PAGE 1
Statistical Needs for a Changing U.S. Economy September 1989 NTIS order #PB90-126954
PAGE 2
Recommended Citation: U.S. Congress, Office of Technology Assessment, Statistical Needs for a Changing U.S. EconomyBackground Paper, OTA-BP-E-58 (Washington, DC: U.S. Government Printing Office, September 1989). Library of Congress Catalog Card Number 89-600753 For sale by the Superintendent of Documents U.S. Government Printing Office, Washington, DC 20402-9325 (Order form can be found in the back of this report.)
PAGE 3
Foreword In 1988 the Office of Technology Assessment published Technology and the American Economic Transition outlining ways that new technologies have redefined options for stimulating economic growth, Several Committees of the Congress had asked OTA to step back from its analysis of specific technologies and describe the combined effect of the changes on the living standards of different American households, on jobs, and on Americas position in the world economy. These questions required OTA to address some very basic questions about the way the economy operates and could operate in the future. The research made extensive use of statistical series from many private and public sources. In many cases standard statistical measures failed to indicate important dimensions of change. Improvements in choice and quality in areas like recreation were often unmeasured while increased spending for health often did not represent an increase in the quality of the amenity purchased. The Subcommittee on Government Information and Regulation of the Senate Committee on Governmental Affairs asked OTA to use the experience gained in the transition study to provide a perspective on areas where better data would improve economic policy analysis. We found that many dimensions of growth and change were not well tracked by the existing statistical system. The problems are greatest where change is most evident: the introduction of radically new technologies like computers and telecommunications equipment, the impact of international trade on the domestic economy, the value of education and training as an input to industry, techniques for evaluating the quality of heath-care and other services. In virtually all cases, the statistical agencies are aware of the problems and are making efforts to correct them. We find, however, that their efforts are hampered by a lack of effective coordination and management-a role that the Office of Management and Budget has the authority to oversee. Some of the problems identified in this study are old issues that recent events have made much more important, such as tracking the effects of international trade or finding ways to adjust service expenditures for inflation. Many of the problems are forced on us by technical change, evident in the difficulty associated with tracking quality changes in computers and other information equipment. In many cases the problems we identify have no easy resolution and the Nation will need to face the fact that uncertainties in key areas exist and, in some cases, are increasing. We also point out some long-standing problems that should not be forgotten. For example, there has never been a coordinated way to report statistics on the quality of life in America. The background paper does not attempt to provide a comprehensive critique of national statistics and does not introduce new research designed to solve the technical problems. It is, instead, designed to show how defects in the existing statistical system can limit our understanding of key economic issues and to demonstrate the ways that better management and coordination of Americas statistical agencies can lead to concrete improvements. OTA acknowledges the generous help of the reviewers and contributors who gave their time to ensure the accuracy and completeness of this report. OTA, however, remains solely responsible for the contents of this background paper. IUs cmge~~, office of ~~o]og As=sment, Technology and the American Economic Tradition choices for t~ F~e;OTA-TEr-283 (Washington, DC: U.S. Government Printing Office, May 1988). iii
PAGE 4
OTA Project Staff Lionel S. Johns, Assistant Director, OTA Energy, Materials, and International Security Division Peter D. Blair, Energy and Materials Program Manager Henry Kelly, Project Director 1 Andrew Wyckoff, Analyst Administrative Staff Lillian Chapman Linda Long Phyllis Brumfield
PAGE 5
Reviewers and Contributors Catherine Anmington Consultant Martin N. Baily The Brookings Institution Mark Baribeau John Hancock Financial Services Anthony J. Barkume Bureau of Labor Statistics U.S. Department of Labor James Bonnen Michigan State University Edwin R. Dean Bureau of Labor Statistics U.S. Department of Labor Edward Denisen The Brookings Institution Faye Duchin Institute for Economic Analysis New York University Paula Duggan Northeast-Midwest Institute Joe Duncan Dun & Bradstreet Corp. Maria Eli Coalition of Service Industries Audrey Freeman The Conference Board Norman Frumkin Consultant Norman Glickman Center for Urban Policy Research Rutgers University Edward Goldfield Committee on National Statistics National Academy of Sciences Gordon Green Bureau of Census U.S. Department of Commerce Hermann Havermann Office of Information and Regulatory Affairs Office of Management and Budget Roger Herriot Bureau of Census U.S. Department of Commerce Chuck Hulten American Enterprise Institute Philip Israilevich Federal Reserve Bank of Chicago Sidney Jones Department of the Treasury Thomas Juster University of Michigan Arnold Packer Hudson Institute Rick Kasten Congressional Budget Office Charles Leedman Organization for Economic Cooperation and Development Linda LeGande Congressional Research Service Dan Luria Industrial Technology Institute Sid Marcus Bureau of the Census U.S. Department of Commerce Larry Mishel The Economic Policy Institute J.R. Norsworthy Rensselaer Polytechnic Institute Robert Parker Bureau of Economic Analysis U.S. Department of Commerce Bruce Phillips The Small Business Administration Karen Polenske Massachusetts Institute of Technology Mellisa Pollak National Science Foundation Barry A. Rappaport Bureau of the Census U.S. Department of Commerce
PAGE 6
Katrina Reut Bureau of Labor Statistics U.S. Department of Labor Gordon Richards National Association of Manufacturers Marc Ross University of Michigan Isabel Sawhill Urban Institute Courtney Slater Slater Hall Information Products Thomas R. Tibbetts Bureau of Labor Statistics U.S. Department of Labor Raymond Vernon Harvard University Louisa Waldstein Klein & Co. Katherine Wallman Council of Professional Associations on Federal Statistics Bruce Walter Bureau of the Census U.S. Department of Commerce Amy Walton Jet Propulsion Laboratory David Weil Klein & Co. Margaret Wigglesworth Coalition of Service Industries Roberton Williams Congressional Budget Office Gaylord Worden Bureau of the Census U.S. Department of Commerce Allan H. Young Bureau of Economic Analysis U.S. Department of Commerce NOTE: OTA is grateful for the valuable assistance and thoughtful critiques provided by the reviewers and conrnbutors. The views expressed in this OTA background paper, however, are the sole responsibility of the Office of Technology Assessment. vi
PAGE 7
Contents Page Part I: Overview of the Federal Statistical System . . . . . . . 1 ANSWERS TO BASIC QUESTIONS . . . . . . . . . 1 STATISTICS FOR A CHANGING ECONOMY . . . . . . . 2 Part II: How Well Can We Answer the Basic Questions? . . . . . . 7 A. How Rapidly is the U.S. Economy Growing? . . . . . . . 7 B. Which Businesses Are Responsible for Growth and Has Growth in the Complexity of the Networks Connecting Different Kinds of Businesses Changed the Interdependence of Businesses? . . . . . . . . . . 13 C. What Is the Impact of Intenational Trade undomestic Producers and Consumers? . . . . . . . . . . . . . 21 D. What Capital and Labor Inputs Are Purchased by Domestic Producers? . 25 E. How Productively Do Domestic Producers Use Inputs? . . . . . 28 F. How Has the Corporate Structure of the U.S. Economy Changed? . . . 31 G. How Does Growth Affect Incomes and Income Distribution? . . . . 34 Part III: Conclusion . . . . . . . . . . . . . . 39 Boxes Box Page A. The Impact of One Economic Statistic: The CPI . . . . . . . . 3 B. The National Income and Product Accounts . . . . . . . . . 9 C. Deflator Series . . . . . . . . . . . . . . . . 10 D. Classifying Businesses . . . . . . . . . . . . . . 13 E. Japans International Input/Output Model Project . . . . . . . . 18 Figures Figure Page l. Index of GNP Growth and the Use of Energy and Materials in the United States . 8 2. Products Carried Per Supermarket. . . . . . . . . . . . 11 3. Shares of GNP in Constant 1982 Dollars . . . . . . . . . . 14 4. Exports and Imports . . . . . . . . . . . . . . . 21 5. Imports Used to Produce Amenity . . . . . . . . . . . . 24 6. Investment in Information Equipment . . . . . . . . . . . 25 7. Productivity . . . . . . . . . . . . . . . . 26 8. Mergers and Takeovers, 1972-1986 . . . . . . . . . . . 35 Table Table Page l. Direct Funding for Major Statistical Programs Fiscal Year 1989 . . . . . 2 2. Distribution of the GNP in 1987 . . . . . . . . . . . . 9 3. BLS Deflator Series for Gross Output . . . . . . . . . . . 20
PAGE 8
Overview Part I of the Federal Statistical System Good public policy demands good information. There may be disagreement about the wisdom of different Federal programs but there is little dispute over the need for adequate data to inform the debate. The information generated by the $2 billion spent this year by Federal agencies on statistical programs is a key resource for government policy makers as well as for private investors, public interest groups, academic researched, and labor organizations (table 1).1 Government statistics play a key role in evaluating and implementing legislation and are often used as indexes in private contracts (see box A). 2 Table 1 does not include a large, hidden cost of national statistics: the time invested by the individuals and businesses that provide the basic data. These costs obviously must be carefully considered in reviewing any proposed changes in statistical efforts. U.S. national statistics are acknowledged to be among the best in the world. But the U.S. economy is changing in ways that make documenting economic performance much more difficult. Business success today rests heavily on efficient management of new technologies and a grasp of the international marketplace. Competitiveness relies on quality, timeliness, and sensitivity to diverse markets. The most important inputs purchased by a business may be research and engineering information and the skills and education of its employees. Many of these factors are extremely difficult to measure. The new dimensions of growth and change have also challenged traditional approaches to economic growth policy. Policies that may have effectively encouraged growth in an era of little international trade may be ineffective or even counterproductive in todays global economy. Economic policy will require the best possible measurement of the factors critical for growth and an awareness of areas where uncertainty prevails. Serving the new needs of policymakers in a time of change will require a coordinated response of the Nations statistical agencies. The present management of the statistical agencies makes such a response difficult. The fault does not lie primarily in the management of individual statistical agencies. These organizations are painfully aware of the problems. The greatest problem appears to be the absence of any central place in government where basic questions about priorities in statistics are being asked, and the lack of effective coordination among statistical agencies. ANSWERS TO BASIC QUESTIONS The most basic questions of economic policy will endure, regardless of the transformations that affect the economy. We will always need to monitor changes in American living standards, determine whether access to the benefits of economic growth are more or less evenly shared, and estimate how the United States compares with other countries. But economic change has made it difficult to address even these fundamental questions with precision. This paper examines eight basic questions and our ability to answer them with currently available statistics: A. B. c. D. E. F. How rapidly is the U.S. economy growing? Which businesses are responsible for this growth, and has growth in the complexity of the networks connecting different kinds of businesses changed the interdependence of businesses? What is the impact of international trade on domestic producers, workers, and consumers? What capital and labor inputs are purchased by domestic producers? How productively do domestic producers use inputs? How does the way U.S. businesses are organized affect economic growth? (i.e., what are the relative contributions of different sized establishments and firms? what is the effect of mergers and acquisitions?) I It sho~d bC no~ that the btiget for fiscal year 1989 is half-a-billion dollars larger than usual because of expenditures for the decennial Casm. See table 1. ZFor exmple, ~ fi=~ yew 1984, 87 ~ent of all F~er~ grants-in-aid to State and local governments were distributed by form~a us~g F~er~ statistics, compared with an estimated two-thirds in fiscal 1975. See U.S. Congress, General &counting Office, Granf Formufas A Carafog of Federal Aid to States and bca/ities, GAO/HRD-87-28 (Washington, DC: U.S. Government Printing Office March 1987), p. 10. In terms of contracts, 2(X) billion to 300 billion dollars worth of contracts are keyed to the Producer Price Index (PPI) through inflation adjustment clauses. l
PAGE 9
2 Table 1Direct Funding for Major Statistical Programs Fiscal Year 1989 (estimate) Millions of dollars Commerce . . . . . . Census . . . . . . Decennial . . . . . Other . . . . . . Bureau of Economic Analysis . Other . . . . . . HAS . . . . . . . Centers for Disease Control . . National Cancer Institute . . Other . . . . . . Agriculture . . . . . . Soil Conservation Service . . National Agricultural Statistics service . . . . . Other . . . . . . Labor . . . . . . . Bureau of Labor Statistics . . Other . . . . . . Interior . . . . . . . U.S. Geological Service . . Other . .. . . . Energy . . . . . . . Environmental Protection Agency . Education . . . . . . Transportation . . . . . Justice . . . . . . . Treasury . . . . . . HUD . . . . . . . Defense . . . . . . Other agencies a . . . . . Total . . . . . . Total without decennial census . $ 705.8 646.2 434.8 211.4 24.7 34.9 290.5 129,0 51.0 110.5 242.3 83.1 64.1 95.1 226.5 190.4 36.1 119.5 78.9 40.6 90.8 69.3 36.1 32.7 30.4 28.2 14.1 8.3 85.7 $1,979.8 $1,545.0 aAgenoy for Mematiinal Development, Consumer Produet Safety Commission, Equal Employment Opportunity Commission, Federal Emergency ManagementAgenoy, Federai Home Losn Bank Board, National #bronautiea and Space Administration, National Science Foundation, Small Businese Administration, National Seienoe Foundation. SOURCE: Office of Management and Budge~Statiitical Programs of the UnitedStatesGovemmenL 1999. G. H. How does growth affect incomes and income distribution? How does growth translate into real improvements in living standards such as better health, an increase in real choice and quality of products, or rewarding employment opportunities? Part II of this report examines the data and the statistical apparatus that is in place to answer these questions endpoints out a number of deficiencies. The last question is not addressed as a separate topic but is touched on throughout. A proper way to address this question remains a major challenge for all statistical work. The material that follows reviews some of the administrative problems that have contributed to the problem. In many cases, of course, the statistical agencies recognize the problems but there are no easy answers (it is much easier for a report like this one to identify faults than to suggest concrete remedies). Better management can not guarantee improvements but can make the search for solutions more productive, and make better use of existing resources. It is OTAs hope that this report can provide guidelines for pursuing important improvements in statistical data gathering and analysis, providing a better understanding of the American economy. STATISTICS FOR A CHANGING ECONOMY While the decentralized system of statistical collection and analysis in the Federal Government has many strengths, the system suffers from the absence of any central organization able to develop a coherent strategy for adjusting to the challenges presented by todays economy. The Office of Information and Regulatory Affairs of the Office of Management and Budget was charged with establishing statistical policy and coordinating statistical efforts in the United States in the 1980 Paperwork Reduction Act. 3 It has not performed this role effectively. 4 While individual statistical agencies have made efforts to work together and solicit the opinions of data users, the absence of effective OMB leadership has left critical gaps. There is no national, systematic effort to articulate priorities in statistics and match budgets to these priorities, to anticipate future needs, to translate the complex and often conflicting objectives of data consumers into a practical set of tasks, or to ensure that the work of individual statistical agencies is adequately coordinated. In particular: There is a pressing need for an organization where fundamental statistical priorities are s~bli~ L~~ 96-511. 4A nm~ of ~r ~vlews have exmin~ deficie~ies in OMBS management of the statistical agencies. See, for example, The Federul Startiticaf System: ]980 rQ 1985, a report prepared by the Baseline Data Corp. for the Congressional Research Sewice, November 1984, pp. 46-67.
