7
The Importance of Data Sharing to Consistent Macroeconomic Statistics

Dennis Fixler and J. Steven Landefeld

Bureau of Economic Analysis

The Bureau of Economic Analysis (BEA) has a unique position in the decentralized U.S. statistical system. BEA produces the national income and product accounts (NIPAs), a comprehensive and consistent double-entry set of accounts for the economy. BEA uses a myriad of data collected from public and private data sources to construct these accounts. In this role, BEA often confronts major inconsistencies in piecing these data together that are not evident from the perspective of the agencies collecting the individual pieces of the economic puzzle. BEA has been described as the canary in the mineshaft for the U.S. statistical system.

The U.S. statistical system has evolved over time in such policy agencies as the U.S. Department of Commerce, Labor, the Treasury, Agriculture, and Energy to provide data and answer questions relevant to the agencies’ missions. Surveys and the legislation supporting them have evolved independently. The result is a diverse set of data using different business registers, different industry classifications for establishments, different concepts and definitions, different timing, and different collection methods.

These differences in survey frames and procedures produce significant quantitative differences in what would appear to be the same measures of economic activity. For example, employment in individual industries as reported by the Labor Department’s Bureau of Labor Statistics (BLS) can differ markedly from that reported by the Department of Commerce’s Census Bureau. Differences exist for wages and salaries across industries, across states, and in the aggregate (see Tables 7-2 and 7-6).



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop 7 The Importance of Data Sharing to Consistent Macroeconomic Statistics Dennis Fixler and J. Steven Landefeld Bureau of Economic Analysis The Bureau of Economic Analysis (BEA) has a unique position in the decentralized U.S. statistical system. BEA produces the national income and product accounts (NIPAs), a comprehensive and consistent double-entry set of accounts for the economy. BEA uses a myriad of data collected from public and private data sources to construct these accounts. In this role, BEA often confronts major inconsistencies in piecing these data together that are not evident from the perspective of the agencies collecting the individual pieces of the economic puzzle. BEA has been described as the canary in the mineshaft for the U.S. statistical system. The U.S. statistical system has evolved over time in such policy agencies as the U.S. Department of Commerce, Labor, the Treasury, Agriculture, and Energy to provide data and answer questions relevant to the agencies’ missions. Surveys and the legislation supporting them have evolved independently. The result is a diverse set of data using different business registers, different industry classifications for establishments, different concepts and definitions, different timing, and different collection methods. These differences in survey frames and procedures produce significant quantitative differences in what would appear to be the same measures of economic activity. For example, employment in individual industries as reported by the Labor Department’s Bureau of Labor Statistics (BLS) can differ markedly from that reported by the Department of Commerce’s Census Bureau. Differences exist for wages and salaries across industries, across states, and in the aggregate (see Tables 7-2 and 7-6).

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop WHY IT MATTERS The implications of these differences in estimates is illustrated in Table 7-1, which summarizes the various data sources used by BEA in constructing one of its sets of accounts. Gross domestic product (GDP) is mainly estimated using data collected by the Census Bureau, while gross domestic income (GDI) is mainly estimated using data collected by BLS, the Census Bureau, and the Statistics of Income (SOI, part of the Internal Revenue Service, IRS). In concept, GDP should equal GDI because all final expenditures should end up as income to households, business, or government. However, because of the differences in the source data used in estimating GDP and GDI, often they are not equal, and the result is the statistical discrepancy. Such discrepancies between GDP and GDI can have large impacts on fiscal and monetary policy. During the latter half of the 1990s, a large and persistent discrepancy arose, with real GDI growing 0.6 percent faster than real GDP (1995-2000). This was important for budget planning because real trend GDP growth is used as the baseline for estimating near-term trend growth in 5-year budget forecasts made by the Office of Management and Budget (OMB) and the Congressional Budget Office. To illus- TABLE 7-1 BEA Summary Account 1—Primary Data Sources (billions of dollars)   Primary Data Source 2004 Income side     Labor compensation BLS $6,693.4 Corporate profits & gov’t enterprises Census Bureau, SOI 973.6 Proprietors’ income and rental income Census Bureau, SOI 1023.8 Interest on assets, taxes, & misc. payments SOI, FRB 1,531.3 Depreciation Census Bureau 1,435.3 GROSS DOMESTIC INCOME   $11,657.5 Statistical discrepancy   76.8 GROSS DOMESTIC PRODUCT   $11,734.3 Expenditure side     Personal consumption expenditures Census Bureau $8,214.3 Gross private domestic investment Census Bureau 1,928.1 Gov’t consumption exp. & gross invest. Gov’t, Census Bureau 2,215.9 Net exports of goods and services Census Bureau, BEA –624.0 GROSS DOMESTIC PRODUCT   $11,734.3

