This essay was prepared for the U.S. Congress, Joint Economic Committee, in October 1997. It is designed to highlight the problems, challenges, and opportunities for improving the nation's economic statistics in a manner useful to policymakers. It is not designed to be a detailed technical road map for the statistical agencies. The measurement issues discussed herein have been one of my professional interests throughout my career. The views expressed in this essay are entirely my own, although they have been heavily influenced by conversations with people, too numerous to mention individually here, in the federal statistical agencies, academia, and the private sector. I would like to mention, however, the many people with whom I worked during my service both as chairman of the Cabinet Working Group on Economic Statistics when I was chairman of the Council of Economic Advisers and as chairman of the CPI Advisory Commission.
Introduction and Overview
The United States' statistical system is one of the finest in the world, staffed by quality professionals. It continuously calculates, produces, and disseminates a vast amount of information about the economy. This information--the core federal government economic statistics--complements a vast array of private sector–generated information. Together this forms our economic informational infrastructure, on which millions of decisions are based every day. Private economic activity, from production and inventory schedules to wage negotiations and private capital budgeting, on the one hand, and federal budget policy (including cost-of-living allowances and monetary policy designed to keep inflation low and stable), on the other, all rely on timely, accurate economic information from the federal statistical system.
Although the rapid pace of change in the structure of the U.S. and world economy is straining the ability of the statistical system to keep up and measure economic performance accurately, and despite the need for improvement, the U.S. statistical system properly enjoys widespread public support and a well-deserved reputation for professionalism and integrity. The statistical systems in some developing countries are considered to be less reliable and more subject to political influence,
reminding us that we should not take the public confidence in the statistical system for granted. That confidence is a valuable social asset and should not be squandered. Although there are suggestions in this report for improvement, they must be done in the context of maintaining this vital social asset.
The current U.S. statistical system is heavily decentralized.1 Some countries have a single statistical agency, usually independent of cabinet departments, that is responsible for the various statistics. This model of organization--Statistics Canada is a prime example--has a variety of advantages and disadvantages relative to our decentralized system, some of which are discussed below. Our statistical system grew up in part by historical accident. As different types of statistics were needed for different purposes, bureaus or agencies were developed to gather data, produce information, disseminate it to the public, oversee the gathering of the information, and provide policy overview and oversight. For example, the consumer price index (CPI) program was initiated early in this century to index the wages of shipyard workers around the time of World War I. Since then, of course, the CPI program in the Bureau of Labor Statistics (BLS) in the Labor Department has evolved, grown, and periodically improved. This particular program, and the statistical system in general, should not be thought of as a static snapshot but rather as an evolving system, a moving target if you will. The statistical agencies continually try to improve their methods, procedures, analyses, and so on. And the various statistics generally published today are based on very different methodologies, and sometimes concepts, than was the case at initiation, often decades ago. In fact, the federal statistical system's major programs will be somewhat different a few years from now than currently, in what is produced, how it is produced, and the concepts being measured.
For example, the BLS has several important improvements and updates under way in the consumer price index program, such as the 1998 decadal revision, which will update the weights, among other things, used for the "market basket" and the consideration of methods to reduce so-called substitution bias. The Bureau of Economic Analysis in the Commerce Department has made major numerous improvements in recent years, including changing from focusing on gross national product (GNP) to gross domestic product (GDP), chain weighting as opposed to using fixed weights, using more advanced and technically improved price and output indexes (the so-called Fisher Ideal Index),2 moving to the system of national accounts (SNA), and initiating work on a variety of satellite accounts.
The major statistical agencies are the Bureau of Labor Statistics in the Labor Department, which has primary responsibility for consumer and producer price information, employment, unemployment, and other labor market data, such as average earnings, to name the most important and widely used; the Census Bureau, which, in addition to the decadal census, collects monthly surveys that provide not only information on a national sample of tens of thousands of households that infers employment, unemployment, labor force participation, and numerous other data but also information on consumer expenditure patterns and important demographic information; and the Bureau of Economic Analysis (BEA) in the Commerce Department, which is responsible for the National Income and Product Accounts (NIPAs) and related data on international transactions. The NIPAs are the core accounts for the national economy. The gross domestic product and its decomposition into real and inflation components, consumption, investment, government expenditures, and exports less imports and detailed information within these subcategories are provided from piecing together large quantities of specific source data, some from the government, some from the private sector.
Other government agencies produce important and useful information, but I will not dwell in detail on them here. The Federal Reserve (nominally private but quasi public) produces data on industrial production and capacity utilization, various measures of the money supply, measures of the flows of funds, and national balance sheets for the economy. Of course, the Federal Reserve is one of the most important consumers of government economic information as it sets monetary policy. The National Agricultural Statistical Service collects and presents substantial data on the agricultural sector of the economy; the National Science Foundation provides information on science and technology indicators. Further, the agencies charged with producing the core economic statistics--the BLS, the BEA, and the Census Bureau--are also charged with producing other data and information, reflecting other priorities.
