengers, L.A. (1994). Mentoring

female engineering students: A model program at the

University of Washington. Journal of Women and Minorities

in Science and Engineering, 1, (2), 123-35.

Covey, S. R. (1992) Principle-Centered Leadership. New

York: FIRESIDE

Fleig-Palmer, M.M. (2009). The impact of mentoring on

retention through knowledge transfer, affective commitment,

and trust. ETD Collection, University of Nebraska –

Lincoln. Research Commons website:

http://digitalcommons.unl.edu/dissertations/AA13366037

Goleman, D. (2005). Emotional Intelligence: 10th

Anniversary Edition; Why It Can Matter More Than IQ .

New York: Bantam Dell

Goleman, D, Boyatziz, R., McKee, A (2008) What makes a

Leader, Emotionally Intelligent Leadership, Harvard

Business Review, HBR Article Collection

Huizing, R. L. (2010) Mentoring together: A literature

review of group mentoring. Paper presented at the

Northeastern Association of Business, Economics, and

Technology Proceedings, State College, PA. Retrieved from

http://www.nabet.us/Archives/2010/NABET%20Proceedings

%202010.pdf

Proceedings of the Pennsylvania Economic Association 52

CLIMATE NEUTRALITY IN THE HIGHER EDUCATION SECTOR: MAKING THE COMMITMENT

By Soma Ghosh

Discussion by Solomon T. Tesfu, Ph.D

Mount St. Mary’s University

SUMMARY

The paper analyzes the factors that influence the

decision to sign the college presidents’ climate

commitment (PCC) by colleges using descriptive

statistics and a probit model.

Analysis is based on data from a sample of 669 PCC

members for descriptive analysis and 303

institutions for probit model estimation.

The results show that PCC signatories are primarily

public, four-year institutions in urban campuses

located in the northeast region.

The results also indicate that public and land grant

institutions are more likely to join the PCC while

research-intensive institutions are less likely to join.

The research question addressed in the paper is

timely and well motivated.

Comments

It appears that the sample of 303 institutions used

for estimation of probit model for adoption consists

of voluntary responders to a survey. Since a

voluntary subsample may not be representative, the

conclusions about all the institutions based on

sample results in the paper may not be valid. For

example, it is stated that “…while public colleges

are more heavily represented in the network, the

results from the probit model do not indicate that

public institutions are more likely to join the PCC.”

(p.14). This is most likely because the sample used

for probit estimation is not representative of all the

relevant institutions. Therefore, It will be more

informative to include and compare the summary

statistics for the entire group and the sub-sample of

the PCC signatories used in probit analysis to see if

the subsample is representative of the entire group

(for example by public-private and urban-rural

divide of the colleges in the sample vs. all the

colleges).

The time of the decision to sign the PCC by some

colleges and the time period for which the data were

collected (2005-2006) for some of the covariates are

different. But it is possible for some of the

characteristics of the institutions to have been

different at the time they signed the PCC and the

correlation between the dependent variable and the

covariates observed at different points in time may

not be fully meaningful.

The author also mentions that potential endogeneity

might be avoided by backdating the data for the

college characteristics. It might be good to explain

how this is the case since it is not apparent. In fact I

believe, quite a few of the covariates like adoption

of Talloires Declaration (TD) and Green Power

Partnership (GPP) are endogenous (decision to join

these could be motivated by similar unobserved

factors which motivated the decision to join PCC

but it doesn't necessarily mean that membership in

TD and GPP are causing membership in PCC.). If

the author’s interest is in just correlations (and not

causation), then endogeneity may not be a concern.

It is stated that“…the results from the empirical

model imply that land grant and research/doctoral

schools are more likely to join the initiative.” (p.14).

But the coefficient for the dummy indicating

research-oriented institution is negative and highly

significant. It might be informative to explain why

this could be plausible.

It might also be useful to re-scale some of the

covariates to make the coefficients neater for

presentation (e.g. endowment/1000, full-time

equivalents/100, etc) and include the number of

observations for probit model in table III.

Proceedings of the Pennsylvania Economic Association 53

QUESTIONS

The author reports that “… the PCC signatories are

primarily white campuses (60%) with a low

percentage of international students (3%), and have

a balanced representation of women (57%)” (p.11).

Are the PCC signatories more ‘white’ than the non-

signatories? Is the proportion of international

students in these campuses ‘low’ compared to the

non-signatories? Is the proportion of women in these

colleges significantly different from the proportion

in the non-member colleges? I believe the paper will

be more informative if these questions are answered

(explained).

Proceedings of the Pennsylvania Economic Association 54

THE U.S. DOLLAR AS AN INTERNATIONAL CURRENCY RESERVE AND ITS CONTINUOUS DEPRECIATION

Ioannis N. Kallianiotis

Economics/Finance Department

The Arthur J. Kania School of Management

University of Scranton

Scranton, PA 18510-4602

ABSTRACT

The current account deficit (because consumption exceeds

production in the U.S., due to movement of U.S. MNCs

abroad), which causes a capital account surplus (capital

inflows in the U.S. to finance the deficit) and the scale of

financing needed to support the U.S. debts (fiscal deficit,

national debt, and the private sector’s and households’ debts)

together with the Federal Reserve’s policy of keeping U.S.

interest rates low to ward off deflation, to stimulate the

financial markets, and to revive growth, which is impossible

without fiscal policy, has revived concerns about a sudden

and sharp depreciation of the U.S. dollar since 2003.

Americans have lost an enormous amount of their purchasing

power and wealth. As a debtor, the U.S. is benefited from the

devaluation of its currency and it seems that the “market”

knows better than us and keeps the dollar “undervalued”, but

this cannot continue for a long time.

INTRODUCTION AND STATE OF THE

ECONOMY

The U.S. dollar has shown a large volatility even before

1971, when the gold exchange standard was abandoned by

President Nixon and continued since 1973, when the

exchange rate became flexible. The value of the dollar can be

seen by looking at the different exchange rate indexes. The

Trade Weighted Exchange Index (USXRI) has during this

period a mean value, 39919.97IUSXR and a standard

deviation, 39571.14USXRI . Its maximum value was 138

(1985:M03) and its minimum value was 69 (2011:M08). In

2002:M02, the index was at a value of 111 and in 2013:04

was 76 (Graph 1). After that date, the dollar continues to

depreciate with respect the other major currencies. The

question, which rises, here, is: What are the causes of this

depreciation? What is the cost and what are the benefits for

the U.S.? Graph 2 shows many factors, which have caused

the dollar’s depreciation. The current account deficit, the

capital account surplus, the huge national debt, the inflation

in the country, the price of oil, and the tremendous

uncertainty, due to the Middle East crises (Iraq and

Afghanistan, and there are new ones coming, soon) have

created for the U.S. and its currency a serious undermining

and creeping chronic instability. Also, the low (zero) interest

rate in the U.S. (enormous liquidity) is depreciating the dollar

and caused the bubble in the financial assets, in the precious

metals, and in the housing market. This was followed by the

financial crisis (August 2007), which affected the U.S.; and if

the Euro-zone had not experienced its unique debt crisis and

its disastrous common currency, the dollar could have been

worse.

The devaluation of the dollar reduces the current account

deficit ( MXCA ), but at the same time the capital

account ( KA ) surplus is falling because international

investors are selling off dollar-denominated investments, due

to their loss of value; then, the CA continues to be equal

with the KA ( KACA ). The value of the dollar is

determined by its demand and supply, as it happened with all

the currencies after 1973. The dollar has become weaker

because its supply exceeds the demand. Foreigners are selling

their products to the U.S. and the money (dollars) that they

receive is invested back to the U.S. by buying bonds and

stocks.1 What will happen if the international investors

(Chinese, Japanese, OPEC, etc.) are unwilling to keep and

start selling their American bonds and stocks? The capital

account surplus will diminish and consequently, the current

account deficit will fall. A devaluation of the dollar at this

point will reduce the current account deficit and equalize it to

the capital account.

Thus, there is a dollar crisis in the world, due to the

enormous level of the U.S. deficits and debt: [Federal

Debt=$16.776 trillion, Social Security Liability=$20.5

trillion, Medicare and Medicaid Contingent Liabilities=$98

trillion, State and Local Governments=$5.71 trillion,

Business Sector Debt=$11.63 trillion, Financial Sector

Debt=$13.6 trillion, Total Personal Debt=$13.22 trillion,

Financial Sector Bail-out=$2.5 trillion, Other Debts=$2.74

trillion: Total Debt (Public and Private)= $184.676 trillion].

The GDP (2013:Q1) was $13.750 trillion. Then, the total

debt is 1,343.1% of the GDP.2 The Federal Reserve Bank

tries to keep the interest rate low to affect positively the

financial markets, but this policy did not help so much the

real economy because we have reached a liquidity trap. This

Fed’s policy is only pro-market and not pro-social. Thus, the

social benefits are insignificant. Also, this policy of

enormous liquidity caused the bubbles in the financial market

and in the housing market and finally, it will induce inflation,

when the unemployment will reach the natural level. The

U.S. dollar has declined from its pick point USXRI=138

(1985:M03) until now USXRI=76 (2013:M04) by more than

-45%. With respect to the euro, the dollar has declined from

Proceedings of the Pennsylvania Economic Association 55

0.8530 $/€ (2001:M06) to 1.6001 $/€ (2008:M04), which is -

87.76%. Now (5/15/2013), it is 1.2921 $/€, a loss of -51.48%

since its pick value.

The depreciation of a currency means a decline in the

purchasing power of the currency, which leads the citizens of

the country to a decline in their living standards. The U.S.

consumers had lost 87% of their purchasing power compared

to the European ones. An investment in dollars has suffered

from this decline. The price of importables is going up with

the same proportion. All commodities that we import (i.e.,

oil, coffee, BMW, French wine, etc.) are setting record

prices, due to the depreciated dollar. The OPEC countries

watch the deprecation of the dollar, which is the currency that

they receive for their sales of oil (except Iran) and they raise

its price to keep their revenue constant. The trade deficit (

0MX ) causes the dollar to depreciate, so exports can be

stimulated and slightly improve the trade account (price

elasticities for exports and imports play a role, here, too). As

a currency declines in value and interest rate parity does not

hold, investors go to other countries with stronger currencies,

i.e., euro, yen, pound, etc.

A forecast from the IMF predicted that China’s economy will

surpass that of the U.S. in five years. The U.S. Fed, to keep

the interest rate close to zero ( %25.0FFi ), is doing the

“quantitative easing” (printing money).3 Then, the value of

the U.S. dollar will continue to be at a low level and inflation

will rise, especially, as long as the employment will be

improved. At the moment, the unemployment rate is still very

high ( %6.7u ). Europeans and other nations are very

critical of this easy money (printing money) policy of the

U.S. This is negatively affecting the competitiveness of the

foreign countries because their currencies are appreciated,

when the dollar is willingly depreciated (“beggar-thy-

neighbor” policy).

It seems that the IMF is interested in a new world reserve

currency. The U.S. dollar was playing this role since World

War II and many resources, like oil, were purchased with the

dollar (petro-dollars). The U.S. was increasing its debt and

countries with surplus (Germany, China, Japan, OPEC

nations, etc.) were financing the U.S. debts. Since 1980s with

all these deregulations in the U.S. financial market, the frauds

in investment banks, the corruption, the housing bubble, and

the financial crisis of 2007-2013, the U.S. dollar continues to

depreciate. The Special Drawing Rights (SRD) and the euro

are competing with the dollar for their share as world reserve

currencies. The G-20, the IMF, the World Bank, and the

Bank of International Settlement are supporting the role of

SDR. The U.S. has lost most of its manufacturing

infrastructure (the most of the U.S. MNCs have become

foreign firms, today) and it is importing much of the natural

resources, oil, high tech, food, clothes, etc. If SDR or euro

will become world’s reserves, the demand for dollar will fall

and the dollar will depreciate further. Thus, it will be very

costly for Americans to pay for their importables; their

purchasing power and their domestic wealth will be

restrained even more with the current financial crisis and the

“fiscal cliff”. There are about 50 million Americans, who

receive Federal Assistance (Medicare, social security,

disability, and unemployment). This is a sign showing the

trend of this “economic power”. As the dollar will be

depreciated, there will be higher cost of imports, more

unemployment, lower income, confinement of wealth, more

home foreclosures, and more bankruptcies.

The U.S. economy was depending on credit to invest,

consume, and grow; but, this “advantage” turned to a big

disadvantage, an unmanageable debt. This economic

expansion, fueled by debt-based capital markets, gave a

temporary advantage of the West (capitalism) toward the

East (communism). Now, with the latest financial crisis and

the debt crises, the West seems to be worse than the East

because it went a step forward. It altered capitalism to a new

system, the globalism and its first social cost is obvious now.

What it will follow is difficult to be predicted from now. The

credit-driven expansion will be restricted in the future and the

debt-driven contraction will take over. The two large U.S.

bubbles (the “dot.com” and the “housing” one) have caused

serious problems to the “laissez-faire, laissez-passer”

economic system. This system recently caused a global

contraction and serious social welfare problems (from

bankruptcies to national destructions to suicides).

Global demand is falling as credit contracts and employment

and income are falling, too. Then, from where the expected

growth will come? The U.S., as the world’s largest debtor,

might have difficulties paying what it owes in the future,

except by rolling its debt forward and borrowing more.

Today, the debt is everyone’s problem because it has reached

an un-payable level. A default by the U.S. will have global

consequences, especially in China, Japan, and the other large

creditors of the U.S.; and of course, to the global peace. The

amount of outstanding U.S. debt (public and private) has

reached a level that can never be paid off. Historically,

inflation destroys the value of money. Debts are paid back

with inflated dollars, a process, which benefits the debtor and

injures the creditor (U.S. usually inflates its way out of debt

by printing what it owes). Depreciation of the dollar is

another option to benefit borrowers at the expense of lenders.

Taxation is a different option that has similar results as the

two previous ones, but already the U.S. middle-class is

paying very high taxes. Corporations, in the U.S. and all over

the world, do not pay taxes. This is another unfair social

policy, too.

Actually, corporations and wealthy people are paying

relatively less taxes compared to the middle class, which is

unfair and unethical and their tax evasion is very high, too.

This illegal capital flight is a large proportion of deposits in

Proceedings of the Pennsylvania Economic Association 56

offshore centers and tax havens.4 Also, GE paid no taxes;

Goldman Sachs paid $14 million in 2010. The GAO reported

in 2008 that “two out of every three United States

corporations paid no federal income taxes from 1998 through

2005”. Companies have become all too astute at paying for

loopholes, which allow them to shift profits abroad or move

their gains (on paper) to foreign low-tax/no-tax nations

(money laundering). As the data show, the change in

corporate taxes — not merely rates, but what they actually

paid — over the past half century are astounding.5 This

practice increases the national debt and the dollar is

depreciated even more.

Also, inflation, through monetary expansion, could raise

aggregate U.S. debt and prices further. With the current

enormous money supply and zero interest rate, we reached a

liquidity trap and this policy has not so far and might fail to

promote growth in the near future. Europe is in trouble; due

to lack of liquidity (the ECB’s overnight deposit rate was 1%

and became 0.75%). Lending has declined drastically in the

U.S. and even more in Europe. The problem is lack of

demand. The question is now: from where will the expected

stimulus come? Troika believes that austerity generates

growth (sic). Of course, if this U.S. monetary expansion had

been used by the economy, we could have experienced

hyperinflation in the U.S. Thus, it is safe for the economy, at

the moment, because the high risk, the low income, and the

high unemployment have made banks reluctant to lend and

people unwilling to borrow. The U.S. has experienced high

rates of inflation in the past, even though that the number one

objective of the Fed is “price stability” ( 0e ). The dual

mandate, it is not “full employment and price stability”, but

“maximum employment and price stability”. The monetary

expansion, the last three years, has far exceeded any previous

ones. Then, what follows, when the unemployment will fall,

it will be a high inflation. This liquidity and uncertainty in

the market has increased the price of gold and silver (another

big bubble in precious metals and a depreciation of the

dollar).

Furthermore, depreciation of the dollar helps partially the

U.S., but not the creditors of the country because their return

is falling from their U.S. investments (translation exposure).

China has linked its currency to the U.S. dollar, so with any

depreciation of the dollar, the yuan is also declining in value

and China becomes the major beneficiary by increasing its

exports. Thus, the peg of the Chinese yuan to the U.S. dollar

prevents the U.S. from altering its trade deficit by currency

devaluation, but it helps the U.S. relative to the other

countries (i.e., Euro-zone), where their currencies are

appreciated.

In 1980s, the U.S. needed Japanese savings to finance

Reagan’s multi-trillion dollar debt-based military buildup

(the famous “star war”).6 If Japanese rates were raised,

Japanese savings would stay at home, but Japanese rates kept

low (with the “help” from the U.S.). These low interest rates

and high earnings of Japanese firms, due to their exports

ignited a speculative frenzy in their stock markets causing the

then largest stock market bubble in history. In 1990, the

bubble collapsed and Japan fell into a deflationary trap, from

which it has never fully emerged.

Today, the economic power has shifted to China from Japan.

The U.S. imports from China are enormous7 and it is

financing its debt with Chinese capital. Thus, the U.S.

dominance is being challenged by China. China has, now,

reduced its holdings of U.S. debt, which will affect the U.S.

deficits, interest rates, and the value of the dollar. A

depreciation of the U.S. dollar by 30% would impact the debt

held by foreigners; their losses would be significant.8 Of

course, after the above economic impacts, geopolitical

considerations are taking place. Expenditures are rising in the

U.S., production is falling, and the need to borrow is

increasing. Then, the total U.S. debt (public and private)

would prove difficult of even being repaid. The future of all

nations might be difficult with all these absurd decisions

(“irrational exuberance”) of the last 30 years. In 1973, when

gold was taken out as a monetary asset, balance sheets of

every central bank of the world suffered, as their dollar-

denominated assets sank in value, in terms of dollar.

Unfortunately, lately, spreads on sovereign debt are rising

and credit default swaps (CDS) reflect the higher premia

being charged to protect against default. Investors compare

risk ( D ) to reward [ )( DRE ] and try to maximize the

reward to variability (RV) ratio of their bonds’ investments

with respect to debt; when the reward is believed to

compensate for the risk (max RV), the bond is bought and the

bet is placed. Some countries are paying 34% interest rate,

due to high austerity measures that have increased the

probability of bankruptcy, as it happened, lately, with Greece

and other Euro-zone countries (PIIGS nations). We hope that

this will not take place (will be prevented) in the heavily

indebted U.S.A.

The most recent literature review on this area is as follows.

Uri and Boyd (1991) found that a devaluation of the U.S.

dollar have caused an increase in output of the agricultural

sector and all producing sectors, except the financial one.

Glain (2003) says that a cheap dollar makes U.S. exports

more competitive, but discourages investors from holding

dollar-denominated assets. Siegman (2008) says that the huge

U.S. deficits are unsustainable and will lead to disruptions in

international financial markets, to a global financial crisis,

and to worldwide recession. Kallianiotis and Bianchi (2009)

show speculation and monetary policy (risk and rate of

return) play a major role in the determination of the exchange

rate between the dollar and euro. Bonitsis (2011) shows that

the introduction of the euro changed the competitiveness in

Proceedings of the Pennsylvania Economic Association 57

Euro-zone nations and made Germany the most competitive

economy. Amstad and Martin (2011) say that after the

financial crisis the Fed increased its liquidity drastically, but

not the ECB, which has affected the value of the currencies.

Nechio (2011) states explicitly that the common target rate

(overnight deposit rate) of the ECB helps only Germany and

does not help at all the peripheral countries of the Euro-zone.

Sharma (2011) argues that the U.S. exploding debt and

deficits have raised concerns about the future of the dollar

and the competitive devaluation is a growing problem

globally. Krugman (2012) recommends fiscal policy for the

U.S. and not only zero interest rate monetary policy by the

Fed.

A THEORITICAL MODEL OF DOLLAR’S

VALUATION

The correlation and the causality tests between the exchange

rate and some macro-variables (Graph 2) reveal that the

exchange rate [e ($/€)] is correlated and affected by the

following variables:

),,,,

,,,,,,,,,(

, EDCDWDOPDiii

DJIACPINDISEYKACAfe

tGBTSFF

tttttttttt

ttt

(1)

where, e=exchange rate, CA=current account, KA=capital

account, Y=income, E=expenditures ( GICE ),

S=saving, I=investment, ND=national debt, CPI=consumer

price index, DJIA=Dow Jones Industrial Average, iFF=federal

funds rate, iS-T=short-term interest rate, iGB=government

bonds rate, OPD=oil price domestic, WD=Iraqi war dummy

(0 before March 2003 and 1 after March 2003), and

EDCD=European debt crisis dummy (0 before September

2009 and 1 after September 2009).

The total national income of the economy (Y) is used for

paying taxes (T), for consumption (C), and for saving (S):

tttt SCTY (2)

The national production (Y) can be used for consumption

(C), for investment (I), for government spending (G), for

exports (X), and a proportion from this aggregate demand

(AD) is imported (M):

tttttt MXGICY (3)

The trade account (TA) of the country can be presented with

the following equation:

tttt

ttt YY

P

PeTA *

32

*

10 )( (4)

where, TA=trade account, P=U.S. price level, P*=foreign

price level, Y=domestic income, and Y*=foreign income.

The capital account can be written as:

ttttttt sfiiCAKA )()( 2*

10 (5)

where, KA=capital account, i=domestic interest rate,

i*=foreign interest rate, f=ln of the forward rate, and s=ln of

the spot exchange rate.

Through substitution of eqs. (2), (3) ,(4), and (5), we receive:

ttt

ttttt

YY

KANDPPe

*

1

3

1

2

11

*

1

0

lnln

ln1

ln1

lnlnln

(6)

where, tKA can be substituted with the tCA , with (*tt ii ),

and with ( tt sf ), too.

Further, we can test the effects of the price of oil (toilP ),

national debt ( tND ), current account ( tCA ), war dummy (

WD ), and European debt crisis dummy ( EDCD ) on the

exchange rate.

t

ttoilt

EDCD

WDCANDPet

5

43210 lnlnlnln

(7)

Thus, eqs. (1), (6), and (7) can be used to determine the

factors that affect the exchange rate (the value of the dollar).

An increase of the direct quoted ($/€) exchange rate ( te )

means a depreciation of the U.S. dollar.

EMPIRICAL RESULTS

It is important to test the above theory by applying data from

the two economies, U.S.A. and Euro-zone. The data, taken

from economagic.com, imfstatistics.org, Eurostat, and

Bloomberg.com are monthly from 1992:01 to 2011:12. They

comprise, spot exchange rate (e), consumer price indexes in

U.S. and EMU (CPI and CPI*), federal funds rate (iFF), 3-

month T-bill rate (iRF), ECB overnight rate (iOND), 3-month

deposit rate (LIBOR) (i3mdl), nominal (Y and Y*) and real

GDP (Q), private consumption (C), private investment (I),

exports (X), imports (M), current account (CA), capital

Proceedings of the Pennsylvania Economic Association 58

account (KA), taxes (T), government expenditures (G),

national debt (ND), personal saving rate (psr), Dow Jones

Industrial Average (DJIA), oil prices (OPD), price of gold

(PGold) for the U.S. and the Euro-zone economies.

First, the correlation coefficients and a Granger causality test

between all these variables are presented in Tables 1 and 2.

There is a very high positive correlation between the

exchange rate (dollar’s devaluation) and the exports, imports,

current account deficit (capital account surplus), U.S. and

foreign income, investment, national debt, U.S. and foreign

prices, taxes, price of oil, consumption, and the war dummy;

also, a negative correlation (dollar’s appreciation) with

respect the domestic and foreign interest rates. The domestic

and foreign income, consumption, and foreign prices cause

the depreciation of the U.S. dollar. The U.S. national debt

causes the depreciation of the dollar, measuring with the

exchange rate index (USXRI), too.

Then, Table 3 shows the estimates of exchange rate by using

eq. (1). Domestic income, consumption, price of oil, the war

dummy, and personal saving rate have a significant positive

effect on the exchange rate (dollar is depreciated). The price

level and the government bond rate have a significant

negative effect on the exchange rate (dollar is appreciated).

Table 4 represents eq. (6). The U.S. national debt, the interest

rate differential, domestic income, and foreign income have a

significant positive effect on the exchange rate (dollar is

depreciated). The foreign price level has a significant

negative effect on the exchange rate (dollar is appreciated).

The foreign price level has a significant effect on the

exchange rate (dollar is depreciated). Table 5 presents the

price and income elasticities (domestic and foreign) of the

current account and the effects of interest rates and forward

discount of the dollar on the capital account.

Now, we do different tests for eq. (7) and the first results of

them are in Table 6, which presents the Augmented Dickey-

Fuller and Phillips-Perron unit root tests. These results show

that only the tnd contains no unit root [ )0(I ]. The other

series are nonstationary [ )1(I ]. Table 7 gives a cointegration

test of the series of eq. (7). It reveals that there is one (1)

cointegrating equation at the 5% level and one (1)

cointegrating equation at the 10% level. Then, the linear

combination of these non-stationary series is stationary; they

are cointegrated. There is a long-run equilibrium relationship

among the variables of eq. (7).

Table 8 displays the regression results of eq. (7). The price of

oil, the national debt, and the war in Iraq are depreciating the

U.S. dollar (spot rate increases). The European debt crisis

appreciates the dollar and depreciates the euro. Table 9

exhibits the correlogram and Q-statistics for testing high-

order serial correlation of the residuals of eq. (7). The

correlogram has spikes at lags up to six and the Q-statistics

are significant at all lags, indicating significant serial

correlation in the residuals. Lastly, Table 10 demonstrates the

serial correlation LM test (Breusch-Godfrey) and rejects the

hypothesis of no serial correlation up to order four. Thus, the

residuals are serially correlated and eq. (7) should be re-

specified before using it for hypothesis tests and forecasting.

In addition, Graph 1 presents the value of the dollar by using

an exchange rate index from BIS. Graph 2 shows the

causality and the two-way causation between the most

important variables in the U.S. economy (

CPIandOPDNDiiiiSKACA TLTSGBFF ,,,,,,,,, ). As

Graph 2 (2 lags) reveals that there is a significant causal

relationship among the economic variables (the arrows show

the direction of causality). The spot exchange rate is caused

by national debt, current account deficit, capital account

surplus, and consumer price index.9 Graph 3 shows the (€/$)

spot exchange rate. The dollar is losing value since 2002 with

the Iraqi invasion of the U.S. (it seems that the Muslim

countries are now investing in Europe).10

Graph 4 reveals the

depreciation of the dollar with respect the gold. In early

1970s, the dollar had a catastrophic downfall. Finally, Table

11 presents the global currency reserves. Since the

introduction of the euro in 1999, the U.S. dollar is losing as a

global currency reserve (from a 71% share, it fell to 61.9%).

CONCLUDING REMARKS

The objective of this analysis is to determine the factors that

have caused the depreciation of the U.S. dollar the last ten

years, and its effects (positive and negative) on trade, wealth,

and social welfare. The factors that have been determined,

here, are the U.S. and European income, the U.S.

consumption, the EMU prices, the U.S. national debt, the

current account, the personal saving rate, the price of oil, the

Middle East turmoil (war dummy), which has increased the

demand for euro, and speculation about the two economies

(U.S. and Euro-zone).11

The exchange rate dynamics is based

on shocks on the economy and on current account, due to oil

prices, debts, and risk, between the U.S. dollar and the euro.

Lately, the U.S. dollar has been losing value with respect the

euro and other major currencies of the world. The last two

years, the dollar is appreciated with respect the euro, due to

the Euro-zone debt crisis (European debt crisis dummy). We

want to see if this depreciation depends on economic shocks

and economic fundamentals or it is just speculation from

individuals and countries, which hold large amounts of

foreign assets denominated in different currencies or due to

the current global financial crisis, recessions, instability, and

the risk that the U.S. might freeze the foreign funds invested

in its assts.

The conclusion from this analysis can be that international

investors are investing in countries with higher return or

lower risk, and safety, depending on their utility function and

Proceedings of the Pennsylvania Economic Association 59

the modern systemic risk. This increase in demand for these

assets increases the demand for currency in that country and

its currency is appreciated; the oil prices, the high risk and

the enormous debts are affecting the currency, too. Before

2001, people were invested in the U.S. and Japan, so the U.S.

dollar and the Japanese yen were appreciated. After 2001,

they invested in the Euro-zone and the U.K. and the dollar

and yen lost their value. Of course, due to, high risk and wars

(in Iraq and Afghanistan) and the creeping ones (in Syria and

Iran), political conflicts, a unique financial crisis, and low

returns, many speculators have invested in euros and other

currencies, instead in dollars denominated assets. After the

last months of 2008, we see a change in this trend because of

the Euro-zone debt problems. The current account is affected

by risk and high debts, too. Historically, the American

governments have frozen the foreign assets inside the U.S.,

when a conflict arises.

Finally, the decline of the U.S. dollar occurs because the

country has failed to correct the macroeconomic distortions

in its economy (production, consumption, self-sufficiency,

international relationships, debts and deficits, low return,

high risk and unemployment, etc.) The response of the two

policy rates (*ONDFF iandi ) on the exchange rate is a

negative one for three months and then, it stays constant,

which means [ euroandSMi s ($)( )] a lot for

3 months (overshoots)12

and then, it is stabilized at a lower

level. Testing the effectiveness of monetary policy on the

exchange rate, we found it non-effective. [

euroSiandSi ONDFF ()($ *)].

13

Taking into consideration the effect of the freezing funds risk

premium (FFRP)14

on the exchange rate, we found that: [

)($ euroandSFFRP ],15

which is reasonable for

our state of the economy, due to the Middle-East crises, the

historic memory with Japan in the past, and the continuation

of the cold war. After all this analysis, we can say that the

dollar could appreciate with respect to euro, if we will have

any other domestic (like, increase of the federal funds rate or

ECB rate, due to fear of inflation) or external shocks on the

two economies (oil prices, Euro-zone member default, new

wars, etc.). Still the forecasting of the exchange rate remains

as a problem and the depreciation of the dollar will continue

if the U.S. socio-politico-economic system stays the same.

1 On May 15, 2013, the U.S. national debt (ND) was $16.776

trillion and foreigners were holding $5.758 trillion, which is 34.32% of the total ND. See, http://www.treasury.gov/resource-center/data-chart-center/tic/Documents/mfh.txt

2 Thus, for the beginning of 2012, the Social Distress Index

(SDI) for the U.S. was:

%91.361,1%03.350,1%58.3%3.8 duSDI

and for the end of March 2013, the SDI was:

%82.353,1%1.343,1%12.3%6.7 duSDI ,

which show that the country is improving a little, but it is still extremely distressful (risky). See, Kallianiotis (2011, p. 344) for this index. The U.S. needs 14 years to pay off its debt, if all the other spending would be zero. Then, it is impossible! 3 The monetary base (MB) from $846 billion in 2007 reached

$3,034.009 billion in 5/1/2013. See, http://research.stlouisfed.org/fred2/series/BASE/ . Also, the Fed is buying $85 billion of securities (government bonds and mortgage back securities) per month. This liquidity keeps the interest rate at zero level and hurts small savers, who face a nominal deposit rate closed to zero and a real one negative for four years. This policy is a redistribution of wealth from savers to banks and the financial market. See, http://www.usatoday.com/story/money/business/2013/05/01/fed-maintains-stimulus/2126381/ 4 See,

http://www.boston.com/business/globe/articles/2004/04/11/most_us_firms_paid_no_income_taxes_in_90s/. But, it was completely unethical for Troika to go against one offshore center, the Cyprus one in March 2013 and not against the big ones. 5 (1) Corporate Taxes as a Percentage of Federal Revenue

were; in 1955: 27.3% and in 2010: 8.9%. (2) Corporate Taxes as a Percentage of GDP; in 1955: 4.3% and in 2010: 1.3%. (3) Individual Income/Payrolls as a Percentage of Federal Revenue; in 1955: 58.0% and in 2010: 81.5%.

See,

http://www.ritholtz.com/blog/2011/04/corporate-tax-rates-then-and-now/. 6 The total military cost of wars since World War I until 2010

for the U.S. was, in constant 2011 dollars: $6,724 billion. The total Iraq-Afghanistan (2001-2010) war cost was $1,147 billion. See, Stephen Daggett, “Costs of Major U.S. Wars”, Congressional Research Service, June 29, 2010. 7 Trade deficit with China: in 2010, it was $273.1 billion; in

2011, it was $295.5 billion; and in 2012, it was $315.054 billion; in 1985 it was only $6 million. See, http://www.census.gov/foreign-trade/balance/c5700.html 8 China currently owns about $1,151.9 billion in U.S. dollar

denominated securities; and a depreciation of the dollar by 30%, it will cost China’s $345.57 billion in losses of its investment. 9 Graph 2, Tables 1, 2, 5, 6, 7, 9, 10, and 11 are omitted, due

to space limitations. All these and the F-statistics of this Granger causality tests and the correlation coefficients between the two variables are available from the author upon request.

Proceedings of the Pennsylvania Economic Association 60

10

See, Kallianiotis and Petsas (2008). 11

Economy of Euro-zone (2011)

(1) GDP = $12,460 billion

(2) Growth of GDP = -0.3%

(3) Inflation rate = 2.7%

(4) Unemployment rate = 10%

(5) Gross External Debt = (N/A)

(6) Public Debt = 86% of the GDP

(7) Budget Deficit = 4.1% of the GDP

(8) GDP as a % of the U.S.A. = 92.83%

Economy of U.S.A. (2011)

(1) GDP = $13,423 billion

(2) Growth of GDP = 1.7%

(3) Inflation rate = 2.98%

(4) Unemployment Rate = 8.5%

(5) Gross External Debt = $8,400 billion

(6) National Debt = $15,251 billion; 113.62% of the

GDP

(7) Budget Deficit = $1,300 billion; 9.68% of the

(8) GDP as a % of Euro-zone = 107.73%

(9) Total Debt (Public and Private)= $115.7

trillion+$40.8 trillion=$156.5 trillion; 1,165.91% of

the GDP

Then, U.S.A. is not doing better than the Euro-zone.

Why so much noise for the Euro-zone? There is no economic

explanation!..

12

See, Dornbusch (1976) and Kallianiotis and Bianchi (2009).

13

The regression is:

467.226,424.0,882.0

)023.0(

947.0

)025.0()016.0()009.0()051.0(

503.1001.0024.0218.0

2

2***

1**********

FSSRR

iis

t

tondFFt tt

14 The freezing funds risk premium must be:

tEUUS FFRPfdiittt $

* , as Kallianiotis and Bianchi (2009)

mention for the U.S. to attract capital from the Muslim countries. 15

These results are as following:

425.647,3,019.0,995.0

)089.0()090.0()018.0(

356.0342.1990.0

)001.0()002.0()002.0()506.0(

001.0007.0005.0357.0

2

2***

1***

1***

$*********

FSSRR

s

fdFFRPis

ttt

tONDt tt

Note: ts = ln of spot exchange rate, *

tONDi = ECB overnight

deposit rate, tFFRP = freezing funds risk premium, t

fd$ =

forward discount of the U.S. dollar, and t = the error term.

Proceedings of the Pennsylvania Economic Association 61

Graph 1

The Depreciation of the U.S. Dollar with Respect the Major Currencies

Note: USXRI = Trade Weighted Exchange Index: Major Currencies: Index March 1973 = 100.

Source: Economagic.com.

Table 3: Factors Affecting the U.S. Dollar (Exchange Rate Determination) [Eq. (1)]

----------------------------------------------------------------------------------------------------------------------------- -----------------------------

Variables teln teln teln teln teln teln

----------------------------------------------------------------------------------------------------------------------------- -----------------------------

C -14.166**

-0.710 -9.930***

-0.824* -9.659

*** -11.965

(6.752) (0.472) (1.201) (0.463) (1.203) (1.981)

tKAln -0.185***

- -0.014 -0.016 -0.033 -0.022

(0.060) (0.067) (0.021) (0.041) (0.042)

tYln 2.497**

- 3.227***

0.164 0.995**

0.867**

(1.197) (1.009) (0.334) (0.404) (0.369)

tCln -2.106***

- -1.707**

0.549* -0.159 0.312

(0.758) (0.861) (0.283) (0.390) (0.455)

tpsr 0.005 - -0.004 0.004* 0.001 0.003

(0.006) (0.008) (0.002) (0.002) (0.002)

tIln 0.277 - 0.021 -0.022 0.170 0.059

(0.216) (0.205) (0.066) (0.163) (0.183)

tNDln -0.025 0.040 0.024 -0.058 -0.034 -0.239

(0.149) (0.054) (0.144) (0.046) (0.142) (0.168)

tPln 1.443 - -0.964 -0.934***

0.209 0.557

(0.901) (0.968) (0.309) (0.743) (0.653)

tDJIAln 0.081 - - - - -

(0.079)

tFFi -0.008 - - - - -

60

70

80

90

100

110

120

130

140

50 55 60 65 70 75 80 85 90 95 00 05 10

USXRI

Proceedings of the Pennsylvania Economic Association 62

(0.031)

tRFi 0.003 - - - - -

(0.035)

tGBi -0.027 -0.024**

- - - -

(0.019) (0.010)

toilPln 0.138***

0.145***

(0.038) (0.020)

WD 0.227***

0.092***

(0.032) (0.022)

1ln te - - - 0.938***

- -

(0.027)

)1(AR - - - - - 0.902***

(0.030)

)1(MA - 1.187***

- - 1.337***

0.206**

(0.071) (0.077) (0.092)

)2(MA - 0.972***

- - 1.289***

-

(0.089) (0.113) 2R 0.866 0.969 0.773 0.977 0.971 0.978

SSR 0.598 0.140 1.002 0.101 0.126 0.098

F 70.480 658.558 70.465 763.622 435.230 692.873

WD 0.284 1.667 0.167 1.783 1.931 1.964

N 156 157 153 152 153 152

----------------------------------------------------------------------------------------------------------------------------- -----------------------------

Note: e = spot exchange rate, X = exports, M = imports, Y = U.S. nominal income, Y*= EMU nominal income, S = personal

saving rate, I = private investment, ND = national debt, CPI = consumer price indexes in U.S. CPI* = EMU consumer price

index, T = taxes, C = private consumption, CA = current account, KA = capital account, i = 3-month T-bill rate, i* = ECB

overnight rate, Poil = price of oil, WD = war dummy, and USXRI = U.S. exchange rate index.

Source: economagic.com, imfstatistics.org, Eurostat, and Bloomberg.com.

Proceedings of the Pennsylvania Economic Association 63

Table 4: Factors Affecting the Devaluation of the U.S. Dollar [Eq. (6)]

----------------------------------------------------------------------------------------------------------------------------- -----------------------------

Variables teln teln teln teln teln teln

----------------------------------------------------------------------------------------------------------------------------- -----------------------------

C -9.372***

-9.780***

-12.350***

-10.287***

-11.706***

-11.006***

(0.634) (0.778) (2.014) (0.647) (2.062) (2.135) *ln tP -1.444

** -1.284

* -0.067 -0.317 -0.263 -0.346

(0.611) (0.720) (0.740) (0.819) (0.755) (0.744)

tPln 0.848 0.423 1.034 0.834 1.109* 0.971

(0.560) (0.671) (0.664) (0.524) (0.663) (0.696)

tNDln 0.180***

0.121 -0.268* 0.095 -0.221 -0.185

(0.061) (0.076) (0.155) (0.062) (0.153) (0.161)

tKAln -0.002 -0.023 -0.029 *tt ii 0.008

** -0.009 -0.012

(0.024) (0.029) (0.038) (0.004) (0.008) (0.008)

tYln -0.364 0.076 0.827***

-0.723***

0.777**

0.905***

(0.266) (0.312) (0.311) (0.260) (0.302) (0.292) *ln tY 0.928

*** 0.818

*** 0.147

** 0.923

*** 0.130

* 0.054

(0.043) (0.053) (0.072) (0.042) (0.070) (0.068)

)1(AR - - 0.910***

- 0.916***

0.907***

(0.030) (0.030) (0.034)

)1(MA - 0.481***

- - - 0.194**

(0.076) (0.090) 2R 0.952 0.963 0.977 0.954 0.977 0.978

SSR 0.210 0.162 0.101 0.204 0.027 0.097

F 475.498 530.866 856.406 489.853 860.797 771.157

WD 1.042 1.718 1.741 1.089 1.696 1.966

N 150 150 149 150 149 149

----------------------------------------------------------------------------------------------------------------------------- -----------------------------

Note: See, Table 3.

Source: See, Table 3.

Proceedings of the Pennsylvania Economic Association 64

Graph 3: The Depreciation of the U.S. Dollar with Respect the Euro

Note: 1/EUS= the exchange rate between dollar and euro (€/$). Source: Economagic.com.

Graph 4: The Depreciation of the U.S. Dollar with Respect the Gold

Note: 1/GOLD=the value of U.S. dollar with respect the price of gold. (troy ounces of gold/$).

Source: Historical Gold Prices-1833 to Present. http://www.nma.org/pdf/gold/his_gold_prices.pdf

0.6

0.7

0.8

0.9

1.0

1.1

1.2

99 00 01 02 03 04 05 06 07 08 09 10 11 12 13

1/EUS

.000

.004

.008

.012

.016

.020

.024

.028

.032

40 45 50 55 60 65 70 75 80 85 90 95 00 05 10

1/GOLD

Proceedings of the Pennsylvania Economic Association 65

Table 8

U.S. Spot Exchange Rate Regression [Eq. (7)]

--------------------------------------------------------------------------

Variables teln teln

--------------------------------------------------------------------------

0 0.897 -2.662***

(0.566) (0.612)

oilPln 0.111***

0.130***

(0.020) (0.021)

tNDln -0.129**

0.252***

(0.066) (0.072)

tCAln 0.686***

-0.116

(0.132) (0.149)

WD 0.264***

0.058**

(0.026) (0.025)

EDCD - -0.026*

(0.026)

)1(MA - 1.257***

(0.076)

)2(MA - 1.048***

(0.100)

2R 0.849 0.969

SSR 0.614 0.126

WD 0.265 1.729

F 190.552 513.154

N 141 141

--------------------------------------------------------------------------

Note: teln = spot exchange rate, oilPln = price of oil, tNDln = ln national debt, tCAln = current account balance, WD =

war dummy, EDCD = European debt crisis dummy, MA = moving average process, 2R =R-squared, SSR = sum of squared

residuals, WD = Durbin-Watson statistic, F = F-statistic, N = number of observations, (*), (**), and (***) = significant at

the 10%, 5%, and 1% level, standard errors in parentheses. All variables are in natural log. Column 2 shows the correction for

the first-order serial correlation of the error term (D-W from 0.265 becomes 1.729).

Source: Economagic.com. Data from 1999:01 to 2010:12.

Proceedings of the Pennsylvania Economic Association 66

REFERENCES

Allen, Franklin and Douglas Gale. 2009. Understanding

Financial Crises. Clarendon Lectures in Finance, New York,

N.Y.: Oxford University Press.

Amstad, Marlene and Antoine Martin. 2011. Monetary

Policy Implementation: Common Goals but Different

Practices. Current Issues in Economics and Finance, Federal

Reserve Bank of New York, Volume 17, Number 7: 1-9.

