1

Introduction

The complexity and dynamism of the U.S. economy make it imperative that accurate, relevant, and timely data about revenues, employment, and other characteristics of U.S. businesses are available for national, state, and local policy makers, corporate business leaders, main-street businesses, trade associations, the media, and researchers. Providing such data in a manner that best uses scarce resources and that minimizes the reporting burden on the business community is a critical priority for the U.S. Census Bureau.

Several years ago, the Bureau launched an effort to reengineer its economic statistics portfolio. The need for change was driven by such factors as budget pressures, the growing difficulty in obtaining responses to business surveys, the growing desire from data users for more detailed information, and the perception that, in the face of increasing competition from privately produced data, official statistics from the Census Bureau are at risk of becoming less relevant.

The Census Bureau’s economic statistics programs include the quinquennial economic censuses and—the subject of this Consensus Study Report—a suite of annual economic surveys, which cover virtually every sector of the economy. The annual surveys have important uses, including benchmarking, or adjusting, estimates from the smaller monthly and quarterly surveys also conducted by the Bureau and providing estimates for the National Income and Product Accounts. The annual surveys were developed at different times and have continued to operate largely independently of each other. As the Bureau has recognized, these surveys need to be made more consistent and relevant to user needs in their content, more

cost effective to operate, and less burdensome on businesses.¹ To provide expert input to its reengineering of the annual surveys, the Census Bureau requested a study by a panel convened under the Committee on National Statistics of the National Academies of Sciences, Engineering, and Medicine: see Box 1-1 for the statement of task to the Panel on Reengineering the Census Bureau’s Annual Economic Surveys.

For context, in this chapter we first review the history of the Census Bureau’s economic statistics programs. We then describe the scope for the panel’s investigation and summarize the work that was conducted to carry out its charge. Then we sketch our vision for a reengineered Annual Business Survey System (ABSS) that, when fully developed, would represent a significant improvement over the current suite of surveys in ways that

___________________

¹ For a broad outline of the challenges and opportunities for modernizing federal economic statistics, see Bostic, Jarmin, and Moyer (2016).

benefit data users, respondents, and the Census Bureau. We conclude by describing the contents of the remaining chapters of the report.

1.1 ECONOMIC STATISTICS AT THE CENSUS BUREAU

1.1.1 Economic Censuses

The U.S. Constitution mandated a decennial census of population, which was first conducted in 1790. From the beginning of the new nation, policy makers recognized the value of additional data, including measures of economic activity. As early as 1810, U.S. marshals and their assistants, who conducted the population enumeration for censuses from 1790 to 1870, were instructed also to provide information about the various manufacturing establishments in their districts. A manufacturing census was conducted together with the population census in 1810, 1820, and from 1840 to 1900. Beginning in 1840, statistics were collected on other economic sectors, such as mining and agriculture. Because of the absence of training and quality control in the field, the quality of these statistics on economic activity was suspect in the period when U.S. marshals collected the data, but the fact that the U.S. Congress continued to mandate their collection, increasing the number of data items from census to census, indicates the importance attached to obtaining this information.²

The Census Bureau was established as a permanent agency in 1902, and its economic statistics programs proliferated in the 20th century. The census of manufactures became a separate census beginning in 1905. The first census of business, covering retail and wholesale trade, was conducted in 1930, and it was soon expanded to include service trades. The economic censuses were suspended during World War II in favor of war-oriented surveys. They resumed with the 1947 Census of Manufactures and the 1948 Census of Business.³ Public Law 80-671, enacted in June 1948, authorized quinquennial “censuses of manufacturing, of mineral industries, and of other businesses, including the distributive trades, service establishments, and transportation” (Bureau of the Census, 1996, p. A-14).

Because of an unfounded belief that the data were readily available from the private sector, no funds were requested or appropriated for data collection in 1954 for the 1953 economic censuses. Outcries from government agencies and the business and academic communities led to the appointment by the Secretary of Commerce of a special review committee,

___________________

² For a description of the evolution of the decennial census, including economic statistics, through 1890, see Wright and Hunt (1900).

³ See https://www.census.gov/history/www/programs/economic/economic_census.html [November 2017].

which recommended that the economic censuses be reinstated, which was done in 1955 with the collection of data for 1954 (Bureau of the Census, 1996, pp. A-14, A-15). Following economic censuses for 1958 and 1963, beginning in 1967, they have been taken every 5 years for years ending in 2 and 7.

