consumer expenditure survey anthology, 2003color, icons, highlighting, and shading were all...

Consumer ExpenditureSurvey Anthology, 2003U.S. Department of LaborBureau of Labor Statistics

September 2003

Report 967

iii

Preface

This is the first in a series of reports presenting bothanalytical articles that use data from the Bureau ofLabor Statistics (BLS) Consumer Expenditure (CE)

Survey and methodological articles that discuss ongoingresearch and issues pertaining to the survey. In the past, theCE Survey Division published a biennial report that includedanalytical articles, standard tables of the most recent CE Sur-vey data, a discussion of expenditure changes, and a de-scription of the survey and its methods. The most recent ofthese was Consumer Expenditure Survey, 1998–99 Report955, published in November 2001. The biennial report will bereplaced by two separate biennial reports that will be pub-lished in alternating years. One will continue the practice ofpublishing tables with recent survey data, a brief discussionof recent changes in expenditures, and a description of thesurvey and its methods. The first of this type of report isConsumer Expenditure Survey, 2000–2001, Report 969, pub-lished in September 2003. The other, of which this is the first,includes both methodological and analytical articles. Themethodological articles are intended to provide data userswith greater insight into both ongoing improvements in thesurvey and issues that are faced in collecting, processing,and publishing information from such a complex survey. Theanalytical articles furnish information on topics of interestpertaining to CE Survey data.

The CE Survey program provides a continuous and com-prehensive flow of data on the buying habits of Americanconsumers for use in a variety of economic analyses and insupport of periodic revisions of the Consumer Price Index.BLS makes data available in news releases, reports, bulle-tins, and articles in the Monthly Labor Review, as well as onCD-ROMs and on the Internet.

This report was prepared in the Office of Prices and Liv-

ing Conditions (OPLC), Division of Consumer ExpenditureSurvey (DCES), under the general direction of SteveHenderson, Chief of the Branch of Information and Analy-sis, and was produced and edited by John M. Rogers, Sec-tion Chief. Articles on research and methodology were con-tributed by Sioux Groves, Chief of the DCES, Jeff Blaha andSally Reyes-Morales of the Division of Price Statistical Meth-ods, Geoffrey Paulin of the Branch of Information and Analy-sis, Linda Stinson of the Office of Survey Methods Research(OSMR), and Nhien To and Jeanette Davis of the Branch ofResearch and Program Development. Analytical articles werecontributed by Abby Duly, George Janini, Eric Keil, LauraPaszkiewicz, and Geoffrey Paulin of the Branch of Informa-tion and Analysis and Neil Tseng of the Branch of Produc-tion and Control.

The material that follows is divided into two sections:section 1 contains articles on survey research and methodol-ogy, and section 2 presents analyses of topics of interest basedon CE Survey data. An appendix includes a general descrip-tion of the survey and its methods and a glossary of terms.

Current and historical CE Survey tables classified by stan-dard demographic variables are available at the BLS Internetsite http://www.bls.gov/cex. Other survey information, in-cluding answers to frequently asked questions, a glossaryof terms, order forms for survey products, and Monthly La-bor Review and other research articles, also is available onthe Internet.

Sensory-impaired individuals may obtain information onthis publications upon request (voice phone: (202) 691–5200,Federal Relay Service: 1–800–877–8339). The material pre-sented is in the public domain and, with appropriate credit,may be reproduced without permission. For further informa-tion, call (202) 691–6900.

v

Contents

Part I. Survey Research and Methodology ...................................................................................................................... 1Creating a “User-Friendly” Expenditure Diary ...................................................................................................... 3Computer-Assisted Personal Interviewing for the Consumer Expenditure Interview Survey ...................... 18Introducing Brackets: Quality in the Consumer Expenditure Interview Survey .............................................. 21Characteristics of Complete and Intermittent Responders in the Consumer Expenditure Interview Survey ..................................................................................................................................................... 25Standard Errors in the Consumer Expenditure Survey ........................................................................................ 30

Part II. Analyses Using Survey Data .................................................................................................................................. 33Consumer Spending for Necessities ...................................................................................................................... 35Consumer Expenditures for Alcohol in 2000 ......................................................................................................... 39The Costs and Demographics of Vehicle Acquisition ......................................................................................... 61Out-of-Pocket Expenditures by Consumer Units with Private Health Insurance ............................................ 67Expenditures on Entertainment ............................................................................................................................... 73Travel Expenditures in 2000 ..................................................................................................................................... 79

Appendix A. Description of the Consumer Expenditure Survey .................................................................................... 83

Page

Consumer Expenditure Survey Anthology, 2003 1

Part I.

Survey Research and Methodology


Creating a “User-Friendly”Expenditure Diary

Linda Stinson is a research psychologist for-merly in the Office of Survey Methods Re-search, Bureau of Labor Statistics.

Nhien To is an economist in the Branch ofResearch and Program Development, Divi-sion of Consumer Expenditure Surveys, Bu-reau of Labor Statistics.

Jeanette Davis is a senior economist in theBranch of Research and Program Develop-ment, Division of Consumer Expenditure Sur-veys, Bureau of Labor Statistics.

LINDA STINSONNHIEN TOJEANETTE DAVIS

Interest in American expenditureshas a long history dating back to the late 1800s, when the Bureau of

Labor Statistics (BLS) first looked atthe economic welfare of our early immi-grants. Today, BLS is mandated to re-port detailed information on all theways in which Americans spend theirmoney. The Consumer Expenditure Di-ary (CED, Diary) is the instrument usedto collect information on the many pur-chases made each week by sampledhouseholds.

When it comes to reporting detailedexpenditure information, not all pur-chases are equally easy to rememberand record. Some expenditures, suchas daily busfare, are often part of a“work commute” mental script and maybe readily recalled. Other purchases, likesodas and snacks from vending ma-chines, tend to be more mundane, bur-ied within the concerns of daily activi-ties, and more easily overlooked. Thediary mode of data collection has longbeen recognized as an especially use-ful tool for collecting daily records ofthese types of frequent, low-saliencepurchases before they are forgotten.The diary also makes it possible to col-lect followup details on purchases thatcan be used to produce the weights forthe Consumer Price Index.1 Such infor-mation would be difficult, if not impos-sible, to collect accurately withoutsome means of recording the purchasesduring the week as they occur.

Over time, numerous economic re-searchers have adopted a diary ap-proach to track household consump-tion, gauge reactions to new productsappearing on the market, and observesocial trends. Through their work, ithas become abundantly clear that dia-ries are useful data collection tools. How-ever, in order to attract and keep respon-dents, a diary must be user friendly andactively engage the respondents’ inter-est in the data-reporting task.

Developing a BLS diaryOver the years, BLS created variousexpenditure diaries with the hope thatthey would produce high responserates and accurate estimates. But evi-dence from numerous research stud-ies, expert reviews, and the reports ofinterviewers and respondents alike hasindicated that these diaries were notparticularly user friendly. From the per-spective of the respondent, the mainproblem with the current CED Diary isthat it is difficult to navigate; neitherits logic nor its structure is apparent.(See exhibit 1.) The respondent mustnavigate both vertically and horizon-tally and must inspect every page thor-

1 For example, reports for grocery itemsneed to include details about the type of pack-aging and whether the item is fresh or fro-zen. Detailed information on clothing in-cludes the gender and age range of the re-cipient. Meals away from home havefollowup details about purchase of alcoholicbeverages.

4 Consumer Expenditure Survey Anthology, 2003

oughly in order to determine how toproceed. In addition, respondents havereported that Diary instructions are noteasy to read or follow. (See exhibit 2.)For example, respondents do not un-derstand some of the words, such as“consumer unit,” used by BLS. Like-wise, the pages used as examples inthe current diary have been reportedto be somewhat overwhelming and,worse, may contribute to, rather thanameliorate, respondents’ confusion.Finally, the large size and landscapedlayout (as opposed to the more typicalbook format) makes it difficult for somerespondents to read and use the diary.

In response to these concerns, theBranch for Research and Program De-velopment in the BLS Division of Con-sumer Expenditure Surveys charteredthe Redesign and Analysis of Diary(RAD) team to develop a more attrac-tive and appealing CED that would beless burdensome to complete. The firststep in the process was to identify themany graphical features that might beused to guide respondents through adiary. Color, icons, highlighting, andshading were all considered as toolsthat could clarify the respondents’ taskand help them report information fullyand accurately.

Working with a contractor, the RADteam developed three prototype diariesthat were ready for evaluation by thespring of 2001. The prototypes weredistinguished by the color of their cov-ers, their internal structure, and theirlength.

Prototype 1 (the peach diary), alsoentitled “Your Daily Notebook,” wasidentical to the current BLS productiondiary, but was reformatted with icons,color, and a portrait, booklike orienta-tion. It was divided into seven days,within which were five major expendi-ture categories. Within each categorywere several subcategories identifyingsubsets of expenditures that should berecorded. Because of its peach-coloredcover, Prototype 1 was referred to asthe peach “current” diary. (See exhibit3.) The copious subcategorization ofexpenditures rendered the peach diarythe longest of the three, at 144 pages.

Prototype 2 (the yellow diary), en-

titled “Track How You Spend YourMoney; also was divided into 7 days.As with the peach diary, all the expen-diture categories and subcategorieswere repeated every day, with tabs in-dicating where each day began. Ex-penditures were recorded on the dayof purchase and under the correct de-scriptive category. The difference be-tween this diary and the peach “Cur-rent” diary was that in the former therewere fewer subcategories within themajor expenditure categories. Becauseof its yellow cover, Prototype 2 wasreferred to as the yellow “day” diary.(See exhibit 4.) With fewer subcatego-ries, it was 132 pages long.

Prototype 3 (the teal diary), entitled“Your Daily Notebook,” was dividedinto four major expenditure categories,instead of the days of the week. Re-spondents recorded purchases underthe correct expenditure category, alongwith the day on which they were pur-chased. Because of its teal cover, Pro-totype 3 was referred to as the teal“parts” diary. (See exhibit 5.) By elimi-nating the repetition of the 7 days, itwas the shortest of the prototypes, atonly 36 pages.

The first step in the process ofevaluating the strengths and weak-nesses of each version of the diary wasto submit all three to knowledgeableBLS staff for review.2 The commentsgenerated by this review processranged from the correction of typos tomore profound concerns about miss-ing data elements and the quality ofthe data. The initial process of internalreview resulted in the elimination of thepeach “current” diary, which was al-most universally disliked because of itslength and complexity. This left theRAD team with two viable prototypes.

Round 1: Evaluation of Proto-types 2 and 3Beginning in June, 2001, copies of theyellow and teal prototype diaries weredistributed to 15 U.S. Census Bureau

interviewers known as field represen-tatives, 90 BLS staff and summer in-terns from the Office of Prices and Liv-ing Conditions and the Office ofSurvey Methods Research, and 11 man-agers and staff from the Census Bu-reau. The prototype diaries were ran-domly assigned, with roughly half ofthe participants receiving a yellow“day” diary and half receiving a teal“parts” diary.

All participants were asked to keepthe assigned diary for their entire con-sumer unit for 1 week. In addition, thefield representatives completed a shortquestionnaire developed by the Cen-sus Bureau, which they mailed to theRAD team at BLS, along with com-ments written in the margins of theirdiaries. All other BLS and Census Bu-reau participants took part in discus-sion groups to talk about their experi-ences using the diary, to identifypotential problems, and to brainstormideas for improvements.

In total, the RAD team conducted12 discussion sessions with 6 to 13participants per session and a small-group interview with three Census Bu-reau managers. In each of the groups,there was a mix of participants, some ofwhom kept the yellow diary, and some,the teal diary. In this way, participantswere able to discuss the relative meritsof the two versions.

The strategy of choosing knowl-edgeable BLS and Census Bureau staffas participants in the first round ofstudy was selected for many reasons.First, it was a way to generate interestin the new diary by disseminating in-formation about proposed changes.Second, it provided BLS subject-mat-ter experts and Census Bureau fieldstaff with an opportunity to commenton the prototypes and to help deter-mine the design of the new diary. Third,it was a chance to draw upon the ex-pertise of those who know what datathe diary should collect and to critiquethe prototypes in light of the estimatesthey would produce.

While each discussion group hadits own unique flavor and focus of in-terest, the comments made throughoutwere remarkably similar. Unanimity on

2 The first rounds of internal BLS evalu-ation included reviews by staff in the Con-sumer Expenditure Program, the ConsumerPrice Index Program, and the Office of Sur-vey Methods Research.


certain key points was highly reassur-ing and made it relatively easy to deci-pher the main themes conveyed in manydifferent ways.

As regards the yellow “day” diary,participants reported that having thediary divided by day of the weekhelped them to recall their purchases.However, at 132 pages, this version wasstill bulky, repetitive, and somewhatdifficult to navigate and use.

The teal “parts” diary was moreproblematic. While it was considerablyshorter and easier to manage, partici-pants reported that they missed theday-of-purchase structure in attempt-ing to recall their expenses. Apparently,these memories were not classified in-ternally by expenditure category, butrather were associated with the activi-ties of the day of the week.

The main results from the first roundof study found their fullest expressionin the following list of recommenda-tions generated by the participants:

• Clarify the instructions, record-ing rules, and definitions for bothprototype diaries; provide a setof “frequently asked questions”(FAQs).

• Eliminate the subcategories andsimplify the recording task in theyellow “day” diary.

• Expand the examples and avoidneedless repetition of examplesin both diaries; use the pages withexamples to convey as much newinformation as possible.

• Organize the teal “parts” diary byday of the week, as done in theyellow “day” diary.

• Make the yellow “day” diary ascompact as possible, with a lengthsimilar to that of the 36-page teal“parts” diary.

• Provide a “mental map”—an over-view of all the major categories—at the beginning of the diary sothat respondents do not have tostudy the entire booklet in order

to understand what lies ahead.

• Tell respondents about any ex-penses that should not be record-ed.

• Use check boxes to collect fol-lowup details, such as the typeof packaging for groceries or thetype of meal eaten away fromhome.

• Make the diary look easy and userfriendly, yet, at the same time,maintain a professional and of-ficial quality.

While these recommendations weredirected specifically toward the devel-opment of a new prototype, other com-ments surfaced that addressed theoverall task of keeping a diary:

1. Keeping a diary is a difficult memorytask .

• It is often difficult to remember torecord expenditures in the diary.

• If expenditures are not recordedclose to the time of purchase, theygenerally become increasinglydifficult to report accurately.

• If a diary is not portable, it is some-times difficult to remember whatwas purchased and what the pricewas by the time one returns home.

2. Reporting for other people is diffi-cult.

• Family members other than the re-spondent are less diligent abouttracking their expenses and repor-ting them than the respondent is.

• Family members other than the re-spondent may become irritatedand annoyed when asked abouttheir spending.

• Adolescents are often uncomfort-able and uncooperative about re-porting their expenditures to theirparents.

• Household members not directlyinstructed by the FR tend to make

more reporting errors.

3. Mathematical calculations are dif-ficult.

• It is often difficult to computeprices (with or without sales tax),even with the aid of a receipt.

• Many respondents are unable tofigure out the price of a purchaseif a receipt for that purchase doesnot clearly specify discountedcoupon amounts and sale prices.

• Rebates also are difficult to com-pute and record .

Taking into account all of this infor-mation, the RAD team turned to expen-diture diaries from other countries forideas on how to apply what waslearned. Many international diaries hadappealing designs, but the diary usedby the Household Budget Survey Pro-gram from the United Kingdom seemedto fit most closely the needs describedby our study participants and an-swered many of their objections. TheU.K. diary included check-box-stylecolumns for followup details, a day-of-the-week structure with only five ma-jor categories each day, and an attrac-tive, yet professional-looking, design.Consequently, the RAD team designeda new “Prototype 4” diary in the samevein as the one from the United King-dom,3 but incorporating additional ben-eficial features specified by BLS par-ticipants. (See exhibit 6.) For example,Prototype 4 included a “mental map”at the beginning of the diary, explain-ing its overall structure (exhibit 7), aswell as expanded example pages (ex-hibit 8) and a series of FAQs address-ing the most common recording diffi-culties that arose during the study(exhibit 9). Among the last were the fol-lowing:

• How detailed should my descrip-tions be?

3 The major categories in BLS Prototype4 are (a) “food and drinks from grocery andother stores,” (b) “catered events and mealplans,” (c) “food and drinks from groceryand other stores,” (d) “clothing, shoes, jew-elry, and accessories,” and (e) “all other prod-ucts, services, and expenses.”


• How should I record multiple pur-chases?

• How should I record prepayments,such as a subway fare card?

• How should I record credit cardpurchases?

• Should I record automatic deduc-tions taken from my paycheck orbank account?

• Should I record typical monthlybills?

• What should I do when I use cou-pons, discount cards, or loyaltycards?

• Can I just give you receipts in-stead of writing the informationdown?

• How should I record items if I don’tknow whether they include tax?

• What if I make a contribution ora charitable donation?

• What about gift certificates or giftcards?

• What do I do about returns andexchanges?

• Should I record subsidized andreimbursed expenses?

• What should I do about shippingand handling costs?

• What’s the difference between aconcession stand and a mobilevendor?

Round 2: Evaluation of Proto-type 4Even though Prototype 4 was devel-oped from information gathered duringthe first round of study, the new de-sign still needed to be evaluated to iden-tify both its strengths and weaknesses.A five-pronged strategy was formula-ted for a second round of study:

• Eight diaries were posted through-out the Division of Consumer witha request for review and comment.

• Fifteen diaries were mailed to the

same Census Bureau FRs whoparticipated from the first roundof study, along with a short ques-tionnaire to target key questionsof interest.

• Fourteen diaries were distributedto a subgroup of BLS staff whoparticipated in the first round ofstudy, so that they could partici-pate in another 2-hour reviewsession comparing the prototypes.

• Fourteen diaries were distributedto staff of the Office of Pricesand Living Conditions and theOffice of Survey Methods Re-search who had not participatedin the first study, so that theycould record their expendituresfor a week and participate in aninterview.

• Twenty diaries were distributedto members of the public, so thatthey could record their expendi-tures for a week and participatein an interview.

During the course of the study, theparticipants mentioned several featuresof the new diary that they especiallyliked and found helpful: (a) The divi-sion of the diary into days of the week,(b) the book’s graphical design and lay-out, (c) the FAQs, (d) the lists of prod-ucts and services used as exampleswithin each major category, and (e) thenew example pages with more sampleentries and information boxes used tohighlight reporting details.

Participants also identified conceptsand instructions that still needed to beclarified:

1. Some participants remained un-sure how to record multiple pur-chases of the same item (for ex-ample, five cartons of yogurt). Toresolve this uncertainty, an ad-ditional FAQ was included: “Howshould I record multiple quanti-ties?”

2. In keeping with the requirementsof the Consumer Price Index, re-spondents were told in the in-

structions not to record expensesincurred when they were awayovernight. However, almost ev-ery participant in the study sup-plied a different interpretation ofwhat being “away overnight”meant. To standardize reports, itwas recommended that this in-struction be clarified and high-lighted in interviewer trainingsessions.

3. The diaries instructed respon-dents to record each meal thatwas eaten as “Food & Drinksfrom Food Service Places” as ei-ther “breakfast, lunch, dinner, orsnack/other.” However, only 72percent of the meals from foodservice places recorded in Pro-totype 2 and Prototype 3 duringround 1 of the study specifiedany one of the four types of mealslisted. Similarly, low percentagealso has been cited as one ofthe flaws of the current CED.One goal of the redesign pro-ject was to reduce the amountof information, including thenumber of records having to dowith meals, that needed to be im-puted because of missing data.Because this same error occurr-ed in a number of diaries kept byCensus Bureau field representa-tives, it was decided that theplace to begin would be with im-proved interviewer training. Inaddition, Prototype 4 was rede-signed to include check boxesfor “breakfast, lunch, dinner, orsnack/other” in order to stan-dardize reporting and reduce theinformation burden on respon-dents. (See exhibit 10.)

4. The Consumer Price Index pro-gram requires additional infor-mation about grocery purchases,including whether the items arefresh, frozen, bottled, canned, orother. An ever-increasing vari-ety of types of packaging, how-ever, makes these distinctionsdifficult to describe and burden-some to use. Many of the par-


ticipants in the study requested moreclarification of these distinctions, andit became clear that two separate ques-tions had become intertwined in theminds of the respondents:

Question 1—• How is the food packaged? That

is, does it come in a can, a bottle,or some other type of packaging?

Question 2—• Is the food fresh, frozen, or in

some other condition?

To make explicit the twin possibilitiesthat fresh food may be packaged (forexample, fresh tomatoes may bewrapped in cellophane) and frozen foodmay be canned (for example, frozen or-ange juice may be sold in a can), thetwo followup questions were placedinto two separate columns together withcheckboxes. (See exhibit 6.)

These and other observartions col-lected during the evaluation phase ofround 2 of the study translated intomany small ideas for correcting minorflaws and tiny oversights—the tradi-tional “tweaking.” The overwhelmingmessage, however, was that Prototype4 is a user-friendly, attractive, and pro-fessional-looking data collection instru-ment.

Next stepsThe final steps in the creation of theuser-friendly expenditure diary involve

• transforming Prototype 4 into animage-scannable document ac-cording to Census Bureau speci-fications,

• updating interviewer training tomirror design changes in the di-ary, and

• conducting a field test to assessthe effect of changes to the diary.

Producing an image scannable docu-ment. Because the Census Bureau hasupdated its system of managing andprocessing paper forms, it is now pos-

sible to move away from the old proce-dure of using microfiche to preservedocuments. The goal is to produce pa-per forms, including diaries, that canbe scanned into an electronic image.Data would be keyed directly from thecomputer image, which would alsoserve as the archived document, replac-ing microfiche.

In order to meet the demands of thisautomated process, the user-friendlydiary must also be converted into a pro-cessing-friendly document. In otherwords, the final formatted diary mustfit the color, font, and size constraintsof the processing system’s specifica-tions. This work has been undertakenby the Census Bureau’s Forms DesignOffice.

Updating interviewer training. As thenew diary prototypes were being de-veloped, it became apparent that cer-tain aspects of the diary-keeping taskneeded more emphasis during inter-viewer training. For instance, BLS sug-gested that interviewer training neededto include more explanations and prac-tice (1) identifying which “overnight”expenses should not be recorded, (2)specifying the different types of meals,and (3) explaining why the diary has aday-of-the-week structure, but the ad-ditional overflow pages do not.

Also, because many of the diary’snew design features would be unfa-miliar to the interviewers, a new train-ing manual and procedures for bothself-study and classroom study neededto be developed. Among the new fea-tures that required instructions werethe following:

• FAQs• example pages with information

boxes• check boxes• pockets for receipts• a daily reminder list

In addition, because the new diarywill incorporate a computerized intro-ductory segment to collect the house-

hold demographic details, new trainingon the computer will be required.

Conducting a field test. In September2002, a field test was scheduled to as-sess the feasibility of using the newuser-friendly diary and to evaluate theeffects upon estimates and responserates. The redesigned diary will beplaced in nine census regions for 4months; it is anticipated that 1,600 com-pleted diaries will be collected. Thesediaries will be analyzed and comparedwith those obtained from a controlgroup, as well as with the regularly pro-duced diaries.

The four main goals of the field testare as follows:

• to determine whether the newuser-friendly diary yields higherresponse rates than those gen-erated by the current productiondiary;

• to test whether there is a statisti-cally significant difference be-tween the estimates produced bythe new diary, and those obtain-tained from the curent produc-tion diary;

• to evaluate the user friendlinessof the new diary in terms of theburden it places on respondents(for example, the length of timethe respondent needs to com-plete the diary and the difficultyrespondents experience in com-pleting it); and

• to test the operation of the com-puterized segments of the datacollection and operational con-trol processes.

Only at the end of these final stepswill we know whether BLS has, in fact,created a user-friendly diary that is atthe same time “processing friendly,”“image friendly,” and “data qualityfriendly.” If the final verdict is affirma-tive, the new user-friendly diary will beimplemented in 2004.


Exhibit 1: Navigation problems in the former BLS expenditure diary


Exhibit 2: Instructions from the current BLS diary


Exhibit 3: The peach “Current” diary


Exhibit 4: The yellow “Day” diary


Exhibit 5: The teal “Parts” diary


Exhibit 6: Prototytpe 4


Exhibit 7: The “mental map” in Prototype 4


Exhibit 8: An example page in Prototype 4


Exhibit 9: The “Frequently Asked Questions” in Prototype 4


Exhibit 10: The “food from food service places” page in Prototype 4


Computer-AssistedPersonal Interviewing forthe ConsumerExpenditure InterviewSurvey

LINDA SIOUX GROVES

Linda Sioux Groves is Chief, Division of Con-sumer Expenditure Surveys, Bureau of LaborStatistics.

Beginning in April 2003, the Con-sumer Expenditure (CE) Inter-view survey will be conducted

by computer-assisted personal inter-viewing (CAPI). The survey will con-tinue to be conducted in person by U.S.Census Bureau interviewers in therespondent’s home; however, the in-terviewer will administer the questionsand record the answers on a laptopcomputer in place of the current paper-and-pencil questionnaire. This articledescribes the process whereby the CECAPI questionnaire was designed anddeveloped and discusses some of thebenefits expected to be realized fromCAPI data collection in the areas ofdata quality, operational efficiency, andopportunities for future improvements.

Design and development of CAPIThe Census Bureau collects the datafor the CE Survey under contract withthe Bureau of Labor Statistics (BLS).The administration of the survey isvery much a collaborative effort be-tween the two agencies. Discussionsand planning regarding converting theCE Interview Survey to CAPI began in1997. The Census Bureau was alreadycollecting several other surveys inCAPI mode at that time, including theCurrent Population Survey and theHealth Interview Survey. However, allof the CAPI surveys being collectedby the Census Bureau, as well as all

the peripheral systems that supportcollection activities, such as the CaseManagement System, had been devel-oped in a DOS computing environment.The availability of new instrument-authoring software and more powerfullaptops led to an early decision that CECAPI would be developed in a MicrosoftWindows environment. The authoringsoftware chosen was Blaise, which wasdeveloped by Statistics Netherlandsand is in wide use in Europe and in otherU.S. survey organizations.

The CE CAPI development projectwas an interagency effort, with man-agement representatives from both BLSand the Census Bureau serving on theCAPI Steering Group (CSG). The steer-ing group developed the strategic planfor the project and chartered numerousworking teams that were then assignedto establish instrument design stan-dards, write specifications, programand test the CAPI instrument and re-lated systems, develop a new CaseManagement System, establish new-interviewer training, and plan a large“dress rehearsal” to assess the impactof CAPI on CE estimates. The steeringgroup approved the project plans foreach of the teams, facilitated communi-cation among teams, and monitoredprogress throughout the project.Among the goals of the Census Bureauwere (1) to use the CE CAPI develop-ment process to set Windows stan-


dards for any future CAPI developmentof other surveys and (2) to create aCase Management System that all ofthe surveys that the Census Bureauadministers could use. The latter aimwas important because Census Bureauinterviewers typically work on manydifferent surveys.

A great emphasis was placed on test-ing during the CE CAPI developmentprocess, with numerous different kindsof tests designed to accomplish vari-ous goals. During the design standardsstage, instrument prototypes were de-veloped and examined by experts fromseveral survey organizations. Once ba-sic design features, such as colors andfonts, were established, developmentof the CE instrument began. The deci-sion was made to program the ques-tionnaire, which consists of 22 sectionsroughly corresponding to different top-ics, in three modules.

The first module consisted of sec-tions of the questionnaire represent-ing as many different design issues aspossible. For example, it included sec-tions in which the interviewer readslong lists of items to the respondent,sections wherein screening questionsare used to skip the respondent to thecorrect set of detailed questions, andsections in which reported data fromthe previous interview are used exten-sively. Functionality testing was doneto ensure that the module met thespecifications. Following this testing,a panel of Census Bureau CE data col-lectors was brought in to test the us-ability of the module. The results fromthese two rounds of testing and thedecisions made on program design is-sues were then applied to the develop-ment of the next module, and the pro-cess was repeated until the entire datacollection instrument was programmed.The early input from data collectorsduring the development process re-sulted in a more “interviewer-friendly”collection instrument.

Concurrently with the developmentof the CAPI instrument, the new CaseManagement System and postcollec-tion processing system were developedand tested. Once all of these pieceswere complete, they were integrated

with the CAPI questionnaire and a sys-tems test was performed. The test wasused to make final adjustments to theCAPI system for the dress rehearsal.

The CE CAPI dress rehearsal beganin January 2002 in all 12 Census Bu-reau regional offices. Lasting 9 months,with three quarterly interviews per re-spondent in a sample of about 3,000households, the purpose of the dressrehearsal was to analyze the impact ofCAPI on response rates and on expen-diture estimates. A secondary purposeof the test was to make final adjustmentsto training and procedures by involvinga larger pool of data collectors.