Bonitsis, Theologos Homer. 2011. Eurozone

Competitiveness: An Analysis of the PIIGS. In 2011

Conference Proceedings Northeast Business & Economics

Association, Rajneesh Sharma (editor), Thirty-Eighth Annual

Meeting, November 3-5, 2011, Sheraton Society Hill Hotel,

Philadelphia, PA, U.S.A.

Dornbusch, Rudiger. 1976. Expectations and Exchange Rate

Dynamics. Journal of Political Economy. 84: 1161-1176.

Eiteman, David K., Arthur I. Stonehill, and Michael H.

Moffett. 2010. Multinational Business Finance, 12th

Edition,

Boston, U.S.A.: Prentice Hall.

Glain, Stephen J. 2003. Stocks Sour on Cheap Dollar World

markets Fall on Concerns over Devaluation Moves. Boston

Globe, September 23: D1.

Hale, David. 2005. The Risk of Dollar Devaluation. The

International Economy. 19 (2), Spring: 32-33.

Heath, Allister. 2004. Pressure Grows on G7 to Agree

Acceleration Dollar Devaluation. The Business, September

26: C2.

Kallianiotis, I. N. 2012a. The Single European Financial

Market and its Uncertain Future. Unpublished manuscript,

University of Scranton, March, pages 51.

Kallianiotis, I. N. 2012b. Privatization and its Anti-Social

Effects on National Economies. Hellas on the Web, February

2, 2012: 1-29. http://www.hellasontheweb.org/2010-04-05-

22-20-08/2010-04-06-19-07-44/12035-privatization-and-its-

anti-social-effects-on-national-economies.

Kallianiotis, I. N. 2012c. Greece’s Imposed Privatization

(Denationalization) and its Effects on Individuals’ Utility and

Social Welfare. Christian Vivliografia, February 20, 2012:

1-13.

http://christianvivliografia.wordpress.com/2012/02/20/greece

s-imposed-privatization-denationalization-and-its-effects-on-

individuals-utility-and-social-welfare/

Kallianiotis, I. N. 2011. Is the Imposed Global ‘Laissez-faire’

Socio-economic System Responsible for the Latest Financial

Crisis. Journal of Business and Economics. 2 (5), May: 325-

353.

Kallianiotis, I. N. 2010. Were the Current Bear Market and

the Recession Predictable? International Journal of Applied

Business and Economic Research. 8 (1): 37-64.

Kallianiotis, I. N. and Karen Bianchi. 2009. Speculation,

Uncertainty, Financial Assets Expected Return and Risk, and

Exchange Rate Determination. Unpublished manuscript,

University of Scranton, November 2009: pages 31.

Kallianiotis, I.N. and Dana M. Harris. 2010. What Went

Wrong with our International ‘Laissez-faire, Laissez-passer’

Economic System? International Research Journal of Finance

and Economics. 50, October: 96-122.

Kallianiotis, I.N. and Iordanis Petsas. 2008. Interdependence

between U.S. and EU Goods, Money, and Foreign markets

and Spillover Effects. Spoudai. 58 (1-2): 115-147.

Krugman, Paul. 2012. What Went Wrong? Folio, The

Magazine of the Graduate Center, The City University of

New York, Winter 2012: 20-23.

Morrissy, John. 2011. Currency Wars are Heating up: Dollar

Policy of Competitive Devaluation Slammed. The Windsor

Star. January 11, 2011: C8.

Nechio, Fernanda. 2011. Monetary Policy When One Size

Does Not Fit All. FRBSF Economic Letter, Federal Reserve

Bank of San Francisco. 6/14/2011: 1-5.

Pindyck, R.S. and Rubinfeld, D.L. 1981. Econometric

Models and Economic Forecasts. New York: McGraw-Hill.

Sharma, Shalendra D. 2011. U.S. Debt, Deficit, and the

Falling Greenback: Does it Mean Currency Wars and an End

to the Dollar’s Reign? SERI Quarterly. 4(3), July: 67-78.

Siegman, C.J. 2008. The Global Recession Risk: Dollar

Devaluation and the World Economy. Choice. 45(5),

January: 868.

Straubhaar, Thomas. 2004. The Euro is our Currency, and

the Dollar is your Problem. Intereconomics. 39(6),

November/December: 286-287.

Uri, Noel and Roy Boyd. 1991. The Impact of a Devaluation

of the U.S. Dollar on the Exports of the U.S. Agricultural

Commodities. Journal of World Trade. 25 (4), August: 5-27.

67

TRADE FLOWS BEETWEEN S. KOREA AND THE U.S.A

Orhan Kara

Department of Economics and Finance

West Chester University

West Chester, PA 19383

ABSTRACT This study examines the trade flows between Korea and the

U.S. after Korea adopted flexible exchange rate system. It is

found that the volatility of the exchange rate has a negative

impact on imports while it has a positive effect on exports.

Furthermore, the Korean exports are highly responsive to

changes in income; whereas, a percent change in income in

Korea leads to a one percent change in imports from the U.S.

Similarly, exports are more sensitive to changes in income

than that of imports. Finally, the relative prices have greater

impact on exports relative to imports.

INTRODUCTION

Korean foreign exchange market faced its first financial crisis

in 2008—2009 since the Asian financial crisis of 1997-1998.

Unlike the Asian finance crisis, until which Korea had an

exchange regime that was either freed or managed floating,

the new crisis took place since Korea adopted flexible

exchange rate system. In addition, the last crisis started in the

U.S., which is one of Korea’s top trading partners. What

followed the last financial crisis was the implementation of

fiscal and monetary policies in both countries. In the U.S.

the budget deficit soared, causing the national debt to reach

one hundred percent of the gross domestic product. The

federal funds rate was reduced to almost zero percent.

Similarly, Korea implemented policies to counter the

negative effect of the downturn. For example, the policy rate

fell to two percent from 5.25% and the Korean government

implemented new measures referred to as the Financial

Market Stabilization measures (Shin and Yoo, 2012).

Not only do stabilization policies affect exports and imports,

they also influence the exchange rates. On the one hand,

government policies cause a change on domestic price level,

leading fluctuations in the relative prices of imports and

exports. On the other hand, monetary policies, especially

interest rate changes, may result in volatility in exchange

rates in the short run. Depending upon the effect, trade

balance may improve or worsen. Therefore, it is important to

know how exports and imports respond to policies and how

sensitive the trade flows to changes in prices, exchange rates,

and volatility. Moreover, for a country such as Korea that has

been an example of export promoting growth strategy, it is

equally significant to find out how and what factors affect the

trade flows and if the trade flows respond any differently in

flexible exchange rate regime. This study aims at answering

those questions by investigating trade flows with export and

import demand functions.

To that and, the next section reviews literature. The section

after that presents the model and methodology. Then the

results are given in the following section. The last section

concludes the study.

LITERATURE REVIEW

An exemplary economic growth of Korea has served as a

model to developing nations, which involved heavy

government intervention with restrictions on trade and capital

flows as the early years of industrialization included

subsidies, credit, and a favorable exchange rate policy to

increase exports (Dornbusch and Park, 1987). When the

Bretton Woods exchange rate system was abolished, Korea

pegged its currency, won, to dollar and strictly regulated the

foreign exchange transactions from 1974 to 1980 (Lee,

2007). When the U.S. currency started appreciating in the

1980s, Korea switched to a controlled, floating effective rate

regime, which consisted of a basket of major trading

partners’ currencies and the Special Drawing Right (Hsing,

2009). After the Plaza Agreement, the Japanese currency, a

major trading partner of Korea, started appreciating, which in

turn increased the value of the won, leading to the Korean

government to allow the exchange rate to fluctuate within a

certain percent band in either direction in 1989 (Hsing,

2009). However, the won kept appreciating, resulting in

trade deficits, the Korean government announced a new

exchange rate policy called market average rate (Black,

1999). Black (1999) called this as a switch to an instrument-

based approach to stabilize the currency as Korea finally

achieved stabilization of the exchange rate. Between 1990

and 1995, the won was allowed to fluctuate by market forces

in a band of 0.4 percent to 2.25 percent (Hsing, 2009). As

the Asian financial crisis started, the Korean government

responded by widening the band to ten percent and

eventually allowing the won float freely on December 17,

1997 (Hsing, 1999; and Lee, 2007). Since then, Korea has a

68

flexible exchange rate system, same as the U.S., which is one

of Korea’s major trading partner.

Although adopting a flexible exchange rate regime has

certain advantages, such as independent monetary policy,

automatic adjustment to trade shocks, and an ability to avoid

speculative attracts as suggested by Frankel (2003), but it

also introduces more volatility in the exchange rate. Since

many countries switched to flexible exchange rate, a rich

literature in the volatility of exchange rates has been

emerged. However, no consensus has been reached as to

whether volatility has a positive or negative impact on trade

flows. A detailed survey of literature, conducted by Ozturk

and Kalyoncu (2009), showed that some studies found

negative effect while other studies discovered a positive

effect of exchange rate volatility on trade flows. For

example, Brada and Mendez (1988), McKenzie and Brooks

(1997), and McKenzie (1999) claimed that volatility

positively affected trade flows while Akhtar and Hilton

(1984), Chowdhury (1993), Vergil (2002), and Chit (2008)

found a negative effect of the volatility on trade. In their

study, Ozturk and Kalyoncu (2009) also examined the

Korean trade by employing Engle-Grainger based

cointegration method from 1980 to 2005, and concluded that

volatility had a significant negative effect. In addition, the

studies on the volatility of exchange rate on the Korean

economy also produced contradicting results (Doroodian,

1999; Doganlar, 2002; Poon, Choong, and Habibullah, 2005;

Park, 2007, Baak, Al-Mahmood, and Vixathep, 2007; Shin

and Yoo, 2012; and Bahmani-0skooee, Harvey, and Hegerty,

2012).

Focusing on the real exchange rates, Doroodian (1999)

investigated the effect of volatility in Korean, Indian, and

Malaysian international trade with quarterly data ranging

from 1973 to 1996. Doroodian (1999) asserted that the

volatility measures used by existing studies had certain

shortcomings and employed the GARCH procedure instead

of measures such as the standard deviation, deviation from

trend, difference between forward and spot rates, Gini mean

difference coefficient, coefficient of variation, and the scale

measure of variability. The empirical evidence illustrated that

Korean exports were negatively affected by the volatility of

the exchange rate. On the other hand, Doganlar (2002)

proxied the volatility with the moving sample standard

deviation of the growth of the real exchange rate and

analyzed the Turkish, Korean, Malaysian, Indonesian and

Pakistani quarterly data between 1980 and 1996. He found

that the volatility of the exchange rates decreased the real

exports (Doganlar, 2002). Similarly, Poon et al. (2005) used

the moving sample standard deviation of the growth of the

real exchange rate for the volatility measure and extended the

analysis to Japan, Singapore, Thailand, Indonesia, and Korea.

They also reached to the same conclusion; exchange rate

volatility was negatively correlated with exports. The inverse

relationship between the volatility in the exchange rates and

trade was also confirmed by Baak et al. (2007) who

investigated the bilateral trade between four Asian countries,

namely Hong King, Korea, Singapore, and Thailand, with the

U.S. and Japan from 1981 to 2004 by using the standard

deviation of the monthly real exchange rates as the measure

of volatility.

However, Park (2007) examined the relationship between

volatility in response to positive and negative shocks by

employing a quantile regression method to the exchange rate

between Korean won and the U.S. dollar and found a

stronger effect of positive shocks. Likewise, Bahmani-

Oskooee et al. (2012) studied ninety-six export industries and

twenty-nine import industries in the short and long run and

concluded that bilateral exports and imports in a majority of

the industries responded positively to volatility in the short-

run. In contrast, Shin and Yoo (2012) could not reach any

significant conclusion as to whether volatility had a negative

or positive effect.

In short, there is no consensus emerged with respect to the

volatility of the exchange and its effect of trade flows.

Studies often produced conflicting results as some found a

positive effect of the volatility on trade flows and others

indicated a negative effect while the rest could not reach any

conclusion. Therefore, this study further contributes by

empirically analyzing the bilateral trade flows between Korea

and the U.S. under the flexible exchange rate regime.

THE MODEL AND METHODOLOGY

In order to determine the effect of the volatility in exchange

rates on the trade flows between Korea and the U.S., we

formulate the functional form to be used in the empirical

analysis. Based on the studies investigating the determinants

of the trade flows, researchers agreed that the following

factors are the main determinants of trade flows: volatility,

relative prices, income, and exchange rates (Kara, 2012).

Therefore, following Bahmani-Oskooee and Kara (2008) and

Kara (2012), functional relation is formulated as follows:

Trade Flows = f(VOL,P, Y, EXC,) (1)

69

where TF is trade flows (imports and exports), P is relative

price index, Y is income, and EXC is exchange rate.

Exchange rate is defined as the number of units of Korean

currency per dollar, which means that an increase in EXC is a

depreciation of the Korean currency.

In order to apply the above functional form to the imports

and exports, we adopted the followings:

ttttt EPD

PMYVOLM ln)ln(lnlnln 4321

(2)

where M is the volume of imports from the USA, VOL is the

volatility, PM = the price of imports, PD = price of domestic

goods, Y = domestic income, E = nominal effective exchange

rate, and is an error term.

ttttt eEPXW

PXYTPVOLaX ln)ln(lnln 4321

(3)

where X is the volume of the Korea’s exports; VOL is the

volatility; (PX/PXW) is the relative price of a country’s

exports (PX) compared to the world export prices(PXW),

YTP is income of the trading partner, E is the nominal

effective exchange rate and e is an error term. ln before the

variables refers to natural logarithm of the variables.

The volatility variable is estimated from the following

formula (Baak et al., 2007):

2

1

2])(1

1ln[ i

tn

tmkikit RERRER

nVOL

(5)

where t is quarter, k is month, RER is reel exchange rate,

k=tm is the first month, and k=tn is the last month of the

quarter. Similar to Baak et al. (2007), the volatility is

computed the standard deviation includes current and the

previous quarter monthly exchange rates.

Equations (2) & (3) outline the long-run relationship between

trade flows and their determinants. In order to test the effect

of the variables, we incorporate the dynamic adjustment into

equations, which requires a specification in an error-

correction format. Following Pesaran et al. (1996, 2001, and

2009), we specify equations (2) & (3) in Autoregressive

Distributed Lag (ARDL) formats as follows:

tutMtEtPD

PM

tYtVOLitMn

ii

itEn

iiit

PD

PMn

iiitY

n

iitVOL

n

iitM

1ln51ln41)ln(31ln21ln1ln1

ln0

)ln(0

ln01

0

ln

(6)

tvtXtEtPXW

PX

tYTPtVOLitXm

ii

itEm

iiit

PXW

PXm

iiitYTP

m

iitVOL

m

iitX

1ln51ln41)ln(31ln21ln1ln1

ln'0

)ln('0

ln'01ln

0''ln

(7)

Again, the variables are defined as before. The error terms

are assumed iid( 0,2 ).

In this study, since we use ARDL approach developed by

Pesaran et al. (1996, 2001 & 2009) due to a relative ease in

estimations, we proceed to the two stages of the estimation.

First, by using an F-test the existence of cointegration is

determined by testing the significance of the lagged

variables. According to the asymptotic properties of the

distribution, an F-test table is provided by Pesaran et al

(2009). A decision is reached by comparing the test statistics

to the two critical values given in the table. When a test

statistic falls between the critical values, the result is

inclusive and other unit-root tests will be used. When a

statistic is larger than the upper bound of the critical values,

the existence of cointegration is accepted, and if test statistic

is below the lower bound, then the existence of cointegration

is rejected.

First, two null hypotheses (stating the non-existence of

cointegration) are constructed to determine the cointegration

among variables in each equation.

H0: 1=2=3=4=5=0 H1: Not H0

H0: 1=2=3=4=5=0 H1: Not H0

70

The above hypotheses are tested according to the test

statistics from the table.

After the completion of the first stage, the second stage

involves estimating the equations, applying error correction,

and making inferences for the estimated coefficients. We

estimate equations (6) and (7) based on the best model

chosen by the R-bar square criterion, Akaika information

criterion (AIC), Schwarz Bayesian criterion (SBC), and

Hannan-Quinn criterion (HQC). In addition, our selection

involves diagnostic statistics.

The quarterly data for this study covers the period from 1998

to 2012 (1998Q1-2012Q4). Part of data used in this study

was extracted from the IMF Financial Statistics, online and

the rest was obtained from IMF Direction of Trade Statistics.

The following data are used in this study:

o Korean imports from the U.S.

o Korean exports to the U.S.

o Unit value of the U.S. exports

o Unit value of exports of Korea

o Unit value of world exports

o Industrial production in Korea

o Industrial production in the U.S.

o Producer prices in Korea

o Nominal Effective Exchange Rates

o Real Effective Exchange Rates

RESULTS

Using quarterly data from 1998 to 2012 for the imports and

exports between Korea and the United States, the equations

(6) and (7) were estimated by applying an error-correction

model, specifically ARDL. As explained above, we obtained

F-statistics for testing the joint significance hypotheses in

equation (8). As the testing can be carried out based on; a) no

intercept, no trend variable, b) intercept and no trend

variable, and c) intercept and trend variables, we performed

six F test analysis. In most cases the F-statistic was

significant. For the case of intercept and no trend variable,

the F statistics were bigger than upper bound critical value,

which led to the existence of cointegration since the null

hypotheses were rejected (Table 1).

After establishing the existence of cointegration, we

proceeded to the next stage and estimated equations (6) and

(7) by first imposing four lags on each first differenced

variable. As a result, two groups of estimates are obtained

for: 1) imports and volatility, relative prices, income, and

exchange rate; and 2) exports and relative prices, income, and

exchange rate. For the import equation, Akaika information

criterion provided better results according to diagnostics

tests, and R-bar criterion indicated best result for the export

equation.

Table 2 illustrates the long run estimates as well as ARDL

estimates for the equation (6). Long run estimates confirm

our expected signs except for the relative price variable. An

increase in volatility reduces the imports, indicating a

negative effect. However, the volatility variable did not turn

out to be statistically significant and when measured in

absolute value, the effect is relatively small. Income variable

is highly statistically significant. Given that we use the log

of the variables, the coefficient can be interpreted as the

elasticity of income. Therefore, a one percent increase in the

Korean income is associated by a 1.057 percent increase in

the imports from the U.S., indicating almost a unit elasticity.

We expected that depreciation in currency makes imports

more expensive, which was supported by the estimated sign

as the exchange rate was defined as the number of Korean

currency per dollar. Ten percent depreciation in won

decrease Korean imports from the U.S. by 0.67 percent.

ARDL estimates indicate that adjustment process is very

short for the relative prices compared to the other variables

which can take as long as four quarters. Based on the

changing signs in the lags, the variables adjust over time to

long run values. For example, if there is an overshoot in one

period, there will be undershooting in the following period.

Diagnostic tests show the robustness of the estimates. For

instance, the value of R-bar squared was relatively high and

F-statistics was highly significant. Under the variables

column in the table, the number in parenthesis shows the

number of lags, which is the same in the other tables in the

article.

Table 3 provides the error correction estimates for the Korean

imports from the United States. In the variables column in

the table, the symbol Δ in front of the variables refers to the

differencing. For example, ΔImports means one period

difference in imports [imports-imports(-1)]. The number after

the variable refers to the further differencing in the variables:

Δimports1 indicates one period difference of ΔImports

[imports(-1)-imports(-2)], and so on. This notation is the

same in all.

Most of the variables are statistically significant in Table 3.

Ecm(-1) stands for the error correction coefficient, which

shows how fast the economy returns to the equilibrium (long

run values) once it is shocked. After a shock occurs (e.g., a

change in exchange rate), an adjustment process takes place

during the transition for the economy returning to its long-run

71

equilibrium values. This speed of this adjustment process is

indicated by the error correction term. We expect the sign of

the error correction coefficient is negative for the following

reason. When the estimated short- run values overshoot the

long-run equilibrium values, then the adjustment is

downward or vice versa (Greene, 2008). The magnitude of

the error correction coefficient is another important feature.

The larger the value of the error correction term in absolute

value, the faster the economy goes back to its long run

equilibrium values. Similarly, smaller value of the error

correction coefficient means a slower adjustment process for

the long-run equilibrium values.

With respect exports, Table 4 presents the long run and

ARDL estimates of the equation (7). All variables are

statistically significant at 95% significance level. Unlike the

import case which has a negative relationship with the

volatility, exports respond positively to the volatility in

exchange rates. Specifically, a one present increase in the

volatility increases Korean exports to the U.S. by 0.19

percent. Given the size of the coefficient for the income

variable, the U.S. imports from Korea is highly sensitive to

income. A one percent increase in the U.S. income leads to

over nine percent increase in the value of goods purchased

from Korea. Due to the high income elasticity, recessions in

the U. S. would cause a relatively big decrease in Korean

exports. Likewise, Korean exports to the U.S. are also

responsive to the changes in the Korean exports prices to the

world export prices. The estimate indicates a 1.81 percent

decrease in the exports if the relative prices go up by one

percent.

Moreover, exchange rate also has a large impact on Korea’s

exports. The estimated exchange rate coefficient has an

expected positive sign, meaning that depreciation in currency

makes the prices of Korean goods cheaper for American

consumers leading to an increase in Korean goods

consumption in the U.S. when won depreciates or dollar

appreciates. In terms of percentage effect, one percent

deprecation in Korean currency is associated with a 3.56

percent increase in the U.S. imports from Korea. Table 4

also gives the ARDL estimates illustrating how long the

response for changes in the variables given by the number of

lags. Given that the number of lags is as high as nine, the

effect of changes in the exchange rate on Korea’s exports to

the U. S. takes up to nine quarters.

The relative long adjustment process is also evident in Table

5 that provides the error correction estimates for the equation

(7). Many variables are, again, statistically significant

including the error correction coefficient (ecm(-1)).

Furthermore, diagnostic statistics show support for the

robustness of the analysis, except for the presence of serial

correlation. We have checked several diagnostic statistics for

the models in this study in addition to a visual examination of

the plot of residuals and fitted values.

CONCLUSIONS

This study analyzed the trade flows between Korea and the

U.S. after Korea adopted flexible exchange rate system. The

effects of exchange rate volatility, income, relative prices,

and exchange rate on trade flows were estimated with Import

and export demand functions. The main conclusions of this

study are as follows. First, exchange rate volatility has a

negative effect on imports, which is in line with the studies

by Akhtar and Hilton (1984), Chowdhury (1993), Doroodian

(1999), Vergil (2002), Baak et al. (2007), Chit (2008), Ozturk

and Kalyoncu (2009). With respect to exports, the volatility

has a significantly positive effect, lending support to

findings by Brada and Mendez (1988), McKenzie and Brooks

(1997), and McKenzie (1999), and Bahmani-Oskooee et al.

(2012). Second conclusion of this study is that Korean

exports to the U.S. are more sensitive to changes in the

exchange rate than that of imports. Third, imports have a

unit income elasticity while exports are very responsive to

income as the income elasticity was about nine in exports.

With respect to external balances, policy makers should

implement policies to counteract the decrease in exports

during downturns in the U. S. economy. Lastly, unlike

imports which are less sensitive to changes in relative prices,

exports are more sensitive to changes in relative prices.

72

TABLE 1. F-TEST RESULTS FOR COINTEGRATION

Equation F-Statistics

Import 7.724

Export 9.321

Critical values for F-Test at 95% Significance Level (intercept, No Trend):

Lower Bound=2.649 & Upper Bound= 3.805

TABLE 2. LONG-RUN COEFFICIENT ESTIMATES FOR IMPORTS

Variables Coefficient Standard Error T-Ratio P-value

Volatility -0.022 0.049 -0.447 0.658

Income 1.057 0.153 6.898 0.000

Relative Prices 0.373 1.699 0.219 0.828

Exchange Rate -0.666 0.389 -1.712 0.095

Constant 20.720 10.032 2.065 0.046

ARDL Estimates for Imports

Coefficient Standard Error T-Ratio P-value

Imports (-1) 0.073 0.117 0.623 0.537

Imports (-2) 0.064 0.108 0.590 0.558

Imports (-3) 0.511 0.117 4.363 0.000

Volatility -0.008 0.017 -0.451 0.654

Income 0.467 0.391 1.196 0.239

Income (-1) 1.375 0.587 2.344 0.024

Income (-2) -0.498 0.528 -0.943 0.352

Income (-3) -0.415 0.527 -0.787 0.436

Income (-4) -0.557 0.304 -1.830 0.075

Relative Prices -3.434 1.266 -2.712 0.010

Relative Prices (-1) 3.566 1.325 2.692 0.010

Exchange Rate -0.516 0.329 -1.567 0.125

Exchange Rate (-1) -0.350 0.388 -0.900 0.373

73

Exchange Rate (-2) 0.424 0.371 1.145 0.259

Exchange Rate (-3) -0.269 0.365 -0.737 0.465

Exchange Rate (-4) 0.476 0.234 2.039 0.048

CONSTANT 7.304 3.986 1.832 0.075

R-Bar-Squared

F-Stat.

Equation Log-likelihood

Akaike Info. Criterion

Schwarz Bayesian Criterion

DW-statistic

0.94041

55.2455

89.433

72.433

55.217

2.316

TABLE 3. ERROR CORRECTION ESTIMATES FOR IMPORTS

Variables Coefficient Standard Error T-Ratio P-value

ΔImports1 -0.575 0.127 -4.539 0.000

ΔImports2 -0.511 0.117 -4.363 0.000

ΔVolatility -0.008 0.017 -0.451 0.654

ΔIncome 0.467 0.391 1.196 0.238

ΔIncome1 1.470 0.390 3.766 0.001

ΔIncome2 0.972 0.389 2.500 0.016

ΔIncome3 0.557 0.304 1.830 0.074

ΔRelative prices -3.434 1.266 -2.712 0.010

ΔExchange Rate -0.516 0.329 -1.567 0.125

ΔExchange Rate 1 -0.631 0.284 -2.225 0.032

ΔExchange Rate 2 -0.207 0.271 -0.765 0.448

ΔExchange Rate 3 -0.476 0.234 -2.039 0.048

ecm(-1) -0.353 0.102 -3.457 0.001

R-Bar-Squared

F-Stat.

Equation Log-likelihood

Akaike Info. Criterion

Schwarz Bayesian Criterion

DW-statistic

0.706

11.397

89.433

72.433

55.217

2.316

74

TABLE 4. LONG-RUN COEFFICIENT ESTIMATES FOR EXPORTS

Variables Coefficient Standard Error T-Ratio P-value

Volatility 0.190 0.058 3.268 0.010

Income Trade

Partner 9.326 1.751 5.328 0.000

Relative Prices -1.810 0.231 -7.831 0.000

Exchange Rate 3.557 0.524 6.784 0.000

Constant -36.241 12.518 -2.895 0.018

ARDL Estimates for Exports

Coefficient Standard Error T-Ratio P-value

Exports (-1) 0.169 0.173 0.978 0.354

Exports (-2) -0.183 0.167 -1.090 0.304

Exports (-3) -0.547 0.181 -3.021 0.014

Exports (-4) 0.571 0.129 4.423 0.002

Exports (-5) -0.395 0.132 -2.997 0.015

Exports (-6) -0.238 0.187 -1.277 0.234

Volatility -0.067 0.021 -3.118 0.012

Volatility (-1) -0.039 0.022 -1.752 0.114

Volatility (-2) -0.011 0.019 -0.584 0.574

Volatility (-3) 0.021 0.026 0.811 0.438

Volatility (-4) 0.092 0.035 2.614 0.028

Volatility (-5) 0.106 0.041 2.587 0.029

Volatility (-6) 0.098 0.037 2.623 0.028

Volatility (-7) 0.052 0.032 1.656 0.132

Volatility (-8) 0.056 0.024 2.352 0.043

Income, Trading

Partner Income

6.896 2.409 2.862 0.019

Income (-1) 5.101 3.180 1.604 0.143

Income (-2) 2.999 2.746 1.092 0.303

Income (-3) 3.397 3.392 1.001 0.343

Income (-4) -9.033 2.934 -3.078 0.013

Income (-5) 9.397 1.958 4.798 0.001

Income (-6) -3.614 1.947 -1.856 0.096

Relative Prices -3.264 0.891 -3.665 0.005

Relative Prices (-1) -1.063 1.302 -0.816 0.435

Relative Prices (-2) 2.073 1.148 1.805 0.104

Relative Prices (-3) -2.502 0.927 -2.700 0.024

Relative Prices (-4) 1.014 0.835 1.214 0.256

Relative Prices (-5) -0.622 0.811 -0.767 0.463

75

Relative Prices (-6) 2.180 0.821 2.657 0.026

Relative Prices (-7) 1.176 1.025 1.148 0.281

Relative Prices (-8) -1.931 0.758 -2.549 0.031

Exchange Rate 2.096 0.620 3.378 0.008

Exchange Rate (-1) 1.133 0.664 1.708 0.122

Exchange Rate (-2) 0.854 0.676 1.263 0.238

Exchange Rate (-3) 1.027 0.774 1.326 0.218

Exchange Rate (-4) -0.120 0.736 -0.163 0.874

Exchange Rate (-5) 1.190 0.775 1.536 0.159

Exchange Rate (-6) -0.678 0.588 -1.153 0.278

Exchange Rate (-7) -0.446 0.575 -0.774 0.459

Exchange Rate (-8) -0.721 0.606 -1.189 0.265

Exchange Rate (-9) 1.442 0.340 4.245 0.002

CONSTANT -58.847 29.471 -1.997 0.077

R-Bar-Squared

F-Stat.

Equation Log-likelihood

Akaike Info. Criterion

Schwarz Bayesian Criterion

DW-statistic

0.962

31.582

137.743

95.743

55.175

3.187

TABLE 5. ERROR CORRECTION ESTIMATES FOR EXPORTS

Variables Coefficient Standard Error T-Ratio P-value

ΔExports1 0.793 0.285 2.786 0.015

ΔExports2 0.611 0.258 2.366 0.034

ΔExports3 0.063 0.209 0.302 0.767

ΔExports4 0.634 0.194 3.261 0.006

ΔExports5 0.238 0.187 1.277 0.224

ΔVolatility -0.067 0.021 -3.118 0.008

ΔVolatility1 -0.414 0.133 -3.107 0.008

ΔVolatility2 -0.425 0.136 -3.133 0.008

ΔVolatility3 -0.404 0.133 -3.032 0.010

ΔVolatility4 -0.313 0.112 -2.797 0.015

ΔVolatility5 -0.206 0.080 -2.592 0.022

ΔVolatility6 -0.108 0.048 -2.266 0.041

ΔVolatility7 -0.056 0.024 -2.352 0.035

ΔIncome 6.896 2.409 2.862 0.013

76

ΔIncome1 -3.146 3.632 -0.866 0.402

ΔIncome2 -0.147 3.434 -0.043 0.967

ΔIncome3 3.250 3.093 1.051 0.312

ΔIncome4 -5.783 2.386 -2.424 0.031

ΔIncome5 3.614 1.947 1.856 0.086

ΔRelative Prices -3.264 0.891 -3.665 0.003

ΔRelative Prices1 -1.389 0.961 -1.445 0.172

ΔRelative Prices2 0.685 0.852 0.803 0.436

ΔRelative Prices3 -1.818 0.925 -1.965 0.071

ΔRelative Prices4 -0.803 0.866 -0.927 0.371

ΔRelative Prices5 -1.426 0.759 -1.879 0.083

ΔRelative Prices6 0.755 0.747 1.010 0.331

ΔRelative Prices7 1.931 0.758 2.549 0.024

ΔExchange Rate 2.096 0.620 3.378 0.005

ΔExchange Rate1 -2.547 1.344 -1.895 0.081

ΔExchange Rate2 -1.693 1.244 -1.361 0.197

ΔExchange Rate3 -0.667 1.082 -0.616 0.548

ΔExchange Rate4 -0.786 0.964 -0.816 0.429

ΔExchange Rate5 0.403 0.718 0.562 0.584

ΔExchange Rate6 -0.275 0.568 -0.484 0.636

ΔExchange Rate7 -0.721 0.440 -1.637 0.126

ΔExchange Rate8 -1.442 0.340 -4.245 0.001

ecm(-1) -1.624 0.354 -4.593 0.001

R-Bar-Squared

F-Stat.

Equation Log-likelihood

Akaike Info. Criterion

Schwarz Bayesian Criterion

DW-statistic

0.802

6.570

137.743

95.743

55.175

3.187

REFERENCES

Akhtar, M. and Hilton, R. S. (1984). Effects of Exchange Rate Uncertainty on German and U.S. Trade. Federal Reserve Bank

of New York Quarterly Review, 9, pp. 7-16.

Baak, S. J., Al-Mahmood, M. A., and Vixathep, S. (2007). Exchange Rate Volatility and Exports from East Asian Countries to

Japan and the U.S. Applied Economics, 39, pp. 947-959.

Bahmani-Oskooee, M. and Kara, O. (2008). Relative Responsiveness of Trade Flows to a Change in Prices and Exchange Rate

In Developing Countries Journal of Economic Development, 33 (1), pp. 147-163

77

Bahmani-0skooee M., Harvey, H; and Hegerty, S. W. (2012). Exchange Rate Volatility and industry Trade Between The U.S.

and Korea. Journal of Economic Development, 37(1), pp. 1-27.

Black, S.W. (1999). Issues in Korean Exchange Rate Policy. Changes in Exchange Rates in Rapidly Developing Countries:

Theory, Practice, and Policy Issues. pp.269-283.

Brada, J. C. and Mendez, J. A. (1988). Exchange Rate Risk, Exchange Rate Regime, and the Volume of International Trade.

KYKLOS, 41, pp. 263-280.

Chit, M. M., (2008). Exchange Rate Volatility and Exports: Evidence from the Asian-China Free Trade Area. Journal of

Chinese Economic and Business Studies, 6,(3), pp. 261-277.

Chowdhury, A. R. (1993). Does Exchange rate Volatility Depress Trade Flows? Evidence from Error Correction Models. The

Review of Economics and Statistics, 76 , pp. 700-706.

Doganlar, M. (2002). Estimating Impact of Exchange Rate Volatility on Exports: Evidence from Asian Countries. Applied

Economies Letters, 9, pp. 859-863.

Doroodian, K. (1999). Does Exchange Rate Volatility Deter International Trade in Developing Countries? Journal of Asian

Economics, 10, pp. 465-474.

Dornbusch, R. & Park, Y. C. (1987). Korean Growth Policy. Brookings papers on Economic Activity. No. 2. pp. 389-444.

Frankel, J. A. (2003). Experience of Lessons From Exchange Rate Regimes in Emerging Markets. NBER Working Papers No

10032.

Hsing, Y. (2009). Modeling the Behavior of the KRW/USD Exchange rate and Policy Implications. Global Economic Review,

38 (2), pp. 205-214.

Greene, W. H., 2008. Econometric Analysis, 6th

Ed. New Jersey: Pearson/Prentice Hall.

Kara, O. (2013). Balancing the U.S. Trade Deficit with China: How Much Does Renminbi Need Appreciating against Dollar?

Journal of International Finance and Economics, 13 (2). pp. 71-82.

Lee. B. (2007). Economic Fundamentals and Exchange Rates Under Different Exchange Rate Regimes: Korean Experience.

Journal of Applied Economics, 10 (1)., pp. 137-159.

McKenzie, M. D., (1999). The Impact of Exchange Rate Volatility on International Trade Flows. Journal of Economic

Surveys, 13(1), pp. 71-106.

McKenzie, M. D. and Brooks, R. (1997). The Impact of Exchange Rate Volatility on German-US Trade Flows. Journal of

International Financial Markets, Institutions, and Money, 7, pp. 73-87.

Ozturk, I & Kalyoncu, H. (2009). Exchange Rate Volatility and Trade: An Empirical Investigation from Cross-country

Comparison. African Development Review, 21(3), pp. 499-513.

Pesaran, M.H., Y. Shin, & R.J. Smith (1996). Testing for the existence of a long-run relationship. In DAE Working paper, V.

9622, Department of Applied Economics, University of Cambridge.

Pesaran, M.H., Y. Shin, & R.J. Smith (2001). Bound Testing Approaches to the Analysis of Level Relationships, Journal of

Applied Econometrics, 6, pp. 289-326.

Pesaran, B., & Pesaran, M.H. (2009), Time Series Econometrics. Oxford: Oxford University Press.

Park, B. (2007). Trading Volume, Volatility, and GARCH Effects in the South Korean won/US Dollar Exchange Market:

Evidence from Conditional Quartile Estimation. The Japanese Economic Review, 58(3), pp. 382-399.

Shin, H. K. and Yoo, B. H. (2012). The Volatility of the Won-Dollar Exchange Rate During the 2008-9 Crisis. Journal of

Economic Development, 37(4), pp. 61-78.

78

Vergil, H. (2002). Exchange rate Volatility in Turkey and Its Effect on Trade Flows. Journal of Economic and Social

Research, 14(1), pp. 83-99.

A SIMPLE MODEL OF BASEBALL DESEGREGATION

Timothy F. Kearney & David Gargone

Department of Business

Misericordia University

Dallas, PA 18612

ABSTRACT Baseball began to desegregate in 1947 when Jackie Robinson

with the Brooklyn Dodgers broke the color line, a process

that would take until 1959. There is considerable research

about the positive impact from integration but no research

about the reasons that led to desegregation. We posit that

baseball’s weak financial position following the Depression

and WWII created pressures to desegregate. Using

attendance as a proxy for financial performance, we find the

more poorly performing team in the seven instances where

teams shared cities predicted which would integrate first.

INTRODUCTION

After decades of intransigence, baseball finally began the

process of desegregation in 1947 when the Jackie Robinson

with Brooklyn Dodgers of the National League and Larry

Doby with the Cleveland Indians of the American League

finally broke the color line. Official segregation had deprived

baseball of a significant pool of baseball talent, as 18 Hall of

Fame players spent time in the segregated Negro Leagues.

As would be expected, tapping that talent pool by major

league baseball (MLB) had significant impacts. In his article

The Cost of Discrimination: A study of Major League

Baseball, “A sizeable and statistically significant relationship

between winning and the number of black players is found

for the 1950s” (Hanssen 1998, p. 603). Nonetheless, MLB

retained segregated teams until 1959 when the Boston Red

Sox finally integrated.

In that article, Hanssen (1998) notes that “Why the color line

was broken remains an open question” (p. 605), and posits

changes in attitudes following WWII. We find that the Great

Depression, the subsequent World War II area and

demographic shifts West and South provided the incentives

for integration. In 1947, there were five cities with 11 teams

which shared MLB teams (Philadelphia A’s/Phillies; Boston

Braves/Red Sox; Chicago Cubs/White Sox; St. Louis

Browns/Cardinals; New York Yankees/Giants/Dodgers).

We consider each of these cities as economic pairs, and find

that the team in the city which was underperforming in terms

of attendance and had other financial difficulties integrated

first.

ECONOMICS OF DESEGREGATION

Gary Becker is well known for his path breaking studies of

labor market discrimination. In his work “The Economics of

Discrimination” (Becker, 1971), Becker identifies three

sources of discrimination: Employer discrimination, co-

worker discrimination and consumer discrimination. There

have been numerous studies of discrimination in sports,

generally showing returns to teams that ended discrimination.

Kahn (1991) surveys the literature on discrimination. He

finds that a number of studies show that teams which

integrated faster had better on-field results as a result. He

concludes this is what would be expected in a market that

eliminates employer discrimination. In terms of co-worker

discrimination, he notes that both members of teams

integrating as well as teams they would oppose balked at

playing on an integrated team/game). Oh and Buck (2012)

looked at wage differentials based on race and marginal

productivity and finds at least some evidence of all three

sources of discrimination in sports. Kuper and Szymanski

(2001, pp. 90-95) finds that there was discrimination, and

that integrating improved team records

Anecdotally, we know that both the Dodgers and Indians (the

first teams to integrate) strongly came down in favor of

integration. Dodger players presented Manager Leo

Durocher with a petition to jettison Robinson. Durocher told

them that “I’m the manager and I say he plays”. (Eig,

Opening Day, p. 44).

Kahn (1991) notes studies which demonstrate that consumer

discrimination was an important consideration. However,

given that integration of the field was followed by integration

of the grandstands, this effect is difficult to estimate.

Hanssen expressed it as “Because of the change in attitudes,

and the complementary rise of the black middle class, the

integration of baseball presented the prospect of a profit to be

made” (p. 605)

The elimination of employer discrimination presumes a free

market, where the inefficiencies of hiring lower ability

workers can be bid away. Baseball as an institution is

somewhat protected, as Major League Baseball (MLB)

enjoys anti-trust protections, which allows baseball to act as a

79

cartel, with their markets protected in general from

competition.

However, in the post WWII era, baseball teams were not

even distributed across the country. In fact, five cities had

multiple teams; Philadelphia, Boston, Chicago and St. Louis

had two teams and New York City had three teams. These

teams are geographically distributed across the eastern part of

the country, covering New England (Boston), the Upper Mid-

West (Chicago), the Mid-Atlantic (Philadelphia and New

York) and baseball’s West (St. Louis), though did not cross

the Mississippi. While anti-trust legislation protected the

leagues from competition, within cities there was strong

competition.

General Demographic Trends Favoring Integration

While the Dodgers and Jackie Robinson are rightfully well

known for integrating major league baseball, it is important

to recognize that American sports – and society – had been

desegregating throughout the 20th century. Baseball was a

late integrator. Of the major league sports in the USA, only

hockey integrated later, not until 1958, when Walter O’Ree

laced up for the Boston Bruins and became known as

“hockey’s Jackie Robinson”. We note that major sports were

integrated long before baseball:

• George Poage was the first American black to win a medal

in the Olympics, in track at the 1904 Olympic Games

• Boxing began to be fitfully integrated in the early 20th

century and by 1908 Jack Johnson won the main prize, the

heavyweight title.

• NCAA football began to be integrated in 1918

• Bobby Marshal integrated the NFL in 1920

Segregation was clearly not the simple result of ignorance by

baseball talent scouts of the abilities of African-American

players. Hall of Fame manager John McGraw tried to sign a

black player as early as 1901. “claiming that he was actually

Native American” (Chadwick, p. 28). Major league owners

were often very knowledgeable about their Negro League

counterparts. Negro League teams such as the New York

Cubans (NY York Giants Polo Grounds), Newark Eagles

(Newark Bears’ Rupert Field, home of the Yankee AAA

team), Homestead Grays (both Pittsburgh Pirates Forbes

Field and Washington Senators Griffith Park), and the East

West All Star Game (Chicago White Sox Comiskey Park)

among others gave white owners a chance to evaluate black

talent.

Importantly, lucrative barnstorming tours between white and

black players were common in the 20th century. In fact, in

the 1930s Future Hall of Fame Negro League pitcher Satchel

Paige was earning some $50k, reportedly second only to

Babe Ruth. (Chadwick, p. 124). Hall of Fame National

League St. Louis Cardinal pitcher Dizzy Dean barnstormed

with Paige from 1934 to 1945. Their relationship was an

important step forward, as “The color coded pairing of starts

gave a human face to the battles between white and black

teams which had been playing out in California for 25 years”

(Tye, p. 94) Importantly in a segregated society, “on the

barnstorming tour, ballparks that normally walled off blacks

let they where they wanted. It brought in white reporters

with white fans” (Tye, p94-95). Bob Feller barnstormed with

Paige, including a 1941 “big money matchup against Satchel

Paige and the (Negro league Kansas city) Monarchs at

Sportsman’s Park in St. Louis” and an important 1946 tour.

