2

What Data Users Want, Need, and Use

There are many users of the economic data produced by the U.S. Census Bureau and other federal statistical agencies, each with somewhat different data needs. One group of users—including officials at the Federal Reserve Board, the Department of the Treasury, the Centers for Medicaid & Medicare Services (CMS), the Office of Management and Budget, and the Congressional Budget Office, as well as many business economists—primarily require national data for assessing current economic conditions and as input to their economic forecasting models. For these users, the accuracy of the data, their timeliness, and their consistency over time all are important, while geographic disaggregation is less so. In contrast, state and local government officials and local chambers of commerce, together with lawmakers who are devising formulas for the allocation of federal funding, care much more about the availability of geographically disaggregated statistics that capture local economic conditions and changes in those conditions over time. And trade associations and others with an interest in particular sectors of the economy seek data that are specific to their industries of interest. Academic researchers make use of published data, but they may also be interested in analyzing underlying survey or administrative microdata.

The information of interest to these disparate users includes data on output, investment, inventories, imports and exports, prices, employment and unemployment, and wages and benefit costs, among other economic statistics. In the U.S. statistical system, there is no single statistical agency that is responsible for all of these data. Rather, the production of relevant data is spread across multiple agencies—primarily, the Census Bureau, the

Bureau of Labor Statistics (BLS), and the Bureau of Economic Analysis (BEA), but also a number of other agencies—and the data production process involves important interactions among the different actors.

The Census Bureau’s Economic Programs Directorate publishes statistics that are directly useful to a range of external data users, but other federal data-producing agencies are perhaps the most important customers for the directorate, and this is especially true of the annual economic surveys that are the focus of this report. Other statistical agencies rely on data from the economic censuses and the annual economic surveys to benchmark their own data series. Internal to the Census Bureau, data from the annual economic surveys play a key role, along with the quinquennial economic censuses, in benchmarking the Bureau’s monthly and quarterly economic indicator series. These indicators include advance monthly sales for retail and food services, monthly wholesale sales and end-of-month inventories, quarterly services revenues, and others.¹ One way to think about the importance of the annual economic surveys is that they provide a regular course correction for other important business statistics, which in turn are critical for ongoing policy, investment, and business planning decisions.

Although the use of the data from the annual economic surveys is thus at present largely indirect, it is possible to imagine a future in which the direct uses of the annual data are considerably expanded. An integrated Annual Business Survey System (ABSS) that collected more comprehensive and more consistent information across the entire economy could be of considerably greater interest to data users than the current disjointed suite of separate annual surveys. A somewhat accelerated and more predictable release schedule also could increase the appeal of an ABSS to users, as could the addition of selected new content. This chapter explores these topics.

2.1 FRAMEWORK

To inform itself about the needs of current and potential users of annual economic data from the Census Bureau, the panel collected information about user needs in a variety of ways. On June 2, 2016, the panel invited three important groups of data users to share their insights in an open meeting. The advance questions circulated to participants are reproduced in Appendix A, along with a table of observations and suggestions from representatives of specific agencies and data user types.

Federal agencies that rely on Census Bureau economic data, both for benchmarking their own statistical data series and as input to modeling and forecasting activities, made up one group of meeting participants. The

___________________

¹ See https://www.census.gov/economic-indicators [October 2017].

four speakers in this group represented BEA, the productivity measurement program at BLS, the Division of Research and Statistics at the Federal Reserve Board, and the National Health Statistics group in the Office of the Actuary at CMS. A second group was made up of private-sector consulting firms, represented by Haver Analytics and IHS Economics. Both of these firms have extensive experience with the use of government economic data and provide sophisticated data services to business clients. In addition, representatives of the National Retail Federation and Powerlytics provided input to the panel. A third group of participants were speakers from four regional and state agencies. Their perspectives were supplemented by interviews with a number of professionals from other state and local agencies.

In addition to the perspectives gleaned through these channels, the panel reviewed an extensive array of materials and documents released over the last decade by key user groups. These included materials from meetings of the Federal Economic Statistics Advisory Committee (FESAC), which advises BEA, BLS, and the Census Bureau; the BEA Advisory Committee; and the Census Scientific Advisory Committee. The panel also reviewed relevant meeting notes, conference proceedings, reports, and other published documents from many organizations, including the Committee on National Statistics of the National Academies of Sciences, Engineering, and Medicine; the Conference on Research in Income and Wealth; the Committee on Economic Statistics of the American Economic Association; the Statistics Committee of the National Association for Business Economics; and the National Academy of Public Administration. These materials included a wide range of recommendations related to improving the economic data available for understanding innovation, globalization, sustainability, and many other topics. Finally, the panel reviewed recommendations related to economic data offered by the United Nations, the International Monetary Fund, OECD, Eurostat, and other international organizations.