The CAPI system was fully imple-mented in April 2003. Interview casesthat began their five-interview cyclewith the paper questionnaire were con-verted to CAPI at that time. In an effortto ease the transition of cases frompaper to CAPI, changes that were an-ticipated in the CAPI questionnairewere largely incorporated into the pa-per questionnaire in advance (during2001). Thus, the content of the paperinstrument and that of the CAPI instru-ment are nearly identical.

CAPI and data qualityThe CE interview is long and complexand usually takes from 60 to 90 min-utes to complete. In addition to col-lecting information on expenditures fora wide range of items, the survey col-lects detailed demographic and incomedata pertaining to consumer unit mem-bers, data on assets and liabilities, anddescriptive information about expen-ditures for classification and bound-ing purposes. The interviewer is re-quired to navigate correctly throughnumerous screening questions and onto the detailed questions, all the whileskipping inapplicable questions. Insome cases, the interviewer is requiredto carry forward information from onepart of the survey to another and makedecisions about which subsequentparts to administer or questions to ask,all on the basis of a complex decisiontable.

Reviews of the collected data revealthat, because of the complexity in-volved, interviewers sometimes make

mistakes in administering the paper-and-pencil interview, resulting in incon-sistencies or gaps in the reported data.If these errors are detected earlyenough, the interviewer may recontactthe respondent to fill in the missingdata. Otherwise, the errors must be re-solved through postcollection editing.

One of the advantages of a CAPIcollection instrument is that many ofthese types of data problems can beeliminated. The logic programmed intothe instrument forces the interviewerto stay on the correct path and doesnot allow questions to be inadvertentlyskipped. For example, in the section ofthe paper questionnaire dealing withproperties owned by the respondent,interviewers ask (1) one set of ques-tions for each new property reported,(2) different sets of questions, depend-ing on what type of mortgage the re-spondent has and whether there arealso home equity loans on the prop-erty, and (3) yet another set of ques-tions if the property was disposed ofor the mortgage payment amountchanged from what was reported in theprevious interview. The CAPI instru-ment will ensure a more seamless flowthrough all of the applicable questionsfor each property. In addition, the in-strument is able to keep track of longlists of items and ensure that the cor-rect set of detailed questions is askedfor each item. As a result, there is muchless postcollection editing and errorresolution with CAPI.

Another way in which CAPI willimprove the quality of the data is byrequiring the interviewer to verify un-usually high or low values with the re-spondent. Range edits are programmedinto the CAPI instrument, based onpreviously reported data. When a valueoutside of the allowable range for aparticular item is entered, an edit mes-sage is triggered, requiring the inter-viewer to explicitly accept the value orchange it, thus checking for typos. Theinterviewer is also allowed, and evenencouraged, to enter textual notes toexplain unusual values. An unusuallyhigh expenditure for dresses or cutflowers, for example, could be accom-panied by the note “Respondent is pre-


paring for daughter’s wedding.” Theinterviewer might also use the note fieldto indicate uncertainty about the clas-sification of an item. A $45,000 expen-diture under “Hobbies,” for instance,might be accompanied by a note “Re-spondent collects antique cars.” Notessuch as these can prove useful to ana-lysts who examine the data later, be-cause items with notes associated withthem are flagged in the data file and thetext of the notes will be stored with thedata. An outlier detection system willautomatically display the notes to thedata reviewer.

Another feature of the CAPI instru-ment is that help screens will be madereadily available to interviewers as theyadminister the questionnaire, ratherthan in a separate collection manualthat might be difficult to consult dur-ing an interview. The CE CAPI helpscreens include examples, such as whatto include under “small household ap-pliances,” and definitions, such asthose of “PPO” and “IPA” with regardto the type of health insurance thateach offers.

Finally, another way in which CAPImay improve the quality of CE data isby allowing new items to be added tothe questionnaire more quickly as theyenter the marketplace. This feature ishighly important to one major user ofCE data—the Consumer Price Indexprogram—in terms of keeping the in-dex as current as possible, as well asbeing important to CE data users in theprivate sector.

Operational advantages of CAPIFrom a survey operations perspec-tive, a CAPI instrument has severaladvantages over paper-and-pencildata collection.

Currently, interviewers send theircompleted CE paper questionnaires tothe Census Bureau’s National Process-ing Center in Jeffersonville, IN. There,the clerical staff checks questionnairesagainst a master control list, appliescodes to certain items (for example, onthe basis of the interviewer’s descrip-tion, the make and model of a vehicle

are coded), and keys in and verifies thedata. With the implementation of CAPI,the data are input directly into the com-puter with no separate keying-in step.Coding is done as the data are entered.(In the case of a vehicle’s make andmodel, the interviewer will select thecorrect description from an alphabeti-cal popup list.) Instead of physicallysending paper questionnaires to a cen-tral location for processing, the dataare transmitted nightly from theinterviewer’s home via a modem. Con-sequently, CAPI data collection shouldmake the data available for tabulationsooner.

Other survey operations also willbe streamlined by the conversion toCAPI. Currently, at the National Pro-cessing Center, clerical staff transcribescertain information from each com-pleted paper questionnaire onto thenext quarter’s blank questionnaire andmails both back to the Census Bureauregional office, which, in turn, mailsthem out to the appropriate interviewerin time for the next collection cycle. Thetranscribed data include inventorieditems, which the interviewer does notrecollect each time, but rather updateswith current information (for example,on properties owned), as well as ex-penditure data collected in the previ-ous period and now used for bound-ing in the current interview. Thesebounding procedures minimize tele-scoping errors that are common in ret-rospective interviews and result froma tendency to report past events in thereference period of the survey. WithCAPI data collection, once these dataare captured in electronic form andthen transmitted, an input file is cre-ated for the next quarter’s interview andis transmitted directly to the inter-viewer’s laptop.

Certain survey management andcontrol functions will also improve un-der CAPI. Field supervisors can easilyreassign cases to a different inter-viewer, if needed, simply by retrans-mitting information about the case. Su-pervisors in the field, as well asheadquarters staff, can get much more

timely reports on the status of datacollection activities than they couldusing paper questionnaires.

Future improvementsRespondent burden is a significant is-sue for the CE Interview survey, likelycontributing to underreporting of ex-penditures and to refusals by respon-dents to participate in later waves ofthe survey. Unfortunately, CAPI willprobably not make the interview anyless burdensome to the respondent,and early indications are that the in-terview may even take slightly longer.

However, future research might per-mit CAPI’s capabilities to be used tostreamline the interview and reduce re-spondent burden. More customizationof the interview could be possible,based, for example, on respondents’characteristics or previously reporteddata. Also, the added flexibility ofCAPI might allow more experimenta-tion with global questions and random-ization of topics, so that not all partsof the questionnaire would need to beasked during each wave of the survey.

CAPI will certainly afford survey re-searchers much more quantitative in-formation about the interview processitself. For example, each CAPI interviewproduces an audit trail that allows oneto “replay” the interview. This can beused to diagnose trouble spots in theinterview, detect whether the inter-viewer jumped around in the instru-ment or followed the default path, andcount how many times help screenswere invoked or warning messageswere suppressed. Similarly, timing datafrom the CAPI instrument can be usednot only to measure overall interviewlength, but also to access how revi-sions to questions affect timing in in-dividual sections. These are valuabletools in the CAPI instrument that arenot available in a paper interview.Through them, investigators can gaina much better understanding of someof the difficulties facing interviewers,and that increased understanding willlead to further improvements in thedata collection process.


Introducing Brackets:Quality in the ConsumerExpenditure InterviewSurvey

GEOFFREY PAULIN

Geoffrey Paulin is a senior economist in theBranch of Information and Analysis, Divi-sion of Consumer Expenditure Surveys, Bu-reau of Labor Statistics.

Nonresponse is a problem insurveys. Some potential par-ticipants may refuse to partici-

pate at all in a survey, while others mayprovide answers to some, but not all,questions asked. For those who par-ticipate at least partially, reasons for notresponding to certain questions mayinclude the sensitivity of the respon-dent to the question asked or simply alack of knowledge on the part of therespondent. One situation in which ei-ther of these two reasons may be citedis when respondents are asked aboutincome levels and sources. Some re-spondents may refuse to answer ques-tions about income because they con-sider the matter too personal todivulge. Others may be willing to an-swer, but may not be able to do so com-pletely, because they lack specific ordetailed knowledge. This is often thecase in “proxy reporting,” wherein therespondent reports income informationfor another member of the consumerunit.1 For example, a parent may notknow precisely the amount of incomeearned by a teenaged daughter who isemployed after school at a neighbor-hood fast-food restaurant.

In the case of complete refusal toparticipate in the survey, little can bedone to obtain information. By contrast,

as regards sensitive questions or lackof knowledge, information may begained by allowing the respondent togive an answer that is not precise. Forexample, a person earning a salary of$300,000 may refuse to divulge that in-formation precisely, but may be com-fortable saying that the salary is“greater than $120,000.” Similarly, theaforementioned parent may not knowthe precise salary of his teenageddaughter, but may know with confi-dence that it is “less than $5,000” peryear. Prior to the second quarter of2001, such information was lost in theConsumer Expenditure (CE) Interviewsurvey, because the respondent couldonly report a value, assert “don’tknow,” or refuse to answer. However,starting in April 2001, respondents weregiven the opportunity to provide anincome range, or “bracket,” when theywere unable or unwilling to give a spe-cific value. This article describes thecollection of income data and the de-velopment of income brackets in theCE Interview survey.

Income data are collected in the sec-ond and fifth interviews for those whoparticipate in those interviews. If theconsumer unit does not complete itssecond interview (for example, if thefamily is unavailable during the surveyperiod or if the family originally resid-ing at the address during the secondinterview has moved away and the new

1 See “Glossary” in Appendix A at the endof this anthology for the definition of a con-sumer unit.


residents are now participating in-stead), the information is collected atthe earliest possible interview (thethird, fourth, or fifth). In either case,incomes are collected for the past year,as determined by the date of the inter-view. For example, a consumer unit in-terviewed in July 2002 would have beenasked to recall income received fromJuly 2001 to June 2002.

Data are collected on severalsources of income. Some of these, suchas data on wages and salaries, are col-lected for members of the consumer unitwho are at least 14 years old. Others,such as information on interest income,are collected for the consumer unit as awhole. In addition to data on “labor”(wage and salary or self-employment)income and “nonlabor” (interest or divi-dend) income, information on othersources (such as alimony, child sup-port, Food Stamps, and welfare income)also is collected. (For a complete list-ing of sources, see the appendix to thisarticle.)

History of bracketing in theInterview surveyIn May 1998, a 2-day seminar was heldat Princeton University to discuss theutility of the CE Survey for measuringpoverty and related issues. During thecourse of the seminar, many ideas forimproving the quality of the data wereproposed. One of these was to investi-gate the use of brackets for collectingdata on income, assets, and liabilities,because these data are important, butfrequently missing. Katharine G.Abraham, Commissioner of the Bureauof Labor Statistics (BLS) at the time,asked her organization’s Division ofConsumer Expenditure Surveys tostudy the feasibility of collecting brack-eted data, starting with the 2000 sur-vey.

In September 1998, a team was char-tered to investigate and recommendstrategies for the implementation ofbracketing if it was deemed feasible.The team had two major questions toanswer: first, does bracketing reducenonresponse in practice? Second,which type or types of brackets, if any,should be used? Starting with a review

of the literature on the subject, the teamdiscovered that, as expected, bracket-ing was useful for collecting data, be-cause respondents with impreciseknowledge could provide at least someinformation. However, one unintendedconsequence described in the literatureis that bracketing can lead to a loss ofprecision, because some respondentswho report bracketed data might havereported actual values if the interviewerhad probed sufficiently.2 In addition,the team reasoned that brackets wouldincrease respondents’ burden, because,without them, a respondent could sim-ply refuse to answer or respond “Idon’t know,” and the next questionwould be asked. With brackets, onceeither of these occurs, the interviewerattempts to collect a bracketed value.Still, the team concluded that bracketswould be useful despite these con-cerns. For example, the loss of preci-sion might be outweighed by an increasein overall response when brackets wereused. Interestingly, the literature alsosupported the hypothesis that brack-ets do not seriously increase respon-dents’ burden: although it is true thatthere is one more question in caseswhere the initial response is “I don’tknow” or a refusal, it also is true that alarge number of those who initially re-spond in either of those ways is sub-sequently willing and able to provide abracketed value.3

Constructing the bracketsGiven that brackets are indeed usefulin data collection, the second questionbecomes operative. The team discov-ered that there are at least two types ofbrackets used in practice: “conven-tional” brackets and “unfolding” brack-ets. With both types, the respondentis first asked for a specific value. If he

or she is unable to provide one, then,in a conventional-bracketing frame-work, the respondent is asked to iden-tify, from a predetermined list, the rangein which the income or asset is likely tofall (for example, less than $5,000;$5,000 to $9,999; $10,000 to $19,999; andso forth). In an unfolding-bracketingframework, the respondent is asked aseries of questions designed to elicitranges of values. For instance, the in-terviewer might say, “Is it at least$5,000?” If the response is “No,” thena range of less than $5,000 would berecorded. If the response is “Yes,” thenthe respondent would be asked, “Is itat least $10,000?” If “No,” then a rangeof $5,000 to $9,999 would be recorded.If “Yes,” the respondent would beasked, “Is it at least $20,000?” If “No,”then a range of $10,000 to $19,999 wouldbe entered. If “Yes,” then a responseof “at least $20,000” would be recorded,and the next question in the surveywould be asked. The team recom-mended that conventional bracketingbe adopted, for a couple of reasons:first, more precise answers would beobtained. (For some sources of income,such as wages and salaries, it is likelythat a large percentage of recipientscould accurately respond that their in-come from those sources was “at least$20,000”; narrower ranges, such as$20,000 to $29,999 and so forth, allow amore precise estimate of the value ofsuch income.) Second, conventionalbrackets were thought to be less bur-densome, because the respondentcould be handed a card with the appro-priate ranges and quickly scan it to findwhich was appropriate for the sourcein question. With unfolding brackets,the respondent might be asked threeadditional questions, instead of one.

Once the type of bracketing wasselected, the next question was whatthe ranges of the brackets should be.One idea was to use standard publica-tion ranges as a guide. For example,data currently are published for fami-lies whose total income is less than$5,000; $5,000 to $9,999; $10,000 to$14,999; and so forth. However, the In-terview survey collects informationfrom a variety of sources, some for each

2 Kennickell, Arthur B., “Using RangeTechniques with CAPI in the 1995 Surveyof Consumer Finances,” on the Internet athttp://www.federalreserve.gov/Pubs/oss/oss2/papers/rangepap0197.pdf, January1997.

3.Juster F. Thomas and James P. Smith,“Improving the Quality of Economic Data:Lessons from the HRS and AHEAD,” Jour-nal of the American Statistical Association,vol. 92, no. 440, December 1997, pp. 1268–1278.


member aged 14 and older, some for theconsumer unit as a whole. The publi-cation ranges may be appropriate forsome sources of income (for instance,wage and salary income), but may notbe appropriate for other sources. Forexample, almost all respondents whoreported interest income reported avalue less than $5,000, so, for thissource, the publication range is toobroad to be meaningful. To determinethe most useful ranges, the distribu-tion of each source was analyzed. Then,through a combination of empirical ex-amination and normative analysis, afew sets of brackets were developed tofit the different kinds of data. The em-pirical examination involved looking atthe percentiles for each source of in-come and seeing where breaks oc-curred. Normative analysis involvedfinding “reasonable” cutoff values forthe data.

Refining the brackets, using theBLS cognitive laboratoryThe next step in the implementationprocess required testing the results inthe BLS cognitive laboratory. At thisstage, a new team was formed that in-cluded a member of the Survey Re-search branch of the Division of Con-sumer Expenditure Surveys and acognitive psychologist from the BLSOffice of Survey Methods Research.Cognitive psychologists are trained inhow respondents perceive certainquestions. That is, when the interviewerasks about interest income, does therespondent correctly perceive what theinterviewer is asking for (such as inter-est earned on checking and savingsaccounts), or might the respondent beconfused and include other sources ofincome (such as dividends fromstocks), or might the respondent evenreport no income received, when, infact, he or she did receive such income,but thought it was something else? Inthe cognitive laboratory, tests are per-formed in which respondents are asked

for their answers and then are debriefedby the psychologist. During the test-ing, the psychologist might ask the re-spondent to define certain terms, tomake sure that the respondent’s defi-nition matches the interviewer’s; or therespondent might be asked questionsabout the survey in general—were thequestions posed easy or difficult tounderstand and answer, for example.

After the brackets were refined onthe basis of findings from the cogni-tive tests, the brackets were ready tobe implemented. Various steps were in-volved in their implementation, includ-ing revising the survey instrument de-signed to collect the data, field-testingthe instrument, and obtaining appro-priate approvals from offices that regu-late Government surveys. Bracketingfinally appeared in the CE InterviewSurvey in the second quarter of 2001.That is, the first respondents to thesurvey who were asked to providebracketed information began their par-ticipation in April 2001.4 Currently, onlyincome brackets have been imple-mented. The original team investigatedthe possibility of using brackets forassets and liabilities as well, but de-cided to start with income only andthen apply any lessons learned there-from to the implementation of assetsand liabilities.

ConclusionsAt present, the first year (2001) of datagathered with the use of brackets hasbeen published, and a new team hasbeen chartered to study how bracketshave changed the collection of incomedata. Among the questions being in-vestigated are the following: are many“don’t knows” and refusals to answer

being converted to bracketed values?Have brackets improved the percent-age reporting various sources of in-come? Has average income reportedrisen as a result of using brackets? andAre there any demographic differencesin the propensity to provide bracketedinformation? As these issues are ana-lyzed, further research results will bepublished documenting the findings.

APPENDIX:Income Sources and

Bracket Ranges

Data on the following sources of in-come are collected for each individualmember of the consumer unit who is atleast 14 years old: Wages or salary; in-come (or loss) from nonfarm business,partnership, or professional practice;income (or loss) from own farm; SocialSecurity or Railroad Retirement Income;and Supplemental Security Income.

The following sources of income arecollected for the consumer unit as awhole: Unemployment compensation;workers’ compensation and veterans’payments, including education; publicassistance or welfare, including moneyreceived from job training grants suchas Job Corps; Food Stamps and elec-tronic benefits transfers; interest on sav-ings accounts or bonds; regular incomefrom dividends, royalties, estates, ortrusts; pensions or annuities from pri-vate companies, the military, or gov-ernment; income (or loss) from room-ers or boarders; income (or loss) frompayments from other rental units; childsupport; regular contributions from ali-mony or other sources, such as per-sons outside the consumer unit; andother money income, including moneyreceived from care of foster children,cash scholarships, fellowships, or sti-pends not based on working.

Table 1 shows the brackets appliedby the interviewer to each source ofincome.

4 Although the initial goal was for imple-mentation in 2000, it became apparent thatto implement bracketing properly would re-quire cognitive testing and other processes.Therefore, the implementation was delayeduntil 2001.


Tab

le 1

. Dat

a co

llect

ed b

y so

urc

e o

f in

com

e an

d in

com

e b

rack

ets

Wag

es a

nd s

alar

ies;

non

farm

bus

ines

s in

com

e;...

......

.Lo

ss1

0–5,

000–

10,0

00–

15,0

00–

20,0

00–

30,0

00–

40,0

00–

50,0

00–

70,0

00–

90,0

00–

120,

000

— o

wn-

farm

inco

me

......

......

......

......

......

......

......

......

......

..4,

999

9,99

914

,999

19,9

9929

,999

39,9

9949

,999

69,9

9989

,999

119,

999

or m

ore

Soc

ial S

ecur

ity a

nd R

ailro

ad R

etire

men

t....

......

......

......

—Le

ss30

0–40

0–50

0–60

0–70

0–80

0–90

0–1,

000–

1,50

0 or

——

Inc

ome

......

......

......

......

......

......

......

......

......

......

......

......

than

300

999

499

599

699

799

899

999

1,49

9m

ore

Sup

plem

enta

l Sec

urity

......

......

......

......

......

......

......

......

..—

0–1,

000–

2,00

0–3,

000–

4,00

0–5,

000–

10,0

00–

15,0

00–

20,0

00–

30,0

00–

40,0

00–

50,0

00 I

ncom

e...

......

......

......

......

......

......

......

......

......

......

......

...99

91,

999

2,99

93,

999

4,99

99,

999

14,9

9919

,999

29,9

9939

,999

49,9

99or

mor

e

Dat

a co

llect

ed fo

r the

......

......

......

......

......

......

......

......

....

0–1,

000–

2,00

0–3,

000–

4,00

0–5,

000–

10,0

00–

15,0

00–

20,0

00–

30,0

00–

40,0

00–

50,0

00 c

onsu

mer

uni

t as

a w

hole

......

......

......

......

......

......

......

.Lo

ss2

999

1,99

92,

999

3,99

94,

999

9,99

914

,999

19,9

9929

,999

39,9

9949

,999

or m

ore

1 Sel

f-em

ploy

men

t inc

ome

only

2 Ren

tal i

ncom

e on

ly

NO

TE

: The

fore

goin

g br

acke

ts s

how

n fo

r sou

rces

col

lect

ed fo

r the

con

sum

er u

nit a

s a

who

le a

re a

lso

colle

cted

for t

he fo

llow

ing

sour

ces

of m

oney

, whi

ch a

re n

ot c

onsi

dere

d in

com

e in

the

CE

Inte

rvie

w s

urve

y: lu

mp-

sum

pay

men

ts re

ceiv

ed fr

om e

stat

es, t

rust

s, ro

yalti

es, a

limon

y, p

rizes

, or g

ames

of c

hanc

e or

from

per

sons

out

side

the

cons

umer

uni

t; an

d sa

les

of h

ouse

hold

furn

ishi

ngs,

equ

ipm

ent,

clot

hing

, jew

elry

, and

pet

s or

oth

er b

elon

ging

s, e

xclu

ding

the

sale

of v

ehic

les

or p

rope

rty.

0In

com

e so

urc

eR

ang

e fo

r b

rack

et—

12

34

1211

105

67

89


Characteristics ofComplete andIntermittent Respondersin the ConsumerExpenditure QuarterlyInterview Survey

Sally E. Reyes-Morales is a mathematical stat-istician in the Division of Price StatisticalMethods, Branch of Consumer ExpenditureSurveys, Bureau of Labor Statistics.

SALLY E. REYES-MORALES The Consumer Expenditure (CE)Quarterly Interview Survey col-lects data from consumer units

(CUs) across the United States. SomeCUs complete all five interviews, oth-ers complete some, but not all, of theinterviews, and some choose not toparticipate in the survey at all. TheseCUs can be called complete respond-ers, intermittent responders, and nonre-sponders, respectively. Do the nonre-sponses of the intermittent respondersand nonresponders affect the pub-lished CE estimates? Are the CUs whostay in the survey for all five interviewsdifferent from those who do not?

To answer these questions, thisstudy uses the CE Interview Surveydata collected from 1997 to 2000. Inthe study, characteristics and expen-ditures of complete responders andintermittent responders are com-pared. Nonresponders are excludedbecause very little information aboutthem is collected.

Background and definitionsThe CE Interview Survey is a rotatingpanel survey in which a random sampleof residential addresses is selected andthe CUs living at those addresses areasked to report their expenditures dur-ing the previous 3 months. The U.S.Census Bureau collects these data forthe Bureau of Labor Statistics. The ran-dom sample of residential addresses isselected by means of systematic sam-

pling, and the CUs at those addressesare interviewed by the Census Bureaufield representatives once per quarterfor five consecutive quarters. After thefifth quarter, the CU leaves the sampleand a new address is selected to re-place it. The CE sample is representa-tive of the total civilian population of theUnited States not living in institutions.

In the initial CE Interview, respon-dents are asked to report all of the ex-penditures they made during the pre-vious month. This interview is usedonly for “bounding” purposes—thatis, to make sure that the expendituresreported in the second through fifthinterviews reflect the correct periods.In the second through fifth interviews,expenditure data are collected for the 3months prior to the interview. Only theexpenditure data collected in the sec-ond through fifth interviews are usedto compute official CE estimates. Datacollected in each quarter are treated in-dependently, so annual estimates donot depend upon any CUs participat-ing for all five quarters.

Following are some of the terms thatwill be used in this article, together withtheir definitions:

Household. The people who occupy ahousing unit. A housing unit is a house,an apartment, a mobile home, a room,or a group of rooms occupied (or in-tended to be occupied) as separate liv-ing quarters.


INTERI. Interview number (1 through5).

INSTAT. Final interview status (01through 19):

01 = Interview

Type A noninterviews:02 = No one home03 = Temporarily absent04 = Refused05 = Other Type A noninterviews

Type B noninterviews:06 = Vacant (for rent)07 = Vacant (for sale)08 = Vacant (other)09 = Occupied by person whose usual09 = residence is elsewhere10 = Under construction (not ready)11 = Other Type B noninterviews

Type C noninterviews:12 = Demolished13 = House or mobile home moved14 = Converted to nonresidential use15 = Merged16 = Condemned17 = Located on military base18 = CU moved19 = Other Type C noninterviews

Interview. An interview is completedby an eligible CU (INSTAT = 01).

Type A noninterviews. An address iswithin the scope of the survey andeligible for interview, but an interviewis not obtained (INSTAT = 02 through05).

Type B noninterviews. An address iswithin the scope of the survey, but isnot eligible for interview (INSTAT = 06through 11).

Type C noninterviews. An address isout of the scope of the survey or per-manently ineligible for the CE sample(INSTAT = 12 through 19).

CU (consumer unit). See “Glossary” inAppendix A at the end of this anthology.

Reference person. See “Glossary” in Ap-pendix A at the end of this anthology.

Consumer units usedIn this study, selected demographiccharacteristics of CUs who completedthe last four interviews (INTERI = 2through 5) were compared with corre-sponding characteristics of those who

did not. To make these comparisons,the universe of CUs from which datawere collected was subdivided, usingthe following criteria:

• Only CUs scheduled to partici-pate in all five interviews be-tween January 1997 and Decem-ber 2000 were used, in order tofollow CUs’ history in the survey.

• Only CUs who completed one ormore of the last four interviews(INTERI = 2 through 5) wereused, because the demographiccharacteristics examined in thestudy are not collected in the firstinterview.

The response rates for the CUs usedin the current study are different fromthe CE response rates published in CEreports, because not all CUs were usedin the study. Table 1 shows the re-sponse rates computed from all recordsin the CE sample, compared with theresponse rates computed from only therecords used in the study. The study’sCUs had higher response rates andlower nonresponse rates than the com-plete universe of CUs had, because thestudy excludes CUs who completednone of the last four interviews (INTERI= 2 through 5).

Typically, Type B and Type C nonin-terviews are not used in response ratecalculations, because they are ineligibleor out of the scope of the survey. Re-sponse rates usually are computed withthe following formula:

Response Rate =

Table 2 shows response rates for CUswho completed the third interview; thethird and fourth interviews; and thethird, fourth, and fifth interviews, giventhat they completed the second inter-view. Of the CUs who completed thesecond interview, 93.1 percent alsocompleted the third interview, 88.7 per-cent completed the third and fourth in-terviews, and 85.9 percent completedthe third, fourth, and fifth interviews.

Demographic characteristics ofcomplete and intermittentrespondersTable 3 compares some demographiccharacteristics of the CUs who com-pleted all of the last four interviews(complete responders) with those ofCUs who did not (intermittent respond-ers). The complete responders tend tohave more members and to be older thanthe intermittent responders and alsoare more likely to be homeowners andmarried couples. The average numberof persons in a complete-responder CUis 2.6, compared with 2.3 for the inter-mittent responders. Likewise, the aver-age age of the reference person in com-plete-responder CUs is greater (50.6,compared with 40.9), the average quar-terly expenditure per CU on all items isgreater ($8,981, as opposed to $7,504),and the average quarterly expenditureper person is greater ($3,442, as against$3,212) than in intermittent-responderCUs. Complete-responder CUs also aremore likely to have both husbands andwives present in the household (57.2percent, compared with 39.8 percent),less likely to be single consumers (25.3percent versus 37.5 percent), morelikely to be homeowners (73.2 percent,as opposed to 41.0 percent), and morelikely to be the only CU living in thehousehold (98.3 percent, comparedwith 87.3 percent).

Table 4 shows some of the same CUcharacteristics, by type of noninter-view. CUs who had one or more Type Bor Type C noninterviews tend to berelatively young (the average age ofthe reference person is 36.0), have fewpeople in them (2.2 persons, on aver-age), have a low average expenditureper CU ($6,863), and have a low aver-age expenditure per person ($3,124).

CUs who drop out of the surveyCUs are considered to have droppedout of the survey permanently whenno more of their interviews are com-pleted with interview status codeINSTAT = 01. These CUs are a subsetof the intermittent responders. The rea-son they have dropped out of the sur-vey can be identified by the interviewstatus code of the first noninterview

100ATypeInterviews

Interviews×

+.


tently, and they completed 1.9 inter-views, on average. The intermittent re-sponders accounted for only 26.9 per-cent of all interviews.