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop trate the impact, according to OMB’s FY 2006 analysis of the “sensitivity of the budget to economic assumptions,” a persistent understatement of real GDP growth by 1.0 percent would result in an overstatement of the projected deficit of $530 billion. Similarly, a persistent understatement of real trend GDP growth could lower the Federal Reserve’s estimate of non-inflationary sustainable growth and signal the need for a tighter monetary policy than necessary. One possible answer to the source of this discrepancy could lie in the recording of stock options, bonuses, and fringe benefits in employee compensation. While there are many sources of the difference between BLS and the Census Bureau payroll data, it is interesting that during the latter half of the 1990s, when stock options and bonuses were growing rapidly, the Census Bureau data rose at a 7.8 percent average annual rate, whereas the BLS data rose at a 7.5 percent average annual rate (1995-2000). Part of this may reflect the recording of stock options. For example, in Washington State—a state with significant stock option activity—the Census Bureau data grew nearly twice as fast (11.5 percent) as the BLS payroll data (6.2 percent) for 2000. If it turned out that stock options were under-reported in the BLS data, it would suggest that the growth rate of GDI might be even higher, thereby focusing additional effort on improving the reporting of final expenditures on services and other less-well-measured components of GDP. Another example of the importance of BEA accuracy is illustrated by its regional data, which are used in the geographic allocation of nearly $200 billion in federal funds. These data are also used by virtually every state for its tax and planning purposes. BEA uses BLS data for these state and local estimates, which are taken from quarterly employment and unemployment tax forms. The differences between the two sets of payroll data across states vary from the BLS set’s being 4.2 percent higher in New Mexico to 9.5 lower in Alaska than what is reported by the Census Bureau. These differences could have a significant impact on the allocation of state Medicaid funds, which uses BEA per capita state personal income to determine the federal share of payments for each state. Differences in growth rates can also have an important impact on state tax projections and spending plans. For example, in New York the $1.2 billion difference in growth in wages and salaries between 2001 and 2002 between BLS and the Census Bureau series would amount to about a $173 million difference in projected income taxes. These are but a few of the examples of the implications for government and business decision makers. In the sections below, the implications for users of estimates ranging from profits and productivity to inflation and offshoring are explored.

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop DATA SHARING Over the years, numerous proposals have been put forth to resolve the problems associated with a decentralized system. One has been the creation of a central statistical office such as those that exist in Canada, Australia, and other countries. The creation of such an entity has not proved popular for various reasons. Consolidation would require extensive budget negotiation and resources to coordinate and implement a process that ensures there is little to no disruption in data production. Furthermore, the current system allows for the specialization that has arguably led to many of the innovations produced by U.S. statistical agencies. Given these difficulties, a practical way to achieve many of the benefits of a central statistical agency without the costs is to permit the sharing of business data among the three general-purpose statistical agencies—BEA, BLS, and the Census Bureau—that produce the bulk of the nation’s economic data. All three agencies have an excellent record of protecting confidential data, have access to various types of tax data, and share various types of data that could be significantly improved by data sharing. A major step forward in allowing data sharing was the passage of the Confidential Information Protection and Statistical Efficiency Act of 2002 (CIPSEA). Section 521 stipulated that business data can be shared for statistical purposes among BEA, the Census Bureau, and BLS. At the time CIPSEA was formulated, it was understood that for data sharing to be completely operational, there would have to be some changes in Section 6103, paragraph (j), of Title 26 (Internal Revenue Code) and the accompanying regulations that govern access to federal tax information (FTI). These changes are necessitated by the facts that much of the Census Bureau information is commingled with FTI and neither BEA nor BLS has the Census Bureau level of access to use such data. Although there have been discussions concerning the formulation of a bill to submit to Congress to bring about the necessary changes in Title 26, to date no bill has been written for submission. The absence of fully implemented data sharing especially affects BEA because it collects few data of its own and relies primarily on the Census Bureau for its data. Data sharing, however, does not just affect the ability of BEA to access Census Bureau data; the inability of BLS and the Census Bureau to share data greatly affects the quality of the data that BEA receives from both agencies. In this chapter we provide examples of how the absence of data sharing affects BEA estimates. The limited access to business tax data has enormous effects on BEA’s ability to access Census data that are commingled with tax data. The Cen-

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop sus Bureau sample frames are constructed from IRS data, and, under current rules, name, address, and employer identification numbers are generally considered tax data. Although in principle BEA has access to corporate tax records in the SOI sample, the Census Bureau does not know the identification of those firms and so BEA has generally not been allowed access to Census records. Without going into the arcane detail, whether BEA has access to corporate Census records that are commingled with tax data is determined by the extent to which the Census Bureau claims that data are based on their own collection and not IRS records. Such a claim is generally made by the Census Bureau in the case of multiunit establishments. Thus BEA cannot access Census Bureau records from single-unit establishments. Finally, because legislation limits BEA access to corporate tax records, BEA cannot access partnership and sole proprietor Census Bureau records, which are collected from tax data–based sample frames. The limited access to tax data also impedes BEA’s use of the Census records to construct sample frames for its international surveys. The impediment is especially problematic in the services area, because many of these providers are not multiunit establishments. In a joint effort by BEA, Census, and the National Science Foundation (NSF) regarding identifying international research and development expenditures, it was discovered that there was considerable difference between Census and BEA sample frames. In this case, BEA had identified many firms that were not in the Census Bureau sample. Below we provide some detailed illustrations of how the absence of data sharing affects BEA estimates. We also discuss how the effect on BEA estimates would affect policy decisions that are based on those estimates. Industry Employment Differences BLS, the Census Bureau, and SOI are the main sources of wage and salary data in the U.S. economy. Figures 7-1 and 7-2 show that the levels and growth of total payroll according to these sources are broadly consistent, but that there are significant differences in magnitudes. Below we focus on the BLS and the Census Bureau data, as those are the two main sources used by BEA.1 BLS prepares comprehensive wage and salary data in its Quarterly 1 The SOI data are composed from a sample of tax returns and therefore are not as comprehensive as either BLS or the Census Bureau data. Furthermore these data are released with a lag.

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop FIGURE 7-1 Payroll data comparison—the Census Bureau, BLS, and SOI levels. FIGURE 7-2 Payroll data comparison—the Census Bureau, BLS, and SOI growth. Census of Employment and Wages Program (QCEW).2 These data are widely used in BEA and are the basis for the wage and salary component of personal income. The Census Bureau also prepares payroll data as part of its Quinquennial Economic Census and Annual Survey programs. These Census data are considered to be less timely than BLS data, but in some areas, such as educational services, membership organizations, and nonprofits, they are considered to be more complete than the QCEW data. 2 These data are commonly referred to as the ES-202 data, the former name of the program.

