18 th international roundtable on business survey frames beijing, china 17 - 22 october 2004 session...

29
18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision to the Industry Classification System Ron S. Jarmin and Edward D. Walker United States Census Bureau

Upload: cory-mcdonald

Post on 25-Dec-2015

216 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

18th International Roundtable on Business Survey FramesBeijing, China

17 - 22 October 2004Session 4 — Classification Issues

Implementing a Major Revision to the Industry Classification System

Ron S. Jarmin and Edward D. WalkerUnited States Census Bureau

Page 2: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

2

North American Industry Classification System—1997

Background North American partners Canada, Mexico, and

the U.S. developed NAICS collaboratively NAICS provides 3-country comparability for

industries The U.S. adopted NAICS in 1997 NAICS is a major structural and conceptual

departure from its U.S. predecessor, the Standard Industrial Classification (SIC)

The Census Bureau’s Business Register implemented NAICS in 1997

Page 3: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

3

SIC

V

ers

us

NA

ICS

OLD U.S. Standard Industrial Classification–1987

NEW North American Industry Classification System–1997

Divisions Sectors No of U.S. Industries

Code Major Groups Title

No of Indus-

tries Code Title New Total

A 01-09 Agriculture, Forestry, and Fishing 58 11 Agriculture, Forestry, Fishing and

Hunting 20 64

B 10-14 Mining 31 21 Mining 0 29 C 15-17 Construction 26 23 Construction 3 28

D 20-39 Manufacturing 459 31-33 Manufacturing 79 474

22 Utilities 6 10 E 40-49

Transportation, Communications and Public Utilities

67 48-49 Transportation and Warehousing 28 57

F 50-51 Wholesale Trade 69 42 Wholesale Trade 0 69 52 Finance and Insurance 23 42 H 60-67 Finance, Insurance, and

Real Estate 53 53 Real Estate and Rental and Leasing 15 24 44-45 Retail Trade 17 72

G 52-59 Retail Trade 64

72 Accommodation and Food Services 10 15

51 Information 20 34

54 Professional, Scientific, and Technical Services 28 47

56 Administrative and Support and Waste Management and Remediation Services

29 43

61 Educational Services 12 17 62 Health Care and Social Assistance 27 39 71 Arts, Entertainment, and Recreation 19 25

I 70-89 Services 150

81 Other Services (except Public Administration) 19 49

[No SIC counterpart; classified in the divisions of establishments managed.]

– 55 Management of Companies and Enterprises 1 3

J 91-97 Public Administration 27 92 Public Administration 2 29 K 99 Unclassifiable – [No NAICS counterpart.] – –

Total 1,004 Total 358 1,170

Page 4: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

4

Characteristic SIC 1987 NAICS 1997

Classification based on a consistent economic concept

NoAd hoc mix of production-

and market-oriented industries

YesProduction processes

Codes

• Alphabetic divisions, A–J

• 2-digit major groups• 3-digit industry groups• 4-digit industries• 6-digit processing

codes

• 2-digit sectors• 3-digit subsectors• 4-digit industry groups• 5-digit NAICS

industries• 6-digit National

industries• 8-digit processing

codes

Structure 10 Divisions

20 Sectors(Disaggregation and

reorganization for services-producing

industries)

Industries 1,004 total

• 1,170 U.S. industries, total

• Net gain of 166 industries

• 358 industries defined separately for the first time (70% in services)

• 300 industries revised substantially

SIC

vers

us N

AIC

S

Page 5: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

5

SIC versus NAICS

SIC/NAICS Relationships 614 SIC industries each go to a single NAICS industry 390 SIC industries consist of 2 subindustry parts or

more that go to different NAICS industries(1,357 such subindustry parts, total)