The average quarterly expenditureis higher for CUs who completed all fourinterviews than for those who did not($8,981, compared with $7,504); theoverall average expenditure was $8,584.To estimate the effect that intermittentresponders have on the CE expendi-ture estimates, the average quarterlyexpenditure per CU can be computedin two different ways by changing theweights used for the intermittent re-spondents. In table 7, the overall aver-age expenditure per CU can be com-puted by weighting the two sets of CUsby the actual number of completedin terviews:

If the response rates could be in-creased so that the intermittent re-sponders completed all four interviews,then those CUs would have completed43,280 (= 4 × 10,820) interviews. If, inaddition, their expenditures are inde-pendent of their (non)participation inthe CE Survey, the weighted averagewould be $8,339, because

The $8,339 figure is a 2.9-percent

after their last completed interview.Table 5 shows that the most commonreason for dropping out of the surveyis “refusal” (23.7 percent), followed by“other” unspecified Type C noninter-views (19.5 percent), “vacant, for rent”(19.4 percent), and “vacant, other” (14.1percent).

Table 6 shows the percentage ofCUs who came back and participatedin the survey after a refusal. Of the CUswhose first refusal was in the secondinterview, only 30.8 percent completedone or more of the remaining inter-views. Of the CUs whose first refusalwas in the third interview, 52.7 percentcompleted one or more of the remain-ing interviews, and of the CUs whosefirst refusal was in the fourth interview,47.4 percent completed the fifth inter-view. Overall, there were 5,554 CUswhose first refusal was in one of inter-views 2 through 4, and 36.8 percent ofthem eventually came back to partici-pate in the survey.

The effect of intermittent respond-ers on CE expenditure estimatesTable 7 shows the total number of in-terviews completed by both the com-plete and intermittent responders.There were 24,860 CUs used in thestudy and 56.5 percent of them com-pleted all four interviews. Those CUsaccounted for 73.1 percent of all inter-views. By contrast, 43.5 percent of theCUs in the study responded intermit-

decrease from the $8,584 calculated thefirst way, indicating that the effect ofintermittent responders on the overallaverage expenditure is relatively small.Moreover, every CU in the CE Surveyhas a weight associated with it, andthe weights include adjustments fornonresponses. As a result of these ad-justments, the 2.9-percent differencecomputed here can be viewed as anupper bound on the true difference;hence, the effect of intermittent re-sponders on the published CE esti-mates is probably considerably lessthan 2.9 percent.

ConclusionsThe study presented in this articlelooked at CE data collected from 1997to 2000 and found that CUs who com-pleted all of the survey’s last four in-terviews (INTERI = 2 through 5) are dif-ferent from CUs who responded onlyintermittently. CUs who completed allfour interviews are larger and older andare more likely to be homeowners andmarried couples than are CUs who re-sponded only intermittently. The studyalso found that the nonresponses ofthe intermittent responders appear tohave a relatively small effect on thepublished estimates. An upper boundon this effect was calculated to be 2.9percent, but, because CU weights inthe CE Survey include adjustments fornonresponses, the actual effect is prob-ably considerably smaller.

280,43 160,56)504,7$ 280,43( )981,8$ 160,56(

339,8$+

×+×= .

702,20 160,56)504,7$ 702,20( )981,8$ 160,56(

584,8$+

×+×= .


Table 1. Response and nonresponse rates for all records, compared with thosefor records from CUs in this study

Number Percent Number Percent

Interviews (I) .................................................... 135,383 65.6 76,862 84.3Type A noninterviews (A) ................................ 32,982 16.0 6,992 7.7

Refusals (R) ............................................... 27,095 13.1 5,272 5.8Other Type A noninterviews ....................... 5,887 2.9 1,720 1.9

Type B noninterviews ...................................... 29,980 14.5 4,852 5.3Type C noninterviews ...................................... 7,994 3.9 2,442 2.7Total ................................................................. 206,339 100.0 91,148 100.0

Response rate of the total sample (I/Total) ..... 65.6 84.3Response rate of the eligible units (I/(I + A)) ....................................................... 80.4 91.7

Refusal rate of the eligible units (R/(I + A)) ......................................... 16.1 6.3

Records from CUsin this study

All records(INTERI = 1–5)

2 2,3 2,3,4 2,3,4,5

CUs who completed the interviews ................. 19,310 16,819 15,145 14,040CUs with at least one Type A noninterview ..... 1,242 1,921 2,309CUs with only Type B or Type C noninterviews 1,249 2,244 2,961Total Interview + Type A ................................... 18,061 17,066 16,349

Probability of completing interview ................... 3 3,4 3,4,5Response rate (I/(I + A)) (percent) ................. 93.1 88.7 85.9

Table 3. Demographic characteristics of complete responders and intermittentresponders

Yes No

Average size of CU ........................................ 2.6 2.3Average age of reference person .................. 50.6 40.9Average quarterly expenditure per CU .......... $8,981 $7,504Average quarterly expenditure per person .... $3,442 $3,212

Husband-and-wife families ............................. 57.2 39.8Husband and wife only .............................. 23.7 15.6Husband and wife with children ................. 29.0 20.9Other husband-and-wife families .............. 4.5 3.3

One parent, own children ............................... 5.5 8.3Single consumers ........................................... 25.3 37.5Other families .................................................. 12.0 14.4

Homeowner .................................................... 73.2 41.0Renter and other ............................................. 26.9 59.0

Single-CU household ...................................... 98.3 87.3Multiple-CU household .................................... 1.7 12.7

Demographic characteristicsDid the CU complete all four interviews

(INTERI = 2–5)?

Item

Table 2. CU response rates, given that the second interview was completed

ItemInterviews

Percent distributions

Type of family:

Housing tenure:

Multiplicity household:


Table 5. Reasons for dropping out of the survey

Refusal .................................................................................... 23.7Other unspecified Type C noninterviews (INSTAT = 19) ........ 19.5Vacant, for rent (INSTAT = 06) ............................................... 19.4Vacant, other (INSTAT = 08) ................................................... 14.1Other Type A (INSTAT = 02,03,05) ......................................... 9.4Other Type C (INSTAT = 12–18) ............................................ 7.3Other Type B (INSTAT = 07,09–11) ........................................ 6.6

Number of CUs ................................................. 14,040 10,820 24,860Percent of CUs ................................................. 56.5 43.5 100.0Number of interviews ........................................ 56,160 20,702 76,862Percent of interviews ........................................ 73.1 26.9 100.0

Average quarterly expenditure per CU ............ $8,981 $7,504 $8,584

Table 4. CU characteristics by type of interview

Total ...................................................................... 76,862 48.0 2.5 $8,584 $3,385Completed all interviews (2–5) ............................. 56,160 50.6 2.6 8,981 3,442

At least one noninterview ...................................... 20,702 40.9 2.3 7,504 3,212At least one Type A noninterview .................... 9,084 47.2 2.5 8,324 3,309

No refusals ................................................. 2,462 46.0 2.4 8,991 3,811At least one refusal .................................... 6,622 47.6 2.6 8,077 3,138

At least one Type B or Type C noninterview (no Type A noninterview) ................................ 11,618 36.0 2.2 6,863 3,124

CharacteristicNumber ofcompletedinterviews

Means, 1997–2000

Age ofreference

person

Number ofpersonsin CU

Quarterlyexpenditure perCU on all items

Quarterlyexpenditure per

person on allitems

Table 6. CUs who came back after a refusal in the Interview survey

2 1,186 2,665 3,851 30.8 69.2 100.03 523 469 992 52.7 47.3 100.04 337 374 711 47.4 52.6 100.0

Total 2,046 3,508 5,554 36.8 63.2 100.0

1These CUs were excluded from the study because they completed none of the lastfour interviews.

Firstrefusal

Cameback

Did notcome back

Total Came back(percent)

Did notcome back(percent)

Total(percent)

1

Category

Did the CU complete allfour interviews

(INTERI = 2–5)? Total

Yes No

Reason Percent

Table 7. The effect of intermittent responders on consumer expenditureestimates


Standard Errors in theConsumer ExpenditureSurvey

JEFFREY L. BLAHA

Jeffrey L. Blaha is a mathematical statisti-cian in the Survey Methods Division, Con-sumer Expenditure Survey, Bureau of LaborStatistics.

Data for the Consumer Expendi-ture (CE) Survey are collectedfrom a sample of all the con-

sumer units (CUs) in the United States.Estimates of the average (mean) annualexpenditure per CU in the CE tables forthe year 2000 were based on a sampleof about 30,000 CUs, out of a total ofabout 109 million in the Nation. Thesemean estimates differ from the truepopulation values because a subset,rather than the whole population, isobserved. Sampling error is the differ-ence between the survey estimate andthe true population value. The mostcommon measure of the magnitude ofthe sampling error is the standard er-ror of the estimate. The standard errorprovides data users with informationabout the variability associated withthe estimate.

Prior to the publication of the 2000data, the CE program made availableseparate tables of standard errors forthe Interview and Diary componentsof the CE Survey. Starting with the 2000data, the CE program began usingtables with integrated data from bothsurveys. Integrated data provide a com-plete accounting of consumer expen-ditures and income, which neither sur-vey component alone is designed todo. The tables, which correspond tostandard integrated tables of CU expen-ditures published in the CE reports andon the CE Web site, are provided bystandard demographic characteristics

(except for region).1 This article givesa summary description of the half-sample replication method used to cal-culate the standard error statistics anddemonstrates the proper interpretationof these statistics.

MethodologyStandard textbook formulas for calcu-lating standard errors assume simplerandom sampling and do not apply tothe CE Survey, because it does not usea simple random sample. Instead, theSurvey uses stratified random sam-pling, with systematic sampling withinthe strata. Hence, a different methodfor calculating standard errors isneeded. Replication methods make upa class of techniques that provides away to produce unbiased and design-consistent estimates of standard errorfor complex survey designs when theusual assumptions are not satisfied.The fundamental idea behind repli-cat ion methods is to selectsubsamples repeatedly from the fullsample, calculate the statistic of in-terest for each subsample, and usethe variability among the subsamplesto estimate the standard error of thefull-sample statistic.

1 The replication methodology used tocalculate the standard errors is designed towork at the national level and is not appli-cable to regional estimates.


the square root of the variance:

The coefficient of variation (CV) isthe standard error expressed as a per-centage of the sample mean estimateand thus is independent of the scale ofmeasurement. The CV is calculated as

These formulas apply to aggregatedcategories as well as individual com-ponent items. In producing a table thatuses integrated data from both surveys,the aggregated categories may be com-posed of items from one survey or theother, or they can be based on inte-grated data from both surveys.

Interpretation of the statisticsThe primary purpose of calculatingstandard errors for the mean estimatesis to provide data users with a measureof the variability associated with theestimates. This variability measureshow close different estimates would beto each other if it were possible to re-peat the survey over and over, usingdifferent samples of CUs. While it isnot feasible to repeat the survey overand over, statistical theory allows thestandard error to be estimated anyway.A small standard error indicates thatmultiple independent samples wouldproduce values that are consistentlyvery close to each other, whereas a largestandard error indicates that multipleindependent samples would producevalues that are consistently not veryclose to each other.

The balanced repeated replicationmethod is used to estimate the stan-dard error in the CE Survey. In thismethod, sampled geographic locationsare divided into 40 groups (calledstrata). The CUs within each stratumare randomly divided into two halfsamples. Half of the CUs are assignedto one half sample, and the other halfare assigned to the other half sample.Because there are 40 strata and 2 groupsof CUs in each stratum, we can com-pute 240 (approximately 18 trillion) dif-ferent estimates of expenditure in whichwe use exactly half of the collecteddata. With this information, we can es-timate the standard error of CE esti-mates by examining how the differentestimates compare with the full-sampleestimate. In the balanced repeated rep-lication method, we use a 44 × 44Hadamard matrix to choose the 44“best” combinations of groups out ofthe 18 trillion possible combinations.

A variance estimate for each cat-egory of item is obtained by first com-puting the mean estimate of the itemfor each replicate, then summing thesquared deviations of the replicatemean estimates from the full-samplemean estimate, and then dividing bythe number of replicates. Thus,

where xi r is a calendar-period estimateof the mean expenditure for item i, us-ing the rth replicate data and xi is thecalendar-period estimate of the meanexpenditure for item i, using the full-sample data.

The standard error is calculated as

( ) ( ) ,441 2

44

1iri

ri xxxV −= ∑

−

.)()SE( ii xVx =

( ) ( ).100

SECV ×=

i

ii x

xx

Table 1 is an extract from one of thestandard published CE demographictables. The table shows the mean esti-mate, standard error, and coefficient ofvariation for a list of expenditure itemsand categories, using integrated datafrom both the Interview and Diary Sur-veys in 2000. For example, the tableshows that the average annual expen-diture by all CUs on personal care prod-ucts and services for 2000 was $563.62,with a standard error of $7.94. Becauseit was impossible to ask all CUs in thecountry how much they spent on per-sonal care products and services, the$563.62 mean figure is an estimate, andwe have a margin of error, usually de-fined as ± 2 standard errors. In this ex-ample, the average annual expenditureon personal care products and serviceshas a margin of error of ± $15.88. Thus,we can say that the average CU prob-ably spent between $547.74 and $579.50($563.62 ± $15.88) annually on personalcare products and services.

Because the CV is the standard er-ror as a proportion of the mean esti-mate, it provides an indication of thespread of the data around the mean.The smaller the CV, the smaller is thespread of the data around the mean.The CV also makes possible compari-sons of the spread of data around themean of different items. For example, inthe 2000 integrated survey, the CV foreducation is 4.55 percent and the CVfor personal care products and servicesis 1.41 percent. Comparing the CVs forthe two items, we can say that thespread of the data around the mean foreducation expenditure is larger than thespread of the data around the mean forpersonal care products and services.


Table 1. Quintiles of income before taxes, annual means, standard errors, and coefficients of variation,Consumer Expenditure, 2000

Mean ................................................... $44,649 $44,649 $7,683 $19,071 $32,910 $53,295 $110,118 (1)SE ....................................................... 517.9 517.9 137.6 87.1 84.6 193.3 1613.4 (1)CV(percent) .......................................... 1.2 1.2 1.8 0.5 0.3 0.4 1.5 (1)

Average annual expenditures Mean ................................................... $38,041.03 $40,234.86 $17,939.45 $26,547.37 $34,713.42 $46,791.00 $75,093.08 $32,059.31

SE ....................................................... 336.7 356.8 399.9 622.2 412.3 626.7 850.8 713.0CV(percent) .......................................... 0.9 0.9 2.2 2.3 1.2 1.3 1.1 2.2

FoodMean ................................................... $5,157.88 5,434.76 2,673.31 4,178.21 5,183.19 6,451.56 8,679.37 4,516.56SE ....................................................... 65.8 77.4 97.0 106.8 143.6 127.7 184.0 93.3CV(percent) .......................................... 1.3 1.4 3.6 2.6 2.8 2.0 2.1 2.1

Alcoholic beveragesMean ................................................... 371.81 422.87 206.4 247.93 366.12 512.9 780.2 253.67SE ....................................................... 15.6 21.2 32.7 24.7 22.3 53.7 75.4 17.9CV(percent) .......................................... 4.2 5.02 15.9 10.0 6.1 10.5 9.7 7.1

HousingMean ................................................... 12,318.51 12,527.38 6,508.78 8,482.33 10,857.48 14,151.75 22,610.61 11,788.92SE ....................................................... 148.8 172.1 202.2 125.9 144.9 267.4 326.7 267.7CV(percent) .......................................... 1.2 1.4 3.1 1.5 1.3 1.9 1.5 2.3

Apparel and servicesMean. .................................................. 1,852.53 2,000.22 843.47 1,298.60 1,612.83 2,261.46 3,980.12 1,500.88SE ....................................................... 38.9 54.3 61.3 72.3 86.6 79.1 198.4 62.1CV(percent) .......................................... 2.1 2.7 7.3 5.6 5.4 3.5 5.0 4.1

TransportationMean ................................................... 7,417.36 7,567.51 3,211.97 5,042.68 7,028.41 9,223.30 13,315.32 6,985.40SE ....................................................... 101.2 110.7 133.4 222.5 234.9 211.0 322.1 211.2CV(percent) .......................................... 1.4 1.5 4.2 4.4 3.3 2.3 2.4 3.0

HealthMean ................................................... 2,065.67 2,120.04 1,469.87 1,987.62 1,964.07 2,312.36 2,864.12 1,919.38SE ....................................................... 30.1 31.6 71.7 63.7 50.6 61.7 64.0 60.3CV(percent) .......................................... 1.5 1.5 4.9 3.2 2.6 2.7 2.2 3.1

EntertainmentMean ................................................... 1,863.50 1,957.63 836.92 1,146.62 1,609.18 2,324.39 3,866.21 1,602.97SE ....................................................... 35.6 39.1 66.4 50.7 64.0 74.7 117.7 70.3CV(percent) .......................................... 1.9 2.0 8.0 4.4 4.0 3.2 3.0 4.4

Personal care products and servicesMean ................................................... 563.62 595.33 318.28 441.98 533.59 698.91 982.91 490.56SE ....................................................... 8.0 9.8 13.1 14.8 22.7 23.0 23.0 14.3CV(percent) .......................................... 1.4 1.7 4.1 3.4 4.3 3.3 2.4 3.0

ReadingMean ................................................... 146.47 156.11 73.04 105.17 135.61 174.92 291.47 118.38SE ....................................................... 2.2 2.3 3.2 4.3 3.9 5.9 7.0 5.0CV(percent) .......................................... 1.5 1.5 4.4 4.1 2.9 3.4 2.4 4.3

EducationMean ................................................... 631.93 635.52 430.25 290.47 393.09 600.05 1,461.94 625.79SE ....................................................... 28.8 35.7 59.5 41.8 48.9 53.5 107.4 56.3CV(percent) .......................................... 4.6 5.6 13.8 14.4 12.4 8.9 7.4 9.0

Tobacco products and smoking suppliesMean ................................................... 318.62 333.3 257.24 316.91 366.31 390.04 335.94 275.75SE ....................................................... 8.1 11.1 13.0 18.3 19.6 16.9 17.4 10.4CV(percent) .......................................... 2.5 3.3 5.1 5.8 5.3 4.3 5.2 3.8

MiscellaneousMean ................................................... 775.78 831.81 364.53 594.94 832.71 1,047.19 1,318.23 619.2SE ....................................................... 19.6 22.8 50.1 53.4 55.0 67.1 55.8 49.6CV(percent) .......................................... 2.5 2.7 13.7 9.0 6.6 6.4 4.2 8.0

Cash contributionsMean ................................................... 1,192.44 1,344.06 332.27 1,162.95 953.01 1,217.29 3,050.11 749.99SE ....................................................... 96.8 116.2 39.0 452.9 125.6 117.0 285.9 128.1CV(percent) .......................................... 8.1 8.6 11.7 38.9 13.2 9.6 9.4 17.1

Personal insurance and pensionsMean ................................................... 3,364.92 4,308.33 413.14 1,250.97 2,877.80 5,424.88 11,556.55 611.86SE ....................................................... 54.7 58.2 28.0 34.0 76.5 110.3 229.4 30.5CV(percent) .......................................... 1.6 1.4 6.8 2.7 2.7 2.0 2.0 5.0

Item Allconsumer

units

Completeincome

reporters

Lowest 20percent

Second-lowest 20percent

Third-lowest 20percent

Fourth-lowest 20percent

Incompleteincome

reporters

Highest 20percent

Complete reporting of income

1 Components of income and taxes are derived from complete income reporters only; see glossary.


Part II.

Analyses Using Survey Data


Consumer Spending forNecessities

Abby Duly is an economist in the Branch ofInformation and Analysis, Division of Con-sumer Expenditure Surveys, Bureau of LaborStatistics.

ABBY DULY The proportion of household 1

spending used to purchase ba-sic necessities is of interest to

policymakers and social researchers asan elementary indicator of economicwell-being. There are several complexi-ties, however, in this application of thedata; for example, the definition of“well-being” itself is not necessarilyuniversal, and, even when the term isdefined, the criteria upon which toevaluate well-being also are subjectiveand debatable. This article does notattempt to address these complexities;rather, data on consumer spending fornecessities are presented in a mannerthat may be interpreted by a variety ofreaders for a variety of uses.

The discussion that follows is or-ganized into three main sections. Thefirst is a description of the data, includ-ing the definition of “necessities” usedin this study and the demographic vari-ables chosen for comparison. The sec-ond section is an evaluation of the Prus-sian mathematical statistician ErnstEngel’s proposition, using data fromthe 2000 Consumer Expenditure (CE)Survey to determine whether the rela-tionship between income and the pro-

portion of expenditures spent on ne-cessities that Engel observed in 1857still holds true. In the third part of thetext, spending on necessities as a shareof total spending is presented for vari-ous additional demographic groups.

Study methodologyThe study uses the expenditure sharestables published in the CE Survey.These tables provide the proportionsof average annual expenditures (or to-tal spending) allocated to various cat-egories of items. The categories of in-terest here are those designated to benecessities: Food, housing, and ap-parel. These three types of expensesare chosen to be consistent with thework done by Engel, which, as previ-ously mentioned, is used as a basis foranalyzing spending for necessities byhouseholds of differing income levels.For consistency, the same definition ofnecessities is used in the comparisonsamong demographic groups. It is im-portant to note that, while food, hous-ing, and apparel are certainly reason-able candidates for necessities in 2000,there have been changes to thesespending categories over time. For ex-ample, within the necessity categoryof food, the allocation among subcom-ponents has shifted such that the shareof the food dollar spent on food awayfrom home (including meals at restau-rants or fast food, carryout, and homedelivery) has grown from 3.0 percent in

1 The basic unit of measurement in theConsumer Expenditure Survey is the con-sumer unit. (See the glossary at the end ofthis anthology for the definition of a con-sumer unit. For convenience, consumer unitand household are used interchangeablythroughout this article.)


First quintile Second quintile Third quintile Fourth quintile Fifth quintile0

5

10

15

20

25

30

0

5

10

15

20

25

30Housing Food Apparel

Percent Percent

1909, to 29.0 percent in 1987,2 to 41.0percent in 2000.3

While data on food and apparel pre-sented here are taken directly from thepublished CE tables, the housing cat-egory is constructed specifically outof two main subcomponents: Shelter4

and utilities. This is an important de-viation from the published data. Thereason is that, arguably, shelter andutilities are the actual necessities ofhousing and that other componentscollected in the CE Survey, such ashousehold furnishings and equipment,are not, in fact, basic goods.

In the next section, necessity sharesare compared across income quintiles,using Engel’s proposition as a guide.The final analyses presented here pro-vide a broad overview of necessityspending by additional demographicgroups: Homeowners and renters, ur-ban consumers and rural consumers,black households and white and otherhouseholds, Hispanic and non-His-panic households, consumer units liv-ing in different regions, and consumerunits of varying compositions.

Spending on necessitiesby income groupIn 1857, Engel observed a relationshipbetween household income and theproportion of total expenditures usedto purchase food, housing, and apparel.He found that, as income increases, theproportion of spending devoted to fooddecreases, while the shares of expen-ditures used to provide housing andapparel remain stable.5 Are the samepatterns visible in the most recent CESurvey? Chart 1 illustrates the shares

of total spending allocated to each ofthe three categories of necessity.

In support of Engel’s proposition,the share of average annual expendi-tures used to purchase food declinesfrom 14.9 percent to 11.6 percent as in-come increases from the third quintileto the fifth quintile. (See table 1.) How-ever, consumer units in the first quintileallocate a smaller proportion of totalspending to food (14.9 percent) thando consumer units in the second quin-tile (15.7 percent), which would seemto violate Engel’s proposition. But, aspublished by the CE Survey in 2000,the average income before taxes of thelowest income quintile is $7,683,whereas the average annual (total) ex-penditures for the same quintile are$17,940.6 Although this sounds con-tradictory, there are some explanationsfor the discrepancy. One is the effectof missing income: even though theresponses of complete income report-ers7 are used, the respondents may nothave provided a complete accountingof all income from all sources. Also,some consumer units in the lowestquintile—retirees and full-time stu-dents, for example—may be able tospend beyond their apparent means byusing loans or cashing in on invest-

ments that are not included as incomein the CE Survey. Therefore, cautionshould be used in interpreting the foodshare of the first income quintile as aviolation of Engel’s proposition.

Expenditure shares for housingclearly decline across income quintiles,as shown in chart 1. While consumerunits in the highest income quintiledevote 22 percent of their total spend-ing to shelter and utility costs, those inthe lowest income quintile spend almost30 percent. This pattern is not the sameone observed by Engel in 1857, and itmay be related to rather large differencesin housing tenure. In 2000, 57 percentof consumer units in the first incomequintile are renters, while 88 percent ofconsumer units in the fifth quintile arehomeowners.

The shares of average annual expen-ditures allocated to apparel are barelydiscernible in chart 1, supporting Engel’sobservation that spending on apparelremains stable across income levels. Infact, the range of apparel shares is lessthan 1 percentage point, from 4.7 per-cent spent by those in the lowest in-come quintile to 5.3 percent spent bythose in the highest income quintile.

Spending on necessitiesby selected demographiccharacteristicsAs mentioned previously, the share oftotal spending allocated to housing ismuch greater for lower income house-holds, and those households are alsomore likely to be renters. Looking at

2 Eva Jacobs and Stephanie Shipp, “Howfamily spending has changed in the U.S.,”Monthly Labor Review, March 1990, pp. 20–27.

3 “Table 1. Quintiles of income before taxes:Average annual expenditures and characteristics,Consumer Expenditure Survey, 2000,” http://www.bls.gov/cexann00.pdf, January 2003.

4 Shelter includes out-of-pocket expendi-tures for mortgage interest and charges, prop-erty taxes, rent, and maintenance and repairservices and commodities.

5 Louis Philips, Applied ConsumptionAnalysis: Revised and Enlarged Edition(Amsterdam, Elsevier Science Publishers,B.V., 1990), p. 103.

6 “Table 1. Quintiles of income beforetaxes: Average annual expenditures and char-acteristics, Consumer Expenditure Survey,2000.”

7 See “Glossary” in Appendix A at the endof this anthology for the definition of a com-plete income reporter.

Chart 1. Shares of average annual expenditures allocated to necessitiesby income quintile, Consumer Expenditure Survey, 2000


the data classified by housing tenure,one readily sees that consumer unitswho rent their homes also devote agreater share of their total expendituresto food (15.0 percent) and apparel (5.4percent) than do their homeowningcounterparts (13.1 percent and 4.7 per-cent, respectively).

Urban consumers spend a higherproportion of their total expenditureson housing (25.7 percent, as opposedto the 21.2 percent spent by consum-ers living in rural areas) and on apparel(5.0 percent, compared with 4.2 percent,respectively). Food, however, makes upa slightly greater proportion of totalspending among rural households (14.1percent) than among urban households(13.5 percent).

Race and Hispanic origin, which arebased on the reference person8 of theconsumer unit, are the next demo-graphic characteristics listed in thetable. Black consumer units spendhigher shares of total expenditures onall three of the necessity categoriesthan do white and other9 consumerunits. The same is true for Hispaniccompared with non-Hispanic house-holds, although the relevant housingshares are not very different, with His-panic consumer units allocating 26.3percent of total spending to housing and

Table 1. Shares of average annual expenditures allocated to necessities, byselected demographic characteristics, Consumer Expenditure Survey, 2000

Characteristic Food Housing1 Apparel

All consumer units ................................................... 13.5 25.2 4.9

Income quintile2

First ............................................................................... 14.9 29.9 4.7 Second .......................................................................... 15.7 25.7 4.9 Third .............................................................................. 14.9 25.0 4.7 Fourth ........................................................................... 13.8 22.9 4.8 Fifth ............................................................................... 11.6 22.0 5.3

Housing tenure Homeowner .................................................................. 13.1 24.2 4.7 Renter ........................................................................... 15.0 28.5 5.4

Type of area Urban ............................................................................ 13.5 25.7 5.0 Rural ............................................................................. 14.1 21.2 4.2

Race of reference person White and other ............................................................. 13.5 24.8 4.8 Black ............................................................................. 14.5 29.3 6.0

Hispanic or non-Hispanic origin of reference person Hispanic ........................................................................ 16.4 26.3 6.3 Non-Hispanic ................................................................ 13.3 25.2 4.8

Region of residence Northeast ...................................................................... 13.8 27.7 5.4 Midwest......................................................................... 13.4 23.3 4.9 South ............................................................................. 13.6 24.3 4.7 West .............................................................................. 13.4 26.4 4.7

Composition of consumer unit Husband and wife only ................................................. 13.2 23.3 4.1 Husband and wife with oldest child under 6 ................. 11.5 26.0 5.1 Husband and wife with oldest child 6 to 17 ................... 13.9 24.4 5.2 Husband and wife with oldest child 18 or older ............ 14.4 22.0 4.9 One parent with at least one child under 18 ................. 14.7 30.0 6.6 Single-person and other consumer units ..................... 13.4 27.9 4.9

1 Shelter plus utilities.2 Complete income reporters only.

non-Hispanics allocating 25.2 percent.There is little variation in the neces-

sity shares of consumer units living indifferent regions. For example, therange of expenditure shares used topurchase food is from 13.4 percent inthe West and Midwest to 13.8 percentin the Northeast. (Households in theSouth spend a comparable 13.6 percenton food). Housing shares across re-gions are more variable, with consumerunits in the Midwest having the low-est share (23.3 percent) of total spend-ing and consumer units in the North-east region having the highest share(27.7 percent).