PAGE 10
Box AThe Impact of One Economic Statistic: The CPI The Consumer Price Index (CPI) measures the increase or decrease in the total price of a set of goods and services (a market basket) representative of a consumers purchases. Constructed as an indicator of inflation, the CPI has a direct effect on nearly every citizen in the United States. At least 8.5 million workers are covered by collective bargaining contracts that link wage rates to changes in the CPI. The payments made to 38 million Social Security beneficiaries, 3.5 million retired military and civil service employees and survivors, 20 million food stamp recipients, and 23 million children who eat lunch at school are also linked to the CPI by law. The Economic Recovery Tax Act of 1981 uses the CPI to prevent inflation-induced tax rate increases (bracket creep). l All told, the Bureau of Labor Statistics estimates that an increase of one percentage point in the CPI could add nearly $5 billion to the Federal budget. 2 IU.S. DqJ art.mem of Labor, Bureau of Labor Statics, BLS Handbook of Methods, Butletm 2134-2, April 1964, pp. 5-6. %hlculatcd by the Ot%ce of Mmagement and Budget for fiscal year 1986. see Es timorty of the Honorable Janet Norwood before the Su txmmittec m Govcrnrnen I Irt.formatim and Regulatiq Cornrnittex on Govcmmcttd Affairs, U.S. Senate, June 12, 1969, p. 2. periodically reexamined in light of the new needs of public and business analysts. The continuing underemphasis on service industries is a clear symptom of this problem. Hard pressed by the demands of mandated publication schedules, the statistical agencies have little time and few resources to do basic research or ask hard questions about priorities. This problem has been exacerbated by budget cuts. While it has a mandate to perform this task, OMB has not given it much priority and has dedicated few resources to the effort. l An effort is needed to evaluate whether statistical efforts match the significance of the problem. For historical reasons, the U.S. Department of Agriculture spends 6.7 times more on statistics than the U.S. Department of Education. It seems unlikely that this is the right ratio l given the transformation underway in the economy. Even worse, no organization in either the executive branch or the Congress has assumed responsibility for asking whether it is the right ratio. The work of the statistical agencies should be more closely coordinated. Much of the output of statistical agencies depends on careful coordination of work in several different agencies. The Bureau of Economic Analysis (BEA), for example, takes price indices from the Bureau of Labor Statistics (BLS), and incorporates data from the Bureau of the Census to develop data about inflation that is, in turn, used by BLS to develop productivity series. Such networks must respond to new priorities in carefully coordinated ways. Budget reductions in one agency can have complex effects on the performance of the integrated system. A clear view of the integrated needs of the statistical agencies is essential if either OMB or the Congress is to make well-informed judgments about budget priorities. A coordinated approach to analysis of the burden statistics impose on users might also reveal ways to produce better data without increasing the burden on respondents. Coordination also requires difficult judgments about how to handle confidential data. Opportunities for using available data without compromising confidentiality may be missed because clearances are not well managed, For example, the BEA staff is not cleared for access to confidential Census data. With adequate management it may be possible to develop tables that facilitate linking data series (e.g., by providing aggregate statistics about employment by establishment size classification) or provide for direct links that do not compromise confidentiality. Many organizations collect statistics for specialized regulatory programs. They have no mandate to contribute to a coherent national statistical program. For example, elimination of Civil Aeronautics Board or the Interstate Commerce Commission data occurred without the realization that the data provided key 5~~~ony of ~~ay sl~~ ~fo~ ~ JO~[ ~~jc c~i[~, U.S. WWCSS, ~~g Qualify Oflhe ~~h~$ Ecottornk Slutiszics, ~CW@ before the Joint Economic Committee, Mar. 17 and Apr. 17, 1986, p. 50; and U.S. Congress, GeneraJ Accounting Office, The Bureau of Ecortomc Anaiysis Should Lad Eflorts To /wq)rove GNP Estimate, GAO/GGD-83-l (Washington, DC: U.S. Government Printing Office, k. 27, 1982), pp. 58-61.
PAGE 11
4 l l information to BEAs income and product accounts. 5 There is reason to suspect that these organizations and nonstatistical operations like OMBs own Office of Federal Contracts and Procurement, the General Services Administration (for data about Federal purchases), and the Internal Revenue Service (IRS) could provide valuable data at comparatively little incremental cost if some care were taken to achieve government-wide efficiencies. The task of coordinating agency statistical work was also assigned to OMB by the 1980 Paperwork Reduction Act. Again, the task has not been given a high priority. Improved management could make the production of data more efficient and faster. A coordinated set of industrial codes, use of a common business list as a sampling frame (i.e., the Business Directory List), and more diligent efforts to use data improvements emanating from other agencies (e.g., BEAs use of BLS deflators in international trade and selected services industries) could improve data quality while possibly lowering costs. There is no formal mechanism that ensures that the needs of consumers of national statistics are reflected in the priorities of the national statistical system taken as a whole. Consumers of data frequently are often forced to work around deficiencies in statistics making heroic assumptions in order to satisfy the pressing demands for policy or business analysis. They may be forced to use data that are a decade out of date or rely on private databases that were not designed for sensitive policy work. The creation of an institutionalized feedback loop, connecting consumers to producers, would make the statistical agencies more aware of deficiencies in the data they produce. The report prepared for Economic Policy Council in 1987 by Wendy Gramm and Robert Ortner l l suggested the creation of such a system but little has been accomplished. 6 In its absence, the BEA, the Bureau of Census and other statistical agencies assemble advisory groups, but in many cases the complexity of the Federal statistical effort makes it difficult for data users to translate their needs into specific recommendations for individual agencies. 7 The feasibility or the cost consequences of different priorities are not easily estimated. With budget constraints, a compromise must be struck between forcing policy analysts to use data that may be many years old or using data that is published comparatively rapidly but might not be as complete or detailed as other users would like. 8 These and other trade-offs are difficult to address at the level of individual agencies. A government-wide perspective is needed. Better use of modern computational and communication equipment would contribute to productivity, The computational systems available to BEA, BLS, and other major statistical services appear to lag far behind the systems available to many o the business service industries that rely heavily on government data. Few Federal agencies have adequate distributed computer workstations or state-of-the-art local area networks. This is a major barrier to productivity (and perhaps to attracting people who expect to be able to use modem equipment). The growing interdependence of economies around the world has increased the need for international cooperative efforts in statistical work. Greater efforts are needed to coordinate U.S. and foreign data, and to identify areas where cooperative research projects in statistical methods would be beneficial. Cooperative efforts are most obviously needed in the statistics of international trade. The United %st.imony of Co urtenay Slatcr before the Joint Economic Committee. U.S. Congress, The Qucafity of the Natiwts Economic Stutisrics, Hearings before the Joint Economic Committee, Mar. 17 and Apr. i7. 1986, p. 50; and U.S. Congress, General Accounting Office, The Bureau of Economic Anafysis Should Lead Eflorts To hnprove GNP Estimate, GAO/GGD-83-l (Washington, DC: U.S. Government Printing Office, Dec. 27, 1982), pp. 58-61. 6s= ~c c$R~ ~ the Wofing Group on ~ @~lty of ~onomic Sttistics to tie &onomic Policy Council chaired by Wendy Gramm ~d Rokt Ormer, April 1987, p. 8. TSi&ey L. JOB, staying on l@ of the Numbers, The Broo&ings Review, Spling 1988, p. 38. 8~ isw~g of @fiinW da~ ~~d in ~ cn~l ~ ongoing ~vi~~ p~ess m ~dit)onal data become available. Such MI effort woldd involve an expansion of the current programs resources.
PAGE 12
5 l Nations has already taken a number of steps to coordinate statistical methods and categories. For example, a new U.N. group is working on international service sector statistics and the harmonized commodity classification system is being implemented. But more can be done. Many of the problems inherent in U.S. statistical agencies are faced by all advanced nations and much could be gained from joint research projects. Efforts should be made to ease the reporting burden and increase the timeliness of the data by taking advantage of commercial computer and communication technologies. It should be possible to improve techniques for electronic gathering of statistics, possibly by allowing companies to transfer data electronically from standard accounting software. For example, the IRS now has a pilot program that allows the electronic filing of tax returns, cutting the processing time by 2 or 3 weeks and reducing errors by a factor of 10. 9 BLS is currently testing touch-tone and voice reporting of establishment data; current results show a significant improvement in the timeliness of the data. The installation of a similar system for industrybased surveys and censuses should reduce the reporting burden on the firms and streamline collection efforts at the statistical agencies. Efforts in this direction require long lead times, and extensive coordination with representatives of the firms that will be affected. It should also be possible to make changes in the way data are delivered. (See OTAs report Informing the Nation for a more detailed discussion on information dissemination. *O) Steps have been taken-the Department of Commerce now has an electronic bulletin board and issues some of its data in a floppy-disk format, the U.S. Geological Survey issues some of its data on a compact-disk, and BLS and the Census Bureau make some data available on floppy disksbut more remains to be done. The proliferation of computers and powerful software has meant that not only has the medium for using data tilted towards an elecl l tronic format, but that the number and variety of users has grown. This shift needs to be weighed against the continued strong demand for printed statistical information. Greater effort needs to be made in coordinating statistical work describing changes in the goods and services available to individual households with the rest of national economic accounting. BEA does a heroic job in collecting and coordinating statistics from the many agencies with data relevant to the standard national accounting framework. But no group is asked to coordinate statistics in a way that provides an integrated look at the way economic change affects different types of households. Many statistics are available on changes in the quality of health care, access to transportation, and quality in education. The statistical system lacks an organization which is charged with ensuring that a complete and balanced picture is available from this data and that links can be drawn between changes in aggregate levels of spending, changes in household spending, and changes in the quality of such things as health care, education, and transportation available to households. Without such a coordinated effort, it proves very difficult to provide a balanced view of the way economic change has, and may, affect the welfare of different American households. Coupled with this is the need for an organization capable of addressing many of the basic challenges presented by an economy in transition. These challenges require a coordinated, patient, and systematic effort to match resources to new demands. They require an organization with the scope to translate emerging priorities into a practical plan for action and the power to ensure that this plan is enacted. Better management of existing resources could undoubtedly improve the quality of and usefulness of U.S. statistics. But there is a limit to the efficiency gains possibleeven with improvements in technology; data collection and compilation is an extremely labor-intensive task, Given the challenges presented by the transformation underway in the 9J~y RO~nfeld, me Electronic Taxtnan, PC World, April 1987, p. 187. l~os. (Imgw, Offi@ of ~~o]ow A~~e~t, ]nfOm~8 1~ NatiO~ Fe&r~ [~or~rion D@emi~lOn in an EkCtrOrlk Age, OTA-CIT-396 (Washington, DC: U.S. Government Printing Office, October 1988).
PAGE 13
6 Nations economy, more resources may well be overlooked. The cost of a poorly run government needed simply to maintain the quality of existing program may be many times higher than the cost of statistical series. Saving money by reducing statistiimprovements to statistical agencies. Unlike other cal budgets can be shortsighted if inadequate data lead to poor management of public programs or government purchases that can be postponed, statisprivate investments. Important opportunities for tics cannot be turned off and ononce a gap is growth may be missed and important dangers created it cannot be easily eliminated.