The scope of the economic censuses continued to expand in the second half of the 20th century. A census of commercial fisheries and surveys of major forms of transportation were added to the 1963 economic censuses, and a census of construction began on a regular basis in 1967. A major expansion of the economic censuses took place in 1992, adding more transportation industries, as well as finance, insurance, real estate, communications, and utilities—a group of industries accounting for more than 20 percent of U.S. output. The economic censuses at present, along with the censuses of agriculture and governments, cover virtually the entire economy, with exceptions only for forestry, agricultural support services, rail transportation, religious organizations, and employment in private households.

1.1.2 Economic Surveys

Censuses have always been expensive, time-consuming, and burdensome to conduct and, consequently, have not been fielded in the United States more often than every 5 years (economic censuses) or every 10 years (population census). Yet users want more timely information. Beginning in the 1930s, theoretical and practical advances made it possible to use probability sample surveys as the basis for statistical information from the Census Bureau and other federal statistical agencies. Such surveys were much less costly and burdensome than censuses, so they could be fielded more often, and, with the use of appropriate probability sampling techniques, the sampling error in survey estimates could be quantified. Probability samples came into use not only for stand-alone surveys but also to reduce the burden in censuses by collecting some items on a sample basis (see Citro, 2014).

The Census Bureau began the Monthly Survey of Wholesale Trade in 1946 and the Annual Survey of Manufactures in 1949, to name just two examples of what is now an extensive portfolio of annual, quarterly, and monthly economic surveys.⁴ The quarterly and monthly economic surveys are fielded with the primary purpose of providing up-to-date key economic indicators as defined in Statistical Policy Directive No. 3 (U.S. Office of Management and Budget, 1985). The annual surveys have considerably larger sample sizes and are fielded to provide more robust estimates that are

___________________

⁴ See https://www.census.gov/econ [November 2017].

used to evaluate and annually readjust the time series of economic indicators from the monthly and quarterly surveys and for other purposes. The annual, quarterly, and monthly surveys generally provide estimates for the nation as a whole, although the Annual Survey of Manufactures and its supplement, the Management and Organizational Practices Survey, provide state estimates, and the Annual Survey of Entrepreneurs provides estimates for both states and large metropolitan statistical areas.

The Census Bureau has other business statistics programs that are intended to provide estimates for states and smaller geographic areas, including not only the economic censuses, but also such programs as County Business Patterns (CBP) and the Longitudinal Employer-Household Dynamics Program (LEHD). These two programs draw information largely from administrative records, such as tax returns and other information that are input to the Bureau’s Business Register (for the CBP) and employment records filed with state employment security agencies (for the LEHD).

1.1.3 Lack of Integration of Economic Censuses and Surveys

The various economic censuses and surveys are vitally important for the nation, but their lack of integration prevents them from being as useful, cost-effective, and minimally burdensome on businesses as they could be. Thus, although the Census Bureau refers to the “economic census,” there are in fact multiple censuses for the various economic sectors (e.g., manufacturing, wholesale trade, construction), which are run largely independently of one another. Similarly, the annual economic surveys are run largely independently, and they differ in many ways beyond those necessitated by different subject matters and target industries. For example, there are differences in sampling methods, frequency of updating of the sampling frame, the unit for sampling and data collection (e.g., establishment or enterprise), the wording of questions and instructions on the same topic (e.g., reporting of total receipts), the roles played by the Business Register and administrative records, the calendar period during which data are collected and for which information is to be reported, the method for determining which establishments or enterprises are to be included with certainty in the sample, and how many years noncertainty businesses are kept in the sample. These differences make it difficult to achieve efficiencies, such as use of the same editing and imputation software, and they place additional burdens on multisector businesses that are included in more than one survey each year.

The Census Bureau’s initiative to reengineer its economic censuses and surveys has as its aim to harmonize and integrate them as much as possible to achieve significant improvements in the relevance, timeliness, cost effectiveness, and data quality, and to reduce the burden on respondents.

The Bureau commendably has already made strides in this direction. The 2017 economic censuses, for example, will all offer electronic reporting as the primary response mode, and several of the annual economic surveys now use a common software platform for editing and imputation. The Census Bureau also recently received approval to field a new Annual Business Survey to provide information on business dynamics, which is jointly sponsored by the National Center for Science and Engineering Statistics (NCSES) at the National Science Foundation and combines several existing surveys (see Chapter 8).

