The Challenges of Producing Economic Data for the 21st Century

Testimony Before the Senate Joint Economic Committee

In his testimony, Keith Hall discusses the economic statistics produced by the federal statistical system. He comments on some of the challenges that the current system is struggling to meet and a handful of specific inadequacies in data coverage.

Congresswoman Maloney, Vice Chairman Brady and Members of the Committee: Thank you for the chance to discuss the economic statistics produced by the federal statistical system. In my testimony, I will talk briefly about some of the challenges that the current system is struggling to meet and then mention a handful of specific inadequacies in data coverage.

Federal economic statistics are important for both policymakers and the public, and the economic data that they represent is a huge bargain for the American people. Such data assist policymakers in the monitoring of the economy and in the development of macroeconomic policy. Beyond business cycle concerns, economic statistics can and should be essential to the development and monitoring of policymaking throughout the federal, state, and local governments, guiding hundreds of billions of dollars in federal spending. For example, the Consumer Price Index is used in determining entitlement payouts, like Social Security benefits, and in setting federal income tax brackets; employment and wage data are used in federal allocations in such programs as the State Children’s Health Insurance Program and Medicaid; and employment cost indexes are used to determine reimbursements under the Medicare Prospective Payment System. For the public, economic data serves the same role as physical infrastructure. Private businesses use statistics to make sales projections, reach investment decisions, adjust contract payments for inflation, and more. And individuals count on reliable economic information to make all kinds of personal economic decisions. Literally millions of people now visit agency websites every month.

The challenges facing federal statistical agencies are significant and many. Like physical infrastructure, statistical systems become obsolete over time. The economy is consistently changing, new industries emerge while old industries restructure and sometimes decline, business practices change, and households change how they make economic decisions. Keeping up the coverage and quality of economic data has been, and is likely to continue to be, constrained by tight budgets and the complexity of data collection and analysis. It has always been a problem that data users often need new information quickly while it takes agencies a long time to design and produce new, high-quality statistics.

In a sense, I know of no economic statistics program that is fully funded. For example, the Bureau of Labor Sta- tistics, or BLS, does not have the best possible data on payroll jobs under the Current Employment Statistics program. Instead, they work to have the best $60 million program possible. With more money, the program could be both improved and expanded, and data users would be thrilled with the result. With less money, agencies need to be free to make sound, professional judgments on how to adjust their programs – for example, which program to cut, whether or not to reduce sample or lower the number of data series, etc. And—as we have seen time and time again—any proposal by a statistical agency to eliminate data causes data users to become upset. Statistical agencies need to be free to discuss the limitations and changing data quality to users. Because survey design, data collection, and data analysis are complex, it is quite easy for data quality to decline without data users’ immediate knowledge. For example, budget issues prevented BLS from updating the housing portion of the Consumer Price Index until 2010. This left a full 40 percent of the index woefully out of date, because it was still based on 1990 census data. We will likely never know how much this led to over- or underestimation of inflation and therefore to over- or underpayments to millions of Social Security recipients.


The nature and scope of economic activities by businesses and households are becoming increasingly complex. The growth of global production chains has sometimes made it hard to categorize companies as either manufacturers or wholesalers. The pace of technological change and product innovation requires constant changes in surveys. As we have just seen in the period running up to the Great Recession, data collectors have had a hard time keeping up with the growth of financial instruments. For households, the greater use of technology in transactions has made it harder for individuals to recall transactions in response to survey questions.

Statistical agencies need to modernize their data collection to better reflect how households and businesses store and use information. Many of the data collection technologies used by federal statistical agencies are outdated and inflexible. For example, far too often data collectors from federal agencies still make personal visits to households and businesses or still collect data by telephone. Not only is this an unnecessary burden to respondents, but it is a very costly mode of data collection. Surveys need to do a better job of accepting electronic records from companies. For example, in the Consumer Price Index program, data collectors still make store visits in which they find and examine products and enter prices in a handheld computer. Instead, there is tremendous potential in collecting this data in electronic form from company headquarters. Once permission is obtained, thousands of transaction prices could be collected at once for sales at hundreds of stores. Research on doing just this is currently underway at BLS. Surveys can also make better use of technology in collecting data in household surveys. Rather than continuing to visit individual households and discuss monthly purchases while sitting in someone’s living room, the Consumer Expenditure Survey program at BLS is researching the use of technology that would allow a household to scan cash register receipts.

Statistical agencies need to improve their use of technology and reduce redundancy in information technology systems. For all agencies that I am familiar with, the development of new systems is both slow and expensive. New business models need to be developed for the delivery of IT systems. For example, the large statistical agencies typically have a number of independent programs, each with its own budget and each with its own independent IT system for data collection and processing. This creates a significant amount of redundancy and raises the overhead cost for agencies. Because these business practices have been in place for decades, they are not easily or cheaply fixed. Similar redundancy can be seen when smaller statistical agencies have their own systems and do not share common IT platforms with each other. The solution, I believe, requires very strong leadership—not only within each agency but across agencies—to move to common platforms and even common data collections and processing systems. This has been done, for example, at Statistics Canada and perhaps at statistical agencies in other countries.