PAGE 14
Part II How Well Can We Answer the Basic Questions? The remainder of this paper examines the statistics available to address a set of important economic policy issues. The discussion is designed to outline some of the challenges faced by the statistical agencies in providing answers to even the most basic economic questions and survey some of the more important problems that have emerged in recent years resulting from inadequate planning, budgets, or coordination. Some of the problems identified are not new, some may be difficult or impossible to resolve completely given even the most perfect system. The purpose of this discussion is not to propose specific solutions to the problems raised but to demonstrate that important problems exist and that there is a need for a coordinated effort to address them. A. How Rapidly is the U.S. Economy Growing? New technology, the pressures of highly competitive domestic and international markets, and changes in the tastes and values of the American market have changed the direction of economic growth in basic ways. The number of pounds of materials and the total amount of energy used by the economy did not increase significantly between 1977 and 1987 even though the total output of the economy measured by the real Gross National Product (GNP) increased 28 percent (see figure 1).1 The products produced by the economy has obviously taken the form of adding more and more value to a given amount of basic materials. Much of this value is difficult to measure with the precision possible in a economy dominated by raw materials. Growth must be measured not only in terms of the number of items produced, but by changes in the quality of products ranging from optic fiber cables to fresh produce. In turn, quality should reflect the growth in the variety of products offered, and the extent to which people are able to purchase products well tailored to their specific tastes and interests. For example, the magazine publication industry, once dominated by large national journals, now has some 11,500 titles. 2 The problem of adjusting for quality becomes more complicated when the product being produced is a service: how do you measure quality changes in legal services? Tracking an economy where growth depends on qualitative factors is obviously more difficult than tracking growth when output is easily weighed or counted. It is necessary to acknowledge the fact that the precision with which we measure economic growth is likely to decline even given the most heroic efforts by statistical agencies. But policies designed to encourage economic growth need to be made with the best possible description of the areas where growth is likely to be important. Measuring Real Economic Growth The primary tool for measuring changes in the size of the economy is the Gross National Product Accounts (GNP) (see box B). The GNP estimates produced by the Bureau of Economic Analysis (BEA) have been the subject of a number of reviews, and improvements are constantly being made, 3 BEA received a total of 75 recommendations from various groups and commissions and was able to implement 51 of them. 4 BEA maintains a long list of additional improvements they would like to incorporate into the accounts. Two kinds of improvements to the GNP have been discussed over the years. The first deal with the basic structure and coverage of the accountse.g. should government spending be treated as consumption or should government spending on roads, airports, research, or education be considered an investments should the value of economic activity that occurs outside of the formal marketplace such as Iu.s, Conwess, office of T&~oIoU Asses~ent, Technology and the American Economic Traruition Choices for the Future, OTA-TET-283 (Wa.shngton, DC: U.S. Government Printing Office, May 1988), p. 277. ZU.S. ~p~cnt of Commerw, BWeau of the Census, Statistical Abstract of the United States, 1989, table 913, p. 549. 3MOS of tie rwent studies we reviewed in c.s. Carson and C. Jaszi, The Use of National Income ~d product Acomts for Public Policy: OU Successes and Failures, U.S. Department of Commerce, Bureau of Economic Analysis, Staff Paper No. 43. Am ~oups ~c~u&d the Nation~ ~comts Revie w Committ=, the Contributors to the Rerrospecf and Prospect, the GNp Data improvement Project, and the Round Table of GNP Users. See Allan H. Young Evaluation of the GNP Estimates, Surve} of Current Business, AugusI 1987, p. 20. 5S= R Ruggle~ ad ND. Rugg]es, ~te~atcd fionomi~ Accounts for the UIUted States, 1947 -80, Survey of cUfrCnt B~inew May 1982; R. Eisner, The Total Incomes System of Accounts, Survey of Current Business, January 1985; and Anthony S. Campagna, Macroeconomics Theory and Policy (Boston, MA: Houghton Mifflin, 1974), ch. 1, pp. 7-21, -7
PAGE 15
8 Figure I-Index of GNP Growth and the Use of Energy and Materials in the United States 150 (100 1973) 140 130 1973 1975 1977 1979 1981 1983 1985 1987 GNP (1982$ ) Primary energy Electricity (kWh ) Materials (Ibs) SOURCE: R. Williams, E. l-arson, and M. Ross, Materials, Affluencs and Energy Uaa,AnmJa/ Review of Erregy, No. 12, 19S7, pp. 99-144. housework or illegal activities be included in the national accounts? 6 The oil crises of the 1970s and growing concern about environmental issues in the 1980s has led to concern that the formal national accounts do not properly reflect changes in natural resource assets or environmental quality and thereby give a misleading view of changes in real national wealth. 7 It would be possible to maintain statistical series that would track resource and environmental issues as addenda to the traditional accounts. More precise tracking of resource and environmental changes would require statistics on the inputs and outputs of different production technologies not available within standard statistical series. 8 The existing statistical system makes it extremely difficult to anticipate the potential impact of emerging technologies. Statistics document changes in average businesses inputs and outputs but provide little information about the performance of facilities using new technology. The accounts do not distinguish between capital investments that simply replace obsolete or worn equipment from capital investments that represent real growth or replacement of old technologies with new. 9 These limitations makes analysis designed to show the net impact of new technology on employment, profitability. job quality, energy use and other factors difficult to track. A second class of improvement, which will be the primary focus of this discussion, deals with more technical issueshow accurate are records maintained within the existing accounting framework. The problem of improving, or indeed even of maintaining the quality of the GNP accounts has been made more difficult in recent years for a number of reasons: l l l Rapid changes in the quality of goods (especially in computers and other information equipment) and rapid growth of service industries makes the problem of adjusting for inflation increasingly difficult. Rapid increases in the number of comparatively small manufacturing establishments (many of which may be subsidiaries of large Fins) and increases in the role of service businesses (service businesses have always been comparatively small establishments), have made census counts more difficult. It is easier and less expensive to obtain accurate data from a small number of large establishments than a large number of small onesif only because the larger establishments keep more precise records. In some instances, data formerly available from accounts provided to regulators in businesses like trucking, railroads, and airlines now must %Mol S. Carson, The Underground Economy: An Introduction, Survey of Current Business, May, pp. 21-37, and July 1984, pp. 106-1 17; Frank de hmiw, An htdirect IMmique for Measuring the Underground Economy: A Note on Revised Data, Survey of Current Business, September 1986, pp. 21-22; Joel F. Houston, The Underground Economy: A Troubling Issue for Policy makers, Business Review, September-@to&r 1987, pp, 3-12; and James D. Smith, Measuring the Informal Economy, The Annals of the American Academy, vol. 493, September 1987, pp. 83-99. 7S= Ro&fl Re~t~, wsst~g AWts: Na~~ ResNces in the National Income Accounts (Washington DC: World Resources ~mitute. 1989); A System of National Accounts, U.N. Statistical Papers Series F-2, 1968; U.N. Department of Economic and Social A.ffsirs, Provisional International Guidelines on the National and Seetoral Balsnee Sheets and Reconciliation Accounts of the System of National Accounts. Statistical Papers Series M60, 1979; U.N. Statistical Office, Future Directions for Work on the System of National Accounts, 1979. sFaye Duehin, Frsmewmk for the Ewluation of Scenarios for the Conversion of Biological Materials and Wastes to Useful Products: An Input-output Approach, presented at the joint session of the American Economics Assoeiatiort/American Association for the Advancement of Science, New York, Dee. 29, 1988. 9Faye ~ch~, 1a&~ysing s~twd fimge in the &xmomy,l~ut-OutputA na,lysis: CurrentDeveloPments, M. Ciaschini (d.) (bndon: Chwmm and Hall, 1988).
PAGE 16
9 Box BThe National Income and Product Accounts The U.S. GNP accounts are constructed in two ways: (i) the product accounts that measure the value of all products and services sold for final consumption by households and the government, as business investment (gross private domestic investment), and net exports (exports less imports), and; (ii) the income accounts that measure the value of all income earned as wages, benefits, profits, and the like (see tab. 2). 1 In principle, the income and product accounts both sum to the GNP. In practice, adjustments need to be made to achieve balance. The product estimates are generally considered most reliable. Even after both sides of the accounts are adjusted using a variety of data sources a small statistical discrepancy remains. It was 0.2 percent of the GNP in 1987 and appears to be declining slowly. 2 It is necessary to recognize that the GNP accounts were never intended to be a complete tool for describing the economy and its limitations must be recognized. Many activities of enormous value do not appear in the accounts (i.e., the value of education received from parents at home, environmental damage resulting from economic activity) largely because they occur outside of the formal marketplace. GNP accounts no longer provide any information about the way income is distributed among households in fact the GNP can grow while the real income of many groups declines. 3 ICarol S. Carson, The History of the U.S. National Income and ProducI Accounts: Development of an Analy~cal Tcwl, Rewew of Income and Weal[h, June 1975. Carol S. Carson, GNP: An Overview of Source Data and Estimadng Merhmts, Swwey of Current Bu.rmess, July 1987. 2A.H. Youg, Evduatmn of the GNP Estunmes, Survey of Curren/ Business, August 1987, pp. 20-21. 3~ he 19703, & BEA had a Progm of non.market ~pec~ of ~onomic wellbeing ~d prexn[d busehold income dMtIhAtlOll ss a part of GNP accounts, but both were el iminated because of budget cuts m the late 1970s and early 1980s. Table 2Distribution of the GNP in 1987 to the BEA, which maintains the GNP ac(percent of total) counts. 10 Income accounts: Compensation of employees . . . . Property-type income a . . . . . Depreciation b . . . . . . . Other c . . . . . . . . . Statistical discrepancy . . . . . Total . . . . . . . . product accounts: Personal consumption expenditures . . Government purchases of goods and services. Gross private fixed investment. . . . Net exports of goods and services . . . 59.2 22.0 10.6 8.2 -0.2 100.0 66.7 20.4 15.7 -2. 8 Total . . . . . . . . 100.0 %oprtetors income, rental income, corporate profits, net interest. bc~lt~ ~nsumption allowances with capital consumption adjustment. Technical Measurement Issues Adjusting for InflationThe question of how to adjust for inflation is probably the most difficult to solve. Yearly measurements of the value of goods and services purchased cannot provide an adequate measure of economic growth. In order to obtain a consistent comparison over time, products valued in current dollars must be revalued into a constant set of prices that adjust not only for the changes in prices, but also for changes in the quality of products. cBu&lness ti~sf& ~yments, indirect busks tax and nontax-llablllty b3S subsidies less current surplus of government enteqmses. The Bureau of Labor Statistics provides the basic SOURCE: U.S. Department of Commerca, Bureau of Eonomic Analysis, data for inflation adjustments primarily by using the National Income and Product *ounts. results of three major surveys (box C); each of which has its share of drawbacks. Some are chronic while be obtained from trade associations or census others result from changes in the economy. surveys. l Perhaps the most vexing problem arises from During the past few years there has been an 8 changes in the quality of the products offered percent real reduction in the funding available for sale. Price adjustments work well only lo~ fjw~ yea 1978, BEAS budget (exc]u~ng ~~sfers) wss $16.2 million (in 1980$). Ln 1986 the budget WaS $14.9 million. U.S. Coni9ess, Gen~al Accounting Office, R&D Funding: Tbe Department of Education in Perspective (GAOEEMD-88-18FS), May 1988. PP 9-11. For a more detail~ discussion of the budgets of the statistical agencies, see National Association of Business Economists, Report of the Statistics Committee of the National Association of Business Economists, February 1988, pp. 10-15, 25.