1.2 STUDY SCOPE

The Panel on Reengineering the Census Bureau’s Annual Economic Surveys, comprising survey specialists, economists, and economic survey data users, was charged to review the design, operations, and products of the Bureau’s suite of annual economic surveys of manufactures, wholesale trade, retail trade, services, and other business activities. The review also included two surveys that are fielded every 5 years as supplements to an annual survey. The panel was to recommend short-term and longer term steps to improve the relevance and accuracy of the data, reduce respondent burden, incorporate nonsurvey sources of data when beneficial, and streamline and standardize Census Bureau methods and processes across surveys. The panel also was asked to consider improvements in data products of value to users with respect to accuracy, relevance, frequency, timeliness, and geographic detail, including the production of subnational estimates. The overall goal is improved information about the nation’s constantly changing economy.

The panel gathered information, deliberated, and prepared this report with its findings and recommendations over a 24-month period beginning in September 2015. Areas considered by the panel included data gaps and redundancies in survey content, the definition of the universe and unit of data collection (especially whether enterprise or establishment), sampling frame, sample design, questionnaire design, data collection procedures, data processing, methods for filling in missing data, and data products and data dissemination methods.

The panel focused its efforts on the Census Bureau’s annual surveys of the private U.S. economy with two exclusions: agriculture, which is surveyed by the National Agricultural Statistics Service, and the government sector, which was reviewed by a previous panel of the Committee on National Statistics (National Research Council, 2007). The 11 surveys covered by the panel are briefly described in Box 1-2; also see Chapter 8. Additional detail and links to websites with more information are in Appendix B.

1.3 THE PANEL’S WORK

1.3.1 Understanding the Landscape

To fulfill its charge, the panel sought to develop a thorough understanding of how the surveys it was asked to examine are used, how they are currently designed and carried out, how estimates based on the surveys are produced, and how those estimates are disseminated. The panel

devoted considerable time to reviewing information about the surveys that Census Bureau staff members provided. Detailed descriptions of each survey were requested and made available prior to the panel’s first meeting. These descriptions later were supplemented with responses to the panel’s lengthy list of specific questions regarding aspects of each survey: sample design, data collection, production of survey estimates, and dissemination of that information. Bureau staff made presentations during each of the panel’s first four meetings and at two meetings of some panel members that focused on the Business Register and the design of samples for the annual surveys.

In the panel’s review of the existing suite of annual economic surveys, two topics that received significant attention were the Business Register and the definition of a statistical unit for data collection and estimation. The Business Register is the master list used as a sampling frame for all of the Bureau’s private-economy censuses and surveys (although there are differences among the surveys in how frequently they refresh their samples). It includes all establishments and enterprises known to the Bureau, both those with employees and those without employees.

In the Annual Survey of Manufactures and the Management and Organizational Practices Survey, the statistical unit of observation is generally the establishment, defined as a single location that carries out an activity, which for these surveys is manufacturing a product (establishments are classified based on their primary activity). In all of the other surveys, the statistical sampling unit is generally the enterprise or company, which may conduct its business at one or more establishments. If a respondent finds it difficult to report on either of these bases, some intermediate entity may be used as the statistical unit, though this is not necessarily recorded on the Business Register. The use of different reporting units for different surveys is one of the challenges associated with harmonizing the annual economic surveys, which is made even more challenging because of the dynamic nature of businesses, which may add, close, or relocate establishments or may change ownership and organizational structure.

In addition to understanding the surveys themselves, an important task for the panel was to identify their various users and potential users, which include businesses and trade associations; state and local governments; and other federal agencies, including several statistical organizations (among them the Bureau of Economic Analysis, the Bureau of Labor Statistics, the Division of Research and Statistics of the Federal Reserve Board of Governors, and the National Health Statistics group in the Office of the Actuary at the Centers for Medicare & Medicaid Services). From each group, the panel sought to learn about the economic surveys data they find to be valuable, how those data are accessed, gaps in available data, and how they compensate for those gaps. In the case of users from other statistical

agencies, the panel sought to learn how these users maintain, update, and integrate the Census Bureau’s economic surveys data with their own statistical products. Many users stated that the Census Bureau’s annual economic data are the “gold standard” for what they measure. Many also indicated, however, that the data are not sufficiently timely and lack product and geographic detail, leading data users to turn to administrative data or to estimates from commercial vendors.

The panel consulted with the staffs of two statistical agencies outside the United States—Statistics Canada and Statistics Netherlands—that have undertaken significant efforts to reengineer their business surveys. The panel learned from their experiences about what worked and what did not. As discussed in Chapter 8, participants in both efforts emphasized the full commitment of top management as the single most important ingredient in the eventual success of their efforts.

