Data Synchronization: Leveraging Existing Business Data to Better Measure the Economy
This column is written to inform ASA members about what the ASA is doing to promote the inclusion of statistics in policymaking and the funding of statistics research. To suggest science policy topics for the ASA to address, contact ASA Director of Science Policy Steve Pierson at firstname.lastname@example.org.
This month’s guest columnist, Adrienne Pilot of the Council of Economic Advisers, describes a major development for the federal statistical agencies. A carefully crafted proposal to Congress would allow the economic statistical agencies to better synchronize their business data, thereby allowing these agencies to improve their economic statistical data while still protecting confidentiality. Contributing Editor
Adrienne Pilot is director of the statistical office at the Council of Economic Advisers. Previously, she held positions with the Office of Management and Budget’s Office of Statistical and Science Policy and the Bureau of Economic Analysis.
-Steve Pierson, ASA Director of Science Policy
An interagency data synchronization proposal developed by the Commerce, Labor, and Treasury departments would yield important improvements in accuracy, reliability, and comparability across the federal economic statistical system—and across national, regional, industry, and international data. If approved by Congress, the revenue-neutral proposal would broaden Bureau of Economic Analysis (BEA) access to Internal Revenue Service (IRS) business tax data and enable the Bureau of Labor Statistics (BLS) and U.S. Census Bureau to better synchronize their business lists while complying with all relevant privacy and confidentiality regulations.
Why Change Is Needed
Current tax law authorizes limited access to federal tax information (FTI) for statistical use. Although the Census Bureau is granted access to FTI for all businesses, BEA is permitted access only to corporate tax data, and BLS lacks access to any federal business tax data. Business tax data (such as company name and address) are used to construct the Census Bureau’s business list; many Census Bureau data products are thus considered to be “comingled” with tax information. Although the Confidential Information Protection and Statistical Efficiency Act of 2002 (CIPSEA) authorizes the Census Bureau, BEA, and BLS to share business information for statistical purposes, companion legislation to amend the IRS code will be necessary to facilitate sharing and synchronization of business tax data among these agencies.
The status quo poses two primary problems for the statistical system. First, growing numbers of U.S. businesses are adopting unincorporated legal forms, such as limited liability companies (LLCs) and partnerships, rather than the corporate legal form. With access restricted to tax data for corporations, BEA must impute increasingly more business income each year.
Second, the BLS and Census Bureau business lists—the foundations of federal business statistics and the sample frames for the majority of federal business surveys—are not in sync. A 2006 study comparing classifications of businesses by BLS and the Census Bureau found 33% of matched single-establishment firms had been assigned different North American Industrial Classification System (NAICS) industry codes on the two lists. Because the Census Bureau list is derived using tax information, the Census Bureau cannot share its list with BLS to reconcile these differences. As a result, classification and coverage inconsistencies persist, affecting major economic statistics, especially those (such as gross domestic product) that integrate data produced by different agencies.
Preserving confidentiality is essential for both FTI and statistical data. IRS Publication 1075 prescribes protocols to safeguard tax data—secure storage, restricted access, proper disposal, and reporting requirements. CIPSEA specifies standardized safeguards for data gathered for statistical purposes. Both regimes provide criminal and civil sanctions, including fines and imprisonment, for unauthorized disclosure or inspection of confidential information. BEA, BLS, and the Census Bureau routinely protect highly sensitive data, including tax data and the market-sensitive economic information these agencies produce.
Consistent with the intent of CIPSEA, the revenue-neutral interagency proposal would amend the Internal Revenue Code—26 USC 6103(j)—to:
Augment BEA’s current access to corporate business FTI data with access to information for sole proprietorships with receipts greater than $250,000 and for all partnerships.
Permit BLS to receive Census Bureau data for businesses (and tax-exempt entities), comingled with limited FTI, so these two agencies may synchronize their business lists. The Census Bureau would provide identifier variables (business name, address, taxpayer identification number, and NAICS industry code) for record matching and selected economic variables (such as total employment and business-level wages) to facilitate investigation of non-matches or questionable matches. (BLS would not have access to FTI for individuals.) The Census Bureau also would provide BLS with sales/revenue by product line data for improvement of industrial price indexes.
Permit state agencies (whose administrative records form the basis of the BLS business list) to receive limited identity variables (business name, address, taxpayer identification number, and industry only) for synchronizing the BLS and Census business lists.
Both IRS and CIPSEA safeguard regimes would apply to any tax information being shared. After researching discrepancies using third-party sources, BLS and the states would introduce only independently verified changes into their lists, so their business lists would never become comingled with shared federal tax information.
Better Data for Better Decisionmaking
This legislative proposal would improve the accuracy and comparability of major, widely used economic statistics. Federal economic statistics are used by private firms––the engines of economic growth––for planning and analysis. Policymakers depend on these measures to understand and guide the economy. Current data gaps and inconsistencies cloud this understanding. Federal agencies use data to target surveys; better data may improve efficiency in data collection. Examples of improvements in accuracy, reliability, and efficiency that may be possible across the federal economic statistical system follow.