The current decentralized system has important advantages, including decentralization for efficiency and effectiveness (the usual economist's predilection) and some competition among the statistical agencies and between them and the private sector, which may generate incentives for greater efficiency and innovation. Virtually all observers agree, however, that the current system faces severe financial, human, and technological constraints that impede improving the quality of economic statistics.
In the last fiscal year, congressional appropriations for statistical agencies amounted to somewhat more than $1 billion, approximately half in the agencies with the largest budgets, the BLS and the Census Bureau. Additionally, the Bureau of Economic Analysis in the Commerce Department has a budget of about $40 million; the National Agricultural Statistical Service, a little more than $80 million; and several other agencies, which have information correlative to or closely related to economic statistics such as the National Center for Educational Statistics, the Energy Information Administration, the National Center for Health Statistics, the Economic Research Service, the Statistics of Income Division of the Internal Revenue Service, the Bureau of Transportation Statistics, and the Bureau of Justice Statistics, all have budgets in the tens of millions of dollars. The work of the different statistics agencies is loosely coordinated by the Office of Management and Budget through the statistical policy branch. Further, statistics are also gathered in other bureaus and agencies that are part of the ongoing functions of the individual agencies.
The amount appropriated for the individual agencies determines their ability to continue to do their current job in an adequate way, not to make major improvements in the quality of the statistics. The agencies have little budget for innovation: intellectual capital, tangible capital, and infrastructure. At a time when most private enterprises in the United States are spending substantially on information technology--hardware, software, and human resources--the modest budgets of the statistical agencies call out for those same resources. (Indeed, many statistical agencies are constrained in their programs to generate traditional statistics in the traditional way on the usual time schedule. It is remarkable that they have been able to make any improvements at all.)
The human capital constraints on the statistical system do not reflect just the aggregate budgets, although the budgets do limit the number of, and promotion to, high-paying skilled jobs competing with analogous ones in the private sector. Much of the knowledge learned on the job in the statistical agencies is what economists call specific human capital. Innumerable technical issues are confined to a particular data series or program. Although some of this human capital is transferable, thereby making the statistician attractive to other employers, that substantial specific human capital is not likely to enhance earnings prospects.
At the entry level, a related phenomenon is the decline in the proclivity to teach the statistical skills and techniques useful in data gathering and dissemination. Both the economics and statistics professions have become more theoretical, spending less time on the practical issues of sampling, data collection, quality of data, and providing professional rewards for those who show great skill in finding, developing, or improving data.
Finally, there are technological constraints on the statistical agencies. Partly because new technology is often expensive in the short run, even though it may ultimately reduce costs considerably in the long term, technological improvements have been modest in the statistical agencies. Some important improvements have been made, including part of the so-called Boskin initiative.3 More electronic reporting and filing of information from data sources is one example, decreasing both the time and the cost to the private sector for supplying the data and, at least in the long run, to the statistical agencies for collecting, processing, and disseminating them. Because the private sector has moved rapidly technologically--for example, using scanners in many types of retail environments, in addition to their own infrastructure hardware and software needs--it may eventually become possible for the statistical agencies to obtain some data from private sources that are reliable, less expensive, and confidential. Whether major expansion of private data collection activities will replace government data collection to any significant degree is an open question. Among the many issues are the confidence, trust, and reliability that private data providers have in government statistical data collectors. For example, some of the information firms provide to the government would be highly valuable to their competitors, and it is unclear that firms would be willing to voluntarily provide such important economic information to a nongovernmental source.
In summary, our decentralized statistical system has advantages and disadvantages relative to a centralized one but is operating under substantial financial, human, and technological constraints. Our statistical system, despite numerous and continuing improvements, is falling short in a variety of areas,4 which are discussed in some detail below. The first question is, so what? Is it that important to improve the quality of the statistics? Can it be done at modest expense? What will be the benefits of doing so? How accurate do these numbers really need to be?