Putting the many suggestions and recommendations for improvements in the Census Bureau’s annual economic surveys into a digestible form required an organizing framework. Statistical agencies in the United States and abroad, building on a seminal paper by Brackstone (1999), have developed useful frameworks for assessing a dataset against user needs. Commonly referred to as “quality frameworks,” they specify a handful of key attributes or dimensions to measure, document, and drive improvement in data quality, broadly conceived, for a statistical program or set of programs. For example, the Eurostat (2000) framework includes relevance, accuracy, timeliness and punctuality, comparability across time and among geographic areas, and consistency with other data series. Biemer and colleagues (2014) present a similar framework and, in addition, break down accuracy into sampling error, where applicable, and seven types of nonsampling error. Daas and colleagues (2011) and Federal Committee on

Statistical Methodology (2013) focus on quality dimensions for data series based on administrative records.

Based on these models, the panel adopted the framework shown in Box 2-1 to organize what we heard about the needs of data users as they pertain to the Census Bureau’s annual economic surveys. We use the five dimensions of accuracy, timeliness, relevance, consistency, and accessibility to structure the remainder of the chapter’s discussion.

2.2 ACCURACY

All of the data users from whom the panel received input, without exception, made clear that they perceive the economic data produced by the Census Bureau to be of high quality, with several referring to Census

Bureau data as the “gold standard” for the things it measures. Although statistics based directly on the Census Bureau’s annual economic surveys are not frequently used by external customers, they are a critical part of the foundation for many U.S. economic statistics that are used much more extensively, meaning that data from the annual surveys are heavily used indirectly. Without annual benchmarking and updating, critical information such as the gross domestic product (GDP) series would become successively less accurate and relevant over the 5 years between the economic censuses.

The Census Bureau’s annual economic surveys are one piece of the Census Bureau’s larger program of economic statistics. The economic censuses are conducted once every 5 years. They cover more than 28 million establishments and provide the most accurate, comprehensive, and detailed data available on the U.S. economy. These data are used to benchmark the country’s input-output tables, which in turn are used to benchmark GDP and other statistical data series. The Census Bureau’s annual economic surveys collect data from many fewer businesses than the economic censuses, but are sufficiently large to provide detailed industry data at the national level. Although not as comprehensive as the economic censuses, they are used as benchmarks for the annual GDP updates, as well as for data series produced by other statistical agencies. Finally, the Census Bureau’s monthly and quarterly economic indicator surveys have much smaller samples. They provide information that is closer to real time but is much less detailed. The indicator estimates are anchored to the annual economic surveys and economic censuses data.

Table 2-1 summarizes the critical economic time series that are benchmarked to, extrapolated from, or otherwise rely on data from one or more of the Bureau’s annual economic surveys. These include such key economic indicators as GDP, the producer price index, industrial production statistics, retail sales, other monthly and quarterly Census Bureau indicators, and indicators from a number of other federal data programs.

The methods used to produce the GDP statistics illustrate the important role played by the annual economic surveys data. When the first estimate of GDP for a quarter is produced, relatively limited information is available. About half of the initial GDP estimate is essentially an extrapolation and about half is based on initial quarterly or monthly Census Bureau data. The initial GDP estimates are revised as more complete and accurate data become available. Early monthly survey results from the Census Bureau are replaced by revised results from more complete tabulations in subsequent months and later by the more comprehensive annual economic surveys with their larger samples and more detailed information. About once every 5 years, when results from the economic censuses are released, the GDP estimates undergo a more comprehensive revision to incorporate those data.

TABLE 2-1 Key Indicators That Rely on Data from the Census Bureau’s Annual Economic Surveys, by User Agency or Organization

The benchmarking of the GDP and other statistics to the annual economic surveys data is critical because of the important uses of the benchmarked data for guiding economic policy and other decisions. For example, GDP estimates are used in the production of revenue and expenditure estimates for use in federal, state, and local budget projections. Errors in the trend in the growth rate for real GDP of as little as 0.5 percent can cause errors in the national output projected over the 5-year federal budget planning horizon of more than $1.7 trillion. Real GDP and the personal consumption expenditures measure of inflation, both derived from the national accounts, are key inputs to the Federal Reserve Board’s data-driven monetary policy. More than $320 billion in federal funds are allocated using regional BEA personal income and GDP data that are benchmarked to and extrapolated from the annual economic surveys.² States and localities throughout the United States use the BEA regional data for budget projections and the allocation of funds. Some 18 states have statutory limits on revenues and spending that are tied to the state personal income or gross

___________________

² For definitions of BEA concepts, see https://bea.gov/glossary/glossary.cfm [June 2018].

product statistics from BEA. Indicators such as GDP and inflation also have a large effect on consumer confidence, spending, and financial markets, affecting more than $56 trillion in household and business investments. Through their effects on interest rates, exchange rates, and stock and bond prices, these data affect everyone who invests in financial markets, buys a home, borrows for a car, or purchases imported goods and services.