80 SIC industries consist of 2 subindustry parts or more that go to NAICS industries in different sectors

401 NAICS industries consist of 2 subindustry parts or more that come from different SIC industries

Overall, relationships may be 1:1, 1:many, many:1, or many:many

Page 6: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

6

NAICS Implementation

The Census Bureau’s Goals Complete transition for the 1997 reference

period Make implementation as complete and

accurate as possible Keep additional cost to a minimum Keep additional response burden to a minimum Provide statistical presentations that ease the

transition for data users

Page 7: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

7

NAICS Implementation

The Census Bureau’s Strategies Use 1997 Economic Census as the

implementation vehicle• Our most comprehensive, broad-based economic

collection• Tightly integrated with the Business Register, which

the census updates• Rich variety of specialized content for assigning

industry codes• Extensive, detailed industry statistics that are optimal

for introducing the new classification system

Page 8: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

8

The Census Bureau’s Strategies—Continued Classify establishments according to both

NAICS and SIC for the 1997 reference period• Requirement for presenting 1997 data on both bases• Additional subindustry detail

– NAICS components of SIC industries– SIC components of NAICS industries

• Bridge SIC coding scheme– SIC-based root (4 digits)– Suffix (2 digits) captures additional subindustry detail– Meets requirement for continued support of SIC-based

surveys through 2001– Fits existing data structures and processing systems– Translates to a distinct NAICS industry and a distinct SIC

industry

NAICS Implementation

1,971 industry/subindustry classes, total

Page 9: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

9Bri

dg

e S

IC I

llu

str

ati

on

OLD Standard Industrial Classification—1987

NEW North American Industry Classification System—

1997 Code Title

Bridge SIC

Code Code Title

5413 Architectural, Engineering, and Related Services

8712 Architectural Services 871200 541310 Architectural Services

0781 (Part) Landscape Counseling and Planning (except Horticultural Consulting)

078120

8748 (Part) Business Consulting Services, NEC (Urban Planners and Industrial Development Organizations)

874820

541320 Landscape Architectural Services

8711 Engineering Services 871100 541330 Engineering Services

7389 (Part) Business Services, NEC (Drafting Services) 738913 541340 Drafting Services

7389 (Part) Business Services, NEC (Home and Building Inspection Services) 738912 541350 Building Inspection Services

8713 (Part) Surveying Services (Geophysical Surveying) 871320

1081 (Part) Metal Mining Services (Geophysical Surveying and Mapping)

108120

1382 (Part) Oil and Gas Field Exploration Services (Geophysical Surveying and Mapping)

138220

1481 (Part) Nonmetallic Minerals Services, Except Fuels (Geophysical Surveying and Mapping)

148120

541360 Geophysical Surveying and Mapping Services

7389 (Part) Business Services, NEC (Map Making Services) 738909

8713 (Part) Surveying Services (except Geophysical Surveying) 871310

541370 Surveying and Mapping (except Geophysical) Services

8734 (Part) Testing Laboratories (except Veterinary Testing Laboratories) 873410 541380 Testing Laboratories

Page 10: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

10

Administrative Data

Important source of data, including industry codes• Business Register maintenance• Economic census observations—i.e., for smaller single

units excluded from direct collection

NAICS implementation by suppliers• Internal Revenue Service: 1998 (available early 1999)• Social Security Administration: 1999• Bureau of Labor Statistics: 1997 – 2000• Methods varied

Role in Census Bureau’s NAICS implementation• Minor for 1997—too late (nonemployers an exception)• Important for Business Register maintenance—1998 and

later

Page 11: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

11

1996 Refiling Collections

Targeted establishments in 80 SIC Industries that had parts going to different NAICS SectorsExample: Metal Mining Services- Part to NAICS Sector 21—Mining- Part to NAICS Sector 54—Professional, Scientific, and Technical Services

Important preparation for economic census collection• 454 questionnaires tailored to fairly narrow industry groups• Content varied markedly from sector to sector

OLD Standard Industrial Classification—1987

NEW North American Industry Classification

System—1997 Code Title

Bridge SIC

Code Code Title

1081 Metal Mining Services

1081 (Part) Metal Mining Services (Except Geophysical Surveying Services) 108110 213114 Support Activities for Metal Mining