That tour, aided by the Flying Tigers Airlines, “provided

blacks with a ‘chance to prove ourselves against white

players’” (Gay, p 224).

In the early-1940s, the owner of the Philadelphia Phillies Bill

Veeck saw the Negro Leagues as a great, untapped area of

baseball talent. In 1943, Veeck made a bid to buy the “down

in the dumps” Philadelphia Phillies. “Veeck’s transformation

plan for the Phils included a secret weapon: Negro Leaguers”

(Eig, pp. 181-182). A few years later, Dodger GM Branch

Rickey recognized “the Negro Leagues contained a gold

mine of big league players “and hence he “saw inefficiency

and exploited it at the expense of his competitors”.

(Bradbury, p 130).

The Economic Downturn Leads to Integration

We contend that the economic pressures of the 1930s sowed

the seeds for integration. Discrimination is expensive;

economic research has shown that “discrimination is

expensive and its cost may reduce its incidence” (Hanssen,

p1998 p. 603). And while baseball as an entity enjoys anti-

trust protections, the structure of baseball in the 1930s

created competition within cities. This created competition

negated to some extent the benefit that monopoly granted.

Baseball had been static since 1908, when the Baltimore

Orioles moved to New York and became the Yankees. In all,

New York had three teams (Yankees, Giants and Dodgers);

two teams shared Philadelphia (A’s,Phillies), St. Louis

(Browns, Cardinals), Chicago (White Sox, Cubs) and Boston

(Braves, Red Sox). Single team cities were found only in

Cleveland, Cincinnati, Pittsburgh, Detroit and Washington.

The economic losses of baseball in the 1930s and WWII era

began to pressure ownership. MLB attendance fell from over

10mn in 1930 to a low of 6mn by 1933. It wasn’t until 1945

that the 10mn mark was reached again. This was matched

by demographic trends. The 1940 census recorded the first

drops in population for Boston, St. Louis and Philadelphia.

The 1950 census saw population peaks for Boston, St. Louis,

80

Philadelphia and Chicago, while the 1960 census saw the

first ever drop in population for New York City.

The 1930s saw the only decade on record where the average

MLB team sales price dropped in real 2002 dollars. In the

1930s, there were four sales, and the average price change

was a 33% reduction or about 4.1% per annum as compared

with a 0.3% annual drop for the Dow Jones Index. (Haupert,

http://eh.net/encyclopedia/article/haupert.mlb).

Our analysis is based on seven pairs in five cities: four cities

(Boston, Philadelphia, St. Louis and Chicago) with two

teams, and the two New York pairs (both the Dodgers and

Giants compared with the later integrating Yankees). We

also consider the Philadelphia situation in 1943when Veeck

intended to integrate (along with the actual 1953 integration

of the Athletics).

By comparing teams within a market, we are able to

recognize important variables which would exert similar

pressure on each team: local demographics including

population changes; state/local pressures on teams to

integrate; income growth and transportation issues.

Financial data for the pre-integration era are notoriously

difficult to find. There are no team salary statistics, nor are

there financial statements. We use attendance as a measure

of financial performance, in the absence of other potential

sources of financial information. However, we also consider

team histories to see if there is evidence of poor management

or financial performance to go along with attendance data.

We hypothesize that financial turmoil arising first during the

Depression and then during World War II sowed the seeds

for some teams to innovate and turn to integration to tap that

‘greatest font of talent. We use both relative attendances in

the 1930s until integration, and also team histories to see if

there is a developing pattern to the pace of integration. The

pattern of significant differences in attendance matching is

well established in the 1930s, and the underperforming team

in fact would ultimately integrate first. We note that the

1930s data predicts that the Phillies would integrate first, and

that the Phillies were stopped by baseball authorities from

doing so. (See Table 1).

World War II

Right after the 1941 Pearl Harbor attack, Commissioner

Landis petitioned the government for approval of the 1942

season. In what has become known as the “Green Light

Letter”, President Roosevelt replied in part “I honestly feel

that it would be best for the country to keep baseball going.

There will be fewer people unemployed and everybody will

work longer hours and harder than ever before. And that

means that they ought to have a chance for recreation and for

taking their minds off their work even more than before.”

Interestingly the President foresaw that “Even if the actual

quality to the teams is lowered by the greater use of older

players, this will not dampen the popularity of the sport”. A

chance to nudge baseball towards integration was bypassed

over the exigencies of the war effort.

But the draft took most white men into the service, leaving a

pool of players too young for military service (including one

15 year old pitching appearance), players who failed to

qualify for military service and players who were too old to

serve and past their baseball primes. These inconsistencies

intersected in with the sacrifices made by black troops during

the war. President Truman made the decision to integrate the

military, which made segregation impossible to keep as a

policy. The 1945 death of Commissioner Landis opened the

door to integration.

Model of Integration

With the door open, we still see an uneven pace of

integration. We posit that competition within markets led the

underperforming team to integrate first. Brooklyn Dodger

GM Branch Rickey famously said that the greatest font of

untapped talent was in the Negro Leagues, and that “The

Negroes will make us winners for years to come, and for that

I will bear being called a do-gooder…” We believe that this

impulse, the need to win and attract fans, led to integration.

We compare both financial histories and relative attendance

records.

Table 2 shows when each team played its first African

American player. We note that of the first seven teams to

integrate, five were in trouble in the 1930s and each of those

teams had changed cities within 10 years or fewer of

integration (Browns 1953; Braves 1953; Athletics 1956;

Dodgers, 1957; Giants 1957).

Team Histories and Relative Attendance

We consider cumulative attendance, from 1930 until

integration (as measured above). The charts below are

measured from 1930 (year 1) forward. In each case, we find

that the team that is underperforming at the box office, and

has a weak financial position, integrated first. We also

consider the team’s history to validate if it in fact had

financial troubles leading up to integration.

Philadelphia Phillies (Attempt to integrate, 1944)

The Phillies franchise foundered from the Depression into the

WWII years. By the early-1930s, deferred maintenance at

the Phillies home field the Baker Bowl was such that “Rather

than use lawn mowers, groundskeepers used goats instead”

(Gershman, p. 144). Into the 1930s, there was considerable

“agitation about the state of the Philadelphia facility in

general and the franchise in general” (Dewey and Acella, p.

81

462). National League owners “had enough votes to force

change in the ownership”, and the team went into new hands

by 1933. In 1938 the Baker Bowl closed, the team moved to

the Athletics home grounds Shibe Park and attendance fell

further. (Dewey and Acella, p. 463).

National League ownership again pressured the Phillies, and

in 1943 owner Nugent seemingly agreed to sell to Bill Veeck,

the impresario and then-owner of the, AAA Milwaukee

Brewers. Veeck made the mistake of informing

Commissioner Landis that he would stock the team with stars

from the Negro Leagues, and the sale was voted down. The

League agreed to a low bid from Mr. Cox, a local business

owner. (Eig, p. 182). Veeck followed through in July 1947

when, as owner of the Cleveland Indians, broke the American

Leagues’ color line with Hall of Fame player Larry Doby,

just months after Jackie Robinson’s introduction to the

National League.

Brooklyn Dodgers (Integrated 1947)

Brooklyn last showed a profit in 1930, and the phones at

times were shut off for non-payment (Dewey and Acella, p.

112). When Walter O’Malley become involved in the team

in 1938, “the Dodger situation...was so frightfully tangled

that no commentator was quite able to explain it. (Robert E.

Murphy, p. 44). Basically, the Dodgers had less money than

the New York Yankees, who performed strongly both on the

field and in the box office. “If the Dodgers wanted a winning

team, he had to tap new talent” (Kuper and Szymanski, p.

95).

St. Louis Browns (Integrated 1947)

St. Louis Brown attendance was the lowest in all of baseball

in the 1930s, and didn’t crack 100k in three separate years.

Grass maintenance was turned over to a goat. (Golenbock,

The Spirit of St. Louis, p. 322). In 1933, when owner Ball

died “No one wanted to buy the team” (Golenbock, p 270).

By 1941, owner Don Barnes ‘concluded he could not

successfully fight the Cardinals” (p 280). The team began

discussion to move to Los Angeles (Dewey and Acella,

p.557). Those plans were interrupted by WWII.

By 1947 when Sportsman’s Park was sold, the park was

‘badly run down, with a leaky roof, broken chairs and a

dilapidated clubhouses…despite a fix up, the Browns drew

only flies” (Golenbock, p. 320). In 1947 the Browns

integrated, just weeks after the Indians.

Boston Braves (Integrated 1948)

In the 1930s the Boston Braves began to sink under its debts

to what was “a desperate situation” (Dewey and Acella,

p.62). In 1933, the team was in financial distress which

ultimately led to a financial restructuring under league

auspices ((William Craig, p. 85).

New York Giants (Integrated 1948)

By the end of the 1930s, the New York Giants were ‘treading

water more seriously than any time…since the start of the

century…the team was beset by uncertainties about the

managerial abilities of the son” (Dewey & Acella, p. 383).

Chicago White Sox (Integrated 1951)

White Sox ownership struggles in the 1930s turned into

“years of wrangles that made the club’s efforts to stay above

water in the standings of almost secondary importance”

(Dewey and Acella, p. 168) This included the team’s

bankers determining that it was a ‘financial risk” in 1940.

This situation was in stark contrast with cross-town rival

Cubs, who were backed by Wrigley family money.

Philadelphia Athletics (Integrated 1953)

In this instance, we consider the post-1943 sale of the

Phillies. As noted, in 1943 the National League forced the

sale of the Phillies to new ownership, ultimately backed by

money from the DuPont Company. With DuPont money

behind the National League squad, the two Philadelphia

teams began to undergo a metamorphosis. Ownership

squabbles “erupted in the boardroom” (Dewey & Acella, p.

485) in 1949. The Mack and Shibe families constantly

fought. A 1950 deal which saw the team go into the hands

of the Macks ‘insured that the franchise would continue to be

run on a shoe string, never sure how the next payroll would

be met” (Jordan, p. 171).

CONCLUSIONS

We believe that the available evidence is that teams which

are poor financially and are doing poorly at the box office

were faced with an incentive to integrate. Once management

decided to integrate, co-worker integration is mooted. And

while consumer discrimination may remain a problem, the

data show that teams which integrated first performed better

on the field and were able to tap into a growing African

American consumer market. We believe our study helps

answer Hanssen’s question about “why do teams integrate”.

We believe it is competitive economic pressures.

82

Table 1: Attendance for Team Pairings from 1930 To 1939

Phillies Athletics Braves Red Sox

2,289,666 4,086,775 3,894,275 5,070,229

Browns Cardinals White Sox Cubs

1,393,827 4,054,118 4,109,937 8,791,668

Giants Yankees

7,516,744 9,089,9536,554,663

Boston

St. Louis Chicago

Brooklyn Dodgers

New York

Philadelphia

Table 2: Player Integration Dates by Team

Player Team Date Player Team Date

Jackie Robinson Dodgers 4/15/1947 Elston Howard Yankees 4/14/1955

Hank Thompson Browns 7/17/1947 Tom Alston Cardinals 4/13/1954

Monte Irvin Giants 7/8/1949 Elston Howard Yankees 4/14/1955

Sam Jethroe Braves 4/18/1950 Pumpsie Green Red Sox 7/21/1959

Minnie Minoso White Sox 5/1/1951 Ernie Banks Cubs 9/17/1953

Bob Trice Athletics 9/13/1953 John Kennedy Phillies 4/22/1957

Larry Doby Indians 7/5/1947 Carlos Paula Senators 9/6/1954

Curt Roberts Pirates 4/13/1954 Ozzie Virgil Tigers 6/6/1958

Nino Escalera Reds 4/17/1954

Figure 1: Philadelphia Phillies & Philadelphia Athletics Attendance (1930-1943)

Phillies: Dark, Athletics: Light

0

1,000,000

2,000,000

3,000,000

4,000,000

5,000,000

6,000,000

7,000,000

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Att

en

dan

ce

Years

83

Figure 2: New York Yankees & Brooklyn Dodgers Attendance (1930-1946)

Dodgers: Dark, Yankees: Light

Figure 3: Saint Louis Cardinals & Saint Louis Browns Attendance (1930-1946)

Browns: Dark, Cardinals: Light

0

2,000,000

4,000,000

6,000,000

8,000,000

10,000,000

12,000,000

14,000,000

16,000,000

18,000,000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Att

en

dan

ce

Years

0

1,000,000

2,000,000

3,000,000

4,000,000

5,000,000

6,000,000

7,000,000

8,000,000

9,000,000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Att

en

dan

ce

Years

84

Figure 4: Boston Red Sox & Boston Braves Attendance (1930-1947)

Braves: Dark, Red Sox: Light

Figure 5: New York Yankees & New York Giants Attendance (1930-1947)

Giants: Dark, Yankees: Light

0

2,000,000

4,000,000

6,000,000

8,000,000

10,000,000

12,000,000

14,000,000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Att

en

dan

ce

Years

0

2,000,000

4,000,000

6,000,000

8,000,000

10,000,000

12,000,000

14,000,000

16,000,000

18,000,000

20,000,000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Att

en

dan

ce

Years

85

Figure 6: Chicago White Sox & Chicago Cubs Attendance (1930-1950)

White Sox: Dark, Cubs: Light

Figure 7: Philadelphia Athletics & Philadelphia Phillies Attendance (1944-1952)

Athletics: Dark, Phillies: Light

0

2,000,000

4,000,000

6,000,000

8,000,000

10,000,000

12,000,000

14,000,000

16,000,000

18,000,000

20,000,000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Att

en

dan

ce

Years

0

1,000,000

2,000,000

3,000,000

4,000,000

5,000,000

6,000,000

7,000,000

8,000,000

1 2 3 4 5 6 7 8 9

Att

en

dan

ce

Years

86

REFERENCES

Becker, Gary S. The Economics of Discrimination. 1971.

Chicago, IL: University of Chicago Press.

Bradbury, J. C. The Baseball Economist. 2007. New York,

NY: Dutton Press.

Chadwick, Bruce. When the Game Was Black and White.

1992. New York, NY: Abbeville Press.

Craig, William J. Boston Braves: A Time Gone By. 2012.

Charleston, SC: History Press.

Dewey, Donald and Nicholas Acocella. 2005. Total

Ballclubs. Wilmington, DE: Sport Classic Books.

Eig, Jonathan. Opening Day. 2008. New York, NY: Simon

& Shuster.

Gay, Timothy. Satch, Dizzy and Rapid Robert. 2011. New

York, NY: Simon & Shuster.

Gershman, Michael. Diamonds. 1993. Boston, MA:

Houghton Mifflin.

Golenbock, Peter. The Spirit of St. Louis. 2000. New York,

NY: Avon Books.

Hanssen, Andrew. The Costs of Discrimination: A Study of

Major League Baseball. The Southern Economic Journal,

vol. 64, no. 3, January 1998: 603-627.

Haupert, Michael J. The Economic History of Baseball.

http://eh.net/encyclopedia/article/haupert.mlb.

Jordan, David M. The Athletics of Philadelphia. 1999.

Jefferson, NC: McFarland & Co.

Kahn, Lawrence. Discrimination in Professional Sports: A

Survey of the Literature. Industrial and Labor Relations

Review, vol. 44, no. 3, April 1991. 395-418.

Kuper, Simon and Stefan Szymanski. Soccernomics. 2012.

New York, NY: Nation Books.

Murphy, Robert E. After Many a Summer. 2009. New

York, NY: Sterling Publishing.

Oh, Derek and Andrew Buck. Consumer Discrimination in

Baseball. Temple University Working Papers.

Overmyer, James. Queen of the Negro Leagues. 1993.

Lanham, MD: Scarecrow Press.

Tye, Larry. Satchel: The Life and Times of an American

Legend. New York, NY: Random House.

87

IMPACT OF MIGRANT REMITTANCES ON EDUCATION OUTCOMES

Tanu Kohli

LAS Global Studies, University of Illinois

03 S. Wright St, MC 301, Champaign, IL 61821

ABSTRACT

Remittances help recipient households to earn extra income

and increase their standards of living over time. These higher

standards are sometimes reflected in greater expenditures

appropriated to the education of each child in the household,

thereby promoting better human development outcomes for

the migrant-sending households. This paper evaluates the

impact of remittances on education expenditures made by

remittance receiving households, and compare these

outcomes with households that do not receive remittances.

The dataset used for this analysis is the 64th

Round of

National Sample Survey conducted by the Government of

India. It is seen that remittance incomes increase the share of

education related expenditures in the household and

education investments in each child.

Keywords: education expenses, schooling, remittances, India

I. INTRODUCTION

Economic migration allows households to expand their

income possibilities and achieve a higher standard of living.

Migration also acts as a tipping point for households to adopt

cultural practices that bring the source communities and

destination communities closer to each other. This paper is an

extension of the existing studies on migration, focusing on

the impact of remittance incomes on the investments in

human capital in the source community. If remittances are

devoted towards higher investments in human capital,

sustained contributions to education in the present will aid

the creation of a better, productive workforce in the future;

thus promoting the economic development of the country in

the long run. The current paper attempts to understand these

long-term human capital investments by studying the

schooling expenses made by remittance receiving households

as compared to non-remittance receiving households. The

dependent variable of interest is the schooling expenditure as

a share of total expenditure that is made by a given

household. The primary objective to see whether remittance

receiving household tend to invest more in child schooling as

compared to other households. The data utilized for this

study is the 64th

round of the National Sample Survey (NSS)

of the Government of India conducted in 2007-08. This data

stands as an outlier to the usual remittance-education studies

focusing on the Mexico-USA migration corridor. The results

from ordinary least square (OLS) analysis show that

remittance receipt has a positive impact on education

expenditures, thus leading to higher human capital outcomes

for these households. The results from instrumental variables

(IV) analysis are inconclusive and warrantee the use of better

instruments to deal with the problem of endogenous

variables. Rest of the paper is arranged as following: section

II presents the literature review; section III introduces the

hypothesis and the model, elaborating on the dataset and the

variables used for the analysis; section IV summarizes the

results of the OLS analysis; section V introduces the

instrumental variables (IVs) and presents the results and;

section VI concludes with data shortages and future work in

this direction.

II. LITERATURE REVIEW

In the past decade, there has been an increased focus on the

impact of migration and remittances on the schooling

outcomes of children in migrant sending households.

Empirical studies address these effects by observing different

parameters of education such as retention rates, academic

performance and gender-based differences in school

enrolments. Studies concentrating on the impact of migration

on schooling outcomes are seen to reflect an ambiguous

impact of parental migration on educational attainment such

as reduction in college aspirations but an increase in

educational aspirations and retention. On the other hand, the

literature concentrating on the impact of remittances on

schooling outcomes usually finds a positive impact of the

receipt of remittances on schooling outcomes. This

discrepancy arises in part because increased expenditure

towards schooling indicates the choices made by remittance

receiving households with respect to building future human

capital; but it is not indicative of the choices made by

households in the long run, with respect to the migration

aspirations of the children. The ambiguity is rendered from

the positive impact of remittances on the ability to make

educational investments (financial effect), combined with a

negative impact of parental absence on the academic progress

of the children (non-monetary effect). McKenzie and

Rapoport (2011) study the effect of Mexican migration to the

United States but find a negative effect of migration on

school attendance and high school completion rates. These

results are indicative of possibility that when a parent

migrates, the onus of taking care of the household falls on the

older child. McKenzie and Rapoport (2011) find support for

88

this argument observing higher workforce participation for

younger males, or a greater rate of migration of these and

greater participation in housework by female children.

Antman (2011) makes similar observations as with respect to

father’s migration from Mexico to the U.S. Her results show

a reduction in study hours and increase in work hours for

both boys and girls, with the effect being more severe on

younger boys and girls, ages 12 to 15 years. Meyerhoefer and

Chen (2011) focus on the impact of parental migration on the

schooling lags created for school children in rural China and

reveal a similar story. The migration of a parent from rural

area to urban area exhibits a greater likelihood of the female

child falling behind by more than half a year of schooling.

On the other hand, the literature focusing on the relationship

of remittances and schooling decisions reflect clear and

positive impact of former on the latter. Edwards and Ureta

(2003) study the impact of remittance incomes on schooling

in El Salvador and find a significantly large impact of

remittances on schooling retention rates. Remittance

receiving households exhibit a lower hazard of dropping out

of school. Acosta (2006) also reaches similar conclusions, as

the receipt of remittances increases the probability of

remittance receiving households to keep children enrolled in

school and reduce the labor force participation of children.

Amuedo-Dorantes and Pozo (2010) study the impact of

remittance receipt on school attendance in the Dominican

Republic by comparing the outcome for households that have

migrants with those that do not have migrants. Isolating the

effect of remittances on children in non-migrant sending

households, the authors predict better schooling outcomes for

these children, compared to the households where one of the

members undertook migration, thus leaving the child

susceptible to hardships. Returns to education and

expectations about future migration also affects school

attendance, with low returns to education in the destination

countries and higher expectation to migrate leading to lesser

school attendance. The positive financial impact of

remittances is thus overcome by the negative, non-monetary

impact of migration. Perhaps an outlier to this generally

positive impact of remittances on education expenses is the

study of Albanian households by Cattaneo (2012) finds that

remittance incomes do not have any impact on education

expenditures. The author attributes this non-preference of

education to spending conditions put forward by migrants to

send remittances and to low returns to education in the

Albanian labor market combined with the importance given

to other more urgent consumption expenditures by the

households.

Three broad conclusions that can be made from the studies

reviewed above. First, the impact of remittances on schooling

outcomes is still a less explored area even though the impact

of remittances on education is more or less predictable.

Second, most of the studies exploring the relationship

between remittances and schooling (or even migration and

schooling) are concentrated in exploring the Mexican

education outcomes, making studies for other countries

virtually negligible. Third, the studies focusing on

remittances and education outcomes focus on the receipt of

remittances and not the amount of remittances. Thus, within

the remittance receiving households, the magnitude to which

remittances effect schooling is not explored and can

definitely be utilized for comparisons of outcomes for

different remittance receiving households.

III. HYPOTHESES AND MODEL

In this paper, the primary dependent variable is the share of

total consumption expenditure devoted to schooling expenses

in the survey year; henceforth referred as share of schooling

expenses. It is measured as the ratio of the household’s

annual schooling expenses to the household’s annual

consumption expenditure. Share of schooling expenses

variable is indicative of the choice made by remittance

receiving households towards higher human capital

investments, as compared to non-remittance receiving

households. While most of the studies reviewed above focus

on schooling enrolment, this data set provides information of

grade of schooling completed only. At any given time thus,

the continued status of school enrolment is not known.

Hence, annual schooling expenses become the most plausible

tool to estimate a household’s preference for education.

Consequently, the following hypotheses is developed- do

remittance receiving households contribute a greater share

of annual consumption expenses towards schooling expenses

and thus, are more conducive to the creation of human

capital.The general form of the education expenditure model

can be summarized as below-

……………………… (1)

Here, the dependent variable is given by, log share of

schooling expenditure = [

] and,

the primary independent variable of interest , is

remittance receipt, a dummy variable which assumes value 1

if the household receives remittances and 0 if the household

does not receive remittances.

Other control variables include economic variables such as

employment status of the head of the household and

employment status of the adults in the household;

demographic variables capture the household and individual

characteristics; education variables measure the educational

attainment of adults in the household and; migration

variables study the strength of influence migrating members

of the household have on the household.

89

Economic variables include employment status of the head

of the household and employment status of adults in the

household. An employed household head can ensure

continued flow of income, enabling the household to spend

more money on tuition and school supplies and satisfying

schooling requirements of each child. If the head of the

household is employed and responds to the economic status

as self-employed or as working in a household enterprise or

as a regular salaried/ wage employee or reported to have

worked in casual wage labor, the dummy takes the value 1. If

however, the employment status of the household head

includes responses such as did not work but was available for

work, attended educational institution, attended domestic

duties, retirees and remittance recipients and disabled, they

were included as unemployed and their employment status is

coded as 0. Since many family structures in India can be

multi-generational families, the head of the household might

not always be employed. For example, the head of the

household can be a retired grandfather with working sons and

daughters. In order to account for this possibility, proportion

of employed adults in the household is also used as an

economic variable. This variable is calculated as the ratio of

number of employed adults to total adults in the household.

The higher the proportion of employed adults in the

households, greater is the assurance that schooling of the

children in the household will not be disrupted.

Demographic variables influence the consumption patterns

of a household via the caste the household belongs to, family

structure of the household, location of the household and;

female participation in household decisions. A household

residing in the rural area will spend lesser of schooling

because of two reasons. First, the concentration of schools is

generally lower in rural areas than in urban areas. Second,

rural areas tend to have more government-run schools that

are completely funded and do not require students to spend

anything extra. Meanwhile in the urban areas, private schools

exist along with government schools which tend to tip the

balance of schooling expenditure in favor of urban areas

further. If the household is in a rural area, the dummy

assumes the value 1, otherwise 0. The caste system in India

reflects the difference in economic opportunities among

households from the reserved castes and the general castes.

Since India has free and compulsory schooling for children

from ages 6 to 14 years, the caste of household will not affect

the enrolment of children in schools. Caste, via inherent

difference in economic opportunities can however, affect the

access to non-tuition education expenses on school supplies.

The household’s caste is thus included as a dummy variable

which equals 1 if the household belongs to any of the

reserved backward castes and 0 otherwise. The dummy

variable for multigenerational family (=1) or not (=0) is

expected to have a negative relationship with share of

schooling expenses, as schooling children can pool their

schooling resources and use them more efficiently. It is also

possible that the presence of a greater number of household

members diverts consumption to other kind of consumption

needs, thus reducing educational expenditures. To further

address decision making process in a multi-generation

family, where spending decisions can be influenced by more

than one parent or couple, a variable measuring the

proportion of adult females in the household is added. It is

calculated as the ratio of total adult women and total adults in

a household. If a larger proportion of adults are women, they

can influence spending decisions with greater bargaining

power. Additionally, three variables addressing the role of

children in the household are added to the analysis. Total

number of school going children is expected to increase the

share of schooling expenses; proportion of female children is

expected to be negatively related to the share of schooling

expenses due to the preference to educate a male child and;

ratio of total children to total members in the household is

expected to exhibit a positive relationship with share of

schooling expenses since the household comprises of more

children than adults, naturally tipping the expenditure in

favor of education expenses.

Education variables in the model are divided in two main

groups- maximum education level of the household and

proportion of educated adult females at the primary,

secondary and graduate levels. The household member who

has completed the highest level of education will influence a

household’s perspective towards education spending.

Maximum education is added in lieu of education level of the

parent in the household as a multigenerational family will

have more than one parent who can influence the spending

decisions of the household. The maximum educational

attainment is divided in three dummy categories- primary

education takes the value 1 if the maximum education

attained by any household member is the completion of

primary schooling (up to 8 years of schooling) otherwise 0;

secondary education takes the value 1 if the maximum

education attained by any household member is completion

of secondary school (9 to 12 years of schooling) otherwise 0

and; graduate education takes the value 1 if the maximum

education attained by any household member is the

completion of graduate or post-graduate education otherwise

0. The reference category for the maximum education

variable is given by no educational attainment or illiteracy of

adults. The second group of education variables account for

the proportion of women educated at each education level.

The education expenditure outcomes will be worse for a

household with a greater proportion of women who

completed primary education than with a household with

greater proportion of women who completed secondary

education or graduate education.

Lastly, the migration variables are used to measure the

strength of relationship between the migrant and the

remittance receiving household. The survey design allows

creating a migration history variable which measures the

average years the household has witnessed migration. This

90

variable is indicative of the changing preferences of a

household that has been exposed to more developed societies.

Since more developed communities also exhibit better human

capital, the effect of the migration history variable on human

capital expenditure in the source community should be

positive. The second migration variable is the proportion of

employed migrants in the household. If there are more

employed migrants, they will remit more money, which in

turn will have a positive impact on the spending abilities of

the household.

The descriptive statistics for all variables is given in Table 1.

The sample breakdown is provided according to remittance

receiving and non-remittance receiving households to get an

estimation of the characteristics of each kind of household.

IV. RESULTS FROM OLS ANALYSIS

Results from the OLS method are summarized in Table 2

with robust standard errors reported in the parentheses and

the results significant at the 1% level. Columns 1 and 2 of

Table 2 summarize the results using different combinations

of independent variables listed above. Variables excluded

from column 2 are employment status of the household head

which is replaced by the proportion of employed adults in the

household to capture the multi-generational household effect.

The multi-generational household dummy in column 1 is

replaced by proportion of adult women in the household in

column 2 to account for the multi-generational effect, as well

as measure the relative bargaining power of females versus

males in the household. Instead of using the total number of

school going children, used in column 1, the ratio of total

children in the household is used in column 2. This variable,

along with the proportion of female adults and proportion of

employed adults can account for the effects of a multi-

generational family. Other variables such as dummy for

remittance receiving household, social characteristics of the

household, proportion of female children in the household

and education and migration variables are included in both

the columns.

It is seen that remittance receipt has a consistently strong

impact on the share of schooling increasing the share of

consumption expenditure devoted to education expenditures

anywhere from 9.6% to 16.6%. Thus, the remittance

experience not only relives credit constraints but also

encourages households to invest in the human capital of the

children in the household in order to secure a better future.

Such tendency might come from the exposure of the migrant

to a better living environment, thus pushing the family left

behind to aspire for similar standards via long run human

capital investments. It is also possible that these households

already give importance to education and the extra income

helps them realize their education goals. Households with an

employed head (column 1) and a larger proportion of

working adults (column 2) seem to have a negative impact on

the two dependent variables by 17.9% and 35.7%

respectively. This result goes against conventional wisdom of

consistency of incomes and higher investments in education.

The proportion of employed migrants in a household also

exhibits a similar negative relationship with the dependent

variables. A greater number of employed migrants would

thus reduce share of education expenditure by approximately

6%. One plausible reason for such unexpected behavior of

the employment variable could be that households prefer that

the children get into the labor force as soon as possible,

instead of investing many years in obtaining education. Such

expectations could lead to lesser investment in schooling. It

is also plausible that if the migrant from the household did

not acquire higher education but is economically successful,

the household might not give importance to education as well

and groom the children to be economic agents instead. For

example, 17% of the households had no literate adult in the

household while 37% households had adults who completed

primary education. On the other hand, only 16.29% of the

household had adults with bachelor degree or higher. Thus,

the household can give more importance to entering the labor

force rather than obtain education.

Among the demographic variables, the coefficients behave in

the expected manner. Households in rural areas tend to spend

approximately 23.9% less on schooling expenses as

compared to a household residing in the urban area. This

difference in spending pattern can arise due to two reasons.

First, rural areas have more government sponsored schools

that do not require any additional investment on schooling or

schooling supplies from the parents. This shrinks the share of

schooling expense in rural areas compared to households in

urban areas where the children might go to private schools

and spend on their own books, extra tuition and other school

fees. Second, rural areas in general have lesser number of

schools, which along with a household’s requirement for

farm and non-farm labor can lead to lesser children enrolled

in schools and thus, lower education expenditure for rural

areas. Households from reserved backward castes devote

12% lesser consumption towards education expenditure than

a non-reserved caste household. As mentioned in section IV

above, this can be due to differences in economic

opportunities of the household because of being a lower caste

household instead of lack of access to education per se. A

multi-generation household also contributes 43% lesser to

education expenditure as compared to a nuclear household.

This however, does not imply that multi-generational

households assign lower importance to human capital. A

more likely explanation is that the household shares

education resources and thus has to spend lesser portion of

the consumption budget on school supplies. For example,

siblings can share school supplies, recycle the same books for

years before discarding them and the teach each other thus

eliminating the need for tutoring.

91

The variables related to children in the household behave

more or less as expected. A larger number of school-going

children in the household leads to a greater share of

consumption expenditure devoted to education expenses

(19.2%). In column 2, the total number of school going

children is replaced by the proportion of children in the

household and it is expected that a greater share of children

in the household will increase the share of consumption on

schooling expenditure. It is seen that with greater proportion

of children, the share of schooling expenditure increases by

almost 119%. Lastly, higher the number of female children in

the household lesser is the share of schooling expenditure

(average 11.8% lesser) in that household. This reflects the

preference for investing in the human capital of a male child

compared to a female child. The general opinion is that while

a male child will have to be the bread-winner of his family in

the future, a female child will be fine without work since she

can get married and secure her future.

Variables related to the education of adults in the household

also behave as expected of them. As the maximum education

obtained by any member in a household increases, the share

of schooling expenditure increased as well. Therefore, if

compared to household with illiterate adults, the share of

expenditure on education is 76.3% to 92.6% more for

households with maximum educational attainment at the

primary level; 130.3% to 149.4% for maximum educational

attainment at the secondary level and; 131.9% to 147% for

maximum educational attainment at the graduate level. As

the household’s education benchmark increases, they lay

higher premium on obtaining schooling for their children. It

is also seen that as more women acquire higher education in

the household, the education outcomes for the children in the

household also improve. Thus, while households where a

greater number of women completed secondary education

spend an average of 31.2% of their consumption expenditure

on education expenses while households with a greater

number of women with graduate education spend an average

of 40.4% of their consumption expenditure on education

expenses.

Migration history of a household has a small but positive

impact of 0.6% on share of education expenditures. Thus, the

exposure to a more developed society encourages households

to reach human capital outcomes similar to those societies.

However, the weak relationship shows that this variable

might not be a crucial determinant of education outcomes.

On the other hand, as the proportion of employed migrants

from a household increases, the schooling expenses incurred

by the household decreases by approximately 5.9%. If the

households see the economic benefits of migration, they

might substitute away from investing in schooling and push

children to become migrants, thus not requiring investments

in schooling. This negative impact of migration but positive

impact of remittances on education aspirations is similar to

the results of studies by Kandel and Kao (2001), Hanson and

Woodruff (2003) and Amuedo-Dorantes and Pozo (2010).

V. RESULTS FROM THE IV ANALYSIS

OLS estimates provide an extremely optimistic picture

regarding the effect of remittance receipt on the dependent

variables. This model however, suffers from potential

endogeneity issues. Remittance receipt will change the

consumption patterns and increase schooling investments, but

in some cases, where the migrant might be a close relative,

remittances might be received specifically to improve

schooling outcomes for the children (to pay for a tutor for a

poor performing child or to buy a computer). The IV analysis

is built on the results presented by OLS regression analysis in

column 1 of Table 2. The Durbin-Wu-Hausman test for

endogenous variables yields an F statistic greater than 10 and

p-value less than 0.05, thus requiring an IV analysis.

The first instrument is district-wise concentration of post

offices, obtained from the Indian Postal Services in the year

2007-08. Post offices allow easy transfer of monies and have

a deeper concentration of branches than commercial banks in

India. The second instrument is the state-wise and sector-

wise unemployment rate in the survey year 2007-08. While a

stronger network of post office will facilitate the transfer of

remittances, the unemployment rates, high unemployment at

the source will encourage the migrant to remit money to the

household. There is however, a strong possibility that high

unemployment will affect the household income, and thus

schooling expenses. To test that unemployment and the

dependent variable do not share a strong relationship, the

correlation between them is calculated. Unemployment and

share of schooling expenses exhibit a weak correlation, thus

allowing the use of these two instruments. The results from

these tests are summarized in Table 3 below. The

independent variables are the same as in Table 2, column 1

and include, remittance receipt (0/1 dummy), employment

status of the head of the household, rural or urban location of

the household, reserved caste status of the household, multi-

generational household and total children and proportion of

female children in the household. Education variable include

maximum education dummies with illiterate as the reference

category and proportion of educated females at each level of

schooling completed. Migration variables include average

years the household has witnessed migration and the

proportion of employed migrants in the household.

IV results show a positive impact of remittance receipt on

share of schooling expenses out of the total household

budget. That is, if the household receives remittances, it tends

to invest 141.5% more than non-remittance receiving

household towards share of education expenses. While this

positive relationship is encouraging, the value of the

coefficient is extremely high, which seems to raise some

concerns. Among other economic variables, an employed

92

household head is expected to devote 3.5% more towards

education expenses out of the total household budget, but this

value is insignificant. This result is expected, as an employed

head will be able to invest more in the educational attainment

of the children. The remaining variables do not behave

differently from the OLS results in Table 2. If the household

resides in a rural area, it will invest 27.5% lesser

consumption expenditure towards schooling and membership

in the reserved caste shows that households contribute 11.1%

lesser towards share of schooling expenses as compared to an

urban household and a non-reserved household respectively.

Multi-generational households spend a lesser portion of their

entire consumption expenditure on schooling expenses which

supports the previous assumption that such households might

be pooling resources and older children, might be helping

their younger siblings which reduces the need to spend more

on education expenses. As the number of school going

children increases, the share of schooling expenditure

increases by approximately 18%. A household where a

higher number of children are female, the share of schooling

expenses reduces by 12.8%.

As the years of completed schooling by a household

increases, the percentage share of consumption expenditure

on schooling expenses also increases. Thus, while households

where at least one adult completed primary school will spend

82.5% more on share of schooling expenses, than household

where none of the adults were educated; for households that

had at least one graduate the share of schooling expenses is

approximately 141% higher. Similarly, as the proportion of

women with completed primary, secondary and graduate

education increase, the share of schooling expenses of the

household increase. Therefore, a household with greater

proportion of female with graduate education would devote

35.9% of the consumption expenditure to schooling, as

compared to a household where the larger proportion of

women completed only secondary education. These

education variables present a picture similar to the

expectations that were set for them earlier in the paper. A

household with higher educational attainment will place

higher premium on schooling.

The two migration variables, migration history and

proportion of employed migrants do not change substantially

from the results reported in Table 2. As a household’s

average year of exposure to migration increases, it invests

approximately 1.1% share towards schooling. In order to test

the validity of this result, migration history of the household

is divided into three periods. If the household has had

exposure to migration in the last five years, the migration

history of the household is short, while medium term

exposure implies an average of five to 10 years since the

household sent a migrant. Breaking down this variable

provides a clearer picture of household’s schooling expenses.

It is seen that a household with short history of migration

would in fact reduce the share of schooling expenses as

compared to a household that has been exposed to migration

for a longer time. That is households with a long term

migration history spend 17.7% more on share of schooling

expenses as compared to a household with recent exposure to

migration. These results seem to indicate that as soon as a

migrant leaves the household, there is a disruption in the

household budget, which would affect the schooling

expenses as well (this disruption could occur if the household

had to divert resources from regular consumption towards

costs of migration). However, as the migrant settles at the

destination, and sends regular remittances, the share of

schooling expenses tend to increase. Antman (2011),

McKenzie and Rapoport (2011) and Meyerhoefer and Chen

(2011) find similar disruptions and reduction in schooling

attainment in migrant sending households. Additionally the

household, witnessing the benefits of migration (especially if

the migrant has high human capital), will tend to increase the

schooling investments of the current generation.

Post-estimation Tests- Over identification tests for the

instruments listed at the end of Table 3 show that the first

model for schooling expenses with share of schooling

expenses as the dependent variable is correctly identified by

using state-wise and sector-wise unemployment rates and

district-wise concentration of post offices as instruments. The

Sargan-Basmann scores are reported in column 1. For the

second model with schooling expenses per child however,

these instruments fail to correctly identify the model (column

2). This warranties the use of alternative instruments that can

better predict the impact of remittance receipt on schooling

outcomes.

VI. SUMMARY AND CONCLUSIONS

The primary objective of this paper was to observe if the

receipt of remittances by surveyed households leads to higher

investments in education in the household. A positive impact

of remittances on education expenditure would mean that not

only remittance incomes enable households to enjoy a higher

level of consumption, but also enable them to enjoy sustained

development by assisting the creation of higher human

capital of the children. The dependent variable chosen to

explore this impact of remittances was the share of schooling

expenditure out of total consumption expenditure. This

variable was chosen as a measure of educational attainment

in the household due to lack of data available on enrolment

rates. The results from the OLS analysis showed that

remittance receiving households devote more money towards

educational expenditure. The OLS model however, was seen

to suffer from endogeneity and to correct for this error, state-

wise sectoral unemployment rates and district-wise

concentration of post offices are introduced as instruments.

The model is recalculated and it is seen that remittance

receipt has a positive impact on the share of schooling

expenses. The model is correctly identified but the value of

both the coefficients is extremely high, which begs for

93

further investigation in terms of better instruments that can

provide more accurate results. Possible instruments could

include the district-level data on natural calamities such as

rainfall, as used by Munshi (2003) or droughts. Many studies

use destination community unemployment rates, but the lack

of data on migrant destination stops the use of this IV.

Table 1 - Descriptive statistics for schooling models

Variable Remittance receiving

households

Non- remittance receiving

households

Mean S.D. Mean S.D.