BEA’s National Income and Product Accounts provide two measures of national economic activity: gross domestic product (GDP, primarily Census Bureau based) and gross domestic income (GDI, primarily BLS based). In theory, the income and product sides of the national accounts ledger should be equal, but they differ in practice due in part to coverage and classification differences in the BLS and Census Bureau source data.
Improving BEA’s measurement of GDP and GDI may result in more accurate forecasts of budget deficits. According to the Office of Management and Budget, consistent understatement of GDP forecasts by 0.5% per year would result in a $1.6 trillion overstatement of the projected cumulative budget deficit over a 10-year budget window. Although GDP exceeds GDI in some years and falls short in others, trends sometimes persist for years. For example, from 1995–2000, real GDP grew 0.5% slower than real GDI, on average, per year. Accurate data are important to the formulation of fiscal policies.
Reducing the discrepancy between GDP and GDI and refining BLS productivity measurement would support the Federal Reserve Board’s formulation of monetary policy. Real trend GDP growth is used to estimate noninflationary sustainable growth; consistent understatement could mistakenly imply the need for tighter monetary policy. Productivity also is considered to evaluate how fast the economy can expand without triggering inflation.
Reconciling differences in BLS and Census Bureau industry classifications would provide a clearer picture of industry-specific growth rates. For example, at the height of the economic expansion in 2007, BEA’s official measure of growth in the computer industry (calculated using payroll data from BLS) was 20%; a simulation by BEA found that, had payroll data from the Economic Census been used instead, the industry’s growth rate would have been measured as 28%. Access to microdata would permit BEA to investigate such inconsistencies to improve the accuracy of industry-specific GDP growth statistics.
Improving the industry sampling frame for producer price indexes would provide more accurate industry output deflators. BLS would use Economic Census revenue by product line data to select the specific items to include in industry “market baskets,” thereby reducing respondent burden, improving sampling efficiency, and producing more accurate price indexes.
Reconciling geographic classifications would improve confidence in the regional income measures used by the federal government to allocate hundreds of billions of dollars to states and used by state governments to forecast revenue. BEA’s state personal income, for example, is used to allocate more than $300 billion in federal funds (including Medicaid) to states and is used by state governments to project tax revenues and plan budgets.
Income source data from BLS and the Census Bureau are often inconsistent. Discrepancies between BLS wage and salary and Census Bureau payroll measures for the private sector vary from state to state. In comparison with Census Bureau data, for example, 2007 total state-level private sector wage and salary data reported by BLS were 6% higher in New Hampshire but 12% lower in Alaska. Although large states may have discrepancies that are small in percent terms (2% in both Michigan and New York), their state-level differences can amount to billions of dollars. Reconciling classification and coverage differences through data synchronization may improve the accuracy and comparability of these estimates, enhancing confidence in their use.
Allowing BEA to supplement its sample frame with Census Bureau identifier information for noncorporate as well as corporate firms would correct an understatement in U.S. trade statistics. As the producer of the official International Transactions Accounts statistics, BEA gathers services trade data through business surveys. The Economic Census includes a question on exports of services, which could help BEA identify firms that engage in services trade. This legislation would permit the Census Bureau to share firm identifier information for noncorporate as well as corporate firms, allowing BEA to develop a more comprehensive universe.
A BEA study comparing data on the export of services found that many firms that reported export revenue on the 2002 Economic Census did not report exports to BEA. (The reverse also held true.) Inconsistent reporting occurred in many important service sectors. In the management and consulting services sector, at least 80% of the firms reporting export revenue to the Census Bureau did not report these exports to BEA. As many as 250 telecommunications firms—with a mean export value of $18 million—were also not being captured in the official statistics. By using Census Bureau information to develop a more comprehensive sample frame, BEA could correct the understatement of U.S. services exports.
By allowing BLS and the Census Bureau to share and synchronize their business data, these agencies could reconcile classification differences, increase coverage, and research reporting differences—all improving the accuracy and comparability of the statistics they produce. Granting BEA access to business FTI for the growing noncorporate segment of businesses (especially in the service sector) will result in improved measures of income and international transactions. The proposed modest changes to the IRS code will be essential for achieving these improvements. By leveraging existing business data, we will be able to gain new insight into our economy, aiding business decisionmakers and policymakers at all levels of government.
Science Policy Actions
The ASA wrote to the U.N. Special Rapporteur regarding mistreatment of statisticians in Argentina.
The ASA signed letters in support of the FY12 budgets for the U.S. Census Bureau, Economic Research Service, National Agricultural Statistics Service, and National Institutes of Health.
The ASA signed a letter in support of reauthorizing and strengthening the current Math and Science Partnership Program at the Department of Education.
The ASA nominated members for the National Assessment Governing Board.