While data users have high regard for the accuracy of Census Bureau economic data, revisions to Census Bureau estimates and to GDP and other data series that rely on Census Bureau data are cited by many of them as a concern. The replacement of Census Bureau monthly and quarterly indicator data with annual survey data is a significant source of revisions to GDP and other data series. Revisions also occur when data from the economic censuses replace the annual surveys data (see, e.g., Landefeld, Seskin, and Fraumeni, 2008).

The Census Bureau already does a good deal to integrate the monthly surveys, the annual surveys, and the economic censuses. For example, the subsamples for the monthly indicator surveys and the sample for the Quarterly Services Survey are taken from the larger samples for the annual surveys. The samples for the annual surveys, in turn, are based on the economic censuses, with various provisions for annual updating (see Chapter 5). In addition, microdata from companies in the monthly and the annual samples are reconciled annually, as are the largest differences in industry-level estimates between the monthly and the annual surveys. The annual surveys, in turn, are reconciled with the 5-year economic censuses. Continuing these efforts and ensuring that the monthly and quarterly economic indicators, the data from a redesigned ABSS, and the economic censuses data are as well-integrated as possible would significantly improve key macroeconomic data that are important to policy.

A suggestion offered during discussion at the panel’s June 2016 meeting (see Appendix A) concerned the availability of quality metrics for the Census Bureau’s annual economic surveys and other economic data. These might include routine publication of historical revisions, coefficients of variation as a measure of sampling error, response rates, and other information. In particular, users suggested, information on historical revisions, including previously published and revised data, would be helpful in understanding overall revisions to the annual data, the extent of biases and other errors, and trends in revisions both over time and across industries. A practice of regularly carrying out and publishing revision studies, including assessments of the sources of revisions (such as changes in questions and coverage errors in surveys compared with the economic censuses), would not only help users, but would also help the Census Bureau identify potential areas for improving the surveys.

CONCLUSION 2-1: The accuracy of data from the Census Bureau’s annual economic surveys is essential to their value for such important uses as benchmarking and extrapolating key macroeconomic statistics, such as gross domestic product. Data users would benefit from further improvements in the accuracy of the annual economic survey data and also from more information on their accuracy, including regularly published information on historical revisions and studies of the sources of and trends in the extent of revisions.

2.3 TIMELINESS

Timeliness may mean different things to different data users. Those concerned with the current state of the economy want data from the Census Bureau and other statistical agencies that are available in as close to real time as possible. Currently, quarterly GDP data are released less than 1 month after the end of the quarter and annual GDP data less than 1 month after the end of the year. Data on monthly retail sales, consumer prices, industrial production, and employment are released with lags of just 15 to 30 days following the end of the reference period, which is the second week of the preceding month. In contrast, data from the Census Bureau’s annual economic surveys for manufactures, retail trade, wholesale trade, and services are released—on average—13 months after the end of the reference year. There are good reasons why the lag in the release of the annual economic surveys data is longer than the lag in the release of other economic statistics. For example, data from the monthly indicator surveys, such as retail and wholesale trade, are based on smaller samples than the annual economic surveys and have data collection and dissemination systems that have been optimized for rapid turnaround. Still, the lags inherent in the release of the annual economic surveys data mean that they can never be a direct source for the current data demanded by a large segment of economic data users.

Although it is important to acknowledge the reasons for the lag, it seems that it would be desirable for the annual economic surveys data to be released more quickly. Releasing these data early enough that they could be used in the annual revisions of other key data series would be especially valuable. With the exception of data from the preliminary Service Annual Survey report, data from the annual economic surveys—including the Annual Survey of Manufactures, the Annual Retail Trade Survey, and the Annual Wholesale Trade Survey—currently are not available in time for the first annual revision of data series, such as GDP, but only in time for the second annual revision.

The Census Bureau’s recent work to accelerate the release of its principal economic indicators illustrates the returns to more accurate and timely

data for GDP and other estimates. The recently created Monthly Advance Economic Indicators report gives data users a significantly earlier glimpse of what is happening across several key aspects of the economy, without diminishing data quality, and it is expected to improve the accuracy of the early GDP estimates.³ In addition to the acceleration of the monthly indicators, the Census Bureau also has begun to release advance Quarterly Services Survey (QSS) estimates 3 weeks ahead of the regular QSS release. The new, accelerated schedule will enable the quarterly data to be incorporated into the second estimate of GDP. BEA simulations suggest that this acceleration will reduce revisions between the second and third quarterly release by an average of 0.1 to 0.2 percentage points (see Jarmin, 2017). The accelerations just described have resulted from improvements in the processes for estimation, imputation, and dissemination.