Annual schooling expenses 5011.67 10040.35 4345.40 10232.46

Share of schooling expenses

in total consumption

expenditure

0.0670 0.0769 0.0556 0.07115

Schooling expense per child 2989.19 6109.25 2650.34 7314.48

Remittance receipt 1 0 0 0

Amount of remittances 27102.13 49732.58 0 0

Employment status of the

head of the household

0.6601 0.4736 0.8582 0.3487

Proportion of employed

adults in the household

0.4722 0.3452 0.6017 0.2729

Rural household 0.6845 0.4647 0.6753 0.4682

Reserved caste 0.6550 0.4753 0.6745 0.4685

Multigenerational household 0.4295 0.4950 0.4750 0.4993

Sex of the head of the

household, male=1

0.6484 0.4774 0.8766 0.3287

Proportion of female adults in

the household

0.6282 0.2493 0.5053 0.1811

Number of school-aged

children (6 years to 17 years)

1.0490 1.3119 1.0450 1.3016

Proportion of female children

in the household

0.4680 03722 0.4582 0.3763

Ratio of total children 0.2913 0.2612 0.2492 0.2224

Maximum education

Primary schooling

Secondary schooling

Graduate education

0.4056

0.3145

0.1723

0.4910

0.4645

0.3776

0.4121

0.3126

0.1808

.4922

.4636

.3849

Education of female members

Primary schooling

Secondary schooling

Graduate education

0.2070

0.1632

0.0636

0.2868

0.3092

0.2084

0.1557

0.1370

0.0561

0.2075

0.2844

0.1948

Migration history 5.9871 5.8646 7.0317 7.0284

Proportion of employed

migrants

0.8398 0.2391 0.3982 0.4494

94

Table 2 - OLS estimates for share of schooling expenses

and schooling expense per child

(1) (2)

Economic Variables

Remittance receiving household 0.1667***

(0.0159)

0.0969***

(0.0166)

Employed household head -0.1791***

(0.1617)

--

Proportion of employed adults in the household -- -0.3579***

(0.0226)

Demographic Variables Household resides in rural area -0.2402***

(0.0153)

-0.2385***

(0.0156)

Household belongs to a reserved caste -0.1259***

(0.1427)

-0.1152***

(0.0146)

Household is multi-generational -0.4302

(0.0133)

--

Proportion of adult women in the household -- -0.1157***

(0.1926)

Total children of school age (6 to 17 years) 0.1928***

(0.0053)

--

Proportion of total children in the household -- 1.1980***

(0.0478)

Proportion of female children in the household -0.1221***

(0.0185)

-0.1157***

(0.0192)

Education Variables I- Maximum education dummies with illiterate as the reference category

Dummy for primary schooling as maximum education 0.7634***

(0.0748)

0.9267***

(0.0740)

Dummy for secondary schooling as maximum education 1.3035***

(0.0757)

1.4940***

(0.0752)

Dummy for graduate education as maximum education 1.3194***

(0.0789)

1.4701***

(0.0790)

Education Variables II- Proportion of educated adult females from each education group

Adult females with primary schooling 0.4102***

(0.0250)

0.2797***

(0.0268)

Adult females with secondary schooling 0.3680***

(0.0272)

0.2544***

(0.0285)

Adult females with graduate education 0.4680***

(0.0482)

0.3404***

(0.0510)

Migration Variables- Migration history of the household 0.0086***

(0.0009)

0.0044***

(0.0009)

Proportion of employed migrants from the household -0.0608***

(0.0182)

-0.0585***

(0.0187)

Number of observations 26436 26436

R-square 0.2171 0.1722

Standard errors are in the parenthesis; ***Significant at 1% level

95

Table 3 - IV estimates for share of schooling expenses and schooling expense per child using

unemployment and district-wise concentration of post offices as instruments

2SLS regressions

Economic variables

Remittance receiving household 1.4155***

(0.2026)

Employed household head 0.0358

(0.0390)

Demographic variables

Household resides in rural area -0.2754***

(0.0176)

Household belongs to a reserved caste -0.1117***

(0.0159)

Household is multi-generational -0.4570***

(0.0156)

Total children of school age (6 to 17 years) 0.1798***

(0.0060)

Proportion of female children in the household -0.1287***

(0.0196)

Education Variables I- Maximum education dummies with illiterate as the reference category Dummy for primary schooling as maximum education 0.8255***

(0.0696)

Dummy for secondary schooling as maximum education 1.3632***

(0.0705)

Dummy for graduate education as maximum education 1.4101***

(0.0751)

Education Variables II- Proportion of educated adult females from each education group

Adult females with primary schooling 0.2644***

(0.0372)

Adult females with secondary schooling 0.2586***

(0.0350)

Adult females with graduate education 0.3591***

(0.0555)

Migration variables

Migration history of the household 0.0114***

(0.0019)

Proportion of employed migrants from the household -0.8276

(0.1255)

Number of observations 26434

First stage correlation tests-

F- statistic 101.66

Prob > F 0.0000

Over-identification tests

Sargan score 2.4832

(p = 0.1151)

Basmann score 2.4818

(p = 0.1152)

Standard errors in the parentheses; *** Significant at 1% level; ** Significant at 5% level

96

REFERENCES

Acosta, Pablo. "Labor Supply, School Attendance, and

Remittances from International Migration: the Case of El

Salvador." World Bank Policy Research Working Paper 3903

(2006).

Amuedo-Dorantes, Catalina, and Susan Pozo. "Accounting

for Remittance and Migration Effects on Children’s

Schooling." World Development 38, no. 12 (2010): 1747-

1759.

Antman, Francisca M. "The Intergenerational Effects of

Paternal Migration on Schooling and Work: What Can We

Learn from Children's Time Allocations?."Journal of

Development Economics 96, no. 2 (2011): 200-208.

Becker, Gary S., Kevin M. Murphy, and Robert Tamura.

"Human Capital, Fertility, and Economic Growth." Journal of

Political Economy 98, no. 5 part 2 (1990): S12-S37.

Cattaneo, Cristina. "Migrants’ International Transfers and

Educational Expenditure: Empirical Evidence from Albania."

(2010).

Das, Saswati, and Diganta Mukherjee. "Role of Women in

Schooling and Child Labour Decision: the Case of Urban

Boys in India." Social Indicators Research 82, no. 3 (2007):

463-486.

Das, Saswati, and Diganta Mukherjee. "Measuring

Deprivation Due to Child Work and Child Labour: A Study

for Indian Children." Child Indicators Research 4, no. 3

(2011): 453-466.

Edwards, Alejandra Cox, and Manuelita Ureta. "International

Migration, Remittances, and Schooling: Evidence from El

Salvador." Journal of Development Economics 72, no. 2

(2003): 429-461.

Hanson, Gordon H., and Christopher Woodruff. "Emigration

and Educational Attainment in Mexico." Documento de

Trabajo del IR/PS. Disponible en http://irpshome. ucsd.

edu/faculty/gohanson/working_papers. htm (2003).

Kandel, William, and Grace Kao. "Impact of Temporary

Labor Migration on Mexican Children's Educational

Aspirations and Performance." International Migration

Review 35, no. (2001): 1205-1231.

Katz, Eliakim, and Oded Stark. Desired Fertility and

Migration in LDCs: Signing the Connection. Center for

Population Studies, Migration and Development Program,

Harvard University, 1985.

Katz, Eliakim, and Oded Stark. "Labor Migration and Risk

Aversion in Less Developed Countries." Journal of Labor

Economics (1986): 134-149.

McKenzie, David, and Hillel Rapoport. "Can Migration

Reduce Educational Attainment? Evidence from Mexico."

Journal of Population Economics 24, no. 4 (2011): 1331-

1358.

Meyerhoefer, Chad D., and C. J. Chen. "The Effect of

Parental Labor Migration on Children’s Educational Progress

in Rural China." Review of Economics of the Household 9,

no. 3 (2011): 379-396.

Munshi, Kaivan. "Networks in the modern economy:

Mexican migrants in the US labor market." The Quarterly

Journal of Economics 118, no. 2 (2003): 549-599.

Stark, Oded, and David E. Bloom. "The New Economics of

Labor Migration." The American Economic Review 75, no. 2

(1985): 173-178.

Stark, Oded, and J. Edward Taylor. "Relative Deprivation

and International Migration." Demography 26, no. 1 (1989):

1-14.

Stark, Oded, and Robert EB Lucas. "Migration, Remittances,

and the Family." Economic Development and Cultural

Change (1988): 465-481.

Yang, Dean. "International Migration, Remittances and

Household Investment: Evidence from Philippine Migrants’

Exchange Rate Shocks." The Economic Journal 118, no. 528

(2008):591-630.

Proceedings of the Pennsylvania Economic Association 97

MANUFACTURING PRODUCTIVITY IN PENNSYLVANIA

James A. Kurre

The Sam and Irene Black School of Business

Penn State Erie, The Behrend College

5101 Jordan Road

Erie, PA 16563-1400

ABSTRACT

Productivity is a crucial issue for any economy, a key

determinant of the standard of living of residents in an area.

And one thing about productivity is clear: it varies

dramatically across places and across industries. So how

does Pennsylvania and its metro areas fare in terms of

manufacturing productivity compared to the nation and other

areas? This paper documents how Pennsylvania’s

manufacturing sector stacks up—or doesn’t—compared to

other states and the nation for the manufacturing supersector

and for selected subindustries. And it further details the

productivity data for PA’s metro areas.

INTRODUCTION

“Whoever could make two ears of corn or two blades of grass

grow upon a spot of ground where only one grew before,

would deserve better of mankind, and do more essential

service to his country than the whole race of politicians put

together.”

-Jonathan Swift, 1667-1745

Three hundred years ago Jonathan Swift understood the

importance of productivity. Economists, too, value

productivity and have focused much research on the topic,

especially the productivity of labor. Spatially, most of the

work has been done at the national or international level, for

example Dall’erba et al (2005) and the NBER Productivity,

Innovation and Entrepreneurship program. The Organization

for Economic Co-Operation and Development (OECD)

publishes rankings of countries based on labor productivity,

and it is clear that productivity varies dramatically across

countries of the world. There has been some work at the sub-

national level, most of it at the state level in the United States

such as Carlino and Voith (1992), Primont and Domazlicky

(2005), Smoluk and Andrews (2005), and Iranzo and Peri

(2006).

This paper focuses specifically on one state—the state of

Pennsylvania. And it goes below the state level to explore

productivity in the state’s metro areas as well. If productivity

is crucial to progress—and to competition, it is important for

us to know how our state and its metro areas perform.

This paper is intended to be descriptive, to present some

aspects of the state of productivity in the state, without

attempts to explain why it is what it is.

PRODUCTIVITY CONCEPTS

Measuring Productivity

Most fundamentally, productivity is “some measure of output

per some measure of input.” We choose to measure

productivity as value added per hour worked by production

workers in manufacturing industries. In their study of

regional comparative advantage, Hill and Brennan (2000) use

a similar measure. We opt not to use a measure of the value

of goods sold, such as value of shipments, since that would

involve double-counting of inputs. As the Census Bureau

says: “Data for cost of materials and value of shipments

include varying amounts of duplication, especially at higher

levels of aggregation. This is because the products of one

establishment may be the materials of another. The value

added statistics avoid this duplication and are, for most

purposes, the best measure for comparing the relative

economic importance of industries and geographic areas.”

(U.S. Census, 2007B)

An economy that produces steel sheets, steel fabrication

(turning the steel sheets into fenders), and automobiles would

have a total value of shipments that double-counts the steel

fabrication and triple-counts the steel itself. Thus the value

of shipments is inflated compared to the true value

produced—the final product, the car. Using value added at

each step in the production process will avoid this problem.

And we focus on the productivity of a single input, labor,

rather than other factors of production. One reason why this

is appropriate is that labor costs account for the lion’s share

of costs for most businesses. And labor is the source of most

income for most Americans. Nationally, employee

compensation accounted for 63.4% of national income in

2007, compared with 12.2% for corporate profits, 8.8% for

proprietors’ income, 5.9% for net interest income, and 1.2%

for rental income. (U.S. BEA, 2013) And employment is a

key focus of government policy, both at the national and the

local level. This is certainly not to say that the other factors

of production aren’t important; we just choose to focus on

this specific factor.

Proceedings of the Pennsylvania Economic Association 98

For this paper we measure productivity using hours of work

by production workers as the denominator. It would have

been possible to calculate productivity as output per worker

instead of per hour of labor, but that would be a less accurate

measure since not all workers work full time. To the extent

that an area’s industries tend to use more part time workers or

to use overtime labor, their measures of labor input (hours of

work) may not correlate closely with employment. For

example, if area A’s firms only hire workers who work 20

hours a week, while area B’s firms only hire workers who

work 40 hours a week, area A would need twice as many

workers to produce the same output as B, although they use

the same amount of labor input (hours of labor). Clearly, use

of output per worker data to measure productivity could be

misleading. For that reason, we elect to use output per hour

as our measure of productivity. Choice of the “per hour”

variable constrains us to use production workers rather than

total employment in this analysis, since hours of work are

only available for production workers.

In fact, measures of productivity per worker and productivity

per hour are highly correlated across metro areas. Kurre

(2004) found a correlation of 0.991 across 327 metro areas in

the 1997 Economic Census, and similar results obtained for

both the “hours” and the “workers” measures in that

regression analysis, suggesting that either approach is

acceptable. 2002 Economic Census data yielded a

correlation coefficient of 0.994 for value added per

production worker and value added per production worker

hour across 273 metro areas (Kurre and Miseta, 2008). And

data from 2007 for the manufacturing supersector (Brunot

and Kurre, 2012) yielded a correlation of 0.995 between

productivity per production worker hour and per production

worker across the 332 Metropolitan Statistical Areas (MSAs)

for which there are data for both. Further examination of the

2007 Economic Census data show that there is a correlation

of 0.973 between value added per worker (including all

employees) and value added per production worker across

those 332 MSAs. And at the state level (including the

District of Columbia as a state), the correlation is 0.960

between productivity per worker and productivity per

production worker, 0.942 between productivity per worker

and productivity per production worker hour, and 0.994

between productivity per production worker and productivity

per production worker hour.

Although this suggests that “employment” or “production

workers” may be an acceptable proxy in the denominator of

the productivity measure, we prefer to use “hours worked”

since that is more appropriate on a theoretical level.

The current paper focuses on productivity for the state of

Pennsylvania as well as its metro areas, so it is necessary to

have a data source that provides geographical data below the

national level. Hammill (2002) explored possible data

sources for metro-level productivity, concluding that the

Geographic Area Series of the Economic Census is the

preferred source. It provides data for all metro areas for a

single point in time from a single data source, permitting the

kind of cross-sectional study that we wish to do. We note

that the Economic Census also provides data at the state and

national levels, which is crucial for purposes of comparison.

Further, the Census Bureau has a well-earned reputation for

the quality of its data, and this is an important factor for any

database.

Of course, the Economic Census is not a perfect database.

We would prefer to have data for as recent a period as

possible, but the Economic Censuses are taken only every

five years. As of this writing, the most recent data are for

2007 and the manufacturing portion of the Geographic Area

Series were released in early 2010, a lag of over two years

from the data point. 2007 was six long years ago, and the

world is surely different now than it was in 2007. The Great

Recession has come and (mostly) gone since then, changing a

lot in our economies. Data for the 2012 Economic Census

are being collected currently, but it will be several years

before those data become available. So 2007 data are what

we have currently. But what the Economic Census lacks in

timeliness, it makes up for in data coverage and detail.

Specifically, it is one of the few sources with input and

output data by industry at the metropolitan level, and that’s

important if we want to know about the Pennsylvania

economy and its metro areas.

Productivity Data

We chose the Economic Census as the source for the key data

for this study since it gives value added and production

worker hours, for detailed manufacturing industries, by metro

area. But as with all spatial and industrial data, there are

tradeoffs between coverage and detail.

Geographically, we selected Metropolitan Statistical Areas

(MSAs) as one unit of analysis since metro areas are defined

to be small economies—actually, labor markets. Not all

MSAs are independent economies; there is undoubtedly some

interdependence among MSAs that are adjacent to other

MSAs--which is the reason for designation of Combined

Statistical Areas (CSAs), after all. However, MSAs are

generally more logical geographical units for economic

analysis than counties (one level down the geographical

spectrum) or states (one level up), both of which use

historical political boundaries that often do not reflect current

economic forces—although they are certainly relevant for

policy purposes. The Economic Census presents data at both

the state and MSA level, making it a logical choice as a data

source.

The 2007 Economic Census uses MSA definitions as of

December 2006 from the Office of Management and Budget

Proceedings of the Pennsylvania Economic Association 99

(Executive Office of the President, 2006). All the PA MSAs

were included in the 2007 Census of Manufactures, as well as

data for the state as a whole. Not all the state’s MSAs had

data for all industries; some industries simply do not exist in

some MSAs. And when an MSA had a small number of

firms in one industry, the data were not disclosed. That led

to varying numbers of observations for different industries.

While the cost of living clearly varies across metro areas, we

do not attempt to adjust the nominal productivity data with a

spatial price index. The argument can be made that

manufacturing firms sell primarily to a national or

international market and thus these firms must compete on

the basis of nominal price—regardless of the local costs of

local inputs. Of course, this implies that low cost-of-living

places may have an advantage as a location for national-

market goods that have low transportation costs.

WHAT THE DATA SAY

The State Level

Productivity, measured as value added per production worker

hour, averaged $126.12 in the U.S. manufacturing sector in

2007. But Table 1 shows that it varied widely across the

states, from more than double the national average in

Louisiana ($259.90) to just two-thirds of the average in

neighboring Arkansas ($84.71). The most productive state

had a value more than three times that of the least productive.

In their 1992 study at the state level Carlino and Voith noted

that the most productive state’s aggregate productivity was

2.3 times that of the least productive state’s, and the range

was only 1.2 for manufacturing. Although their method for

measuring productivity was quite different from that used in

this study, their productivity index varied significantly less

across states than the measure used in this study.

So how is Pennsylvania doing? Unfortunately, the state’s

productivity performance is a bit disappointing. Despite its

strong manufacturing tradition, Pennsylvania’s productivity

is $118.13, 6.3% below the national average, and ranking #27

of the 51 “states”. Of PA’s neighbors, Delaware, New

Jersey, Maryland, and New York all beat the national

average, while Ohio and West Virginia did worse than

Pennsylvania. While Delaware’s productivity is about 54%

higher than PA’s, West Virginia’s is about 10% less.

The Metro Level

Productivity varies even more widely at the metro level.

Table 2 shows productivity in the sixteen metro areas that lie

wholly or partially in Pennsylvania, as well as in selected

other MSAs. Perhaps the most notable fact from Table 2 is

the extremely wide range of metro productivity values, from

$637 an hour in Cheyenne WY to just $43 an hour in El

Centro CA. This 15-fold range across MSAs is much wider

than across the states.

Only four of PA’s metro areas surpass the national average

for productivity, with Philadelphia at the forefront beating the

national average by nearly 25%. Laggard Johnstown’s

productivity is only 51% of the national average, and only

41% of Philly’s. More detail about productivity at the metro

level will be presented in a later section. But first, it’s

necessary to discuss industry disaggregation.

The Industry Level

Of course, a key question is “why?” The answer must start

with the fact that “manufacturing” is not the same in all of

these areas. As mentioned above, not all industries exist in

all areas. “Manufacturing” is hardly a homogeneous industry.

In Erie it includes locomotives and plastics, but in Pittsburgh

it is robotics and medical devices. Of course, these

individual industries can all have very different levels of

productivity. This means we need to disaggregate the

manufacturing supersector to see what industries comprise

“manufacturing” in each area.

But does productivity vary much across industries? If not,

perhaps this is a moot question. Table 3 shows the range of

productivity across 3-digit NAICS industries for the U.S. and

Pennsylvania. In fact, the range across industries is even

greater that it is across metro areas—from a low of $48.86 in

apparel manufacturing (NAICS 315) to $828.16 in petroleum

and coal products (324), a 17-fold range. Clearly, the

industry mix of any economy is going to have a major impact

on its overall manufacturing productivity.

At the 3-digit level, Pennsylvania’s productivity patterns

follow those of the nation for the most part; their correlation

coefficient is 0.85. But that leaves room for some significant

differences, too. The range across 3-digit industries is from

$38.64 in leather products (316) to $409.66 in chemical

manufacturing (325), an 11-fold range—compared with the

17-fold range nationally.

On the high end, PA’s productivity exceeds national levels in

ten of the 21 3-digit industries. Relative to the national

industry levels, PA’s most productive industry is primary

metal manufacturing (NAICS 331), with a productivity that is

about 29% higher than the national level in that industry.

This is followed by food manufacturing (311), with a 25%

premium, and textile mills (313), with a 22% premium. But

the PA industry with the highest absolute level of

productivity, in dollar terms, is chemical products (325). At

$409.66 per hour, it is 3½ times the average PA productivity,

and is about 6% higher than the level of the industry in the

nation as a whole. Clearly a winner for the state!

Proceedings of the Pennsylvania Economic Association 100

On the other end of the scale relative to the U.S., PA’s

productivity in petroleum and coal products (324) is only

39% of the national level. But since this is the industry with

the highest productivity at the national level, PA’s value still

registers $324.46 per hour, about 2.7 times the state average,

despite the 61% reduction from this industry’s national

average. Relative to its competition in other states, it appears

that this industry doesn’t fare so well, but compared to other

industries in the state, it makes a major contribution to

incomes.

More of the story is revealed if we delve further into industry

disaggregation. The striking difference between U.S. and PA

productivity levels in petroleum/coal products (324) is

explained partially by digging down to the 5-digit NAICS

level. Table 4 shows the 5-digit components of that industry

and its sister industry, chemicals (325). It is clear that there

is some dramatic variation within these 3-digit industries. At

the national level, within NAICS 324 the productivity of the

petroleum refining industry (32411) is more than six times as

high as that of the paving and roofing materials industry

(32412), and more than three times as high as the remainder

of the 3-digit industry (32419). In productivity terms, these

industries are clearly quite different. Similarly in the

chemicals industry (325), petrochemical manufacturing

(32511) with productivity at a phenomenal $2,144 per hour is

3.7 times the next closest industry, pharmaceutical and

medicine manufacturing (32541) which itself has a notably

high productivity at over $582 per hour. Both of these are far

greater than the $97 per hour of the lowly synthetic fibers and

filaments industry (52522). Clearly, the 3-digit industries are

far from homogeneous when it comes to productivity.

All of this implies that industrial productivity analysis should

be done at the greatest level of industrial detail possible. But

there’s the rub. The greater the level of industry detail, the

more likely it is that the data will be unavailable due to the

legislation-imposed confidentiality responsibilities of the

Census Bureau. Table 5 shows the actual number of NAICS

industries in each level of industrial detail, as well as the

number of industries for which productivity data are

available at the national and state level. At the 3-digit level,

data are available for all 21 industries for both the nation and

PA. But once we get down to the 6-digit level, even at the

national level it is necessary to conceal data for a few

industries. And for the state, data are available for only 277

of the 472 officially defined 6-digit industries. It appears that

105 of the missing values are due to nondisclosure of the data

by the Census Bureau, and the other 90 are cases where the

industry simply doesn’t exist in the state. If we go to smaller

levels of geography, such as metro areas, the nondisclosure

problem becomes considerably more problematic, as we’ll

see below. At the micropolitan or county level, it is rare to

have good data for detailed industries. This clearly makes

the analysis more challenging.

Productivity by Metro Area, for Disaggregated Industries

As mentioned above, Pennsylvania’s metro areas are a

diverse lot, with quite different industry mixes. Table 6

shows productivity data for each MSA at the 3-digit level, as

well as for the manufacturing supersector. The number of

empty cells in this table clearly shows how difficult it is to

get comprehensive metro data that are industrially detailed,

even at the relatively aggregated 3-digit level. Of the 21 3-

digit industries, Lancaster had data for 12 industries and

Scranton for 11. All the other MSAs had data for

significantly fewer industries. And none of the PA MSAs

had data for four of the 21 3-digit industries.

At the supersector level, only four of the sixteen metro areas

had higher productivity than the U.S. average. Philadelphia’s

$157 an hour was 24% higher than the national average.

Allentown was 15% higher than the U.S. average, Pittsburgh

13%, and New York 12%. All of the other twelve had less

than average productivity, with Johnstown pulling up the rear

with productivity 49% less than the nation’s.

Given this, it is not surprising that most of PA’s metro areas

had lower productivity than the national average in each

industry, too. But in food manufacturing (211), five of the

MSAs surpassed the industry average. In fact Harrisburg’s

$198 an hour was 87% higher than the national average for

that industry. Another notable PA industry is machinery

manufacturing (333) where four of nine PA MSAs had higher

than average productivity, with Lebanon’s productivity 77%

higher than the nation’s. Lebanon also surpassed the national

average in fabricated metal products (332), where it had more

than double the national average productivity. And given its

primary industrial cluster, perhaps it is not surprising that

State College surpassed the national average in the printing

and related support activities (323) by nearly 90%. Erie,

unfortunately, did not surpass the national average in any of

the seven industries for which data are available.

Which Industries Are Pennsylvania’s Productivity Stars?

Table 7 shows industries that have productivity higher than

the state’s average of $118.13 per production worker hour in

manufacturing. The industry with the highest value is

biological product manufacturing (325414) at $786.64 per

hour. There were twenty establishments in this industry

employing 2,533 production workers in Pennsylvania in

2007, and they generated over $4.4 billion of value added.

Six other industries from the chemical products (325)

category are in the top eleven highest productivity six-digit

industries. So it’s no surprise that chemical products (325)

tops the three-digit list with a productivity of $409.66.

Chemical products industries also make up three of PA’s top

eight four-digits, and six of the top ten five-digits.

Proceedings of the Pennsylvania Economic Association 101

In 2007 the state had 624 establishments in the chemical

products (325) category, employing over 20,000 production

workers who produced over $17.3 billion of value added.

That made it the largest 3-digit industry in the state in terms

of value added, although food manufacturing (311) had over

2.5 times as many production workers and fabricated metals

(332) had more than 3.3 times as many.

Does PA Specialize in High Productivity Industries?

One question that arises from these data is whether the state

specializes in the manufacturing industries where it has high

productivity.

Table 8 shows productivity per production worker hour in 3-

digit manufacturing industries in PA and the U.S., and

relative to U.S. productivity in each industry. The final

column shows the manufacturing location quotient (LQ) for

each industry. This is calculated as the industry’s percent of

total manufacturing production worker employment in PA

compared to its percent in the U.S.:

LQi = % of PA’s mfg prdn workers in industry i . (1)

% of U.S.’s mfg prdn workers in industry i

An LQ greater than one implies that PA has a greater share of

its manufacturing activity in this industry than does the

nation, and is one of the state’s manufacturing specialties.

Eight of the 21 3-digit industries fall into this category, led

by primary metal manufacturing (331) which has nearly

double the national concentration.

The question above asks about “high productivity”, but this

can be interpreted in two ways. First, there is the absolute

sense; which industries have productivity values greater than

the PA average? Table 9 presents the same data as in the

previous table, but sorted by PA productivity, in dollars. Of

the seven industries with productivity above the PA average,

only three (43%) have LQs greater than one. Of the fourteen

industries with less than average productivity, five (36%)

have LQ’s greater than one. Alternatively, of the eight

industries for which PA has an LQ greater than one, only

three (38%) have productivity greater than the average. Of

the thirteen industries for which PA has an LQ less than one,

four (31%) have productivity greater than the average. These

numbers imply a slightly greater than average concentration

in higher productivity industries. In fact, the LQ for the

seven higher-than-average productivity industries taken

together is 1.05 (with 33.0% of manufacturing production

workers in PA, and 31.3% in the U.S. in these seven

industries.)

So the answer to the question posed above is “yes”, but only

a little bit.

A second way to answer this question would be to look at

industries in which PA’s productivity is higher than the

nation’s. Table 10 presents the same data as the previous two

tables, but this time ranked by PA’s productivity as a percent

of U.S. in each industry. Of the ten industries which had

higher productivity than the nation in their respective

industries, four of them (40%) had LQs greater than one. For

the eleven industries which had lower than national

productivity, we also find that four (36%) had LQs greater

than one. As with the absolute productivity measure, these

numbers imply a slightly greater than average concentration

in higher productivity industries. In fact, the LQ for the ten

higher-than-national-average productivity industries taken

together is 1.07 (with 47.2% of manufacturing production

workers in PA, and 44.3% in US in these ten industries.)

Again the answer to the question posed above is “yes”,

Pennsylvania does tend to specialize in its high relative

productivity industries, but only a bit.

CONCLUSIONS

This paper has intentionally been descriptive, and did not

seek to explain high or low productivity levels. It has

presented productivity data, in the form of value added per

production worker hour, for the state of Pennsylvania, both

for the manufacturing supersector and for various levels of

industry disaggregation. It also presented data for

Pennsylvania metro areas, both at the supersector level and

for 3-digit industries.

Some key findings are:

► Productivity varies dramatically across states and metro

areas of the nation.

► Productivity varies dramatically across industries within

manufacturing.

► Because of this, it is necessary to perform productivity

analysis at the greatest level of industry detail possible.

► Pennsylvania is “middle of the pack” compared to other

states in manufacturing productivity. Many of our near

neighbors are doing better, some significantly better.

► A handful of PA’s metro areas have higher productivity

than the national average, but ¾ of them lag the national

average, by significant amounts in some cases.

► PA and its metro areas are also home to some industries

with quite high levels of productivity, both in absolute and

relative terms.

► Pennsylvania tends to specialize in its high productivity

industries, but only to a small extent.

Proceedings of the Pennsylvania Economic Association 102

► Data problems become more severe the greater the level

of industrial detail, and the finer the level of geographical

detail.

Proceedings of the Pennsylvania Economic Association 103

Table 1

Manufacturing Productivity* by State, 2007

Table 2

Manufacturing Productivity* for PA and Selected Other Metro Areas, 2007

*Value added per production worker hour, in dollars.

Rank State Productivity % of U.S. Rank State Productivity % of U.S.

1 Louisiana $259.90 206.1 26 Illinois 118.44 93.9

2 Wyoming 210.60 167.0 27 Pennsylvania 118.13 93.7

3 Delaware 181.46 143.9 28 Ohio 115.45 91.5

4 Hawaii 172.31 136.6 29 Kentucky 113.58 90.1

5 Texas 171.65 136.1 30 Minnesota 112.92 89.5

6 New Mexico 158.77 125.9 31 Michigan 112.41 89.1

7 Oregon 154.11 122.2 32 Tennessee 112.10 88.9

8 Arizona 153.74 121.9 33 Oklahoma 110.35 87.5

9 Washington 153.14 121.4 34 Kansas 109.42 86.8

10 Connecticut 148.06 117.4 35 Wisconsin 107.80 85.5

11 Montana 146.44 116.1 36 North Dakota 107.00 84.8

12 Massachusetts 141.71 112.4 37 West Virginia 105.97 84.0

13 New Jersey 140.28 111.2 38 South Carolina 104.03 82.5

14 North Carolina 138.44 109.8 39 Maine 102.85 81.6

15 California 137.40 108.9 40 Georgia 101.11 80.2

16 Maryland 136.32 108.1 41 Alabama 100.84 80.0

17 Nevada 130.85 103.8 42 New Hampshire 95.18 75.5

18 Colorado 130.42 103.4 43 Idaho 94.45 74.9

19 Indiana 129.98 103.1 44 South Dakota 93.90 74.5

20 New York 127.69 101.2 45 Rhode Island 93.71 74.3

UNITED STATES 126.12 100.0 46 Alaska 92.93 73.7

21 Iowa 125.69 99.7 47 Mississippi 90.84 72.0

22 Utah 123.51 97.9 48 Vermont 90.07 71.4

23 Missouri 121.45 96.3 49 District of Columbia 86.76 68.8

24 Florida 120.99 95.9 50 Nebraska 85.09 67.5

25 Virginia 120.66 95.7 51 Arkansas 84.71 67.2

*Value added per production worker hour, in dollars.

Rank of

16 PA

MSAs

Rank of

332

MSAs

Metro Area Productivity % of U.S.

1 Cheyenne, WY $637.64 505.6

2 Lake Charles, LA 455.52 361.2

3 Alexandria, LA 426.35 338.1

1 57 Philadelphia-Camden-Wilmington, PA-NJ-DE-MD 157.41 124.8

2 75 Allentown-Bethlehem-Easton, PA-NJ 145.52 115.4

3 83 Pittsburgh, PA 142.18 112.7

4 85 New York-Northern New Jersey-Long Island, NY-NJ-PA 140.58 111.5

UNITED STATES 126.12 100.0

5 120 Scranton--Wilkes-Barre, PA 121.39 96.3

6 124 Williamsport, PA 119.79 95.0

7 128 Harrisburg-Carlisle, PA 118.96 94.3

8 129 Lancaster, PA 118.49 94.0

Pennsylvania 118.13 93.7

9 130 York-Hanover, PA 118.00 93.6

10 191 State College, PA 101.18 80.2

11 208 Altoona, PA 96.79 76.7

12 258 Youngstown-Warren-Boardman, OH-PA 86.49 68.6

13 264 Erie, PA 84.35 66.9

14 286 Lebanon, PA 78.36 62.1

15 308 Reading, PA 68.69 54.5

16 318 Johnstown, PA 64.26 50.9

330 Gadsden, AL 46.65 37.0

331 Kokomo, IN 46.48 36.9

332 El Centro, CA 42.58 33.8

Proceedings of the Pennsylvania Economic Association 104

Table 3

Manufacturing Productivity* for U.S. and PA, for 3-digit Manufacturing Industries, 2007

*Value added per production worker hour, in dollars.

Table 4

5-digit Detail for the Petroleum/Coal Products and Chemicals Industries

Table 5

Availability of Productivity Data by Level of Industry Detail

NAICS Industry US PA PA % of U.S.

31-33 Manufacturing $126.12 $118.13 93.7

311 Food manufacturing 105.60 131.45 124.5

312 Beverage and tobacco product manufacturing 458.20 279.08 60.9

313 Textile mills 60.86 73.95 121.5

314 Textile product mills 60.16 63.23 105.1

315 Apparel manufacturing 48.86 55.61 113.8

316 Leather and allied product manufacturing 53.91 38.64 71.7

321 Wood product manufacturing 49.18 48.16 97.9

322 Paper manufacturing 120.05 137.04 114.2

323 Printing and related support activities 71.94 72.16 100.3

324 Petroleum and coal products manufacturing 828.16 324.46 39.2

325 Chemical manufacturing 384.93 409.66 106.4

326 Plastics and rubber products manufacturing 73.75 73.99 100.3

327 Nonmetallic mineral product manufacturing 94.78 92.06 97.1

331 Primary metal manufacturing 120.40 154.71 128.5

332 Fabricated metal product manufacturing 77.07 75.85 98.4

333 Machinery manufacturing 110.04 103.81 94.3

334 Computer and electronic product manufacturing 230.30 149.76 65.0

335 Electrical equipment, appliance, and component manufacturing 104.23 88.21 84.6

336 Transportation equipment manufacturing 131.12 112.22 85.6

337 Furniture and related product manufacturing 60.60 69.09 114.0

339 Miscellaneous manufacturing 118.06 114.81 97.3

NAICS Industry U.S. PA PA % of U.S.

324 Petroleum and coal products mfg $828.16 $324.46 39.2

32411 Petroleum refineries 1,211.49 509.06 42.0

32412 Asphalt paving, roofing, and saturated materials mfg 190.69 136.30 71.5

32419 Other petroleum and coal products mfg 378.60 141.00 37.2

325 Chemical mfg 384.93 409.66 106.4

32511 Petrochemical mfg 2,144.42

32512 Industrial gas mfg 402.78 321.26 79.8

32513 Synthetic dye and pigment mfg 253.18 439.53 173.6

32518 Other basic inorganic chemical mfg 334.07 172.21 51.5

32519 Other basic organic chemical mfg 288.59 607.00 210.3

32521 Resin and synthetic rubber mfg 279.89 168.92 60.4

32522 Artificial and synthetic fibers and filaments mfg 96.60

32531 Fertilizer mfg 276.87 175.10 63.2

32532 Pesticide and other agricultural chemical mfg 522.81 165.37 31.6

32541 Pharmaceutical and medicine mfg 582.36 619.85 106.4

32551 Paint and coating mfg 237.33 292.40 123.2

32552 Adhesive mfg 162.03 179.01 110.5

32561 Soap and cleaning compound mfg 491.01 143.16 29.2

32562 Toilet preparation mfg 335.96 130.01 38.7

32591 Printing ink mfg 161.64

32592 Explosives mfg 109.81 227.01 206.7

32599 All other chemical product and preparation mfg 197.19

Empty cells mean either that industry does not exist in that area or the data are nondisclosed.

"2-digit" 3-digit 4-digit 5-digit 6-digit

Max number of NAICS industries: 1 21 86 184 472

United States 1 21 86 184 469

Pennsylvania 1 21 79 145 277

Nondisclosed 0 0 6 29 105

Nonexistent in PA 0 0 1 10 90

Proceedings of the Pennsylvania Economic Association 105

Table 6

Productivity in Pennsylvania’s Metro Areas

NAICS Industry

U.S

.

PA

All

en

tow

n

Alt

oo

na

Erie

Ha

rris

bu

rg

Jo

hn

sto

wn

La

nca

ster

Leb

an

on

New

Yo

rk

Ph

ila

delp

hia

Pit

tsb

urg

h

Rea

din

g

Scra

nto

n

Sta

te C

oll

eg

e

Wil

lia

msp

ort

Yo

rk

Yo

un

gst

ow

n

Nu

mb

er

of

MS

As

wit

h d

ata

Nu

mb

er

of

MS

As

gre

ate

r th

an

US

valu

e:

31-33 Manufacturing $126.12 $118.13 $145.52 $96.79 $84.35 $118.96 $64.26 $118.49 $78.36 $140.58 $157.41 $142.18 $68.69 $121.39 $101.18 $119.79 $118.00 $86.49 16 4

311 Food mfg 105.60 131.45 53.97 78.11 197.61 111.26 183.30 54.44 102.34 121.72 95.53 136.88 10 5

312 Beverage and tobacco product mfg 458.20 279.08 0 0

313 Textile mills 60.86 73.95 93.96 1 1

314 Textile product mills 60.16 63.23 0 0

315 Apparel mfg 48.86 55.61 0 0

316 Leather and allied product mfg 53.91 38.64 0 0

321 Wood product mfg 49.18 48.16 46.41 47.99 59.23 3 1

322 Paper mfg 120.05 137.04 65.30 74.19 2 0

323 Printing and related support activities 71.94 72.16 53.14 59.64 136.56 3 1

324 Petroleum and coal products mfg 828.16 324.46 0 0

325 Chemical mfg 384.93 409.66 304.69 1 0

326 Plastics and rubber products mfg 73.75 73.99 67.91 54.62 69.46 61.70 73.12 100.25 115.97 7 2

327 Nonmetallic mineral product mfg 94.78 92.06 129.10 138.50 83.26 70.82 4 2

331 Primary metal mfg 120.40 154.71 113.30 81.25 2 0

332 Fabricated metal product mfg 77.07 75.85 74.94 65.18 69.16 48.98 62.72 155.83 71.80 75.19 71.96 78.53 10 2

333 Machinery mfg 110.04 103.81 89.47 83.99 81.73 125.86 128.72 194.24 98.05 122.97 78.08 9 4

334 Computer and electronic product mfg 230.30 149.76 84.02 112.88 238.90 121.97 4 1

335 Elect eqpt, appliance, & component mfg 104.23 88.21 79.34 120.60 161.33 3 2

336 Transportation equipment mfg 131.12 112.22 85.83 48.02 44.17 88.22 4 0

337 Furniture and related product mfg 60.60 69.09 57.30 87.50 50.33 3 1

339 Miscellaneous mfg 118.06 114.81 52.77 120.39 118.89 3 2

Number of 3-digit industries with data 21 21 2 6 7 5 3 12 5 2 1 5 1 11 2 3 3 1

Number of 3-digits > US value 10 1 1 0 1 2 4 2 0 0 2 1 5 1 2 1 1

Boldface means a value greater than the U.S. value.

Proceedings of the Pennsylvania Economic Association 106

Table 7

Pennsylvania’s High Productivity Industries

NAICS Industry Productivity NAICS Industry Productivity

31-33 Manufacturing $118.13 SIX-DIGIT

THREE-DIGIT 325414 Biological product (except diagnostic) mfg $786.64

325 Chemical mfg 409.66 336391 Motor vehicle air-conditioning mfg 553.70

324 Petroleum & coal products mfg 324.46 325131 Inorganic dye & pigment mfg 551.48

312 Beverage & tobacco product mfg 279.08 325412 Pharmaceutical preparation mfg 526.27

331 Primary metal mfg 154.71 324110 Petroleum refineries 509.06

334 Computer & electronic product mfg 149.76 334510 Electromedical & electrotherapeutic apparatus mfg 495.74

322 Paper mfg 137.04 312111 Soft drink mfg 332.74

311 Food mfg 131.45 333295 Semiconductor machinery mfg 327.59

FOUR-DIGIT 325120 Industrial gas mfg 321.26

3254 Pharmaceutical & medicine mfg 619.85 325510 Paint & coating mfg 292.40

3251 Basic chemical mfg 445.28 325998 All other miscellaneous chemical product & preparation mfg 254.06

3241 Petroleum & coal products mfg 324.46 311613 Rendering & meat byproduct processing 249.98

3112 Grain & oilseed milling 317.13 311320 Chocolate & confectionery mfg from cacao beans 236.84

3221 Pulp, paper, & paperboard mills 314.51 311821 Cookie & cracker mfg 229.42

3121 Beverage mfg 279.35 334514 Totalizing fluid meter & counting device mfg 228.20

3122 Tobacco mfg 277.96 331111 Iron & steel mills 223.73

3255 Paint, coating, & adhesive mfg 252.36 327310 Cement mfg 219.98

3345 Navigational, measuring, electromedical, & control instruments mfg 232.50 337125 Household furniture (except wood & metal) mfg 215.92

3119 Other food mfg 230.58 334290 Other communications eqpt mfg 211.67

3259 Other chemical product & preparation mfg 229.11 312112 Bottled water mfg 211.21

3311 Iron & steel mills & ferroalloy mfg 222.64 334515 Instrument mfg for measuring & testing electricity & electrical systems 203.68

3342 Communications eqpt mfg 189.73 334517 Irradiation apparatus mfg 202.77

3113 Sugar & confectionery product mfg 188.49 334516 Analytical laboratory instrument mfg 202.18

3253 Pesticide, fertilizer, & other agricultural chemical mfg 173.14 311111 Dog & cat food mfg 197.03

3111 Animal food mfg 168.76 325132 Synthetic organic dye & pigment mfg 192.34

3252 Resin, synthetic rubber, & artificial synthetic fibers & filaments mfg 168.49 332913 Plumbing fixture fitting & trim mfg 191.86

3314 Nonferrous metal (except aluminum) production & processing 148.59 333921 Elevator & moving stairway mfg 190.06

3313 Alumina & aluminum production & processing 143.03 334220 Radio & TV broadcasting & wireless communications eqpt mfg 188.69

3333 Commercial & service industry machinery mfg 138.01 311822 Flour mixes & dough mfg from purchased flour 187.55

3274 Lime & gypsum product mfg 137.65 339114 Dental eqpt & supplies mfg 186.05

3256 Soap, cleaning compound, & toilet preparation mfg 136.70 327999 All other miscellaneous nonmetallic mineral product mfg 184.88

3391 Medical eqpt & supplies mfg 128.73 334210 Telephone apparatus mfg 183.28

3118 Bakeries & tortilla mfg

128.06

334513 Instruments & related prdcts mfg for measuring, displaying, &

controlling industrial process variables 181.55

3332 Industrial machinery mfg 126.78 325520 Adhesive mfg 179.01

3353 Electrical eqpt mfg 122.46 311330 Confectionery mfg from purchased chocolate 176.23

3312 Steel product mfg from purchased steel 119.81 311520 Ice cream & frozen dessert mfg 176.05

3339 Other general purpose machinery mfg 119.04 333516 Rolling mill machinery & eqpt mfg 174.04

FIVE-DIGIT 337214 Office furniture (except wood) mfg 173.46

32541 Pharmaceutical & medicine mfg 619.85 325314 Fertilizer (mixing only) mfg 172.59

32519 Other basic organic chemical mfg 607.00 325188 All other basic inorganic chemical mfg 172.21

32411 Petroleum refineries 509.06 333512 Machine tool (metal cutting types) mfg 172.19

32513 Synthetic dye & pigment mfg 439.53 331315 Aluminum sheet, plate, & foil mfg 166.31

32512 Industrial gas mfg 321.26 324122 Asphalt shingle & coating materials mfg 166.06

32551 Paint & coating mfg 292.40 325320 Pesticide & other agricultural chemical mfg 165.37

31211 Soft drink & ice mfg 287.39 339112 Surgical & medical instrument mfg 162.25

31132 Chocolate & confectionery mfg from cacao beans 236.84 324191 Petroleum lubricating oil & grease mfg 154.98

33451 Navigational, measuring, electromedical, & control instruments mfg 232.50 333319 Other commercial & service industry machinery mfg 153.68

32599 All other chemical product & preparation mfg 227.01 335314 Relay & industrial control mfg 152.72

33141 Nonferrous metal (except aluminum) smelting & refining 223.53 333618 Other engine eqpt mfg 150.41

33111 Iron & steel mills & ferroalloy mfg 222.64 334519 Other measuring & controlling device mfg 149.62

32731 Cement mfg 219.98 334417 Electronic connector mfg 149.18

33994 Office supplies (except paper) mfg 214.85 327420 Gypsum product mfg 146.96

33429 Other communications eqpt mfg 211.67 331210 Iron & steel pipe & tube mfg from purchased steel 146.44

33142 Copper rolling, drawing, extruding, & alloying 201.36 334512 Auto. environmental control mfg for resdntl, comrcl, & appliance use 144.38

33422 Radio & television broadcasting & wireless communications eqpt mfg 188.69 333414 Heating eqpt (except warm air furnaces) mfg 139.72

33391 Pump & compressor mfg 185.96 335313 Switchgear & switchboard apparatus mfg 137.09

33421 Telephone apparatus mfg 183.28 331314 Secondary smelting & alloying of aluminum 136.86

32552 Adhesive mfg 179.01 336413 Other aircraft parts & auxiliary eqpt mfg 136.42

31133 Confectionery mfg from purchased chocolate 176.23 333220 Plastics & rubber industry machinery mfg 136.16

31152 Ice cream & frozen dessert mfg 176.05 311119 Other animal food mfg 134.50

32531 Fertilizer mfg 175.10 331112 Electrometallurgical ferroalloy product mfg 133.76

32518 Other basic inorganic chemical mfg 172.21 311911 Roasted nuts & peanut butter mfg 133.75

32521 Resin & synthetic rubber mfg 168.92 333992 Welding & soldering eqpt mfg 133.39

31111 Animal food mfg 168.76 339920 Sporting & athletic goods mfg 133.22

32532 Pesticide & other agricultural chemical mfg 165.37 327410 Lime mfg 133.04

32742 Gypsum product mfg 146.96 326122 Plastics pipe & pipe fitting mfg 132.68

33121 Iron & steel pipe & tube mfg from purchased steel 146.44 333518 Other metalworking machinery mfg 132.49

32561 Soap & cleaning compound mfg 143.16 326130 Laminated plastics plate, sheet (except packaging), & shape mfg 132.13

33131 Alumina & aluminum production & processing 143.03 333131 Mining machinery & eqpt mfg 130.68

32419 Other petroleum & coal products mfg 141.00 325620 Toilet preparation mfg 130.01

33331 Commercial & service industry machinery mfg 138.01 311812 Commercial bakeries 129.96

32412 Asphalt paving, roofing, & saturated materials mfg 136.30 324199 All other petroleum & coal products mfg 129.23

33322 Plastics & rubber industry machinery mfg 136.16 339113 Surgical appliance & supplies mfg 128.02

33992 Sporting & athletic goods mfg 133.22 311340 Nonchocolate confectionery mfg 126.33

32741 Lime mfg 133.04 333298 All other industrial machinery mfg 124.53

32613 Laminated plastics plate, sheet (except packaging), & shape mfg 132.13 333111 Farm machinery & eqpt mfg 124.24

32562 Toilet preparation mfg

130.01

331491 Nonferrous metal (exc copper & aluminum) rolling, drawing, extruding,

& alloying 124.06

33911 Medical eqpt & supplies mfg 128.73 335932 Noncurrent-carrying wiring device mfg 123.91

33329 Other industrial machinery mfg 126.82 333294 Food product machinery mfg 122.10

31134 Nonchocolate confectionery mfg 126.33 333314 Optical instrument & lens mfg 121.65

33531 Electrical eqpt mfg 122.46 313230 Nonwoven fabric mills 121.43

33313 Mining & oil & gas field machinery mfg 121.45 335121 Residential electric lighting fixture mfg 120.54

31323 Nonwoven fabric mills 121.43 315232 Women's & girls' cut & sew blouse & shirt mfg 120.24

33639 Other motor vehicle parts mfg 118.44 322291 Sanitary paper product mfg 120.08

327993 Mineral wool mfg 119.20

324121 Asphalt paving mixture & block mfg 118.39

327125 Nonclay refractory mfg 118.20

Proceedings of the Pennsylvania Economic Association 107

Table 8

Productivity and Location Quotients in 3-digit Industries

Table 9

Productivity and Location Quotients in 3-digit Industries, Ranked by PA Productivity

PA U.S.