PAGE 17
10 Box CDeflator Series l The Consumer Price Index (CPI) is used to compare prices of a fixed market-basket of products and services purchased by consumers -about two-thirds of the GNP. Surveyors are given a precise list of products, an item might be a jar of peanut butter, and check the prices of these items in a scientifically selected set of retail establishments throughout the country. Changes in the prices of the selected products are used to estimate price changes in most areas of consumer spending. They are also used to estimate price changes for many areas of government spending. The ratio of priced changed for a product in a given year to the price charged in a baseline year (e.g., 1982) is called the deflator for that year. l The Producer Price Index (PPI) is used to adjust the value of most products purchased as an investment by businesses and business inventory changes. This index uses techniques similar to those used in the CPI but in this case surveyors ask individual businesses for the prices of their products. l The International Price Program measure price changes for products imported and exported from the united states. Separate series are provided for housing and other structures and for some other products. The PPI is used to deflate some government purchases of goods and services for which a reasonable analogy can be found in the private sector and which are not covered by BEA price series on defense expenditures and compensation paid to government employees. 2 IIW u,s. Nfi~ -C ~d fi~~[ &XOUXMS: Revised Estimates, Survey o/Curren( Bwiness, July 1988, table 7, pp. 31-33. 21bid when quality changes occur comparatively variety of different types of peanut butter in slowly. But in today's economy, growth must increasingly be measured in terms of product quality and consumer choice. Quality adjustments are comparatively easy when a new feature is added to a familiar product. If a new car is sold with white side-wall tires as standard equipment, for example, the Consumer Price Index (CPI) can be adjusted to reflect the extra marginal cost of these tires. Other kinds of quality adjustments are more difficult-both conceptually and mechanically. Electronic products are perhaps the most striking example of change. New kinds of television receivers, home computers, telephones, and a variety of other devices redefine consumer electronic markets yearly. Even more troubling, the inflation adjustments miss an important dimension of quality that seems to have proven appeal to the American market: the value of growing choices available to consumers. While the BLS surveyor accurately assesses price changes in the 12-ounce jar of peanut butter, a grocery store customer may be selecting the jar from an enormous 1989 while having few choices in 1983. For example, large groceries have also introduced many other amenities (salad bars, bakeries, fish specialties, etc.) without significant increases in product cost. Is this an improvement in quality? Consumers and grocery store owners seem to think so, Generic brand offerings have declined while stores offering 20,000 or more different products have prospered (see figure 2). 11 The value of variety embodied in grocery purchases is not captured by the BLS surveyors. As a result, BLS may measure a price increase while in fact the real cost of the bundle of goods and services purchased at groceries may have decreased because sophisticated technology and management techniques make variety increasingly inexpensive. It has become increasingly difficult to develop good measures of changes in the quality of products purchased as capital equipment by businesses (7 percent of GNP). In particular, new information equipment presents the most serious challenges. Working with IBM, BEA has attempted to find a way to adjust the prices 11N B&ly ~d Ro~~ J. Gordon, me productivity Slowdown, Me~~ment ISSUCS, and the ExplosioII of COmpWiX power, Brookings Papers on Economic Aaivify, vol. 2, 1988, p. 412.
PAGE 18
Figure 2Products Carried Per Supermarket Thousands 2 0 16 10 0 1948 1967 1967 1972 1977 1982 1986 SOURCE: M. Baily and R. Gordon, The Productivity Sbwdown, Measurement ksues, and the Explosion of Computer Power, = Brookings Pspws on Economic Activity, vol. 2, 1988, p. 413. of computers for quality changes. 12 The deflator uses an index not strictly tied to the price of products, but instead makes adjustments by comparing specific characteristics over time, such as how many million of instructions are executed per second (MIPS) or how memory capacity has changed, to get an indication of the change in the value of computer power. 13 Not surprisingly the change has been dramatic. For example, the price of one megabyte of main memory fell by a factor of 20 between 1972 and 1984. 14 Changes of this magnitude have a large affect on the deflator. While the deflator for all producer durable equipment (PDE) was 1.078 in 1987 (1982=1.00), the deflator for the office, computing, and accounting machinery (OCAM) category was 0.55. 15 This single deflator has an enormous effect on the real purchases of PDE measured, the real growth of GNP, and the productivity growth rate of manufacturing. For example, PDE in constant l 1982 dollars increased at an annual growth rate of 13.8 percent between 1983 and 1987. If the OCAM deflator had been the same as the average of the other parts of the PDE deflator, growth would have been only 4.3 percent. 16 The process used to reflect quality improvements in computers has not been extended to other high-technology equipment like semiconductors and communications equipment, although work is beginning. As a result, the output price deflators for microelectronics actually rose from 1972 to 1982 while the index for computers fell drastically .17 This inconsistency could lead a researcher unaware of the problem to erroneously conclude that productivity gains in the computer industry were achieved without, or in spite of, corresponding gains in the semiconductor industry. 18 Similar problems exist in many areas where new products are radically different than the ones they replace. Measuring changes in the quality of computers, however, can seem easy in comparison with the challenge of measuring changes in the quality of services. Services are a growing fraction of the GNP before adjustments are made for inflation. Changes in the quality of health care (10 to 11 percent of the GNP) and education (7 percent of the GNP) are poorly measured or not measured at all. Fundamental conceptual problems must be confronted in developing deflators for these sectors. BLS has research projects underway in health care and other difficult service sectors, A deflator for health care should, in principle, adjust for changes in the quality of care received. But while it may be possible to develop a deflator for a specific medical test or a specific set of medical IZR. ale, Y.C. cttm, J.A. Bsrquin-stollctnan, E. Dulberger, N. Helvacian, and J.H, Hedge, *@ality-Adjusted Price bdexes for COrnPUter ~mes~~ and Sekctcd Peripheral Equipment, WV9 of Current Buriness, January 1986, pp. 41-50; David W. CartWright, ( Lrnproved Deflation of Purchases of computas, ~wey of Current Business, March 1986; and David W. Cartwright and Scott D. Smith, Deflators for Purchases of Computers in GNP: Revised and Extended Estimates, 1983 -1988, WV9 of Current Bwiness, November 1988, pp. 22-23. 13c01e, et ~., op. cit., pp. 41-50. Iak, et al., op. cit., p. XT. 15u-s. -mt of berm, B~u of ~~c Analysis, National income ~d product Acc~ts, ~blc 5.7. l~~id., tables 5.6 ~ 5.7. 17s~w.~ J~g ~ J-R. N~~fiy, sc~e ~~ies, ~~g *CS Snd ~ws~~ ~~uctivity Grow: A Study of ~hnolOgy k the Us. Micnx kctronics and Computer Industries, Wchnical Report 02-88, Center for Science and Ikdmology Policy, School of Management, Rensselaer Polytechnic lttstitute, August 1988, p, 13. 18 Ibid., p. 12.
PAGE 19
12 procedures, it is difficult to measure whether the patients health has benefited from additional tests that may be administered. Changes in the quality of structures (8.4 percent of the GNP), particularly nonresidential structures, are measured very poorly .19 The value of new housing is adjusted using 10 measures of housing quality. 20 The value of nonresidential structures is adjusted for inflation using a mixture of the residential deflator and a standard set of building inputs. 21 There are many reasons to believe that the construction deflators prepared in this way do not provide an accurate measure of changes in product quality. The residential deflator, for example, does not reflect the addition of new amenities such as dishwashers, energy-efficient improvements, and landscaping which have become common 22 The use of a standard mix of labor, material, and equipment inputs, is used to represent nonresidential buildings ranging from warehouses to hospitals has obvious deficiencies. Changes in quality and productivity improvements are not incorporated. The Canadians, who do a much more thorough job of measuring construction quality, estimate that the price of construction products rose only 3.5 percent more than the Canadian GNP average between 1967 and 1986 while the U.S. estimates show construction prices rose 15 percent 23 Because the price more than the U.S. average. index is so high, output of the construction industry tends to be over adjusted for inflation, resulting in an underestimate of real output. Adjustments for inflation can lead to misleading measures of growth rates. For example, the deflator for computers is so much lower than the deflator for the economy, a significant fraction of the percentage growth of GNP in constant dollars results from the methods, particularly the selection of a fixed base year, used to adjust for inflation-not an increase in current spending for computers. The deflators can thereby create a distorted view of growth rates. For example, growth in producer durable equipment between 1982 and 1988 measured in constant 1987 dollars averaged 5.9 percent per year but was 8.4 percent per year when measured in constant 1982 dollars. The BEA is examining alternative ways to express GNP growth in preparation for the comprehensive revision scheduled for 1990. 24 One way to avoid distortions due to different base-year weights is to compute growth rates by determining the constant dollar growth rate for each product separately and weighting each product by the average current dollar sales of each product for the first year. 25 It would also be useful to develop explicit addendum accounts that report changes in production and use of physical commodities whenever such data is available. Many of these series are already maintained to develop deflator series. Published series consistent with the national accounts would provide a view of changes in demand for energy, materials, and other countable products and services that would be a valuable addition to constant dollar measures. 1~Oblern is Cornplicti by the fact that structures are very heterogeneous products produced by a diverse industry composed of gener~ly sm~l fins. See Committee on Construction Productivity, Building Research Board, Commission on Engineering and fkchnical Systems, National Research Council, Con$trucdonProducrivity (Washington DC: National Academy of Sciences Press, 1986); and P. PiePer, The Measurement of Structures Prices: Retrospect ad Prospect, 50th Anniversary Conference on Research in Income and Wealth, NBER, May 12-14, 1988. %oor q number of stories, number of bathroom, presence of central air-conditioning, type of psrking facility, type of foundation, geographic region, metropolitan location, presence of fireplaces, and lot size. See price Index of New One-Family Houses Sold (Bureau of the Census), various issues. zl~ ~ ~ns~ction G., a l~ge builder of c.ommerci~ and industrial structures, estimates the cost plus profit using a standard mix Of inputs. See Revised Deflators for New Construction, 1947-73, Survey of(krenr Business, August 1974. ootnote 11, pp. 402-406. 22B~]y ~ Gordon, op. Cjt.g f mB~ly ~ @rdon, op. cit., footnote 11, PP. @2~. ZqAIlm H. yo~, Alternative Measures of Real GNP, Survey o~Currenf Business, April 1989, PP 27-34. MS= N~O~ ~come and Product Accounts, op. cit. footnote 15, table 8.1.
PAGE 20
1 3 Box D-Classifying Businesses Measuring the output of industries is contingent upon defining the industry itself. The fundamental identifying system for industry classifications is the Standard Industrial Classification (SIC) Code-the official existence of an industry. After 15 years, several false starts, and the input of over a thousand public and private opinions, the SIC was revised in 1987. Nevertheless, several observers think that this effort was incremental in nature and that a more complete overhaul is needed. l The concern is that emerging industries that typically enjoy tremendous growth can not be tracked because they are not identified by the SIC system. Instead their growth is lumped into a broader category that obscures the source of the change or a category that is a grab bag of leftovers such as SIC 7389 Miscellaneous Business Services which includes everything from meter readers to yacht brokers. Even though the 1987 SIC revision created three new 4-digit categories for computer equipment like storage devices, terminals, and computer equipment n.e.c. (not elsewhere classified), it still left ail computer manufacturing (minis, micros, and mainframes) under one 4-digit SIC. Similarly all eating places (McDonalds to the 21 Club) are under one 4-digit SIC (no change from 1972). Meanwhile, footwear gets broken into rubber and non-rubber categories (at the 3-digit level) and then under non-rubber footwear there are four, four-digit categories: mens footwear (except athletic), womens footwear (except athletic), house slippers, and footwear, except rubber n.e.c. 2 Although attempts to retain consistency over time are important, additional detail for large and growing industries at the 4-digit level seems warranted. INati~ As=iatim of Fhsiness hmomists, Report of the Staustics Comtm [tee of the Nationat Association of Business Ecmmrnk, February 1988, p. 17; Statunent of Courten ay Slater before the Subcommi [tee on Govemrnenl Lnfotma[ion and Regulation, Commi ttee on Govemtne ntal Affairs, U.S. Senate, May 15, 1989, p. 5; National Academy of Scicnees, Committee on National Statistics, Statidcs Abcwf Service Industries (Washington, DC: National Academy of Sciences), 19S6, p. 10. @ffkx of Management and Budget, Standard Industrial Classifiiatwn Manual, 1987 (WashingtorL DC: U.S. Gov emment Printing Office). B. Which Businesses Are Responsible for for future growth, and it is needed to understand the Growth and Has Growth in the Complexity of way businesses operate together as parts of complex the Networks Connecting Different Kinds of Businesses Changed the Interdependence of Businesses? While many difficulties are encountered in measuring the gross output of a modem economy, even greater problems are faced when attempts are made to trace this output to the activities of different kinds of businesses. Just defining and classifying a business is a difficult task (see box D). The contribution each business type makes to GNP is important for a number of reasons: it is needed to assess rates of innovation and productivity growth in different types of business, it is needed to understand which kinds of economic activity are likely to be the basis production networks .40 Data published by BEA indicate that the share of the GNP provided by natural resource industries, such as farming and mining, has decreased sharply since 1950 while output from the service sector, particularly services that play a transactional role in the economy such as finance, communication, and business services, has increased (see figure 3). 27 BEAs data show that manufacturings contribution to GNP has stayed remarkably stable at roughly 20 to 22 percent of the GNP over the 36-year period from 1950 to 1986 when measured in constant 1982 dollars. 28 This estimate has been the subject of a ~Jerome A. Mark, Problems Encountered ift Measurin g Singleand Mtdtifactor Productivity, Monrtsfy Lu60r Review, December 1986, p. 6; and Edwin Dean and Kent Kunze, Recent Changes in the Growth of U.S. Multifactor Productivity, IUottrhly Lubor Review, May 1988, p. 20. zTBusin~s ~i~s aS defm~ in the irtput/owput accounts includes activities such as consulting, law, advertising, and cOmp@r WViCe5. to name a few. 28National ~come and Product Accounts, op. cit., foomote 15, table 6.2.