1081 (Part) Metal Mining Services (Geophysical Surveying Services) 108120 541360 Geophysical Surveying and Mapping

Services

Page 12: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

12

1996 Refiling Collections

Single Units Special survey

• 215,829 single units• December 1996 mailing• March 1997 follow-up

Questionnaire• 1 page• 17 versions for groups of

related industries• Listed bridge SIC

industry/ subindustry descriptions

Respondent asked to select one description that best fit the business’ primary activity

Page 13: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

13

1996 Refiling Collections

Multiunits Done by means of the 1996 Company Organization Survey

(COS) Sample modified to include companies with establishments in

80 industries targeted for refiling• 50,698 enterprises with 1.1 million establishments, total• 67,973 establishments targeted for refiling

Presented inventory of establishments and requested updates to industry code and other items for each one

21

01-2345678 Enter codefrom Insert K

Any Mining Services Company

XYZ Enterprises

123

456 Any Street

Any City AB 43210

9876543210 00 108100 10 40 000

Page 14: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

14

1996 Refiling Collections

Multiunits—Continued Inventory list directed

respondent to an industry coding insert• 11 versions for groups of

related industries• Listed bridge SIC industry/

subindustry descriptions Respondent asked to select

one description that best fit the business’ primary activity and recorded the code on the inventory list

Page 15: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

15

1997 Economic Census Collections

Standard Forms Sent to

• All multiunits• Larger single units• A sample of smaller single units

454 versions tailored to groups of related industries Content

• Basic economic measures: sales/receipts/revenue, payroll, employment, and other items that varied by sector

• Specialized items important for assigning bridge SIC codes—e.g.,– Detail of sales/receipts/revenue (by class)– Kind of business (bridge SIC industry/subindustry descriptions)– Materials consumed (by class)– Selling characteristics– Employees by occupation

3.4 million establishments

Page 16: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

16

1997 Economic Census Collections

Sector-Specific Classification Forms Sent to smaller, partially classified single units—1.2

million 28 versions tailored to groups of related industries Content

• Kind of business– Bridge SIC industry/subindustry descriptions with check boxes– Respondent asked to pick the one description that best fit the

business’ primary activity

• Some versions had other classification inquiries—e.g.,– Method of selling– Selling characteristics

Retail trade and wholesale trade

Page 17: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

17

1997 EconomicCensus Collections

General Classification Form Sent to unclassified single units

of any size—0.4 million Content

• Business or activity inquiry– Bridge SIC industry/subindustry

descriptions with check boxes– Respondent asked to pick the

one description that best fit the business’ primary activity

• Sources of sales, receipts, or revenue

• Materials consumed (manufacturers)

• Class of customer Once classified, larger single

units were sent a standard form as described earlier

Page 18: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

18

1997 Economic Census Classification Methods

Industry coding edits• Applied complex rules• Considered a variety of classification data from standard forms

Clerical coding• Used industry coding procedures and reference material• Primary uses:

– Classification form responses that required interpretation– Referrals from industry coding edits– Industry descriptions from nonemployers’ tax data

Analytical coding• Used industry expertise and reference material• May have contacted respondent for clarification• Primary uses:

– Difficult cases referred by clerical coding units– Classification errors identified by macro-analytical review of data

products

Page 19: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

19

1997 Economic Census Classification Methods—Continued

NAICS classifications from 1998 tax data—nonemployers only

Statistical modeling procedure• Parameters

– Based on subpopulation of establishments with a complete 6-digitbridge SIC

– Determined relative distribution of units from each 2-, 3-, 4-, or 6-digit bridge SIC to the corresponding set of 6-digit bridge SICs

• Method– Assumed similar bridge SIC distributions for partially and

completely classified establishments– Preserved the distributions observed for completely classified

establishments– Used 3 uniformly distributed digits from the establishment’s EIN as

a distribution index– Assigned a 6-digit bridge SIC code to each establishment by

referring its SIC and distribution index to the distribution parameters

Page 20: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

20

1997 Economic Census Classification Methods—Continued

Statistical Modeling Procedure Illustrations Parameters for Bridge SIC Modeling Procedure

Distribution within SIC Code (In)

SIC Code (In)

Cumulative Distribution Parameter (> Index )

Bridge SIC Code

(Out) Percent Cumulative Percent

Corresponding NAICS Code

870000 178 871100 17.8% 17.8% 54133

870000 248 871200 7.0% 24.8% 54131

Example 1 870000 249 871310 0.1% 24.9% 54136

870000 278 871320 2.9% 27.8% 54137

870000 1000 Etc. 72.2% 100.0% Etc.