Accelerating the release of the annual economic surveys data to be better aligned with the production schedules for other key economic indicators similarly could help to improve the accuracy of those data series. For example, if the relevant annual economic surveys could be available in early May of the year following the reference year, they could be used in the annual GDP revision. If this acceleration could be accomplished through process improvement without diminishing data quality, as was achieved for the monthly economic indicator surveys, according to Census Bureau staff in a meeting with the panel, it would substantially improve the accuracy of the GDP and other estimates benchmarked to the annual economic surveys.

It may not be possible to complete the full annual survey data releases as quickly as users might like, but it might be feasible to release key topline data on an accelerated schedule, with additional data provided in a later second release. Some participants in the June 2016 meeting mentioned that this would be helpful to them. Another option might be to release preliminary annual economic surveys estimates, perhaps informed by auxiliary information, to provide users with an early sense of the data and then to release final estimates at a later date. Both ideas seem worth exploring, and they are discussed further in Chapter 5.

A related point is that data users view the release schedule for the annual economic surveys data as being inconsistent and opaque. Several of the data users who participated in the June 2016 meeting requested the creation of a visible and easily accessed calendar of annual data releases. They also requested better communications about revisions made to the data.

CONCLUSION 2-2: Accelerating the release of data from the Census Bureau’s annual economic surveys could significantly reduce revisions to gross domestic product and other macroeconomic and

___________________

³ See https://www.census.gov/econ/indicators/index.html [October 2017].

industry data series. Users would value early releases that contained either preliminary estimates or a limited set of statistics, providing that they proved sufficiently reliable regarding the direction and relative magnitude of measured changes. Users also would value the announcement of scheduled release dates for the annual survey estimates on an easily accessible calendar.

2.4 RELEVANCE

Data from the Census Bureau’s annual economic surveys are only one component of a larger set of data sources relied on by users who seek information about the economy to inform their decisions. These surveys should not be expected to provide all of the economic information that users need. Yet even with that caveat, users view the annual economic surveys as not sufficiently detailed to meet their meets. Based on comments from the participants in the panel’s June 2016 meeting and on its reading of pertinent materials the panel has identified some significant limitations and gaps in the topical coverage of existing economic data that a new ABSS could help to close.

2.4.1 Geographically Disaggregated Statistics

For users from state and local agencies, more geographically disaggregated data were the clear priority. Most of the major annual economic surveys provide only national data: the Annual Survey of Manufactures and the Manufacturers’ Unfilled Orders Survey do provide data by state, but only the Annual Survey of Entrepreneurs provides data for substate areas, specifically the 50 largest metropolitan statistical areas. Speakers from state and local agencies made clear that data at the national or even the state level are of little help to them. Ideally, they would like data that could be aggregated into custom local geographies, such as a user-specified collection of counties, and custom industries, such as a user-specified grouping of industries to represent, for example, advanced manufacturing, the green economy, or biotechnology. Given the lack of geographic detail available in the annual economic surveys, these users often turn to data from other sources, most notably, private data vendors who produce modeled estimates for detailed geographies and industries that rest on proprietary methods and a variety of underlying information sources. The business participants in the meeting seconded the interest in data that are more geographically disaggregated, along with expressing an interest in more industry and product detail.

Currently, the Census Bureau provides geographically disaggregated data mainly through its County Business Patterns program, which rests on information derived from the administrative data in the Business Register,

the American Community Survey, and the decennial census. Not coincidentally, these data are also some of the key inputs into the modeled local data estimates produced by private data vendors. Although it is unlikely that the Census Bureau could ever meet all of the desires of users for very local and customized data, participants in the June 2016 meeting did suggest that the Census Bureau could be doing more to leverage “big data” and private-public partnerships to create estimates of economic activity that better meet data users’ needs for timely and disaggregated data. The panel acknowledges that such partnerships may not be easy to develop, but it believes the idea merits consideration.

2.4.2 More Detailed Data by Industry and Other Characteristics

Another common theme in users’ comments, as noted above, was the need for greater industry detail—this was mentioned by representatives from BEA and BLS, as well as meeting participants from state and local agencies and the business community. The lack of detail in currently available data is especially severe for services. At present, data are available for only 406 industry categories for services, a sector that accounts for nearly 60 percent of GDP; by comparison, data are available for 614 industry categories for manufacturing, a sector that accounts for just 12 percent of GDP. Greater detail on rapidly growing and highly productive industries, such as information technology, was identified specifically by users as important for keeping up with changes in the economy. A related comment was that some users consider the North American Industry Classification System (NAICS) to be out of date and would like to see it revised more frequently, though this obviously is not a change that could be effected by the Census Bureau on its own.