PAGE 21
14 Figure 3-Shares of GNP in Constant 1982 Dollars Percent of GNP 2 5 20 Manufacturing Transactions I 10 I Natural resources I considerable amount of criticism during the past year. 29 Manufacturings stability appears to be inconsistent with other events in the economy such as a huge trade deficit in manufactured goods, lagging investment in plants and equipment, and the loss of 2 million manufacturing jobs between 1979 and 1986. 30 BEA has responded to some of the criticisms and has undertaken an effort to revise the constant dollar value-added by industry series (termed gross product originating) .31 Although firm conclusions cannot be drawn about biases in the present series until the revised series is available, it appears likely that manufacturings growth in output from 1979 to 1985 will be revised downward. 32 Assuming that the gross national product of the economy is properly measured, overestimating the contribution of one business sector (e.g., manufacturing) necessarily is balanced by an underestimate of the contribution of other sectors (e.g., business services). Understanding Linkages Tracing the source of growth in the economy to individual business sectors requires an ability to describe the complex business networks that now operate in virtually every part of the economy. The transformation of the American economy can be seen in the growing complexity of these networks and the service businesses needed for their efficient operation. 33 Understanding the complex patterns of interdependence that result from these networks is critical for understanding the way national policies can affect economic performance. Imports that affect manufacturing industries have a strong indirect effect on the service firms that supply these industries. The prosperity of services and manufacturing depends as never before on the quality of infrastructures like communications and a responsive transportation system. In an effort to cut costs many firms have begun to specialize and purchase products and services from other specialized fins. Manufacturing firms may, for example, purchase legal, bookkeeping, or janitorial services from outside suppliers rather than performing these activities in-house. 34 As a result of this growth in subcontracting, and the wider geographical dispersion that it entails, service sector businesses have thrived because of the increased need for financing, legal assistance, consulting services, communication, transportation, and wholesale and retail trade. These transactional and distributional service sectors represent the two fastest growing parts of the economy over the past two decades. Separately, each of the two groups contributes more to the GNP than all of manufacturing. The basic source of information about the way businesses depend on each other is the input/output Wechnologyandthe Ameriean Economic Tramition: Choices for the Future, op. cit., footnote 1, pp. 168-175; Lawrence R, Mishel, Manufacturing Numbers: How Inaccurate Statistics Conceal U.S. Industrial Deeline (Washington, DC: Economic Policy Institute, April 1988); Edward F, Denisen, Estimates of Productivity Change by lndust?y (Washington, DC: The Brookings Institution, 1989); Robcn Kutmcr, U.S. Industry is Wasting AwayBut Official F@ures DuIt Show It, Business Week, May 16, 1988; The Factory Rebound maybe More Fantasy Than Fact, Business Week, Dee. 12, 1988, p. 98., and Anthony Harris, Figures Calculated to Deceive, Financial Times, July 11, 1988, p. 11; Toddi L. Gufncr, U.S. Economic Statistics Off the Mark, National Journal, Sept. 3, 1988, p. 2200. % Lawrenec R. Mishcl, The Late Great Debate on Deindustrialization, Challenge, Janwq/February 1989, p. 35. 31U.S ~~at of c~em, Bureau of Economic Analysis. Gross Product by Industry: Comments on Recent Criticisms, Survey of Current Bwiness, July 1988, p. 132. 32~id. p. 133; ~~el (1989), ~, ci~, fm~~ 30, p. ~, Baily and Gordon, op. Cit,, footnote 11, p. 367; and Deniscm, op. cit., footnote 29, p. 23 and p. 37. ~JTeC~~~ ~ t~ tiric~ Economic Transition, op. cit., chs. 44 and 5, pp. 143-177. 34John T~&t@, *C~u= Senlces ~d~es: Wy ~e They Growing SO Rapidly? Monthly f,ubor Review, December 1987, pp. 314.
PAGE 22
15 (I/O) tables, compiled by BEA. These tables show what each of 537 industries purchase from other industries. Input/output data provide an essential tool for tracking the effects of new technologies; they alone provide a detailed description of how new technologies affect the inputs needed by different businesses. I/O tables also provide a key tool for monitoring the performance of the complex business networks that are coming to dominate the U.S. economy, showing how the value of products sold to final consumers combine the skills and technologies of the retail, wholesale, transport, production, and natural resource industries. They also provide a key insight into the way the rapidly expanding service industries are used by other businesses. Since so much detailed information is needed, it takes many years to create input/output tables using current methods. The most current benchmark table, called a benchmark because it is largely based on quinquennial industrial censuses, dates to 1977 and was published in 1984. The 1982 benchmark table will not be published until later in 1989. Unless something changes we will be using the recession year of 1982 as a benchmark until 1994. BEA produces input/output tables for years between benchmarks using a variety of approximation techniques. Until the past few months, the updated annual tables were published 6 years after the year for which they apply (e.g., the 1983 table was published in 1989). 35 Partly in an effort to fully integrate the gross product originating series with the I/O tables, the BEA has accelerated the process of constructing annual I/O tables and a 1986 table will be published late in 1989. Although more up-to-date, the annual input/ output accounts suffer from an industrial classification scheme that has a strong manufacturing bias. Of the 85 industries, 52 are dedicated to manufacturing, 15 are services, 12 are in natural resource, and 6 are other. This occurs even though only a fifth of the GNP is attributable to manufacturing. The end result is great detail on the production of wooden boxes, about two-one hundredths of a percent of GNP, while the private health, education, and social service industries, about 8 percent of GNP, are lumped together into one category. Obviously, these classifications present a severe constraint on analysis. Another limitation with the more timely annual tables is that they are forced to use the 1977 benchmark table as the basis for scaling, limiting the ability to track areas where the economy has changed rapidly. Unfortunately these are often precisely the areas where most policy analysis focuses. Difficulties in tracking the role of services is an important example. The fastest growing intermediate input in the economy, and particularly to manufacturing, is the purchase of a group of services collectively called business services, which contains services like accounting, advertising, legal help, computer services, and temporary help services. The 1982 input/output table shows business services as the third largest intermediate input to manufacturing above commodities like steel, rubber, paper, and transportation. 36 The basic data source on intermediate inputs for the manufacturing sector, the quinquennial Census of Manufacturing, does not collect data on purchased business services. The smaller, sample-based Annual Survey of Manufacturing collects data on only a few purchased services such as repair and communication services once every five years. This means that a number of approximation techniques were used even to establish the benchmark 1977 input/output table. Attempts to scale up from this benchmark to a more recent year are highly approximate 37 It is extremely difficult to track many technical changes resulting from greater purchases of services by businesses. For example, employment in temporary help agencies (one component of the 1/0 business services sector) grew by 70 percent from 1982 to 1984. 38 This out-sourcing has 35u.s. ~artmcmt of commerce, Bureau of Economic Analysis, Survey of Current Business, Annual Input-Output kcounts of the U.S. Economy, 1983, Fe1989, pp. 21-36; U.S. Depsrtmatt of Commerce, Bureau of Economic Analysis, Survey of Current Business, Annual Input-Output Accourttsof tk U.S. Economy, 1982, April 1988; U.S. Department of Commerce, Bureau of Economic Analysis, Survey of Current Business, InputOutput Accotmts of tbe U.S. Economy, 1981, Jsnusry 1987. Mu+S. ~-at of c~~, B~~ of ~nomic ~ysis, Suney @Curren( B~iness, AnntI~ Input-output Accounts of the U.S. )%OnOmy, 1982, April 1988, pp. 3146. 37Gross Roduct by Industry: Comments on Recent (Micisms. op. cit., footnote 31, p. 132. 38* L. C=y ~ ~m L. H~]~er, Employment Groin in ~ ~mporq Help ~dlls~, Monthly hhr Review, April 1986, p. 37.
PAGE 23
16 obviously reshaped business networks and affected the difference between sales and value-added contributed by the business, but the impact is almost impossible to trace. The fast growth of business services means that it is likely that they have not been fully accounted for, causing an overestimate of manufacturings contribution to GNP that has grown in severity recently .39 Data on the services used by businesses at the establishment level are difficult for Census to obtain because services like accounting and advertising are often purchased by corporate headquarters while the questionnaires go to individual establishments. The managers in the establishment often do not know how much corporate advertising is done in their interest. This problem could be reduced if the information was collected at the headquarters of the firm and then allocated to the fros individual establishments through an imputation scheme. There are also defects in the way the input/output statistics can track the performance of the new kinds of transportation systems required by a flexible, highly interconnected economy. 40 One important source of data is a Census Bureau product called the Commodity Transportation Survey which measured what types of manufacturing output were transported by a particular mode of transportation: truck, rail, air, or water. As the economy shifts toward a system of flexible production networks that relies more on just-in-time inventories, quick reactions to competitors, and better responses to consumer demand, transportation data like those provided in the Commodity Transportation Survey (CTS) are of considerable importance in tracking this change. 41 Due to methodological problems, the 1982 CTS was postponed to 1983 and conducted in a modified form. Because of the deficiencies found in the quality, the 1983 CTS was never published. 42 The 1987 CTS was canceled due to methodological problems and budget constraints. 43 The input/output tables are important for many reasons other than computing accurate estimates of value-added in each type of business. They are also used extensively in preparing detailed tables throughout the National Income and Product Account. 44 The Bureau of Labor Statistics relies on input/output data to generate its industry-level multifactor productivity series, 45 construct the producer price index, 26 and estimates of what occupations will be in demand in the future. 47 Input/output is at the heart of the Department of Agricultures projections of agricultural output and the Department of Energys estimates of energy use. Since input/output statistics are so important, and BEA tables are often many years out of date, private analytical firms and other Federal statistical agencies have developed their own updated tables using a variety of methods. The Bureau of Labor Statistics estimates its own updates of input/output tables independently of BEA. Because of this independent effort there are disagreements between BLS and 39&tim, op. cit., fOOmOti 299 P. 47 @U.S. Congress, General Accounting Office, The Bureau of Economic Analysti Skndd Lead Eforts to Improve GNP Estimate, GAO/OGD-83-l (Washington, DC: U.S. Government Printing Office, Dec. 27, 1982), p. 58. 41Ro&~ H. Hayad R~ch~dr~ hikm~, Manufacturings tisis: New lldmologies, Obsolete ~gSIIk~i0f15, Harvard Business Review, September-October 1988, pp. 77-85. 42u.s, cm-, ~lu of ~~oloa A=ment, Tr~portatwn of ~a~rd~~ Afaterids, OTA-SET-3@$ (Washington, DC: U.S. Government Printing Offke, July 1986), p. 44. 4SK.R. polem~, RelevmU of U.S. Re@on~ Statistics, pm~nt~ at the ~eric~ Association for the Advancement of Science, Jan. 19, 1989, p. 3. Warwn, op. cit., foomote 6, p. 112. qS~w~ km ~d Kent K-, *R~ent ~anges in the Growth of U.S. Multifactor Productivity, A40nthty Labor Review, May 1988, p. 20. 46A-G. clm ad Willlm D, Thom~, New wei@t s~t~ king us~ in producer Price lndcx,~ont~fy Lubor Review, Augu~ 1987, p. 12; d Robert Gaddie and Maureen Zoner, New Stage of Process Price System Developed for the Producer Price Index, Mwtthly Lubor Review, April 1988, pp.3-16. dTu,s. ~mmt of~, Bweau of ~rs~stics, B~&o~mic GrowlhMOdelSYSlern UsedforProjectio~ to 1990, Bulletin 2112, April 1982, p. 2,
PAGE 24
17 BEA estimates of industrial output. 48 The Forest Service, a division of U.S. Department of Agriculture, creates another set of updated input/output tables for its own use (IMPLAN). 49 Even the Department of Commerce, the source of the input/ output tables, contracts out to consultants for updated input/output tables. 50 In an era when computers are easier to use and vastly more powerful, delays in creating input/ output tables should be growing shorter. Significant problems remain in data communication. Until 1985 the massive amounts of data from the Bureau of the Census needed to construct the input/output accounts were delivered to the Bureau of Economic Analysis in printed form. Every number had to be reentered by hand and rechecked. In the last few years computer tapes have been delivered to BEA. Unfortunately the data are still not in the form needed by BEA, but instead, are simply a digital representation of the pages that formerly appeared in printed form. The tapes contain tables with lines, headings, and notes, which must be removed in a laborious process. After the irrelevant characters are removed, the arduous process of converting the data to forms useful for input/output work can begin. As the process proceeds, it is often discovered that some data items may not be provided by Census for reasons of confidentiality. In part because the Census does not have money set aside for retabulation, BEA can seldom get additional information from the Census Bureau (e.g., data aggregated in a way that does not reveal confidential information). Other nations manage to produce detailed input/ output data much faster than we dothough possibly with less accuracy-in part because 1/0 plays a more fundamental role in policy making in these countries. Japan already has input/output tables based on data collected in 1985. And the Japanese government is involved in creating an input/output model for a major portion of the international enconomy. 51 (See box E for a description of this effort.) The United Kingdom has a benchmark table for 1984. China is about to complete a 1987 table. 