31-33 Manufacturing $118.13 $126.12 93.7

311 Food mfg 131.45 105.60 124.5 0.92

312 Beverage and tobacco product mfg 279.08 458.20 60.9 0.92

313 Textile mills 73.95 60.86 121.5 0.83

314 Textile product mills 63.23 60.16 105.1 0.77

315 Apparel mfg 55.61 48.86 113.8 0.85

316 Leather and allied product mfg 38.64 53.91 71.7 0.69

321 Wood product mfg 48.16 49.18 97.9 1.09

322 Paper mfg 137.04 120.05 114.2 1.19

323 Printing and related support activities 72.16 71.94 100.3 1.32

324 Petroleum and coal products mfg 324.46 828.16 39.2 1.33

325 Chemical mfg 409.66 384.93 106.4 0.90

326 Plastics and rubber products mfg 73.99 73.75 100.3 1.04

327 Nonmetallic mineral product mfg 92.06 94.78 97.1 0.99

331 Primary metal mfg 154.71 120.40 128.5 1.94

332 Fabricated metal product mfg 75.85 77.07 98.4 1.17

333 Machinery mfg 103.81 110.04 94.3 0.98

334 Computer and electronic product mfg 149.76 230.30 65.0 0.78

335 Electrical equipment, appliance, and component mfg 88.21 104.23 84.6 1.30

336 Transportation equipment mfg 112.22 131.12 85.6 0.59

337 Furniture and related product mfg 69.09 60.60 114.0 0.82

339 Miscellaneous mfg 114.81 118.06 97.3 0.96

Mfg LQNAICSProductivity*

PA % of US

*Value added per production worker hour, in dollars.

Industry

PA U.S.

31-33 Manufacturing $118.13 $126.12 93.7

325 Chemical mfg 409.66 384.93 106.4 0.90

324 Petroleum and coal products mfg 324.46 828.16 39.2 1.33

312 Beverage and tobacco product mfg 279.08 458.20 60.9 0.92

331 Primary metal mfg 154.71 120.40 128.5 1.94

334 Computer and electronic product mfg 149.76 230.30 65.0 0.78

322 Paper mfg 137.04 120.05 114.2 1.19

311 Food mfg 131.45 105.60 124.5 0.92

339 Miscellaneous mfg 114.81 118.06 97.3 0.96

336 Transportation equipment mfg 112.22 131.12 85.6 0.59

333 Machinery mfg 103.81 110.04 94.3 0.98

327 Nonmetallic mineral product mfg 92.06 94.78 97.1 0.99

335 Electrical equipment, appliance, and component mfg 88.21 104.23 84.6 1.30

332 Fabricated metal product mfg 75.85 77.07 98.4 1.17

326 Plastics and rubber products mfg 73.99 73.75 100.3 1.04

313 Textile mills 73.95 60.86 121.5 0.83

323 Printing and related support activities 72.16 71.94 100.3 1.32

337 Furniture and related product mfg 69.09 60.60 114.0 0.82

314 Textile product mills 63.23 60.16 105.1 0.77

315 Apparel mfg 55.61 48.86 113.8 0.85

321 Wood product mfg 48.16 49.18 97.9 1.09

316 Leather and allied product mfg 38.64 53.91 71.7 0.69

*Value added per production worker hour, in dollars.

NAICS IndustryProductivity*

PA % of US Mfg LQ

Proceedings of the Pennsylvania Economic Association 108

Table 10

Productivity and Location Quotients in 3-digit Industries, Ranked by PA Productivity as % of U.S.

1Thanks to Economic Research Institute of Erie Research Assistant Brittany Martinelli for able assistance in preparation of the

databases for this research.

PA U.S.

31-33 Manufacturing $118.13 $126.12 93.7

331 Primary metal mfg 154.71 120.40 128.5 1.94

311 Food mfg 131.45 105.60 124.5 0.92

313 Textile mills 73.95 60.86 121.5 0.83

322 Paper mfg 137.04 120.05 114.2 1.19

337 Furniture and related product mfg 69.09 60.60 114.0 0.82

315 Apparel mfg 55.61 48.86 113.8 0.85

325 Chemical mfg 409.66 384.93 106.4 0.90

314 Textile product mills 63.23 60.16 105.1 0.77

326 Plastics and rubber products mfg 73.99 73.75 100.3 1.04

323 Printing and related support activities 72.16 71.94 100.3 1.32

332 Fabricated metal product mfg 75.85 77.07 98.4 1.17

321 Wood product mfg 48.16 49.18 97.9 1.09

339 Miscellaneous mfg 114.81 118.06 97.3 0.96

327 Nonmetallic mineral product mfg 92.06 94.78 97.1 0.99

333 Machinery mfg 103.81 110.04 94.3 0.98

336 Transportation equipment mfg 112.22 131.12 85.6 0.59

335 Electrical equipment, appliance, and component mfg 88.21 104.23 84.6 1.30

316 Leather and allied product mfg 38.64 53.91 71.7 0.69

334 Computer and electronic product mfg 149.76 230.30 65.0 0.78

312 Beverage and tobacco product mfg 279.08 458.20 60.9 0.92

324 Petroleum and coal products mfg 324.46 828.16 39.2 1.33

Mfg LQ

*Value added per production worker hour, in dollars.

NAICS IndustryProductivity*

PA % of US

Proceedings of the Pennsylvania Economic Association 109

REFERENCES

Brunot, Justin A and James A. Kurre. 2012. Manufacturing

Productivity: How Much Does It Vary Across Metro Areas

and Why? Paper presented at the 66th Annual Conference of

the Association for University Business and Economic

Research, Honolulu HI, Oct 2012.

Carlino, Gerald A. and Richard Voith. 1992. Accounting for

Differences in Aggregate State Productivity. Regional

Science and Urban Economics, v. 22, #4, pp 597-617.

Dall’erba, Sandy, Yiannis Kamarianakis, Julie LeGallo, and

Maria Plotnikova. 2005. Regional Productivity Differentials

in Three New Member Countries. What Can We Learn from

the 1986 Enlargement to the South? The Review of Regional

Studies, v. 35, #1, pp. 97-116.

Executive Office of the President, Office of Management and

Budget. 2006. Update of Statistical Area Definitions and

Guidance on Their Uses OMB Bulletin 07-01. December

18, 2006.

Hammill, Michael L. 2002. Productivity Growth in Erie

Manufacturing Through Time. Erie PA: Economic Research

Institute of Erie, Penn State Erie. December 2002. Available

online at www.ERIEdata.org.

Hill, Edward W. and John F. Brennan. 2000. A

Methodology for Identifying the Drivers of Industrial

Clusters: The Foundation of Regional Competitive

Advantage. Economic Development Quarterly, v. 14, #1, pp.

65-96.

Iranzo, Susana and Giovanni Peri. 2006. Schooling

Externalities, Technology And Productivity: Theory And

Evidence From U.S. States. NBER Working Paper 12440,

August 2006.

Kurre, James A. 2004. Determinants of Productivity

Differences Across American Metro Areas. Presented at the

58th Annual AUBER Fall Conference, Tucson AZ , October

17-19, 2004.

Kurre, James A. and Edward R. Miseta. 2008 Determinants

of Productivity Differences across Metro Areas for

Manufacturing Industries. Presented at the 47th Annual

Meeting of the Southern Regional Science Association,

Arlington, VA, March 2008.

National Bureau of Economic Research. 2013. Productivity,

Innovation and Entrepreneurship Program. Online at:

http://www.nber.org/programs/pr/

Primont, Diane F. and Bruce R. Domazlicky. 2005. Which

Matters Most to the Estimation of Efficiency and

Productivity Growth in State Manufacturing: Method or

Measurement? Review of Regional Studies, v. 35, #2, pp.

117-138.

U.S. Bureau of Economic Analysis. 2013. NIPA Table

1.12. National Income by Type of Income.

U.S. Census Bureau. 2007A. 2007 Economic Census. Table

ECO731SG1: Manufacturing: Summary Series: General

Summary: Industry Statistics for Industry Groups and

Industries: 2007.

U.S. Census Bureau. 2007B. 2007 Economic Census

Methodology: Manufacturing (sector 31-33), page 6.

Available online at:

http://www.census.gov/econ/census07/pdf/meth/meth_31.pdf

.

Proceedings of the Pennsylvania Economic Association 110

FORCE-USING FIRMS IN THE COMPETITIVE EQUILIBRIUM

WHEN THE ONLY PUBLICLY PROVIDED GOOD IS ENVY GRATIFICATION

Johnnie B. Linn III

Division of Business

Concord University

Athens, WV, 24712

ABSTRACT

Envy is a public bad because if another’s possessions

engender envy in one person, they will engender envy in all

other persons similarly situated. Gratification of envy is

therefore a public good. Government maximizes solely this

public good subject to a budget constraint in extracting

capital from firms. In the equilibrium solution, the tax on

firm size is regressive. Firms have an incentive to become

larger. As a consequence, the tax base out of which envy is

gratified, and the amount of envy gratified, becomes larger.

INTRODUCTION

Envy is unhappiness that arises from the well-being of others.

It can augment a utility function otherwise based purely on

self-interest as an extra argument, denoting the well-being of

someone else, making a negative contribution to utility

(Zizzo, 2007). Envy thus implies more than the mere desire

to take from another, because it is gratified even if that which

is taken is not received by the person having the envious

desire. Envy thus has the attribute of a public bad, for if

particular assets exist that engender envy, they engender envy

in all persons that are envious. Gratification of envy, or

schadenfreude, is therefore a public good. A traditional

function of government is to provide public goods. This

paper explores a competitive equilibrium in the presence of a

government whose only function is the gratification of envy.

Frank (1997) coined the term positional externality to denote

the dependence of utility on relative consumption. Positional

externalities are analytically no different than environmental

pollution, and Frank proposes a tax on consumption

analogous to a tax on pollution (p. 1843). It is the relative

consumption, not consumption per se, that is the target of the

tax, so the tax would necessarily be progressive. Frank

further likens the optimization of consumption as having the

same public-good attribute as military preparedness:

Because each individual's consumption affects the

frame of reference within which others evaluate

their own consumption, this frame of reference

becomes, in effect, a public good. The

uncoordinated consumption decisions of individuals

are not more likely to result in the optimal level of

this public good than the uncoordinated actions of

individuals are likely to result in an optimal level of

military preparedness. The progressive consumption

tax is a simple policy measure that can help mould

the frame of reference in mutually beneficial ways

(p. 1844).

This paper takes as given that the function of government is

to forcibly provide public goods that the people want without

making a moral judgment as to whether what the people want

is morally appropriate, but also that those being taxed will

use force to ameliorate the government’s taxing power. The

resultant tax regime may therefore not be progressive.

Envy can take various forms in what is envied. The object of

envy can be all the possessions of all others, whether or not

the others are of higher or lower station than the envier, or all

the possessions of only those who are of higher station than

the envier, or only the possessions among those of higher

station than the envier that exceed the level of possessions of

the envier. The third form will be used in this paper to avoid

the complications of change in rank of individuals in regard

to their station.

THE MODEL

Labor is assumed to be homogenous in the population and

each individual is endowed with one unit of labor. Capital is

the object of envy and the envy engendered by each

individual is a function of that individual’s holding of capital.

Individuals may hire capital and guards and operate their own

firms, or may work as a guard, or may work as a tax

collector. Production requires only capital and one

entrepreneur, and force requires only labor. Following Linn

(2007), a ratio rule will be used for the allocation of winnings

by force.

We consider a horizon encompassing n individuals embedded

in a world market in which labor and capital can be

purchased at constant wages. Individuals are given a

designation j according to their rank in holdings of capital.

Envy extends only to the holdings of others who can be

“seen”, that is, those within the horizon. The amount of envy

engendered by individual j is

∑

∑

(1)

Proceedings of the Pennsylvania Economic Association 111

where the wage for capital is r. The amount of envy

engendered by a unit of capital is proportional to its value to

its soon-to-be dispossessed owner, or r. The jth

set of terms

in the first summation is individual j’s self-envy which will

be zero.

The Government’s Optimization Problem

The government’s objective function is to maximize the

value of a quantity of capital extracted from all individuals in

proportion to the envy engendered by them, less outlays for

tax collectors, thus:

∑ ( ∑

)

∑

(2)

where the government deploys Tj tax collectors paid a wage

w into arena j occupied by firm j and extracts a proportion tj

of the value of the firm’s capital. The taxes are weighted by

the amount of envy engendered in Equation (1) by each firm.

Taxes are levied only in arenas that have output. The other

arenas are empty, having been vacated by individuals who

have been hired as guards or as tax collectors. The

government expects the firms to anticipate its actions and

regards their variables as fixed. The government also regards

the number of individuals in the horizon to be unaffected by

the deployment of tax collectors.

Under a ratio rule, the government’s own-force elasticity for

its winnings is

(3)

The government’s force function in elasticized form for its

tax collectors is

(4)

The first order condition for the government in a given arena

is, employing the chain rule with Equations (3) and (4),

( ) ( ∑

)

(5)

If we sum the government’s first-order conditions over all n

arenas, we obtain

∑ ( ) ( ∑

)

∑

(6)

At equilibrium, setting government outlays for tax collectors

equal to the value of capital extracted, we obtain

∑ ( ) ( ∑

)

∑

(7)

or

∑ ( )( ∑

)

∑

(8)

For the government to balance its budget, it must break even

in at least one arena, so

|

(9)

The Firm’s Optimization Problem

The objective function for the firm in a given arena is

(10)

where Kj is the amount of capital hired and Gj is the number

of guards hired. In calculating the firm’s first-order

conditions, the firm expects the government to anticipate all

the firms’ actions, so the government’s variables, and the

amount of envy gratification to be generated for each

individual, are regarded as fixed. The firm’s production

function in elasticized form for its output is

(11)

The firm’s first-order condition for its capital is

( )

(12)

The force technologies for the firm and the government are

assumed to be the same, so the firm’s force elasticity with

respect to its guards is . Under the ratio rule, the firm’s

cross-force elasticity with respect to a tax rate is, when the

firm and the government are the only users of force,

Proceedings of the Pennsylvania Economic Association 112

(13)

The first-order condition for the firm’s guards is

( )

(14)

As in the competitive equilibrium in the absence of

government as set out in Linn (2007), we expect to find

stable competitive equilibria here only if is less than unity

and is greater than unity.

The Second-Order Conditions

The government’s matrix of second derivatives for its tax

collectors is a diagonal matrix each of whose diagonal

elements is the second derivative for a particular arena. The

second-order sufficiency condition is met if each of the

diagonal elements is negative.

The firm’s matrix of second derivatives for its capital and

guards is

(

( )

( )

)

where

(15)

The second-order sufficiency condition is met when the

diagonal elements are negative and the determinant

( )

|

( )

( )

|

(16)

is greater than zero. For r we may substitute in its value from

the first-order condition for Kj to get a further simplification

( )

( )

| ( ) ( )

( ) |

(17)

The second-order condition for the firm will not be met if the

tax rate is too small or or are excessively large.

We will return to the second-order conditions when we

consider simplified solutions in the neighborhood of the

long-run competitive equilibrium.

The Long-Run Competitive Equilibrium

In the long-run competitive equilibrium, the surviving firms

will earn a normal profit w*, the amount that its proprietor

could earn not only as a guard or a tax collector, but also the

amount of additional envy gratification that the former

capitalist would receive by entering the ranks of the non-

holders of capital. In the long-run competitive equilibrium,

labor is fully employed, so each person hired as a guard

supplies one unit of labor to one firm. After substituting for

the outlays to capital and guards in the profit function, we

obtain

( ) [ ( ) ( )]

(18)

The government must be breaking even in at least one arena.

For that arena, making use of Equation (9) and the guards’

first-order condition, we obtain

(19)

Let us suppose that in this arena, the number of guards and

number of tax collectors is the same. Since they have the

same force technology, the resultant tax rate would be 0.5.

Only for a value of unity for is such a solution possible.

For all values of greater than unity, only values of tj

greater than 0.5 are possible in arenas where the government

breaks even.

Progressivity or Regressivity of the Tax Regime

We now consider whether the tax regime across firms of

different sizes is progressive or regressive from the short-run

perspective of the firms. We take the derivative of equation

(18) with respect to tj and obtain

[ ( ) ]

( )

(20)

We set to zero and back-substitute for w

* from Equation

(18):

( ) ( )

( )

(21)

After cancellation of the expressions involving (1 - ), the

quantity can be factored out of the remaining terms

Proceedings of the Pennsylvania Economic Association 113

leaving a quadratic relation involving only tj. Its positive root

is

√

(22)

The value of Yj is maximized at this tax rate, so the firm size

and tax rate are directly correlated (progressive) as lower

levels of the tax and the firm size and tax rate are inversely

correlated (regressive) at higher levels of the tax.

For a tax rate other than 0.4142, there are two tax rates

compatible with a given size. The government will not allow

all firms to select a tax rate less than 0.4142, because for the

government to balance its budget in an arena the tax rate

must be at least 0.5. Firms that are sitting on the right hand

side of the firm size curve as a function of the tax rate, as

shown in Figure 1, will have an incentive to move up and left

on the curve, increasing their size and reducing their tax rate.

The government will lose surplus funds from the arenas of

such firms, so it will be forced to balance its budget in all

arenas, and since all firms employ identical factors, they will

all be the same size. The tax rate faced by all firms will be

regressive.

THE SOLUTION WHEN ALL FIRMS ARE ALIKE

Let the lowest-ranking, or only, firm in the horizon have the

designation m. The government’s objective function from

Equation (2) is

∑

(23)

Its first-order condition is

(24)

Setting the government’s revenue in arena m equal to its

outlays there,

(25)

The second derivative for the government in each arena is

identical to the firms’ second derivative for guards, as seen in

the lower right element of H in Equation (15), except that the

G’s are replaced with T’s. The second-order sufficiency

condition for the government is

(26)

or, from Equation (25),

(27)

which is met for all value of greater than unity.

The second order sufficiency condition for the firm will be

met if

|

|

(28)

In the long-run equilibrium, each firm defending against the

tax collectors will have one proprietor and an integral number

of guards, and there will be n – m + 1 such firms. Firms with

larger numbers of guards will be favored in the competitive

equilibrium because, from Equation (14), capital and guards

are positively correlated, and more capital means a lower tax

rate, since all firms are sitting on the right hand side of their

firm size curves. For a horizon of a given size, larger firms

means fewer firms. In the long run, there will be only one

firm in the horizon, and in Equation (28), the m’s will be

replaced with n’s. Figure 2 shows how the zone where

Equation (28) is satisfied varies with horizon size.

A greater horizon size selects for greater robustness of the

firm and a larger range of and in which the competitive

equilibrium is stable.

IMPLICATIONS

A stable competitive equilibrium where envy gratification is

maximized has been found. Contrary to what might be

wished for, the associated tax regime is regressive. The

absence of a progressive tax regime does not imply that

maximum envy gratification has not been achieved. This

model is built on the basest of human emotions, thus is likely

to be empirically robust. A government that wants to provide

more to the people than envy gratification can use this model

to improve upon.

Proceedings of the Pennsylvania Economic Association 114

Figure 1. Short-run firm size as a function of the tax rate on capital.

Figure 2. Zone of long-run stability as a function of horizon size.

REFERENCES

Frank, R.H. 1997. The Frame of Reference as a Public Good,

The Economic Journal, 107 (November), 1832-1847.

Linn, J. B. 2007. Force-Using Firms in the Competitive

Equilibrium when Public Goods are Absent, Pennsylvania

Economic Review, 15(1): 21-34.

Zizzo, D.J. 2007. The Cognitive and Behavioral Economics

of Envy, in Richard H. Smith (ed.), “Envy: Theory and

Practice”, Oxford University Press (Affective Science

Series), forthcoming.

0

0.4

0.8

1.2

1.6

2

0 0.2 0.4 0.6 0.8 1

Y

t

1

1.2

1.4

1.6

1.8

2

0 0.2 0.4 0.6 0.8 1

n 2

n 3

n

Proceedings of the Pennsylvania Economic Association 115

THE EFFECTS OF INCREASES IN CIGARETTE PRICES ON SMOKING BEHAVIORS:

ESTIMATES USING MSA AS A NATURAL EXPERIMENT

Zhen Ma

Department of Business

Misericordia University

301 Lake Street

Dallas, PA 18612

ABSTRACT

This paper examines the impact of the large increases in

cigarette prices after the Master Settlement Agreement

(MSA) on cigarette consumption for smokers by estimating

dynamic panel data models. I use system GMM estimator

with fixed effects to address the unobserved heterogeneity

and initial conditions issues. I find that older smokers are

virtually nonresponsive to price changes, and younger

smokers are more price sensitive; women smokers might be

more price responsive; and the overall estimated price

elasticity for the full sample of smokers is -0.26. In addition,

smoke-free air laws do not show significant effects.

I. INTRODUCTION

Since the release of the Surgeon General’s report on smoking

and health in 1964, the health hazard caused by smoking has

been more and more recognized. Smoking harms almost

every organ of the body; and it remains the single largest

preventable cause of death in the United States. Each year,

there are 393,600 estimated deaths caused by direct

smoking1. Secondhand smoking is no less harmful.

Secondhand smoke contains more toxic chemicals than does

smoke inhaled from a cigarette. Currently, there are 49,400

estimated deaths of nonsmokers annually from secondhand

smoking2. For half a century, the federal and state

governments have been engaged in anti-smoking campaigns.

Major policy instruments include widespread dissemination

of information on the health consequences of cigarette

smoking, ban of broadcast advertising of cigarettes,

restrictions on smoking in public places and private

workplaces, increased cigarette excise taxes and more

(Chaloupka and Grossman 1996). Over the past decade or

so, America has seen the historically largest increases in

cigarette excise taxes. Since January 1 1997, with the

exception of Florida, Mississippi, Missouri, North Dakota,

South Carolina and Virginia, all other states and the District

of Columbia have increased their taxes on cigarettes at least

once. The following are the number of states as of 2009 that

raised cigarette taxes by 25 cents or more per pack in

nominal terms from the previous year: 3 states in 1997 and

1998, 1 in 1999, 2 in 2000, 7 in 2002, 18 in 2003, 8 in 2004,

9 in 2005 and 6 in 2006 through 2009. Among these, 13

states had one-time increases of a dollar or more at least

once. As of 2009, Rhode Island has the highest state excise

tax of $3.46 per pack, followed by $2.75 in New York

(Orzechowski and Walker 2009). Also over the past decade,

an increasing number of states have passed stronger smoke-

free air laws. As of March 2011, 27 states have enacted

statewide bans on smoking in all enclosed public places,

including private workplaces, restaurants and bars, whereas

there was none prior to the year 2002.

In November 1998, 46 states and the four major tobacco

companies reached the Master Settlement Agreement which

stipulated that the tobacco companies pay the states $206

billion over the next 25 years to compensate most states for

Medicaid expenses for treating tobacco-related illnesses. The

other four states (Florida, Mississippi, Texas and Minnesota)

settled with the tobacco companies individually. The tobacco

companies are financing these payments by increasing

cigarette prices. As a result, the cigarette prices went up by

45 cents per pack3, or 19.5%, nationwide immediately after

the settlement, and continued to rise over the following five

years in many states due to substantial increases in their

excise taxes on cigarettes.

In this paper, I investigate how cigarette consumption among

smokers has been affected by these unprecedented increases

in cigarette prices and the changes in anti-smoking

regulations (the full prices of smoking)4. I estimate system

GMM estimator with fixed effects to address the unobserved

heterogeneity and initial conditions issues associated with

estimation of dynamic panel data models.

II. BACKGROUND AND LITERATURE

Raising cigarette taxes has been one of the most effective

means to prevent and reduce smoking. There is an extensive

literature documenting the impact of changes in cigarette

taxes or prices on the demand for cigarettes. Estimates of

price elasticity of demand for cigarettes vary from -0.05 to -

0.60.

Lewit and Coate (1982) is one of the earliest works that

examine the impact of cigarette excise taxes on reducing

smoking. The unique contribution of their study is that it

controls for possible cigarette bootlegging by eliminating the

individuals that reside in an area where the price of cigarettes

is higher than another price found within a 20-mile-wide

Proceedings of the Pennsylvania Economic Association 116

band around it. Using 1976 Health Interview Survey, they

estimate a price elasticity of -0.16 for adults, -0.15 for the 20-

25 age group, -0.03 for the 26-35 age group, and -0.30 for 36

years of age and older. For smokers, price elasticity is -0.10

for all adults, -0.20 for the 20-25 age group, -0.04 for the 26-

35 age group, and -0.15 for adults over 35 years of age.

Following the idea of Lewit and Coate (1982) on accounting

for cigarette bootlegging, Wasserman et al (1991) uses a

sample from the National Health Interview Survey (NHIS)

between 1970 and 1985. They dummy out the individuals

who boarder on an area that has lower-priced cigarettes,

rather than deleting them. Their estimates of price elasticities

of demand for cigarettes among adults range from 0.06 in

1970 to -0.23 in 1985, with the elasticities becoming

increasingly negative over time, but relatively lower than the

results from Lewit and Coate (1982).

Using repeated cross sections from Second National Health

and Nutrition Examination Survey (NHANES2) 1976-1980,

Chaloupka (1990) uses two staged least squares (2SLS) to

estimate demand equations derived from the Becker-Murphy

rational addiction model. He finds that, for men, the price

elasticity of demand for cigarettes centers on -0.60, while

women are virtually unresponsive to changes in cigarette

prices. Chaloupka (1991), using the same data, empirical

models, and estimation methods, finds that the price elasticity

of demand for the full sample ranges from -0.36 to -0.27.

The price elasticity of demand for cigarettes by only current

smokers ranges from -046 to -0.30.

Townsend et al. (1994) employ the British data General

Household Survey 1972-1990 and find that women are more

responsive to changes in cigarette prices. Their estimates of

price elasticity of demand for cigarettes for men and for

women are -0.08 and -0.23, respectively.

Hersch (2000), using the Current Population Survey (CPS)

Tobacco Use Supplements from September 1992, January

1993, and May 1993, finds that price elasticity of demand for

cigarettes to be -0.46 for men, and -0.38 for women.

Farrelly et al. (2001) estimate the two-part model with

National Health Interview Survey (NHIS) 1976-1993. They

find that the overall price elasticity for the full sample is -

0.15. They also estimate the price elasticities across gender

and age groups. They find that the demand elasticities for

cigarettes are -0.21 and -0.32 for men and for women,

respectively, -0.55 for ages 18-24, -0.53 for ages 25-39, and -

0.08 for ages 40 and older.

Stehr (2007) uses a sample of 1.3 million observations drawn

from the 1985-2000 Behavioral Risk Factor Surveillance

System (BRFSS), and finds that, by interacting dummy

variables for sex, race, ethnicity, age and income quartile

with cigarette taxes and state fixed effects, women are more

responsive to cigarette taxes than men, as opposed to what

previous studies have indicated. Specifically, the demand

elasticities are -0.09 for men and -0.12 for women.

All these studies use repeated cross-sectional data and fail to

control for reinforcement effects. Reinforcement of an

addictive good means that past consumption increases current

consumption by raising the marginal utility of current

consumption. Due to the addictive nature of cigarettes, past

consumption will reinforce its current consumption. This

paper advances the literature by using panel data to

consistently estimate the price elasticity of demand for

cigarettes allowing for reinforcement.

III. CONCEPTUAL FRAMEWORK AND EMPIRICAL

MODEL

The model starts with an individual’s utility in period t as a

function of the consumption of cigarettes )( tC and the

composite good )( tG given consumption of cigarettes from

last period )( 1tC . The individual maximizes utility:

)|,( 1 ttt CGCuU

subject to a budget constraint:

t

G

tt

C

tt GpCpY

where tY is income in period t , C

tp and G

tp are prices for

cigarettes and the composite good, respectively. Prices of

cigarettes are the full prices that include monetary prices and

policies that regulate the consumption of cigarettes.

Maximizing utility subject to the budget constraint yields the

demand function for the individual. The dependent variable

can be either amount of cigarettes consumed or a

dichotomous indicator for smoking. The following empirical

model for the probability of smoking for individual i in

period t will be estimated

titi

c

ti

c

tititi xpolpcc ,,3,2,11,1, (1)

where the error term ti , consists of the unobserved individual

fixed effects iu and the idiosyncratic disturbances tiv ,

( ),, tiiti vu . tic , is an indicator for smoking

participation or the amount of cigarettes consumed, 1, tic is

smoking status for the previous period, c

tip , is the money

prices for cigarettes, c

tipol , is the policy variable for

smoking, tix , is a vector of exogenous social-demographic

variables including family income, age, gender, race,

Proceedings of the Pennsylvania Economic Association 117

ethnicity, education, employment and marital status. 1

measures the reinforcement effect of past smoking behavior

on current smoking, and it is expected to be positive. 1 and

2 , the coefficients on the money price and policy variables,

should be negative based on the law of demand.

Estimation of equation (1) requires longitudinal data.

Difficulty arises in obtaining unbiased and consistent

estimates of the state dependence due to the correlation

between the time invariant unobserved individual

heterogeneity iu and the lagged smoking behaviors. Serial

correlation in the idiosyncratic disturbances tiv , also

complicates the estimation. Even when coefficients on

lagged dependent variables are not of direct interest,

estimating them consistently may be crucial for recovering

consistent estimates of other parameters (Bond 2002).

IV. DATA

The data are from the 1999, 2001, 2003, 2005, 2007 and

2009 waves of the Panel Study of Income Dynamics (PSID).

Conducted by the Survey Research Center, Institute for

Social Research, University of Michigan, the PSID is a

longitudinal study of a nationally representative sample of

U.S. individuals and the family units in which they reside.

The initial wave of the PSID was administered in 1968.

Follow-up interviews were conducted annually until 1996

and biennially thereafter. The health behavior measures such

as alcohol and cigarette consumption have been collected

since 1999.

4.1 Dependent Variables

The research objective is to find out whether smoking

behavior is significantly affected by changes in cigarette

prices. Two dependent variables are used: smoking

participation and the average number of cigarettes per day the

individual consumes. The values of the two dependent

variables are based on the answers from the respondents to

the questions “Do you smoke cigarettes?” and “On the

average, how many cigarettes per day do you usually

smoke?” Figure 1 shows the trends in cigarette consumption

among smokers.

4.2 Independent Variables

To control for the reinforcement effects of past smoking,

variables for lagged smoking participation and lagged

cigarettes consumption are included.

The PSID provides a wide range of demographic and

socioeconomic variables. The estimation of the model

controls for age, sex, race, ethnicity, educational attainment,

family income, household size, marital and employment

status. Three categories of race are used: white, black and

other races. White is the omitted group. Four categories of

educational attainment are created: college, some college,

high school and less than high school. College is defined as

16 or more years of schooling; some college is defined as 13

to 15 years of schooling; high school is 12 years of schooling

and less than high school is 11 or fewer years of schooling.

College is chosen to be the omitted group. Family income is

the total family income in thousands of dollars deflated to

1999 dollars from the previous year of interview. Household

size is the number of persons living in the same family unit.

For marital status, a dichotomous indicator for marriage is

constructed. Similarly for employment status, a dichotomous

indicator is used which is equal to one if the individual is

working now or only temporarily laid off or on

sick/maternity leave, and zero otherwise.

The key variables of the model are the costs of consuming

cigarettes. The costs consist of monetary costs and non-

monetary costs. Cigarette prices come from the state level

weighted average prices per pack in the Tax Burden on

Tobacco (Orzechowski and Walker 2009). All prices are

deflated to 1999 dollars. Non-monetary cost refers to

policies and regulations that increase the degree of

inconvenience for consuming cigarettes. The estimation uses

a smoking ban index that is constructed based on the

smoking restriction decisions of the smoke-free air laws in

the following 12 locations: Government worksites, private

worksites, child care centers, health care facilities,

restaurants, recreational facilities, cultural facilities, public

transit, shopping malls, public schools, private schools, and

free standing bars. Smoking restrictions for some locations

are coded 0, 1,2 and 3, with 0 being no restrictions against

smoking and 3 being smoking banned at all times; other

locations are coded from 1 to 5 in the similar fashion. The

smoking ban index for each state of each year is the sum of

the numerical codes. The price and policy variables are

merged with the PSID data based on the respondent’s state of

residence and the year of interview. The law of demand

predicts negative signs cigarette prices and the smoking ban

index. Table 1 shows the summary statistics of the

independent variables.

V. ESTIMATION

Equation (1) will be estimated as a linear dynamic panel data

model with fixed effects using the two-step system

Generalized Method of Moment (GMM).

Due to the addictive nature of cigarettes, 1, tic is positively

correlated with time invariant preferences for smoking, tiu , .

Therefore, treating all other variables as strictly exogenous,

Proceedings of the Pennsylvania Economic Association 118

OLS estimation of equation (1) will yield upward biased and

inconsistent estimates.

A system of the first differenced equation and the original

equation in levels are estimated

titi

c

ti

c

tititi vxpolpcc ,,3,2,11,1, (2)

titi

c

ti

c

tititi xpolpcc ,,3,2,11,1,

tic , is smoking participation or number of cigarettes

consumed; 1, tic is a predetermined variable: past

participation in smoking 1, tic is not correlated with current

shocks to smoking tiv , , but is affected by the past ones.

System GMM uses the lagged levels of the predetermined

variable as instruments for their first-differenced form and

uses the first differenced predetermined variable (now

exogenous to the fixed effects) to instrument them in levels.

Let tiX , be a vector of all exogenous variables in first

differences ),,,( ,,,, ti

c

ti

c

titi xpolpX . I use tiX ,

to

instrument tiX , in the difference equation and to

instrument themselves in the level equation. One period

lagged price and policy variables can serve as the additional

instruments for the predetermined variables5. Define

),( 1,1,,

c

ti

c

titi polpLP . Instrument matrices are given by6

6,64321

5,5321

4,421

3,31

000

000

000

000

iiiiii

iiiii

iiii

iii

di

LPXcccc

LPXccc

LPXcc

LPXc

Z (3)

6,64321

5,5321

4,421

3,31

000

000

000

000

iiiiii

iiiii

iiii

iii

li

LPXcccc

LPXccc

LPXcc

LPXc

Z (4)

li

di

siZ

ZZ

0

0 (5)

The moment conditions are7

0)( QZE s (6)

where

U

UQ ,

iT

i

i

i

v

v

v

U

4

3

and

iT

i

i

i

v

v

v

U

4

3

VI. RESULTS

This section reports the estimation results. Separate

regressions are done for the full sample, as well as for sub-

samples stratified by age group and sex8. Table 2 reports

OLS parameter estimates. Table 3 contains the GMM

estimates. The p-values associated with the AR (1) and AR

(2) tests are also reported Table 3. Both test the validity of

the instruments. To take into account of any time-specific

common trends, year fixed effects are included in all

regressions.

In Table 2, lagged cigarette consumptions are positive and

significant at the 1% level for all sample specifications.

Since this lagged term is positively correlated with the

unobserved individual fixed effects in the error terms, these

estimates might be upward biased and inconsistent. As such,

estimates of other key parameters might also be biased and

inconsistent. Indeed, the coefficients on cigarette prices have

unstable signs across columns, and are all insignificant.

In Table 3, all coefficients on the lagged cigarette

consumptions are lower in magnitude than those in Table 1.

This is in compliance with the expectation that OLS

overestimates the state dependence. Coefficients on cigarette

price are all negative, as expected, and are significant for

younger adults and females. Smoke-free air laws do not

show any significant effects.

Table 4 reports estimates of price elasticity of demand for

cigarettes. The elasticity is -0.43 for the age group of 50 and

younger, and -0.04 for 51 year-olds and older. For males and

females, their elasticities of demand are -0.27 and -0.49,

respectively. The price elasticity for the full sample of

smokers is -0.26.

The validity of the instruments depends on the assumption of

no autocorrelation (AR) in the error terms. Table 4 also

reports the P-values of the AR tests. The null hypothesis of

the AR test is no autocorrelation. Due to first-differencing in

the GMM estimator, there will be first-order autocorrelation

in the errors. Thus, the AR tests essentially test for second-

order autocorrelations. The P-values indicate that the sub-

sample of female smokers fails the autocorrelation test.

VII. CONCLUSIONS

The MSA has led to unprecedented increases in cigarette

prices in the U.S. since 1998. In this paper, I use data from

Proceedings of the Pennsylvania Economic Association 119

the PSID to examine how large increases in cigarette prices

affect cigarette consumption by smokers. I estimate the price

elasticity of demand for cigarettes for the full sample of

smokers, and for the subsamples stratified by age group and

sex. To my knowledge, all prior studies utilize repeated

cross-sectional data. I take advantage of the panel features of

the PSID, and apply the consistent system GMM estimator to

estimate a dynamic panel data model that accounts for the

reinforcement effects of past consumption on current

consumption. There are three major findings. (1) Older

smokers are virtually nonresponsive to price changes, and

younger smokers are more price sensitive. This may be

because younger smokers have been smoking for a shorter

time and are less addicted to smoking; they are less

financially stable and are affected more by higher prices. (2)

There is limited evidence in this paper that women smokers

might be more price responsive. However, the validity of

this claim is weakened by the failure in the autocorrelation

test. (3) The overall estimated price elasticity for the full

sample is -0.26, lower than the -0.40 to -0.60 range in the

literature. One explanation for this reduced price elasticity

could be that smokers who do not quit when facing

substantially higher prices are likely to be more addicted.

The regression analyses seem to indicate that raising price or

taxes is a more effective policy instrument for reducing

demand for cigarettes among smokers than smoke-free air

laws. This does not mean, however, that smoke-free air laws

are not effective in altering smoking behaviors. They appear

insignificant in the regressions because there is not much

variation in the smoke-free air laws within states during the

time of this analysis. These laws would induce occasional

smokers to quit, or deter nonsmokers from starting smoking

by increasing nonmonetary costs of smoking or by

influencing preferences for smoking.

A shortcoming of this research is that it does not take into

consideration of the effects of future prices on current

cigarette consumption. According to the rational addiction

framework developed by Becker and Murphy (1988),

consumers are forward-looking, and current consumption is

also affected by the anticipated future prices. However,

including a lead price will leave only three waves of data to

use. It would be more feasible to test rational addiction as

more data become available.

11

12

13

14

15

16

17

18

19

1999 2001 2003 2005 2007 2009

Fig.1: Trends in Cigarette Consumption

all age 50 & younger age 51 & older males females

Proceedings of the Pennsylvania Economic Association 120

Table 1. Independent Variables Descriptive Statistics

Variables Definition Mean

Standard

Deviation

Cigarette Price

Price of a pack of 20 cigarettes adjusted by consumer

price index to 1999 dollars

3.42 0.53

Smoking Ban

Index

Numerical values created from the smoking restriction of

the smoke-free air laws

17.87 13.66

Age Age of respondent

43.94 11.61

Male Dichotomous indicator for male

0.47 0.50

White Dichotomous indicator for white

0.61 0.49

Black Dichotomous indicator for black

0.30 0.46

Other Race Dichotomous indicator for race other than white or black

0.10 0.29

Hispanic Dichotomous indicator for Hispanic

0.05 0.21

College Dichotomous indicator for having college degree or higher

0.23 0.42

Some College Dichotomous indicator for having some college

0.14 0.35

High School

Dichotomous indicator for having high school diploma or

equivalent

0.35 0.48

Less than High

School

Dichotomous indicator for not having completed high

school

0.28 0.48

Family Income

Total family income of previous year in thousands of

dollars discounted 1999 dollars

68.24 53.69

Household

Size Number of persons residing in the household

2.92 1.51

Married Dichotomous indicator for being married

0.62 0.48

Employed Dichotomous indicator for employed 0.70 0.46

The maximum sample size is 8394. Some variables may have fewer observations due to non-

response.