Other areas for which meeting participants said additional detail would be valued include the categories used in collecting expense data; distinguishing between domestic and international revenue sources; and breakdowns between exporting and nonexporting enterprises and foreign and domestically owned enterprises. More detailed data would improve the accuracy of input-output tables since narrower industries generally are more homogeneous, and the patterns of activity that the input-output coefficients summarize therefore are more likely to be stable over time (see Box 2-2). Such estimates are important to the production of GDP statistics, industry projections and impact studies, productivity analysis, regulatory analysis, analysis of globalization, and a wide range of other analytic work by academia, business, and policy. In addition, data users expressed interest in more detail on revenue sources and expenses and more detailed breakdowns of published statistics by company characteristics.

CONCLUSION 2-3: The Census Bureau’s current annual economic surveys do not satisfy data users’ strongly expressed interests in more detailed information for geographic areas, for services industries, and for new and expanding industries, especially in information technology.

2.4.3 Gaps in Subject Coverage

In addition to lacking geographic detail and enough industry detail, the current annual economic surveys provide no information for some important industries and enterprise or establishment characteristics. One obvious gap in coverage by industry, which was mentioned by the business participants at the June 2016 meeting, is the lack of data for construction. The panel’s recommended ABSS ideally would cover mining and construction as well as the industries for which an annual survey currently exists.

Gaps in the coverage of U.S. economic statistics have been a major theme at various meetings of data users over the last decade, including the impact of globalization on economic statistics and the measurement of intellectual property. Addressing these gaps will depend at least in part on changes in the annual economic surveys. Our reading of the materials (see, e.g., Houseman and Mandel, 2015) suggests five areas with the most important unmet needs for economic data: innovation, the effects of globalization, structural change, sustainability and welfare, and the effects of financial crises.

1. Better measures of innovation. The slowdown in productivity growth in the United States and other developed economies is

3. a major source of policy concern (see, e.g., Bernanke, 2013; Cardarelli and Lusinyan, 2015; Council of Economic Advisers, 2014; Feldstein et al., 2017; Gordon, 2013), yet there is little agreement on the relative importance of the various possible factors behind the slowdown. Researchers and statistical agencies have responded by going beyond the old approach of treating all growth that cannot be accounted for by measured factors as a residual measure of the impact of innovation, to more directly measuring innovation and its impact on growth (see, e.g., Corrado, Hulten, and Sichel, 2009; United Nations et al., 2009). This work has been constrained, however, by the data that are available, and much remains to be done.

4. Better measures of the effects of globalization. Offshoring, global value chains, and the globalization of production more generally have become key features of today’s economic landscape. Existing measures of international trade, production, value added, investment, and the distribution of income fail to adequately capture the effects of these developments: they tend to highlight the costs of globalization and fail to measure the large positive contributions of global trade to economic growth, productivity, and standards of living (see, e.g., Houseman and Mandel, 2015).
5. Better and more up-to-date measures of structural change. Technological change, globalization, and other forces have caused significant changes in the structure of the economy in recent decades. Examples include the falling share of U.S. employment in manufacturing, the increasing importance of health care and other services sectors, and the apparent rise of the “gig” economy, such that work seems increasingly to be performed by independent contractors who take on a series of short-term engagements rather than by employees who have a continuing relationship with an employer. Conventional aggregate measures, such as measures of labor and multifactor productivity, may no longer provide consistent measures over time on the state of the economy (see, e.g., Council of Economic Advisers, 2013; Feldstein et al., 2017).
6. Expanded measures of sustainability and welfare. Questions have been raised about how well existing statistics measure the sustainability of economic growth and the changing distribution of the gains from growth. Factors behind these concerns include the apparent slowing in the potential growth of the U.S. economy, climate change, energy prices, the need to replace and expand the aging U.S. infrastructure, growing health care spending, income dependency burdens, and rising inequality. Some of the data gaps revealed by these changes are the need for better measures of in-

5. vestments in infrastructure, productivity, natural resources and the environment, human capital, health care and health care spending, and the distribution of income (see, e.g., Krueger et al., 2009; Landefeld et al., 2010; Stiglitz, Sen, and Fitoussi, 2009).

6. Gaps revealed by the financial crisis. Another key statistical challenge is the need to develop measures that fill the gaps identified by the 2008 financial crisis (see, e.g., International Monetary Fund and Financial Stability Board, 2009, 2017). Policy makers in the United States and abroad did not have adequate data on the risks that the financial system confronted. For the future, better measures are needed to monitor macro-prudential regulation and industry risk.

Together with other federal statistical agencies, the Census Bureau has been taking steps to address these and other data gaps. For example, the Bureau has redesigned what is now called the Business R&D, and Innovation Survey, sponsored by the National Center for Science and Engineering Statistics at the National Science Foundation.⁴ To take another example, although much of the work to fill the data gaps that were revealed by the financial crisis will be the responsibility of the Department of the Treasury and the Federal Reserve Board, the Census Bureau is doing important work to expand coverage of the financial sector consistent with the coverage of other services.