Chart 2 depicts the shares of aver-age annual expenditures allocated tonecessities by the composition of theconsumer unit.The household typesselected for this analysis are husbandand wife only, husband and wife withthe oldest child under 6 years of age,husband and wife with the oldest childbetween the ages of 6 and 17, husbandand wife with the oldest child aged 18or older, and single parents with at leastone child under the age of 18 years.(Table 1 also provides data for single-person and other consumer units.) Thechart indicates that single parents de-vote greater proportions of their totalspending to food (14.7 percent), hous-ing (30.0 percent), and apparel (6.6 per-cent) than do other types of household.Also, the age of the oldest child in thehousehold is inversely related to theshare of total spending allocated tohousing and directly related to theshare allocated to food. Interestingly,the expenditure share for food is greaterfor husband-and-wife-only consumerunits (13.2 percent) than for those withyoung children (11.5 percent). This dif-ference is attributable to a decline infood away from home, as parents ofyoung children may not eat outside ofthe home as often, or in restaurants asexpensive, as do couples without chil-dren.10

8 See the glossary at the end of this anthol-ogy for the definition of reference person.

9 The “other” race group includes NativeAmericans, Alaska Natives, Asians, and Pa-cific Islanders.

10 The expenditure shares for food at homeare roughly equivalent for husband-and-wifeconsumer units (7.5 percent) and householdswith children under 6 years of age (7.2 per-cent). However, the former allocate 5.7 per-cent of total spending to food away from homewhile the latter allocate just 4.3 percent.


0

5

10

15

20

25

30

35

0

5

10

15

20

25

30

35

Food Housing Apparel

Percent

Oldest child between ages 6 and 17, inclusive

Oldest child age 6 or younger

Husband and wife only

Single parentOldest child 18 or older

Percent

In sum, this article has presented avariety of data on spending for neces-sities as a proportion of total expendi-tures, from the 2000 Consumer Expen-diture Survey. With respect to Engel’sproposition, the expected trends areobserved for food and apparel, while acontradictory decrease in housingshares occurs as income increases.Necessity spending also varies amongconsumer units with different demo-graphic characteristics.

Chart 2. Shares of average annual expenditures allocated to necessitiesby consumer unit composition, 2000


Consumer Expendituresfor Alcohol in 2000

GEOFFREY PAULIN

Geoffrey Paulin is a senior economist in theBranch of Information and Analysis, Divi-sion of Consumer Expenditure Surveys, Bu-reau of Labor Statistics.

In 2000, per capita consumption ofalcoholic beverages was 24.9 gal-lons, mostly in the form of beer (21.7

gallons).1 That same year, according tothe Consumer Expenditure (CE) Survey,the average consumer unit2 reportedexpenditures of $372 for alcoholic bev-erages; that is, about $1 was spent onalcohol for every $8 spent on food athome.3 Other recent studies have citedsimilar figures, as well as health andsocial concerns, as reasons for study-ing the consumption of alcoholic bev-erages.4 These studies examine eitherthe consumption of a specific beverageby a specific group or the consumptionof alcohol in countries other than theUnited States. By contrast, this articlefocuses on U.S. domestic consumer ex-penditures on alcohol in 2000—specifi-cally, the demographic patterns involved,the mean weekly expenditure on alco-hol, the probability of purchase of alco-

hol either at home or away from home(such as a drink at a restaurant or bar),and the type of alcohol purchased(beer, wine, or other alcohol, such aswhiskey).

The DataData for the CE Survey are derived fromtwo sources: The Interview survey,which is a rotating-panel quarterly re-call survey, and the Diary survey, inwhich respondents record all their ex-penditures during the 2-week surveyperiod. Data from the two sources areintegrated into tables for analysis andsubsequent publication. The data forthis article are taken from the Diary com-ponent of the 2000 CE Survey. In thepublished CE Survey, one item—alco-holic beverages purchased on trips—is taken from the Interview component.However, this item (which is collectedsolely in the Interview survey) ac-counts for only about $34, or less than10 percent of average total expendituresfor alcohol in 2000, so it is safe to ex-clude it from the current analysis. Us-ing only Diary data also allows the re-gression results (described later) to becompared with the expenditure dataexamined herein.

Caution should be exercised in at-tempting to interpret some of the datashown. Expenditures for alcohol aresubject to a great deal of “allocation”during the publication process. Thatis, when a respondent records “expen-ditures for alcohol” or “meal at restau-

1 Statistical Abstract of the United States,2002 (U.S. Census Bureau, 2002), p. 130,table 197, “Per Capita Consumption of Se-lected Beverages by Type: 1980 to 2000.”

2 See the glossary at the end of this an-thology for the definition of a consumer unit.

3 Consumer Expenditure Survey, 1999–2001, Report 966 (Bureau of Labor Statis-tics, April 2003), table A, “Average annualexpenditures of all consumer units and per-cent changes,” p. 3.

4 J. R. Blaylock and W. N. Blisard, “WineConsumption by U.S. Men,” Applied Econom-ics, May 1993, pp. 645–51; and MohamedAbdel-Ghany and J. Lew Silver, “Economicand Demographic Determinants of CanadianHouseholds’ Use of and Spending on Alco-hol,” Family and Consumer Research Jour-nal, September 1998, pp. 62–90.


rant, including alcohol,” but providesno details on the type or amount of al-cohol purchased, the expenditure is es-timated on the basis of the total expen-diture reported by the respondent foralcohol or the meal at the restaurant,together with an allocation factor thatis in turn based on responses fromthose who record specifically what waspurchased. At the aggregate level, thistechnique presumably has little impacton total expenditures for alcohol, but itcould cause a larger share of those ex-penditures to be accounted for by ei-ther beer, wine, or other alcoholic bev-erages than is actually the case; inaddition, at the individual-record level,a consumer unit might show expendi-tures for beer, wine, and other alcohol,even though that consumer unit pur-chased only one of those items. Forexample, suppose a respondent pur-chases beer for $10 and records a $10expenditure for alcohol. Then, becausethe fact that all $10 went for beer is notrecorded, the consumer unit mightshow expenditures of $7 for beer, $2 forwine, and $1 for other alcohol, assum-ing allocation factors of 70 percent forbeer, 20 percent for wine, and 10 per-cent for other alcohol. The actual num-ber of records created through alloca-tion as opposed to reporting varies bythe type of alcohol purchased. (For ex-ample, 43 percent of beer-at-home re-ports5 are the result of allocation, com-pared with 76 percent of wine-at-homereports and 92 percent of other-alco-hol-at-home reports.) Overall, about 46percent of expenditures reported forspecific types of alcohol are created byallocation from general reports of alco-hol either at home or away from home.

MethodologyThis article investigates expendituresin several ways. First, expenditure val-ues and the percent of consumer unitsthat report purchasing alcohol (that is,the percent reporting) are examined forthree demographic categories: Incomequintile, age of reference person, andsex of reference person for single con-sumers only.6 The analysis is extendedthrough the use of logistic regression,or “logit,” a technique that enables oneto predict the probability that an event(in this case, the purchase of alcohol)will occur, given that certain conditions(in this case, demographic characteris-tics) are held constant. By means ofregression analysis, it is possible toisolate relationships between thesecharacteristics and the probability ofpurchase of some kind of alcohol. Forexample, the probability of purchasingwine rises steadily with income and in-creases with age until the referenceperson is 45 to 54, after which it de-creases with age. Given that incomealso increases with age until the refer-ence person is 45 to 54 and starts todecrease with age thereafter, it is diffi-cult, in the absence of regression analy-sis, to say which characteristic—age

or income—is more strongly related tothe purchase of alcohol. Logit is usedto estimate the probability of purchas-ing alcohol in general, as well as that ofpurchasing alcohol at home, away fromhome, or both. Logit also is used topredict the probability of purchasingbeer, wine, or some other alcoholic bev-erage. (The appendix to this article de-scribes the use of logit in more detail.)

Except for the data in the logit analy-ses, the data used in this article areweighted to reflect the population. (Thereasons why the data employed in thelogit analyses are not weighted will bepresented shortly.) The data used inthe article also are limited to consumerunits whose reference person is at least21 years old—that is, the legal age topurchase alcohol in the United States.(Those under the legal age may be morelikely than those who are at least 21years old to omit expenditures for alco-hol from their diaries.) Specific incomedata (such as mean values and quintileassignments) are derived from com-plete income reporters only, unless oth-erwise specified.7 For best results, fami-lies that reported income losses (forexample, through self-employed busi-ness loss or rental property loss) alsoare excluded from the sample.8

5 The CE Survey uses the terminology “athome” and “away from home” to describeplaces at which goods are purchased, ratherthan where they are ultimately consumed.For example, when an expenditure is reportedfor “food at home,” it means that the foodwas purchased at a grocery store or similarvendor, rather than at a restaurant, cafete-ria, or bar. The food purchased may havebeen consumed elsewhere—for example, aperson buys fruit and takes some to the of-fice for lunch or packs a sandwich for thechild’s lunch at school. Even though the foodwas not eaten in the home, the food waspurchased at a grocery store and is therefore

7 See “Glossary” in Appendix A at the endof this anthology for the definition of com-plete income reporter.

8 The income used in the CE Survey re-sults is found by summing the value of allsources of income reported. When lossesoccur, the negative income is added to thetotal (or the loss is subtracted, depending onhow one looks at it), which has the result ofartificially lowering total income. Sometimes,the losses are large enough to cause total in-come to be negative. Losses make compari-son across consumer units difficult. For ex-ample, a family in which one member re-ceives $50,000 in salary appears to have thesame income as another family in which onemember receives $75,000 in income, but inwhich another member incurs a loss of$25,000. Both consumer units have $50,000in income, according to the survey results,but each may have different spending pat-terns; the losses may be temporary and an-ticipated, for example, causing the consumerunit incurring the losses to spend differentlythan the unit that regularly receives $50,000in income. Including the loss could substan-tially increase the variance for the incomedata and could also bias parameter estimatesin the regression section. For these reasons,consumer units reporting losses are omittedfrom the sample.

designated as “food at home.” Similarly, whena person has a pizza delivered from a localrestaurant, the amount paid is classified as anexpenditure for “food away from home,”despite the fact that the pizza was eaten infront of the living room television. The rea-son is that the vendor was a restaurant. Withalcoholic beverages, the same rules apply. Anexpenditure for beer, wine, or other alcoholthat is purchased from a grocery, liquor, orconvenience store is considered an expendi-ture for “alcohol at home,” even though thepurchaser may have taken the bottle of wineto a dinner party or taken the beer to a localpark to drink at a picnic or while watching asoftball game. In the case of alcohol, how-ever, it is not likely that alcohol classified as“away from home” would have been con-sumed inside the home, because restaurantsand bars usually restrict alcohol purchased tobe consumed on the premises. For consis-tency with the classifications used in the CESurvey, the terms “at home” and “away fromhome” will be used in this article to describeexpenditures for alcohol, regardless of wherethe alcohol was actually consumed.

6 See “Glossary” in Appendix A at the endof this anthology for the definitions of refer-ence person and income quintile.


Demographic analysisBy any measure shown in table 1, beeris the most popular form of alcoholpurchased by the average consumerunit. Whether one looks at percent re-porting or mean weekly expenditure,beer is at the top of the list, both athome and away from home. However,this ranking changes when one looksat the mean weekly expenditure of onlythose consumer units reporting pur-chases of alcohol, a figure that can becalculated by dividing mean weeklyexpenditure by percent reporting. Inthis case, the largest average expendi-ture for all consumer units is for wineat home ($23.29). Other alcohol at homeis second ($19.36), with beer at home adistant third ($16.39). In contrast, thelargest expenditure for alcohol awayfrom home is for other alcohol ($12.08).The smallest expenditure obtained byusing this measure is that for wine awayfrom home ($9.73).

Income. As one might expect, expendi-tures for alcohol increase with income.(See table 2.) This statement holds trueregardless of the type of alcohol pur-chased and regardless of whether it isalcohol at home or away from home.What is more interesting is the rate ofincrease with income. For example,while the fifth income quintile spendsabout 3.5 times as much for alcohol asdoes the first income quintile, it spendsonly 2.7 times as much for alcohol athome, compared with more than 7.1times as much for alcohol away fromhome. When the types of alcohol pur-chases are analyzed, the ratios of thefifth to the first income quintile rangefrom 1.6 (for beer at home) to 9.2 (forother alcohol away from home).

The percent reporting follows a simi-lar pattern. For alcohol at home, thepercent reporting for the fifth quintile(29.1 percent) is more than double thepercent reporting for the first quintile(11.9 percent). For alcohol away fromhome, the differences across quintilesare even more dramatic, ranging from6.9 percent for quintile 1 to 25.6 per-cent for quintile 5. The smallest rangeis for other types of alcohol at home,which only doubles from the lowest to

the highest quintile (2.3 percent to 5.1percent). The largest range in absoluteterms is for beer away from home (6.1percent to 22.6 percent). However, thepercent reporting other alcohol awayfrom home is still more than 6 timeshigher for the fifth quintile (11.8 per-cent) than it is for the first (1.8 percent).

Age. In all cases, expenditures for alco-hol away from home rise with age up toa point and then decline. (See table 2.)The pivotal age group is the one whosereference persons are 35 to 44 yearsold. For alcohol at home, wine followsthe pattern, except that expenditurespeak for those aged 45 to 54. However,expenditures for beer and other (thatis, nonwine) alcohol at home actuallydecline with age. For beer at home, ex-penditures range from a high of $5.48for the under-25 group to a low of $0.65for the 75-and-older group, a decreaseof 88 percent over that entire age range.Stated another way, the youngestgroup spends 8.4 times as much for beerat home as does the oldest group. Thepercent of those reporting expendituresfor beer at home follows a similar pat-tern: nearly 1 in 4 consumer units in theyoungest group report such expendi-tures, compared with fewer than 1 in 20consumer units in the oldest group.Most other expenditures for alcoholicbeverages follow the same pattern forpercent reporting, peaking either for theunder-25 group or the 25- to 34-year-old group. The lone exception is wine:the percent reporting expenditures forwine peaks with the 45- to 54 year-oldgroup (13 percent), and the group withthe lowest percent reporting is againthe 75-and-older group (6 percent). Thepercent reporting wine away from homeis only about 4 to 5 percent for thoseunder 65, but decreases for those aged65 and older (of whom less than 2 per-cent report such expenditures).

Singles. Single individuals spend theirmoney differently than do nonsingles.(See table 3.) Interestingly, though,when the data are classified by the sexof the reference person, it becomesclear that single men spend more, onaverage, than do nonsingles (of both

sexes) for all alcoholic beverages, ex-cept wine at home, while single womenspend less than non-singles on all al-coholic beverages (including wine athome). The same pattern holds for thepercent of consumer units reportingexpenditures on alcohol. That is, exceptin the case of wine at home, single menhave the largest percent reporting, fol-lowed by nonsingles and then singlewomen. The difference also affects thetotal percent reporting expenditures forwine generally, but here single men runa close second (10.3 percent reporting)to nonsingles (10.7 percent reporting),with fewer single women reporting pur-chases (6.2 percent).

Predicted probabilitiesGiven the similarity in trends for expen-ditures for alcoholic beverages at homeand for those away from home (for ex-ample, percent reporting increasessteadily with income for both types ofpurchase), logit is used only to ana-lyze total purchases of beer, wine, andother alcohol once the probability ofpurchase for alcohol in general is ex-amined by type of purchase. Accord-ingly, the first set of analyses to followexamines the probability of purchasingalcohol in general. The rest of the analy-ses examine probabilities of purchas-ing specific beverages. In other words,what is the probability of purchasingalcohol at home as opposed to the prob-ability of purchasing alcohol away fromhome? What is the probability that aconsumer will purchase both alcoholat home and alcohol away from home,rather than one or the other? What isthe probability of purchasing beer,wine, or other alcohol? The results ofthe logits, used to answer these ques-tions, should be interpreted with cau-tion. Those who did not purchase al-cohol may have chosen not to do sofor any number of reasons, includingthe fact that they had enough liquor inthe cabinet to last for the week duringwhich they filled out the diary or thatthey may be persons who choose notto consume alcohol on any occasionat all. Because it is not possible to dis-tinguish “potential” purchasers from“nondemanders” in the Diary survey,


the answers can be interpreted to pre-dict only the probability of actual pur-chase during the previous week, ratherthan the probability of actual use (ornonuse) of alcohol by the consumerunit over longer periods.

Also, unlike the data in the previ-ous section, the logit results here arenot weighted to reflect the population.Previous experience has shown thatweighting logistic regressions for thatpurpose yields parameter estimatessimilar to the unweighted results, butwith much smaller standard errors. Thishas the effect of making every param-eter estimate appear to be statisticallysignificant. Therefore, to be conserva-tive in the estimates, unweighted re-gressions are used to estimate prob-abilities of purchase in this article.

In using regression analysis, a “con-trol group” is standardly identified toserve as a reference point for the analy-sis. In this article, parameter estimatesthat have negative coefficients are pre-dicted to have lower probabilities ofpurchase than the control group, whilethose with positive coefficients have ahigher predicted probability of pur-chase than the control group. Here, thecontrol group consists of consumerunits whose reference person (1) is 35to 44 years old; (2) reports income inthe middle quintile; (3) is a single, white,non-Hispanic male employed as a man-ager or professional receiving a wageor salary; (4) owns his home, but paysa mortgage; and (5) is living in the ur-ban South. Comparisons with the con-trol group are made by changing onecharacteristic at a time; for example, inattempting to find the relationship of re-gion of residence to purchases of alco-hol, one assumes that all characteristicsof the members of the group to be testedare identical to those of the members ofthe control group (that is, every memberof each group is a single, white, non-Hispanic male, aged 35 to 44 years old,with an income in the middle quintile,and so forth), except that the membersof the group to be tested live in the North-east instead of the South. Such compari-sons are known as “ceteris paribus” com-parisons in economics—comparisons inwhich “all else is held equal.”

General purchases of alcohol. Theprobability of purchasing alcohol forthe general adult population appearsto follow the trends already described,at least with respect to age, income, andsex of the reference person. That is, thepredicted probability of purchase,which is about 38 percent for the con-trol group, is highest for the youngestgroup (46 percent) and lowest for theoldest group (22 percent). Similarly, theprobability of purchase is lowest forthe first income quintile (29 percent)and highest for the fifth (50 percent).Single women are less likely to pur-chase (23 percent) than are single men(38 percent).

The logit regressions also allowcomparisons across a variety of othercharacteristics. For example, ethnicityappears to have little relationship to theprobability of purchasing alcohol ingeneral: the parameter estimate for “His-panic” is small in magnitude (–0.0628)and is not statistically significant. Race,by contrast, appears to play a role inprobability of purchase: black andAsian consumers have much lowerprobabilities of purchase than do whiteconsumers, and those of other racesappear to be similar to Asians in theirpurchasing behavior. (The coefficientassociated with “other race” is nearlyequal to that of Asians, while it is notstatistically significant.) Occupationhas a less strong relationship: althoughpersons in technical, sales, or servicepositions and those in agriculturalfields (forestry and farming) have posi-tive, statistically significant coeffi-cients, no other working group is pre-dicted to be statistically significantlydifferent from salaried (or wage-earn-ing) managers and professionals in theirpurchases of alcohol in general. Ofthose who do not work, retirees have afairly small coefficient that is not sta-tistically significant. The long-term un-employed9 have a large, but not statis-

tically significant, negative coefficient,indicating that they are a lot less likelyto purchase than are managers andprofessionals. The sample size for thisgroup is small, so it is difficult to saywhether the negative relationship isindicative of the general population inthe group. However, those who are notworking for reasons other than thatthey are a member of the long-term un-employed (for example, they may beattending school, working without pay,too ill to work, or doing something else)also have a large negative coefficientthat, this time, is statistically significant.The predicted probability of purchasefor this group is 31 percent, comparedwith 38 percent for managers and pro-fessionals. Finally, the South appearsto be the region with the lowest prob-ability of purchasing alcohol (38 per-cent); persons in other regions havepredicted probabilities ranging from 44percent to 46 percent. Rural men areabout 9 percent less likely than theirurban counterparts to purchase alco-hol. (That is, their predicted probabil-ity of doing so is 29 percent, about 9percentage points lower than that ofurban single men.)

Probabilities for specific purchases ofalcohol. The remaining sets of regres-sion results are for specific types ofalcohol purchase—at home, away, orboth; and for beer, wine, or other alco-hol. Once again, several demographiccharacteristics appear to be related tothe probability of purchase. For ex-ample, the probability of purchasingalcohol at home is negatively relatedto age, as is the purchase of alcohol ingeneral. The youngest age group hasa 30-percent predicted probability ofpurchase at home compared with a 12-percent probability for the oldestgroup. The coefficients for each ofthese groups are statistically significantat the 99-percent confidence level, asare all of the age coefficients, with theexception of the 25- to 34-year-old agegroup (significant at the 95-percentlevel) and the 45- to 54-year-old agegroup (not statistically significant). In-come, by contrast, is positively relatedto the purchase of alcohol at home,

9 The survey question on occupation asksat what profession the person earned themost money in the previous year. If the ref-erence person received unemployment in-surance and then did not work or worked onlysporadically, the person could be reported tohave “earned” the most through unemploy-ment.


ranging from 18 percent for the lowestquintile to 29 percent for the highest.Interestingly, the presence of childrenor a single adult woman in the homeappears to lower the probability of pur-chasing alcohol at home. Single men(the control group) have a predictedprobability of purchase of 24 percent,while single women have only an 11percent probability. Single mothershave an even lower predicted probabil-ity: 9 percent. Husband-and-wife fami-lies with children have a lower prob-ability of purchasing alcohol at home(20 percent) than the 24-percent prob-ability of single men. Families with ahusband and wife only, however, witha 23-percent probability of purchasingalcohol at home and a coefficient thatis not statistically significant, are simi-lar to single men in that type of pur-chase. Like husband-and-wife-onlyfamilies, other-husband-and-wife fami-lies in which children are present havea predicted probability of purchase ofalcohol at home of 23 percent, with acoefficient that is not statistically sig-nificant.) Here, too, ethnicity appearsto play no role in the probability of pur-chase, but race does: both black andAsian families have a lower predictedprobability of purchase (18 percent)than that of the control group, and bothcoefficients are statistically significantat the 99-percent confidence level. Fami-lies of other nonwhite races appear tohave a similarly lower probability (17percent), but their coefficient is not sta-tistically significant. Occupation alsoappears to play a role: technical, sales,and service workers (29 percent), aswell as blue-collar workers (28 percent),have slightly higher probabilities ofpurchasing alcohol at home than domanagers and professionals (24 per-cent); however, agricultural workers (40percent) and armed-service workers (41percent) have substantially higherprobabilities of purchase. Work status,by contrast, plays less of a role: theself-employed, with a probability ofpurchase of 24 percent, are not statisti-cally significantly different from wageor salaried families, and, although retir-ees are predicted to have a higher prob-ability of purchase (29 percent) than

wage or salaried families, those who areunemployed or who are not workingfor another reason are not statisticallydifferent from wage or salaried families.Region plays a role (the Northeast hasthe highest predicted probability ofpurchasing alcohol at home, 28 per-cent), as does degree of urbanization(with rural “control” families 7 percentless likely than similar urban families topurchase). Finally, the purchase of al-cohol away from home is also positivelyrelated to the purchase of alcohol athome. The coefficient is positive andsignificant at the 99-percent level.However, it is so small (0.0173), that itis economically not significant in its re-lationship to the probability of pur-chase.

For purchases of alcohol away fromhome, the findings are similar, but notidentical. First, the probability of pur-chase is lower (21 percent) for the con-trol group in this case than it is for theprobability of purchase of alcohol athome (24 percent). Second, the prob-ability of purchase of alcohol away fromhome is higher for 25- to 34-year-oldsthan for those under 25, but it peaksfor the former (at 27 percent) and de-clines with age thereafter. It is also posi-tively related to income, but the rangeof predicted probabilities is wider (from14 percent to 33 percent) than it is foralcohol purchased at home. Althoughhusband-and-wife-only families are notstatistically significantly different fromsingle men in respect of purchasing al-cohol away from home, all other typesof family are. Single women have a 16-percent predicted probability of pur-chase, compared with 21 percent forsingle men. The presence of childrenalso appears to be related to the prob-ability of purchase, with single fathers,single mothers, and husband-and-wifefamilies with their own children only allhaving a lower probability of purchas-ing alcohol away from home (12 per-cent) than single men without children.Other husband-and-wife families withchildren have a higher probability ofpurchase (16 percent), but it is stilllower than that for single men. Perhapsthis is because the other members ofthe consumer unit also are likely to be

adults (such as the parent or sibling ofone of the spouses), and, therefore, theadditional adults contribute to the to-tal probability of purchasing alcoholaway from home. Unlike its weak rela-tionship to alcohol purchased at home,ethnicity now is strongly related to theprobability of purchase. Hispanics (15percent) have a much lower probabil-ity of purchase than do non-Hispanics(21 percent); the same is true for Asians(16 percent) and, especially, blacks (11percent). Region makes a difference,but now the Midwest is the region withthe highest predicted probability ofpurchase (26 percent). Rural families arestill less likely to purchase (18 percent),and the purchase of alcohol at homealso makes a statistically, but not eco-nomically, significant difference in theprobability of purchasing alcohol awayfrom home.

The probability of purchasing bothalcohol at home and alcohol away fromhome is only about 12 percent. Theprobability of purchasing both appearsto be negatively related to age: theyoungest group (those under 25) hasthe largest coefficient, but it is not quitesignificant at the 95-percent confidencelevel. Taken at “face value,” though(that is, without regard to statistical sig-nificance), the predicted probability forthe youngest group is 16 percent, com-pared with 5 percent for the oldestgroup (75 and older). The positive rela-tionship to income still holds, with thepredicted probability of purchase rang-ing from 7 percent to 20 percent. Again,the presence of children or a singlewoman appears to lower the probabil-ity of purchasing alcohol for both pur-poses. Single women have a predictedprobability of purchase of 5 percent,while single mothers have an evenlower 3-percent probability. The low-est probability of all, however, is thatfor single fathers: 2 percent. Marriedcouples whose children are biologicallyrelated to both parents or have beenjointly adopted by them have a 6-per-cent probability of purchasing both al-cohol at home and alcohol away fromhome. This probability, although largerthan that for single parents, is still onlyabout half the predicted probability for


single men (12 percent). Hispanics alsohave a lower predicted probability ofpurchase (9 percent) than do non-His-panics (12 percent), but race lowers thepredicted probability even more: bothblacks and Asians are about half aslikely (6 percent) as whites to purchaseboth alcohol at home and alcohol awayfrom home. Finally, neither occupationnor region plays a major role in the pre-dicted probability of purchase. Ruralconsumers (9 percent) appear to be lesslikely than urban consumers (12 per-cent) to purchase alcohol for both pur-poses, but the coefficient is significantonly at the 10-percent confidence level.

It is also interesting to examine pre-dicted probabilities for purchasing spe-cific types of alcohol. Although, inthese regressions, the same variablesare retained as predictors of probabil-ity, three new independent variables areadded to each equation. The first twoare binary variables and indicate thatthe purchaser purchased some othertype of alcohol than the type understudy. For example, in predicting theprobability of purchasing beer, the firstbinary variable describes whether theconsumer unit did or did not purchasewine, and the second variable de-scribes whether the consumer unit didor did not purchase other alcohol. Inpredicting the probability of purchas-ing wine, the first binary variable de-scribes whether the consumer unit didor did not purchase beer, and the sec-ond describes whether the consumerunit did or did not purchase other alco-hol. And in predicting the probabilityof purchasing other alcohol, the firstbinary variable describes whether theconsumer unit did or did not purchasebeer, and the second describes whetherthe consumer unit did or did not pur-chase wine. The third term is an “inter-action term” indicating that the con-sumer unit purchased both remainingtypes of alcohol, given the particulardependent-variable alcohol. (For ex-ample, if the probability of purchasingbeer is being predicted, the interactionterm will be equal to unity if the con-sumer unit purchased both wine andother alcohol, but will be equal to zeroif the consumer unit bought only wine

or other alcohol or bought neither winenor other alcohol.) These variables areadded to the analysis to see whetherdifferent types of alcohol are “substi-tutes” or “complements,” at least interms of their probability of purchase.Once again, the total sample includesall consumers who purchased at leastsome type of alcohol during the weekthey filled in the diary.