52 Computing Business Output by Sector Each business in a network delivering a final product or service to a consumer adds some value (value-added) to the product. The sum of value-added in all businesses in the United States equals the GNP. The input/output tables described above provide the basic tool needed to see how much value each business in the economy contributes in constant dollar terms.53 The value-added by a business can be computed by subtracting the value of all products (both goods and services) purchased by a business from total sales (or gross output) .54 In this sense, value-added is a better indicator of performance than sales (gross output) because it shows the contribution made by the company-not the aggregate value of the companys contribution and the value of inputs produced by suppliers. An automobile company can, for example, decide to purchase components from abroad instead of producing them internally. Its sales 4SF~r~x~Pl~, ~ BLS 1982 ti~of ~souqut fortheoffice, computing, Snd ~uting mwhines industry (SIC 3572,3573,3574,3576, d 3579) was over a billion dollars less than the BEA 1982 estimate. Other discrepancies occur in industries such as eating and drinking places (SIC 58) and non-metallic minerals, except fuels (SIC 14), but exact SIC matches between the two series are difficult. U.S. Department of Commerce, Bureau ofi%onomic Analysis, Survey occurrent Business, Annual Input-Output Accounts of the U.S. Economy, 1982, April 1988, p. 35, and U.S. Department of Labor, Bureau of Labor Statistics, Ofilce of Economic Growth, Output and Employment Database, January 1988. A similar discrepancy occurs between Federal Reseme Board estimates of production (the Index of Industrial Production) and those issue-d by Department of Commerce (sales adjusted for inventories). See Jeffkey A. Miron and Stephen P. Zeldes, Production, Sales, and the Change in Inventories: An Identity that Doesnt Add Up, Working paper No. 2765, National Bureau of Economic Research, Cambridge, MA., November 1988. 4!l~1f E. Siv~d D~el E. Chww]le, *A com~son of ~tu~ c~~s in Employment ~d kcornc With predictions Using IMPLAN Models, presented at the Western Regional Science Association, Feb. 19-22, 1989, San Diego, CA, p. 4. S~swr A. Davis, ~rnbutions of Exports to U.S. Employment: 1980 -1987, U.S. Department Of commerce, ~~rnatioti Tr~ ~inis~tion! Trade Research Division, Rojea DTR4)14-89, March 1989, p. 22; and Ken Young, Ann Lawson, and Jennifer Duncan, U.S. Department of Commerce, Office of Business Analysis, Trade Ripples Across U.S. Industries, January 1986, p. 9. SIM$ Sam, J~s Miniof ~t~~m~ Trtie ~d ~u~, compilation of ~ kt~tion~ hput-output lhble, paper presented at the OECD Workshop on Intonational I-O Tables and performance Analysis of Structural Adjustment, Dec. 14, 1988, Paris, France. 52K.R. po]~e, Relev~ceof U.S. Re@on~ St~stics,paPr pre~nt~ at~~eric~ A~iationfortie Advancement of Science, Jan. 19,1989, p. 7. ssGross ~Xt ~ l.nd~~: comments on Recent Criticisms, op. cit., footnote 31, p. 132. ~N~on~ ~~y of ScienWs, tit= ~ N~o~ Stmstics, Me~ure~~ ~ /nterPret@n of pr~~ffv~, (W~iXtgtort, ~: National Academy of Sciences, 1979), p. 65. 20-014 0 89 2
PAGE 25
18 B OX EJapans International Input/Output Model Project In an effort to build an analytical capability for analyzing bilateral and multilateral economic issues and conflicts, analyzing the economic impact of international economic activities, and clarifying the magnitude of international interdependence, the Japanese are engaged in creating an international input/output table. At the cost of a million dollars a year for 6 years, the Ministry of International Trade and Industry (MITI) directed project will construct a series of I/O tables that connect the economies of Japan, the United States, the United Kingdom, France, the Federal Republic of Germany, South Korea, Malaysia, Singapore, Thailand, the Philippines, Indonesia, China and Taiwan. 1 Its scheduled to be completed in 1992. Because the model is based on 1985 data, data for the United States had to be estimated by a private contractor. Similarly, since the United States does not have data on the use of imports by industry-an important characteristic of the model-the Japanese had to estimate it by surveying Japanese firms about which U.S. industries buy which type of goods from them. In the case of the developing Pacific Rim countries who in some cases lack a strong statistical system, the Japanese are collecting and organizing the data as a part of Japans foreign aid to those countries. The Japanese have stated that they will use the input/output tables to: make international comparisons of industrial structures, l evaluate the results of a given countrys protectionism, determine the economic effects of direct overseas investment, analyze the impact of changing crude oil prices, and evaluate the effect of fostering the development of new industries in specific countries. 2 Obviously, the data could bean important competitive tool in identifying and targeting key industries. As the world gets increasingly carved up into large trading-blocks (i.e., U.S. and Canada and Europes 1992 agreement), the input/output tables could also be used to evaluate the costs and benefits of a Pacific Rim trading agreement. llt ~ ~~ ~~ ~ ~ Jap~ ~ $1 million ~ yon ~S intmtiml effort which is a supplement to Iheir main domestic effort, Lhe 1989 fkd year budget for the cmtire U.S. inputbtttput effat (BEAs Interindustsy Division) was $1.4 milliom %. Sate, Japans Ministry of Intemadonal Trade and Industry, Compilation of an International Input-output Table, paper presented at he OECD workshop on IutuMtt OMI I-O Tables and Pedamuscc AnaJysis of Structural Ac@strmmt, Dec. 14, 1988, Pans, France, p. 2,3. could remain unchanged while its domestic valuediate inputs in complex ways. On close examinaadded and employment declines. (In fact, about 70 tion, there are deficiencies in much of the data that percent of the value of the components of Chryslers are currently used. The largest of these include vehicles are manufactured by outside suppliers. 55 ) problems with: Similarly, the company could decide to purchase advertising, payroll accounting, software developl ment, and other services from specialty firms instead of doing this work with internal staff. This could also leave total auto sales unchanged while reducing value-added and employment in the automobile l industry. Value-added in constant dollars can be computed l by subtracting an industrys intermediate inputs from its output after both have been adjusted for inflation. Doing this with precision obviously redata showing the links connecting different kinds of businesses (including data showing what businesses purchase from other businesses) the way products and services are adjusted for inflation (many of these problems have already been discussed) the way imported intermediate inputs are treated (this issue is treated in greater detail in the next section on international trade). quires an enormous amount of data. It also requires Data limitations in these areas make it extremely high quality data for the entire economy since difficult to trace national output to specific busibusinesses depend on each other through intermenesses with much precision--particularly when it is sSKeVi~ ~~~, fici~ ~~~g by ~ U.S. A~~obile ~us~, report prepared for the Congressional Research Service, Washington, DC, Nov. 8, 1985, p. 2.
PAGE 26
important to remove the effects of inflation. In many cases it is necessary to use gross output or sales instead of value-added in analysis of specific industries; this is the case with estimates of industrial productivity. 56 Adjusting for Inflation In principle, the value-added in each type of business can be computed using input/output tables. The inputs purchased by each business are deflated separately and the total deflated value of intermediate inputs is subtracted from a deflated level of industry sales (or gross output). This technique, called double-deflation, is recognized as a preferred method by the Department of Commerce because of its use of a consistent set of price indexes. 57 The problem of adjusting total GNP accounts for inflation were discussed earlier. Developing deflators for value-added by industry (also referred to as gross product originating) compounds the problem since much more detailed and extensive information is needed. Services are a major problem since about two-thirds of all services are purchased by businesses as intermediate inputs and not by households or the government as a final product. 58 In practice, however, incomplete data mean this technique can only be used in 29 percent of the 1986 GNPthe manufacturing, farm, and construction sectors. 59 A variety of scaling techniques and other methods are used for the rest of the economy. 60 Again, services are the most difficult area. The intermediate inputs of service industries are very poorly documented from survey-based series. 61 For many kinds of services, no direct data on intermediate inputs are available. Even after the current dollar data are compiled, adjusting for inflation is difficult because of the lack of adequate price series for services. BEA calculates deflators for several service sectors by extrapolation using jobs as a proxy for increases in quantity. By definition, this means that no labor productivity growth can occur, resulting in an overstatement of the rise of prices in this industry and a subsequent underestimate of the quantity of services purchased as inputs. 62 The underestimate of inputs means that the value-added is overstated. 63 Estimates of value-added in construction suffer both from poor estimates of total industry output in constant dollars (for reasons discussed earlier) and from poor estimates of inputs. Since there is no detail on prefabrication by construction suppliers, inputs are double counted. For example, inputs such as wood are counted once when they are purchased by a business making prefabricated (pre-hung) doors and are counted again for a second time when the prefabricated doors are purchased as an input by a construction firm building a house. 64 This results in an over counting of inputs and subsequently an underestimate of constructions value-added. BLS has a deflator series for the gross output of a number of service businesses (some not completely incorporated into the BEA accounts) and has research projects underway in a number of areas (communications, semiconductors, computer programming services, medical services, and banking). Double-deflation is also limited by the fact that deflators for gross output are available only for 72 percent of the GNP (table 3). BLS computes deflators for a number of service industries not presently used by BEA. The services for which precise deflator series are available, however, are primarily those where output is comparatively easy to quantify (e.g., electricity). In other service sectors attempts are made to count or quantify output using measures such as the number of checks processed or %Me~~e~# & Interpretation Of Productivity, Op. Cit., fOOtnOtC 54, P. 67. 57~1(3 ~~, Grosa Roduct by Industry, Survey of Current Buriness, vol. 67, No. 4, April 1987, p. 27. SSU.S. ~p~at Ofcomerce, B~~ of &onomlc ~~ysis, ~~rvey ofc~~renfB~i~ess. 1984, mC InpUt-GutpUt Structure of the U.S. Economy: 1977, vol. 64, No. 5 (hhy), pp. 42-84. 59c~c-~ ~m N~on~ ~ome ~ ~uct &XXXIMS, op. cit., foo~om 15, table 6.1. @Peterson, op. cit., foomote 57, p. 27, 61~ ~me ex~nt ~ is a rat ad growing probl~ Since a good pofiion of intermti~ ~rvke kput data usd to come from records kept on regulated industries that have been or are being deregulated. ~~mn, op. cit., footnote 29, p. 52. s3Grosa Product by Industry: Comments on Recent Criticisms, op. cit., footnote 31, p. 133. 64BA1Y ad Gord~, op. cit., fOOtnOU 11 ~ P.
PAGE 27
20 Table 3BLS Deflator Series for Gross Output Percent of GNP covered Gross output deflators based on price indices . 72 Manufacturing Mining Some Services Gross output deflated by an index based on 6 hours or cost indexes . . . . . Nonresidential structures etc. No BLS deflator series. . . . . . General government . . . . . (10) Owner-occupied housing . . . . Rest of the world . . . . . . (1) Households and institutions . . . . (4) NOTE: Totat may not be 100 pement due to rounding. SOURCE: J.A. Mark, Problema Encountamd in Measuring Singleand Multtfdor Produetivlty, *fh& LdXW /?ev&w, Deeember 1-, p. 5. the number of operations performed. 65 But these techniques have clear limitations in times of rapid technical change. 66 Adjusting for TradeAnother difficulty in assigning value-added to different industries is the fact that the current input/output tables do not distinguish between imported and domestically produced products used as intermediate inputs. Imported inputs are recorded as a category of final demand. This accounting convention is adequate for aggregate measures of GNP. But without additional information the data cannot be used to track many important effects of international trade on the U.S. economy-such as the way trade affects the output of different kinds of businesses. First, BEA uses the deflator for domestic industries to adjust intermediate inputs purchased even though some of these inputs are imported. 67 This had the effect of underestimating the value of intermediate inputs when the dollar was strong in the mid1980s and underestimating their value when the dollar weakened. During the mid 1980s, using faster rising domestic price series on imported inputs, the level of intermediate inputs was overstated and the level of value-added or output was understated. 68 BEA estimates that correcting for this problem would drop the growth rate of manufacturing by half a percent or more per year from 1979 to 1985. 69 That correction is nearly equivalent to eliminating the increase in output of the electrical equipment industrythe second largest contributor to manufacturings growth over the period. 70 Second, some products purchased intermediate inputs by businesses are in fact products produced by foreign subsidiaries. In many cases U.S. components are shipped abroad for assembly and then reimported for final testing and sales. This kind of production is encouraged by Items 806 and 807 of the U.S. Tariff Code requiring duties only on the difference between the value of components exported and products imported by a multinational firm. The Census Bureau reports that many firms fail to report inputs received from overseas affiliates as costs of business. 71 As a result, the value of the input gets credited as value that was added domestically, overstating the true amount of U.S. production in that industry .72 Although this problem has always existed, the upsurge in intermediate inputs coming from foreign sources makes this a growing problem. And potentially it is a large problem. In 1985, nearly a third of all U.S. exports were exports from U.S. companies to overseas affiliates. Over a fifth of all imports to the United States came from these overseas U.S. affiliates. 73 For some industries like 65J~e A. Mo~~ ~~tivi~ in ~i~ ~~~es: ~ BLS Expakmce, paper presented at the Symposium on Technology ~d tk services Induntrtca Hanover, NH, August 1987, p. ~. 66Fm ~-p]e, ~ nw~ of ~=tions p~~ fails to include developments such as increases in branch btiing m~ ~ssible ~u@ ~ widespread uae of automatic teller machines (ATMS). Baily and Gordon, op. cit., footnote 11, pp. 399-400. 67~ ~~m ~ ~is occurs in he defl~on of ~~eum Where the imported price iS used. % decline in the vak of the dollar since 1985 should have the reverse effect of underestimating the change in manufacturings output. w~m -t ~ ~d~: COIIMXICIUS on Recent &hicisms, op. cit., footnote 31, p. 132. 7~me1~~ mew, whi~ ~cl~ ~ cmw~r indu~, wss the largest contributor. National Income and product Wounts, OP. Cit-, footnote 15, table 6.2. 71~ d~~g ~f~ ~ is tie c~~ of ~uf~~m. ~ JohtI P. @VO1d, Possible Improvements in Industrial Statistics for 1987, paper -M to tk MtiSOV Cmmittee of the Americart Economic Association and the American Marketing Association at the Joint A&iwry committee, Oct. 9-1o, 1986, p. 5. 72~~. 73J~ ~~ ~~~ ~ T* Num~~ Hi~, ~ross t~ hard, October 1987, p. 12.