During its conversations with staff at Statistics Canada, the panel heard about their Account Manager Program, which includes 10 individual enterprise portfolio managers supported by 6 analysts who cover about 320 enterprises, typically the largest and most important enterprises for the production of economic statistics. The account managers act as focal points for business contacts, profile the businesses to understand each company’s business and accounting structures, and ensure that the business register accurately represents the corporate structure. The account managers also coordinate and assist in the survey collection process, visiting the enterprises as needed to resolve complex issues. This model for managing relationships with large respondents differs from the current practice of the Census Bureau and, as discussed in more detail in Chapter 3, significantly influenced the panel’s thinking.

1.3.2 A New Approach for the Future—An Integrated System

Early in the panel’s deliberations, the idea of an integrated ABSS to replace the current suite of annual instruments began to take hold. William Bostic, then associate director for economic programs at the Census Bureau, presented this concept to the panel at its first meeting in October 2015 and again at its June 2016 meeting with data users. As a group, the users viewed the suggestion favorably, seeing the potential for improved consistency, timeliness, and relevance.

Some on the panel were initially skeptical, noting that the different surveys in the current suite reach different types of respondents and serve a variety of different purposes. Over time, however, the panel concurred with this idea. More specifically, as elaborated in the recommendations detailed later in the report, the panel envisions an ABSS that would contain a core set of questions asked of all respondents, with appropriate industry cus-

tomization, plus modules containing industry-specific and topical questions. It would have a rotating panel sample drawn from a redesigned comprehensive Business Register that links establishments, companies, enterprises, and alternative reporting units, which would provide for more efficient sample design, data collection, editing, estimation, and dissemination while at the same time improving the quality and timeliness of the data received and reducing the reporting burden on businesses. It would have a single application programming interface to access data, which would allow for improved dissemination. An integrated ABSS, together with account managers, also would allow the Census Bureau to better manage its relationships with large and complex businesses.

An ABSS would have the goal of providing more timely and accurate information currently collected in 8 of the 11 surveys within the panel’s purview (see Box 1-2, above): ASM, M3UFO, MOPS, ARTS, AWTS, SAS, ACES, and ICTS. Two of the other surveys, COS and SQ-CLASS, would be part of a reengineered system for updating the Business Register. The Annual Survey of Entrepreneurs, which pertains to people who own businesses, rather than the businesses themselves, would remain separate, with design changes as appropriate to complement an ABSS. Indeed, as noted in Box 1-2 above, the new ASE is being combined with several other surveys in a joint Census Bureau–NCSES program to improve measurement of business dynamics in the United States.

An ABSS would include two other important components: (1) the use of small-area estimation techniques to produce subnational estimates, which are much desired by users; and (2) greater use of administrative records to achieve improvements in quality, timeliness, and burden reduction. Although response to the annual economic surveys, like the economic censuses, is required by law, the annual economic surveys have not been immune to the well-documented phenomenon of declining response to surveys of all kinds. The Census Bureau has been able to maintain response rates for the surveys at relatively high levels—70 to 80 percent—but only with increased costs.⁵ Responding businesses also do not always complete questionnaires fully. Consequently, it is imperative to expand the use of already collected administrative information to reduce burden and thereby increase the likelihood that businesses will respond with completed questionnaires (see National Academies of Sciences, Engineering, and Medicine, 2017, Ch. 2).

___________________

⁵ Communication from Kevin Deardorff, U.S. Census Bureau, during the panel’s October 29–30, 2015, meeting.

1.3.3 Report Structure

The remainder of the panel’s report proceeds systematically to assess and make recommendations for the Census Bureau’s annual economic surveys by following the steps in the process for designing any data collection program that is anchored in a survey. Table 1-1 outlines the basic steps, from prefield to field to postfield activities. For each step, survey designers need to identify potential sources of error (e.g., misspecification of concepts to be measured, coverage errors from missing or duplicate entries in the sampling frame, reporting errors by respondents), so that the design can anticipate and minimize the various errors to the extent possible.

The first step in the survey design process, as shown in Table 1-1, is to specify the data needed from the survey. Chapter 2 examines the Census Bureau’s annual economic surveys program from the perspective of data users. What do users want, need, and use? What are the gaps and limitations in currently available data? How can the tension between users’ interest in consistent time series and the need for the surveys to adapt to changes in the economy best be resolved? This chapter presents seven conclusions and offers one recommendation.