Should we be satisfied with inaccurate economic statistics? Actually, if the estimates are fairly close and obtaining better measures is enormously expensive, perhaps so. After all, Newtonian physics got us to the moon; we did not need relativity. Most of the time, when the economy is operating in pretty good shape, small errors in the economic statistics would be unlikely to give rise to a huge mistake. For example, it is unlikely that small errors in inventory data would give rise to a miscalculation by the Federal Reserve of the probability of a recession. But a large error in exactly this data series did give the wrong signals in the mid-1970s, leading to a long delay in recognizing how severe the downturn was and perhaps to a monetary policy that was less than optimal. Some economic statistics have great ramifications, even if they are only off a little. Inflation, or the change in the cost of living, is difficult to measure in a complex dynamic market economy. But if the inflation measure is off by what would be thought of as a little in terms of macroeconomic consequences, the consequences for the federal budget compounded over a few years would be enormous because one-third of the federal budget is escalated by changes in the consumer price index, as are the tax brackets. In this case even a small error is consequential, unless it is almost sure to be reversed quickly. Even small or modest errors, if they cumulate and compound, can make an immense difference to private decision making, the federal budget, monetary policy, and citizens' understanding of economic reality.5
To understand where we need to improve, it is useful to go back to the some of the most basic concepts in economics--the price and quantity of products, in an industry or sector of the economy, and the entire economy and the circular flow of income and product that generates the revenue out of which wages, interest, and dividends are paid. Each of these basic economic concepts might appear easy to measure accurately and track through time, for we are all used to a supply and demand diagram defined over a homogeneous product and period of time, with a unique price for the product characterizing the equilibrium of the market. Many issues arise, however, not the least of which is that defining many types of products is becoming harder and harder in the information age. Indeed, there are 1.7 million uniform product codes for bar scanners; a single WalMart store might hold forty thousand separately priced items, and a typical large supermarket, thirty thousand. In a given year, that WalMart might introduce a thousand or more new products; many others may leave, never to reappear; and a much larger number may come in new forms, shapes, sizes, colors, and other characteristics that purport to improve quality and sometimes do. In this context--when some products change frequently, when an enormous number of new products is introduced all the time, and when the quality of existing products is changing rapidly--the apparently simple act of defining the product is complex. A larger and larger number of such products, and economic activity in general, are more and more difficult to measure. When a large fraction of production and consumption was agriculture and food, those problems, while conceptually similar, were not nearly as severe. In an economy that produces mostly goods such as tons of steel and bushels of wheat, and in which a large part of output is produced by a relatively modest number of firms, economic activity is relatively easier to measure and information relatively easier to come by. But in an economy in which a majority of output is services, where there is far less concentration of firms and quality change is endemic, measurement issues are acute. Indeed, it is likely that we are relatively better at measuring the more traditional sectors of the economy than those that are evolving rapidly. This has profound ramifications for public policy and for the allocation of statistical agency resources in the information age. A strong case could be made that funds need to be provided, or freed up from other use, for more conceptual work, more research and development, so to speak, on better measures of the hard-to-measure, conceptually difficult, rapidly evolving sectors of the economy and that fewer resources, if necessary, be allocated to the data on easier-to-measure traditional sectors of the economy.
The next section discusses some of the major issues in the quality of economic statistics--especially those engendered by the rapid change in the economy--followed by sections discussing making various measurements more closely approximate the economic concepts to which they apply, what statistics will be affected and what qualitatively is likely to result therefrom, generic recommendations, and a conclusion.
The Importance of Economic Statistics in the Information Age
Timely and accurate economic information has always been important. But because technology, especially information technology, may be making patterns of work, production, inventory control, and investment more time sensitive and more condensed, more "just in time," timely and accurate economic information is perhaps becoming even more important. As mentioned above, private economic decisions often rely on various types of federal government (and private) economic statistics both to form a basic background of analysis on which decisions are made and to provide information on changing trends, on which specific decisions often are based. This applies to firms and households in the private sector and economic decisions ranging from mutual fund portfolio investment of households, job and consumption decisions of households, business firms' production and inventory decisions, and capital spending decisions, to name but a few. Of course, these decisions drive the economy.
Other than the daily spectacle of the stock and bond market responding to each economic statistical release, we are perhaps most familiar with the potential impact of economic statistics on government policy decisions and programs. The Federal Reserve clearly analyzes, evaluates, and uses economic statistics, particularly those on inflation and variables potentially correlated with future movements in inflation. Most economists would agree that Federal Reserve monetary policy exerts a substantial influence on the economy and that even small improvements in monetary policy timing or direction have aggregate economic impacts that are consequential.
Budget policy is also driven by economic statistics. First, in the formulation of the budget in the Executive Branch (and its early evaluations in Congress), economic forecasts and projections, developed by the troika (Council of Economic Advisers, Treasury, and Office of Management and Budget) in the Executive Branch and the Congressional Budget Office for Congress, are the basic underpinnings of budget projections. The level, growth rate, and composition of spending and revenues depend heavily on the growth of real GDP, inflation, interest rates, employment and unemployment, and the composition of aggregate income between wages and profits. These forecasts, in turn, are based partly on historical patterns and trends and partly on recent information contained in the most current economic statistics.
Budget policy is affected by economic statistics in a second way: many programs and tax code features are tied to economic statistics. The most obvious are the cost-of-living allowances (COLAs) tied to the CPI for Social Security, tax brackets, and so on. But other statistics matter as well (e.g., grants to states and localities) and are often based in part on them.
Regulatory policy and trade policy are also heavily driven by economic statistics. For example, numerous bilateral trade disputes are based on, and made more contentious by, the fact or perception of widening bilateral trade deficits. Of course, economists generally believe that, although unfair trade practices should be eliminated, they are only a small contributor to bilateral trade deficits and that, in any event, bilateral trade deficits are far less important to the nation's economic condition than the multilateral position of the current account balance. However, because trade imbalances tend to occur in specific sectors and products that affect different regions, and therefore producers and workers, disproportionately serious attention is paid to them and perceived or actual unfair practices. But the trade statistics have serious flaws. In recent years, for example, Canada has claimed our exports to it vastly exceed our official statistics; economists believe that our official statistics overstate and China's understate the bilateral deficit between the two countries. To put matters in perspective, adding up the current account positions of all the countries in the world should produce a balance of zero since what one country sells abroad, another buys, and double-entry bookkeeping should ensure a zero balance. Of course, the statistics are not entirely accurate; the entire world trading system is running a large deficit. I leave it to others to infer with whom this deficit occurs. In any event, both the bilateral trade negotiations, which can spill over into foreign policy and other considerations with many countries, and the regional and multilateral trade arrangements are heavily affected by the trade statistics. It is becoming more and more difficult to measure trade accurately, as discussed below.