871000 639 871100 63.9% 63.9% 54133

Example 2 871000 890 871200 25.1% 89.0% 54131

871000 894 871310 0.4% 89.4% 54136

871000 1000 871320 10.6% 100.0% 54137

871100 1000 871100 100.0% 100.0% 54133

Procedure

1. Identify the first row that matches on SIC code (in)

2. Compare the establishment’s 3-digit index (based on EIN) to the cumulative distribution parameter

• If INDEX < PARAMETER, assign corresponding bridge SIC code (out)

• Else, advance to next row and repeat Step 2

Illustrations of the Bridge SIC Modeling Procedure

Example ID

EIN (Index based on positions 7 – 9)

SIC Code (In)

Bridge SIC Code (Out)

1 0123456248 123456248 870000 871310

2 0123456789 123456789 871000 871200

Page 21: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

21

1997 Economic CensusData Products

Note: Data are shown for selected industries used in the paper’s illustrations.

Comparative Statistics for the United States on a 1987 SIC Basis—1997 [Includes only establishments with payroll. Data are in current dollars and have not been adjusted for inflation.]

N – Comparable data not available

Establishments Receipts ($1,000)

Paid employees Annual payroll

($1,000) SIC 1987 SIC Description

1997 1992 % chg 1997 1992 % chg 1997 1992 % chg 1997 1992 % chg

1081 Metal mining services 203 251 -19.1 341,888 350,441 -2.4 3,066 2,973 3.1 110,070 104,612 5.2

1382 Oil and gas exploration services

1,197 1,473 -18.7 818,607 964,629 -15.1 7,039 12,930 -45.6 215,996 423,687 -49.0

1481 Nonmetallic minerals services, except fuels

172 N N 190,942 188,932 1.1 1,874 N N 63,551 N N

7389 Business services, not elsewhere classified

69,376 52,375 32.5 62,276,780 32,885,901 89.4 867,462 523,650 65.7 17,597,943 9,783,317 79.9

8711 Engineering services 52,526 41,834 25.6 88,180,688 65,245,236 35.2 730,048 657,609 11.0 35,337,890 27,246,839 29.7

8712 Architectural services 20,602 17,875 15.3 16,988,338 11,244,379 51.1 146,702 121,675 20.6 6,468,524 4,408,064 46.7

8713 Surveying services 9,025 8,418 7.2 3,453,489 2,280,177 51.5 56,880 45,324 25.5 1,712,316 1,089,694 57.1

8734 Testing Laboratories 5,488 4,540 20.9 6,442,964 4,763,614 35.3 82,024 70,462 16.4 2,708,782 1,998,829 35.5

8748 Business consulting services, not elsewhere classified

17,853 12,628 41.4 8,687,728 4,573,223 90.0 77,341 52,456 47.4 3,191,884 1,766,156 80.7

Page 22: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

22

1997 Economic CensusData Products

NAICS Industry

SIC 1987 Components

NAICS-Based Industry Statistics for the United States with Distribution Among 1987 SIC-Based Industries—1997 [Includes only establishments with payroll.]