The Census Bureau also can play an important role in addressing other identified data gaps. For example, with respect to the measurement of globalization, better data on company characteristics, such as whether the company’s ownership is foreign or domestic or whether the company is an exporter or a nonexporter, could help a great deal to improve the world and regional input-output tables used to measure value-added trade. The understanding of globalization also could be informed by better data on foreign sourcing by U.S. companies. Supplements to the Census Bureau’s annual economic surveys would be one way (although not the only way) to collect this information. To take another example, the Census Bureau could collect data on contracts with individuals who are not employees or contracts with other businesses for services previously performed in-house, which would help to shed new light on ongoing changes in the structure of economic activity. The Annual Survey of Manufactures collects some of this kind of information, but the other industry sector surveys (retail trade, wholesale trade, and services) do not. Again, a supplement to the Census Bureau’s annual economic surveys would be a possible vehicle for doing this.

___________________

⁴ See https://www.nsf.gov/statistics/srvyindustry [October 2017].

CONCLUSION 2-4: Major gaps in existing economic data identified by data users include measures of innovation, measures of the pattern and effects of globalization, measures of structural change in the economy, measures of sustainability, and measures for the financial sector. Some of these gaps, although not likely all of them, could be addressed in the redesign of the annual economic surveys into an integrated Annual Business Survey System.

2.5 CONSISTENCY ACROSS DOMAINS AND OVER TIME

In spite of the changing global economy, data users want consistent data. This means data that are consistent across geographic areas, across industries, across data products, across agencies, and over time. Consistency across geographic areas and across industries is critical for allowing data users to make meaningful comparisons along those dimensions. Adherence to international standards and guidelines—such as the Harmonized Commodity Description and Coding Systems,⁵ the System of National Accounts 2008 (United Nations et al., 2009), and the Balance of Payments and International Investment Position Manual, Sixth Edition (International Monetary Fund, 2013)—is an important element of data consistency. As outlined in the Guidelines on Integrated Economic Statistics (United Nations, 2013), comparisons between business statistics and macroeconomic data and across countries become difficult if national statistical agencies are not measuring economic activity in the same way.

Consistency in measurement over time is also important. For many data users, economic data are valued primarily because they are informative about trends in the economy. Assessing trends of course requires that that the measurements in question be consistent from one year to the next.

As presently configured, data from the annual economic surveys do not fare well in regard to consistency. As described by Census Bureau staff in presentations and materials provided to the panel, the 11 annual economic surveys reviewed by the panel use different item definitions, reporting units, methodologies, processes, and systems and other infrastructure. Moreover, the data are released with different time lags. The Bureau’s website tries to explain how the data differ among the individual datasets for the various surveys, but these differences, as well as inconsistencies in presentation on the Bureau’s website, make information difficult to access and compare.

Among the specific requests related to data consistency made by participants in the June 2016 meeting, the BEA representative sought greater consistency with international standards and guidelines. The BLS representative

___________________

⁵ See https://unstats.un.org/unsd/tradekb/Knowledgebase/Harmonized-Commodity-Description- and-Coding-Systems-HS [October 2017].

wanted to see data reported in a format that is common across surveys and also comparable to the reporting of data for the economic censuses. The CMS representative specifically expressed a desire for the presentation of data as continuous time series across the annual economic surveys, rather than in separate tables for each year, a point that other participants generally supported.

There is, of course, a tension between the desire for data that evolve as the economy changes and the desire for consistent time series. This tension can be seen, for example, in regard to the suggestion that there should be more frequent revisions to NAICS, as changes to the industry classification structure inevitably imply a certain amount of discontinuity. This kind of discontinuity was a lesson learned during the statistical agencies’ transition from the old Standard Industrial Classification System to the updated NAICS in the late 1990s. While the introduction of NAICS was hailed by data users as a great step forward, the loss of time series and cross-sectional continuity was seen as a glaring implementation problem. However, with careful planning, the impact of these kinds of discontinuities on the data can be minimized. Including plans for the construction of bridge tables, higher-level aggregates, and historical time series resting on a new coding structure will be important in any reengineering of the annual economic surveys.

CONCLUSION 2-5: The Census Bureau’s annual economic surveys lack consistency with respect to definitions, reporting units, methodologies, processes, and systems and other infrastructure. Recognition of this fact was an important part of the motivation for this study. Many of these inconsistencies cause problems for data users. Lack of consistent data over time is another weakness in the current surveys.