Beer. As mentioned earlier, beer is themost popular alcoholic beverage. Theparameter estimate associated with theintercept is –1.1944, indicating that thecontrol group’s predicted probabilityof purchasing beer is 23 percent. Theprobability of purchase is strongly re-lated to age, declining from 29 percentfor the youngest group (under 25) to10 percent for the oldest group (75 andolder). The probability of purchasealso is related to income, although onlythe lowest and highest quintiles havestatistically significant coefficients.The probability for the lowest quintileis 17 percent, compared with 27 per-cent for the highest quintile. Single menare again the most likely to purchasebeer (23 percent), single women (12 per-cent) and single mothers (9 percent)the least likely. Married couples with-out children are not different fromsingle men to a statistically significantdegree, but when children are addedto the family, the probability of pur-chase drops slightly, to 17 percent.When ethnicity and race are consid-ered, only blacks (16 percent) are sig-nificantly different from the controlgroup. Among salaried workers, occu-pation makes a difference, with techni-cal, sales, and service workers (28 per-cent), blue-collar workers (30 percent),agricultural workers (35 percent), andmembers of the armed services (38 per-cent) all having higher predicted prob-abilities of purchasing beer than domanagers or professionals (23 per-cent). Neither the self-employed nornonworkers are significantly differentfrom wage and salaried workers, al-though retirees appear to have a higherprobability of purchasing beer (28 per-cent) than do wage and salaried work-ers. (The coefficient is positive, but

statistically significant only at the 90-percent level.) The Midwest has thehighest probability of purchase (29percent), and the purchase of wine (57percent) or of some other alcohol (65percent) strongly increases the prob-ability of the purchase of beer. How-ever, the purchase of both wine andanother alcohol does not significantlyincrease the probability beyond whatis predicted when the coefficient forpurchasing wine alone and that forpurchasing another alcohol alone areincorporated into the equation. (Thatis, without including the interactioneffect, a member of the control groupwho purchases both wine and anotheralcohol has a predicted probability ofpurchasing beer of 89 percent. Whenthe interaction term is incorporated, theprobability rises to 91 percent. This 2-percent difference is not statisticallysignificant, because the coefficient forthe interaction term is not statisticallysignificant.)

Wine. The probability of purchasingwine is much lower than the probabil-ity of purchasing beer: only 1 in 20 con-sumer units (5 percent) in the controlgroup is predicted to buy wine duringthe week its respondent fills out thediary. Age does not appear to bestrongly related to the purchase ofwine, although 45- to 54-year-olds havethe only statistically significant coeffi-cient and thus the highest predictedprobability of purchase of any agegroup. However, at 6 percent, this dif-ference is not economically significant.The probability of purchasing wine in-creases with income, although only thehighest quintile has a statistically sig-nificant coefficient associated with it.Once again, without regard to statisti-cal significance, the lowest quintile hasa predicted probability of purchase of4 percent, compared with a predictedprobability of purchase of 7 percent forthe highest quintile. Family type is notrelated to the purchase of wine to a sta-tistically significant degree, whileethnicity is perhaps weakly related: thepredicted probability for Hispanics (4percent) is different from the probabil-ity for non-Hispanics (5 percent) only


at the 10-percent confidence level.However, blacks (4 percent) and Asians(3 percent) do have statistically signifi-cant coefficients at the 95-percent con-fidence level. (The coefficient forAsians actually is significant at the 99-percent confidence level.) Occupationplays little role; although blue-collarworkers have the lowest predictedprobability of purchasing wine (3 per-cent) of all working consumers. Simi-larly, those who are not working for rea-sons other than retirement orunemployment have a lower probabil-ity than other groups (3 percent). Re-gion plays little role in predicting theprobability of purchasing wine, but ru-ral consumers also are less likely (3percent) than urban consumers (5 per-cent) to purchase. However, both thepurchase of beer (18-percent probabil-ity) and the purchase of other alcohol(17-percent probability) substantiallyincrease the probability of purchasingwine. Nevertheless, purchasing bothbeer and some other alcohol adds littleto the probability of purchasing abovewhat purchasing beer or another alco-hol alone adds.

Other alcohol. As with wine, the pre-dicted probability of purchasing otheralcohol is low—only 4 percent for thecontrol group. However, demograph-ics play a larger role in predicting theprobability of purchasing some otheralcohol than wine, in that more coeffi-cients are statistically significant.

Although age does not have a sta-tistically significant relationship to theprobability of purchasing some alco-hol other than wine or beer, both thefourth and fifth income quintiles (6 per-cent) are more likely to purchase thanis the control group. Family type playsa role as well, with female-headed con-sumer units having lower predictedprobabilities (3 percent for singlewomen and 2 percent for single moth-ers) than do single-male households.In addition, husband-and-wife coupleswith their own children only have alower predicted probability of purchas-ing some other alcohol (2 percent) thanhave single men. Hispanics and Asiansboth have lower predicted probabilities

(2 percent) than do white non-Hispan-ics (4 percent). In respect of occupa-tion, only blue-collar workers have astatistically significant coefficient, witha predicted probability of purchase of3 percent. By region, only the Midwesthas a statistically significant coeffi-cient, raising its probability of purchas-ing some other alcohol to 5 percent.Once again, the predicted probabilityof purchase rises sharply when eitherbeer (21 percent) or wine (16 percent)is purchased, but purchasing both beerand wine has no additional effect onthe probability of purchasing someother alcohol than is accounted for byincluding the coefficients for purchas-ing beer and wine separately. (That is,the expenditures on alcohol of thosewho purchase beer, but not wine, orwine, but not beer, are not statisticallysignificantly different from those whopurchase both beer and wine.)

SummaryThis article has examined expendituresfor alcohol from several perspectives,including mean weekly expenditures,percent reporting expenditures, andpredicted probability of purchase forconsumers with different demographiccharacteristics. Expenditures for alco-hol are analyzed both by place of pur-chase (at home or away) and by typeof alcohol purchased (beer, wine, andother alcohol, such as whiskey). Con-sistent with national sales figures, beerappears to be the most popular form ofalcohol purchased, both at home andaway from home. Beer has the largestaverage weekly expenditure for all con-sumer units and the largest percent ofall consumer units reporting the pur-chase of alcohol. However, when theaverage expenditure for those who ac-tually purchase alcohol is examined,wine has the largest average expendi-ture, followed by other alcohols.

Expenditures for alcohol at home risesubstantially with income and decreasewith age. The exception is expendituresfor wine at home, which peak for con-sumers aged 45 to 54. Expenditures foralcohol away from home also rise withincome, but, like expenditures for wineat home, rise with age to a point and

then decline. Regardless, single menspend more on alcohol than do singlewomen, with nonsingles in the middlefor expenditures on all alcoholic bever-ages except wine at home, for whichnonsingles spend the most, on aver-age, followed by single men.

When characteristics are held con-stant by means of regression analysis,the trends in the predicted probabilityof reporting appear generally to matchthose described for the observed per-cent reporting. Other characteristicsalso appear to be related to the pur-chase of alcohol, including race andethnicity, occupation, and region ofresidence. However, the parameter es-timates associated with these variablesare not always statistically significant,especially for specific categories ofcharacteristics. (For example, with re-gard to the purchase of specific typesof alcohol, Asians are predicted to beless likely than whites to purchasewine, but the Asian coefficient for thepredicted purchase of beer is not sta-tistically significant.) Also, the prob-ability of purchasing one type of alco-hol is strongly related to the purchaseof another type of alcohol. For instance,consumers who purchase wine or someother alcohol are more likely to pur-chase beer as well, but the coefficientfor the purchase of both wine and an-other alcohol is not statistically signifi-cant, indicating that there is no “addi-tional effect” on the probability ofpurchasing beer when both wine andanother alcohol are purchased than iscaptured by including the effects ofwine and other purchases of alcoholseparately.

APPENDIX:

The Use of Logistic Re-gression (LOGIT) as aProbability Predictor

Logistic regression, or “logit,” is oftenused to predict the probability that anevent will occur, based on a series of


in general yields

P = exp(–0.4741)/[1 + exp(–0.4741)] = 0.384.

However, suppose one wanted toknow the predicted probability forsingle women instead of single men.That probability is

P = exp(–0.4741 – 0.7493)/[1 + exp(–0.4741 – 0.7493)] = 0.227.

The coefficient for single women(–0.7493) is simply added into the equa-tion as appropriate.

observed variables. In this approach,the probability of incurring expendi-tures for alcoholic beverages awayfrom home, given a series of demo-graphic characteristics, is examined.

One of the advantages of logit isthat the coefficients are easily con-verted into probabilities without hav-ing to resort to special tables or othermeans of calculation. The formula forsuch a probability is

Pj = exp(a + b1X1j + … + bnXnj)/[1 +exp(a + b1X1j + … + bnXnj)],

where b1,…,bn are parameter estimatesand X1j,…, Xnj are characteristics for thejth unit.

In the simplest example in this study,suppose one wants to calculate theprobability of purchasing alcohol awayfrom home for the control group de-scribed in the text of this article (that is,single men in the middle-income group,and so forth). Because all the indepen-dent variables in this case are binary,the only coefficient of concern is thatfor the intercept. In other words, usingthe results for the purchase of alcohol


Number of consumer units ...................... 80,020,767 13,215,599 14,720,627 14,628,126 14,613,513 14,653,034 8,189,868

Sample size ............................................. 11,276 1,727 2,010 2,063 2,138 2,202 1,136

Income before taxes (complete reporters only, except where designated otherwise) ................ $48,248 $8,914 $20,191 $34,647 $55,141 $118,611 $7,576

Age of reference person .......................... 49.3 56.4 50.7 45.6 44.5 45.3 57.3

Percent

Family type: ..............................................Husband and wife only ....................... 21.5 8.9 22.4 21.8 23.7 25.8 27.4Husband and wife, all children under 18 ........................................... 20.3 4.8 12.0 19.3 30.9 38.1 11.2Husband and wife, at least one child 18 or older ................................ 6.4 1.8 2.9 6.2 8.1 11.4 8.7Single parent (male) ........................... 7 0.5 0.4 1.3 1.0 0.2 0.3Single parent (female) ........................ 5.4 8.7 10.5 5.9 2.2 0.6 4.7Single man .......................................... 12.3 20.0 17.6 13.4 9.4 5.9 5.6Single woman ..................................... 15.7 42.3 16.2 11.8 6.6 2.1 19.9Other family ........................................ 17.7 13.0 18.1 20.2 18.2 16.0 22.3

Ethnic origin:Hispanic .............................................. 9.1 10.9 13.8 11.4 7.5 4.4 5.4Non-Hispanic ...................................... 90.9 89.1 86.2 88.6 92.5 95.6 94.6

Race:White ................................................... 83.4 78.3 81.4 84.6 84.7 88.5 82.1Black ................................................... 12.4 18.0 14.6 12.0 11.2 6.2 13.7Asian ................................................... 3.4 2.7 2.7 2.8 3.3 5.1 3.4Other race .......................................... 0.8 1.0 1.3 0.6 0.8 0.2 0.8

Occupation:Works for wage or salary: .................. 65.1 36.5 57.0 74.7 84.8 85.2 37.9

Managers and professionals ........ 20.3 4.4 8.3 16.8 28.2 48.2 9.9Teachers ....................................... 3.7 1.3 1.8 4.6 6.4 5.6 1.7Technicians, sales, and services ...................................... 25.6 21.5 29.8 32.2 29.4 20.9 14.2Blue collar ...................................... 14.2 8.5 15.1 18.9 19.4 9.6 12.0Agriculture (farming, forestry, or fishing) ....... 0.9 0.8 1.9 1.6 0.6 0.2 0.1Armed services ............................. 0.4 0.0 0.1 0.6 0.8 0.7 0.0

Self-employed ..................................... 5.0 4.1 4.6 5.9 3.1 5.9 7.4Not working: ....................................... 32.1 59.3 38.6 19.3 12.2 9.0 54.5

Retired ........................................... 19.9 37.9 27.7 12.1 7.7 3.7 41.0Unemployed .................................. 2.4 1.1 0.0 1 0.2 0.0 0.2Other not working ......................... 9.8 20.3 10.9 7.2 4.3 5.3 13.3

Housing tenure:Homeowner: ....................................... 66.8 48.3 56.9 61.4 75.2 86.7 73.2

Has mortgage ............................... 41.8 14.0 24.5 39.2 59.3 75.5 30.5Owns without mortgage ................ 25.0 34.3 32.4 22.2 15.9 11.2 42.7

Renter ................................................. 33.2 51.7 43.1 38.6 24.8 13.3 26.8

Region of residence:Northeast ............................................ 19.6 17.5 17.6 22.3 18.5 20.3 22.8Midwest .............................................. 24.1 21.2 22.8 24.3 28.0 23.6 25.2South .................................................. 35.1 40.3 39.5 32.7 32.0 30.8 35.5West ................................................... 21.2 21.0 20.1 20.7 21.5 25.3 16.5

Degree of urbanization:Urban ................................................. 86.9 82.2 85.4 85.5 88.3 91.3 88.8Rural ................................................... 13.1 17.8 14.6 14.5 11.7 8.7 11.2

All consumerunits

(21 andolder)

Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5Incompletereporters

Table 1. Purchases of alcohol by income quintile and selected demographic characteristics, 2000


Percent reporting

Purchase of alcohol:Alcohol, total ....................................... 29.0 16.7 22.9 30.9 35.8 43.2 18.6

At home ......................................... 20.4 11.9 16.9 22.3 24.8 29.1 13.4Away from home ........................... 14.6 6.9 9.2 14.4 19.8 25.6 8.5Both types purchased2 .................. 6.0 2.1 3.2 5.8 8.8 11.5 3.3

Beer: ................................................... 23.7 13.1 19.8 26.1 28.9 35.2 13.7At home ......................................... 14.4 8.4 13.5 17.3 17.0 18.7 8.5Away from home ........................... 12.8 6.1 8.1 12.4 17.3 22.6 7.4

Wine: ................................................... 9.9 4.3 5.5 9.0 12.4 19.3 7.3At home ......................................... 7.0 3.3 3.9 6.1 8.4 13.9 5.5Away from home ........................... 3.7 1.2 1.9 3.3 5.0 7.7 2.1

Other alcohol: ..................................... 8.3 4.0 4.9 7.4 11.2 15.1 5.8At home ......................................... 3.7 2.3 2.3 3.6 4.8 5.1 3.9Away from home ........................... 5.3 1.8 3.0 4.4 7.2 11.4 2.3

Mean weekly expenditure

Alcohol, total ............................................. $7.05 $3.72 $4.09 $6.49 $9.22 $13.15 $3.94At home ......................................... 4.71 3.05 3.07 4.51 5.37 8.35 3.01Away from home ........................... 2.34 0.67 1.02 1.98 3.85 4.80 0.93

Beer: ................................................... 3.70 2.24 2.68 4.18 4.93 5.48 1.70At home ......................................... 2.36 1.83 2.06 2.92 2.67 2.96 1.15Away from home ........................... 1.34 0.41 0.62 1.26 2.26 2.52 0.55

Wine: ................................................. 1.98 0.93 0.77 1.12 2.34 5.00 1.37At home ......................................... 1.63 .83 0.65 0.85 1.76 4.19 1.22Away from home ........................... 0.36 0.10 0.12 0.27 0.58 0.81 0.15

Other alcohol: ..................................... 1.36 0.55 0.64 1.19 1.95 2.67 0.87At home ......................................... 0.72 0.39 0.36 0.74 0.94 1.20 0.64Away from home ........................... 0.64 0.16 0.28 0.45 1.01 1.47 0.23

All consum-er units(21 andolder)

Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5Incompletereporters

Table 1. Purchases of alcohol by income quintile and selected demographic characteristics, 2000

1 Less than 0.5 percent.2 This group is included in both alcohol-at-home and alcohol-away-from-home groups. When the figure shown is subtracted from

the at-home and the away-from-home totals, the total percent reporting alcohol is obtained.


Number of consumer units ............................. 80,020,767 4,271,663 14,262,057 18,057,721 14,844,186 10,265,925 9,498,725 8,820,490

Sample size ................................................ 11,276 591 2,073 2,517 2,093 1,477 1,287 1,238

Income before taxes (complete reporters only, except where designated otherwise) ................................. $48,248 $24,207 $46,818 $60,703 $61,814 $49,729 $33,191 $22,659

Age of reference person ................................ 49.3 22.5 29.8 39.4 49.2 59.4 69.4 80.5

Percent

Family type: Husband and wife only ............................. 21.5 12.2 12.5 8.9 19.1 37.6 42.7 28.3 Husband and wife, all children under 18 ...... 20.3 12.7 36.9 40.6 17.4 4.6 0.3 3

Husband and wife, at least one child 18 or older .................................... 6.4 0.0 0.2 5.6 15.0 9.5 6.3 3.1

Single parent (male) ................................. 0.7 0.4 0.7 1.5 0.8 0.2 3 3

Single parent (female) .............................. 5.4 9.9 8.9 10.1 5.3 0.3 3 3

Single man ............................................. 12.3 24.2 12.3 10.9 11.9 10.6 10.0 14.9 Single woman ......................................... 15.7 14.3 9.3 6.1 10.4 18.5 24.9 42.4 Other family ........................................... 17.7 26.3 19.1 16.3 20.0 18.6 15.9 11.3

Ethnic origin: Hispanic ................................................ 9.1 11.1 16.6 9.6 9.6 5.3 4.6 3.7 Non-Hispanic .......................................... 90.9 88.9 83.4 90.4 90.4 94.7 95.4 96.3

Race: ......................................................... White .................................................... 83.4 82.2 78.8 82.8 82.1 84.3 86.9 90.0 Black .................................................... 12.4 12.8 14.0 13.7 12.7 12.9 11.1 7.8 Asian .................................................... 3.4 4.8 6.2 2.1 4.0 2.8 1.7 2.1 Other race ............................................. 0.8 0.2 1.0 1.4 1.2 1 0.3 0.1

Occupation: Works for wage or salary: ......................... 65.1 89.9 86.7 84.3 79.7 59.5 22.6 4.4

Managers and professionals ................ 2.3 15.7 26.9 27.8 25.3 19.8 8.3 1.4Teachers .......................................... 3.7 3.4 5.3 3.3 6.2 4.9 0.5 0.1Technicians, sales, and services ......................................... 25.6 50.7 33.4 30.5 31.5 21.0 10.3 2.5Blue collar ........................................ 14.2 15.7 18.9 21.9 15.9 12.8 3.3 0.4Agriculture (farming, forestry, or fishing) ............. 0.9 3.5 1.4 0.1 0.8 0.6 0.2 3

Armed services ................................. 0.4 0.9 0.8 0.7 3 0.4 3 3

Self-employed ......................................... 5.0 2.0 3.2 5.3 6.1 6.4 6.1 3.9 Not working: ........................................... 32.1 8.2 10.2 9.3 14.2 34.0 71.3 91.6

Retired ............................................. 19.9 0.4 0.1 0.2 1.2 18.3 65.1 86.3Unemployed ...................................... 2.4 3 0.3 0.3 0.3 1 0.5 3

Other not working .............................. 9.8 7.8 9.8 8.8 12.7 15.7 5.7 5.3

Housing tenure: Homeowner: .......................................... 66.8 15.5 47.5 66.2 73.8 73.8 83.4 78.1

Has mortgage ................................... 41.8 11.9 42.3 58.2 58.1 58.1 24.9 9.8Owns without mortgage ...................... 25.0 3.6 5.2 8.0 15.7 15.7 58.5 68.3Renter ............................................. 33.2 84.5 52.5 33.8 26.2 26.2 16.6 21.9

Region of residence: Northeast ............................................... 19.6 9.5 18.9 18.6 19.7 21.8 21.6 23.0 Midwest ................................................. 24.1 22.8 23.5 25.1 23.9 22.2 24.3 26.3 South .................................................... 35.1 38.6 32.7 35.2 35.0 36.4 38.2 31.8 West

West ..................................................... 21.2 29.1 24.9 21.1 21.4 19.6 15.9 18.9

Degree of urbanization: Urban .................................................... 86.9 91.3 88.2 86.9 88.6 85.4 80.4 88.1 Rural ..................................................... 13.1 8.7 11.8 13.1 11.4 14.6 19.6 11.9

Allconsumerunits (21

and older)

Under25 45–54 65–7425–34 35–44 55–64

75 and older

Table 2. Purchases of alcohol by age group and other selected demographic characteristics, 2000


Percent reporting

Purchase of alcohol: Alcohol, total ........................................... 29.0 34.6 36.5 31.7 31.5 27.4 21.3 14.3

At home ........................................... 20.4 26.4 26.1 22.6 22.5 18.3 14.2 9.2Away from home ................................ 14.6 14.9 19.4 16.0 15.6 14.9 9.9 7.1Both types purchased2 ....................... 6.0 6.7 9.0 6.9 6.6 5.8 2.8 2.0

Beer: ...................................................... 23.7 31.3 32.1 27.1 24.9 21.3 14.8 9.9At home ........................................... 14.4 23.2 21.4 17.5 14.4 10.6 7.6 4.6Away from home ................................ 12.8 13.8 16.7 14.1 14.2 13.0 8.2 6.1

Wine: ..................................................... 9.9 9.1 11.1 10.4 12.7 10.0 7.0 5.9At home ........................................... 7.0 5.8 7.5 7.1 9.3 7.1 5.5 4.3Away from home ................................

Other alcohol: .......................................... 8.3 8.7 10.3 8.9 8.9 8.6 6.5 4.3At home ........................................... 3.7 4.8 3.4 3.8 3.2 4.5 4.2 2.6Away from home ................................ 5.3 4.7 8.0 5.8 6.2 4.9 2.8 2.0


Alcohol, total ............................................... 7.05 9.65 8.18 8.57 7.60 6.69 4.72 2.81At home ................................................ 4.71 7.46 5.18 5.33 5.10 4.35 3.78 2.12Away from home .................................... 2.34 2.19 3.00 3.24 2.50 2.34 0.94 0.69

Beer: .................................................... 3.70 6.85 4.95 4.58 3.91 3.09 1.56 1.05At home ........................................... 2.36 5.48 3.24 2.77 2.50 1.69 0.97 0.65Away from home ................................ 1.34 1.37 1.71 1.81 1.41 1.40 0.59 0.40

Wine: .................................................... 1.98 1.34 1.66 2.29 2.40 2.23 2.13 1.05At home ........................................... 1.63 1.04 1.23 1.79 2.00 1.86 2.00 .92Away from home ................................ 0.36 0.30 0.43 0.50 0.40 0.37 0.13 0.13

Other alcohol: ......................................... 1.36 1.46 1.57 1.70 1.29 1.37 1.03 0.71At home ........................................... 0.72 0.94 0.71 0.77 0.60 0.80 0.81 0.55Away from home ................................ 0.64 0.52 0.86 0.93 0.69 0.57 0.22 0.16

1 Less than 0.5 percent.2 This group is included in both alcohol-at-home and alcohol-away-from-home groups. When the figure shown is subtracted from the at-home and the away-

from-home totals, the total percent reporting alcohol is obtained. 3 No data reported.

Allconsumerunits (21

and older)

Under25 45–54 65–7425–34 35–44 55–64

75 and older

Table 2. Purchases of alcohol by age group and other selected demographic characteristics, 2000


Number of consumer units ............................................................... 80,020,767 9,882,436 12,584,190 57,554,141

Sample size ...................................................................................... 11,276 1,365 1,708 8,203

Income before taxes (complete reporters only, except where designated otherwise) .... $48,248 $35,788 $22,042 $56,108

Age of reference person ................................................................... 49.3 48.2 60.0 47.1

Percent

Family type:Husband and wife only ................................................................ 21.5 1 1 29.8Husband and wife, all children under 18 ..................................... 20.3 1 1 28.2Husband and wife, at least one child 18 or older ........................ 6.4 1 1 8.9Single parent (male) .................................................................... 0.7 1 1 0.9Single parent (female) ................................................................. 5.4 1 1 7.6Single man ................................................................................... 12.3 100.0 1 1

Single woman .............................................................................. 15.7 1 100.0 1

Other family ................................................................................. 17.7 1 1 24.6

Ethnic origin:Hispanic ....................................................................................... 9.1 5.7 3.1 11.0Non-Hispanic ............................................................................... 90.9 94.3 96.9 89.0

Race:White ............................................................................................ 83.4 85.6 84.8 82.7Black ............................................................................................ 12.4 10.9 12.0 12.8Asian ............................................................................................ 3.4 2.9 2.8 3.6Other race ................................................................................... 0.8 0.6 0.4 0.9

Occupation:Works for wage or salary: ........................................................... 65.1 65.4 45.6 69.4

Managers and professionals ................................................. 20.3 20.7 15.0 21.4Teachers ................................................................................ 3.7 2.4 4.6 3.8Technicians, sales, and services .......................................... 25.6 23.7 22.3 26.6Blue collar ............................................................................... 14.2 16.4 3.4 16.2

Agriculture (farming, forestry, or fishing) ................................. 0.9 1.7 0.3 0.9Armed services ...................................................................... 0.4 0.5 3 0.5

Self-employed .............................................................................. 5.0 6.4 2.6 5.3Not working: ................................................................................ 32.1 28.0 52.0 25.4

Retired .................................................................................... 2.4 0.1 0.2 0.3Other not working .................................................................. 9.8 8.2 9.2 10.2

Housing tenure:Homeowner: ................................................................................ 66.8 49.4 58.2 71.6

Has mortgage. ....................................................................... 41.8 26.2 20.5 49.1Owns without mortgage ......................................................... 25.0 23.2 37.7 22.5Renter .................................................................................... 33.2 50.6 41.8 28.4

Region of residence:Northeast ..................................................................................... 19.6 19.8 21.3 19.2Midwest ....................................................................................... 24.1 23.5 27.4 23.5South ........................................................................................... 35.1 35.0 32.0 35.8West ............................................................................................ 21.2 21.7 19.3 21.5

Degree of urbanization:Urban ........................................................................................... 86.9 90.5 89.0 85.8Rural ............................................................................................ 13.1 9.5 11.0 14.2

Singles onlyAll consumerunits (21 and

older)Not single

Men Women

Table 3. Purchases of alcohol by marital status and other selected demographic characteristics, 2000


Percent reporting

Purchase of alcohol:Alcohol, total ................................................................................ 29.0 34.4 16.9 30.7

At home .................................................................................. 20.4 25.0 9.3 22.0Away from home .................................................................... 14.6 17.4 10.1 15.2Both types purchased2 ........................................................... 6.0 8.0 2.5 6.5

Beer: ............................................................................................ 23.7 30.0 12.0 25.2At home .................................................................................. 14.4 19.8 4.4 15.7Away from home .................................................................... 12.8 15.3 8.6 13.3

Wine: ............................................................................................ 9.9 10.3 6.2 10.7At home .................................................................................. 7.0 5.7 4.3 7.8Away from home .................................................................... 3.7 5.7 2.2 3.7

Other alcohol: .............................................................................. 8.3 11.2 4.6 8.6At home .................................................................................. 3.7 4.4 2.2 3.8Away from home .................................................................... 5.3 7.7 2.7 5.4


Alcohol, total ...................................................................................... 7.05 10.44 2.79 7.40At home ....................................................................................... 4.71 6.08 1.69 5.14Away from home ......................................................................... 2.34 4.36 1.10 2.26

Beer: ............................................................................................ 3.70 5.80 1.38 3.86At home .................................................................................. 2.36 3.57 0.67 2.53Away from home. ................................................................... 1.34 2.23 0.71 1.33

Wine: ............................................................................................ 1.98 2.43 0.80 2.16At home .................................................................................. 1.63 1.55 0.65 1.85Away from home .................................................................... 0.36 0.88 0.15 0.31

Other alcohol: .............................................................................. 1.36 2.21 0.61 1.38At home .................................................................................. 0.72 0.96 0.37 0.76Away from home .................................................................... 0.64 1.25 0.24 0.62

1 Not available.2 This group is included in both alcohol-at-home and alcohol-away-from-home groups. When the figure shown is subtracted from the

at-home and the away-from-home totals, the total percent reporting alcohol is obtained.3 No data reported.