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop In addition, Census data on wages are generally recognized as providing a better industry distribution of aggregate wages, and incorporating these data into BEA estimates offers a unique opportunity to create greater consistency among the BEA industry accounts’ measures of gross output, intermediate inputs, and value added. The decision, however, to choose one set of data over the other has implications for the measurement of value added in the industry accounts, which can be traced out through examining the estimates prepared as part of the annual industry accounts. For some industries, the differences in the level of employment are significant. Table 7-2 identifies the differences in levels between BLS and the Census wage and salary data for 2002, an economic census year.3 The primary explanations for the differences are that the Census Bureau and BLS have different sample frames of establishments and that establishments in both frames are not always classified in the same industry. The Census Bureau and BLS are currently engaged in a project that seeks to study this source of difference and explore other sources of differences. Before this project could be undertaken, approval from the IRS had to be obtained. The presentation at the workshop by James Spletzer (BLS) and Paul Hanczaryk (Census Bureau) provided details of the study (see Chapter 2 of this volume). As shown in Table 7-2, there are many relatively large differences among industries in which estimates are available from both BLS and the Census Bureau. In the case of oil and gas extraction, the Census payroll estimate is about 50 percent lower than the BLS estimate. In addition, the Census Bureau estimate for all of manufacturing is about 15 percent—or roughly $100 billion—lower than the BLS estimate. In contrast, Census payroll estimate for management of companies and enterprises is about 63 percent—or over $70 billion—higher than the BLS estimate.4 Because employment and wage data are used in several places in the national accounts, we will now show how BEA estimates would be different if the Census data were used instead of the currently used BLS data for manufacturing and a few other industries in the computation of value-added. Although the current-dollar growth rate could change by as much as 2.0 percentage points (e.g., computers and electronic products), Table 7-3 shows the relative rankings for the selected industries tended to be relatively stable. 3 The Census payroll data used are from the U.S. Census Bureau web site as of April 1, 2005. BLS wage data are consistent with the 2004 annual revision to the national income and product accounts and the 2004 annual revision to the annual industry accounts. 4 This pattern may suggest a different classification treatment of head company offices by the Census Bureau and BLS.

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop TABLE 7-2 Differences in BLS Wages and the Census Bureau Payroll by NAICS Industry, 2002 (in thousands of dollars) 1997 NAICS Codes Industry Name BEA Wagesa   All Industries 4,968,131,000   Private industries 4,119,730,000 11 Agriculture, forestry, fishing, and hunting 31,815,000 111, 112 Crop and animal production (“Farms”) 17,685,000 113, 114, 115 Forestry, fishing, and related activities 14,130,000 21 Mining 30,788,000 211 Oil and gas extraction 11,455,000 212 Mining, except oil and gas 10,470,000 213 Support activities for mining 8,863,000 22 Utilities 40,094,000 23 Construction 272,418,000 31, 32, 33 Manufacturing 675,523,000 33, 321, 327 Durable goods 441,182,000 321 Wood product manufacturing 17,585,000 327 Nonmetallic mineral product manufacturing 20,674,000 331 Primary metal manufacturing 23,209,000 332 Fabricated metal product manufacturing 59,742,000 333 Machinery manufacturing 57,050,000 334 Computer and electronic product manufacturing 98,359,000 335 Electrical equipment and appliance manufacturing 20,630,000 3361, 3362, 3363 Motor vehicle, body, trailer, and parts manufacturing 58,705,000 3364, 3365, 3366, 3369 Other transportation equipment manufacturing 38,954,000 337 Furniture and related product manufacturing 18,232,000 339 Miscellaneous manufacturing 28,042,000 31, 32 (excluding 321 and 327) Nondurable goods 234,341,000 311 ,312 Food product manufacturing 60,356,000 313, 314 Textile and textile product mills 14,525,000 315,316 Apparel manufacturing 10,751,000 322 Paper manufacturing 25,611,000 323 Printing and related support activities 27,061,000 324 Petroleum and coal products manufacturing 7,632,000 325 Chemical manufacturing 57,293,000 326 Plastics and rubber products manufacturing 31,112,000 42 Wholesale trade 280,745,000 44, 45 Retail trade 360,341,000 48, 49 Transportation and warehousing, excluding postal service 162,206,000 481 Air transportation 30,550,000 482 Rail transportation 11,824,000 483 Water transportation 2,888,000 484 Truck transportation 47,917,000