Regulation policy is also heavily affected by various economic statistics, especially those concerning particular industries. In this information age, it is becoming increasing difficult to monitor more and more products, which are increasingly difficult to measure in terms of prices and costs. But, like most economists, I believe that deregulation has generally been a major success and that the benefits from deregulation dwarf the problems in obtaining accurate data with which to continue regulation in the few areas where it is still desirable. Deregulation has led, however, to a substantial reduction in the availability of data on previously regulated industries from regulatory bodies.
The most important reason to improve the quality of economic statistics in the information age is to provide citizens with an accurate picture of where we are, where we are going, and how we compare with other nations. Our most fundamental measure of changes in the standard of living is becoming less accurate because more of the economy is becoming harder to measure. We are systematically and gradually getting further away from economic reality relative to our own history and relative to countries whose economies are evolving less rapidly than ours. An extremely inaccurate picture of the U.S. economy could emerge that would substantially affect citizens' beliefs about the performance of the economy and their desires to change various aspects of it.
That an increasing share of the population is aware of, and affected by, economic statistics is indicated by the increasing dollar volume and share of the population participating in financial markets (e.g., in mutual funds). Variable-rate loans raise the pocketbook impact of economic statistics because so many financial factors affecting households are tied to economic statistics. Not just private economic factors affect an increasing number of households; an increasing dollar volume, and share of the population, is affected by federal tax and expenditure programs that are indexed to economic statistics such as changes in the consumer price index. This is true of about one-third of federal budget outlays and the tax brackets, to name the most prominent examples. An increasing dollar volume of Medicaid and other federal grants to states and localities is allocated based on regional economic statistics, among other factors. And, finally, there is increasing access and exposure to economic statistics in the print and television media--particularly the proliferation of business programs on cable television and business magazines and trade journals--and the rapid growth of such information and analysis on the Internet.
The Information, or "New," Economy and Some of Its Implications
Increasingly we hear of a "new" economy. However, a complex, dynamic, flexible market economy, of which the United States' is the biggest and best performing, is always undergoing substantial change: people enter and leave the labor force, new firms are created and others fail, new products and processes are generated, quality is improved in others, and incomes generally rise. Broad social trends interact with this economic performance to produce other structural changes, along with those wrought by factors such as technology and globalization, such as the increased labor force participation of married women in the last quarter century. But if we just take a few economic trends that have either continued, accelerated, or at least begun in recent years and are often thought to be either part of the information age, or greatly affected by it, among those that a serious economist would draw attention to are the shifts
- From domestic markets to global markets
- From monopolies or regulated duopolies and oligopolies to competition
- From wages being the almost exclusive form of labor compensation to greater importance of fringe benefits and to ownership in companies
- Of investment from plant to equipment and increasingly to intellectual property
- From a single job (always overstated historically), career, or skill to mobility and the apparent importance of lifelong learning on the job
Likewise, there have been important policy trends and intellectual paradigm shifts both in the United States and globally. These include shifts from
- Central planning, state ownership, and industrial policy to privatization and markets
- Regulation to deregulation or regulation with market incentives
- The notion that higher inflation can permanently reduce unemployment and that the cost of inflation is low to the idea that stable low inflation is an important pillar of maximizing long- term economic growth
- The idea that high tax rates do not damage the economy by discouraging productive economic behavior to the widespread acceptance that low or at least modest tax rates are economically desirable
- Tariff and nontariff barriers to more open markets and free trade agreements
- Deficit finance to attempts to balance the budget, at least during prosperous peacetime
The Quality of Economic Statistics
The U.S. statistical system is among the finest in the world, with a dedicated staff of professionals. Our statistical system is good but not good enough. It needs to be, and can be, improved. The statistical agencies themselves have recognized this for many years and have made, and continue to make, improvements they deem desirable and for which they have the resources. The agencies themselves, however, are severely limited in both what they can do and what they dare to propose, given severe financial, human resource, and capital constraints. Further, there is a large gap between the economic statistics that are being produced and disseminated and what is actually going on in the economy (at least it appears that this is the case). The problem is more the rapid change in the economy than problems in the statistical system per se. Among the many problems in the rapidly changing economy from the standpoint of the statistical system (although I should emphasize that almost every one of these is the flip side of a coin that produces immense benefits for the vast majority of Americans) are the following.