Statistical agencies need to modernize their data dissemination. The data collected and analyzed by statistical agencies are paid for by taxpayers, and the output of these agencies belongs to them. Frankly, agencies seem to sometimes forget that the data are not exclusively for the use of economic policymakers. It is therefore an important part of the mission of each agency to make sure that their information is available to everyone and in an understandable and useable form. This burden has increased in recent years with the decline of newspapers and newspaper coverage of economic data. Fortunately, the Internet has revolutionized data dissemination and analysis, and its potential for data collection is great. However, data users currently have to work much too hard to navigate the statistical system and dozens of independent websites to get information. Agencies need to con- tinue to use the Web and take advantage of new, evolving forms of communication for data dissemination. For example, to my knowledge only the Census Bureau has begun to significantly use the various forms of social media to discuss their agency and their data.


The U.S. economy is huge, complex, and ever-changing. As a result, there are many inadequacies in the statistical data available. I will mention just a few of the larger gaps that I am familiar with.

There is a significant gap in the data coverage of services. For decades, the statistical system focused primar- ily on goods. However, the service sector for many years was larger than the goods sector, and it has also grown faster. This is true for every wealthy economy in the world. In 2007, the service sector was responsible for over 80 percent of total U.S. employment, and it has been responsible for essentially 100 percent of job growth over the past 40 years. There has been significant progress in services coverage, but it is still quite incomplete. The Great Recession was perhaps a good reminder that we need to fix this. In past recessions, job loss was centered on the goods sector of the economy; in fact, the service-sector job loss has often been minimal and occasionally, as with the 2001 recession, there was no service-sector job loss at all. During this recession, for the first time ever, more than half of the job loss has been in services.

There is a significant gap in data coverage of international trade. This particularly centers on trade in services, and this should be a significant concern for the U.S. It is widely recognized that the U.S. has a significant comparative advantage in service activities, particularly relative to developing countries. Yet trade agreements have focused primarily on trade in goods, and there is a strong view held by many trade economists that there are significant untapped markets in developing countries that could be opened up for U.S. services companies. The lack of data on trade in services has almost certainly led to a real lack of research on the potential benefits of liberalizing trade in services. With respect to import prices, budget cuts in recent years have led to lower, rather than higher, cov- erage of services.

A potential data shortcoming that has received a good deal of attention over the past few years is in the quality of data on import prices. There is legitimate concern that import prices have been underestimated. When U.S. companies switch purchases from a U.S. producer to an imported intermediate product, they often do so for lower prices. If the full decline in prices is not captured by the import price program, then import quantities are under- estimated and real GDP, which focuses only on domestic production, will be overestimated. Similarly, productivity in manufacturing will appear higher than it really is. At the moment, the solution appears to be the development of a new survey that focuses on the prices that companies pay for intermediate products from either domestic or foreign companies.

There are a number of shortcomings in the measurement of consumer prices. In fact, the Consumer Price Index as it currently exists is a bare-bones measure of consumer prices. Prices are collected for a single, average bundle of goods and services. This mix is an effort to represent the average for the entire U.S. However, since different groups of people, like the elderly for example, have different consumption patterns than other groups do, this index can be misleading. In the case of the elderly, tens of billions of dollars of Social Security benefits are allocated based on cost of living adjustments that do not necessarily represent the mix of goods and services that older people consume. Also, data is only collected in urban areas, despite the fact that there can be significant differences in prices in rural areas. And last, there is insufficient data collected for cost of living estimates at the city, state, or regional level. This limitation creates limitations for other economic data. For example, the census measurement of welfare is not adjusted for cost of living differences in different areas of the country.

And, last, I want to mention difficulties with the unemployment rate as a measure of labor market slack. Although the unemployment rate is consistent with an international standard that is followed by most government statistical agencies, it has some long-recognized limitations that have made it one of the most widely criticized economic statistics in the world. Its design follows three basic concepts:

  1. People with jobs are employed.

  2. People who are currently jobless, actively looking for jobs, and available for work are unemployed.

  3. People who are neither employed nor unemployed are not in the labor force.

The labor force statistics are intended as a measure of the current supply of labor, so defining the unemployment rate as the share of the labor force without work gives us, in principle, a measure of how much supply exceeds demand for labor. During a recession, demand for labor declines as economic activity declines, leaving labor market slack. Under these circumstances, there is no real reason for the supply of labor to diminish. In fact, if anything, we would expect an increase in labor supply as incomes fall. Because the labor force is defined as those currently and actively looking for work, when the unemployed become discouraged and decide to stop actively looking for work until the economy improves, the supply of labor appears to decline, reducing the unemployment rate. Similarly, coming out of a recession, when the economic news improves, more jobless become active in their job search, and the labor supply appears to increase. A better measure of labor supply, and therefore of labor slack, would not change through the business cycle.