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop     Percent Differenced   BLS Wagesb Census Payrollc BEA and Census BLS and Census BLS and Census Differencee     — —   3,923,090,541   — —   24,146,183   — —   15,862,753   — —   8,283,429   — —   30,557,227   — —   11,269,829 5,564,811 –51.4 –50.6 –5,705,018 10,321,353 8,987,397 –14.2 –12.9 –1,333,956 8,966,044 6,707,242 –24.3 –25.2 –2,258,802 39,895,551 43,493,804 8.5 9.0 3,598,253 260,841,814 254,000,182 –6.8 –2.6   670,676,772 573,401,510 –15.1 –14.5 –97,275,262 437,547,486 370,407,941 –16.0 –15.3 –67,139,545 16,952,331 15,909,908 –9.5 –6.1 –1,042,423 20,543,618 17,933,376 –13.3 –12.7 –2,610,242 23,246,080 21,508,667 –7.3 –7.5 –1,737,413 59,352,280 57,361,374 –4.0 –3.4 –1,990,906 56,689,509 49,470,768 –13.3 –12.7 –7,218,741 98,045,569 64,314,150 –34.6 –34.4 –33,731,419 20,479,516 17,957,015 –13.0 –12.3 –2,522,501 58,579,129 50,331,680 –14.3 –14.1 –8,247,449 38,446,534 31,231,174 –19.8 –18.8 –7,215,360 18,107,133 17,364,837 –4.8 –4.1 –742,296 27,105,787 27,024,992 –3.6 –0.3 –80,795 233,129,286 202,993,569 –13.4 –12.9 –30,135,717 59,649,421 52,334,562 –13.3 –12.3 –7,314,859 14,501,506 12,333,814 –15.1 –14.9 –2,167,692 10,360,588 8,567,969 –20.3 –17.3 –1,792,619 25,744,232 21,336,257 –16.7 –17.1 –4,407,975 26,457,610 25,738,613 –4.9 –2.7 –718,997 7,891,082 6,202,508 –18.7 –21.4 –1,688,574 57,322,150 44,032,801 –23.1 –23.2 –13,289,349 31,202,697 32,447,045 4.3 4.0 1,244,348 276,607,852 249,986,560 –11.0 –9.6 –26,621,292 348,909,029 296,215,722 –17.8 –15.1 –52,693,307 146,810,674   — — –146,810,674 30,180,386   — — –30,180,386 10,869   — — –10,869 2,793,556 3,031,880 5.0 8.5 238,324 46,824,531 47,833,730 –0.2 2.2 1,009,199

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop 1997 NAICS Codes Industry Name BEA Wagesa 485 Transit and ground passenger transportation 8,996,000 486 Pipeline transportation 3,272,000 487, 488, 492 Other transportation and support activities 39,802,000 493 Warehousing and storage 16,957,000 51 Information 189,736,000 511 Publishing including software 58,394,000 512 Motion picture and sound recording industries 18,258,000 513 Broadcasting and telecommunications 84,838,000 514 Information and data processing services 28,246,000 52 Finance and insurance 370,088,000 521, 522 Federal Reserve banks, credit intermediation and related services 132,010,000 523 Securities, commodity contracts, investments 112,344,000 524 Insurance carriers and related activities 119,830,000 525 Funds, trusts, and other financial vehicles 5,904,000 53 Real estate, rental, and leasing 71,785,000 531 Real estate 51,015,000 532,533 Rental and leasing services and lessors of intangible assets 20,770,000 54 Professional and technical services 415,422,000 5411 Legal services 80,297,000 5415 Computer systems design and related services 84,251,000 5412-5414, 5416-5419 Other professional, scientific and technical services 250,874,000 55 Management of companies and enterprises 117,147,000 56 Administrative and waste services 193,525,000 561 Administrative and support services 180,230,000 562 Waste management and remediation services 13,295,000 61 Educational services 74,446,000 62 Health care and social assistance 472,214,000 621 Ambulatory health care services 209,724,000 622, 623 Hospitals and nursing and residential care facilities 217,119,000 624 Social assistance 45,371,000 71 Arts, entertainment, and recreation 51,526,000 711, 712 Performing arts, museums, and related activities 24,724,000 713 Amusements, gambling, and recreation 26,802,000 72 Accommodation and food services 153,922,000 721 Accommodation 40,764,000 722 Food services and drinking places 113,158,000

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop     Percent Differenced BLS and Census Differencee BLS Wagesb Census Payrollc BEA and Census BLS and Census 7,787,298 7,575,497 –15.8 –2.7 –211,801 3,277,932 3,082,558 –5.8 –6.0 –195,374 39,101,445 34,489,516 –13.3 –11.8 –4,611,929 16,834,658 18,689,122 10.2 11.0 1,854,464 188,758,526   — — –188,758,526 58,307,089 64,712,028 10.8 11.0 6,404,939 17,879,785 12,516,040 –31.4 –30.0 –5,363,745 84,664,461 88,624,463 4.5 4.7 3,960,002 27,907,191 27,686,444 –2.0 –0.8 –220,747 356,371,058   — — –356,371,058 131,188,066 124,076,870 –6.0 –5.4 –7,111,196 108,325,327 101,285,387 –9.8 –6.5 –7,039,940 110,965,984 120,683,183 0.7 8.8 9,717,199 5,891,681   — — –5,891,681 68,801,129   — — –68,801,129 48,110,832 41,911,444 –17.8 –12.9 –6,199,388 20,690,296 18,706,319 –9.9 –9.6 –1,983,977 390,450,138   — — –390,450,138 69,875,728 69,939,404 –12.9 0.1 63,676 83,897,952 72,168,495 –14.3 –14.0 –11,729,457 236,676,458   — — –236,676,458 117,462,176 190,807,531 62.9 62.4 73,345,355 191,825,310   — — –191,825,310 178,563,429 195,425,035 8.4 9.4 16,861,606 13,261,881 12,178,484 –8.4 –8.2 –1,083,397 64,700,545   — — –64,700,545 456,030,369   — — –456,030,369 204,320,753 203,716,200 –2.9 –0.3 –604,553 215,390,850 212,480,514 –2.1 –1.4 –2,910,336 36,318,766 36,090,970 –20.5 –0.6 –227,796 47,050,671   — — –47,050,671 24,652,961 24,057,801 –2.7 –2.4 –595,160 22,397,710 21,069,716 –21.4 –5.9 –1,327,994 142,208,429   — — –142,208,429 36,805,629 34,874,261 –14.4 –5.2 –1,931,368 105,402,801 92,632,794 –18.1 –12.1 –12,770,007