The Growth of Hard-to-Measure Services
The American economy has produced more services than goods for more than fifty years, and the percentage continues to grow. The growth of services primarily reflects the shift in demand as people become richer, live longer and as technology enables us to provide those services, such as greatly improved medical care. Many services are more difficult to measure than traditional goods. Perhaps the most difficult problems revolve around health care, where we tend to count procedures or measure outlays, rather than define specific outcomes and measure and price these accordingly. When arthroscopic surgery replaces invasive surgery, or medication replaces the need for surgery, it greatly improves the quality and lowers the real price of repairing a knee or ulcer but is beyond the scope of the current statistical system to measure in a timely and accurate fashion.
It is not just the generic growth of services, but, as discussed above, the rapid introduction of new products. To be sure, many of the new products are similar to the old ones; an even larger number is marketed as new when they contain virtually no substantive change at all. But many important new products are introduced not only every year but every month. To take a few examples in the last few decades, cellular telephone service has freed up some segments of the population from spatial confines, so that, for example, working mothers can keep in touch with an aged parent or a sick child as they commute to work. VCRs have greatly enhanced the range of options for movie viewing and shifting entertainment and informational sources to accommodate the time schedules of different households; microwave ovens have greatly assisted in reducing the time necessary to prepare meals, which has been enormously important, especially in an age when the fraction of two-income couples has risen substantially; the personal computer has revolutionized many business and household processes and procedures. The Internet provides new forms of communication, such as E-mail, for example, for parents to keep in touch with children away at college and so forth.
But we do not do a very good job of identifying, measuring, and estimating the value and pricing of new products as they enter our economy. A substantial fraction of the improvements in standards of living have likely come by the generation of new products over time, not just by producing and consuming more of what we have always produced and consumed.6 Otherwise, we would all be spending radically higher fractions of our incomes on bread and potatoes. Obviously, new products are not new to the last decade or two or to the information age. But it appears that the introduction of new products has become a steady, if not perfectly predictable, feature of the information age and that new ways to locate them, estimate their value, and price them must be a major priority for our statistical system.
Some products improve, others may deteriorate, but, on balance, the overwhelming bulk of products either experiences quality improvements or stays the same.7 Improvements in energy efficiency, safety, durability, and so forth are the hallmark of a large number of products, especially consumer electronics.8 These improvements appear to be systematically underreported in our statistical system despite the attempts of statisticians to keep up with the explosive pace of quality change.
Technology and Innovation
Much of the new products and quality change comes from technology and innovation, and the technology moves so rapidly that in some sectors of the economy it is impossible for our statistical agencies, which are geared to decadal revisions, biennial updates, and so on, fully to keep up. If it is true that this is a source of a large fraction of improvements in standards of living in the United States, far greater attention needs to be paid to this sector. Correspondingly, the technology and innovation make obsolete not only the existing products we are counting or pricing but also some of the methods we use to collect information on them. The technology and innovation may mean that some products are distributed in a way that we don't measure, such as Internet commerce.
The transition toward a knowledge-based economy, particularly scientific and technological, knowledge playing a larger role throughout the economy, is a hallmark of the U.S. economy. Simultaneously, economists have focused increasingly on technology, human capital, R&D, innovation, and ideas to understand contemporary economic growth. The measurements of these factors and activities are incomplete, at times primitive, at best piecemeal. Renewed attention must be given to improving science and technology indicators, human capital flows and stocks, and the like.
The changing time use of households is also enormously important. We have episodic survey information on this, but how, why, and when households spend their time doing what is of enormous importance in evaluating economic performance. The amount of hours devoted to work, child rearing, shopping and so forth changes within the economic environment. For example, the widespread trend in retailing in recent years toward malls and discount outlets leads to lower prices and perhaps also some decline in correlative services. Much more, more frequent, and better data on household uses of time are enormously important, especially in evaluating health care outcomes. An expensive procedure that saves a week in the hospital, three weeks lost work, and a year of rehabilitation needs to be evaluated appropriately. Unless substantial information of this sort is integrated into our economic statistics, we could be led sorely astray. Equally important is expanding the information on worker mobility and job transition.
The Growth of International Trade
The growth and changing mix of international trade are also important. The foreign trade penetration, or average propensity to export or import, of the U.S. economy has doubled in recent decades. This has been a source of growth for the United States and the world, as trade is a large positive sum game. But the decline of tariffs and nontariff barriers has decreased the availability of statistics that were more readily available when international trade was more heavily regulated.
One great success of the U.S. economy is an environment that generates many new firms and jobs. The new firms are important not only because some of them become large and produce important products and employment opportunities but also because they provide competition for existing firms, thereby forcing existing firms to innovate and lower costs so that consumers benefit from lower prices. Our information about the "birth and death" of firms is spotty at best.
Financial Innovation and Changing Payment Methods
Financial innovation and changing methods of payment make it difficult to track transactions in the economy. As Internet commerce becomes widespread, it will become easier to track such transactions, but there are numerous other examples.
Changes in the Organization of Production and Distribution
The changing organization of production and distribution, especially with respect to time, space, and required physical inputs, makes measurement difficult. People are spared the necessity of physically having to go to, say, a bank or other financial institution to conduct financial transactions, but the same is increasingly true of other forms of economic activity, and these are much more likely to fall through the cracks in our measurement system.