Chapter 3 turns to the question of the universe that the annual surveys should sample, which requires considering the fundamental question of the definition of a business. The chapter addresses how the definition relates to the statistical units currently sampled by the annual economic surveys, the use of the Business Register as a sampling frame, improved methods for keeping the register up to date, including the use of information fed back from the annual surveys, and an Account Manager Program for larger enterprises.

Chapter 4 describes the data collection processes for the current surveys and discusses the potential for administrative records and other data sources to reduce the need for survey questions, the design and testing of survey questionnaires, and the modes for obtaining survey responses. This chapter has a particular focus on the burden imposed on survey respondents. The chapter also addresses new data content that users want that would be useful and feasible to include in the annual economic surveys and, ultimately, an ABSS.

Chapter 5 addresses the approaches to sampling and the production of estimates in the current system, paying particular attention to inconsistencies in the methods currently used and recommending more standardized and efficient approaches. The chapter also discusses methods that may be able to provide greater geographic and industry detail and produce preliminary estimates for key data items in response to users’ expressed needs.

Chapter 6 considers the process of editing data as they flow from economic reporting units to the Census Bureau for each of the annual economic surveys and procedures used for disclosure protection and quality control.

TABLE 1-1 Stages in the Design of a Data Collection Program Anchored in a Survey and Component Activities

SOURCES: Adapted from Snijkers and colleagues (2013); this builds on the Generic Statistical Business Process Model of the U.N. Economic Committee for Europe (2013).

Chapter 7 discusses standardization of data products, data access tools that need to be available to users, data archiving, and the need for feedback loops from users to the Census Bureau to facilitate continuous improvement of data dissemination.

Finally, Chapter 8 presents the panel’s recommendation for an integrated ABSS and outlines how the Census Bureau can manage the transition to an ABSS from the current state of largely disparate annual economic surveys.

Appendix A presents the responses from data users who spoke with or wrote to the panel on data needs and other observations for the annual economic surveys. Appendix B presents descriptions of each of the surveys reviewed by the panel. Appendix C provides a detailed matrix of information about the surveys, which permits comparison of such features as sample design and statistical unit. The printed version of the report lists the survey characteristics included in the matrix; the full matrix is available online only. Lastly, Appendix D lists biographical sketches of panel members and staff.

1.4 REFERENCES

NOTE: All URL addresses were active as of November 2017.

Bostic, W.G., Jr., Jarmin, R.S., and Moyer, B. (2016). Modernizing federal economic statistics. American Economic Review, 106(5, May), 161–164.

Bureau of the Census. (1996). History of the 1992 Economic Census. Washington, DC: U.S. Department of Commerce. Available: https://www.census.gov/history/pdf/1992econhistory.pdf.

Citro, C.F. (2014). From multiple modes for surveys to multiple data sources for estimates. Survey Methodology, 40(2), 137–161. Available: http://www.statcan.gc.ca/pub/12-001-x/2014002/article/14128-eng.pdf.

National Academies of Sciences, Engineering, and Medicine. (2017). Innovations in Federal Statistics: Combining New Data Sources While Protecting Privacy. Washington, DC: The National Academies Press. doi: https://doi.org/10.17226/24652.

National Research Council. (2007). State and Local Government Statistics at a Crossroads. Panel on Research and Development Priorities for the U.S. Census Bureau’s State and Local Government Statistics Program. Committee on National Statistics, Division of Behavioral and Social Sciences and Education. Washington, DC: The National Academies Press. doi: https://doi.org/10.17226/12000.

Snijkers, G., Haraldsen, G., Jones, J., and Willimack, D. (2013). Designing and Conducting Business Surveys. Hoboken, NJ: John Wiley & Sons.

U.N. Economic Committee for Europe. (2013). Generic Statistical Business Process Model (GSBPM), Version 5.0. Geneva, Switzerland. Available: http://www1.unece.org/stat/platform/display/metis/The+Generic+Statistical+Business+Process+Model.

U.S. Office of Management and Budget. (1985). Statistical Policy Directive No. 3: Compilation, release, and evaluation of principal federal economic indicators. Federal Register, 50(186), 38932–38934. Available: https://www.whitehouse.gov/sites/whitehouse.gov/files/omb/assets/OMB/inforeg/statpolicy/dir_3_fr_09251985.pdf.

Wright, C.D., and Hunt, W.O. (1900). The History and Growth of the United States Census. Prepared for the U.S. Senate Committee on the Census. Washington, DC: U.S. Government Printing Office. Available: https://www.census.gov/history/pdf/wright-hunt.pdf.