NAICS SIC Pt Description Estab-

lish- ments

Receipts ($1,000)

Paid employees

Annual payroll

($1,000)

541330 Engineering services 52,526 88,180,688 730,048 35,337,890

8711 Engineering services 52,526 88,180,688 730,048 35,337,890

541360 Geophysical surveying & mapping services

587 1,087,786 9,905 445,595

1% of 1081 20 Geophysical surveying services only for metal mining, contract basis

21 3,783 41 1,101

63% of 1382 20 Geophysical surveying services for oil & gas fields, contract basis

213 518,667 2,907 104,681

4% of 1481 20 Geophysical surveying services for nonmetallic minerals (excluding fuels)

17 8,313 62 2,877

16% of 8713 20 Geophysical surveying 336 557,023 6,895 336,936

Pt – Part

The symbol is used as a link to the 1992 figures shown in Comparative Statistics. Note that there are links only for SIC industries, not for NAICS industries.

Comparable NAICS/SIC

Note: Data are shown for selected industries used in the paper’s illustrations.

Page 23: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

23

1997 Economic CensusData Products

1987 SIC-Based Industry Statistics for the United States with Distribution Among NAICS-Based Industries—1997 [Includes only establishments with payroll.]

SIC NAICS Pt Description Estab-

lish- ments

Receipts ($1,000)

Paid employees

Annual payroll

($1,000)

8712 Architectural services 20,602 16,988,338 146,702 6,468,524

541310 Architectural services 20,602 16,988,338 146,702 6,468,524

8713 Surveying services 9,025 3,453,489 56,880 1,712,316

51% of 541360 10 Geophysical surveying & mapping services (pt)

336 557,023 6,895 336,936

95% of 541370 10 Surveying services 8,689 2,896,466 49,985 1,375,380

Pt – Part

Comparable SIC/NAICS

SIC 1987IndustryNAICS Components

Note: Data are shown for selected industries used in the paper’s illustrations.

Page 24: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

24

Time Series Considerations

The adoption of NAICS causes a disruption in industry level time series data used by policymakers and researchers

Also affects the confidential micro data sets used at Census and its Research Data Centers

Page 25: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

25

Time series (cont.)

Example of the problem

Table 5. Sample Bridge Between SIC 3578 and NAICS

NAICS Description

Value of Shipments

($1,000) Share of

Shipments Paid

Employees Share of

Employment

SIC 3578 – Calculating & Accounting Machines, Except Electronic Computers

2,014,806 100% 7,683 100%

333313 Office Machinery Manufacturing 144,380 7% 966 13%

334119 Other Computer Peripheral Equipment Manufacturing

1,870,426 93% 6,717 87%

Source: Bayard and Klimek (2003)

Page 26: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

26

Time series (cont.)

Historical NAICS Classification Methodology• Bayard and Klimek (2003) develop a method to

recursively assign NAICS codes to historical Census of Manufactures data

• 4 ways to assign codes1. Product Code based assignment2. One to One SIC – NAICS mappings3. Longitudinal establishment linkages4. Statistical Model

Page 27: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

27

Time series (cont.)

Uses, status and next steps• Methodology has been used to develop NAICS based

time series data for Plant Capacity and Industrial Production

• Also used for Census Benchmarking exercise for Retail and Wholesale surveys

• Current plans to recode Longitudinal Business Database (i.e., all establishments in the Bureau’s business register going back to 1975)

Page 28: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

28

Lessons Learned

Minor discrepancies between final NAICS specification and census classifications Causes:

• NAICS refinements continued into early 1998– Detailed industry definitions– Determining industry placement for 35,000 specific activities

• December 1997 census mailout required questionnaire content had to be made final during the first half of 1997

Result: Some industries published on an “as collected” basis

Lessons:• Expect last-minute changes• As much as possible, allow industry classification system to

stabilize before implementing it• Be prepared to adapt to definitional uncertainty

Page 29: 18 th International Roundtable on Business Survey Frames Beijing, China 17 - 22 October 2004 Session 4 — Classification Issues Implementing a Major Revision

29

Lessons Learned

1996 multiunit refiling produced low item response rate for reclassification inquiry Causes:

• Use of traditional collection instrument with minimal change– Designed primarily for collecting company organization

information– Not optimal for collecting classification updates

• Classification insert did not work well Result: costly follow-up and supplementary

classification activity Lesson: A better design was needed

Data users will demand time series continuity