The rapid changes in the economy have also led to difficulties in classifying and defining sectors, industries, and commodities so as to provide an up-to-date and relevant picture of economic activity to public and private decision makers. Input/output tables are generated infrequently in the U.S. economy. Indeed, Japanese trade negotiators appeared to have more up-to-date input/output tables on the American economy than did their American counterparts. Our ability to make these kinds of classifications is enormously important. For example, the changing nature of compensation described above has occurred for some workers, so we need to update our statistical systems to reflect these broader measures of compensation. Data collected only for production workers should be collected for all workers because, among other things, production workers are a decreasing fraction of the total workforce. The classification of investment versus consumption is particularly worrisome since the nature of investment has changed so much over time. For example, software is the largest investment for a growing number of American firms, but it only counts as part of investment if it is preinstalled in the hardware, a judgment that was made decades ago, when most software was preinstalled in mainframe hardware. Although that is no longer the case, it appears that, from this source alone, about 1 percent of GDP of investment is not being counted annually. Likewise, because conceptual and knowledge-based goods, and services are harder to measure than traditional goods, and because of the increasing importance of human capital in our economy, attention must be focused on measuring this factor of production.
What Statistics Will Be Affected?
Improving the quality of economic statistics in the information age requires a program that will change the way we collect, aggregate, and disseminate our economic statistics (as mentioned above). It requires R&D on what are products? what are markets? what is investment? and so on, as well as major new initiatives in identifying, valuing, and pricing new products, new firms, and quality change. If this is carried out over a span of several years in a coordinated, high-quality fashion, the following results are likely: Real gross domestic product (GDP) and its growth will be affected considerably. Some of what is now being called price increases is really quality change, and the overstatement of price increases9 in consumer prices, as well as in other segments of GDP, feeds into the national income accounts, attributing too much nominal consumption and nominal GDP to prices and not enough to real output.10
Correspondingly, since productivity is output per worker, if real output and its growth are understated, then the level and growth rate of productivity are being understated as well. There are, of course, many other issues of measuring productivity beyond just the price deflators.
As identified by the CPI Commission and many others, the price level is rising more slowly than the official inflation measures, especially the CPI, suggest. Thus, the price level and the rate of inflation are likely to be revised downward. The CPI Commission estimated an upward bias in consumer inflation using a cost-of-living concept of a little more than 1 percentage point a year.11 Part of this bias will be eliminated with the successful completion of BLS initiatives under way or proposed.
Everything that is "deflated" (i.e., adjusted for inflation) will change as well: real wages, real median family income, poverty rates, real GDP, and so on.12 Although each of these data series has other issues and problems, some perhaps working in the opposite direction, the deflation by an overstated price index leads the deflated series to be understated. But many other dimensions of these statistical series can be, and need to be, improved.
As noted above, the composition of GDP between investment and consumption--the most fundamental measure of what society is putting aside to make itself more productive and wealthier in the future--will be affected with improved classifications (e.g., software not preinstalled in hardware as investment rather than consumption). The measures of the factors of production, capital, labor, human capital, and other inputs will be improved as well with more-comprehensive, timely, better, more-conceptually accurate data series, although, again, let me reiterate that many improvements have been made, for example, the changes in the calculation of capital input by the BEA. Combined with time use and other studies and data following the same households over time, we will perhaps gain a better insight into labor markets than the snapshots we currently get from the employment and unemployment surveys.
These are the core economic statistics; I do not mean to ignore others, for example, those on international trade and other sectors of the economy, from health to transportation to technology to science to education. Indeed, they are sometimes the feedstock on which some other data are developed. But as a general proposition, the core economic statistics can and should be greatly improved.
Among the procedures useful here will be updating census and BLS sample frames and enlarged survey instruments to capture new industries and products, births and deaths of new companies, changes in accounting practices, changes in methods of payment, and shifts in the organization of the production, distribution, and consumption of goods and services, including offshore consumption. The Consumer Expenditure Survey (CES) is the most urgent of several priorities.
Our existing measures focus on investment in physical plant and equipment. The increasing importance of computer software, research and development, and other intellectual property--as well as the increasing importance of our natural and environmental resources--illustrate how out of date the existing measures can be.
As discussed above, the increasing rate of technological change, obsolescence, and new product introductions in such rapidly growing areas as computers and high-tech products exacerbates existing problems in measuring depreciation and the value of capital.
The increasing international nature of trade and finance, the integration of trade and financial markets, technological innovation, the increasing volume of trade accounted for by transactions between multinational companies and their overseas subsidiaries, and the increasing volume of direct transactions between U.S. and foreign residents that bypass traditional channels have made it increasingly problematic to collect data purely from domestic sources. The proliferation of new financial institutions and new goods and services require updating sample frames and survey instruments to cover international transactions in such newly emerging categories as financial derivatives, computer software, and credit card services. All of these issues are a product of, affected by, or in turn affect what has come to be called the information age or information technology. But information technology can not only generally improve dissemination and production but also improve that of the economic statistics. For example, investments in information technology have already allowed the public greatly increased and more timely access to economic statistics through the Internet and other electronic gateways. Further investments in information technology can be used to reduce respondents burdens in data collection through electronic filing and to increase the accuracy, efficiency, and timeliness of data processing and transfer.