PAGE 28
21 1 6 10 6 0 6 -1 0 -16 -2 0 Figure 4-Exports and Imports (percent of GNP in constant 1982 dollars) Percrnt of GNP 1 u I i I 1 I I 1 1 I I 1 I I I i I I I I I I I I I I I I I I I I I I I I [ I I I 1960 1966 1960 1966 1970 1976 1980 1986 SOURCE: U.S. Department of Commeree, Bureau of Economc Analysis, National Income and Product Accounts, tables 1.2 and 4.4. semiconductors where potentially half of the value of the chip is added overseas in the assembly process, this could be an especially important problem. 74 C. What Is the Impact of International Trade on Domestic Producers and Consumers? The speed by which the U.S. economy has become immersed in international trade and accumulated a huge trade deficit is a striking example of an economy in transition. The share of GNP held by exports has grown by almost a third from 1970 to 1986; import penetration increased by 76 percent. (See figure 4.) From the end of World War II to 1983 the amount of imports purchased on a per capita basis (real 1982 dollars) slowly crept from $300 to $1,500. By 1987, the amount was $2,300----a 50 percent increase in 4 years. This phenomenon is not limited to a few select products but now affects virtually every industry. Global production networks have redefined the nature of trade. Direct foreign investment 75 by the U.S. investors overseas increased by 43 percent between 1980 and 1987; foreign investment in the United States tripled during the same time. 76 There has been particularly sharp growth of worldwide production associated with direct investment leading to an upsurge in intrafirm trade where a 7~ovoni, op. cit., f(XXllOtC 7 ] P. 3 75~fm~ ss owning 10 percent or more of an enterprise. 76u~ome]y, ~is &ta ~ ~v~l~le only in Cment ~IIMS, a Shortcoming ~SCUS~ ~low. !$CX J~eS K. Jackson, American Direct Investment Abroad, Congressional Research Service, Aug. 8, 1988, p. 2.
PAGE 29
22 division of a multinational corporation trades with another affiliate. 77 It is obviously important that we understand how foreign trade influences the performance of the domestic economy. For example, the effects of trade on jobs, inflation, and our dependence on foreign sources, are the root of many policy decisions. While data about international trade have improved significantly in recent years with the development of improved reporting and good deflator series, more improvement is possible. The absence of any detail about trade in the input/output accounts, for example, means that an uncomfortable number of assumptions must be made to see which industries are affected directly and indirectly by changes in the U.S. trade position. 78 The explosive growth in the volume and complexity of trade makes this a major challenge. And there is reason to believe the quality of trade statistics may decline in the next few years. Trade data are largely a by-product of data collected for administrative or regulatory purposes .79 The administrative need associated with trade data is the assignment of duties, tariffs, or quotas on particular products from particular countries. But increasingly, trade between countries have become more free, undermining the motivation of Customs to collect data on those products. 80 From 1970 to 1986, the share of imports (based on value) subject to duties has fallen by 50 percent. 81 Data on U.S. trade with Canada (about 20 percent of all U.S. trade) may become very unreliable when tariff barriers between the two nations are removed. Data on trade with individual members of the European Economic Community (EEC) may be difficult to obtain once internal EEC tariff barriers are removed at the end of 1992. 82 New mechanisms for collecting trade data must then be found to prevent the quality of trade statistics from declining. Some form of international cooperation on trade statistics is likely to be desirable (possibly essential) to prevent a major deterioration in the quality of data. Tracking Trade Volume When the number of export and import documents more than doubles over a decade, reaching 15 million in 1987, just reporting the level of exports and imports becomes a daunting task. 83 This huge jump in the volume of transactions has contributed to deficiencies in the data. For instance, in the mid-1980s the dramatic increase in imports made it impossible for the government to collect accurate monthly trade statistics. So much time was required to process trade data that in 1985 anywhere from 35 to 53 percent of the official import statistics in any single month actually represented a carryover from previous months. 84 Monthly trade data are, of course, also needed to measure GNP accurately. The discovery of one carryover forced the estimate of growth in the gross national product (GNP) during the last quarter of 1984 to be revised downward-from 4.3 to 0.6 percent. 85 Another revision of GNP tied to trade data occurred in the second quarter of 1986 where the new estimate reversed the direction of growth from a positive rate to a negative one, resulting in the first quarter of negative growth in the current economic expansion-a fact that went largely unnoticed. 86 Obviously, revisions of such magnitude can send a nJ~ S* h~c, ~@a-Fi~ TrA: h Update, The New England Economic Rw2?w, May-June 1987, p. 47. 78w K. Yomg, ~ Law~n, md Jennifer Dman, Trade Ripples Across U.S. Industries, wotking paper, U.S. DePment of c~e~t ffi= of Business Analysis, Januaty 1986, pp. 57-61, or ttte appendix of Technology and the American Econornic Tramition, OP cit.. foomo~ 1 v for ex~Pl~ of assumptions employed to circumvent this ~bletn, pp. 469470. 79N~on~ ACAY Of ~ie~es, Coanmhtee on National statistics, proposal for a Panel Study on Foreign Trade Statistics, July 1987, p. 6. ~eport on me Working Group on h Quality of Economic Statistics, op. cit., footnote 6, part 1, p. 10. S]U.S. ~pmat of ~~eme, B-u of tie -u, s~fi:icuf Ab~tract of the u@ed ~tu~e~, ]9&7, p. 778, ltile 1351, ~ @ant W. Gardner and Kent P. Kimbrou~ The Behavior of U.S. Tariff Rates, The American Economic Review, March 1989, vol. 79, No. 1, pp. 212-214. sxJ~ph W< ~cm, l ~e s~ti~cs @er: S@tiCS and 1992 in Europe, Business ECOnOrrUCS, JIIIY 1989, pp. 52-53. 83U.S. Congtlxa, General Accounting Offke, Merchandise Trade Statistics: Some Observations, GAO~CE-89-lBR, (Washington, DC: U.S. Governtnent Printing Office, April 1989), p. 40. ~Slater, op. cit., footnote 5, PWt 1. P. 53. Ssslwr, op. cit., foomote 5, part 1. p. 53. 86 U S ~mat of co-=, B~u of ~~ic ~~ysis, Sumey Of c~re~ f?~i~ss, July 1988 and July 1986, table 1.2.
PAGE 30
23 false signal to policy makers about the strength or weakness of the economy 87 The Bureau of the Census is aware of these shortcomings and improvements are gradually being made by automating reporting and reducing carryovers by releasing trade figures 45 days after the close of the month. By July of 1988, import and export carryovers had been reduced to 5 percent of the value of imports and exports. 88 But nagging problems remain. An illustration is provided by looking at the gaps in data between the worlds two largest bilateral trading partners: Canada and the United States. U.S. exports to Canada are consistently underestimated. 89 In 1986 the U.S. trade deficit with Canada was revised downward by 42 percent from $22.9 to$13.3 billion. 90 Since June 1987, U.S. exports to Canada are based on the more relaible Canadian reports of imports from the United States. 91 These data are likely to become less accurate if, as is likely, Canadian interest in precise tracking declines in a free-trade regime. A report by the General Accounting Office (GAO) on Merchandise Trade Statistics pointed out many other discrepancies in trade data. Over the last 13 years GAO found that U.S. export data consistently fell short of foreign import data. 92 Overall, the shortfalls rose from $5.7 billion in 1975 to $18.7 billion in 1986, falling to 13 billion in 1987 after the use of Canadian data was implemented. 93 The report concluded that: there is a strong possibility that U.S. exports are not fully counted; as a result, the U.S. merchandise trade deficit possibly has been overstated for the past several years. Trade in Production Networks The growth in investments around the globe is indicative of the emergence of worldwide production networks. For example, the assembly of the Ford Escort involves glass from Canada, fan belts from Denmark, radiators from Spain, steering wheels from England, and tires from Norway. In all, parts are made or assembled in 16 countries on three different continents. 95 In 1985, more than a quarter of all Digital Equipment Corporations (DEC) sales were sales to its overseas affiliates. % These global production networks transcend the idea of the sovereign state, obscuring the notion of what is 1313upposedly, ti overly optil~lc early estimate of 1984 GNP fooled the Federal Reseme Bank into limiting the expansion of the money supply, helping to push the value of the dollar to its 1985 high mark. Testimony of Courtenay Slater before the Joint Economic Committee. Slater, op. cit., footnote 5, part 1, p. 94. EEMerctie Tr~e Statistics: Some Observatwns, op. cit., foomote 83, p. 22. 89s1a~r, op. cit., foomote 5* X 1 ~ P 56. %lational Academy of Sciences, Committee on National Statistics, Proposal for a Panel Study on Foreign Trade Statistics, p. 6. g~Merc_e Tr~e Stattitics: Some Observatwm, op. cit., footnote 83. p. 42. 92w q~y hcjdd the Un]tti !Natcs m@r trading partners: Canada, Japan, West Germany, France, The United Kingdom, and the Netherlands. The Netherlands was the only country where U.S. exports were not consistently short of recorded imports from the United States. Sec h4erchundise Trade Statistics: Some Observatiaw op. cit., footnote 83, p. 30. gJMerc-e Tra& st~~tics, Some Observations, op. cit., footnote *3, pp. 30-32. g4Me?cWe T?~e Statistics: Some Observations, op. cit., foomo~ 83. p. 2. 95Bq BIUCSWM and Bennett Harrison, The Deindustrialization of America (New York, NY: Basic Books), 1982, p. 177. %Hein, op. cit., footnote 73! P. 12
PAGE 31
24 meant by foreign trade. 97 It is as if GMs sourcing of parts from a division in New York to an assembly plant in Michigan was considered international trade. Certainly, international, intrafirm transactions are a much different variety of trade than what is connoted by the traditional meaning of trade. This changing nature of trade requires a reexamination of how the statistics are collected, presented, and interpreted. For example, when the International Trade Commission (ITC) tries to calculate the import penetration rate of the computer industry, how should it treat the large intrafirm imports from DECs Scotland plant to its Maynard, Massachusetts plant? Is this a threat to the U.S. industry? Should the sale price of a Chrysler Mini Van complete with a Mitsubishi engine (24 percent of Mitsubishi is owned by Chrysler) be counted as domestic production? 98 Tied to this is the question of whether or not these intrafirm transactions are being priced at the armslength market rates or at internal, transfer prices? The Internal Revenue Service under section 482 of the Internal Revenue Code requires that these transactions be valued at market rates, but in many cases the product being sold is a unique component that does not have a corollary in the market. This problem is even more vexing in the cases of intangible, intermediate service sector products. The BLS does not include intracompany transfer prices in its international price series, preferring instead to include only a subset of intrafirrn prices designated as being arms-length. In any case, if in fact these intrafirm transactions are not priced at market rates then our knowledge about the true level of imports and exports is diminished further. Existing data makes it difficult to trace the effects of imports or exports through the economy. Making some reasonable assumptions about how trade is used as an intermediate input, it is possible to see that imported commodities affect businesses throughout the economy (figure 5). 99 Without more precise data, however, it is impossible to make firm estiFigure 5-imports Used to Produce Amenity (directly and indirectly) Transportation Clothing & pers care Federal defense Exports Recreation leisure Food Housing Government (n.o.c.) Health Personal buo. & comm Education o 5 10 16 20 26 Percent of spending on unonlty SOURCE: U.S. Congress, Offke of Technology Assessment, Tmhr?ology and the American Economic Transition: Choices for the Future, OTA-TET-283 (Washington, DC: U.S. Government Printing Office, May 1968), p. 296. mates of the problems that might be created by a sudden change in trade--such as a change in the import quotas of steel on the U.S. auto industry, a disruption of the oil supply because of a war in the middle east, or the effect low-priced or dumped semiconductors would have on the U.S. computer industry. It is even difficult to determine the vulnerability of the Department of Defense (DoD) to import disruptions since DoD purchases products are at the end of long and complex production networks. 100 The Census Bureau is attempting to ask more detailed questions about imports used in production, but in many cases the establishments simply do not know the national origin of the products they purchase since the products have passed through many middlemen. Direct Investment Shifts in the Capital Account the value of U.S. assets owned overseas net of foreign owned assets in the United Statesare also distorted by a failure to s= ROlXII B. Reich, Mcm&s Only, The New Republic, June 26, 1989, pp. 14-18. 98 Autmotive News, 1987 M~ket Data B~k ~ cit~ in office of ~ho&y Assessment, Technology @d the American Econor?dc Transition, Op. cit., footnote 1, part 1, p. 326. 99s= Tec~fo~ ad the krican Economic Transition, op. cit., foomote 1, pm 1, p. 469. 1OOs= us, ConmSS, offIce of ~hnolo~ As~ssm~nt, Holding the Edge, OTA-ISC-421, Aptil 1989.