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop NAICS 2000 Code Industry Code Description CBP QCEW $ dif. % dif.   Total 3,879.4 3,889.0 9.6 0.2 11 Forestry, fishing, hunting, and agric. support 4.7 23.2 18.5 395.7 21 Mining 22.1 29.7 7.6 34.3 22 Utilities 40.7 37.9 –2.7 –6.7 23 Construction 239.9 245.8 5.9 2.4 31 Manufacturing 644.0 743.8 99.8 15.5 42 Wholesale trade 270.1 276.8 6.7 2.5 44 Retail trade 302.6 335.8 33.2 11.0 48 Transportation and warehousing 125.6 147.2 21.6 17.2 51 Information 209.4 212.3 2.9 1.4 52 Finance and insurance 346.8 333.8 –13.0 –3.7 53 Real estate, rental, and leasing 59.2 64.4 5.2 8.8 54 Professional, scientific, andtechnical services 362.0 395.4 33.4 9.2 55 Management of companies and enterprises 211.4 124.0 –87.4 –41.4 56 Admin, support, waste management, remediation services 210.3 185.4 –24.9 –11.8 61 Educational services 61.9 55.6 –6.4 –10.3 62 Health care and social assistance 431.4 394.7 –36.8 –8.5 71 Arts, entertainment and recreation 43.2 45.1 1.9 4.4 72 Accommodation and food services 125.6 134.2 8.6 6.9 81 Other services (except public administration) Mean (excluding Total and 11) 109.9 92.7 –17.2 –15.6 0.9 NOTE: BLS QCEW data prior to 2001 have been backcasted to NAICS 2002 using NAICS reports from employers in the first quarter of 2001. Data for 2001 and 2002 also use NAICS 2002. CBP data are based on NAICS 1997. data. This project was conducted under the authority of the International Investment and Trade Act and CIPSEA. No Title 26 data were used in the linking operation or subsequent tabulations or reports for this study; neither BEA nor the Census Bureau data sets used for the project contained such data, as all original FTI were replaced by respondent data for the Census Bureau surveys being linked. The Census Bureau informed the IRS of the project to alleviate any questions or concerns the IRS might have. The project demonstrated that it is feasible to link the Census Bureau and BEA survey data, and that by linking the data an integrated data set on the domestic and international dimensions of R&D can be created. Table 7-10 compares the NSF data with the BEA data for U.S. parent companies and therefore examines only a subset of U.S. firms—U.S. affiliates of foreign companies were not included. Despite the smaller BEA uni-

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop 2001 2002 CBP QCEW $ dif. % dif. CBP QCEW $ dif. % dif. 3,989.1 3,952.2 –36.9 –0.9 3,943.2 3,930.8 –12.4 –0.3 4.8 23.6 18.8 393.2 5.0 24.1 19.2 385.0 25.0 31.9 6.9 27.7 24.0 30.6 6.6 27.5 41.9 39.3 –2.6 –6.2 41.8 39.9 –1.9 –4.7 247.2 260.2 13.0 5.2 247.3 260.8 13.5 5.5 617.7 704.1 86.4 14.0 580.4 670.7 90.3 15.6 275.9 279.6 3.7 1.3 262.5 276.6 14.1 5.4 314.8 344.1 29.3 9.3 320.7 348.9 28.2 8.8 129.5 149.8 20.3 15.6 127.3 146.9 19.6 15.4 207.1 205.8 –1.3 –0.7 188.1 188.8 0.7 0.4 373.6 359.4 –14.2 –3.8 372.7 356.4 –16.3 –4.4 64.0 66.9 2.9 4.6 65.2 68.8 3.6 5.5 374.4 403.7 29.3 7.8 368.8 390.5 21.7 5.9 213.1 118.4 –94.8 –44.5 204.8 117.5 –87.3 –42.6 221.4 189.1 –32.3 –14.6 212.2 191.8 –20.4 –9.6 67.1 60.4 –6.7 –10.0 72.0 64.7 –7.3 –10.1 465.7 425.5 –40.3 –8.6 499.2 456.0 –43.1 –8.6 46.1 45.2 –0.9 –2.0 47.7 47.1 –0.7 –1.4 128.6 138.1 9.5 7.4 131.1 142.2 11.1 8.5 115.2 97.7 –17.5 –15.2 –0.7 118.9 101.0 –17.9 –15.1 0.1 SOURCES: Census Web site (5/23/05): http://censtats. census.gov/cgi-bin/cbpnaic/cbpsel.pl; and BLS Web site (5/24/05): ftp://ftp.bls.gov/pub/special.requests/cew/. verse, the table shows that in some industries BEA data indicate a far higher level of R&D expenditures than that for all U.S. firms— pharmaceuticals and medicines, for example. At the micro level, there were 11 cases in which the BEA and Census data for total R&D spending for the matched U.S. parent companies differed by more than $500 million. There is a substantial difference in the collected data for manufacturing and nonmanufacturing. The substantially lower number for non-manufacturing may result from the fact that R&D expenditures for nonmanufacturing firms are relatively more difficult to define and identify; so this area is more likely to be affected by differences in treatment. The study also demonstrated some of the main benefits of data sharing—in the improvement of sample frames and the quality of reported data. For example, as a result of the project, the Census Bureau added over 500 companies to the sample for the Survey of Industrial Research