The primary recommendation is to develop organizational, substantive, financial resources and procedures to move the statistical system to approximate more closely the underlying economic concepts more often, more rapidly, and more continuously.13 For example,
- New products need to be brought into price indexes and measures of real output on a more timely basis; the prices need to be measured before they fall rapidly, which is typical of the early stages of the product cycle, and value must be assigned to the new products for which there is a sizable consumer surplus due to their introduction.14 Surveys need to be done more frequently, and a broader mix of goods and services must be sampled than is already done, to get the new goods in as quickly as possible and to measure consumption patterns and their distribution more accurately. As noted above, expanding the CES is a major priority.
- The issue of quality change, which has been around for a long time, was mentioned in the Stigler Report, the CPI Commission Report and was the lead item in the Boskin Initiative in 1989. Simply put, much of our data comes as the product of price and output--as revenue and expenditure. We have to separate the price and quantity measure; over time, if quality is improving or deteriorating, we must account for that. Thus, if something lasts twice as long and costs 10 percent more, we should count that as a large, real-quality adjusted price decline and real output increase. There are a number of methods for doing so, the most widely accepted of which is called hedonic price measurement and quality adjustment.15
- We need to get closer to actual transactions. Perhaps scanner data will provide a major improvement. We need to get our concepts straight; too often we measure inputs rather than outputs. In some industries, for example, output is measured by input costs, so productivity growth is automatically zero.
- There is usually a trade-off between timeliness and accuracy, as more data become available later and new studies are done shedding light on how to improve the data. For example, in revising the national income and product accounts, many products are cycled between statistical revisions and thus missed completely. Revisions or updates traditionally done every few years or decades may need to be done more frequently. Agencies must have a continuous update arm to complement their core data collection and analysis.
- Consistency across statistics, agencies, and the private sector should be an important goal. For many reasons, complete consistency will be impossible and, in some cases, too costly to attain. But because we have similar concepts being measured in different programs--sometimes within the same agencies, sometimes across agencies--we should do what we can to reconcile the differences. That may result in a single series or in parallel series with more than just a few technical experts being aware of what those differences are.
- Private versus public collection and dissemination of data must be reevaluated. Some movement has been made in this regard, for example, privatizing the leading indicators series, but it is more than just what is government doing that could be done by the private sector as or more efficiently or effectively. The public statistical agencies can learn from private sector methods, and if we really do believe in private sector methods, we ought to be prepared to implement them in the public sector, even when they involve some short-run investment, for example, in information technology.
- We need to encourage thinking on organizational issues out of the traditional box and moving more rapidly. Although most statistical agencies have some research divisions, all of which have made important contributions to improving the statistics, far more needs to be done in this arena. This is not just an issue of larger research budgets for the agencies or academe and more interaction among the private sector, academe, and government statistical agencies, although that is desirable. But research into what is actually going on in the economy, and the implications for the statistical agencies, needs to be a high priority financially, organizationally, substantively, and in the leadership of the agencies. Finally, serious attention must be paid to the pros and cons of reconstituting the statistical agencies in a Statistics America,16 an independent agency headed by a high-level chief statistician that would merge the functions of the various agencies, achieve whatever scale economies and organizational efficiencies could eventually be attained, improve consistency, data sharing, and so on, and minimize the risks of losing competition and other benefits of decentralization.
- All statistical agencies need to completely review the adequacy, or gap, of various data series as they relate to the underlying concepts being measured.
- Sampling procedures, time frames, and so on need to be reassessed as necessary, especially to accommodate the rapid change in the economy by sampling more items--products, outlets, households, firms, or whatever--more frequently.
- A permanent external mechanism for evaluating the quality of statistics should be established. Members should serve on a staggered, rotating basis. This could be done by an independent professional entity such as the American Economic Association, the National Academy of Sciences, or the National Bureau of Economic Research, together with significant resource commitment.
- An expanded R&D budget should develop new methods and procedures beyond the current framework of each of the data programs and each of the agencies.
The return to improved economic statistics--whether better early signals for monetary policy, which might assist the Federal Reserve in meeting its policy objectives, more accurate budgetary programmatic implementation in COLAs, or more timely and accurate information for the millions on millions of decisions made by American households and firms every day--is likely to be substantial. That being said, it is obvious that except for an occasional high-profile event, such as the poor inventory statistics data that had to be dramatically revised in the midst of the early 1970s recession and that likely worsened the policy response, it is difficult to quantify exactly which decisions will be changed and exactly in what manner, with exactly what payoff. And, of course, the private sector develops alternative information sources and substitute vehicles to deal with the lack of timeliness and accuracy. These, however, are expensive, and reducing the private sector costs would surely result in private sector savings greatly in excess of the modest government expenditures that might prove necessary, at least in the short term, to improve the quality of economic statistics. But from my perspective, by far the most important payoff will be a more accurately informed citizenry. Statistics will always be abused, as well as used, by people with particular self-interests to promote or oppose. But, because they fail to measure the concept that they purport to measure, we should not--advertently or inadvertently--be putting out data we know to be either inaccurate or misleadingly labeled.