Table 2: OLS Parameter Estimates for Smokers Only

Full

Aged 50 and

younger

Aged 51 and

older Males

Females

Independent Variable (1) (2) (3) (4)

(5)

Lagged Cigarette

Consumption 0.571***

0.558***

0.572***

0.592***

0.534***

(0.009)

(0.011)

(0.017)

(0.013)

(0.012)

Cigarette price -0.115

0.096

-0.516

-0.364

0.092

(0.195)

(0.229)

(0.001)

(0.330)

(0.228)

Smoking ban index -0.005

-0.009

-0.001

0.006

-0.015*

(0.039)

(0.009)

(0.014)

(0.013)

(0.009)

Number of observations 5684 3939 1745 2658 3026

All regressions include year fixed effects.

*significant at 10% level, **significant at 5% level, ***significant at 1% level

Proceedings of the Pennsylvania Economic Association 121

Table 3: GMM Parameter Estimates for Smokers Only

Full

Aged 50 and

younger

Aged 51

and older Males

Females

Independent Variable (1) (2) (3) (4)

(5)

Lagged Cigarette

Consumption 0.175***

0.153***

0.206***

0.081

0.356***

(0.045)

(0.045)

(0.080)

(0.056)

(0.063)

Cigarette price -1.059

-1.728**

-0.158

-1.252

-1.798*

(0.895)

(0.781)

(1.485)

(1.458)

(1.012)

Smoking ban index .0023

-0.032

0.024

0.023

-0.007

(0.022)

(0.027)

(0.005)

(0.039)

(0.0037)

Age

-0.317

-0.173

0.071

-0.730*

0.924***

(0.425)

(0.323)

(0.79)1

(0.438)

(0.281)

Male

6.381

2.403

-22.264

-

-

(11.010)

(8.980)

(31.779)

-

-

Black

-2.903

3.117

19.362

0.447

-0.146

(8.536)

(7.912)

(23.856)

(54.767)

(9.563)

Other race

5.99

1.681

11.681

-104.335

2.375

(8.692)

(7.541)

(27.340)

(125.755)

(6.667)

Hispanic

28.839

27.035

-15.157

-32.684

13.059

(22.41)

(20.967)

(55.462)

(235.539)

(19.747)

Some college -4.071

-3.505

12.530

9.555

12.304*

(8.33)

(6.308)

(17.453)

(18.834)

(6.49)

High school

-1.278

4.726

-45.671*

-15.656

-5.915

(6.018)

(4.753)

(23.365)

(31.434)

(4.814)

Less than high school 13.46**

4.587

11.211

6.197

0.074

(5.561)

(6.125)

(15.285)

(38.800)

(5.102)

Family income 0.009*

0.008

0.004

0.005

0.011

(0.005)

(0.008)

(0.005)

(0.011)

(0.008)

Household size 0.817

0.771*

-0.336

0.445

1.463

(0.499)

(0.428)

(1.415)

(0.801)

(0.737)

Married

-1.043

0.281

-0.439

0.358

-0.527

(1.615)

(1.553)

(7.930)

(2.540)

(2.010)

Employed

0.317

0.924

-1.307

1.273

-0.899

(0.614)

(0.661)

(1.316)

(0.906)

(0.809)

p-value AR(1) test 0.000

0.000

0.000

0.000

0.000

p-value AR(2) test 0.443

0.359

0.972

0.903

0.038

Number of observations 5685 3943 1742 2657 3028

Windmeijer finite-sample corrected standard errors are in parentheses.

All regressions include year fixed effects.

*significant at 10% level, **significant at 5% level, ***significant at 1% level

Table 4: Estimates of Price Elasticity of Demand for Cigarettes for Smokers Only

Full Aged 50 and younger

Aged 51 and

older Males Females

Price Elasticity of Demand -0.26 -0.43 -0.04 -0.27 -0.49

The elasticity is calculated at the mean of cigarette prices of 3.42.

Proceedings of the Pennsylvania Economic Association 122

ENDNOTES

1 Source: Center for Disease Control and Prevention (CDC).

“Tobacco Use: Targeting the nation’s leading killer.”

http://www.cdc.gov/chronicdisease/resources/publications/A

AG/osh.htm. Accessed on July 11, 2011.

2 Source: Facts about secondhand smoke.

http://www.co.dakota.mn.us/HealthFamily/HealthyLiving/S

moking/ShSfacts.htm. Accessed on July 15, 2011.

3 Adjusted to 1999 dollars.

4 Because non-smokers should not be affected by increases in

cigarette prices, they are excluded in this study.

5 These are used as IV style instruments in addition to the

GMM style instruments.

6

diZ is the instrument matrix for the differenced equation

for each individual; liZ is the instrument matrix for the level

equation for each individual; siZ is the instrument matrix for

the system of equations for each individual.

7

sZ ,iU and

iU are stacked matrices across individuals.

8 Balanced panels are used.

REFERENCES

Arellano, M., & Bond S. “Some tests of specification for

panel data: Monte Carlo evidence and an application to

employment equations.” The Review of Economic Studies,

April 1991, 58, pp. 277 – 297.

Arellano, M., & Bover, O. “Another look at the instrumental-

variable estimation of error- components models.” Journal of

Econometrics, 1995, 68, 29-52.

Becker, G., Grossman, M., & Murphy K. “An empirical

analysis of cigarette addiction.” The American Economic

Review, 1994, 84(3): 396–418.

Becker, G., & Murphy, K. “A theory of rational addiction.”

The Journal of Political Economy, 1998, 96(4): 675-700.

Blundell, R., & Bond, S. “Initial conditions and moment

restrictions in dynamic panel data models.” Journal of

Econometrics, 1998, 87: 115–143.

Bond, S. “Dynamic panel data models: a guide to micro data

methods and practice.” Institute for Fiscal Studies, WP09/02,

2002.

Bond, S., & Windmeijer, F. “Finite sample inference for

GMM estimators in linear panel data models.” Institute of

Fiscal Studies, WP02/x, 2002.

Cameron, C., & Trivedi, P. Microeconometrics: Methods and

Applications. New York, NY: Cambridge University Press,

2005

Chaloupka, F. “Men, Women, and Addition: the Case of

Cigarette Smoking.” NBER working paper #3267, 1990.

Chaloupka, F. “Rational Addictive Behavior and Cigarette

Smoking.” The Journal of Political Economy, 1991, 99(4):

722-742.

Chaloupka, F.J., & Grossman, M. “Price, tobacco control

policies and youth smoking.” National Bureau of Economic

Research, Inc, NBER Working Papers #5740, 1996.

Chaloupka, F.J., & Warner, K.E. “The Economics of

Smoking.” In The Handbook of Health Economics, ed.

Joseph P. Newhouse and Anthony J. Cuyler, 2000.

Chamberlain, G. “Panel data.” In Handbook of Econometrics,

Griliches Z, Intriligator M (eds), vol. 2. North Holland:

Amsterdam, 1984, 1247–1318.

Chay, Kenneth Y. and Dean Hyslop (1998), “Identification

and Estimation of Dynamic Binary Response Panel Data

Models: Empirical Evidence using Alternative Approaches.”

Center for Labor Economics, Working Paper No. 5, UC

Berkeley.

Farrelly, M., Brady, J., Pechacek. T., Woollery, T.,

“Response by Adults to Increases in Cigarette Prices by

Sociodemographic Characteristics.” Southern Economic

Journal, 2001, 68(1): 156-165.

Greene, W. Econometric Analysis (5th ed). Upper Saddle

River, NJ: Pearson Education, Inc. 2003.

Grossman, M., et al. “Policy watch: Alcohol and cigarette

taxes.” The Journal of Economic Perspective, autumn 1993,

Vol. 7, No. 4, pp. 211-222.

Hajivassiliou, V. “Smooth simulation estimation of panel

data LDV models.” Unpublished Manuscript, Yale

University, 1990.

Hayashi, F. Econometrics. Princeton, NJ: Princeton

University Press, 2000.

Heckman, J. 1981. “The incidental parameters problem and

the problem of initial conditions in estimating a discrete

time–discrete data stochastic process.” In Structural Analysis

of Discrete Data and Econometric Applications, ed. C. F.

Proceedings of the Pennsylvania Economic Association 123

Manski and D. L. McFadden, 178–195. Cambridge, MA:

MIT Press.

Hersch, J., “Gender, Income, Levels, and the Demand for

Cigarettes.” Journal of Risk and Uncertainty, 2000, 21(2/3):

263-282

Holtz-Eakin, D., Newey, W., & ROSEN, H. “Estimating

Vector Autoregressions with Panel Data.” Econometrica,

1988, 56, 1371-1395.

Hsiao, C. Analysis of Panel Data. New York, NY:

Cambridge University Press, 2003.

Keane M. “A computationally practical simulation estimator

for panel data.” Econometrica, 1994, 62, 95-116.

Kenkel, D., Mathios, A., & Pacula, R. “Economics of youth

drug use, addition and gateway effects.” Addiction, 2001, 96,

151-164.

Markowitz, S., & Tauras, J. “Substance use among

adolescent students with consideration of budget constraints.”

Review of Economics of the Household, 2009, 7:423–446.

Nickell, S. “Biases in dynamic models with fixed effects.”

Econometrica, 1981, 49(6): 1417-26.

Orzechowski, W. & Walker R. The tax burden on tobacco:

Historical compilation. Arlington, Virginia, 2009.

Pacula, R. “Economic modeling of the gateway effect.”

Health Economics, 1997, 6: 521–524.

Picone, G., Sloan, F., & Trogdon, J. G. “The effect of the

tobacco settlement and smoking bans on alcohol

consumption.” Health Economics, 2004, 13(10), 1063–1080.

Roodman, D. “How to do xtabond2: an introduction to

“Difference” and “System” GMM in Stata.” Center for

Global Development Working Paper Number 103, December

2006.

Stewart, M. “The Inter-related Dynamics of Unemployment

and Low-Wage Employment.” Journal of Applied

Econometrics, April 2007, Vol.22, No. 3, pp. 511-531.

Townsend, J., Roderick, P., Cooper, J., “Cigarette Smoking

by Socioeconomic Group, Sex, and Age: Effects of Price,

Income, and Health Publicity.” British Medical Journal,

1994, 309: 923-927.

Windmeijer, F. “A finite sample correction for the variance

of linear two-step GMM Estimators.” Journal of

Econometrics, 2005, 126: 25-51.

Wooldridge, J. Econometric Analysis of Cross Section and

Panel Data. MIT Press: Cambridge, MA, 2002.Wooldridge,

J. “Simple solutions to the initial conditions problem in

dynamic, nonlinear panel data models with unobserved

heterogeneity.” Journal of Applied Econometrics, 2005, 20:

39–54.

Proceedings of the Pennsylvania Economic Association 124

MORTGAGE CASH FLOW ANALYSIS AND PRICING USING CAS (COLLATERAL ANALYSIS SYSTEM)

Stephen M. Mansour and Riaz Hussain

The University of Scranton

Kania School of Management

Monroe and Linden Sts.

Scranton, PA 18510

ABSTRACT

The Collateral Analysis System (CAS) was developed as an

analytic database software product primarily serving the

mortgage industry. In this paper we will examine the cash

flow analysis and pricing tools. CAS uses an object-oriented

scripting language to generate cash flows which can be

discounted and priced accordingly. We will look at several

simple examples which can be built upon to create more

complex models.

HISTORY

During the Savings and Loan Crisis of the 1980’s mortgage

companies often acquired loans from several different

sources due to bankruptcies and reorganizations. The data

were stored in differing formats on computer tapes. Many

of these companies hired “tape-crackers” to extract and

reformat the data to company specifications. CAS was

originally developed to facilitate this process. Later on

database queries, reporting and analytical tools were

developed in CAS. The first commercial release of CAS

was in 1996. Currently there are over 40 companies using

the product.

CAS is involved in all facets of the mortgage industry,

including investment and commercial banking, origination,

servicing, finance, accounting and bond insurance. CAS is

used for tape-cracking and importing data, reporting, edit and

error analysis, exporting data, underwriting and automated

data scrubbing, inventory monitoring, secondary marketing,

shadow servicing, pricing, research and analysis, and cash

flow analysis. CAS is written in Dyalog APL to take

advantage of its array-processing capability, graphical user

interface, object-oriented model and quick development time.

CASH FLOW ANALYSIS

CAS generates cash flows from various user inputs including

balance, interest rate, term, prepayment rate, day-count

method, rate type (Fixed or Adjustable), amortization term

and/or payment. These inputs allow the CAS user to analyze

fixed-rate, adjustable-rate, balloon and interest-only

mortgages using various prepayment scenarios including the

CPR (Constant Prepayment Rate) and PSA (Public Securities

Administration) models. Default models including MDR

(Monthly Default Rate) and SDA (Standard Default

Assumption) can also be included; users can also assign loss

severity and recovery periods. Cash Flows generated

include Remaining Balance, Scheduled Principal, Actual

Interest, Voluntary Prepayments, Servicing Fees, Principal

Recoveries and Losses, New Defaults, Loans in Foreclosure

and Lost Interest. Cash flows can be generated for individual

mortgages or for a mortgage pool.

From previously generated cash flows, CAS will calculate

price, yield, modified and Macaulay duration average Life as

well as other industry-specific measurements. Various

prepayment and default assumptions can be set up

dynamically allowing the user to observe the price sensitivity

as well as the sensitivity of other measurements to changes in

the interest rate, prepayment rate, and default assumptions.

Error! Reference source not found. shows the basic logic

flow from mortgage-level data to pool sensitivity analysis.

GENERATING CASH FLOWS

To Generate Cash Flows using the object model, you must

first create a CashFlow object:

CF <- CashFlow.New ''

The left arrow (<-) evaluates the expression on its right and

assigns the result to the name on the left. Thus A<-2+2

assigns the value to the name A.

The Generate method takes two inputs, the coupon and the

balance. It assumes a term of 360 months (30 years). Thus

to amortize $100,000 AT 5.25% enter:

D <- CF.Generate 5.25 100000

The result of Generate is a multi-dimensional object, which

contains the cash flows. To make it readable we can convert

it to a table which is known as a RecordSet. The rows of the

table will be time periods, and the columns will be cash

flows.

R <- D.MakeRecordSet 'Time'

Proceedings of the Pennsylvania Economic Association 125

Since the result is a RecordSet, we can use the Show method

to display the cash flows

R.Show ''

The results are shown in Error! Reference source not

found.

The rows of the RecordSet are the 360 time periods, and the

columns are the following Cash Flows:

PERFBAL – Performing Balance (Balance at beginning of

Period)

ACTAM – Actual Amortization (Principal)

EXPINT – Expected Interest

If we wanted to use model a 15-year mortgage, we could

reset the Periods property to a different term by inserting the

following statement before the Generate method:

CF.Periods <- 180

The record set would now have only 180 records representing

time periods.

Prepayments and Defaults

Most mortgages do not pay as scheduled – Some property

owners pay off early due to sale of the property or

refinancing; others may default on the mortgage. To handle

these situations we introduce the SetPrepaymentRate and

SetRateOfDefault methods.

The simplest type of PrepaymentRate is the CPR – Constant

Prepayment Rate. We can generate CashFlows with an

annualized prepayment rate of 6% by inserting the following

statement immediately before the Generate method:

CF.SetPrepaymentRate 'CPR' 6

After running the script, the display will have a new column:

VOLPREPAY – Voluntary Prepayments

If we are dealing with non-Agency mortgages, we need to

concern ourselves with default rates. The simplest type of

default rate is CDR – Constant Default Rate. To model an

annualized default rate of 0.6% we insert the following

statement:

CF.SetRateOfDefault 'CDR' 0.6

Loss severity is the proportion of the balance which is lost to

the lender upon foreclosure. We will set the Loss Severity at

20%.

CF.LossSeverity <- 20

Now if we run the script, the display will have eight new

columns:

NEWDEF – New Defaults

ADB – Amortized Default Balance

FCL – Foreclosures

AMDEF – Amortization from Defaults

ACTINT – Actual Interest (This will differ from Expected

interest)

LOSTINT – Lost Interest (ACTINT – EXPINT This value

will be negative)

PRINLOST – Principal Lost (This value will be negative)

PRINRECOV – Recovered Principal

Variable rate (ARM)

Generating Cash Flows for an adjustable-rate mortgage

(ARM) requires that we set up the floating rates ahead of

time. The Floating Rates vary over time are depend on the

value of an underlying index. To do this we need to create a

TimeSeries Object containing the index.

IDX <- TimeSeries.New 'Index'

Since the index values are time-dependent we need to set the

Start Date for January 1, 2005:

IDX.StartDateTime <- 20050101 IDX.Periods

<- 360

To simulate rates between 4 and 6% set the Floor and Ceiling

properties:

IDX.Floor <- 4

IDX.Ceiling <- 6

Now generate “LIBOR” rates starting at 5.25%

IDX.Generate 'LIBOR' 5.25

Alternatively, we can simply copy in rates from a Data Table

using the Import method:

IDX.Import 'ARMINDEX'

Now we need to create a second TimeSeries object to reflect

the Floating Rates used in the calculation:

C<-TimeSeries.New 'FloatingRate'

Now set the contract provisions of the loan. The next reset

date is June 1, 2006, reset frequency is six months, reset cap

is 2% and margin is 3%:

Proceedings of the Pennsylvania Economic Association 126

C.ResetDate <- 20050601 C.ResetFrequency

<- 6

C.ResetCap <- 2

C.Margin <- 3

Next, we need to link the floating rate time series to the Index

time series we created previously.

C.Indextable <- IDX

C.Index <- 'LIBOR'

C.StartDate <- 20050101

C.Generate 'ARM1' 4.39

Next, create a Cash Flow object.

CF.CashFlow.New ''

Since the monthly payments will vary with an ARM, add a

new Cash Flow, PANDI

(Principal and Interest)

CF.AddFlow 'PANDI'

Instead of using a constant rate, substitute the floating-rate

time-series object for the coupon rate to generate cashflows

for a $100,000 mortgage:

D <- CF.Generate C 100000

The following two statements will allow us to view the cash

flows:

R <- D.MakeRecordSet 'TIME'

R.Show ''

The results are shown in Error! Reference source not

found.. Observe that the monthly payment (PANDI) changes

periodically in response to the changes in coupon payments

due to the variable rate.

Commercial Mortgages

Some Commercial mortgages have terms which are more

complex. These include interest-only mortgages whose cash

flows are similar to corporate bonds, and balloon mortgages

which amortize like traditional fixed-rate mortgages for a set

period, then require the balance to be paid off at a specific

time. Commercial mortgages often use different day count

methods instead of the standard 30/360 day count that is used

for residential mortgages. Hybrid mortgages combine

interest only, fixed-rate and balloon mortgages. A script to

generate a hybrid mortgage is displayed below, showing the

initial interest-only period of 3 months, the amortization

period of 30 years which determines the monthly payment

and the remaining term of 10 years after which the balloon

payment is due:

CF <- CashFlow.New ''

CF.Periods <- 360

CF.DayCount <- 'ACT/360'

CF.RemainingTerm <- 120

CF.StartDate <- 20050101

CF.IOPeriod <- 3

CF.Payment <-12000

D <- CF.Generate 5.25 2000000

R <- D.MakeRecordSet 'TIME'

R.Show ''

Aggregation of Cash Flows

If we want to generate cash flows for a pool of mortgages, we

need to link the Cash Flow object to a RecordSet, which

contains the individual loans. First we must create the

RecordSet: RS <- RecordSet.New ''

RS.GetCurrentFile ''

Error! Reference source not found. shows the first few

records of a RecordSet containing loan-level data

Now create a new cash flow object, using the record set as

the argument. CF <- CashFlow.New RS

Individual properties can be modified using the SetProp

method; the right argument has two elements; the name of

the property, and the name of a field in the record set:

CF.SetProp 'LoanID' 'LOANID'

CF.SetProp 'Coupon' 'RATE'

CF.SetProp 'Balance' 'CURBAL'

CF.SetProp'Periods' 'CALCRTM'

To generate total cash flows use field names for the coupon

and balance:

D<-CF.Generate'RATE' 'CURBAL'

To generate cash flows for individual loans you must include

the field containing the Loan ID. This is not recommended

for large RecordSets because of memory constraints.

D <- CF.Generate 'LOANID' 'RATE'

'CURBAL'

PRICING

In order to price individual loans, one must first create a

pricing object from the same record set used to create the

cash flows:

Proceedings of the Pennsylvania Economic Association 127

PX <- Pricing.New RS

The pricing object allows you to calculate the price from the

yield or the yield from the price. To calculate the price, you

first need to determine the yield. For example, to create a

field containing the yield assuming a base index of 4.25,

enter the following statements:

YLD<-RS.Evaluate'4.25+Margin'

YLD.Name <- 'YIELD'

YLD.Format <- 'F9.5'

RS.AddField YLD

Now set the yield and other properties for the pricing object:

PX.SetYield 'YIELD'

PX.SettlementDate <- 20050413

To calculate the Price, apply the Analyze method to the Cash

Flow object.

PD <- PX.Analyze CF

To display the pricing information, select the bond

measurements of interest:

PD <- PD.Select 'BONDMEAS =

PRICE,YIELD,AVGLIFE,MODDUR'

PR <- PD.MakeRecordset 'COLL'

PR.Show ''

To update the database with loan level pricing information:

PR.Key <- 'YIELD'

RS.Key <- 'LOANID'

RS.JoinInPlace PR

RS.SetCurrentFile ''

Price Sensitivity

In addition to loan-level pricing, CAS enables us to price

mortgage pools which are the basis for mortgage-backed

securities. We can analyze how prices of mortgage-backed

securities respond to changes in interest rates. Instead of

using static analysis, we observe that the prepayment rate

changes with respect to yield. Suppose the current yield is

5.25%. Let us assume that if the yield increases 25 basis

points, the prepayment rate will decrease from 6% CPR to

3% CPR, and if the yield decreases 25 basis points, the

prepayment rate will increase to 12%. The script to

calculate the price and duration is listed below: CF <- CashFlow.New ''

CF.SetPrepaymentRate 'CPR' (12 6 3)

DC<-CF.Generate 5.25

P<-Pricing.New ''

P.SettlmentDate<-20050101

P.SetYield 5 5.25 5.5

DP <- P.Analyze DC

R<- DP.MakeRecordSet 'PRE'

R<-R.Select '' 'PRICE,YIELD,MODDUR'

R.Show ''

CAS will produce the following output:

PRICE YIELD MODDUR

101.261 5.000 4.938

100.000 5.250 6.970

97.837 5.500 8.488

To find the effective duration, calculate

= 6.849 Note that the effective duration is smaller than the modified

duration of 6.970 because of the embedded call option. (See

Fabozzi et. al. 1994 pp. 170-171).

CONCLUSION

Thus, CAS can take loan level data, convert it into aggregate

cash flows for the pool with various prepayment and default

assumptions, and price the pool accordingly. In addition to

pricing, CAS can perform other calculations such as average

life and duration. If the price of a pool is known, then CAS

can figure out the yield.

The recent mortgage crisis has resulted in the sale of many

mortgage portfolios, often at a discount. The cash flow and

pricing tools in CAS have been used extensively by some of

our clients as an aid in pricing deals based on the valuation of

mortgage pools.

Proceedings of the Pennsylvania Economic Association 128

Figure 1: Pricing Object Model Flow Chart

Figure 2: The first few rows of the mortgage amortization table

TIME PERFBAL ACTAM EXPINT

20130523 100,000.00$ 114.70$ 437.50$

20130623 99,885.30$ 115.21$ 437.00$

20130723 99,770.09$ 115.71$ 436.49$

20130823 99,654.38$ 116.22$ 435.99$

20130923 99,538.17$ 116.72$ 435.48$

20131023 99,421.44$ 117.23$ 434.97$

20131123 99,304.21$ 117.75$ 434.46$

20131223 99,186.46$ 118.26$ 433.94$

20140123 99,068.20$ 118.78$ 433.42$

Proceedings of the Pennsylvania Economic Association 129

Figure 3 - Cash Flows for an ARM

TIME PERFBAL ACTAM EXPINT PANDI

20130523 100,000.00$ 134.34$ 365.83$ 500.17$

20130623 99,865.66$ 134.83$ 365.34$ 500.17$

20130723 99,730.83$ 135.32$ 364.85$ 500.17$

20130823 99,595.51$ 135.82$ 364.35$ 500.17$

20130923 99,459.70$ 136.31$ 363.86$ 500.17$

20131023 99,323.38$ 94.64$ 528.90$ 623.53$

20131123 99,228.75$ 95.14$ 528.39$ 623.53$

20131223 99,133.61$ 95.65$ 527.89$ 623.53$

20140123 99,037.96$ 96.16$ 527.38$ 623.53$

20140223 98,941.80$ 96.67$ 526.87$ 623.53$

20140323 98,845.13$ 97.18$ 526.35$ 623.53$

20140423 98,747.95$ 68.39$ 678.89$ 747.29$ Figure 4 - Record Set containing loan level data

LOANID NAME CITY STATE CURBAL RATE RTERM PANDI

10066 PAUL MANSOUR CHARLOTTESV VA 41764.73 7.000 333 500.00

10322 WILLIAM A BLOCK HADDON HEIG PA 86432.00 7.000 332 467.04

11098 ROBERT A. LOWRY SARASOTA FL 66469.19 6.875 333 446.71

12161 TYRONE R SMITH CHARLESTON SC 55716.77 6.875 334 374.45

13250 HENRY G PROSACK HARRISONBUR VA 75973.15 6.875 333 511.09

13268 ALBERT V. GIAMBALVOHARRISONBUR VA 76175.39 7.500 333 543.99

Figure 5 – Pricing Information at the Loan Level

COLL PRICE YIELD AVGLIFE MODDUR

10066 -$ 7.20 0 0.0000

10322 97.64$ 7.25 18.03953 9.1741

11098 96.92$ 7.20 18.10344 9.2350

12161 96.46$ 7.25 18.10344 9.2071

13250 96.47$ 7.25 18.03814 9.1906

13268 115.78$ 6.00 18.36883 9.9074

13854 96.47$ 7.25 18.03814 9.1906

13862 96.46$ 7.25 18.10344 9.2071

14191 96.47$ 7.25 18.03814 9.1906

REFERENCES

The Bond Market Association, 1990 Standard

Formulas for the Analysis of Mortgage-Backed

Securities and Other Related Securities

Fabozzi, 1997 Handbook of Fixed Income Securities,

Fifth Edition , JohnWiley & Sons

Fabozzi, 1994 The Handbook of Mortgage-Backed

Securities , Sixth Edition, McGraw-Hill

Fabozzi, Ramsey and Ramirez 1994 Collateralized

Mortgage Obligations, Structures and Analysis, 2nd

ed., FJF Associates

Mansour, P. and Mansour S., Using APL Expressions

in Database Operations, APL 1998 Conference

Proceeedngs, 22-26

Mansour, S., 2000 Houses, Windows and DOHRs

ACM SIGAPL APL Quote Quad, Vol. 30 Issue 4,

June 2000, 145-152

Proceedings of the Pennsylvania Economic Association 130

DISCUSSANT COMMENTS

THE GEOGRAPHIC CONCENTRATION OF ECONOMIC ACTIVITY

Stephen M. Mansour

The University of Scranton

Kania School of Management

Monroe and Linden Sts.

Scranton, PA 18510

The concepts in this paper are somewhat esoteric so it is

somewhat difficult to follow what the author is trying to

accomplish. There are some interesting historical data on

land use; mainly that manufacturing and industry has become

more dispersed and that farmland has become more

concentrated. However, the paper fails to discuss in detail

the consequences of these trends.

Thomas Hylton mentions in his book “Save Our Land, Save

Our Towns”, that since the 1950s, Pennsylvania has lost in

excess of 4 million acres of farmland. He claims this is due

mainly to misguided public policies that have encouraged

urban sprawl. The economic consequences of this include

excessive infrastructure costs for roads, bridges, water and

sewer lines to suburban and exurban communities. These

costs are often subsidized by the government. Social

consequences include include a loss of community and the

devastation of inner cities in the 1970’s.

Ironically, Detroit, which is in a manufacturing-dense county,

was destroyed by the very industry which made it an

industrial power. The automotive industry allowed people to

live farther from the industrial centers where they worked, so

the middle class taxpayers were able to move out.

Another issue is the author’s use of the Gini coefficient

which measures inequality. This can be misleading because

it looks at relative wealth rather than absolute wealth (or

concentration in this case). To illustrate this point, the late

Prime Minister Margaret Thatcher responded to Parliament

member Simon Hughes’ question in 1990 about the wealth

gap in Britain:

“What the honorable member is saying is that he would

rather that the poor were poorer, provided that the rich

were less rich. So long as the gap is smaller, they would

rather have the poor poorer. You do not create wealth

and opportunity that way. You do not create a property-

owning democracy that way.”

Is increased concentration of agriculture good or bad? The

“locavore” movement encourages the sale and consumption

of food produced within a 100 mile radius. Farmer’s markets

are a primary example of this. This would suggest that the

concentration of agriculture is not a good thing. On the

other hand more efficient farm yields per acre allow us to use

land more efficiently and for purposes other than farming.

Proceedings of the Pennsylvania Economic Association 131

CREDENCE GOODS AND STATE MANDATED VEHICLE SAFETY INSPECTIONS:

HOW NON-PROFIT INSPECTION SERVICES CAN CORRECT FOR MARKET FAILURE

John McCollough,

Department of Economics

Lamar University

Beaumont, TX 77710

ABSTRACT

The problems associated with asymmetric information and

credence goods are a common worry to consumers who

require the services of technicians with expert knowledge. A

study was designed to see if the concerns of consumers were

justified. This paper reports the empirical results of this

study. Specifically, the test looks at two different samples of

vehicle owners and the repair costs associated with a vehicle

state safety inspection. In one sample the vehicles were

inspected by a ‘non profit’, state affiliated inspection station

while in the second sample the vehicles were inspected by a

‘for profit’ vehicle inspection station. The results suggest that

those vehicles inspected at a ‘for profit’ inspection station

had higher repair costs than those vehicles inspected by a

‘non profit' vehicle’ station.

INTRODUCTION

The issue of credence goods is a special case of asymmetric

information and as such, it can lead to market failure. More

specifically, credence goods deal with service goods

provided by a technician with expert knowledge and,

furthermore, the consumer’s knowledge is much less than the

technician’s (Rasch and Waibel, 2012). There are many

examples of this and it is quite common. Because of this

asymmetric knowledge between the technician and the

consumer, there can be an incentive for the service provider

to exploit the consumer, and as a result of this exploitation a

market failure can arise. With respect to vehicle repairs,

Schneider, (2012) estimates the welfare loss of this market

failure at $8.2 billion. It could be that a majority of

technicians within any one particular trade are honest, but if

perceptions spread among consumers that this trade group

has a proclivity toward exploitation then less service will be

demanded by consumers than is socially optimal (Dulleck

and Kerschbamer, 2006). As a result, this market failure may

even result in environmental damage as consumers decide to

forgo the chance of being exploited and dispose of a product

that could have been repaired for further reuse (McCollough,

2010).

There are many common examples of credence goods. As

suggested above, one example would be services provided by

an auto mechanic. Indeed, vehicle repairs rank first in

customer complaints. However, there are many other

examples which range from services provided by your local

roofer or plumber or even services provided by the medical

profession. Because of the consumer’s reliance on the

mechanic’s expert knowledge, the consumers are at an

information disadvantage and can, therefore, be easily

exploited. Exploitation can take the form of overcharging for

service, providing more service than is needed, or perhaps

even charging for services that never took place (Darby and

Karni, 1973,Webbink,1978).

The objective of this paper is to find evidence of this type of

exploitation and to find out if the consumers’ fears and

suspicions are justified. This paper also attempts to quantify

the exploitation. In addition, this paper will also show how

state-affiliated agencies, acting in the role of safety inspectors

without profit motives can play an important role in

correcting the market failure associated with credence goods.

The hypothesis set out in this paper is that repair costs

associated with vehicle safety inspections provided by quasi-

governmental agencies will be statistically less than repair

costs associated with inspections provided by privately

owned, ‘for-profit’, service stations. The reason for this is

that for- profit service stations have an incentive to provide

more service than is required. For example, a ‘for profit’

service station might require brake work or perhaps a tire

replacement when, in fact, these services are not really

needed in order for the vehicle to pass inspection. Worse yet,

the ‘for-profit’ service station might require certain repair

work before the vehicle can pass the safety inspection, but

then never provide the service, charging the customer for

work that never took place. On average, any difference in

repair costs should represent the cost of the market failure.

A test was designed which compares the repair costs

associated with vehicle safety inspection for residents from

the state of Pennsylvania and for residents from New Jersey.

In Pennsylvania, vehicle owners must have their vehicle

inspected by a ‘for profit’ service station, while in New

Jersey the residents can choose to have their vehicle

inspected by either a ‘for profit’ service station or a ‘not for

profit’ vehicle inspection station. When New Jersey

residents have their vehicle inspected by a ‘non profit’, state-

affiliated, vehicle inspection station, the vehicle is actually

inspected by a private firm that has been contracted out to

Proceedings of the Pennsylvania Economic Association 132

perform all state safety vehicle inspections. Safety inspectors

do not work as state employees, rather they work for the firm

which provides the inspections. Neither the firm nor the

inspectors have a profit motive. These inspectors tell the

vehicle owner what needs to be fixed before the vehicle can

pass inspection. The inspection stations are prohibited from

performing any repairs. The vehicle owner will then fix the

problem at a service station of his or her choosing and then

come back to the state-affiliated inspection station for an

inspection sticker as proof that the vehicle passed inspection.

LITERATURE REVIEW

Due to the nature of credence goods, and the fact that the

expert knowledge required to perform the service is

asymmetric, the services provided can often times be price

insensitive with low price elasticity’s (Peppers and Rogers,

2006). The lower the price elasticity for the service, the

easier it is for disreputable service providers to take

advantage of the consumer. The literature typically cites lack

of competition as the cause for price insensitivity.

Geographic locations are a prime determinant in how

competitive vehicle repairs and vehicle inspection services

are. Typically, the denser the geographic location, the more

competition there is and, hence, the ‘switching costs’ are low.

In other words if it is easy for consumers to find other service

providers then this makes it more difficult for service

providers to overcharge customers.

Rasch and Waibel (2012) state that overcharging for vehicle

repairs occur more frequently in less densely populated, non-

competitive locations. They find that non-competitive, low

density locations just off the interstate overcharge since there

is less chance of repeat business. Customers at these

locations are mainly one time customers just passing through.

They conclude that in more dense geographic locations

where competition is higher, service providers are dependent

on repeat business.

As service providers seek repeat customers, protection of

their reputation can “discipline” service providers, especially

when there is a possibility of repeat business by customers

(Schneider, 2012). However, Hubbard (2002, pg 466) warned

that “Incentives are weaker when consumers are naive about

sellers’ private objectives, believe that sellers are

homogeneous, or when switching costs are high.” As an

example of this, Hubbard (1998) finds that independently

owned service stations are more likely to pass vehicles for

inspection than chain store service stations, new car

dealerships, and tune-up shops, because the latter work on

commission whereas the independent shop is motivated by

repeat business. He also finds that the more inspectors there

are at a service station, the more likely a vehicle is to fail. In

addition, Schneider (2012) states that higher quality service

can be provided by those technicians looking for repeat

business, but the prospect for repeat business must be likely.

In a follow-up study, Hubbard (2002) finds that the

reputation effect does pay off. More specifically, he finds that

consumers are 30% more likely to utilize a service station in

the future if that service station had passed the vehicle for

inspection in the recent past. Biehal (1983) also finds that

consumers make choices with respect to auto repair services

based on previous experiences with repair facilities.

However, with respect to annual state vehicle safety

inspections, the desire for repeat business can actually create

a moral hazard problem. For example, Hubbard (1998) found

that in California private inspection facilities pass vehicles at

twice the rate of state inspection facilities, except in cases

when the emission repairs are covered under a warranty for

late model, low mileage vehicles that are being inspected at

new car dealerships. Interestingly, Hubbard also found that

the inspection failure rate was even slightly lower when the

service provider was located in a more competitive location,

and this he attributes to ‘low switching costs’ and the ease

with which to obtain a second opinion.

Schneider (2012) finds that initial diagnostic fees are lower

for possible repeat customers. This suggests that when

reputation was important, the service provider charged a

lower up front diagnostic fee but Schneider (2012) found no

difference in repair recommendations, repair prices or the

number of legitimate repairs when the service mechanic was

trying to protect his reputation. Schneider concludes that the

ability of consumers to discipline service providers with the

possibility of repeat business is ‘fruitless’.

So, how can consumers protect themselves from

unscrupulous service providers? One way is to obtain a

second opinion. However, second opinions are usually

expensive with respect to either money or time (perhaps

both) for the consumer and the service provider, particularly

when it is cheaper to provide the diagnosis and the repair

service together as opposed to the repair service and

diagnosis taking place separately (Emmons, 1997). In

addition, it is unclear to the consumer if a proper diagnosis

was even performed. Pesendorfer and Wolinsky (2003)

suggest that in competitive markets with competitive prices

efforts by the service provider to provide proper diagnosis

might be sub-optimal.

Therefore, barring a second opinion it is difficult for

consumers to determine if they actually were taken advantage

of because who, other than the service provider, can really

judge if service was required or not. With respect to vehicle

repairs mandated by an annual vehicle inspection, second

opinions can be costly to the consumer, particularly with

respect to time. Typically a consumer must leave their

vehicle for half a day or more with the service provider

giving the second opinion. Typically, this means alternative

transportation must be arranged. Customers then find

themselves in a dilemma. If the vehicle inspection station

Proceedings of the Pennsylvania Economic Association 133

does repair work themselves, which is usually the case in

Pennsylvania, then the customer must decide to either go

ahead or trust the inspector to do the repairs while the vehicle

is still queued up. Or, does the customer take the vehicle in

for a second opinion, requiring additional time and expense

on the part of the customer.

Another common strategy is that the consumers can ask for

the old part back when a part was replaced. This helps to

prevent fraudulent billing and overcharging for work that was

not performed. But it still does not prevent ‘over-treatment’,

which is providing more repairs than necessary (Dulleck and

Kerschbamer, 2006)

States can help to protect consumers from fraudulent repairs

by requiring licensing or certification of service providers.

Unfortunately, this can increase barriers to competition

which reduces competition when more competition should be

encouraged. Often time consumer publications and rating

agencies such as AAA membership clubs or Angie’s list

allow consumers to undertake information searches to help

sort out reliable service providers from the unreliable ones.

However, Biehal (1983) suggests that consumers are not as

proactive in their information search as they need to be. For

example, from a survey of customers who recently had their

vehicles repaired Biehal found that 31.7% of the respondents

felt as if their bills were unreasonable and one-fourth were

dissatisfied with their service. But, the level of customer

dissatisfaction decreased as the amount of customer external

information for auto repair services increased.

Whether or not a repair will fall into the hands of a reputable

or disreputable mechanic, the consumer needs to weigh the

expected benefits against the cost of a repair. The expected

benefits of a repair includes both an expectation that a repair

will be completed correctly as well as an expectation that the

product will have its useful life extended. The more trust a

consumer has in a mechanic (either because the mechanic is

certified, or has an excellent “word of mouth” reputation, or

has been endorsed by a rating agency) the higher the

expectation that a repair will be made correctly. If this

expectation is low then it is more likely that the consumer

will forgo the repair in favor of choosing to replace the

product (McCollough, 2010, pg 189)

EMPIRICAL TEST

An empirical test is designed to see if the repair bills

associated with vehicle state inspections are statistically

different for vehicle owners who go to a ‘not-profit’, state-

affiliated inspection station as opposed to those who go to a

‘for-profit’, private inspection station. Obviously, if the

repair bills at the private inspection stations are statistically

higher than this would suggest evidence of market failures

from an asymmetric information and credence goods point of

view. In other words, according to the literature, technicians

with asymmetric and expert information regarding the repair

and maintenance of a product are thought to have an

incentive to cheat the customer and are in fact doing so. This

then would suggest that there is role for government with

respect to vehicle inspection services because the government

would be able to cut back on unnecessary repair bills by

providing an initial diagnostic and inspection service as in the

case of New Jersey’s vehicle safety inspection program. On

the other hand, if repair bills associated with vehicle

inspections from a private vehicle inspection station are not

statistically different from the state-affiliated vehicle

inspection station, then this might suggest that there is no

need for governments to be involved in the vehicle inspection

business. Finally, if the repair charges associated with vehicle

inspections are statistically less at private vehicle inspection

stations than for the state-affiliated vehicle inspection

stations, then one might conclude that the reputation effect is

at work and that private inspection stations are working hard

to keep customers satisfied. However, it is very troubling to

think that private inspection stations could possibly be

overlooking necessary and important repairs at inspection

time because they are afraid of losing potential long term

clients. On the other hand, it is just as unsettling to think that

the state-affiliated inspectors could be overlooking necessary

and important repairs which are being caught by private

sector inspectors.

The data for the empirical test was taken from the 2005 BLS

annual consumer expenditure survey. In this data set

households are chosen at random from around the country

and the head of the household keeps a bi-weekly diary on

their day to day expenditures. In addition, the head of

household responds to a detailed monthly survey with respect

to purchases that are not routine and do not occur on a daily

or weekly basis. During the interview the respondents are

asked to list their vehicle repair expenditures as well as

annual vehicle registration fees and vehicle inspection fees.

Additional information is also collected on the vehicle’s

make, age and mileage, as well as to whether or not the

vehicle was purchased new or used.

Respondents to the survey from both New Jersey and

Pennsylvania were chosen for the empirical test. Both New

Jersey and Pennsylvania have a vehicle inspection program.

However, the major difference between the two states is that

Pennsylvania requires its residents to have their vehicle

inspected once a year by a private inspection station. These

privately owned service stations could be a large chain of

service stations, or proprietarily owned service station, or

maybe even a car dealership. In this case, the vehicle owner

pays the ‘for-profit’ vehicle inspection station a fee for the

state inspection. If a repair is required to pass the inspection,

the vehicle owner can then opt to have the inspection station

do the repairs or have a different service station do the work.

The car is then re-inspected, and if it passes, the vehicle gets

its annual inspection sticker. In general, many vehicle owners

Proceedings of the Pennsylvania Economic Association 134

simply choose to have the original inspection station perform

the repairs since it more convenient and usually will save

time.

New Jersey residents, on the other hand, can opt to have their

vehicle inspected by one of many state-affiliated vehicle

inspection stations located around the state or by a privately

run inspection station. In the past the state of New Jersey

would actually provide inspection services as an alternative

to the privately run inspection stations. However, at the time

of this data set, New Jersey no longer provided the inspection

service themselves. Instead, they sub-contract this service out

to a private firm. This firm is prohibited from performing any

repairs; they only provide the inspection service. Therefore,

in the case of New Jersey, there is no incentive to cheat the

vehicle owner at a state-affiliated inspection station by

requiring unnecessary repairs. If the inspectors at the state-

affiliated stations find a problem with the vehicle, then the

owner must go to a service station of his or her choice, have

the problem fixed and return to the inspection station for a

final inspection and inspection sticker. There is no charge to

New Jersey residents who use this inspection service, but

there is an inspection fee for those New Jersey residents who

opt to use the inspection service of a private inspection

station in New Jersey.