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop TABLE 7-8 Private Annual Payroll Data—Census Bureau (CBP) and BLS (QCEW): Growth     Growth Rate (%)     1999 NAICS CBP QCEW Diff. Ab. Diff. Code Industry Code Description   Total 7.4 7.7 0.3 0.3 11 Forestry, fishing, hunting, and agric. support 3.4 3.8 0.4 0.4 21 Mining –4.4 –4.7 –0.3 0.3 22 Utilities 3.3 2.8 –0.5 0.5 23 Construction 10.4 10.5 0.1 0.1 31 Manufacturing 3.0 3.5 0.5 0.5 42 Wholesale trade 6.9 7.2 0.3 0.3 44 Retail trade 8.3 6.8 –1.5 1.5 48 Transportation and warehousing 7.5 6.9 –0.5 0.5 51 Information 16.0 16.3 0.4 0.4 52 Finance and insurance 8.0 8.5 0.5 0.5 53 Real estate, rental, and leasing 8.4 6.5 –2.0 2.0 54 Professional, scientific, and technical services 12.1 17.3 5.2 5.2 55 Management of companies and enterprises 9.5 7.5 –2.0 2.0 56 Admin, support, waste management, remediation services 11.9 9.8 –2.0 2.0 61 Educational services 8.8 7.4 –1.4 1.4 62 Health care and social assistance 3.5 3.9 0.4 0.4 71 Arts, entertainment, and recreation 9.4 8.2 –1.1 1.1 72 Accommodation and food services 6.7 6.9 0.2 0.2 81 Other services (except public administration) 6.2 5.5 –0.7 0.7   Minimum     –2.0 0.1   Mean     –0.2 1.0   Maximum     5.2 5.2 and Development. For more information, see a report on the findings of the project—“Research and Development Link Project: Final Report” at www.bea.gov/bea/di/FinalReportpublic.pdf. How Data Sharing Could Help A large part of BEA’s job is adjusting the various data for differences in timing, concepts, and definitions. However, this is often difficult because, for the most part, BEA does not have access to the underlying

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop 2000 CBP QCEW Diff. Ab. Diff. 9.1 8.2 –0.9 0.9 –2.7 4.0 6.7 6.7 5.3 5.7 0.4 0.4 3.3 6.9 3.6 3.6 9.5 9.6 0.1 0.1 2.9 5.9 2.9 2.9 8.0 8.3 0.2 0.2 7.3 6.8 –0.5 0.5 7.6 6.9 –0.7 0.7 23.0 14.0 –9.0 9.0 10.7 10.2 –0.5 0.5 9.5 8.4 –1.1 1.1 16.3 9.1 –7.2 7.2 9.9 9.4 –0.4 0.4 14.8 11.0 –3.9 3.9 8.9 8.5 –0.3 0.3 5.4 6.1 0.6 0.6 9.6 10.1 0.4 0.4 7.4 6.8 –0.6 0.6 7.8 6.7 –1.0 1.0     –9.0 0.1     –0.6 2.1     6.7 9.0 microdata. If armed with full data-sharing capability, BEA, BLS, and the Census Bureau could explore and resolve differences in the activities of major companies or in their classification to various industries and regions. The agencies could also compare data to investigate and resolve persistent differences, such as the reporting of bonuses and stock options, the capitalization of computer investment, the impact of differences in timing, and the differences in company practices with respect to the writing-down of inventories or to the treatment of pensions and other fringe benefits.

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop     Growth Rate (%) NAICS 2001 CBP QCEW Diff. Ab. Diff. Code Industry Code Description   Total 2.8 1.6 –1.2 1.2 11 Forestry, fishing, hunting, and agric. support 2.3 1.8 –0.5 0.5 21 Mining 13.2 7.7 –5.6 5.6 22 Utilities 3.2 3.7 0.5 0.5 23 Construction 3.0 5.9 2.8 2.8 31 Manufacturing –4.1 –5.3 –1.3 1.3 42 Wholesale trade 2.1 1.0 –1.1 1.1 44 Retail trade 4.0 2.5 –1.6 1.6 48 Transportation and warehousing 3.1 1.8 –1.3 1.3 -51 Information 1.1 –3.1 –2.0 2.0 52 Finance and insurance 7.7 7.7 –0.1 0.1 53 Real estate, rental, and leasing 8.1 3.9 –4.2 4.2 54 Professional, scientific, and technical services 3.4 2.1 –1.3 1.3 55 Management of companies and enterprises 0.8 –4.5 –5.4 5.4 56 Admin, support, waste management, remediation services 5.3 2.0 –3.3 3.3 61 Educational services 8.4 8.7 0.4 0.4 62 Health care and social assistance 7.9 7.8 –0.1 0.1 71 Arts, entertainment, and recreation 6.8 0.3 –6.5 6.5 72 Accommodation and food services 2.4 2.9 0.5 0.5 81 Other services (except public administration) 4.9 5.3 0.5 0.5   Minimum     –6.5 0.1   Mean     –1.5 2.0   Maximum     2.8 6.5 NOTE: BLS QCEW data prior to 2001 have been backcasted to NAICS 2002 using NAICS reports from employers in the first quarter of 2001. Data for 2001 and 2002 also use NAICS 2002. CBP data are based on NAICS 1997. The limited access also affects BEA’s ability to study observed anomalies in the Census Bureau data. The following are some examples of observations that BEA would like to study. There are substantial differences in the reported payrolls from the Census Bureau and BLS, by area. For example, in Washington state, between 1999 and 2000, the Census Bureau reports an 11.5 percent increase (more than $9 billion) while BLS reports a 6.2 percent increase ($5 billion). There are many possible reasons for the discrepancy, and data sharing