All consumerunits (21 and

older)Not single

Men Women

Table 3. Purchases of alcohol by marital status and other selected demographic characteristics, 2000

Singles only


Intercept ........................................................................................... –0.4741 0.1047 20.4963 <0.000

Age of reference person (35 to 44):Under 25 ..................................................................................... 0.3076 0.1053 8.5411 0.003525 to 34 ....................................................................................... 0.2466 0.0659 13.9996 0.000245 to 54 ....................................................................................... –0.1135 0.0678 2.7993 0.094355 to 64 ....................................................................................... –0.3159 0.0828 14.5675 0.000165 to 74 ....................................................................................... –0.4296 0.1100 15.2484 <0.000175 and older ................................................................................ –0.8070 0.1300 38.5615 <0.0001

Income quintile (quintile 3):Quintile 1 ..................................................................................... –0.4375 0.0878 24.8015 <0.0001Quintile 2 ..................................................................................... –0.2861 0.0739 14.9804 0.0001Quintile 4 ..................................................................................... 0.1532 0.0685 5.0008 0.0253Quintile 5 ..................................................................................... 0.4824 0.0727 44.0884 <0.0001Incomplete income reporters ..................................................... –0.4096 0.0920 19.8067 <0.0001

Family type (single man):Husband and wife only ............................................................... –0.0925 0.0791 1.3676 0.2422Husband and wife, own children only ......................................... –0.4797 0.0814 34.6906 <0.0001Other husband and wife with children ........................................ –0.1951 0.1057 3.4057 0.0650Single father ................................................................................ –0.2906 0.2414 1.4494 0.2286Single mother .............................................................................. –1.0723 0.1316 66.4027 <0.0001Single woman ............................................................................. –07493 0.0905 68.5886 <0.0001Other family ................................................................................ –0.2896 0.0795 13.2701 0.0003

Ethnic origin of reference person (non-Hispanic):Hispanic ...................................................................................... –0.0628 0.0779 0.6499 0.4201

Race of reference person (white):Black ........................................................................................... –0.5253 0.0810 42.0712 <0.0001Asian ........................................................................................... –0.3847 0.1096 12.3303 0.0004Other race .................................................................................. –0.3502 0.2669 1.7217 0.1895

Occupation of reference person (manager or professional, wage or salaried):

Teacher ...................................................................................... –0.0911 0.1173 0.6039 0.4371Technical, sales, or services ..................................................... 0.1292 0.0626 4.2587 0.0390Blue collar ................................................................................... 0.0346 0.0751 0.2122 0.6451Agricultural .................................................................................. 0.4531 0.2136 4.5014 0.0339Armed services .......................................................................... 0.2514 0.2837 0.7854 0.3755Self-employed ............................................................................. 0.0129 0.1073 0.0144 0.9046Retired ........................................................................................ 0.0723 0.1059 0.4662 0.4948Unemployed long term ............................................................... –0.6179 0.6308 0.9596 0.3273Not working, other reason .......................................................... –0.3305 0.0950 12.1095 0.0005

Housing tenure (homeowner with mortgage):Homeowner no mortgage ........................................................... –0.0477 0.0674 0.4992 0.4798Renter ......................................................................................... –0.0155 0.0575 0.0723 0.7880

Region of residence (South):Northeast .................................................................................... 0.2978 0.0639 21.6987 <0.0001Midwest ...................................................................................... 0.2903 0.0602 23.2703 <0.0001West ........................................................................................... 0.2285 0.0600 14.5163 0.0001

Degree of urbanization (urban):Rural ........................................................................................... –0.4238 0.0842 25.3511 <0.0001

Intercept ........................................................................................... –1.1579 0.1168 98.2728 <0.0001

Age of reference person (35 to 44):Under 25 ..................................................................................... 0.3185 0.1138 7.8368 0.0051

Alcohol, total

Standard error Chi-square Pr > chi-square

Logit resultsCharacteristic (control group value in parentheses) Parameter

estimate

Table 4. Parameter estimates and other results of the logit regressions on alcohol purchase patterns, 2000

Alcohol at home


25 to 34 ....................................................................................... 0.1773 0.0720 6.0620 0.013845 to 54 ....................................................................................... –0.0441 0.0745 0.3503 0.554055 to 64 ....................................................................................... –0.3031 0.0934 10.5272 0.001265 to 74 ....................................................................................... –0.4555 0.1263 13.0060 0.000375 and older ................................................................................ –0.8446 0.1518 30.9599 <0.0001

Income quintile (quintile 3):Quintile 1 ..................................................................................... –0.3573 0.0995 12.8953 0.0003Quintile 2 ..................................................................................... –0.2210 0.0823 7.2202 0.0072Quintile 4 ..................................................................................... 0.0545 0.0759 0.5164 0.4724Quintile 5 ..................................................................................... 0.2399 0.0802 8.9408 0.0028Incomplete income reporters ..................................................... –0.3521 0.1038 11.5177 0.0007

Family type (single man):Husband and wife only ............................................................... –0.0521 0.0875 0.3543 0.5517Husband and wife, own children only ......................................... –0.2185 0.0891 6.0163 0.0142Other husband and wife with children ........................................ –0.0531 0.1156 0.2113 0.6457Single father ................................................................................ –0.3249 0.2718 1.4284 0.2320Single mother .............................................................................. –1.1369 0.1574 52.1439 <0.0001Single woman ............................................................................. –0.9314 0.1089 73.1263 <0.0001Other family ................................................................................ –0.1672 0.0872 3.6803 0.0551

Ethnic origin of reference person (non-Hispanic):Hispanic ...................................................................................... 0.1099 0.0830 1.7526 0.1855



Teacher ...................................................................................... –0.1204 0.1363 0.7798 0.3772Technical, sales, or services ..................................................... 0.2559 0.0697 13.4980 0.0002Blue collar ................................................................................... 0.2208 0.0823 7.1906 0.0073Agricultural .................................................................................. 0.7617 0.2191 12.0920 0.0005Armed services .......................................................................... 0.7945 0.2840 7.8288 0.0051Self-employed ............................................................................. –0.0083 0.1229 0.0045 0.9464Retired ........................................................................................ 0.2674 0.1217 4.8284 0.0280Unemployed long term ............................................................... –0.1551 0.6299 0.0606 0.8055Not working, other reason .......................................................... –0.1162 0.1054 1.2149 0.2704

Housing tenure (homeowner with mortgage):Homeowner no mortgage ........................................................... –0.0192 0.0761 0.0636 0.8009Renter ......................................................................................... 0.0035 0.0635 0.0030 0.9567

Region of residence (South):Northeast .................................................................................... 0.2379 0.0710 11.2192 0.0008Midwest ...................................................................................... 0.1854 0.0674 7.5557 0.0060West ........................................................................................... 0.1813 0.0665 7.4371 0.0064

Degree of urbanization (urban):Rural ........................................................................................... –0.4292 0.0964 19.8406 <0.0001

Type of alcohol purchased:Alcohol for consumption away from home........................................ 0.0173 0.0022 61.4922 <0.0001

Intercept ........................................................................................... –1.3053 0.1314 98.6242 <.0001

Age of reference person (35 to 44):Under 25 ..................................................................................... 0.1957 0.1380 2.0110 0.156225 to 34 ....................................................................................... 0.3044 0.0824 13.6458 0.0002

Alcohol away from home



estimate


Alcohol at home—Continued


45 to 54 ....................................................................................... –0.1808 0.0864 4.3757 0.036555 to 64 ....................................................................................... –0.2171 0.1045 4.3120 0.037865 to 74 ....................................................................................... –0.3970 0.1441 7.5897 0.005975 and older ................................................................................ –0.5690 0.1716 10.9909 0.0009

Income quintile (quintile 3):Quintile 1 ..................................................................................... –0.4909 0.1225 16.0642 <0.0001Quintile 2 ..................................................................................... –0.3016 0.1011 8.8954 0.0029Quintile 4 ..................................................................................... 0.2862 0.0873 10.7341 0.0011Quintile 5 ..................................................................................... 0.6007 0.0910 43.5773 <0.0001Incomplete income reporters ..................................................... –0.3696 0.1280 8.3396 0.0039

Family type (single man):Husband and wife only ............................................................... –0.0995 0.0984 1.0234 0.3117Husband and wife, own children only ......................................... –0.6775 0.1035 42.8473 <0.0001Other husband and wife with children ........................................ –0.3196 0.1351 5.5944 0.0180Single father ................................................................................ –0.6523 0.3300 3.9073 0.0481Single mother .............................................................................. –07275 0.1724 17.7999 <0.0001Single woman ............................................................................. –0.3575 0.1131 9.9967 0.0016Other family ................................................................................ –0.3951 0.1023 14.9156 0.0001

Ethnic origin of reference person (non-Hispanic):Hispanic ...................................................................................... –0.4688 0.1153 16.5228 <0.0001



Teacher ...................................................................................... –0.0630 0.1384 0.2073 0.6489Technical, sales, or services ..................................................... –0.0965 0.0757 1.6263 0.2022Blue collar ................................................................................... –0.3175 0.0966 10.8138 0.0010Agricultural .................................................................................. –0.4291 0.3224 1.7721 0.1831Armed services .......................................................................... –0.4104 0.3782 1.1774 0.2779Self-employed ............................................................................. –0.0111 0.1305 0.0073 0.9319Retired ........................................................................................ –0.2224 0.1377 2.6092 0.1062Unemployed long term ............................................................... –11.5682 201.4000 0.0033 0.9542Not working, other reason .......................................................... –0.6136 0.1330 21.2917 <0.0001



Degree of urbanization (urban): ......................................................Rural ........................................................................................... –0.2208 0.1095 4.0682 0.0437

Type of alcohol purchasedAlcohol for consumption away from home........................................ 0.0181 0.0016 125.3391 <0.0001

Intercept ........................................................................................... –1.9918 0.1829 118.5713 <0.0001

Age of reference person (35 to 44):Under 25 ..................................................................................... 0.3589 0.1880 3.6434 0.056325 to 34 ....................................................................................... 0.3085 0.1132 7.4340 0.006445 to 54 ....................................................................................... –0.1963 0.1205 2.6563 0.1031



estimate


Alcohol away from home—Continued

Alcohol at home and alcohol away from home


Alcohol at home and alcohol away from home—Continued

55 to 64 ....................................................................................... –0.3268 0.1507 4.7008 0.030165 to 74 ....................................................................................... –0.8113 0.2330 12.1242 0.000575 and older ................................................................................ –0.8939 0.2811 10.1150 0.0015

Income quintile (quintile 3):Quintile 1 ..................................................................................... –0.6653 0.2013 10.9182 0.0010Quintile 2 ..................................................................................... –0.3272 0.1535 4.5440 0.0330Quintile 4 ..................................................................................... 0.3343 0.1249 7.1651 0.0074Quintile 5 ..................................................................................... 0.5858 0.1295 20.4516 <0.0001Incomplete income reporters ..................................................... –0.3699 0.1991 3.4501 0.0632

Family type (single man):Husband and wife only ............................................................... –0.1450 0.1363 1.1324 0.2873Husband and wife, own children only ......................................... –0.7011 0.1415 24.5382 <0.0001Other husband and wife with children................................................. –0.2276 0.1837 1.5344 0.2154Single father ................................................................................ –1.7053 0.7267 5.5062 0.0189Single mother .............................................................................. –1.3269 0.3028 19.2059 <0.0001Single woman ............................................................................. –0.8691 0.1854 21.9819 <0.0001Other family ................................................................................ –0.4285 0.1423 9.0715 0.0026

Ethnic origin of reference person (non-Hispanic):Hispanic ...................................................................................... –0.3771 0.1595 5.5871 0.0181Race of reference person (white): .............................................Black ........................................................................................... –0.8317 0.1873 19.7237 <0.0001Asian ........................................................................................... –0.7441 0.2331 10.1877 0.0014Other race .................................................................................. –0.3315 0.5238 0.4007 0.5268


Teacher ...................................................................................... –0.3262 0.2108 2.3936 0.1218Technical, sales, or services ..................................................... –0.0185 0.1044 0.0315 0.8591

Blue collar ................................................................................... –0.2494 0.1349 3.4196 0.0644Agricultural .................................................................................. –0.0917 0.4091 0.0503 0.8226Armed services .......................................................................... 0.3739 0.3985 0.8807 0.3480Self-employed ............................................................................. –0.0689 0.1903 0.1310 0.7174Retired ........................................................................................ –0.0978 0.2171 0.2032 0.6522Unemployed long term ............................................................... –11.2583 288.2000 .0015 0.9688Not working, other reason .......................................................... –0.6011 0.1988 9.1399 0.0025



Degree of urbanization (urban):Rural ........................................................................................... –0.2788 0.1624 2.9457 0.0861

Intercept ........................................................................................... –1.1944 0.1202 98.7746 <0.0001

Age of reference person (35 to 44):Under 25 ..................................................................................... 0.2794 0.1160 5.8050 0.016025 to 34 ....................................................................................... 0.2044 0.0733 7.7706 0.005345 to 54 ....................................................................................... –0.2487 0.0776 10.2743 0.001355 to 64 ....................................................................................... –0.4444 0.0965 21.2240 <0.000165 to 74 ....................................................................................... –0.6851 0.1323 26.8169 <0.000175 and older ................................................................................ –1.0011 0.1579 4.1924 <0.0001



estimate


Beer


Income quintile (quintile 3):Quintile 1 ..................................................................................... –0.3618 0.1002 13.0291 0.0003Quintile 2 ..................................................................................... –0.1609 0.0830 3.7578 0.0526Quintile 4 ..................................................................................... –0.0241 0.0784 0.0944 0.7587Quintile 5 ..................................................................................... 0.1772 0.0832 4.5310 0.0333Incomplete income reporters ..................................................... –0.5580 0.1101 25.6931 <0.0001

Family type (single man):Husband and wife only ............................................................... –0.1248 0.0908 1.8898 0.1692Husband and wife, own children only ......................................... –0.3699 0.0919 16.1972 <0.0001Other husband and wife with children ........................................ –0.2430 0.1226 3.9307 0.0474Single father ................................................................................ –0.6450 0.2922 4.8720 0.0273Single mother .............................................................................. –1.0858 0.1514 51.4280 <0.0001Single woman ............................................................................. –0.8272 0.1072 59.5817 <0.0001Other family ................................................................................ –0.2681 0.0905 8.7658 0.0031

Ethnic origin of reference person (non-Hispanic):Hispanic ...................................................................................... 0.1383 0.0862 2.5710 0.1088



Teacher ...................................................................................... –0.0312 0.1386 0.0506 0.8220Technical, sales, or services ..................................................... 0.2266 0.0724 9.7906 0.0018Blue collar ................................................................................... 0.3630 0.0846 18.3947 <0.0001Agricultural .................................................................................. 0.5969 0.2294 6.7688 0.0093Armed services .......................................................................... 0.6854 0.3022 5.1442 0.0233Self-employed ............................................................................. 0.0699 0.1261 0.3075 0.5792Retired ........................................................................................ 0.2404 0.1279 3.5300 0.0603Unemployed long term ............................................................... –0.4230 0.6913 0.3744 0.5406Not working, other reason .......................................................... –0.0920 0.1084 0.7199 0.3962


Region of residence (South):Northeast .................................................................................... 0.2833 0.0738 14.7555 0.0001Midwest ...................................................................................... 0.3047 0.0693 19.3346 <0.0001West ........................................................................................... 0.1180 0.0693 2.9012 0.0885


Type of alcohol purchased:Purchased wine .......................................................................... 1.4732 0.0826 318.0120 <0.0001Purchased another alcohol ........................................................ 1.8090 0.0953 360.5648 <0.0001Purchased wine and another alcohol ......................................... 0.2146 0.1938 1.2265 0.2681

Intercept ........................................................................................... –2.9972 0.1783 282.7000 <0.0001

Age of reference person (35 to 44):Under 25 ..................................................................................... 0.2933 0.1799 2.6567 0.103125 to 34 ....................................................................................... 0.0643 0.1094 0.3451 0.5569

45 to 54 ....................................................................................... 0.3258 0.1072 9.2425 0.002455 to 64 ....................................................................................... 0.2068 0.1334 2.4022 0.121265 to 74 ....................................................................................... 0.0941 0.1800 0.2729 0.601475 and older ................................................................................ 0.0200 0.2115 0.0090 0.9246



estimate


Beer—Continued

Wine


Income quintile (quintile 3):Quintile 1 ..................................................................................... –0.3014 0.1582 3.6306 0.0567Quintile 2 ..................................................................................... –0.2527 0.1330 3.6102 0.0574Quintile 4 ..................................................................................... 0.1887 0.1137 2.7517 0.0972Quintile 5 ..................................................................................... 0.4614 0.1158 15.8821 <0.0001Incomplete income reporters ..................................................... 0.1291 0.1490 0.7513 0.3861

Family type (single man):Husband and wife only ............................................................... 0.0787 0.1300 0.3666 0.5449Husband and wife, own children only ......................................... 0.1395 0.1345 1.0765 0.2995

Other husband and wife with children ........................................ 0.1119 0.1690 0.4380 0.5081Single father ................................................................................ 0.2328 0.3951 0.3472 0.5557Single mother .............................................................................. –0.0507 0.2258 0.0504 0.8224Single woman ............................................................................. 0.0302 0.1515 0.0397 0.8420Other family ................................................................................ –0.0376 0.1345 0.0781 0.7798


Race of reference person (white):Black ........................................................................................... –0.2846 0.1397 4.1493 0.0417Asian ........................................................................................... –0.6004 0.1989 9.1125 0.0025Other race .................................................................................. –0.1145 0.4709 0.0591 0.8079


Teacher ...................................................................................... –0.1149 0.1780 0.4167 0.5186Technical, sales, or services ..................................................... –0.1094 0.0955 1.3115 0.2521Blue collar ................................................................................... –0.5779 0.1294 19.9599 <0.0001Agricultural .................................................................................. –0.6064 0.4275 2.0118 0.1561Armed services .......................................................................... –0.7662 0.5178 2.1895 0.1390Self-employed ............................................................................. –0.1024 0.1652 0.3845 0.5352Retired ........................................................................................ –0.1846 0.1684 1.2029 0.2727Unemployed long term ............................................................... –0.9262 1.1455 0.6537 0.4188Not working, other reason .......................................................... –0.6230 .1689 13.6054 0.0002

Housing tenure (homeowner with mortgage):Homeowner no mortgage ........................................................... 0.0162 0.1098 0.0218 0.8827Renter ......................................................................................... –0.2146 0.0967 4.9199 0.0265

Region of residence (South):Northeast .................................................................................... 0.1805 0.1014 3.1686 0.0751Midwest ...................................................................................... –0.1860 0.1003 3.4416 0.0636West ........................................................................................... 0.1555 0.0960 2.6234 0.1053


Type of alcohol purchased:Purchased beer .......................................................................... 1.4781 0.0823 322.3686 <0.0001Purchased another alcohol ........................................................ 1.4416 0.1669 74.6139 <0.0001Purchased beer and another alcohol ......................................... 0.2136 0.1931 1.2230 0.2688

Intercept ........................................................................................... –3.1624 0.1913 273.2494 <0.0001

Age of reference person (35 to 44):Under 25 ..................................................................................... –0.0362 0.1899 0.0363 0.848925 to 34 ....................................................................................... 0.1603 0.1150 1.9423 0.163445 to 54 ....................................................................................... –0.1186 0.1200 0.9763 0.323155 to 64 ....................................................................................... –0.0669 0.1460 0.2098 0.646965 to 74 ....................................................................................... 0.1273 0.1915 0.4417 0.506375 and older ................................................................................ –0.0402 0.2326 0.0299 0.8628

Another alcohol



estimate


Wine—Continued


Income quintile (quintile 3):Quintile 1 ..................................................................................... –0.2224 0.1700 1.7113 0.1908Quintile 2 ..................................................................................... –0.1768 0.1425 1.5397 0.2147Quintile 4 ..................................................................................... 0.3201 0.1233 6.7399 0.0094Quintile 5 ..................................................................................... 0.4278 0.1287 11.0468 0.0009Incomplete income reporters ..................................................... 0.0696 0.1686 0.1703 0.6799

Family type (single man):Husband and wife only ............................................................... –0.1649 0.1327 1.5440 0.2140Husband and wife, own children only.................................................. –0.7449 0.1433 27.0278 <0.0001Other husband and wife with children................................................. –0.1470 0.1794 0.6718 0.4124Single father ................................................................................ 0.2641 0.3785 0.4868 0.4854Single mother .............................................................................. –0.5883 0.2501 5.5334 0.0187Single woman ............................................................................. –0.4545 0.1621 7.8643 0.0050Other family ................................................................................ –0.2688 0.1363 3.8901 0.0486


Race of reference person (white):Black ........................................................................................... –0.1530 0.1496 1.0460 0.3064Asian ........................................................................................... –0.5739 0.2180 6.9304 0.0085Other race .................................................................................. 0.2898 0.4528 0.4098 0.5221


Teacher ...................................................................................... –0.1699 0.2024 0.7042 0.4014Technical, sales, or services ..................................................... –0.1351 0.1050 1.6558 0.1982Blue collar ................................................................................... –0.4466 0.1387 1.3688 0.0013Agricultural .................................................................................. –0.1272 0.4046 0.0988 0.7533Armed services .......................................................................... –0.1766 0.4906 0.1296 0.7189Self-employed ............................................................................. 0.0741 0.1778 0.1735 0.6770

Retired ............................................................................................. –02470 0.1856 1.7713 0.1832Unemployed long term ............................................................... 0.6229 0.8580 0.5272 0.4678Not working, other reason .......................................................... –03382 0.1806 3.5049 0.0612


Region of residence (South):Northeast .................................................................................... –0.1086 0.1173 0.8579 0.3543Midwest ...................................................................................... 0.2482 0.1059 5.4901 0.0191West ........................................................................................... 0.1782 0.1063 2.8119 0.0936


Type of alcohol purchased:Purchased beer .......................................................................... 1.8210 0.0950 367.7814 <0.0001Purchased wine .......................................................................... 1.4733 0.1662 78.6136 <0.0001Purchased beer and wine .......................................................... 0.1604 0.1918 0.6989 0.4032



estimate


Another alcohol—Continued


The Cost andDemographics of VehicleAcquisition

LAURA PASZKIEWICZ

Laura Paszkiewicz is an economist in theBranch of Information Analysis, Divisionof Consumer Expenditure Surveys, Bureauof Labor Statistics.

Transportation costs make up alarge part of a consumer’s bud-get. Consumer Expenditure (CE)

Survey data for 2000 indicate that 88percent of all consumer units1 eitherowned or leased a vehicle,2 and expen-ditures for leasing and purchasing (thelatter defined as a net outlay) vehiclesmade up almost 10 percent of the aver-age consumer unit’s total expenditures.

In April 1991, the CE Survey beganto ask for detailed information on theleasing of vehicles. Since that time, theincidence of leasing a vehicle increasedsteadily before tapering off in recentyears. With the introduction of the moredetailed data, it is possible to investi-gate the factors that contribute to aconsumer’s decision to lease a vehicle,as opposed to purchasing it. The mainfactor contributing to this decision isthe varying cost of each option. Usingrecent CE Survey data, this article ex-amines the initial and monthly costsinvolved in leasing a vehicle, purchas-ing a new vehicle, and purchasing aused vehicle. The article presents de-tails on the demographic breakdown ofconsumers who lease, buy new, or buyused vehicles.

MethodologyThe sample used for this article includesall Interview survey participants from1999 or 2000 who reported a new lease3

or purchase of a vehicle in the year inwhich the interview took place. (Inother words, the sample consists of allparticipants in the 2000 Survey wholeased or purchased a vehicle in 2000,as well as all participants in the 1999Survey who leased or purchased a ve-hicle in 1999.) Respondents who re-ported using the vehicle for business,or, alternatively, receiving complete orpartial payment for the vehicle by anemployer are excluded from the sample.

Costs involved in leasing versus buy-ing are investigated. Average down-payments and monthly payments arecompared, as are the average durationsover which payments are made. Theinvestigation further includes analysesof leasing and buying by the follow-ing demographic characteristics: Age,race, gender, income quintile, geo-graphical region, and type of area (ur-ban vs. rural).

BackgroundThe increase in the frequency of leas-ing vehicles in recent years has beencaptured in a new section of the CESurvey added in April 1991. Leases

1 See the glossary at the end of this an-thology for the definition of consumer unit.

2 In the published CE data, vehicles aredefined as cars, trucks (including minivans,vans, sports utility vehicles (SUVs), andjeeps), and other vehicles (motorcycles andaircrafts). Henceforth, the term vehicle willencompass only cars and trucks.

3 The time at which a lease is started isdetermined by the year in which the firstpayment was made on the lease.


made up 2.7 percent of all new acquisi-tions of vehicles4 in 1991. By 1997, leas-ing had reached 10.0 percent of all re-cent car acquisitions. (See chart 1, toppanel.) After 1997, the incidence of leas-ing began to decline, falling to 7.2 per-cent of vehicle acquisitions in 2000.With the increasing popularity of leas-ing during 1991–97, new-vehicle pur-chases decreased as an overall percent-age of vehicle acquisitions, droppingfrom 27 percent in 1991 to 21 percent in1997. After 1997, the incidence of new-vehicle purchases began to rise, reach-ing 25.6 percent of vehicles acquired in2000. Used-vehicle purchases made up70.4 percent of all acquisitions of ve-hicles in 1991. From 1991 to 1996, thepercentage rose to 72.7 percent, afterwhich it began to fall. In 2000, used-vehicle purchases made up 67.2 per-cent of all vehicle acquisitions.

In 1999 and 2000, the total percent-age of consumer units acquiring a ve-hicle was just under 4 percent of theentire population. Of those who re-ported a recent acquisition, 66 percentbought a used vehicle, 26 percentbought a new one, and the remaining 8percent leased a vehicle. (See table 1.)

CostsOne of the factors involved in choos-ing a method of acquisition is cost.Among the costs incurred in acquiringa vehicle are downpayments and month-ly payments for leasing and purchas-ing.

Overall, 81 percent of new-vehiclepurchasers financed their purchases,compared with 56 percent of used-ve-hicle purchasers. The CE Survey asksquestions regarding the amount ofdownpayments and monthly paymentsfor purchased vehicles of those re-spondents in the sample who financedthe vehicle and have monthly pay-ments remaining. Of those purchaserswho financed, 87 percent of new-ve-hicle purchasers and 79 percent ofused-vehicle purchasers had paymentsremaining.

Downpayments are a good indica-tor of upfront costs for acquiring a ve-hicle—costs that could dictate whetherto lease a vehicle, buy a used vehicle,or buy a new vehicle. Lessees paid$868,5 on average, as a downpayment,about 76 percent of the amount that aused-vehicle purchaser paid as adownpayment ($1,147) and only 30 per-cent of the amount that a new-vehiclepurchaser put for a downpayment($2,914). (See table 2.) The maximumdownpayment was $8,500 for lessees,$37,000 for new-vehicle purchasers,and $19,000 for used-vehicle purchas-ers. These data suggest that the initialcosts for leasing an automobile arelower than the costs for purchasing ei-ther a new or used vehicle. The differ-ence in downpayments can be partiallyexplained by the main difference be-tween leases and purchases: with leases,the downpayment is for a service; withpurchases, the downpayment is for anasset.

Monthly costs also could be a fac-tor in deciding whether to lease, buynew, or buy used. The average monthlypayment was $353 for lessees, $399 fornew-vehicle purchasers, and $273 forused-vehicle purchasers. Thus, al-though lessees have a lower monthlypayment than do new-vehicle purchas-ers, they still have a higher monthlypayment than do used-vehicle pur-chasers.

The amount of time it takes to payoff a loan or to complete a lease alsocould have an effect both on a person’sdecision to lease, buy new, or buy usedand on the total cost of the vehicle. Onaverage, new-vehicle buyers made 54monthly payments, used-vehicle buy-ers 43, and lessees 39. The most com-mon term for leasing was 3 years, and50 percent of lessees chose that term.For new-vehicle purchasers, the mostcommon term for financing was 5 years,and 55 percent of new-car purchaserschose that term. For used-vehicle pur-chasers, a number of terms were com-mon, but the top two were 5-year terms

(chosen by 27 percent) and 4-year terms(selected by 24 percent).

Demographic analysisThe demographic analysis in this sec-tion examines the entire sample of thoseacquiring a vehicle in 1999 or 2000, in-cluding consumers who financed theirvehicles and those who did not. (Seetable 1.)

Income. Consumers who purchasedused vehicles had the least income, onaverage. The average annual income(based on complete income reporters6)of someone who bought a used vehiclewas $48,004, compared with $72,992 forlessees and $69,875 for new-vehiclepurchasers. Overall, nearly 30 percentof those who recently acquired a ve-hicle were in the highest incomequintile; the 30-percent figure was morethan that for the first and second in-come quintiles combined.7 The percent-age of used-vehicle purchases de-creases and the percentages of bothnew-vehicle purchases and leasings ofvehicles increases as one proceedsfrom a lower income quintile to a higherincome quintile.

Among the consumer units thatbought or leased a vehicle, those in thelowest income quintile were the mostlikely to buy a used car (80.9 percent).In comparison, only 54.1 percent of autopurchasers in the fifth quintile boughtused vehicles. Almost 36 percent ofthose leasing or buying in the highestincome quintile bought a new car; thefigure was 10 percentage points higherthan that of the fourth income quintileand more than 20 percentage pointshigher than that of the lowest incomequintile.

Age.8 Twenty-eight percent of thoseacquiring vehicles in 1999 and 2000

4 New acquisitions of vehicles include pur-chases of new vehicles, purchases of used ve-hicles, and leases of vehicles.

5 In computing these averages, those whorecently acquired a vehicle and reported nodownpayment were counted as having zerodollars for a downpayment.