Other than New Jersey residents having the option to go to a

state-affiliated inspection station, there are two other

important differences between New Jersey and Pennsylvania.

First, in New Jersey residents are required to have their

vehicle inspected only once every 2 years as opposed to

Pennsylvania where vehicles are inspected annually.

Secondly, New Jersey does not mandate vehicle inspections

for vehicles that are five years old or newer.

Only New Jersey and Pennsylvania respondents who had

reported owning only one vehicle were selected for the

empirical test. The reason for this is that from the data set it

is impossible to determine which vehicle was being inspected

for households with two or more vehicles. New Jersey

respondents were also deselected from the data set if their

vehicle was newer than 5 years old since, as previously

stated, those vehicles are not required to have an inspection.

It is important to keep in mind that New Jersey residents are

only required to have their vehicles inspected every other

year, while in Pennsylvania the vehicle must get inspected

once a year. Therefore, the only difference, for example,

between a 2005 Honda CRV owned by a New Jersey resident

and a 2005 Honda CRV owned by a Pennsylvania resident is

that the New Jersey vehicle had not been inspected in two

years while the Pennsylvania vehicle was just inspected the

year before. Because of this fact one would expect that repair

bills for Pennsylvania vehicles to be half the amount of repair

bills for New Jersey vehicles. To correct for this discrepancy

the repair costs reported by Pennsylvania residents were

doubled.

A total of 28 vehicles from New Jersey and 120 vehicles

from Pennsylvania were selected for the empirical test. The

difference in the number of observations between the two

states resulted from the facts stated above, which are vehicles

newer than 5 years old do not need to be inspected in New

Jersey and vehicles in Pennsylvania are inspected twice as

often as those vehicles in New Jersey.

Various vehicle repair bills were cumulated and totaled per

survey respondent. The total of the repair bills per vehicle

constitutes the explanatory variable. The following types of

repairs were included as the explanatory variable; brake

work, tire repair, tire purchases and mounting, front end

alignment, wheel balancing and wheel rotation, steering or

front end work, electrical system work, engine repair or

replacement, exhaust system work, engine cooling system

work, clutch or transmission work, motor tune-up, battery

purchase and installation, and finally other vehicle services,

parts, and equipment. Survey respondents reported other

types of vehicle repairs. However, these repairs were most

likely not associated with passing a vehicle inspection, such

as air conditioning repair, tune-up, body work, or radio

repair.

The empirical test is modeled as follows.

C = b1(S) + b2(F/D) + b3(N) + b4(R) + b5(MSRP) + b6(Y) +

b7(M*A)

The variables are defined as follows:

C = This is the total of repair bills during the month of the

vehicle owner’s annual state inspection as in the case of

Pennsylvania residents or during the month of the annual car

registration for New Jersey residents.

S = A value of ‘0’ was assigned to vehicles owned by New

Jersey residents and a value of ‘1’ was assigned to vehicles

owned by Pennsylvania residents.

F/D = A value of ‘0’ was assigned to a vehicle if it was

manufactured by a foreign car manufacturer such as Toyota

or Volvo. A value of ‘1’ was assigned to the vehicle if it was

manufactured by a domestic car manufacturer such as Ford.

These values were assigned regardless of where the vehicle

was manufactured.

N = A value of ‘0’ was assigned if the vehicle was purchased

by the owner as new and a value of ‘1’ was assigned to the

vehicle if the owner purchased it as used.

R = This is a general vehicle reliability index found on the

website ‘autos.msn.com/research/vip/Reliability.aspx?year’

for that specific make, model, and year.

Proceedings of the Pennsylvania Economic Association 135

MSRP = This is the manufacturer’s suggested retail price

found on the website

‘autos.msn.com/research/vip/Reliability.aspx?year’ for that

specific make, model, and year

Y = This represents the average number of miles driven per

year.

M = This represents the total number of miles on the vehicle.

The results of the regression analysis are exhibited below.

The purpose of this paper is to find empirical evidence in

support of problems associated with asymmetric information

and credence goods. Therefore, determining the factors that

explain vehicle repair costs associated with state inspections

is not the reason for running the empirical tests. Rather the

empirical test is more narrowly focused than that, and that is

to find out whether or not the cost of vehicle repairs

associated with a state inspection is statistically different

depending on whether the vehicle is inspected by a private

inspection station or a state-affiliated inspection station.

Therefore, in this regression the explanatory variable of

interest is the state variable.

Table 1 reports the regression results for the empirical model.

The regression results show that the state variable is positive

and significant at the 6.8% level. The positive coefficient of

$166.17 suggests that those vehicles inspected by privately

owned service station can expect to have, on average, an

additional $166.17 in repair bills during the month of their

state inspection. This amount is relevant on a bi-annual basis

since, as stated above, repair bills for Pennsylvania residents

were doubled to account for the fact that they are inspected

twice as often as New Jersey vehicles Since New Jersey

residents are having their vehicles inspected every other year,

then they can expect to save $166.17 every time they go in

for a state inspection.

The finding from the regression analysis supports the

literature on credence goods and asymmetric information,

meaning that service technicians with superior and expert

knowledge over the customer have an incentive to cheat the

customer, and they are, in fact, doing so. This ‘cheating’ or

even the belief that vehicle owners will be cheated is what

creates the market failure.

It should be pointed out that Poitras and Sutter (2002) have

reported the results of a similar study which looks to see if

vehicle inspections can increase vehicle repair cost. They use

a dataset of 733 vehicle inspections for vehicles that were 12

years or older in 50 different states between the years of 1953

– 1967. They find that state inspections do not increase

repair costs (ie, repair revenue for the inspecting facility).

The difference in these results and the results reported here

are most likely attributed to the time period used in both

studies, the type of vehicles used in the study, and the fact

that this study categorizes vehicles inspected into those

inspected by a for-profit or a non-profit inspection facility.

The regression results also show vehicles produced by

foreign manufacturers have statistically higher repair costs

during the month of their state inspection at the 9.3% level of

significance. This result suggests that repair costs associated

with vehicles from foreign manufacturers is $129.63 higher

than for vehicles from domestic manufacturers. There could

be a number of reasons for this result. First, it could be the

case that parts cost more for vehicles from foreign

manufacturers as opposed to domestic manufacturers.

Second, for whatever reason, it could be that vehicles from

foreign manufacturers are slightly more complicated for

American mechanics to work on. Perhaps American

mechanics have a good deal more experience and training

with vehicles made by domestic manufacturers.

The only other variable that was statistically significant was

total mileage. It was positive and significant at the 5.2%

level. The value of the coefficient suggests that for each

addition mile on the vehicle, owners can expect to pay an

additional $.002. This result should be the most intuitive of

all the explanatory variables. The higher the mileage on the

vehicle, the more it cost to maintain

The remaining variables all turned out to be insignificant,

including the number of miles driven over the most recent

year. The coefficient of determination was .229. The

consumer expenditure survey data lacked one or two other

relevant pieces of information which could have increased

the coefficient of determination. This would be the labor

rates charged per vehicle repair shop and information

regarding each shop’s productivity. However, as stated

above, the primary focus of this paper is to find empirical

evidence on the problems associated with asymmetric

information and credence goods.

The data set yields socio-economic characteristics on the

survey respondents. However, it is interesting to point out

that each of these socio-economic characteristics was highly

insignificant. Meaning that, on average, a survey

respondent’s race, gender, age, education level or income

level was insignificant in determining how much they paid

with respect to vehicle repair bills in the month of their state

inspection. Suspicions that a vehicle owner was being taken

advantage of based on their gender, race, age, etc. were

unfounded in this empirical test.

CONCLUSION

Credence good types of services provided by technicians that

are characterized as yielding asymmetric information leave

the consumers at an information disadvantage. This creates

an opportunity for unscrupulous service providers to take

Proceedings of the Pennsylvania Economic Association 136

advantage of the consumer. Indeed, from time to time one

hears stories in the news media of consumers being taken

advantage of. As a result, a market failure arises. An

empirical test was designed and reported on in this paper to

see if the consumer’s fears are warranted.

The empirical test in this paper looks at vehicle owners who

have had their vehicles inspected by either a for-profit

inspection service station or a state affiliated, ‘non-profit’

inspection station. The results from this test indicate that if

your vehicle is inspected by a state-affiliated, ‘non-profit’,

inspection station rather than a ‘for-profit; inspection station,

the vehicle repair bills will be less. The amount as reported

from the regression results is $166.17 on a bi-annual basis.

This is a meaningful savings for owners who utilize the ‘non-

profit’, state affiliated inspection stations. However, there are

a number of factors to consider when interpreting these

results. First, there is the cost of providing vehicle inspection

services. Those vehicle owners in New Jersey do not directly

pay an inspection fee whereas a direct fee is paid to the ‘for

profit’ inspection station in Pennsylvania (when

accumulating repair costs, the cost of the vehicle inspection

fee was not included). The bi-annual savings of $166.17 in

repair bills from New Jersey residents have to be compared

to the cost of running the vehicle inspection stations in New

Jersey. The costs of running the state-affiliated’ inspection

stations are funded by the New Jersey taxpayers.

Secondly, we do not know for sure if vehicle owners in

Pennsylvania are getting more thorough and higher quality

inspections. Perhaps the state-affiliated inspectors from New

Jersey are simply shirking their duties and passing vehicles

that, in reality, do require some repair and maintenance.

There is no way to tell for certain. Although it is beyond the

scope of this paper, it might be possible to look at highway

traffic accident data and see if there is a correlation between

increased accidents and vehicles inspected by state-affiliated

inspection stations. However, the literature with respect to

this topic is inconclusive and shows conflicting results

between vehicle safety inspections and their effectiveness at

preventing accidents (For example, see Fosser 1992, White

1985, or Merrell, D., Poitras, M. & Sutter, D. 1999)

Finally, it need not be that inspections stations that provide

diagnostic inspection services only and none of the required

repairs do not have to be state-affiliated or state operated.

This type of service could just as easily be provided by the

private sector, and perhaps the private sector could perform

the services more efficiently than the state affiliated or state

inspection station. And then it might just be possible that

these private sector companies could run their services more

efficiently than the state affiliated inspection company.

Tables

Table 1 – Regression Results for the Empirical Model

Variable Coefficient “t” – value Significance

State 166.17 1.84 .068

Foreign or domestic 129.63 1.692 .093

Purchase new or used 29.19 .394 .694

Reliability index -21.50 -1.154 .251

MSRP .005 1.009 .315

Annual mileage -.005 -1.225 .223

Total mileage .002 1.956 .052

R-sq .229

REFERENCES

Biehal, G. J. 1983. Consumers' prior experiences and

perceptions in auto repair choice. The Journal of Marketing.

3, summer 1983: 82-91.

Darby, M. & Karni, E, 1973. Free Competition and the

Optimal Amount of Fraud. Journal of Law and Economics.

16, April 1973: 67-88.

Dulleck, U., & Kerschbamer, R. 2006. On doctors,

mechanics, and computer specialists: The economics of

credence goods. Journal of Economic Literature. 44, March

2006: 5-42.

Emons, W. 1997. Credence goods and fraudulent experts.

The Rand Journal of Economics. 28, Spring 1997: 107-119.

Fosser, S. 1992. An experimental evaluation of the effects of

periodic motor vehicle inspection on accident rates. Accident

Analysis & Prevention. 24 December 1992: 599-612.

Proceedings of the Pennsylvania Economic Association 137

Hubbard, Thomas 1998, An Empirical Examination of Moral

Hazard in the Vehicle Inspection Market. Journal of

Economics. 29 Summer 1998: pg 406 – 426.

Hubbard, Thomas 2002. Ho Do Consumers Motivate

Experts? Reputational Incentives in an Auto Repair Market.

Journal of Law and Economics. 45. No. 2, Part 1. Oct. 2002:

437-468

McCollough, J. 2010. Consumer Discount Rates and the

Decision to Repair or Replace a Durable Product: A

Sustainable Consumption Issue. Journal of Economic Issues.

44, March 2010: 183-204.

Merrell, D., Poitras, M., & Sutter, D. 1999. The effectiveness

of vehicle safety inspections: An analysis using panel data.

Southern Economic Journal. 65, January 1999: 571-583.

Peppers, D., & Rogers, M.2006. CUSTOMER

SEGMENTATION STRATEGIES. Pricing on Purpose:

Creating and Capturing Value, 197.

Pesendorfer, W., & Wolinsky, A. 2003. Second opinions and

price competition: Inefficiency in the market for expert

advice. The Review of Economic Studies. 70(2), 2003: 417-

437.

Poitras, M., & Sutter, D. 2002. Policy ineffectiveness or

offsetting behavior? An analysis of vehicle safety

inspections. Southern Economic Journal. 68, April 2002:

922-934.

Rasch, A., & Waibel, C. 2012. What drives fraud in a

credence goods market?-Evidence from a field experiment

(No. 03-07). Cologne Graduate School in Management,

Economics and Social Sciences.

Schneider, H. S. 2012. Agency problems and reputation in

expert services: Evidence from auto repair. Journal of

Industrial Economics. 60, September 2012: 406-433

Webbink, D. W. 1978. Automobile Repair: Does Regulation

or Consumer Information Matter?. The Journal of Consumer

Research. 5, December 1978: 206-209.

White, W. T. 1986. Does periodic vehicle inspection prevent

accidents?. Accident Analysis and Prevention.

18, February 1986: 51-62

Proceedings of the Pennsylvania Economic Association 138

SOME POSSIBLE REASONS FOR THE IRRATIONAL CHOICE OF GIFT CARDS

David Nugent

Robert Morris University

Moon Township, PA 15108

ABSTRACT

This paper is a proposal for a study that will address the

purchase of gift cards. Topics include economic theory that

suggests that the utility of a gift card is less than the utility of

an equal amount of cash. Reasons for the choice of cash gifts,

gift cards and tangible gifts are addressed, with a focus on

gift cards. A number of gift-card related hypotheses are

presented. A questionnaire is proposed that will include

questions related to the hypotheses. Analyses of the

questionnaire data will determine if data is consistent with

the hypotheses.

INTRODUCTION

This paper is a proposal for a study that will address the

question of why people purchase gift cards even though an

equal gift of cash would provide greater utility to most

recipients. Types of gifts given include cash, gift cards and

tangible gifts. In recent years, the purchase of gift cards has

become popular. Card Hub (2013) reports that gift card sales

in 2011 were approximately $99 billion.

Economic theory suggests that a gift of cash should provide

more utility to a recipient than a gift card. Economic

textbooks (Samuelson, 1980; Schiller, 2008) suggest that if a

person had some amount of cash, that person would use the

money to purchase a “market basket” of products and

services that would maximize utility. The market basket

would consist of a combination of housing, food,

transportation, clothing, appliances, etc. that is consistent

with the person’s tastes and preferences. Different people

have different tastes and preferences. Accordingly, market

baskets of purchases also differ. A set of purchases that

maximizes utility for one person will likely not maximize

utility for most other people. If a person has cash, that person

has the choice of buying whatever will maximize utility. But,

if that same person has a gift card for a particular store or

restaurant, it is likely that utility will be less than if the

person had cash that could be spent anywhere. To give a gift

card that that is likely to provide the recipient with less utility

than an equal gift of cash seems to be irrational.

Evidence that at least some recipients would prefer cash to

gift cards includes the market for gift cards. ABC News

(2012) reported that there are several gift card exchange

websites where holders of gift cards can sell their gifts cards

for amounts less than face value. Offenberg (2007) conducted

a study of gift card auctions on eBay for 25 merchants.

Results showed that average sales prices were approximately

15% less than the face value of the gift cards. When fees and

other expenses are considered, the effective discount was

approximately 20%.

REASONS FOR GIVING TANGIBLE GIFTS

Before addressing reasons for choosing gift cards, consider

the decision to give either cash or tangible gifts. For reasons

similar to the preceding argument that a gift card has less

value to a recipient than cash, a tangible gift is likely to have

less value to a typical recipient than the cash spent to

purchase the gift. Waldfogel (1993) conducted a study in

which gift recipients were asked to estimate the cost of gifts

received, and to estimate the value to them, measured as

either the maximum that the recipient would pay for the gifts,

or the minimum that the recipient would accept in lieu of the

gifts. The results indicated that value fell short of cost by

between ten percent and one-third.

Although the monetary value to recipients may be less than

the costs of gifts, it is commonly recognized that the

significance of a gift can be more than simply the money

spent. Studies that address gift-giving (Belk, 1976; Caplow,

1982; Camerer, 1988; Davies, Whelan, Foley and Walsh,

2010; Sherry, 1983; Solnick and Hemenway, 1996; Webley

and Wilson, 1989 ) suggest that the topic is complicated.

Reasons may include expressing affection, reinforcing and

improving personal relationships, facilitating social bonding,

and provide a long-lasting, tangible reminder of the gift-

giver. In choosing gifts, a gift-giver will expend time and

effort to select and purchase an appropriate, thoughtful

expression of an understanding of a recipient’s likes and

dislikes. A resulting gift may be appreciated by a recipient

not because the gift has monetary value, but instead because

of the symbolism that it represents, or because it has

Proceedings of the Pennsylvania Economic Association 139

sentimental value or because it serves as a remembrance of

the giver. In some cases, a gift may have little or no monetary

value. For example, suppose that a 5-year old girl were to

give her grandmother a home-made piece of pottery. A piece

of pottery made by a 5-year old probably has zero market

value, but to the grandmother it could be a prized possession

prominently displayed in her china cabinet.

In contrast to tangible gifts, a gift of cash will likely not

provide a lasting reminder of the gift-giver. A gift of cash

may be viewed as impersonal and as not entailing time and

effort and thought. Imagine if a group of people were to

gather to exchange gifts on a gift-related holiday. If each

person were to give each other person a $50 bill, the result

would be each person ending the day where they started.

Alternately, if some people gave $50 bills and others gave

$20 bills, the inequality of gifts would be immediately

apparent. If the goal of gift exchange were a reciprocal

exchange of equal value, those giving the smaller cash gifts

would be embarrassed and possibly viewed as not fulfilling

the expectation of reciprocity.

The exchange of tangible gifts may give rise to uncertainty

regarding cost that may obscure differences between gifts. If

two people exchange gifts that may or may not have cost $50

each, the potential embarrassment of unequal gifts is

diminished. If each gift entailed time and effort and

thoughtfulness to select and purchase, the significance of

unequal cost would be further diminished.

In comparing cash gifts, gift cards, and tangible gifts, gift

cards have many of the same shortcomings as cash. The cost

of the gift card is explicitly stated. If people were to

exchange gift cards, any inequality of gifts would be

apparent. The recipients of a gift card may perceive that the

purchase of a gift card required little time and effort.

If a person is considering the purchase of a tangible gift, the

non-monetary aspects of the gift may make it an appropriate

choice. However, if a person is considering the purchase of a

gift card, economic theory, as addressed in the preceding,

suggests that a gift card is an irrational choice. A gift of cash

would be more likely to maximize recipient utility.

In some circumstances cash gifts are made. For example,

parents sometimes give substantial amounts of cash to

children and grandchildren with no expectation of equal

reciprocity of cash gifts. In such cases, cash gifts are

appropriate. Imagine, in contrast, if a parent were to give

each child and grandchild a $10,000 gift card to a national

restaurant chain. It seems likely that each recipient would

prefer the cash.

HYPOTHESES

Consider next possible reasons why a gift-giver may make

the seemingly irrational choice of gift cards.

Discounts and Rebates for Purchasers of Gift Cards

The argument that a gift card purchase is irrational is based

on the assumption that the cost of a gift card is the same as

the face value of the gift card. If a purchaser were to receive a

rebate, or if a gift card could be purchased at a discount, or if

some other benefit were available to purchasers, then the

purchase of a gift card may provide the purchaser an

incentive to give a gift card rather than cash. An example of

such an incentive would be the discounts on gasoline

provided by the Giant Eagle supermarket chain. Giant Eagle

(2013) has an incentive program that allows customers 10

cents per gallon discount for up to 30 gallons on Giant Eagle

gasoline for each $50 that a customer spends at Giant Eagle.

Qualified purchases include the purchase of gift cards for

more than 150 stores and restaurants.

If a customer had a gas tank large enough to hold 30 gallons,

the potential discount would be $3.00 ($0.10 X 30 = $3.00).

Accordingly, the purchaser of a $50 gift card could

potentially save on gasoline an amount equal to 6% ($3.00 /

$50.00) of the cost of the gift card.

If the equivalent monetary value to the recipient falls short by

more than the discount or other benefit received by the

purchaser, the net benefit of a gift card would be negative.

If modest incentives influence purchasers of gift cards, the

following can be hypothesized:

H1: Gift-givers are more inclined to purchase

gifts cards if an incentive such as a

discount or rebate is available.

Reduction of Social Risk

The complicated nature of selecting an appropriate tangible

gift may make gift selection difficult. Austin and Huang

(2012) suggest that if a gift-giver lacks knowledge of a

recipient’s needs and preferences, the choice of a gift may

Proceedings of the Pennsylvania Economic Association 140

expose a gift-giver to social risk. A gift-giver may feel

anxiety about choosing the wrong gift, and may be inclined

to reduce exposure to social risk by choosing a gift card.

The preceding leads to the following hypothesis:

H2: Gift-givers are more inclined to purchase

gift cards if their knowledge of the needs

and preferences of recipients is limited.

Possible Preference For Gift Card Characteristics

Although social risk may make a gift-giver inclined to choose

a gift card instead of a tangible gift, it could similarly be

argued that social risk could make a gift-giver inclined to

give cash instead of a tangible gift. Reasons for a preference

for gift cards may be characteristics that make gift cards

seem more appropriate than cash.

A recipient may perceive the selection of a gift card as

entailing some thoughtfulness by the gift-giver. The purchase

of a gift card may entail some time and effort, particularly if

it is for a store or restaurant that the buyer knows that the

recipient regularly patronizes.

Valentin and Allred (2012) suggest that liquidity may

influence the selection of a gift card. If a gift-giver wants to

provide a recipient with a gift that can be readily used to

acquire products or services that the recipient values, but the

gift-giver does not want to simply give cash, a gift card may

be the preferred choice.

This leads to the following hypothesis:

H3: Gift-givers are more inclined to purchase

gift cards if their objective is to provide

recipients with a wide range of product

choices.

Reduction of Recipient Guilt and Regret

Studies that address the emotional aspects of purchases

(Burnett and Lunsford, 1994; Keinan and Kivetz, 2008;

Kivetz and Keinan, 2006) suggest that unneeded or

extravagant purchases may cause a person to feel guilt. A

prudent person may refrain from extravagant purchases in

favor of more responsible purchases. However, over time,

people who have denied themselves luxuries and extravagant

purchases may feel regret because of what they missed.

If a person were to give a gift card for luxury products and

services, the gift card would provide the recipient with an

excuse to indulge in something that ordinarily would be an

unnecessary extravagance. By giving such a gift card, the

gift-giver would be relieving the recipient of the guilt and

regret that otherwise would arise.

This leads to the following hypothesis:

H4: Gift-givers are more inclined to purchase

gift cards if their objective is to provide

recipients with excuses to indulge in

extravagant luxury products.

RESEARCH DESIGN

At this time the design of the research instrument is not

complete. The general approach to data collection will entail

the completion of questionnaires by subjects who have

knowledge of buying and receiving gift cards. Potential

subjects may include college students.

Questions may include:

Have you received gifts cards?

Have you purchased gift cards to give to others?

In choosing gift cards, have you chosen a particular card

because the purchase resulted in a discount, rebate or other

incentive?

Have you given gift cards instead of tangible gifts because

you did not know the kind of tangible gifts that recipients

would prefer?

Have you given gift cards that you felt would provide the

recipients with a wide range of redemption choices?

Have you given gift cards for extravagant luxury products

because you wanted the recipient to have an excuse to

indulge?

Do you consider a gift card to be a more appropriate gift than

cash?

Questionnaire items may also include responses on a scale

that could be quantified for analysis purposes. For example, a

Proceedings of the Pennsylvania Economic Association 141

question might ask subjects to state the degree to which they

feel that cash is an appropriate or inappropriate gift. Another

question might ask subjects to state the degree of inclination

to either purchase a gift card or to give cash.

Analysis of results will entail determining whether

questionnaire responses are consistent with the hypotheses.

CONCLUSIONS

This paper is a proposal for a study that will address the

question of why people purchase gift cards even though an

equal gift of cash would provide greater utility to most

recipients. Economic theory suggests that if a person has

cash, a person will purchase a mixture of products and

services that maximizes utility. A gift card for a specific store

or restaurant limits a person’s choices and likely would result

in less utility than if the person had received cash.

The study addresses three categories of gifts: cash, gift cards

and tangible gifts. For reasons similar to the argument against

gift cards, tangible gifts are likely to provide less utility than

cash. However, for a variety of reasons such as reinforcing

personal relationships, social bonding and other social

factors, tangible gifts may be viewed as more appropriate

than either gift cards or cash. Gift cards generally do not

provide for the social facilitation that tangible gifts provide.

A number of hypotheses are presented that address the topic

of the choice of gift cards. Topics addressed include

discounts and rebates for purchasers of gift cards, reduction

of social risk, possible preference for gift card characteristics,

and reduction of recipient guilt and regret.

This paper proposes that data will be gathered through a

questionnaire that will include questions related to choice of

gift cards. Analyses of the data will determine if data is

consistent with the hypotheses.

REFERENCES

ABC News Website, Hate That Gift Card? Trade It, Sell It,

www.abcnews.go.com/Business/hate-gift-care-trade-

sell/story?id=18083809 retrieved 5/27/2013.

Austin, Caroline Graham and Lei Huang, 2012, First Choice?

Last Resort? Social Risks and Gift Card Selection, Journal of

Marketing Theory and Practice Volume 20, No 3, pp. 293-

305.

Belk, Russell W., 1976, It’s the Thought that Counts: A

Signed Digraph Analysis of Gift Giving, Journal of

Consumer Research Volume 3, pp. 155-162.

Burnett, Melissa S. and Dale A. Lunsford, 1994,

Conceptualizing Guilt in the Consumer Decision-Making

Process, The Journal of Consumer Marketing” Volume 11,

No 3, pp. 33-43.

Camerer, Colin, 1988, Gifts as Economic Signals and Social

Symbols, American Journal of Sociology Volume 94, pp.

S180-S214.

Caplow, Theodore, 1982, Christmas Gifts and Kin Networks,

American Sociological Review Volume 47, No 3, pp. 383-

392.

Card Hub Website, Gift Card Market Size,

www.cardhub.com/edu/gift-card-market-size/ retrieved

5/27/2013.

Davies, Gary, Susan Whelan, Anthony Foley and Margaret

Walsh, 2010, Gifts and Giving, International Journal of

Management Reviews Volume 12, pp. 413-434.

Giant Eagle Website,

www.gianteagle.com/Save.fuelperks/fuelperks-Rules-

Regulations/ retrieved 5/22/2013.

Keinan, Anat and Ran Kivetz, 2008, Remedying Hyperopia:

The effects of Self-Control Regret on Consumer Behavior,

Journal of Marketing Research Volume XLV, pp. 676-689.

Kivetz, Ran and Anat Keinan, 2006, Repenting Hyperopia:

An Analysis of Self-Control Regrets, Journal of Consumer

Research Volume 33, Issue 2, pp. 273-282.

Offenberg, Jennifer Pate, 2007, Markets: Gift Cards, Journal

of Economic Perspectives Volume 21, No 2, pp. 227-238.

Samuelson, Paul A., 1980, Economics, New York: McGraw-

Hill.

Schiller, Bradley R., The Economy Today, New York:

McGraw-Hill Irwin.

Proceedings of the Pennsylvania Economic Association 142

Sherry, John F., 1983, Gift Giving in Anthropological

Perspective, Journal of Consumer Research Volume 10, pp.

157-168.

Solnick, Sara J. and David Hemenway, 1996, The

Deadweight Loss of Christmas: Comment, The American

Economic Review Volume 86, No 5, pp. 1299-1305.

Valentin, Erhard K. and Anthony T. Allred, 2012, Giving and

getting Gift Cards, Journal of Consumer Marketing Volume

29, No 4, pp. 271-279.

Waldfofgel, Joel, 1993, The Deadweight Loss of Christmas,

The American Economic Review Volume 83, No 5, pp.

1328-1336.

Webley, Paul and Richenda Wilson, 1989, Social

Relationships and the Unacceptability of Money as a Gift,

Journal of Social Psychology Volume 129, Issue 1, pp. 85-

91.

Proceedings of the Pennsylvania Economic Association 143

SERVICE QUALITY IN THE U.S. AIRLINE INDUSTRY: FACTORS AFFECTING CUSTOMER SATISFACTION

Logyn Pezak

Rose Sebastianelli

Department of Operations & Information Management

Kania School of Management

University of Scranton, Scranton, PA 18510

ABSTRACT

Airlines have consistently received low customer satisfaction

ratings, with the industry ranking third from the bottom in

terms of the American Customer Satisfaction Index (ACSI).

Using data obtained from the U.S. Department of

Transportation’s Air Travel Consumer Report, we examine

the relationship between ACSI ratings and a number of

explanatory variables related to service quality. Regression

results indicate that in addition to time period and percentage

of on-time arrivals, ACSI ratings are significantly impacted

by four specific complaint categories. These four categories

are complaints related to (1) reservations, ticketing and

boarding, (2) baggage, (3) customer service, and (4)

disability.

INTRODUCTION

Airlines are notoriously noted for poor service quality. In

2012, the airline industry had the third lowest American

Customer Satisfaction Index (ACSI) rating among all

industries (American Customer Satisfaction Index, 2012). Its

ACSI rating actually decreased 6.9% from 1995, the first

year it was measured. While technological advances over this

period would presumably have led to improved service

quality and higher levels of customer satisfaction, this has not

been the case for most U.S. airlines. Airlines are

simultaneously faced with satisfying customers that expect

lower prices, better service quality, and higher comfort levels

while operating profitably in a competitive business

environment with a number of economic challenges (J.D.

Power and Associates, 2012). In order to be successful,

however, airlines must understand that financial well-being

depends on sales, and sales depend on meeting and/or

exceeding customer expectations. Airlines that recognize the

importance of satisfying customers are those that will

continue to be financially viable and profitable in the long

term.

This paper reports the findings of an empirical study

designed to determine what factors are significant in

explaining customer satisfaction in the U.S. airline industry.

We use ACSI ratings for U.S. airline companies as the

measure of customer satisfaction. While a number of factors

affect customer satisfaction, our focus is on the service

quality provided by airlines. By doing so, we limit attention

to factors over which airline companies have some control.

Most of the independent variables included in our study

involve specific categories of airline customer complaints.

Complaints play a large part in any satisfaction rating system.

For example, The Wall Street Journal annually ranks airlines.

Its rankings showed that Southwest is at No. 1 with the

lowest number of customer complaints among U.S. airlines

in 2012. Consequently, determining which customer

complaint issues have the most impact on customer

satisfaction should help airlines to better manage service

quality and meet the expectations of its customers.

RELEVANT LITERAURE

A number of studies have examined service quality within

the airline industry. Most recently, Degirmenci et al. (2012)

used the SERVQUAL scale to measure customer satisfaction

levels among Turkish Airline passengers. The surveys used

were designed according to Skytrax, the most accepted and

prestigious official airline quality 5-star rating system. The

results of this study showed that meals and passenger

transferring services have the highest impact on customer

satisfaction. Huang (2009) also employed a marketing

approach to examine service quality in the airline industry.

Based on a sample of airline passengers in Taiwan, the study

explored how companies’ responses to “other-customer

failures” influenced customer satisfaction. Customers may

become dissatisfied with a company because of how other

customers act or behave. Employee efforts and

compensation were considered as contributing factors in this

study. An important finding was that companies can benefit

from encouraging dissatisfied customers to voice their

complaints as it gives them the opportunity to make amends

and identify root causes responsible for service failures.

The terrorist attacks of September 11, 2001, are perhaps the

most important singular event to impact the U.S. airline

industry. Cunningham, Young, and Lee (2004) examined its

effect on customer perceptions of airline service quality using

the SERVPERF scale. They found that although the number

Proceedings of the Pennsylvania Economic Association 144

of airline trips declined following 9/11, there was no

statistically significant change in overall satisfaction levels

among airline passengers. These results suggest that events

perceived as being “outside” the direct control of airlines

may have limited to no effect on customer satisfaction levels.

Our study most closely resembles that of Rhodes and

Waguespack (2000) in that they too examined complaint data

(obtained from the Department of Transportation’s Air

Travel Consumer Report). Their goal was to use these data

to evaluate the level of service quality both within the airline

industry and between airline companies. Based on data

collected for the period from 1987 to 1996, they found that

airline quality was improving, with Southwest having the

lowest complaint rate (3 complaints per 10,000 departures) of

all airlines. Their study also found that many airline

passengers associate service quality with safety. They did

not link complaint data, however, to any measure of customer

satisfaction.

RESEARCH STUDY

Objectives

The main objective of this empirical study is to identify

factors that have a significant impact on customer satisfaction

in the U.S. airline industry. Primarily, we examine the

relationship between customer satisfaction ratings (ACSI

rating) of U.S. airlines and a set of explanatory variables,

most representing the specific categories of complaint data

compiled by the U.S. Department of Transportation. In

addition, we also examine the differences in mean ACSI

ratings among the U.S. airlines included in our study.

Variables and Data Collection

The secondary data for this study were obtained from two

sources: The American Consumer Satisfaction Index (ACSI)

and the Department of Transportation’s Air Travel Consumer

Report. Data were collected for seven U.S. airlines

(American, Continental, Delta, Northwest, Southwest,

United, and US Airways) that were in operation during the

14-year period from 1998 to 2011. Two airlines (Northwest

and Continental) were not in operation during the entire

period as a result of mergers.

The dependent variable is the annual ACSI rating for each

airline. The ACSI rating is a national benchmark of customer

satisfaction produced by a private company based in Ann

Arbor, Michigan. It aims to quantify customer satisfaction

and quality perceptions and relate them to firm financial

performance (Brecka, 1994). It is based on a sample of 250

customer interviews, with more than 70,000 interviews

conducted annually using random samples via telephone and

email (Carey, 2012). The ACSI quantifies how perceived

quality, perceived value, and customer expectations affect

each other in addition to how they affect customer

satisfaction.

Data for the independent (explanatory) variables were

obtained from the Air Travel Consumer Report, a report

published monthly by the U.S. Department of

Transportation. With the exception of Percentage of On-

Time Arrivals, all independent variables represent specific

passenger complaint categories. Please refer to Table 1 for a

list of the variables used in this study and a brief description

of each.

Several adjustments to the data were required in order to

make them comparable across airlines and with the time

frame of the ACSI rating. To make the complaint data

comparable across airlines, it was necessary to account for

differences in company size and number of enplaned

passengers. This was done by computing an annual index for

each complaint type per airline per year. This was achieved

by dividing the number of complaints in a particular category

in a given year by the total number of complaints for that

airline in the same year. This was then multiplied by the

complaints per 100,000 enplanements (given in the report)

for that airline and year. These indexes made it possible to

compare specific complaint categories across airlines as the

index now represented the number of a specific type of

complaint per 100,000 enplaned passengers.

The ACSI ratings for each airline are published in June and

reflect the previous 12 months (i.e., a particular year’s ACSI

rating reflects data from June of the previous year to May of

the current year). However, the annual complaint data

(independent variables) obtained from the Air Travel

Consumer Report represent calendar years (i.e., January

through December). In order to make the time periods

comparable between the dependent variable (ACSI rating)

and the independent variables, we adjusted the ACSI ratings

so they too represented a calendar year. To do this,

proportions were taken to combine two years of ACSI data

into one representing a standard calendar year. For example,

the ACSI rating for the year 2000 in this study is actually

five-twelfths of the ACSI rating for 2000 and seven-twelfths

of the ACSI rating for 2001.

Proceedings of the Pennsylvania Economic Association 145

Data Analysis and Results

Before determining which explanatory variables have a

significant impact on the ACSI ratings of U.S. airlines, we

first examine if (and how) the mean ACSI ratings might

differ among the seven airline companies in our study. Given

that the annual airline ACSI ratings may be influenced by

economic and/or other factors related to time period, we use

year as the blocking variable. Consequently, we perform an

analysis of variance (ANOVA) using one factor (airline) and

one block (year), the response variable being annual ACSI

rating. The ANOVA results appear in Table 2.

Taking into account the variation in ACSI ratings explained

by time period (the blocking variable) we can reject the null

hypothesis that the mean ACSI ratings are the same across all

seven airlines. Further analysis reveals that the mean ACSI

rating for Southwest is significantly higher than for all other

airlines included in our study. Moreover, Continental was

found to have a significantly higher mean ACSI rating than

the remaining six airlines. In addition, American and Delta,

while not significantly different from each other, had

significantly higher mean ACSI ratings compared with

Northwest, United and US Airways.

Stepwise regression was used to fit a multiple regression

model with annual airline company ACSI rating as the

dependent variable and thirteen potential independent

variables (the twelve listed in Table 1 and Year). Stepwise

regression is an automatic model building procedure that

employs a forward selection method for adding variables to

the model with a “backward” look to drop any that become

insignificant. The data used to fit the model spanned the 14-

year period from 1998 to 2011. However, as a result of

mergers, data for Northwest and Continental were available

only through 2009 and 2010, respectively. The data were

found to satisfy the basic assumptions for regression.

Regression results are reported in Table 3.

Of the thirteen potential independent variables presumed to

be related to ACSI ratings, six are found to be statistically

significant (at α = 0.05). These are the four complaint

categories of Reservations/Ticketing/Boarding, Baggage,

Customer Service, and Disability as well as Year and

Percentage of On-Time Arrivals. The model explains almost

65% of the variability in ACSI ratings as indicated by the

resulting R2 value of .648. As expected, the coefficient

associated with the percentage of on-time arrivals is positive.

Also as expected, the coefficients associated with the various

complaint categories are negative, with the exception of

Customer Service. The correlation matrix of all variables

shows that Customer Service complaints are indeed

negatively associated with ACSI ratings (r = -0.395), but it

also shows that Customer Service complaints are positively

correlated with all three other complaint categories included

in the regression model (the correlations are 0.641 with

Reservations, Ticketing, Boarding, 0.688 with Baggage, and

0.444 with Disability). Consequently, in the presence of the

other variables in the model, Customer Service complaints

have the effect of increasing ACSI ratings. This is because

the decrease in ACSI ratings caused by Customer Service is

already being accounted for by the other complaint categories

included the model.

DISCUSSION, LIMITATIONS, IMPLICATIONS

Our study shows that in addition to the percentage of on-time

arrivals, complaints concerning reservations / ticketing /

boarding, baggage, disability, and customer service have a

statistically significant impact on ACSI ratings of airlines.

Year was also found to be significant, indicating that factors

related to time period, perhaps economic or environmental in

nature, also affect customer satisfaction ratings. These

findings suggest that airline companies can improve

satisfaction among its customers by offering superior service

in these particular areas. Reducing the number of customer

complaints in these specific categories airlines improves the

ACSI rating. However, this does not mean that airlines

should discourage customers from complaining. As

previously mentioned, studies suggest that airlines can

benefit from encouraging dissatisfied customers to complain

because it gives them an opportunity to make amends and

identify root causes (Huang 2009). It is important, however,

for them to act on the complaints and work to eliminate the

root causes for problems that result in dissatisfaction. This is

what leads to fewer complaints, improved service quality,

and increased levels of customer satisfaction.

Another finding from this study is that Southwest has a

statistically significantly higher mean ACSI rating than all

other airlines. This corresponds to Southwest having

substantially fewer complaints per 100,000 customers.

Boxplots of ACSI ratings for specific airline companies over

time show increased variability, suggesting that airline

companies are increasingly differentiating themselves in

terms of meeting and/or exceeding customer expectations

(see Figure 1). This leads to the question: What makes

Southwest so much better than its competitors? Southwest is

the only so called “low-cost carrier” to be included in this

Proceedings of the Pennsylvania Economic Association 146

study. In the past, a low-cost carrier was defined as having a

one-class cabin, high aircraft utilization, all internet sales and

low fares (Low-cost Carriers Become Harder to Define,

2008). That definition, however, is beginning to become

obsolete with the influx of so many low-cost carriers

worldwide. Today it is harder to classify airlines as such

because of the narrowing gap in cost between the “low-cost”

companies and the legacy companies. However, as this study

dates back to 1998, the fact that Southwest was originally

perceived as a “low-cost” carrier may help explain why its

customer satisfaction is higher than other airlines. Since

customer satisfaction depends on the fulfillment of customer

expectations, perhaps its customers’ expectations are lower

than those of the more “expensive” carriers. Southwest is

also recognized as having a very pleasant workforce (The

Rise of Southwest Airlines, Advance! Business Consulting).

“Pleasant” is not always a term used to describe any airport

experience, but Southwest aims to keep its employees happy

in order to keep its customers happy. Its mission places a

high value on customer satisfaction (The Mission of

Southwest Airlines, Southwest.com.). It is evident that this

customer-driven strategy has been quite successful for

Southwest Airlines.

As with all research studies, ours is not without limitations.

First, surely there are other factors not included here that

impact customer satisfaction in the airline industry. Second,

our study relies on complaint data. It is likely that complaint

data seriously underestimates the true level of customer

dissatisfaction. Many customers simply do not complain.

For example, in 2011 there were 16.2 mishandled baggage

reports per 100,000 customers yet there were only 1.13

baggage complaints per 100,000 customers. The difference in

these numbers shows that either airline passengers were

unaware that they could file a complaint with the Department

of Transportation or just did not bother to do so. Finally, we

use the ACSI rating as a proxy for customer satisfaction

rather than measuring it directly.

CONCLUDING REMARKS

Even with its limitations, our study supports the notion that

airlines can (and should) improve service quality in an effort

to increase customer satisfaction. While a number of factors

affecting customer satisfaction may not be under the control

of airlines (e.g., airport issues, weather, fuel prices, and the

economy), our results suggest that by paying attention to a

few specific service areas that are under their control, namely

those related to reservations, ticketing and boarding,

baggage, customer service, and disability, may help airlines

keep more of its customers satisfied. By offering improved

service in these key areas, and making customer satisfaction

a high priority, an airline may be able to gain the competitive

edge needed to remain viable and profitable well into the

future.

Proceedings of the Pennsylvania Economic Association 147

Table 1: Independent (Explanatory) Variables

Variables Descriptions

Flight Problems Cancellations, delays, or any other deviations from schedule, whether planned or

unplanned.

Reservations, Ticketing,

Boarding

Airline or travel agent mistakes made in reservations and ticketing; problems in

making reservations and obtaining tickets due to busy telephone lines or waiting in

line, or delays in mailing tickets; problems boarding the aircraft (except oversales).

Fares Incorrect or incomplete information about fares, discount fare conditions and

availability, overcharges, fare increases and level of fares in general.