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop 2002 1998-2002 CBP QCEW Diff. Ab. Diff. CBP QCEW Diff. Ab. Diff. –1.2 –0.5 0.6 0.6 4.5 4.2 –0.3 0.3 3.9 2.2 –1.7 1.7 1.7 2.9 1.2 1.2 –4.2 –4.3 –0.1 0.1 2.2 0.9 –1.3 1.3 –0.2 1.4 1.7 1.7 2.4 3.7 1.3 1.3 0.0 0.3 0.2 0.2 5.6 6.5 0.8 0.8 –6.0 –4.7 1.3 1.3 –1.1 –0.3 0.8 0.8 –4.8 –1.1 3.8 3.8 2.9 3.8 0.8 0.8 1.9 1.4 –0.5 0.5 5.4 4.3 –1.0 1.0 –1.7 –1.9 –0.2 0.2 4.0 3.4 –0.7 0.7 –9.2 –8.3 0.9 0.9 6.4 4.2 –2.2 2.2 –0.2 –0.8 –0.6 0.6 6.5 6.3 –0.2 0.2 1.9 2.8 0.9 0.9 6.9 5.4 –1.6 1.6 –1.5 –3.3 –1.8 1.8 7.4 6.0 –1.3 1.3 –3.9 –0.8 3.2 3.2 3.9 2.7 –1.2 1.2 –4.2 1.4 5.6 5.6 6.7 6.0 –0.7 0.7 7.2 7.1 –0.1 0.1 8.3 7.9 –0.4 0.4 7.2 7.2 0.0 0.0 6.0 6.2 0.2 0.2 3.4 4.0 0.6 0.6 7.3 5.6 –1.7 1.7 1.9 3.0 1.1 1.1 4.6 4.9 0.3 0.3 3.2 3.4 0.2 0.2 5.5 5.2 –0.3 0.3     –1.8 0.0     –2.2 0.2     0.7 1.2     –0.4 0.9     5.6 5.6     1.3 2.2 SOURCES: Census Bureau web site (5/23/05): http://censtats. census.gov/cgi-bin/cbpnaic/cbpsel.pl and BLS Web site (5/24/05): ftp://ftp.bls.gov/pub/special.requests/cew/. with access to tax data would help get at the cause. For example, one source of the difference could be differences in the recording of stock options. By knowing the companies in the state, it would be possible to check with firm reports about reported stock options and thereby reconcile any difference. Such huge differences in the payroll numbers affect the estimation of GDI. BEA obtains monthly data from the Census Bureau for the manufacturing sector based on the M3 (Manufacturers’ Shipments, Inventories and Orders) survey. However, because participation in this survey is vol-

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop TABLE 7-9 Comparison of the Census Bureau and BLS Data for Foreign-Owned Establishments, 1992   Census Bureau BLS BLS Value as Percentage of Census Bureau Value   Number of Establishments Number of Employees Number of Reporting Units Number of Employees Number of Establishments Number of Employees All industries 102,958 4,944,157 106,041 4,747,637 103 96 Agricultural services, forestry, and fishinga 139 5,814 139 4,265 100 73 Mining   120,782 1,640 102,814 102 85 Construction 3,322 2,305 90,866 189 97   Manufacturing 81 2,004,947 13,076 1,930,135 102 96 Transportation and public utilitiesb 231,638 3,792 222,999 97 96   Wholesale trade 91 513,012 34,999 491,578 186 96 Retail trade 3,190 26,756 853,158 71 100   Finance, insurance, and real estate 401,018 9,558 360,287 83 90   Services 2,775 12,899 676,091 85 94   Otherc 9 1,116 569 3,137 n.m. n.m. NOTE: n.m. = not meaningful. aExcludes agricultural production of crops and livestock. bThe Census Bureau data exclude railroad transportation. cFor the Census Bureau: consists of private education and noncommerical establishments; for BLS: consists of nonclassifiable est ablishments. SOURCES: The Census Bureau data: Foreign Direct Investment in the United States: Establishment Data for 1992 available on BEA’s Web site at http://www.bea.gov/bea/ai1.htm#BEACENS. BLS data: BLS news release: “Employment and Wages in Foreign-Owned Businesses in the United States, Fourth Quarter 1992,” October 1996.

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop TABLE 7-10 Comparison of NSF R&D Expenditures by All U.S. Companies with BEA R&D Expenditures by U.S. Parent Companies, 2001 (millions of dollars or percentage)   NSF: All U.S. Companies BEA: U.S. Parent Companies Parent Companies as Percentage of all U.S. Companies All industries 198,505 143,017 72 Manufacturing 120,705 115,118 95 Food 1,819 914 50 Beverage and tobacco products 152 469 309 Textiles, apparel, and leather (D) 125 n.a. Wood products 182 (D) n.a. Paper, printing and support activities (D) (D) n.a. Petroleum and coal products (D) 1,002 n.a. Chemicals 17,892 31,927 178 Basic chemicals 1,876 1,742 93 Resin, synthetic rubber, fibers, and filament (D) 2,972 n.a. Pharmaceuticals and medicines 10,137 23,169 229 Other chemicals (D) 4,045 n.a. Plastics and rubber products (D) 929 n.a. Nonmetallic mineral products 990 339 34 Primary metals 485 484 100 Fabricated metal products 1,599 554 35 Machinery 6,404 8,561 134 Computer and electronic products 47,079 38,356 81 Computers and peripheral equipment (D) 7,727 n.a. Communications equipment 15,507 14,526 94 Semiconductor and other electronic components 14,358 11,114 77 Navigational, measuring, electromedical, and control instruments 12,947 4,158 32 Other computer and electronic products (D) 832 n.a. Electrical equipment, appliances, and components 4,980 2,008 40 Transportation equipment 25,965 25,147 97 Motor vehicles, trailers, and parts (D) 18,183 n.a. Other (D) 6,964 n.a. Furniture and related products 301 128 43 Miscellaneous manufacturing 6,606 2,570 39 Nonmanufacturing 77,799 27,899 36 Mining, extraction, and support activities (D) 411 n.a. Utilities 133 59 44