6 See “Glossary” in Appendix A at the endof this anthology for the definition of com-plete income reporter.

7 See “Glossary” in Appendix A at the endof this anthology for the definition ofquintiles of income before taxes.

8 Both the age and race variables refer tothe age or race of the reference person—theperson first mentioned when the respondentis asked, “Start with the name of the personor one of the persons who owns or rents thehome.”


were in the 35-to 44-year-old agebracket, although that age group madeup just 22 percent of the population.Both the 25- to 34-year-old age groupand the 45- to 54-year-old age groupalso made up large portions of thoseacquiring vehicles. Each of the twogroups accounted for more than 20percent of all acquisitions, yet made upless than 20 percent of the population.The oldest group (75 and older) ac-quired the fewest vehicles, with only2.6 percent of acquisitions, much lessthan their 9.6-percent share of the popu-lation.

The average age was 44 for con-sumers leasing vehicles, 47 for consum-ers buying new vehicles, and 42 forconsumers buying used vehicles. Asthese data imply, the probability ofchoosing a used car over a new car de-creases with age. The incidence of leas-ing ranges from 7.9 percent to 8.6 per-cent for those 25 to 64, peaking in theage range from 55 to 64. Consumersunder the age of 25 or over the age of75 who acquired a vehicle were the leastlikely to lease, with 5.4 percent and 4.9percent, respectively, doing so.

Gender. In order to examine the statis-tics on vehicle acquisitions by gender,the sample was divided into subsetsthat include only single-member con-sumer units. The CE Interview surveycollects expenditure data for all mem-bers within the consumer unit com-bined, not for each member separately.By using single consumer units instead,a differentiation can be made betweenthe expenditures of men and those ofwomen.

Men acquired a slightly larger per-centage of vehicles than their share ofthe single population in 1999–2000. Thefigures were 58 percent and 54 percent,respectively.

Results of this portion of the studysuggest that men and women acquirevehicles differently. A total of 9.6 per-cent of single men in the sample leasedvehicles, 20.6 percent bought new ve-hicles, and 69.9 percent purchased usedvehicles. By contrast, 11.5 percent ofsingle women leased vehicles, 36.9 per-cent bought new vehicles, and 51.5

percent bought used vehicles.Even though single women acquired

a smaller percentage of vehicles thantheir share of the population, they pur-chased a greater percentage of new ve-hicles and leased a greater percentageof vehicles than their population share.In particular, women made up 46 per-cent of the singles population, yet pur-chased 56.9 percent of all new vehicles,and leased 46.9 percent of all leasedvehicles, among singles.

Region. Acquisitions of vehicles varyby region. With 31 percent and 16 per-cent, respectively, of total acquisitions,consumer units in the South and theNortheast acquired smaller percent-ages of vehicles than their populationshares in 1999–2000: 35 percent and 19percent of the total U.S. population. Bycontrast, with 27 percent and 25 per-cent, respectively, of vehicle acquisi-tions, consumer units in the Midwestand the West acquired higher percent-ages of vehicles than their populationshares of 24 percent and 22 percent ofthe total U.S. population.

Consumers acquiring vehicles in theNortheast were more likely to leasethan were those in the West, at 12.6percent, in contrast to 4.8 percent. Con-sumer units in the West were morelikely to buy a used vehicle, with 66percent of those acquiring vehiclesdoing so, compared with 58 percent ofthose in the Northeast. The Northeastand the West both had about 30 per-cent of their vehicle-acquiring popula-tion reporting a purchase of a new ve-hicle. The Midwest and the Southvaried only slightly in the three kindsof acquisitions: in the Midwest, 9 per-cent of those who acquired vehiclesleased them, 23 percent bought themnew, and 69 percent purchased themused; in the South, 8 percent leasedtheir vehicles, 25 percent purchasedthem new, and 67 percent bought themused.

Type of area (urban vs. rural). Con-sumers in urban and rural areas eachacquired roughly the same percentageof vehicles as their population share.The methods of acquisition that con-

sumers in the two areas chose, how-ever, were considerably different.

Consumer units in urban areas weremore likely to lease or buy a new ve-hicle than were those in rural areas.Among consumer units acquiring ve-hicles, 3.3 percent of those living in ru-ral areas leased their vehicles, whereas8.5 percent of those living in urban ar-eas did so. Almost 27 percent of con-sumer units in urban areas bought newvehicles, compared with 22.6 percentof those in rural areas. Someone livingin a rural area was more likely to buy aused car (71.4 percent) than was some-one in an urban area (64.8 percent).

Race.9 The CE Survey has four racecategories: White; black; Asian or Pa-cific Islander; and American Indian,Aleut, or Eskimo.

Persons of Asian or Pacific Islanderheritage accounted for just 3.1 percentof the population acquiring vehiclesand were the most differentiated interms of the three ways of acquiringthem, compared with the other races. Alittle more than half of their populationacquiring vehicles bought a used ve-hicle, 42 percent purchased a new ve-hicle, and the remaining 7 percent leaseda vehicle. Among the remaining racialgroups, the most similar in terms ofvehicle acquisition method was thewhite population, which accounted formost (88 percent) of the populationacquiring vehicles: among whites, 65.5percent bought used vehicles, 26.5 per-cent purchased new ones, and 8 per-cent leased vehicles.

The black population and the Ameri-can Indian, Aleut, and Eskimo popula-tion were most different from the groupof Asian and Pacific Islander descentin their distribution over the three kindsof arrangements for acquiring a vehicle.The two populations were similar toeach other in having the lowest per-centage of leases and new-vehicle pur-chases and the highest percentage of

9 Both the age and race variables refer tothe age or race of the reference person, theperson who was first mentioned when therespondent is asked, “Start with the name ofthe person or one of the persons who ownsor rents the home.”


used-vehicle purchases. Among blackconsumer units acquiring vehicles, 5.3percent leased, 19.6 percent purchaseda new vehicle, and 75.2 percent boughta used vehicle. Among American Indi-ans, Aleuts, and Eskimos, 4.2 percentleased a vehicle, 16.5 percent purchaseda new vehicle, and 79.4 percent pur-chased a used one.

ConclusionThe 1999–2000 CE Survey data on ve-hicle acquisition indicates that, over-all, purchasing used vehicles is themost common method of acquiring avehicle. This is because it is typicallyless expensive to purchase a used ve-hicle than it is to buy a new vehicle orlease a vehicle. By contrast, cost is not

the predominant factor in choosing topurchase a new vehicle over leasingone. Even though leasing a vehicle isfinancially less of a burden comparedwith purchasing a new vehicle, the nextmost common method of acquiring avehicle is purchasing new vehicles.Leasing remains the least commonmethod.

The 1999–2000 data also suggestthat the choice of a vehicle acquisitionmethod varies by age, race, gender, in-come level, region, and degree of ur-banization. The largest differences oc-cur with respect to income levels,gender, and race.

In addition to demographic differ-ences and various expenses involvedin the decision to lease a vehicle, pur-

chase a new vehicle, or purchase aused vehicle, several other factors en-ter into the decision as well. The avail-ability of leases or new vehicles in dif-ferent regions may affect the frequencywith which one can obtain a lease orfind a suitable vehicle to purchase.Further, the desirability of owning anasset may spur an individual to pur-chase rather than lease. If, instead, thevehicle’s intended use is most impor-tant to a person, then leasing might bepreferred. Finally, the types of vehiclesavailable under a lease may impel aconsumer to lease rather than buy: if aconsumer can drive a luxury car by leas-ing it for the same cost as purchasinga standard car, he or she may prefer tolease.


Table 1. Percent of consumer units reporting vehicle acquisitions, by type of acquisition, selected consumer unitcharacteristics, 1999–2000

All ......................................................................................... 100.0 100.0 7.73 26.15 66.12

Income:1

Quintile 1 ......................................................................... 20.0 9.3 4.7 14.4 80.9Quintile 2 ......................................................................... 20.0 15.3 4.0 19.9 76.0Quintile 3 ......................................................................... 20.0 20.4 5.5 20.0 74.5Quintile 4 ......................................................................... 20.0 26.0 7.2 26.2 66.6Quintile 5 ......................................................................... 20.0 29.0 10.2 35.8 54.1

Region:Northeast ........................................................................ 19.3 17.2 12.3 29.0 58.7Midwest .......................................................................... 23.6 26.8 8.1 21.9 70.0West .............................................................................. 34.9 34.2 4.7 28.5 66.8South .............................................................................. 22.2 21.8 8.5 25.5 66.1

Degree of urbanization:Urban .............................................................................. 87.6 85.6 8.5 26.8 64.8Rural .............................................................................. 12.4 14.5 3.3 22.6 74.1

Race:White .............................................................................. 83.8 87.4 8.0 26.5 65.5Black .............................................................................. 12.1 8.9 5.3 19.6 75.2American Indian, Aleut, Eskimo ...................................... 1.0 1.1 4.2 16.4 79.4Asian or Pacific Islander ................................................. 3.1 2.7 7.4 42.3 50.3

Age:Under 25 ......................................................................... 7.5 7.2 5.4 16.9 77.725–34 .............................................................................. 17.6 21.5 7.9 22.5 69.635–44 .............................................................................. 22.3 27.5 8.1 22.4 69.445–54 .............................................................................. 19.6 23.4 8.2 26.1 65.755–64 .............................................................................. 12.7 11.4 8.6 37.1 54.365–74 .............................................................................. 10.7 6.4 5.8 39.1 55.175 and older .................................................................... 9.6 2.6 4.9 41.6 53.5

Gender:Male ................................................................................ 54.3 57.7 9.6 20.6 69.9Female ............................................................................ 45.7 42.4 11.5 36.9 51.5

1 Percentage represents the percent of complete reporters.

Among groups Within groups

Percent ofgeneral

population

Percentof all

acquisitionsLeased Bought new Bought used

Table 2. Costs and term of vehicle acquisitions, by type of acquisition, 1999–2000

Leased .............................................. $353 $ 868 39Bought new1 ...................................... $399 $2,914 54Bought used ...................................... $273 $1,147 43

1 The bought new and bought used categories represent vehicles that were financedand still had payments remaining.

Averagemonthlypayment

Average lengthof term

(months)Average

downpayment


1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

Year

64

66

68

70

72

74

76

78

Percent of all acquisitions

64

66

68

70

72

74

76

78


Purchase of used vehicle

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

Year

20

22

24

26

28

30


20

22

24

26

28

30


Purchase of new vehicle

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

Year

0

2

4

6

8

10

12


0

2

4

6

8

10

12


Lease of vehicle

Chart 1. Trends in vehicle acquisition methods, 1991-2000


Out-of-PocketExpenditures byConsumer Units withPrivate Health Insurance

Eric J. Keil is an economist in the Branch ofInformation and Analysis, Division of Con-sumer Expenditure Surveys, Bureau of LaborStatistics.

ERIC J. KEIL Although managed-care healthplans have been around for quite some time, rising medi-

cal costs in the 1970s, along withchanges in Federal law, set the stagefor increased interest in such plans. Asa result, health maintenance organiza-tions (HMOs) have grown steadily inpopularity since the 1970s, while thepopularity of more traditional fee-for-service health plans has declined.1 In-creases in health care costs continueto stir national debate and have prom-pted much criticism of current methodsof dealing with high-cost health care.Although many solutions to the prob-lem have been proposed, no significantchanges have occurred.

In this article, data from the 1999 and2000 Consumer Expenditure (CE) Inter-view surveys are used to show thatthere are differences in certain out-of-pocket medical expenditures between

consumer units insured through HMOsand those insured through fee-for-ser-vice plans. Demographic characteris-tics of consumer units are examined aswell, to aid in our understanding ofspending patterns with regard to healthinsurance.

Study methodologyThe sample for this study was restrictedto those consumer units who com-pleted all four quarterly interviews. Allinterviews must have occurred betweenJanuary 1999 and December 2000. Inaddition, these consumer units musthave had private health insurance forat least one quarter during the periodin which they were interviewed. Be-cause the CE Interview survey doesnot match medical expenditures withthe health plans responsible for cover-ing them, the sample was further re-stricted either to those consumer unitswho had one private health plan or tothose whose multiple plans were all ofone type, either HMO or fee for ser-vice. This strategy allowed consumerunits to be grouped into two separatecategories: Those with HMO coverageor those with fee-for-service coverage.In either case, it was possible for a con-sumer unit to have a member who wasalso covered by Medicare or Medicaid.

Health care expenditures from theCE Interview survey are out-of-pocketexpenditures. They consist of expen-ditures paid for medical services, prod-

1 Consumer expenditure data show in-creasing expenditure levels and percentagesreporting for HMO insurance. (In 1984, av-erage annual expenditures were at $15 with 3percent reporting; by 1993, they stood at$110 with 10 percent reporting; and in 2000,expenditures reached $254 with more than20 percent reporting.) The data also showdecreasing expenditure levels and percent-ages reporting for fee-for-service insurance:in 1997, expenditures were $100 with 8 per-cent reporting; by 2000, they reached $77with 5 percent reporting. Due to changesmade to the Interview survey in 1996, it isnot practical to show fee-for-service expen-diture levels or percentage reporting prior to1977.


ucts, and supplies that are net of anypayments or reimbursements fromhealth insurance plans, governmentprograms, or any other third-party pay-ers.2

DefinitionsTwo definitions are essential to an un-derstanding of the material presentedin this article:

Health maintenance organization.There are two basic types of HMOs.The first is the group or staff type, inwhich the participant goes to a centralfacility (a group health center) to re-ceive care. The second type is the in-dependent practice association (IPA),in which providers work from their in-dividual offices and are referred to asprimary care physicians. Expenses inthis type of plan are usually covered infull, or there is a modest copayment atthe time of the visit.

Fee-for-service plan. Commercialhealth insurance plans encompassboth traditional fee-for-service plansand preferred provider organizations.In these plans, a fee is charged for eachmedical service rendered or for all medi-cal equipment purchased. In traditionalfee-for-service plans, participants re-ceive medical care from the providersthey choose. The plan reimburses ei-ther the provider or the individual forsome or all of the cost of care received.Participants in a preferred provider or-ganization are given a list of doctorsfrom which they may choose. If theychoose to go to one of the doctors onthe list, the amount of expenses cov-ered is higher than if they had gone toa doctor who is not on the list.

The impact of health insurance onmedical expendituresAs with most products and services,health care expenditures are affectedby interactions between prices andquantities demanded. One major differ-ence is that health insurance acts as athird-party payer for health-related

products and services. This aspect canalter expenditure levels both directlyand indirectly. In brief, the presence ofhealth insurance can affect medical ex-penditures in the following ways:

1. Differences in payment and ben-efit structures between the twotypes of health plans can lead todirect differences in the out-of-pocket component of health carespending. In other words, givenconsumer units with identicalmedical consumption, those withHMO insurance may pay less foreach medical bill in comparisonwith those with fee-for-service in-surance. This difference effec-tively lowers the out-of-pocketprice of heath care to HMO mem-bers, which, in turn, tends tolower expenditure levels for health-care-related items or services.

2. The aforementioned differencesin payment and benefit structurescan lead indirectly to differentspending patterns between par-ticipants in the two types of healthinsurance plans. A consumer unitwho expects to have high medi-cal bills might decide to select in-surance that will cover more ofthe costs. In addition, lower out-of-pocket costs may have an ef-fect on the quantity of medicalitems and services demanded.Because HMO insurance planscover a larger proportion of thebill, one might expect higher us-age by those consumer unitswith that type of insurance. Thedifferent spending patterns trans-late into a higher quantity de-manded by consumer units inHMO plans, which, in turn, tendsto raise expenditure levels, all elseheld constant.

3. Administrative differences affectthe selection of a health plan. Aconsumer unit who anticipatesusing medical services with great-er frequency might seek an in-surance plan with a low adminis-trative burden or one that allowsmore flexibility in choosing pro-

viders. These considerationsmay tend to counteract eachother. A common assumption isthat HMOs tend to require lesspaperwork, whereas fee-for-ser-vice health plans offer greaterflexibility in choosing physiciansor other health care services. Theoverall effect on expenditures isdifficult to determine.

Health care expenditures by typeof insuranceThe CE Survey collects comprehensivespending data for medical goods andservices as well as detailed informationregarding insurance coverage, includ-ing the type of health plan and the out-of-pocket costs for premiums. The Sur-vey classifies these expenditures into17 categories. (See table 1.) Summingup the medical expenditure componentsreveals that total out-of-pocket medi-cal spending was significantly higher,on average, for those who had fee-for-service insurance, than for those whohad HMO coverage ($2,315 per year,as opposed to $1,789). Of the 17 cat-egories, 6 were found to be significantlydifferent between the two groups ofconsumer units.

Differences were noted for healthcare insurance, physicians’ services,laboratory tests and x rays, hospitalservices other than room, prescriptiondrugs and medicine, and dental care.In each case, expenditures were greaterfor consumer units in fee-for-servicehealth plans.3 Table 1 shows that thelargest difference in annual out-of-pocket spending, in absolute terms,was for health care insurance ($159):4

consumer units with fee-for-service in-surance paid $1,029 per year, on aver-

2 Cash reimbursements paid directly tothe consumer unit are reported only infre-quently in the CE Survey.

3 The following medical expenditure itemswere found not to be statistically differentbetween the two types of health plans: Pur-chase of eyeglasses and accessories, includinginsurance; purchase of medical or surgicalequipment for general use; purchase of sup-portive or convalescent medical equipment;hearing aids; eye exams, treatment, or sur-gery; services by other medical profession-als; hospital room and meals; care in a con-valescent or nursing home; other medical careservices; rental of medical or surgical equip-ment for general use; and rental of support-ive or convalescent equipment.

4 Health insurance expenditures includethose captured by payroll deductions.


age, while those with HMO insurancepaid $870. Other significant differencesin spending included physicians’ ser-vices ($210 for fee-for-service plans,$129 for HMOs), laboratory tests and xrays ($38, compared with $15), hospitalservices other than room ($68 and $37),prescriptions drugs and medicines($329 and $236), and dental services($311, as opposed to $265).

A similar analysis shows that con-sumer units with fee-for-service insur-ance had a higher percentage report-ing for several medical expenditurecategories. In this article, percent re-porting is defined as the percentage ofconsumer units having at least one, butpossibly more, expenditures during theyear they were interviewed. Table 2shows that there were significant dif-ferences in percent reporting for labo-ratory tests and x rays (23 percent forfee-for-service plans, 13 percent forHMOs), hospital services other thanroom (16 percent and 13 percent), pre-scription drugs and medicines (80 per-cent and 75 percent), dental care (51percent, compared with 48 percent),purchases of medical or surgical equip-ment (4 percent, as opposed to 2 per-cent), and eye exams, treatment, or sur-gery (32 percent and 28 percent).

Although the percentage reportingfor all medical expenditures was higherfor the fee-for-service group, the num-ber of reported expenditures per medi-cal expenditure item was generallyhigher for the HMO group. Significantdifferences in reported expenditurewere noted for physicians’ services(13,113 for HMO plans, 11,176 for fee-for-service arrangements),5 prescrip-

tion drugs (26,871, compared with24,088), dental care (6,449 and 5,748),and eyeglasses and accessories (2,445and 1,909). The number of reported ex-penditures was higher for the fee-for-service group only for lab tests and xrays (1,451, as against 940).6

Demographic differences betweenthe two insured groupsA demographic analysis shows that thetwo groups of insured were similar withrespect to age, income, family size, andthe number of children living in theconsumer unit. There was no statisti-cally significant difference between in-comes ($43,226 for those in HMO plans,$43,728 for fee-for-service partici-pants), but there were slight differ-ences with respect to age, family size,and number of children. Althoughthere was a statistical difference in age,it was small, with an average age of 50for the fee-for-service group and 48 forthe HMO group. Similarly, consumerunits with fee-for-service plans, onaverage, were composed of 2.6 per-sons, of which 0.80 were children; con-sumer units with HMO insurance com-prised 2.7 persons, of which 0.91 werechildren. The demographic differencesbetween these two groups may not belarge enough to be considered a con-tributing factor in expenditure differ-ences.

Looking at distributions of insuredconsumer units by age of the referenceperson, one can see that there weremore units with HMO insurance in the

group aged 25 to 54, but more consumerunits in fee-for-service plans in theupper age categories. 7 (See chart 1.)The distributions of insured consumerunits with respect to their size do notshow much difference (chart 2), but thedistributions with respect to numbersof children in the unit indicate that therewere more fee-for-service consumerunits with no children than HMO unitswith no children. (See chart 3.)

In sum, out-of-pocket expendituresand spending patterns vary betweenfee-for-service and HMO health plans.Significant expenditure differences ex-ist for health care insurance, physi-cians’ services, lab tests and x rays,hospital services other than room, pre-scription drugs, and dental care. Ineach case, consumer units with HMOinsurance had lower out-of-pocket ex-penditures for these items. They alsohad a lower percentage reporting formany of the items, but a higher numberof actual reported expenditures withinitem categories. The higher frequenciesfor reported expenditures may be a re-sult of perceived lower costs. Consum-ers who have HMO insurance gener-ally incur lower out-of-pocket medicalcosts despite a higher number of re-ported expenditures. Their lower medi-cal expenditures may be more the re-sult of differences in plan benefits. Thedemographic makeup of the twogroups of insured is similar with respectto income, age, family size, and the num-ber of children in the consumer unit.Although some of the differencesfound are statistically significant, theyare nonetheless small.

6 Care must be taken in evaluating per-centages of consumer units reporting a medi-cal expenditure, as well as the total volumeof expenditures, because medical goods andservices that are completely paid for by athird party are not recorded in the CE Inter-view survey. .

5 All figures in parentheses in this para-graph are in millions.

7 Consumer units whose reference personis eligible for Medicare also can have mem-bers who are insured through a fee-for-ser-vice or HMO plan.


Under 25 25 to 34 35 to 44 44 to 54 55 to 64 65 to 74 75 and older

Age of reference person

0

5

10

15

20

25

30

Percent

0

5

10

15

20

25

30

Percent

HMOFFS

One to three Four to five Six and up

Number of persons in consumer unit

0

10

20

30

40

50

60

70

80

Percent

0

10

20

30

40

50

60

70

80

Percent

HMOFFS

None One Two to three Four and up

Number of children in consumer unit

0

10

20

30

40

50

60

Percent

0

10

20

30

40

50

60

Percent

HMOFFS

Percentages of consumer units participating in health maintenance organizations(HMO) and fee-for-service (FFS) health care plans, by age of reference person,Consumer Expenditure Survey, 1999-2000

Chart 1.

Chart 3.

Chart 2. Percentages of consumer units participating in health maintenance organizations(HMO) and fee-for-service (FFS) health care plans, by number of persons in theconsumer unit, Consumer Expenditure Survey, 1999-2000

Percentages of consumer units participating in health maintenance organizations(HMO) and fee-for-service (FFS) health care plans, by number of children in theconsumer unit, Consumer Expenditure Survey, 1999-2000


Table 1. Average annual health care expenditures by type of insurance, Consumer Expenditure Interview Survey, 1999–2000

Total medical expenditures .......................................................................... $2,314.71 $1,789.24 1 $525.47

Health insurance.......................................................................................... 1,028.67 870.45 1 158.22

Physicians’ services .................................................................................... 210.14 128.65 1 81.49Lab tests and x rays .................................................................................... 37.88 14.63 1 23.25Hospital services other than room .............................................................. 67.64 37.34 1 30.30Prescription drugs and medicine ................................................................. 329.02 236.47 1 92.55Dental care .................................................................................................. 310.96 265.42 1 45.55

Eyeglasses and accessories, vision insurance.......................................... 73.09 77.72 – 4.63Purchase of medical or surgical equipment for general use ....................... 2.93 2.16 0.77Purchase of supportive or convalescent medical equipment ..................... 3.52 6.21 – 2.69Hearing aid ................................................................................................... 21.77 13.68 8.09Eye examinations, treatment, or surgery .................................................... 42.75 44.94 – 2.19Services by other medical professionals .................................................... 58.49 38.91 19.58Hospital room and meals ............................................................................. 56.42 31.24 25.18Care in a convalescent or nursing home .................................................... 58.70 9.05 49.66Other medical care services ....................................................................... 11.56 11.28 .28Rental of medical or surgical equipment for general use ............................ .50 .62 – .11Rental of supportive or convalescent equipment ........................................ .69 .51 .18

1 Significantly different at the 95-percent confidence level.

Medical expenditure itemFee-for-service

insuranceHealth maintenance

organizationDifference in means

Table 2. Percentage reporting medical expenditures, Consumer Expenditure Interview Survey, 1999–2000

Health insurance.......................................................................................... 73 72 1

Physicians’ services .................................................................................... 70 67 3Lab tests and x rays .................................................................................... 23 13 1 10Hospital services other than room .............................................................. 16 13 1 3Prescription drugs and medicine ................................................................. 80 75 1 5Dental care .................................................................................................. 51 48 1 3

Eyeglasses and accessories, vision insurance.......................................... 34 35 – 1Purchase of medical or surgical equipment for general use ....................... 4 2 1 2Purchase of supportive or convalescent medical equipment ..................... 4 3 1Hearing aid ................................................................................................... 3 3 0Eye examinations, treatment, or surgery .................................................... 32 28 14Services by other medical professionals .................................................... 16 15 1Hospital room and meals ............................................................................. 9 8 1Care in a convalescent or nursing home .................................................... 1 1 0Other medical care services ....................................................................... 6 4 2Rental of medical or surgical equipment for general use ............................ 1 1 0Rental of supportive or convalescent equipment ........................................ 1 1 0

1 Significantly different at the 95-percent confidence level.

Medical expenditure itemPercent reportingfee-for-service

plan

Percent reportinghealth maintenance

organization

Difference inpercent reporting


Table 3. Frequencies of health care expenditures, Consumer Expenditure Interview Survey, 1999–2000

Health insurance....................................................................................... 34,612 41,852 1 – 7,240

Physicians’ services ................................................................................. 5,748 13,113 1 – 1,937

Lab tests and x rays ................................................................................. 1,451 940 1 511Hospital services other than room ........................................................... 1,124 1,062 62Prescription drugs and medicine .............................................................. 24,088 26,871 1 – 2,783Dental care ............................................................................................... 5,748 6,449 1 – 701

Eyeglasses and accessories, vision insurance....................................... 1,909 2,445 1 – 536Purchase of medical or surgical equipment for general use .................... 225 186 39Purchase of supportive or convalescent medical equipment .................. 195 234 – 39Hearing aid ................................................................................................ 212 216 –4Eye examinations, treatment, or surgery ................................................. 1,771 1,924 – 153Services by other medical professionals ................................................. 1,836 1,857 – 21Hospital room and meals .......................................................................... 544 628 – 84Care in a convalescent or nursing home ................................................. 127 81 46Other medical care services .................................................................... 344 345 –1Rental of medical or surgical equipment for general use ......................... 76 90 – 14Rental of supportive or convalescent equipment ..................................... 80 97 – 17

1 Significantly different at the 95-percent level.

Medical expenditure itemFrequency of

reporting, fee-for-service plans

(in milions)

Frequency ofreporting, health

maintenanceorganizations(in millions)

Difference infrequency of

reporting


Expenditures onEntertainment

Neil Tseng is an economist in the Branch ofProduction and Control, Division of Con-sumer Expenditure Surveys, Bureau of LaborStatistics.

NEIL TSENG O ver the past half-century, the increase in incomes and decline in hours worked have al-

lowed American consumers to enjoymore leisure time and increase theirspending on entertainment. In 2000,spending on entertainment by Ameri-can consumers totaled approximately$203 billion (see table 1), almost 3 timesthe amount that Americans spent oneducation. Using data from the Con-sumer Expenditure (CE) Survey, thisarticle looks at the level of expenditureson entertainment, its share of nationalaggregate expenditures, and the waysin which selected demographic groupsallocate these expenditures. The articlehighlights entertainment expendituresby consumer units 1 in 2000, classifiedby age of the reference person,2 in-come quintiles 3 of complete incomereporters,4 and education of the refer-ence person.

The CE Survey divides entertain-ment expenditures into four categories:Fees and admissions; televisions, ra-dios, and sound equipment; pets, toys,and playground equipment; and other

entertainment supplies, equipment, andservices. Fees and admissions, whichaccounted for 28 percent of entertain-ment expenditures in 2000, include ex-penses for out-of-town trips, fees forrecreational lessons, and the cost ofadmission to sporting events, culturaland theatrical events, the movies, andspecial events, such as live musicalperformances. Television, radios, andsound equipment accounted for 33 per-cent of entertainment spending andinclude color televisions, DVD players,VCRs, CD players, video game con-soles and software, videotapes anddiscs, and speakers and various otherhome theater sound systems. Pets,toys, and playground equipment ac-counted for 18 percent of entertainmentspending and includes toys, games,and playground equipment; hobbiesand tricycles; and pet food, veterinar-ian services and pet services. Otherentertainment supplies, equipment, andservices accounted for 21 percent ofentertainment spending and includes“volatile” expenditures, such as therental or purchase of recreational ve-hicles and the purchase of boats. Ex-penditures on many of the items in thecategory tend to fluctuate from year toyear, chiefly because, each year, rela-tively few consumers purchase theseexpensive items (such as a boat with amotor or a motorized camper) and in-creases or decreases in the percentageof consumers purchasing the items can

1 See “Glossary” in Appendix A at the endof this anthology for the definition of a con-sumer unit.

2 See “Glossary” in Appendix A at the endof this anthology for the definition of refer-ence person.

3 See “Glossary” in Appendix A at the endof this anthology for the definition of quintilesof income before taxes.

4 See the glossary at the end of this an-thology for the definition of complete in-come reporter.


have a large effect on the mean expen-diture. For example, consumer unitswho reported no expenditures on mo-torized recreational vehicles arecounted as spending $0.00. In 1999, 0.33percent of consumer units reportedpurchasing a motorized recreationalvehicle, and they spent an average of$171, whereas, in 2000, the percent re-porting was 0.24 percent, and the aver-age amount spent was $82.