2.6 ACCESS TO SURVEY DATA

Data users often operate in high-pressure environments that depend on analysis and insights, not just data, and they often have constrained resources. Consequently, users need access tools that make it easy for them to find the data they need and to manipulate and visualize those data. Current Census Bureau data dissemination and access tools, in particular, American FactFinder,⁶ are viewed by users as hard to work with and inefficient. Many users at the June 2016 panel meeting provided specific suggestions, based on their experiences elsewhere in the public and private sectors, about how to improve users’ data access experience with Census Bureau prod-

___________________

⁶ See https://factfinder.census.gov [July 2018].

ucts. These include suggestions for improving the Census Bureau website and the American FactFinder data access tool (or its replacement), data downloading and manipulation options, better access to metadata, and training and help options such as YouTube tutorials, webinars, live-chat availability, and easy access to subject-matter experts (see Chapter 7 for further discussion).

In recent years, there has been growing interest by researchers in access to the Census Bureau’s microdata. Important analyses have been carried out by researchers working on projects, which have been carefully screened and approved, carried out in a secure environment using microdata from the Business Register and the Annual Survey of Manufactures. The microdata collected as part of a new ABSS likely would be of considerable interest to researchers, especially if those data could be matched with other information and over time (see discussion of creating longitudinal panels in Chapter 5). Recommendations for expanding access to business microdata for research purposes, while protecting confidentiality, lie outside the panel’s charge, but it would be appropriate for the Census Bureau to keep microdata analysis needs in mind as the redesign of the annual economic surveys proceeds.

CONCLUSION 2-6: Many data users seek expanded and improved access to the Census Bureau economic data they use. Users would value both improved data dissemination tools and user-friendly interactive data manipulation tools. Users also would value training and assistance in using these tools, such as through webinars. For a subset of data users, expanded access to microdata, understanding that such access would need to be in a secure environment, is a priority.

2.7 CONSULTATION WITH DATA USERS ON AN ABSS

During the June 2016 panel meeting, the idea of an ABSS was introduced to data users. They responded favorably to this suggestion, seeing the advantages of enhanced consistency, timeliness, and relevance that a well-implemented redesign of the annual economic surveys could provide. Process improvements implemented in conjunction with the introduction of an ABSS also could help the Census Bureau to better meet data users’ demands for accuracy, timeliness, relevance, consistency, and accessibility.

A major redesign such as the introduction of an ABSS inevitably will involve tradeoffs, so that as development proceeds, it will be important to develop mechanisms for obtaining regular user feedback. Such feedback would be useful both for users and for the Census Bureau. The implementation feedback loop could be folded into the agenda of groups that already

exist, such as the Census Bureau’s Scientific Advisory Committee and FESAC, which jointly advises BEA, BLS, and the Census Bureau. Alternatively, an outside organization could organize formal or informal working groups to fill this role and, possibly, to provide input on other components of the Census Bureau’s economic programs. There is precedent for this approach at other statistical agencies—for example, the Energy Information Administration regularly seeks input from a Committee on Energy Statistics organized under the auspices of the American Statistical Association.⁷

CONCLUSION 2-7: Based on a brief description of an Annual Business Survey System (ABSS), the data users consulted by the panel were supportive of the general idea and of the associated goals of improving accuracy, timeliness, relevance, and consistency of annual economic data. As decisions about an ABSS are made, ongoing consultation with users will be important to ensuring that a redesign of the current annual economic surveys meets their needs.

RECOMMENDATION 2-1: The Census Bureau should develop a process for regularly updating the data user community about plans for an Annual Business Survey System (ABSS) that will reengineer the annual economic surveys and obtain feedback from that community about its development. Consultation with users should include explicit consideration of ways to minimize the potential effects of an ABSS on breaks in time series.

2.8 REFERENCES

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

Bernanke, B. (2013). Economic Prospects for the Long Run. Remarks at Bard College, MA. Available: http://www.federalreserve.gov/newsevents/speech/bernanke20130518a.pdf.

Biemer, P., Trewin, D., Bergdahl, H., and Lilli, J. (2014). A system for managing the quality of official statistics. Journal of Official Statistics 30(3), 381–442. doi: https://doi.org/10.2478/jos-2014-0022.

Brackstone, G. (1999). Managing data quality in a statistical agency. Survey Methodology 25(2), 139–149. Available: https://unstats.un.org/unsd/dnss/docViewer.aspx?docID=194.

Cardarelli, R., and Lusinyan, L. (2015). U.S. Total Factor Productivity Slowdown: Evidence from the U.S. States. Working Paper WP/15/116. Washington, DC: International Monetary Fund. Available: https://www.imf.org/external/pubs/ft/wp/2015/wp15116.pdf.

Corrado, C., Hulten, C., and Sichel, D. (2009). Intangible capital and U.S. economic growth. Review of Income and Wealth, 55(3), 661–685.

___________________

⁷ See http://ww2.amstat.org/committees/commdetails.cfm?txtComm=CCNARS03 [October 2017].