Refunds Problems in obtaining refunds for unused or lost tickets, fare adjustments, or

bankruptcies.

Baggage Claims for lost, damaged or delayed baggage, charges for excess baggage, carry-on

problems, and difficulties with airline claims procedures.

Customer Service Rude or unhelpful employees, inadequate meals or cabin service, treatment of

delayed passengers.

Disability Civil rights complaints by air travelers with disabilities.

Advertising Advertising that is unfair, misleading or offensive to consumers.

Discrimination Civil rights complaints by air travelers (other than disability); for example,

complaints based on race, national origin, religion, etc.

Animals Loss, injury or death of an animal during air transport provided by an air carrier.

Other Frequent flyer, smoking, tours credit, cargo problems, security, airport facilities,

claims for bodily injury, and others not classified above.

Percentage of On-Time

Arrivals

A flight is counted as "on time" if it operated less than 15 minutes after the

scheduled time shown in the carriers' Computerized Reservations Systems (CRS).

Proceedings of the Pennsylvania Economic Association 148

Table 2: ANOVA Results

Source Degrees of Freedom

Sums of Squares

Mean Square F Statistic P-value

Airline 6 1843.90 307.316 60.95 < 0.0001

Year (Block)* 11 151.34 13.758 2.73 0.0060

Error 66 32.77 5.042

Total 83 2328.00 *Data used in ANOVA are for the years 1998-2009. Because of mergers, data were not available for Northwest

after 2009 and for Continental after 2010.

Table 3: Multiple Regression Results

Model Coefficient T-ratio P-value

Constant -709.2 -2.80 0.006

Reservations, Ticketing, Boarding

-45.680 -6.57 0.000

Baggage -11.787 -2.50 0.014

Customer Service 16.159 4.50 0.000

Percentage of On-Time Arrivals

0.2824 2.58 0.012

Disability -45.11 -3.03 0.003

Year 0.3796 3.01 0.003

Proceedings of the Pennsylvania Economic Association 149

Figure 1: Boxplots of ACSI Ratings by Airline and Over Time

US AirwaysUnitedSouthwestNorthwestDeltaContinentalAmerican

80

75

70

65

60

55

Da

ta

Boxplot of American, Continental, Delta, Northwest, Southwest, ...

20112010200920082007200620052004200320022001200019991998

80

75

70

65

60

55

Da

ta

Boxplot of 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, ...

Proceedings of the Pennsylvania Economic Association 150

REFERENCES

American customer satisfaction index. 2012. Benchmarks by

Industry: Ann Arbor, MI.

Carey, S. 2013. U.S. Airlines scored poorly in consumer

survey. The Wall Street Journal. January.

http://online.wsj.com/article/SB10001424052702303703004

577475002478830004.html

Cunningham, L.F. & Young, C.E. 2004. Perceptions of

airline service quality: pre and post 9/11. Public Works

Management and Policy. 9(10): 10-25.

Degirmenci, E., Basligil, H., Bolat, A., & Ozdemir, Y. 2012.

Customer satisfaction measurement in the airline services

using SERVQUAL. Open Access Scientific Reports. 1(5): 1-

9.

Air travel consumer report. 1998-2011. Department of

Transportation Office of Aviation Enforcement and

Proceedings, Washington, D.C.

Huang, W. 2010. Other-customer failure. Journal of Service

Management. 21(2): 191-211.

2012 North America airline satisfaction study.

.http://www.jdpower.com/content/press-

release/aOGunkG/2012-north-america-airline-satisfaction-

study.htm

Low-cost carriers become harder to define. Flight Global.

Airline Business. 19, May 2008. Retrieved from

http://www.flightglobal.com/news/articles/low-cost-carriers-

become-harder-to-define-223826/

McCartney, S. 2013. Believe it or not, flying is improving.

The Wall Street Journal.

Rhoades, D.L. & Waguespack, B. 2000. Service quality in

the U.S. airline industry: variations in performance within

and between airlines and the industry. Journal of Air

Transportation World Wide. 5(1): 60-77.

The Rise of Southwest Airlines. Advance! Business

Consulting.

http://www.advancebusinessconsulting.com/advance!/strategi

c-alignment/strategic-alignment-business-cases/the-rise-of-

southwest-airlines.aspx

Proceedings of the Pennsylvania Economic Association 151

WHAT AFFECTS NEW ZEALAND WINE PRICES? ESTIMATION OF THE EFFECTS OF SENSORIAL,

REPUTATIONAL, OBJECTIVE, AND QUALITY FACTORS IN THE HEDONIC PRICE MODEL

Angela M. Rowland

Indiana University of Pennsylvania

ABSTRACT

This paper applies the hedonic price model to determine the

factors that influence the bottle price of New Zealand wines.

This study primarily focuses on sensorial, reputational,

objective, and quality factors and how they influence the

bottle price. Included are wine vintages from 1995 to 2012

from every main wine-producing region of New Zealand. All

data and ratings were taken from wine expert Robert Parker’s

website. Results indicate that though the Parker Rating is

significant in determining bottle price in other countries, it

only explains a small percentage of the variation in bottle

price for New Zealand wines. Various sensorial and objective

characteristics are more statistically significant to bottle price

than others when regional and quality attributes are

controlled for, regardless of Parker Ratings. This suggests

that, though the Parker Rating does have a significant effect

on bottle prices of New Zealand wines, objective,

reputational, and sensorial elements are also significantly

correlated.

INTRODUCTION

Background

A generally unknown fact about wine is that it is the most

traded liquid commodity in the world, second only to the

global oil trade (Sechrist, 2012). A social lubricant that has

withstood the test of time, it actually may have been one of

the catalysts for the creation of civilization. While many

theories attribute the birth of civilization to the growing of

grains and the need to gather in large groups to protect

immobile resources, there are some that claim grapes and

wine began this process instead. While this will most likely

always be a debate for speculation, the fact that wine has

been around since the beginning of civilization remains. Not

only is it used leisurely for pleasure in social situations, but it

also can have significant ritualistic and spiritual connotations

in various cultures and religions, particularly those found in

western societies. A commonly known use of wine in a

religion is the Eucharist in Christianity, but many other

religions use it as well. Various sects of Judaism, for

instance, celebrate with Kosher wines such as Manischewitz

(Kramer, 2010).

Wine Consumption

Over the past 10 years global wine-consumption has risen

10%, roughly 1% per year, which fell short of the predicted

20% due to the deteriorating economic climate (AFP with

Wine-Searcher Staff, 2013). According to the Beverage

Information Group, wine-consumption has consistently

increased every year for the past 16 years in the United States

alone, which is the predominant wine-consuming nation by

both value and volume. China’s wine-consumption is on

track to becoming second globally by 2016 and already has a

significant impact on global wine-consumption (AFP with

Wine-Searcher Staff, 2013). Wine-consumption in traditional

wine-producing areas like France, Italy, Spain, and the

United Kingdom, however, has dramatically decreased due to

the recent financial crises. Despite these decreases and the

recent recessions faced around the world, global consumption

is expected to grow 5% between 2012 and 2016, compared to

the 3% increase from 2007 to 2011 (AFP with Wine-Searcher

Staff, 2013). United States wine consumption is expected to

rise at least 40% between 2012 and 2016, and Chinese wine-

consumption is expected to rise at least 12% in the same

period (AFP with Wine-Searcher Staff, 2013). These

increases are largely due to a progressively more competitive

marketplace with new brands and competitive prices. In

contrast, European wine demand is expected to continue its

downward trend, partially due to the economic crisis and

partially due to changing drinking preferences (AFP with

Wine-Searcher Staff, 2013). Today, wine demand is so high

that nations around the world have begun to challenge the

traditional wine-growing countries like France, Italy, and

Spain and have begun developing their own reputations as

competitive wine-makers. One of the most recent of these

upstarts in the wine industry is New Zealand, whose wines

have become particularly popular in Australia, China, and the

United States.

New Zealand Wines

Prior to the 1960s, New Zealand’s grape cultivation on any

scale was nonexistent. Early grape-growers produced low-

quality grapes that made even worse wines, leading to a

dead-end for New Zealand wineries (Sechrist, 2012). The

technology and skills to grow quality grapes in the New

Zealand terroir simply was not available. Persistence

prevailed, however, and by the 1980s around 14,000 acres of

land were being used to grow grapes for the purpose of wine-

making (Kramer, 2010). With new developments in grape-

growing knowledge and technology across the world,

however, New Zealand grape-growers began experimenting

and updating their methods of both growing grapes and

making wine. Early attempts ended poorly with prices for

New Zealand wine plummeting because of the terrible

quality of its wines (Sechrist, 2012). White wines, such as

Proceedings of the Pennsylvania Economic Association 152

Sauvignon Blanc and Chardonnay, were the most successful

until the 1990s when the South Island’s red wines started to

meet globally competitive standards (Sechrist, 2012). Today,

New Zealand’s Pinot Noir, Cabernet Sauvignon, and Merlot

hold their own even against traditional French Bordeaux

wines (Kramer, 2010).

The issue in question, however, is what characteristics make

a wine desirable to consumers, in this case from New

Zealand. The topic examined in this paper is the determinants

of bottle price using the hedonic price model for New

Zealand wines in the global market for 2012. Various studies,

such as Bicknell and MacDonald (2012), Oczkowski (2001)

and Anderson and Schamel (2003), analyze the effects of

ratings and reputation on wine prices in New Zealand. There

has also been some research into the effects of sensorial

characteristics on New Zealand wines, such as that conducted

by Combris, Lecocq, and Visser (1997). Studies such as

Bicknell and MacDonald (2012) have been conducted on the

influences of a particular type of grape variety or wine on

bottle price.

This study will broaden the research to include reputational,

sensorial, quality, and objective variables to create a more

complete breakdown of influences on New Zealand wine

prices. As of yet, no other study to my knowledge has

analyzed New Zealand wines for years later than 2010, and

no paper later than 2003 conducts a comprehensive study for

New Zealand such as this one. Much of the information I

gathered came from world-renowned wine critic Robert

Parker’s website and database, Robert Parker’s Wine

Advocates. Robert Parker is famous for his ability to

influence bottle prices with his ratings and reviews

(Anderson & Schamel, 2003).

This study seeks to provide better insight into what

characteristics make a bottle of wine desirable to consumers,

particularly in determining how significant Parker Ratings

are to New Zealand bottle prices. This could in turn be used

by individuals and businesses to make wiser purchases when

buying New Zealand wines and by New Zealand vineyards to

produce more wines with those particular desired attributes.

If the Parker Rating were significant for New Zealand wines,

then wineries would be able to determine what characteristics

are correlated with the highest Parker Ratings and target

those characteristics. If the Parker Rating were not significant

for New Zealand wines, then wineries could still use the

information from this study to see which characteristics are

the most significant influences on bottle price.

Outline of Paper

This study attempts to determine the significant objective,

sensorial, reputational, and quality factors that influence the

bottle price of New Zealand wines in the hedonic price

model. The second section will review the literature used in

this study to identify the factors associated with bottle price.

The descriptive statistics, data, and variables used will be

discussed in the third section, and the fourth will detail the

econometric model and methodology utilized in this study.

The fifth section will discuss the regression equations and

results of the various models. Conclusions and implications

of the findings will be examined in the final section.

LITERATURE REVIEW

Various studies have undertaken the task of analyzing the

factors that influence bottle prices of wine. Oczkowski

(1994) conducted one of the first empirical wine studies,

adopting the hedonic price model estimating the log-linear

function of Australian premium wines. A similar method was

also used by Combris et al. (1997) in analyzing the hedonic

prices of Bordeaux wines of France. They explained that the

determinants of wine prices are not necessarily easy to define

because of their qualitative nature. Results indicated that the

market prices for Bordeaux wines were primarily determined

by characteristics found right on the bottle. Combris et al.

(1997) focused on the impact of both reputational and

sensorial characteristics on bottle prices.

Oczkowski (2001) claimed that when single indicators are

used for reputation and quality they ultimately contain

measurement error because, as shown by Cliff and King

(1996), expert wine tasters’ evaluations differ. This problem

can be averted if quality and reputation are treated as latent

constructs, which can only be reflected by multiple observed

indicators. Thus, Oczkowski (2001) used factor analysis to

consistently estimate hedonic prices in the presence of

attributes measured with error by using expert ratings by

Robert Parker. He found that theoretical and empirical

evidence both indicated that wine prices depend upon quality,

reputation, and objective characteristics. In the application to

Australian premium wines, reputational factors proved

significant to bottle price while quality factors were

insignificant. This theory was fortified by Bicknell and

MacDonald (2012), who noted that wine is a good with

highly differentiated experiences with limited availability of

information prior to consumption. Consumers are forced to

use reputation and expert ratings as proxies of a wine’s value

to make purchasing decisions. Thus, Bicknell and

MacDonald (2012) used hedonic price analysis to generate

implicit prices for a sensorial quality rating and regional

reputation for premium New Zealand wine.

Benfratello, Piacenza, and Sacchetto (2004) were the first to

use a combination of the different groups of variables to

estimate the market price of Italian premium wines. As

suggested by Oczkowski (2001), consumers of wine

frequently face imperfect information, especially concerning

sensorial characteristics. One cannot determine the sensorial

characteristics prior to purchasing the wine. Thus, other types

of information provided by quality, reputation, and objective

Proceedings of the Pennsylvania Economic Association 153

data fill the void. Using price as the dependent variable, and

objective, sensory, reputational, and quality characteristics

for the independent variables they develop new methods of

explaining wine price using a single producer’s reputation.

They found that reputation outperformed taste significantly in

estimating price, but both were significant.

Arguea and Hsiao (1993) examined the econometric issues

found in estimating hedonic price functions by analyzing the

U.S. market for automobiles. They found that consumers are

ambivalent to focusing on every minute detail. This can also

be applied to the wine market, as the average consumer of

wine likely would not notice the small details in a particular

bottle. For instance, consumers are more likely to think a

wine tastes fruity as opposed to associating a flavor with a

particular spice like black pepper or anise. Dummy variables

were used by Arguea and Hsiao (1993) and were found to be

highly significant. Not including them would have created

substantial omitted variable bias. There was some degree of

multicollinearity; however, they claimed that

multicollinearity is to be expected with dummy variable

groups and as long as the equations are homoskedastic it is

not a problem. Therefore, this study also uses dummy

variables for sensorial characteristics, such as flavors from

the wine wheel and body style, along with other variables

such as growing region and grape variety.

Similarly, Anderson and Schamel (2003) conducted a

combined study of changes in hedonic price functions of

Australian and New Zealand wines from 1992 to 2000 with a

focus on regional reputations and expert ratings. The

dependent variable was the logged price of a bottle of wine.

They also tested a log-log form and a linear form with the

log-linear model giving the best results. For their independent

variables they included dummies for red varieties, white

varietes, growing region, and whether a wine is a classic or

not. They also included Parker Rating, star rating, vintage

rating, winery rating, and point rating. Anderson and

Schamel (2003) concluded that in New Zealand, vintage

ratings were significant and fairly constant over time.

Varietal and regional differences were far less significant

from 1992 to 2000.

DATA

This study is a cross-sectional analysis that uses the hedonic

price model to test for a relationship between the bottle price

and a wine’s objective, quality, reputation, and sensorial

characteristics. All data was taken from Robert Parker’s

website, Robert Parker’s Wine Advocates. Anderson and

Schamel (2003) use this website as the main source for their

data as well. Robert Parker is a renowned wine enthusiast

whose critiques are internationally valued and trusted by

wineries and consumers alike (Oczkowski, 2001). The data

pertains to all major wine-producing regions in New Zealand,

including vintages ranging from 1995 to 2012. A map of

these regions, along with common grape varieties found in

each region, can be found in Figure 1 of the Appendix. Only

red and white wines are included, as including rose would

create bias and be inconclusive because only 3 or 4 bottles

were available for the sample.

Data taken from Robert Parker’s website includes wine

ratings (PARKRAT), which range from 0 to 100; 0 being the

worst and 100 being the best. Descriptions of these ratings

can be seen in Table 1 of the Appendix. Any rating below a

60 is deemed an unacceptable wine and is not included on his

website, and thus not in this study. Also taken from his

website are the sensorial characteristics, which are used to

distinguish between different flavors and textures present in

different wines. Arguea and Hsiao (1993) state that it is

unlikely that the average consumer will distinguish between

flavors on a more specific scale than the basic categories of

the aromatic wine wheel, thus this study limits the sensorial

variables to the broadest flavor categories consumers

typically detect. These categories are taken from the Davis

Wine Aroma Wheel, which can be found at The Wine Cellar

Insider (Leve, 2010).

Combris et al. (1997) found that the quality measures used by

wine tasters are primarily explained by sensorial

characteristics. Since wine tasters are essential in determining

the value of a particular wine, sensorial variables are clearly

important in determining the market price of wine and I will

use them in this study. Bicknell and MacDonald (2012)

indicated that both regional reputation and quality ratings

vary between grape varieties, using three different varieties.

Therefore, I will be including grape varieties to account for

this variation, although I include six varieties.

Oczkowski (2001) revealed that quality, reputational and

objective variables are potentially significant to the bottle

price of wine; therefore, they will be included in the sample.

Though they focused on Australian wines, the same can be

applied to New Zealand wines. Benfratello et al. (2004) used

a similar approach but included sensorial characteristics as

well. They also included variables such as alcohol gradation,

which was unavailable for the majority of the wines in my

sample and thus excluded from this study. Producer

reputation was also incorporated, which is unavailable for

this dataset.

Anderson and Schamel (2003) also incorporated reputational,

quality, and objective variables, but for both Australian and

New Zealand wines and excluding any sensorial variables.

They created an overall examination of what influences

bottle prices of New Zealand wine rather than isolating a

particular characteristic that only influences a part of the

bottle price. This study concentrates most heavily on quality

and reputational variables, using objective variables mostly

as controls to reduce omitted variable bias. Prior to this

study, however, no study to my knowledge has included

Proceedings of the Pennsylvania Economic Association 154

reputational, quality, and objective characteristics along with

sensorial characteristics such as flavor for New Zealand

wines. Thus, this is a broadly inclusive study of New Zealand

wines for 2012 prices.

Sensorial Variables

Sensorial variables used in this study include flavors, color,

and body. Flavors are divided into dummy variables for the

general categories taken from the wine wheel: fruity,

herbal/vegetal, nutty, caramel, woody, earthy, chemical,

pungent, oxidized, microbiological, floral, and spicy. These

are FRUITY, HERBVEG, NUTTY, CARAM, WOODY,

EARTHY, CHEM, PUNG, OXID, MICROB, FLORAL, and

SPICY, respectively. Though PUNG, CHEM, and OXID are

primary categories on the aromatic wine wheel, they were

condensed into one category OTHERFLV. Wine color is also

dummied for red and white; COLORR and COLORW,

respectively. Rose wines were not included due to lack of

data availability. Similarly, dummy variables were used for

body, which is divided into full, medium, and light:

BODYFULL, BODYMED, and BODYLGT, respectively.

For the same reasons that PUNG, CHEM, and OXID were

condensed, so were medium-full and medium light.

Medium-light was combined with light, and medium-full was

combined with full. Most wines fell between full and

medium, very few falling into the medium-light and light

categories.

Reputational Variables

Reputational variables include grape variety and wine region.

Grape varieties were dummied, including Sauvignon Blanc,

Chardonnay, Riesling, Syrah, Pinot Gris, Cabernet

Sauvignon, Pinot Noir, Merlot, Proprietary Blends (usually

similar to Bordeaux blends), Cabernet Franc, Viognier,

Montepulciano, Gewurztraminer, Nebbiolo, and Malbec.

Some of these were condensed for not having a large enough

share in the sample. Thus, the variables included are only

PINNOIR, SYRAH, PROPBLE, CHARD, and SAUVBLA

with all other varieties represented by OTHERVAR.

Dummy variables were created including Auckland,

Gisborne, Hawkes Bay, Wairarapa, Nelson, Waipara, Otago,

and Marlborough. These are represented by AUCKLE,

GISBOR, HAWKESB, WAIRAR, NELSON, WAIPAR,

OTAGO, and MARLB, respectively. They are the main

wine-producing regions; however, there are some specific

regions within these that have been incorporated into the

larger, inclusive regions. For instance, Central Otago is a part

of Otago, but some wines from Central Otago are specifically

labeled as such rather than just being labeled as being from

Otago. The same conditions apply to Clevedon, Kumue,

Ponui, Waiheke Island, and Matakana in Auckland, Esk

Valley in Hawkes Bay, Prophets Rock, Waitaki, Felton Road,

Surveyor Thompson, and Bannockburn in Otago, Canterbury

and Hakataramea Valley in Waipara, Solstone and

Martinborough in Wairarapa, and Moutere in Nelson. Some

of these, such as Waiheke Island, may soon be classified as

their own wine-producing regions, but currently they are still

included under the main regions listed. Again, because of

small sample sizes, NELSON and GISBOR are condensed

into OTHERREG.

Quality Variables

The main quality variable included in this study is the Parker

Rating (PARKRAT). Quality ranges are 96 to 100 which are

extraordinary wines, 90 to 95 are outstanding, 80 to 89 are

barely above average to good, 70 to 79 are average, 60 to 69

are below average, and any wine rated below 60 is

unacceptable. Table wines are not included in this study as

they are typically not of a high enough quality to be rated.

Objective Variables

Objective variables included in this study are vintage and

shelf life. Vintage (VINTAGE) describes the year a particular

wine was produced, subtracting the vintage year from 2012.

For example, if a wine was bottled in 1992, the vintage

would be 20. Shelf life (SHELF) states how many years the

wine can be stored as of 2012 before it is no longer a

desirable buy. Negative numbers for this variable indicate the

wine is past its drinking time. Not all bottles have data for

this variable. A distinction between classic and not could not

be found for New Zealand wines, and thus this variable was

excluded.

Expected Signs

A table of all expected signs can be found in Table 2 of the

Appendix. There are several variables that one would expect

to have a positive sign in relation to bottle price, including

Parker Rating and shelf life. The higher the Parker Rating,

the better the wine, and the higher the price is likely to be.

The longer the bottle will last indicates a higher quality wine

and thus would also be expected to be positive. Similarly, the

longer a wine can be stored, the more willing a buyer may be

to spend more money on it.

Many of the variables used in this study, however, have

ambiguous signs, including body type, grape variety, wine

region, and flavors. These could be either positive or negative

depending on the preferences of the consumers, which will

be reflected in the bottle price. For instance, Pinot Noir is one

of the most well-known grape varieties in New Zealand, and

therefore buyers may be more willing to pay higher prices for

a Pinot Noir than a Nebbiolo. Benfratello, et al. (2004) used

the most well-known grape variety as the omitted condition

in their study, which in this case would be Pinot Noir.

According to the literature available, there is no particular

established method for choosing the omitted condition for

Proceedings of the Pennsylvania Economic Association 155

grape variety. Therefore, I simply used OTHERVAR as the

omitted condition and left the main specific grape variety

variables in the model. I used a similar method for the region

and flavor dummy variables, using OTHERREG and

OTHERFLV as the omitted conditions. BODYLGT was the

omitted condition for body style variables, and COLORR for

wine color.

Descriptive Statistics

Complete descriptive statistics for all variables can be found

in Table 3. The most interesting statistics are as follows. The

average bottle price of New Zealand wine in this sample was

$57.08, with a maximum of $275.00 and a minimum of

$38.00. The average vintage age in this sample was 4.5 years,

with a maximum of 17 years and a minimum of 1 year. The

average shelf life of the wines included in this sample was

4.2 years, with a maximum of 13.0 years and a minimum of -

10.0 years, meaning that bottle was past its drinking prime by

approximately 10 years. In this sample, 92% of the wines

have a fruity flavor and none had an oxidized flavor. The

region from which the largest proportion of the sample

originates was Auckland with 22.5% of the sample. The

region with the smallest proportion was Nelson with 2.0%

ECONOMETRIC MODEL

This study uses ordinary least squares (OLS) regression

analysis as done in the similarly comprehensive study by

Anderson and Schamel (2003). The original hypothesized

equation is as follows:

This model represents a combination of Anderson and

Schamel (2003) and Oczkowski (2001). Table 1 exhibits the

variables included in each category in the equation above.

Anderson and Schamel (2003) particularly influenced the

form of the dependent variable (which is logged), dummy

variables for region, grape variety, flavors, body, stage, class,

and color. Unlike Anderson and Schamel (2003), this study is

cross-sectional rather than a panel, which is more similar to

Oczkowski (2001). I conduct a more focused analysis,

however, emphasizing the effects of the chosen independent

variables on only New Zealand wines, intending to update

the research for present-day wines. I also include sensorial

characteristics with the other dummy groups, unlike past

studies concerning New Zealand wines.

Econometric Issues

The first and most prevalent issue in this study was the

availability of data. Other studies have included winery

ratings or regional ratings, which were unattainable within

the constraints of this study. Furthermore, some of the bottles

used in this sample did not have data for each of the

variables, decreasing the sample size depending upon the

model. This was particularly prevalent with the body

variables. Additionally, the expectation that the stochastic

error term does not have a constant variance across

observations is confirmed by the results of White tests. The

null hypothesis of homoskedasticity is rejected for all

regressions. As a result, all regressions are corrected for

heteroskedasticity.

RESULTS

The results for all models appear in Table 4 of the Appendix.

As expected, PARKRAT remained positive and significant

across all models. Model 1 was a simple regression using

only PARKRAT in the estimation of LNPRICE. PARKRAT

is significant at the 1% level in this model, with an adjusted

R-squared of 0.037. Though this model obviously contains

omitted variable bias, it does suggest that the Parker Rating is

significantly correlated with the bottle price of New Zealand

wines. This is consistent with the findings in Anderson and

Schamel (2003). However, it clearly explains only a small

portion of the story in understanding the influences on bottle

prices for New Zealand wines. Anderson and Schamel (2003)

relied more heavily upon Parker Rating. This may have been

necessary for the combined examination of Australian and

New Zealand wines, but was less useful for New Zealand

alone for 2012 wines. Anderson and Schamel (2003) also

used winery rating and regional rating, however, which may

have an influence on the significance of the Parker Rating in

their models. The significance of the Parker Rating to New

Zealand wine prices is also consistent with the results

Oczkowski (2001). Though Oczkowski (2001) used different

sources for the various ratings he used, all ratings proved

significant in the various models. Thus, all subsequent

models contain PARKRAT.

Model 2 includes PARKRAT as well as VINTAGE and

SHELF. This functioned as the base equation for all future

models. Also as expected, the coefficients on VINTAGE and

SHELF remain positive and significant across all models

after Model 1. Both VINTAGE and SHELF are significant at

the 1% level, while PARKRAT is significant at the 5% level.

The inclusion of VINTAGE and SHELF caused the adjusted

R-squared to increase to 0.165. Oczkowski (2001) also used

variables equivalent to vintage and shelf life, which proved

significant in all models.

Each of the following models contains a different group of

dummy variables. Several models containing multiple

dummy groups were tested, but were excluded due to

multicollinearity between groups. Model 3 includes wine

regions along with the variables from Model 2. For this

model, the adjusted R-squared is 0.175. Interestingly,

Proceedings of the Pennsylvania Economic Association 156

PARKRAT’s significance falls to the 10% level, but

VINTAGE and SHELF remain significant at the 1% level. Of

the region variables, only AUCKLE and OTAGO are

statistically significant at the 10% level. Despite these low

individual significance levels, F-test results indicate that as a

group the region variables are significant to the model.

AUCKLE, MARLB, OTAGO, WAIPAR, and WAIRAR

tend to see higher bottle prices compared to OTHERREG,

while HAWKESB is correlated with lower bottle prices than

OTHERREG. Though AUCKLE and OTAGO are

statistically significant for New Zealand wine prices in 2012,

this is no indication that they would be significant for other

years. Anderson and Schamel (2003) conducted a panel

study, which showed the change in significance of the

various regions over time. They found that Auckland most

frequently was significantly correlated with higher bottle

prices, however. This indicates that Auckland tends to

produce wines for which consumers tend to be willing to pay

higher prices.

Model 4 replaces the region variables with the flavor

variables, resulting in an adjusted R-squared of 0.192.

NUTTY is significant at the 10% level, while EARTHY and

WOODY are significant at the 5% level. Though none of the

other flavors are statistically significant, as a group they are

significant to the model based on F-test results. CARAM,

EARTHY, FLORAL, FRUITY, HERBVEG, SPICY, and

WOODY are all positive, while MICROB and NUTTY are

negative. This reveals that consumers tend to pay higher

prices for the former flavors than they would for

OTHERFLV, which includes pungent and oxidized. On the

other hand, they are more willing to pay higher prices for

OTHERFLV than they are for MICROB and NUTTY. Some

flavor preferences may change over time with wine

“fashion,” while others may remain consistent over time. A

panel study delving into the changes in preferences for

particular sensorial characteristics could be a point of interest

for a future study.

Flavors are exchanged for grape varieties in Model 5, with an

adjusted R-squared of 0.206. The only variety not significant

is SUAVBLA, which is also negative in sign. This initially

seemed odd because New Zealand is known for its

Sauvignon Blanc. However, this actually makes sense if

approached from the perspective that New Zealand wineries

produce more Sauvignon Blanc than most other varieties

(probably because of its popularity), which would eventually

trigger downward pressure on the bottle price in comparison

with OTHERVAR. CHARD is also negative, but it is

significant. Sauvignon Blanc and Chardonnay also had a

consistently negative signs in Anderson and Schamel (2003).

This suggests that consumers are inclined to pay lower prices

for these two white New Zealand wines compared to the

wines included in OTHERVAR. PROPBLE is positive and

significant at the 10% level, and PINNOIR and SYRAH are

positive and significant at the 1% level. This indicates that

consumers are willing to pay higher prices for these three red

New Zealand wines than for those other wines included in

OTHERVAR. Grape varieties changed in significance over

time Anderson and Schamel (2003), but Pinot Noir and Syrah

were significant at the 5% level in most years.

Model 6 contains the variables for body style, including

BODYFULL and BODYMED in the regression with

BODYLGT as the omitted condition. With an adjusted R-

squared of 0.223, BODYFULL is significant at the 5% level,

while BODYMED is insignificant in the model. Though

there is obviously some level of omitted variable bias, F-test

results indicate that the body style variables are jointly

significant to the model overall and are thus included. The

coefficients of BODYFULL and BODYMED are both

negative, suggesting that consumers will pay higher prices

for wines with light bodies than for full or medium-bodied

wines. The last model, Model 7, includes only COLORW

with COLORR being the omitted condition. With an adjusted

R-squared of 0.200, COLORW is negative. This suggests that

consumers tend to pay higher prices for red wines rather than

white wines.

Adjusted R2 values ranged from 0.23 to 0.69 in the literature,

the panel studies typically yielding higher values than the

cross-sectional studies. Since this study is a cross-sectional

analysis, the adjusted R2 values obtained of roughly 0.17 to

0.22 fall closer to the lower end of this range. When

considered in groups rather than individual characteristics,

the flavor variables are less statistically significant compared

to region, grape variety, body, and color. This may be due to

the fact that flavors cannot be experienced until after the

bottle is already purchased. Combris et al. (1997) argued that

most consumers base their purchases upon the information

found on the bottle. They must then rely heavily upon

reputational, quality, and objective variables to attempt to

make wise purchasing decisions. Flavors, on the other hand,

are used to determine the rating of a wine such as those by

wine tasters like Robert Parker. Therefore, the flavor

variables are significantly correlated with wine prices, but

other characteristics take precedence in determining bottle

price because they are tangible at the time of purchase.

CONCLUSION

This study suggests that sensorial, objective, reputational,

and quality variables are all correlated with bottle price. Most

of the variables in this study were included because they

were found to be important the existing literature and were

necessary to control for different conditions that could

influence individual bottle prices, even if they do not affect

all bottles. This study found that while Parker Rating is

significant in determining bottle prices for New Zealand

wines, it explains only a small portion of the variation in

prices. Vintage, shelf life, region, flavors, grape variety,

body, and color are all significantly correlated with New

Proceedings of the Pennsylvania Economic Association 157

Zealand wine prices. The particular regions, varieties, body

styles, flavors, and colors of significance depend upon the

sample taken and year.

In the case of the time period considered in this study,

Auckland and Otago had a statistically significant and

positive correlation with higher bottle prices compared to

Nelson and Gisborne. Similarly, earthy, nutty, and woody

flavors were significantly correlated with bottle price,

whether positive or negative, in relation to oxidized and

pungent flavors. Pinot Noir, Proprietary Blends, and Syrah

were significantly correlated with higher prices as opposed to

the many varieties included in OTHERVAR, while

Chardonnay and Sauvignon Blanc were negatively

correlated. For instance, a sturdy Syrah would most likely

fetch a higher price than a more temperamental Nebbiolo,

which usually requires significant periods of aging. In the

case of body style, light-bodied wines were correlated with

higher prices than those for medium and full body styles.

White wines also seemed to receive lower prices than red

wines in this time period. The significance of the particular

variables within the dummy groups may change over time or

with different samples.

Implications of Results

Wineries in regions that tend to receive lower prices for their

wines can try to adopt some of the characteristics found in

the regions that tend to command higher prices to increase

their own profits. These results indicate the flavors, body

styles, colors, and grape varieties for which consumers seem

to be the most willing to pay more money. Wine producers

could also try to increase the shelf life of their wines, which

would subsequently increase the probability of being able to

sell older vintages. Any of these methods would be viable

options for increasing profits and improving customer

satisfaction. Combris et al. (1997) indicated that consumers

frequently choose their purchases based upon information

found directly on the bottle. Therefore, wineries could use the

information in this study to better understand what

information they should include on their bottle labels and

descriptions of their wines. Essentially, while the Parker

Rating is important, it is not the be-all-end-all for New

Zealand wine sellers when setting bottle prices. They would

do better focusing on producing wines that last for long

periods of time with flavors consumers prefer as opposed to

catering to wine tasters’ preferences.

(I would like to thank Dr. James Jozefowicz, for all the time

and effort you put in to helping me with my paper. You will

never know how much your constant support is valued and

appreciated. Thank you also to Dr. Yaya Sissoko, my

discussant, Brian Foster-Pegg, for your input, and my former

partner, Grey Berrier, for the moral support and answering

random questions at all hours of the night)

Proceedings of the Pennsylvania Economic Association 158

APPENDIX

Figure 1 Sechrist (2012)’s Map of New Zealand Wine Regions

Table 1 Parker Rating Ranges and Descriptions

Rating Descriptions

96-100 An extraordinary wine of profound and complex character displaying all the attributes expected

of a classic wine of its variety. Wines of this caliber are worth a special effort to find, purchase,

and consume.

90-95 An outstanding wine of exceptional complexity and character. In short, these are terrific wines.

80-89 A barely above average to very good wine displaying various degrees of finesse and flavor as

well as character with no noticeable flaws.

70-79 An average wine with little distinction except that it is a soundly made. In essence, a

straightforward, innocuous wine.

60-69 A below average wine containing noticeable deficiencies, such as excessive acidity and/or tannin,

an absence of flavor, or possibly dirty aromas or flavors.

50-59 A wine deemed to be unacceptable.

Proceedings of the Pennsylvania Economic Association 159

Table 2 Expected Signs and Definitions for all Variables

Dependent Variable: Logged Bottle Price of New Zealand Wines

Independent Variable Definition Expected Sign

PARKRAT Parker Rating from eRobertParker.com +

Objective Variables

VINTAGE 2012 minus the year the bottle was produced +

SHELF Shelf life +

Regions

AUCKLE Dummy variable for the region of Auckland +/-

HAWKESB Dummy variable for the region of Hawkes Bay +

MARLB Dummy variable for the region of Marlborough +

OTAGO Dummy variable for the region of Otago +/-

WAIPAR Dummy variable for the region of Waipara +/-

WAIRAR Dummy variable for the region of Wairarapa +/-

OTHERREG†

Omitted condition for region variables – includes Nelson and Gisborne N/A

Flavors

CARAM Dummy variable for the flavor caramel +/-

EARTHY Dummy variable for the flavor earthy +/-

FLORAL Dummy variable for the flavor floral +/-

FRUITY Dummy variable for the flavor fruity +/-

HERBVEG Dummy variable for the flavor herbaceous/vegetal +/-

MICROB Dummy variable for the flavor microbiological +/-

NUTTY Dummy variable for the flavor nutty +/-

SPICY Dummy variable for the flavor spicy +/-

WOODY Dummy variable for the flavor woody +/-

OTHERFLV† Omitted condition for flavor variables – includes oxidized and pungent N/A

Grape Varieties

CHARD Dummy variable for the grape variety Chardonnay +/-

PINNOIR Dummy variable for the grape variety Pinot Noir +/-

PROPBLE Dummy variable for the grape variety Proprietary Blend +/-

SAUVBLA Dummy variable for the grape variety Sauvignon Blanc +/-

SYRAH Dummy variable for the grape variety Syrah +/-

OTHERVAR† Omitted condition for grape variety variables N/A

Body Style

BODYFULL Dummy variable for full body style +/-

BODYMED Dummy variable for medium body style +/-

BODYLGT† Omitted condition for body style variables N/A

Colors

COLORW Dummy variable for white wine +/-

COLORR† Omitted condition for color variables N/A

† indicates the omitted conditions for each dummy group, which were dropped from the regressions

N/A: Not Applicable because they are the omitted conditions

Proceedings of the Pennsylvania Economic Association 160

Table 3 Descriptive Statistics

Mean Median Maximum Minimum Std. Dev. Observations

PRICE 57.08333 50.00000 275.0000 38.00000 24.93101 492

PARKRAT 90.54158 91.00000 97.00000 82.00000 2.067164 493

VINTAGE 4.488844 4.000000 17.00000 1.000000 2.643903 493

SHELF 4.241279 4.000000 13.00000 -10.00000 3.044363 344

AUCKLE 0.083164 0.000000 1.000000 0.000000 0.276411 493

HAWKESB 0.225152 0.000000 1.000000 0.000000 0.418107 493

MARLB 0.199187 0.000000 1.000000 0.000000 0.399795 492

OTAGO 0.249493 0.000000 1.000000 0.000000 0.433159 493

WAIPAR 0.064909 0.000000 1.000000 0.000000 0.246615 493

WAIRAR 0.129817 0.000000 1.000000 0.000000 0.336444 493

OTHERREG 0.048682 0.000000 1.000000 0.000000 0.215420 493

CARAM 0.237323 0.000000 1.000000 0.000000 0.425874 493

EARTHY 0.375254 0.000000 1.000000 0.000000 0.484680 493

FLORAL 0.225152 0.000000 1.000000 0.000000 0.418107 493

FRUITY 0.929006 0.000000 1.000000 0.000000 0.257076 493

HERBVEG 0.326572 0.000000 1.000000 0.000000 0.469436 493

MICROB 0.267748 0.000000 1.000000 0.000000 0.443236 493

NUTTY 0.260437 0.000000 1.000000 0.000000 0.260437 493

SPICY 0.432049 0.000000 1.000000 0.000000 0.495864 493

WOODY 0.423935 0.000000 1.000000 0.000000 0.494682 493

OTHERFLV 0.139959 0.000000 1.000000 0.000000 0.353101 493

CHARD 0.107505 0.000000 1.000000 0.000000 0.310069 493

PINNOIR 0.519270 0.000000 1.000000 0.000000 0.500136 493

PROPBLE 0.152130 0.000000 1.000000 0.000000 0.359511 493

SAUVBLA 0.029398 0.000000 1.000000 0.000000 0.166275 493

SYRAH 0.093306 0.000000 1.000000 0.000000 0.291157 493

OTHERVAR 0.093306 0.000000 1.000000 0.000000 0.291157 493

BODYFULL 0.393258 0.000000 1.000000 0.000000 0.489161 356

BODYMED 0.568627 0.000000 1.000000 0.000000 0.495963 357

BODYLGT 0.036415 0.000000 1.000000 0.000000 0.187582 357

COLORW 0.196755 0.000000 1.000000 0.000000 0.397949 493

COLORR 0.801217 1.000000 1.000000 0.000000 0.399490 493

Proceedings of the Pennsylvania Economic Association 161

Table 4 OLS Regressions of Hedonic Price Model

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

CONSTANT 1.212

(1.753)

1.250

(1.351)

1.627

(1.626)

1.353

(1.396)

0.677

(0.686)

0.545

(0.572)

0.612

(0.649)

PARKRAT 0.031***

(3.999)

0.027** (2.592) 0.0221* (1.954) 0.024** (2.232) 0.033***

(2.933)

0.036***

(3.264)

0.035***

(3.273)

VINTAGE 0.024***

(2.751)

0.0260***

(2.798)

0.032***

(3.388)

0.019** (2.089) 0.026***

(2.657)

0.017* (1.936)

SHELF 0.038***

(4.269)

0.040***

(4.221)

0.037***

(4.082)

0.031***

(3.305)

0.044***

(4.266)

0.030***

(3.227)

AUCKLE 0.176* (1.881)

HAWKESB -0.068 (0.102)

MARLB 0.068

(1.085)

OTAGO 0.113* (1.867)

WAIPAR 0.029

(0.416)

WAIRAR 0.091

(1.355)

CARAM 0.048

(1.311)

EARTHY 0.088** (2.565)

FLORAL 0.059

(1.548)

FRUITY 0.045

(0.411)

HERBVEG 0.048

(1.403)

MICROB -0.001

(0.033)

NUTTY -0.106* (1.932)

SPICY 0.049 (1.590)

WOODY 0.070** (2.234)

CHARD -0.109**

(2.522)

PINNOIR 0.100***

(2.840)

PROPBLE 0.103* (1.798)

SAUVBLA -0.027

(0.327)

SYRAH 0.148*** (2.626)

BODYFULL -0.151**

(2.107)

BODYMED -0.038

(0.525)

COLORW -0.160***

(4.893)

R-Squared 0.039 0.172 0.197 0.220 0.224 0.236 0.209

Adjusted

R-Squared 0.037 0.165 0.175 0.192 0.206 0.223 0.200

F-statistic for

Group Joint

Significance

NA NA 5.831 2.257 4.448 36.708 14.974

N 492 343 342 343 343 296 343

Note: Calculated t-statistics in parentheses are based on White heteroskedasticity-consistent standard errors.

* = significant at the 10% level

** = significant at the 5% level

*** = significant at the 1% level

Proceedings of the Pennsylvania Economic Association 162

References available upon request from Angela M. Rowland.

Proceedings of the Pennsylvania Economic Association 163

Author Index

Alt, Ashley M. 16

Armstrong, Thomas O. 28

Balagyozyan, Aram 34

Byun, Chong Hyun C. 41

D’Angelo, Dana 48

Epstein, Susan 48

Gargone, David 78

Ghosh, Soma 52

Giannikos, Christos 34

Hussain, Riaz 123

Kallianiotis, Ioannis N. 54

Kara, Orhan 67

Kearney, Timothy F. 78

Kohli, Tanu 87

Kurre, James A. 97

Linn III, Johnnie B. 109

Ma, Zhen 114

Mansour, Stephen M. 123, 129

McCollough, John 130

Mona, Kyoko 34

Nugent, David 137

Oberkofler, Daniel R. 16

Pezak, Logyn 142

Rowland, Angela M. 150

Sebastianelli, Rose 142

Tesfu, Solomon T. 52