AgeIn 2000, the share of total aggregateentertainment spending accountedfor by consumer units with referencepersons in two age groups—thoseunder 35 and those 55 and older—was smaller than their populationshare. The under-35 group accountedfor 25 percent of the total population,but spent 22 percent of the total of$203 billion that U.S. consumers al-located to entertainment in 2000,whereas those 55 and older had apopulation share of 33 percent andspent 25 percent of the total amountallocated to entertainment. Con-sumer units with reference personsin the age group from 35 to 54 had apopulation share of 42 percent, butaccounted for more than half of thetotal of $203 billion dollars spent onentertainment.

As regards the individual catego-ries of entertainment, persons under theage of 35 and those 55 and older spentless on entertainment than their popu-lation share in all four categories of en-tertainment, whereas those between theages of 35 and 54 spent more than theirpopulation share on each of the cat-egories. Although those under 35 madeup 42 percent of the population, theirshare of spending on the four subcat-egories of entertainment was as follows:Fees and admissions, 55 percent of theaggregate entertainment share; TVs,radios, and sound equipment, 50 per-cent; pets, toys, and playground equip-ment, 53 percent; and other entertain-ment supplies, equipment, andservices, 55 percent.

EducationThis section examines consumer units

in two broad categories of educationalattainment. The first, those who didnot graduate from college, comprisesfour classes: Those who did notgraduate from high school, highschool graduates, high school gradu-ates with some college, and thosewith an associate’s degree. The sec-ond category, college graduates,consists of two classes: Those witha bachelor’s degree and those with amaster’s, professional, or doctoraldegree. Consumer units with refer-ence persons who did not graduatefrom college had a population shareof 74 percent and accounted for 60.5percent of the aggregate expenditureson entertainment, whereas collegegraduates had a population share of26 percent, yet accounted for 39.5percent of the aggregate expenditureon entertainment.

Of the 4 subclasses making up thegroup who did not graduate from col-lege, 3 had an aggregate expenditureshare that was lower than their popula-tion share: Those who did not gradu-ate high school, 8 percent expenditureshare, compared with 17 percent popu-lation share; high school graduates, 24percent expenditure share, as opposedto 29 percent population share; andhigh school graduates with some col-lege, 20 percent expenditure share and21 percent population share. Only thosewith associate’s degrees had a spend-ing share exceeding their populationshare (9 percent, compared with 8 per-cent). These statistics are evidencethat an increase in education level leadsto an increase in average income, en-abling the more educated to spend moreon leisure and recreation. Average in-comes for the four classes were as fol-lows: Those who did not graduate highschool, $23,329; high school graduates,$36,134; high school graduates withsome college, $38,837; associate’s de-gree, $50,060). Among the collegegraduates, those with a bachelor’sdegree and those with advanced de-grees had aggregate expenditureshares of 25 percent and 15 percent,respectively, and population sharesof 17 percent and 9 percent. Thesefigures are likely attributable to the

fact that, as their education levels in-creased, so did their incomes, pro-viding them with more discretionaryincome to spend on entertainment.

Income quintilesAn examination of spending on enter-tainment by income quintile reveals thatthe proportion of aggregate expendi-tures allocated to entertainment rangedfrom 9 percent by the lowest quintile to40 percent by the highest quintile. Theaggregate amount spent on entertain-ment by complete income reporters was$158 billion. Not surprisingly, consumerunits in the highest quintile contributedthe most to each of the four categoriesof entertainment expenditure. Theseconsumer units spent more than $22billion on fees and admissions; approxi-mately $17 billion on televisions, radios,and sound equipment; $10 billion onpets, toys, and playground equipment;and $13 billion on other entertainmentsupplies, equipment, and services. Toput the figures in perspective, the $22billion spent on fees and admissionswas more than twice the amount spentby consumers in the fourth incomequintile and almost 7 times the amountspent by those in the first quintile.

The proportion of total aggregateentertainment expenditures allocated tofees and admissions ranged from nearly7 percent for those in the lowest quint-ile to more than 50 percent for those inthe highest quintile. For pets, toys, andplayground equipment, expendituresranged from 7 percent for those in thelowest quintile to 37 percent for thosein the highest quintile. Total entertain-ment expenditures allocated to otherentertainment supplies, equipment, andservices ranged from 8 percent for thosein the lowest quintile to almost 38 per-cent for those in the highest. Althoughthe lowest quintile contributed only 11percent toward televisions, radios, andsound equipment, while the highestcontributed 33 percent, the 11-percentfigure accounted for the largest shareof the bottom quintile’s expenditureson entertainment. Apparently, the cat-egory may be the main form of enter-tainment for those in the lowest incomequintile.


In sum, consumers spent approxi-mately $203 billion on entertainment in2000, with about $56 billion going tofees and admissions; $68 billion to tele-vision, radios, and sound equipment;$36 billion to pets, toys, and play-ground equipment; and $43 billion toother entertainment supplies, equipment,

and services. Those with associate’s orhigher degrees accounted for 49 percentof the aggregate expenditure on enter-tainment, well above their populationshare of 34 percent. Consumer unitswith reference persons between theages of 35 and 54 had a population shareof 42 percent, but accounted for 53 per-

cent of the aggregate expenditure onentertainment. Finally, consumer unitswith reference persons in the highestincome quintile had a population shareof 20 percent, but accounted for 40 per-cent of the aggregate expenditure onentertainment.

Table 1. Average annual entertainment expenditures and aggregate expenditures, by age of reference person, ConsumerExpenditure Survey, 2000

Aggregate Total entertainment ................................................................... $10,687 $203,712 100.0 100.0 Fees and admissions ............................................................ 2,911 56,308 100.0 100.0 Televisions, radios, and sound equipment ........................... 3,618 67,999 100.0 100.0 Pets, toys, and playground equipment .................................. 1,904 36,452 100.0 100.0 Other entertainment supplies, equipment, and services ...... 2,254 42,910 100.0 100.0

Under age 35 Total entertainment ................................................................... 1,485 44,530 21.9 24.9 Fees and admissions ............................................................ 366 10,923 19.4 24.9 Televisions, radios, and sound equipment ............................ 577 16,796 24.7 24.9 Pets, toys, and playground equipment .................................. 262 8,092 22.2 24.9 Other entertainment supplies, equipment, and services ...... . 280 8,719 20.3 24.9

Aged 35 to 54 Total entertainment ................................................................... 4,695 107,837 52.9 41.9 Fees and admissions ............................................................ 1,352 31,082 55.2 41.9 Televisions, radios, and sound equipment ........................... 1,485 34,135 50.2 41.9 Pets, toys, and playground equipment .................................. 835 19,210 52.7 41.9 Other entertainment supplies, equipment, and services ...... 1,023 23,410 54.5 41.9

Aged 55 and older Total entertainment ................................................................... 3,024 51,300 25.2 33.2 Fees and admissions ............................................................ 828 14,302 25.4 33.2 Televisions, radios, and sound equipment ............................ 980 17,068 25.1 33.2 Pets, toys, and playground equipment .................................. 545 9,149 25.1 33.2 Other entertainment supplies, equipment, and services ...... 671 10,781 25.1 33.2

Averageannual

expenditure

Aggregateexpenditure(in millions)

Aggregateshare

(in percent)

Populationshare

(in percent)Age of reference person and type of expenditure


Table 2. Average annual entertainment expenditures and aggregate expenditures, by education of reference person,Consumer Expenditure Survey, 2000

Total entertainment ....................................................................... $896 $15,948 7.8 16.5 Fees and admissions. .............................................................. 132 2,365 4.2 16.5 Televisions, radios, and sound equipment ............................... 418 7,480 11.0 16.5 Pets, toys, and playground equipment ..................................... 192 3,354 9.2 16.5 Other entertainment supplies, equipment, and services .......... 54 2,749 6.4 16.5

Total entertainment ....................................................................... 1,519 48,475 23.8 29.2 Fees and admissions ............................................................... 298 9,516 16.9 29.2 Televisions, radios, and sound equipment ................................ 566 18,088 26.6 29.2 Pets, toys, and playground equipment ..................................... 303 9,660 26.5 29.2 Other entertainment supplies, equipment, and services .......... 351 11,211 26.1 29.2




Total entertainment ............................................................................. 3,011 29,476 14.5 8.9 Fees and admissions ............................................................... 1,227 11,937 21.2 8.9 Televisions, radios, and sound equipment ................................ 787 7,684 11.3 8.9 Pets, toys, and playground equipment ..................................... 500 4,958 13.6 8.9 Other entertainment supplies, equipment, and services ......... 498 4,897 11.4 8.9

Averageannual

expenditure


Aggregateshare

(in percent)

Populationshare

(in percent)

Did not graduate high school (Income before taxes = $23,329)

High school graduate (Income before taxes = $36,134)

High school graduate with some college(Income before taxes = $38,837)

Associate’s degree (Income before taxes = $50,060)

Bachelor’s degree (Income before taxes = $64,201)

Master’s, professional, or doctoral degree (Income before taxes = $84,438)

Education of reference person and type of expenditure


Averageannual

expenditure


Aggregateshare

(in percent)

Populationshare

(in percent)

Table 3. Average annual entertainment expenditures and aggregate expenditures, by quintiles of income before taxes,Consumer Expenditure Survey, 2000

Lowest quintile Total entertainment ................................................................... $837 $13,545 8.6 20.0 Fees and admissions ............................................................ 198 3,227 7.4 20.0 Televisions, radios, and sound equipment ............................ 363 5,902 11.1 20.0 Pets, toys, and playground equipment .................................. 122 1,946 6.9 20.0 Other entertainment supplies, equipment, and services ...... 154 2,470 7.5 20.0

Second quintile Total entertainment ................................................................... 1,147 18,527 11.7 20.0 Fees and admissions ............................................................ 250 4,100 9.4 20.0 Televisions, radios, and sound equipment ............................ 465 7,551 14.2 20.0 Pets, toys, and playground equipment .................................. 239 3,780 13.4 20.0 Other entertainment supplies, equipment, and services ...... 192 3,096 9.4 20.0

Third quintile Total entertainment ................................................................... 1,609 25,986 16.4 20.0 Fees and admissions ............................................................ 331 5,408 12.4 20.0 Televisions, radios, and sound equipment ............................ 590 9,571 18.0 20.0 Pets, toys, and playground equipment .................................. 337 5,532 18.9 20.0 Other entertainment supplies, equipment, and services ...... 351 5,475 16.9 20.0

Fourth quintile Total entertainment ................................................................... 2,324 37,476 23.7 20.0 Fees and admissions ............................................................ 547 8,897 20.4 20.0 Televisions, radios, and sound equipment ............................ 782 12,709 23.9 20.0 Pets, toys, and playground equipment .................................. 422 6,714 23.8 20.0 Other entertainment supplies, equipment, and services ...... 573 9,156 27.8 20.0

Highest quintile Total entertainment ................................................................... 3,866 62,579 39.6 20.0 Fees and admissions ............................................................ 1,349 22,025 50.5 20.0 Televisions, radios, and sound equipment ............................ 1,071 17,441 32.8 20.0 Pets, toys, and playground equipment ................................. 656 10,466 37.1 20.0 Other entertainment supplies, equipment, and services ...... 790 12,647 38.4 20.0

Quintile of income and type of expenditure


Table 4. Average annual expenditures of different demographic groups and shares spent on entertainment, byeducation, age, and income quintile, Consumer Expenditure Survey, 2000

EducationDid not graduate high school ...................................................... $23,386 3.8High school graduate ................................................................. 32,447 4.7High school graduate with some college .................................... 35,999 4.9Bachelor’s degree ....................................................................... 50,785 5.5Master’s, professional, or doctoral degree ................................. 60,527 5.0

Age Under 25 ..................................................................................... 22,543 4.825 to 34 ....................................................................................... 38,945 4.835 to 54 ....................................................................................... 45,655 5.155 and older ................................................................................ 32,937 4.5

Income quintile Lowest ........................................................................................ 17,940 4.7Second ....................................................................................... 26,550 4.3Third ........................................................................................... 34,716 4.6Fourth ......................................................................................... 46,794 5.0Highest ....................................................................................... 75,102 5.1

Average share ofexpenditure spent on

entertainment(in percent)

Averageannual

expenditureCharacteristic


Travel Expendituresin 2000

GEORGE JANINI Although most of the averageconsumer’s spending budget is devoted to food, housing,

and transportation, consumer unitsthat went on trips in the year 2000 spentan average of $875 in travel expenses.The total amount spent on travel by allconsumers was roughly $32 billion.This article uses data from the 2000Consumer Expenditure (CE) Interviewsurvey to look at spending on trips andvacations by various demographicgroups.

MethodologyTravel expenditures in the CE Surveyare broken down into five main groups:Transportation, food, lodging, enter-tainment, and the purchase of gifts.Transportation expenditures include allcosts of traveling to and from the des-tination, as well as transportation costsincurred on the trip. All modes of trans-portation, such as plane, boat, ship, car,taxi, truck, motorcycle, and camper, areconsidered. Food expenditures includeall food and alcohol consumed on thetrip. Lodging expenses encompass thecosts for hotels, motels, cottages, trailercamps, and other lodging on the trip.Entertainment expenditures include alltypes of entertainment, such as admis-sion to sporting events, parks, muse-ums, and tours, as well as any type offee related to these events. Gift expen-ditures are the total cost of all giftspurchased on the trip for persons out-side the consumer unit.

George Janini is an economist in the Branchof Information and Analysis, Division ofConsumer Expenditure Surveys, Bureau ofLabor Statistics.

Age, income, and the compositionof the consumer unit are the character-istics used in the comparison. The dataare reported as both average annualexpenditures and aggregate expendi-tures, for each of the spending groups.The data are annual average amountsspent during the year 2000 on all tripsand not the amount spent per trip. Av-erage and aggregate expenditures aregiven only for those consumer unitsthat actually reported a trip in 2000. Allaggregate amounts were estimated withweights derived from the CE Survey.Excluded from the survey are all busi-ness-related expenditures for which theconsumer unit is reimbursed.

Expenditures on travelOverall, consumer units that went ontrips in 2000 spent an average of $352on transportation, $204 on food, $66on entertainment, $76 on gifts, and $177on lodging. These figures aggregatedto about $13 billion spent on transpor-tation, $7.6 billion on food, $2.4 billionon entertainment, $ 2.8 billion on gifts,and $6.5 billion on lodging. Out of ap-proximately 109 million consumer units,34 percent, or 37 million units, reportedtaking a trip or vacation in the year 2000.

AgeThe highest percentage of trip takerswas posted by the group aged 45 to 54,with 38 percent reporting a trip. Thelowest percentage was that of thegroup aged 65 and older, 27 percent.


This group, however, had the highestaverage expenditures on trips of anyof the age groups. It is interesting tonote that the group, consisting mainlyof retirees, spent an average of 4 per-cent of its total average annual expen-ditures on trips and vacations, abouttwice the share spent by most of theother age groups. However, the 65-and-older group did not account for thehighest share of aggregate trip expen-ditures. That distinction went to thegroup aged 45 to 54, with 24 percent ofaggregate trip expenditures in 2000.The group aged 35 to 44 spent almostas much, followed by the 65-and-oldergroup at 19 percent, with the groupsaged 25 to 34 and 55 to 64 each ac-counting for 15 percent. The groupaged 25 and under spent the least, ac-counting for only 4 percent of total tripexpenditures. Except for consumer unitsin the two lowest age groups, aggre-gate expenditure shares were in pro-portion to the population share of thegroup.

IncomeFully 58 percent of consumer units withreported incomes1 over $50,000 took a

trip or vacation in 2000, almost doublethe share of consumer units with re-ported incomes of less than $25,000.With more discretionary income at theirdisposal, higher income consumerunits would be expected to spend moreon travel and trips than lower incomegroups. Consumer units in the highestincome bracket, $75,000 or more, sig-nificantly outspent those in all otherincome groups and almost doubled theaverage spending on trips and vaca-tions of the next highest incomebracket, those reporting income rang-ing from $50,000 to $75,000. Not sur-prisingly, consumer units with reportedincomes of $75,000 or more accountedfor 41 percent of aggregate trip expen-ditures in 2000, whereas the travel ex-penditure of all of the other reportedincome groups combined was 52 per-cent. The classifications by income arebased on complete reporters only,which account for 74 percent of all ofconsumer units.

Composition of the consumerunitA few selected types of consumer unitsare included in this article: Husband-and-wife-only consumer units, hus-bands and wives with children youngerthan 17, single-persons, and one-par-ent consumer units. Forty-two percentof husband-and-wife-only consumerunits reported taking a trip, comparedwith 20 percent of single-person units.Thirty-six percent of husband-and-wifeconsumer units with children younger

than 17 reported taking a trip, as did 24percent of one-parent consumer units.In all, husband-and-wife consumerunits (with and without children) madeabout half of aggregate trip expendi-tures in 2000.

Overall, consumer units reportingincomes of $35,000 or more accountedfor 76 percent of total travel expendi-tures, while making up only 35 percentof the population. Looking at the databy age reveals that the highest spend-ers, on average, were aged 65 and older,while the lowest were under the age of25. The youngest group did not spendmuch, on average, on trips, but did havea relatively high percentage of trip tak-ers. By comparison, the group aged 65and older had the lowest percentage oftrip takers, but spent the most money,on average, on trips. As regards a com-parison of expenditure shares withpopulation shares, the age groups olderthan 35 had similar overall travel expen-ditures and habits. The age groups 35and under had far lower expenditureshares compared with their populationshares. Even though single consumerunits made up 43 percent of the popu-lation, they accounted for just 22 per-cent of aggregate expenditures. By con-trast, husband-and-wife consumerunits and single consumer units ac-counted for 40 percent of the popu-lation, but 58 percent of aggregateexpenditures.

Tables 1, 2, and 3 give an annualsummary of travel expenditures, by se-lected categories.

1 The distinction between complete andincomplete income reporters is based, in gen-eral, on whether the respondent providedinformation on his or her major sources ofincome, such as wages and salaries, self-em-ployment income, and Social Security in-come. In the survey, across-the-board zeroincome reporting was designated as invalid,and the consumer unit thus reporting wascategorized as an incomplete reporter.


Table 3. Average annual travel expenditures, selected types of consumer unit, Consumer Expenditure Interview survey,2000

Number of consumer units (thousands) ............................. 109,367 22,805 20,687 46,948 6,132Population share (percent) .................................................. 100 21 19 43 6Percent of group that reported a trip .................................... 34 42 36 20 24Aggregate travel expenditures (billions) .............................. $32.3 $10.1 $7.2 $6.5 $0.9Share of aggregate expenditures (percent) ........................ 100 34 24 22 3Average annual expenditures

Total trip expenses ......................................................... $875 $1,049 $957 $675 $642Transportation ........................................................... 352 425 370 281 244Food .......................................................................... 204 244 231 149 158Entertainment ............................................................ 66 75 85 43 56Gifts ........................................................................... 76 81 64 85 58Lodging ...................................................................... 177 225 207 118 126

ItemAll

consumerunits

Husbandand

wife only

Husband andwife withchildrenyoungerthan 17

Singleperson

Oneparent

Table 2. Average annual travel expenditures, by pretax income, Consumer Expenditure Interview survey, 2000

Number of consumer units (thousands) ................................ 109,367 31,543 10,759 12,392 11,337 15,424 28,067Population share (percent) ..................................................... 100 29 10 11 10 14 26Percent of group that reported a trip ....................................... 34 22 39 44 64 53 7Aggregate travel expenditures (billions) ................................. $32.3 $2.8 $2.3 $3.9 $6.5 $12.4 $4.4Share of aggregate expenditures (percent) ........................... 100 9 8 13 22 41 7Average annual expenditures

Total trip expenses ............................................................ $875 $404 $557 $718 $886 $1,510 $1,077Transportation .............................................................. 352 177 221 282 358 585 475Food ............................................................................. 204 105 139 173 208 342 212Entertainment ............................................................... 66 26 41 50 69 120 79Gifts .............................................................................. 76 35 51 71 77 135 66Lodging ......................................................................... 177 60 106 143 175 328 245

Pretax income

Lessthan

$25,000

$25,000to

$35,000

$35,000to

$50,000

$50,000to

$75,000

$75,000or

more

Incom-plete

reportingof

income

ItemAll

consumerunits

Table 1. Average annual travel expenditures, by age of reference person, Consumer Expenditure Interview survey, 2000

Number of consumer units (thousands) .............................. 109,367 8,306 18,887 23,983 21,874 14,161 22,155Population share (percent) ................................................... 100 8 17 22 20 13 20Percent of group that reported a trip ..................................... 34 36 34 34 38 36 27Aggregate travel expenditures (billions) ............................... $32.3 $1.2 $4.7 $7.6 $7.9 $4.9 $6.0Share of aggregate expenditures (percent) ......................... 100 4 15 23 24 15 19Average annual expenditures

Total trip expenses .......................................................... $875 $392 $717 $922 $973 $970 $1,025Transportation ............................................................ 352 170 300 374 365 383 428Food ........................................................................... 204 106 177 218 233 228 212Entertainment ............................................................. 66 33 59 74 70 72 67Gifts ............................................................................ 76 31 54 69 91 88 106Lodging ....................................................................... 177 52 127 187 214 199 212

ItemAll

consumerunits Under

2525–34 35–44 45–54 55–64 65 and

older

Age of reference person


Appendix A: Descriptionof the ConsumerExpenditure Survey

The current Consumer Expendi-ture (CE) Survey program beganin 1980. Its principal objective

is to collect information on the buyinghabits of American consumers. Con-sumer expenditure data are used in vari-ous types of research by government,business, labor, and academic ana-lysts. The data are required for peri-odic revision of the Consumer PriceIndex (CPI).

The survey, which is conducted bythe U.S. Census Bureau for the Bureauof Labor Statistics, consists of twocomponents: A Diary, or recordkeeping,survey completed by participating con-sumer units for two consecutive 1-weekperiods; and an Interview survey inwhich expenditures of consumer unitsare obtained in five interviews con-ducted at 3-month intervals.

Survey participants record dollaramounts for goods and services pur-chased during the reporting period, re-gardless of whether payment is madeat the time of purchase. Expenditureamounts include all sales and excisetaxes for all items purchased by theconsumer unit for him- or herself or forothers. Excluded from both surveysare all business-related expendituresand expenditures for which the con-sumer unit is reimbursed.

Each component of the survey que-ries an independent sample of con-sumer units that is representative of theU.S. population. In the Diary survey,about 7,500 consumer units are sampledeach year. Each consumer unit keeps a

diary for two 1-week periods, yieldingapproximately 15,000 diaries a year. Theinterview sample is selected on a rotat-ing-panel basis and yields reports for7,500 consumer units each quarter. Eachconsumer unit is interviewed once perquarter, for five consecutive quarters.Data are collected on an ongoing basisin 105 areas of the United States.

The Interview survey is designedto capture expenditure data that respon-dents can reasonably recall for a pe-riod of 3 months or longer. In general,the data captured report relatively largeexpenditures, such as spending on realproperty, automobiles, and major ap-pliances, or expenditures that occur ona regular basis, such as spending onrent, utilities, and insurance premiums.Including global estimates of spend-ing for food, it is estimated that about95 percent of expenditures are coveredin the Interview survey. Expenditureson nonprescription drugs, householdsupplies, and personal care items areexcluded. The Interview survey alsoprovides data on expenditures incurredon leisure trips.

The Diary survey is designed tocapture expenditures on small, fre-quently purchased items that are nor-mally difficult for respondents to recall.Detailed records of expenses are keptfor food and beverages—both at homeand in eating places—tobacco, house-keeping supplies, nonprescriptiondrugs, and personal care products andservices. Expenditures incurred awayfrom home overnight or longer are ex-


cluded from the Diary survey. Althoughthe diary was designed to collect infor-mation on expenditures that could notbe recalled easily over a given period,respondents are asked to report all ex-penses (except overnight travel ex-penses) that the consumer unit incursduring the survey week.

Integrated data from the BLS Diaryand Interview surveys provide a com-plete accounting of consumer expen-ditures and income, which neither sur-vey component alone is designed todo. Data on some expenditure items arecollected in only one of the compo-nents. For example, the Diary surveydoes not collect data on expendituresfor overnight travel or information onthird-party reimbursements of con-sumer expenditures, as the Interviewsurvey does. Examples of expendituresfor which reimbursements are excludedare medical care; automobile repair; andconstruction, repairs, alterations, andmaintenance of property.

For items that are unique to one orthe other survey, the choice of whichsurvey to use as the source of data isobvious. However, there is consider-able overlap in coverage between thesurveys. Because of this overlap, inte-grating the data presents the problemof determining the appropriate surveycomponent from which to select expen-diture items. When data are availablefrom both survey sources, the morereliable of the two (as determined bystatistical methods) is selected. As aresult, some items are selected from theInterview survey and others, from theDiary survey.

Population coverage and the defi-nition of components of the CE Surveydiffer from those of the CPI. Specifi-cally, consumer expenditure data coverthe total population, whereas the CPIcovers only the urban population. Inaddition, home ownership is treated dif-ferently in the two surveys. Actual ex-penditures of homeowners are reportedin the CE Survey, whereas the CPI usesa rental equivalence approach that at-tempts to measure the change in thecost of obtaining, in the rental market-place, services equivalent to those pro-vided by owner-occupied homes.

Interpreting the dataExpenditures are averages for con-sumer units with specified characteris-tics, regardless of whether a particularunit incurred an expense for a specificitem during the recordkeeping period.The average expenditure for an itemmay be considerably lower than theexpenditure by those consumer unitsthat actually purchased the item. Theless frequently an item is purchased,the greater is the difference betweenthe average for all consumer units andthe average for those purchasing theitem. Also, an individual consumer unitmay spend more or less than the aver-age, depending on its particular char-acteristics. Factors such as income, theages of family members, geographiclocation, taste, and personal prefer-ence also influence expenditures.Furthermore, even within groups withsimilar characteristics, the distribu-tion of expenditures varies substan-tially. These points should be con-sidered in relating reported averagesto individual circumstances.

Users of these survey data alsoshould keep in mind that prices formany goods and services have risensince the survey was conducted. Forexample, rent, as measured by the CPI,rose 8.2 percent between 2000 (annualaverage index) and September 2002.

In addition, sample surveys are sub-ject to two types of error: Sampling andnonsampling. Sampling errors occurbecause the data are collected from arepresentative sample rather than theentire population. Nonsampling errorsresult from the inability or unwilling-ness of respondents to provide correctinformation, differences in interviewers’abilities, mistakes in recording or cod-ing, or other processing errors.

Glossary

Consumer unit. Members of a house-hold related by blood, marriage, adop-tion, or some other legal arrangement;a single person living alone or sharinga household with others, but who isfinancially independent; or two or morepersons living together who share re-sponsibility for at least two out of the

three major types of expenses: Food,housing, and other expenses. Studentsliving in university-sponsored housingare also included in the sample as sepa-rate consumer units.

Reference person. The first membermentioned by the respondent whenasked to “Start with the name of theperson or one of the persons who ownsor rents the home.” It is with respect tothis person that the relationship ofother members of the consumer unit isdetermined.

Total expenditures. The transactioncosts, including excise and sales taxes,of goods and services acquired duringthe interview period. Estimates includeexpenditures for gifts and contributionsand payments for pensions and per-sonal insurance.

Income. The combined income earnedby all consumer unit members 14 yearsor older during the 12 months preced-ing the interview. The components ofincome are wages and salaries; self-employment income; Social Securityand private and government retirementincome; interest, dividends, and rentaland other property income; unem-ployment and workers’ compensationand veterans’ benefits; public assis-tance, Supplemental Security Income,and Food Stamps; rent or meals orboth as pay; and regular contribu-tions for support, such as alimonyand child support.

Complete income reporters. In gen-eral, a consumer unit who provides in-formation on at least one of the majorsources of its income, such as wagesand salaries, self-employment income,and Social Security income. Even com-plete income reporters may not providea full accounting of all income from allsources.

Quintiles of income before taxes. Fivegroups with a similar number of com-plete income reporters, ranked in as-cending order of income. Incompleteincome reporters are not ranked and areshown separately in the quintiles-of-income tables.

consumer expenditure survey anthology, 2003color, icons, highlighting, and shading were all...

Documents