The resistance to such change, especially by special interests, is likely to be substantial. Powerful interest groups have a vested interest in particular statistics and may benefit from currently built-in inaccuracies. For example, the widely known overstatement in the cost of living in the current procedures to measure changes in the consumer price index overindexes many government spending programs. The recipients of those spending programs, many of whom are not well off, have a strong vested interest in resisting such an improvement.
In the end, however, the social benefit-cost ratio from the improved quality of economic statistics is likely to be extremely high. First, because the cost of making sensible changes is likely to be modest, perhaps negligible in the long term. And, second, because a myriad of social and private benefits will flow from a consistently more accurate and timely set of economic statistics.
1 See Norwood (1995) for a detailed discussion and history.
2 See Diewert (1981) for a discussion of index numbers.
3 As I was chairman of the Cabinet Working Group on Economic Statistics, the resulting budgetary initiatives and appropriations in the early 1990s, as well as the substantive, improvements, have come to be called the Boskin Initiative. Actually, my colleagues in the Cabinet departments and statistical agencies did most of the work and thus deserve the credit.
4 Recently, much attention has been focused on the issues surrounding improving the consumer price index (CPI). See Boskin, Dulberger, Gordon, Griliches, and Jorgenson (1996, 1997), Greenspan (1997), Abraham (1997), Diewert (1997). Some of these issues have been of concern for decades (see Stigler 1961).
5 See Boskin and Jorgenson (1997) for a discussion of how even small errors in the CPI can have immense budgetary effects and ramifications for understanding economic progress.
6 See Nordhaus (1996).
7 See Gordon and Griliches (1997).
8 See Gordon (1990).
9 See Boskin, Dulberger, Gordon, Griliches, and Jorgenson (1996).
10 See Boskin and Jorgenson (1997).
11 See Boskin, Dulberger, Gordon, Griliches, and Jorgenson (1996).
12 See Boskin and Jorgenson (1997).
13 More rationale and detail in the case of the CPI can be found in Boskin, Dulberger, Gordon, Griliches, and Jorgenson (1996).
14 See Hausman (1997).
15 See Gordon and Griliches (1997) and Moulton and Moses (1997).
16 Norwood (1995) makes the case for (with thoughtful analysis of the case against) a combined statistical agency or at least system and structure. Norwood was an influential member of the Cabinet Working Group on Economic Statistics, but at that time she concluded on balance against centralization. Her thoughtful recent analysis carries considerable weight in my current thinking on this fundamental issue.
Abraham, Katharine G. 1997. "The CPI Commission: Discussion." American Economic Review, May.
Boskin, Michael J., and Dale W. Jorgenson. 1997. "Implications of Overstating Inflation for Indexing Government Programs and Understanding Economic Programs." American Economic Review Papers and Proceedings.
Boskin, Michael J., E. Dulberger, R. Gordon, Z. Griliches, and D. Jorgenson. 1996. "Toward a More Accurate Measure of the Cost of Living." Final Report to the U.S. Senate Finance Committee, December 4.
Boskin, Michael J., E. Dulberger, R. Gordon, Z. Griliches, and D. Jorgenson. 1997. "The CPI Commission: Findings and Recommendations." American Economic Review Papers and Proceedings.
Diewert, W. Erwin. 1981. "The Economic Theory of Index Numbers: A Survey." In A. Deaton, ed., Essays on Theory and Measurement of Consumer Behavior in Honor of Sir Richard Stone, pp. 163–208.
Diewert, W. Erwin. 1997. "Comment on CPI Commission Report." American Economic Review Papers and Proceedings, May.
Gordon, Robert J. 1990. The Measurement of Durable Goods Prices. Chicago: University of Chicago Press.
Gordon, Robert J., and Zvi Griliches. 1997. "Quality Change and New Products." American Economic Review Papers and Proceedings.
Greenspan, Alan. 1997. Testimony before the U.S. Senate Committee on Finance, January 30.
Hausman, Jerry. 1997. "Comment on CPI Commission Report." American Economic Review Papers and Proceedings.
Moulton, Brent R., and Karin E. Moses. 1997. "Addressing the Quality Change Issue in the Consumer Price Index." Brookings Papers on Economic Activity.
Nordhaus, William D. 1996. "Do Real-Output and Real-Wage Measures Capture Reality? The History of Light Suggests Not." In T. Bresnahan and R. J. Gordon, eds., The Economics of New Goods. Chicago: University of Chicago Press.
Norwood, Janet. 1995. Organizing to Count. Urban Institute.
Stigler, George, ed. 1961. "The Price Statistics of the Federal Government." In Report to the Office of Statistical Standards, Bureau of the Budget. New York: National Bureau of Economic Research.