Council of Economic Advisers. (2013). The Recent Slowdown in Health Care Cost Growth and the Role of the Affordable Care Act. Available: https://obamawhitehouse.archives.gov/blog/2013/11/20/new-report-council-economic-advisers-recent-slowdown-health-care-cost-growth-and-rol.

Council of Economic Advisers. (2014). The Annual Report of the Council of Economic Advisers. Available: http://www.whitehouse.gov/sites/default/files/docs/full_2014_economic_report_of_the_president.pdf.

Daas, P., Ossen, S., Tennekes, M., Zhang, L.-C., Hendriks, C., Foldal Hauagen, K., Bernardi, A., Cerroni, F., Laitila, T., Wallgren, A., and Wallgren, B. (2011). List of Quality Groups and Indicators Identified for Administrative Data Sources. BLUE-ETS deliverable 4.1. Luxembourg: Eurostat.

Eurostat. (2000). Assessment of the Quality in Statistics. Doc. Eurostat/A4/Quality/00/General/Standard report. Available: http://www.unece.org/fileadmin/DAM/stats/documents/2000/11/metis/crp.3.e.pdf.

Federal Committee on Statistical Methodology. (2013). Data Quality Assessment Tool for Administrative Data. Statistical Policy Working Paper 46. Subcommittee on Statistical Uses of Administrative Records. Washington, DC: U.S. Department of Commerce. Available: https://s3.amazonaws.com/sitesusa/wp-content/uploads/sites/242/2014/04/DataQualityAssessmentTool.pdf.

Feldstein, M., Syverson, C., Groshen, E.L., Moyer, B.C., Aizcorbe, A.M., Bradley, R., and Friedman, D.M. (2017). Symposia, “Are Measures of Growth Biased?” The Journal of Economic Perspectives 31(2, Spring), 145–210.

Gordon, R. (2013). U.S. productivity growth: The slowdown has returned after a temporary revival. International Productivity Monitor 25, 41–47. Available: http://economics.weinberg.northwestern.edu/robert-gordon/files/ConfDisc/USProductivityGrowth.pdf.

Houseman, S.N., and Mandel, M., eds. (2015). Measuring Globalization: Better Trade Statistics for Better Policy. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research. doi: https://doi.org/10.17848/9780880994903.

International Monetary Fund. (2013). Balance of Payments and International Investment Position Manual, Sixth Edition. Available: https://www.imf.org/external/pubs/ft/bop/2007/bopman6.htm.

International Monetary Fund and Financial Stability Board. (2009). The Financial Crisis and Information Gaps. Report to the G-20 Finance Ministers and Central Bank Governors. Available: https://www.imf.org/external/np/g20/pdf/102909.pdf.

International Monetary Fund and Financial Stability Board. (2017). The Financial Crisis and Information Gaps—Second Phase of the G20 Data Gaps Initiative—(DGI1) Second Progress Report. Available: http://www.imf.org/external/np/g20/pdf/2017/092117.pdf.

Jarmin, R. (2017). Operational Excellence: Accelerating the Census Bureau’s Economic Indicators Improves Accuracy, Timeliness of Commerce Data. Washington, DC: U.S. Department of Commerce. Available: https://www.commerce.gov/news/blog/2017/01/operational-excellence-accelerating-census-bureaus-economic-indicators-improves.

Krueger, A.B., Kahneman, D., Schkade, D., Schwarz, N., and Stone, A. (2009). National time accounting: The currency of life. In A.B. Kreuger (Ed.), Measuring the Subjective Well-Being of Nations: National Accounts of Time Use and Well-Being (Ch 1, pp. 9–86). Chicago: University of Chicago Press.

Landefeld, J.S., Moulton, B.R., Platt, J.D., and Villones, S.M. (2010). GDP and beyond—measuring economic progress and sustainability. Survey of Current Business 90(4, April), 12–25. Available: https://bea.gov/scb/toc/0410cont.htm.

Landefeld, J.S., Seskin, E.P., and Fraumeni, B.M. (2008). Taking the pulse of the economy: Measuring GDP. Journal of Economic Perspectives 22(2, Spring), 193–216. Available: https://www.aeaweb.org/articles?id=10.1257/jep.22.2.193.

Stiglitz, J.E., Sen, A., and Fitoussi, J-P. (2009). Report by the Commission on the Measurement of Economic Performance and Social Progress. Available: http://ec.europa.eu/eurostat/documents/118025/118123/Fitoussi+Commission+report.

United Nations. (2013). Guidelines on Integrated Economic Statistics. Statistics Division. Available: https://unstats.un.org/unsd/nationalaccount/docs/IES-Guidelines-e.pdf.

United Nations, European Commission, International Monetary Fund, Organization for Economic Cooperation and Development, and World Bank. (2009). System of National Accounts 2008. Available: https://unstats.un.org/unsd/nationalaccount/docs/SNA2008.pdf.