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop   NSF: All U.S. Companies BEA: U.S. Parent Companies Parent Companies as Percentage of all U.S. Companies Construction 320 (D) n.a. Trade 24,372 (D) n.a. Transportation and warehousing 1,848 12 1 Information (D) 9,514 n.a. Publishing 13,760 6,452 47 Newspaper, periodical, book, and database 649 (D) n.a. Software 13,111 (D) n.a. Broadcasting and telecommunications (D) 796 n.a. Telecommunications (D) 782 n.a. Other (D) 14 n.a. Other information (D) 2,266 n.a. Finance, insurance, and real estate (D) 624 n.a. Professional, scientific, and technical services 27,704 10,348 37 Architectural, engineering, and related services 3,386 18 1 Computer systems design and related services 9,154 8,929 98 Other 15,164 1,401 9 NOTES: (D) = suppressed to avoid disclosure of data of individual companies; n.a. = not available. SOURCES: R&D spending by all U.S. companies: Research and Development in Industry: 2001 available on the NSF web site at http://www.nsf.gov/statistics/nsf05305/htmstart.htm; R&D spending by U.S. parent companies: U.S. Direct Investment Abroad: Operations of U.S. Parent Companies and Their Foreign Affiliates, Revised 2001 Estimates available on the BEA’s web site at http://www.bea.gov/bea/ai/iidguide.htm#link12b. untary and some firms decide not to participate, BEA does not know the extent of participation. A recent example is the decision by a major producer of semiconductors to terminate its participation, which represented a huge erosion in the representativeness of the surveys. The Annual Survey of Manufacturers, however, is mandatory. Thus BEA must wait until the annual data are available before it can check the estimates based on the monthly data. If BEA had access to the M3 data, then it could identify the firms responsible for missing data and possibly estimate the missing information from publicly available sources such as company reports. Publicly available sales data from company reports could aid in the estimation of missing shipment data from a company that did not provide

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop such information in the M3 survey. The ability to estimate such missing information would serve to reduce revisions to GDP. Relatedly, the M3 surveys provide inventories data to BEA, and BEA does not always know whether a company has reported an inventory adjustment to the Census Bureau in the same way that the company has entered it on its financial accounts. A few years ago the press reported a major write-down of inventory by a major producer of information technology equipment. Without knowing how the firm entered the write-down in its Census Bureau report, BEA had no way of checking if the inventory adjustment was accurately reflected both on the product side of the national accounts (inventories is a component of investment) and on the income side (the valuation of inventories affects corporate profits). In fact, there was a large adjustment to inventories. How Data Sharing Would Help Cope with Disasters The massive destruction wrought by Hurricane Katrina is having a significant impact on the ability of the statistical agencies to collect economic data in the affected regions. As a result of the disappearance of sample units, estimates of retail trade, construction, employment and wages, and other components of the principal economic indicators will contain many imputations. Data sharing would allow a combining of data that would enable the statistical agencies to better impute missing values. For example, in the absence of complete business list reconciliation between the Census Bureau and BLS, data sharing would allow one of the agencies to find alternative establishments that might serve as proxies for missing establishments and thereby provide a straightforward imputation. In addition, the Census Bureau sales values might be used by BLS to impute prices for its price indexes. The ability to share data would also enable the statistical agencies to examine each other’s establishment-level imputations to see if they suit an agency’s needs. For example, the BEA regional program would have access to the BLS establishment imputations for QCEW to see if their needs are met. In short, data sharing would allow the statistical agencies to economize resources to efficiently handle disruptions to the usual production of economic statistics. REFERENCES Morisi, T. 2003 Recent changes in the National Current Employment Statistics Survey. Monthly Labor Review June. Moyer, B.C., M.A. Planting, M. Fahim-Nader, and S.K.S. Lum 2004 Preview of the comprehensive revision of the annual industry accounts. Survey of Current Business 84(March):38-51.

OCR for page 91
Improving Business Statistics Through Interagency Data Sharing: Summary of a Workshop National Science Foundation, Division of Science Resources Statistics 2005 Research and Development in Industry: 2001. (NSF #05-305.) Arlington, VA: National Science Foundation. Available: http://nsf.gov/statistics/nsf05305/front.htm [accessed May 2006]. Strassner, E.H., G.W. Medeiros, and G.M. Smith 2005 Annual industry accounts: Introducing KLEMS input estimates for 1997-2003. Survey of Current Business 85(September):34. U.S. Bureau of Economic Analysis, International Economic Accounts 2005a Foreign Direct Investment in the United States: Establishment Data for 1992. (Census data.) Available: http://www.bea.gov/bea/ai1.htm#BEACENS [accessed May 2006]. 2005b U.S. Direct Investment Abroad: Operations of U.S. Parent Companies and Their Foreign Affiliates, Revised 2001. Available: http://www.bea.gov/bea/ai/iidguide.htm#link12b [accessed May 2006]. U.S. Bureau of Economic Analysis and U.S. Census Bureau 1992 Foreign Direct Investment in the United States: Establishment Data for 1987. Washington, DC: U.S. Government Printing Office. U.S. Census Bureau 2003 County Business Patterns, NAICS. Available: http://censtats.census.gov/cgi-bin/cbpnaic/cbpsel.pl [accessed May 2006]. U.S. Census Bureau, Bureau of Economic Analysis, and National Science Foundation Division of Science Resources Statistics 2005 Statistics Research and Development Data Link Project: Final Report. Available: http://www.bea.gov/bea/di/FinalReportpublic.pdf [accessed May 2006]. U.S. Department of Labor, Bureau of Labor Statistics 1996 Employment and Wages in Foreign-Owned Businesses in the United States, Fourth Quarter 1992. (News release.) Washington, DC: U.S. Department of Labor, Bureau of Labor Statistics. 2002 Quarterly Census of Employment and Wages Program Data. Available: ftp://ftp.bls.gov/pub/special.requests/cew/ [accessed May 2006].