PAGE 32
25 Figure 6Investment In Information Equipment (percent of all producers durable equipment) 4 0 90 20 10 0 1960 1966 1960 1966 1070 1976 1060 1906 SOURCE: U.S. Department of Commerce, Bureau of Economic Analysis, National Irteome and Product keounts, table 5.7. adjust for changes in value over time. 101 Direct investments by the United States overseas and by foreigners in the United States 102 are valued at book value-the price paid for at the time of purchase. No attempt is made to adjust for inflation. Since 70 percent of U.S. direct investments owned overseas were purchased by 1980 at relatively low prices while over two-thirds of foreign direct investments in the United States were acquired after 1980 at higher prices, the net position of the United States is greatly underestimated. 103 Estimates indicate that the failure to adjust for inflation undervalues U.S. overseas investments by $200 billion to $400 billion. 104 In terms of policy, this undervaluation means that the rate of return associated with those foreign assets has been grossly overestimated and the alarm over the United States becoming the largest net debtor in the world might be misplaced. 105 In addition to a more accurate valuation process, researchers argue that more detailed data on inputs, type of labor employed, and layoffs are needed for a more complete analysis of the effect of foreign direct investment on the U.S. economy. l06 D. What Capital and Labor Inputs Are Purchased by Domestic Producers? Up to this point this discussion has focused almost entirely on different ways of measuring the output of the economy and how to assign this output to the activities of different kinds of domestic products. It is important, however, to determine the extent to which the economy is growing only because of increases in the amounts of labor and capital and the extent to which growth which can be traced to better management or use of technology. Measuring the productivity of the economy, or output per unit of input, requires good information about both inputs and outputs. Rapid economic changes have always been associated with major changes in capital investment. While investment in steam engines, railroads, telephones, and electric generating facilities dominated earlier periods of economic change, the most critical investments in the present period seem to involve information and intelligenceboth human and artificial. It proves difficult to measure either kind of input. A sharp growth in purchases of information equipment is documented in the income and product accounts. Over 40 percent of all business investment in durable equipment is now in information processing equipmentdouble the 1979 share and four times the 1972 share (see figure 6). It is difficult to know how precise these measures are. The loi~ Norm~ j. G]ickm~ ad hugISS p. Woodwud, The NW Competitors (New York, NY: Basic Books), 1989, Appendix A; Roben E. Lip=y, Changing Patterns of Internatiomd Investment In and By the United States, National Bureau of Economic Research Working Paper No. 2240, May 1987, p. 47; and Measuring the U.S. International Investment Position, Survey of Currerat Business, June 1989, p. 40. l~D~t inves~ent ~ def~ed ~ the ownership, acquisition, or establishment directly or indirectly by a person-individual association, corporation, government, etc. of 10 percent or more of the voting securities of a foreign enterprise. See James K. Jackson, American Direct Investments Abroad: How Much are hey Worth? Congressional Research Service, July 25, 1988, p. 1. mJames K. Jackson, American Direct Investments Abroad: How Much are They Worth? Congressional Research Setvice, July 25, 1988, p. 2. l~Jackson, op. cit., footnote 103, p. 10. lC6Jac~, op. Clt-, fmmae 103, p. I and p. 6. when forms of international investment other than direct investment such m go~dt b~k loans, ad h of which have valuation problem~ securitie~ reviewed for measurement problems, the Department of Commerce contends that it is likely that the aggregate international investment position has shown a substantial decline reflecting the large cumulative U.S. current account deficit. &e Messurin g the U.S. International Investment Position, Survey of Currenr Business, June 1989, p. 40. l~ljc~m and Woodward, op. cit., footnote 101, p. 280. lmN~mSI ~~tne and product Accounts, op. cit., foomote 15, table 5.7.
PAGE 33
26 difficulty of measuring the constant dollar value of computer purchases, however, has already been discussed. While the inflation adjusted value of the capital equipment purchased may be difficult to measure with precision, it seems likely that a significant change in the composition of what is bought has occurred. Such a large shift over such a short time period means that not only has the composition of capital investment undergone a dramatic change, but also the nature of the firms that are buying the equipment. Increasingly, industries not typically associated with capital equipment, like insurance, banking, and retailing, are becoming large users. In 1982, department and grocery stores bought more computers and peripheral equipment than the aircraft and guided missile industry .* These large investments in information processing equipment, all along the chain of production, are indicative of an economy becoming more responsive to the needs of customers and the challenges of competitors. 109 Accounting Conventions Changes in the economy have reopened an ancient debate about whether to count investment in the education and training of people in the same way that physical capital is treated. Workers with good educations, and skilled in adapting to new tasks, are the most important inputs for most businesses. The traditional national income and product accounts, however, continue to treat education as a form of consumption even though economists have long understood it as a form ofinvestment.11 Similar problems are encountered in government spending. Unlike accounting conventions used for private businesses, government purchases of capital goodsroads, airports, public buildings, and other facilitiescontinue to be treated as current government expenditures (current accounts) and not investments (capital accounts), even though many of these facilities obviously provide a critical infrastructure Figur87-ProductMty (output per hour) 100 SOURCE: U.S. Department of Labor, Bureau of Labor Statmtke, Fkw?dbod( of Labor SkNisfics, Bulletin 2217, June 19S5, table 91. for economic growth. Since government investments are excluded by definition, figures on net investment, the stock of capital, and governments role as an investor in the economys infrastructure and a holder of assets is obscured. This limits analysis that attempts to understand the links between types of government investment and economic growth. The United States is the only major advanced economy that doesnt separate its government expenditures into current and capital accounts. 111 Although these accounting conventions are not new, their existence has drawn more attention in an era of large budget deficits. According to one estimate, the amount of useful capital investment made by the government is roughly equal to half of the $155 billion annual Federal deficit. 112 The role of Federal Government investment in promoting economic growth also takes on increased importance as international competitors explicitly base economic policies on government investment strategies. Industry Purchases of New Capital Equipment Statistical problems are encountered even using the most conventional definitions of capital. The 10SU.S. ~-ent of Cmeme, Bureau of Census, Gener~ RcpO~ on lndusfriu/ Organization, October 1986, table 8, p. 294. 1~=11 Johnqon ~d pad R. Lawrence, Beyond ve~c~ Integrati~-~e Ri~ of tie Hue-Adding p~ershlp, The Harvard Business Review, hdy-AuguSt 1988, pp. 94-101. 110H~ bvin, Mapping the Economics of Education, Educarwnal Researcher, May 1989, pp. 13-16. 11 l~c~l J. ~skjn, ~mtic~ ~d ErnplfiC~ ISSUeS in the M~urernent, Evaluation, and Interpretation of postwar U.S. !$aving, in &rvings Ond Capiral Forrnano n, Gerard Adams and Susan Watcher, eds. (Lexington, MA: hxington Books, 1986), p. 19. 1 IZRoben Elsner, *Divergences of Measurement ~d ~eory ad Some Implications for &onomic policy, The American Ecorwrrtic Review, Mach 1989, p. 5; see Charles L. Schultz.e, Of Wolves, lkrmites, and Pussycats, The Brookings Review, Summer 1989, for an opposing view, pp. 26-33.
PAGE 34
27 most recent statistics providing a comprehensive estimate of which businesses purchased which types of capital equipment (the capital flows tables) date to 1977.113 Sometime in 1990, the 1982 capital flows tables will be released, but given the peculiarity of that recession year, the usefulness of the data on capital investment will be limited. Even if 1982 were not a recession year, the fast pace of change in capital investments that has occurred since 1982 leaves a poor understanding of how the economy has evolved. For example, from private sources we know that the number of personal computers in the workplace has jumped from 1.2 million units in 1981 to nearly 13 million in 1986, but we dont know which industries are using them or how .114 Recent criticisms that U.S. industry has not concentrated on the process side of production 115 and has hollowedout 116 need to be analyzed using data that reflect which industries are purchasing what type of capital goods and how this equipment is being applied. Recently, the Bureau of the Census has produced a report that attempts to rectify this situation for the manufacturing sector and goes further by breaking the data down by size and age of the firm, but there is no provision to make this an ongoing survey or to extend it to other sectors. 117 Interestingly, the main funding source for this survey was the Defense Logistics Agency of the Department of Defense. A one-time snapshot is useful, but it is static and it does not begin to answer questions about how investments in capital goods change in an economy that is increasingly dynamic. Capital Stock These issues deal with accounting for new purchases of capital equipment; another, possibly more fundamental issue, involves accounting for the existing stock and vintage of existing capital. Currently, estimates of capital stock are derived from a perpetual inventory system that assumes an average useful life, a set retirement distribution, and a particular efficiency rate. 118 Estimates of the existing stock can be calculated by knowing what was purchased, when, and how fast it depreciates. By and large, service lives and discard rates are based on educated guesses and tax law provisions. 119 Except for nuclear fuel, railroad equipment, and autos, the service lives for private equipment were estimated from industry studies conducted during the 1970s. 120 These estimates work reasonably well in periods of economic tranquility. But unexpected events like the two oil shocks, 121 the severe recession of the early 80s, 122 the advent of fierce foreign competition, and the arrival of new types of capital equipment, like computers, have undoubtedly changed the efficiency, depreciation, and discard rate of equipment and have caused a radical realignment of relative prices from what prevailed in the early seventies. 123 The rapid obsolescence of capital equipment used in the design and creation of semi-conductors is an example of an important industry that barely even existed in the early seventies. A periodic rebenchmarking of the exiting capital stock would reanchor the perpetual inventory system and allow a fine tuning of the perpetual 1 Jerry Silverstein, New Structures and Equipment by Using Industry, 1977, Survey o~Currenr Business, November 1985, pp. 26-35. 1 ldFu~ Cmputing k. cit~ in tie U.S. ~-ent of commerce, Bureau of the Census, StattiticalAbstractof the United States: 1988, table 1286, p. 726. 115~c~l L. ~~m ~ Mchad K. ~er, ROIXXI M. S O 1 OW and the MIT Commission on Industrial productivity, Made in America. Regairdng the Productive Edge, (Cambridge, MA: The MIT press), 1989. 11% tem hollo#+ut* refers to the prxtice of contracting out for finished goods and setvices that used to be produced internally, reducing the function of the company to primarily marketing and distribution. The Hollow Corporation, Business Week, Mar. 3, 1986, pp. 57-85. I ITU.S. Department of commerce, Bureau of Census, Man@cruring Technology Z988, SMT(88)-1, May 1989. 118cha]eS R. H~ten, ~ Me~~rnent of &plt~, fo~cornlng In tie proc~ings from t~ Confennce on Research in kome and Wealth, Washington, DC, May 12-14, 1988, edited by Ernst Bemtt and Jack Tnplett, July 1988, p. 44. 119~w~ ~, -. Dmou@, ad MW N~f, Altemat.ive Measures of Capital Inputs in Japanese Mmufxturing, foficoming in A Comparison of Productivity in Japan and the United States, Nationzl Bureau of Economic Research, (University of Chicago Press), p. 3. 1~.s. ~p~mt of Cmeme, B~u of &onomic ~~ysis, ~~ed Preproducibfe Tangfi/e Wealth in the United States, 1925-88, (Washington, DC: Government Printing Offk.c, 1987), p. xxi. 121s= MN. B~ly, pr~uctivlty ~d he Servims of Capital and Labor, The Brookings p@erS on ECOnO@c Acrivify, vo~.1, 1981. 122s= sum G. Powem, The Role of Capiti ~sc~d~ in Multif~tor Productivity Me~~ement, Mo~~y Mor Revi~, June 1988, pp. 27-31. 123~ sl~ficmce of Some of ~e= fxtom ~ estimates of capll~ we ~xout~ by Huten et, al. s= ch~]es R. I-fdten, James W. Robertson, ad Frank C. Wykoff, Energy Obsolescence, and the Productivity Slowdown, in Technology and Capital Formation, edited by Date W. Jorgenson and Ralph Landau (MIT Press 1989), pp. 225-258.
PAGE 35
28 inventory estimation process. 124 Canada has recently begun an effort to measure capital stock for exactly this purpose. 125 Education as Investment Traditional accounting conventions do not treat expenditures on education or training as a form of investment. 126 This is not a trivial exclusion because total education expenditures are nearly 10 percent of the GNP. Currently, data suggest that a company is investing if it purchases a new machine, but not if it pays for the employee training needed to use that machine efficiently. Indeed, the total national investment in education probably exceeds net private purchases of equipment like machine tools and computers. Even if we decided to treat education as an investment, however, we dont have any clear idea of how much we are investing. Estimates of corporate investment in education and training range from $66 billion to $210 billion. 127 GM boasts that it has the largest, private educational program in the world. n Federal data on training are very incomplete. Estimates of private investment in education are based primarily on private surveys, which are acknowledged to be inadequate. 129 If education outlays are to be included in capital accounting, a number of problems must be overcome. Perhaps most difficult is measuring the value of education. However, even the most conservative estimates would more accurately reflect the new realities of an education-based economy. Labor Inputs Data on the actual number of hours worked by Americans are available from BLS. But all hours are not equal. Several attempts have been made to adjust for the quality of labor by examining changes in skills and levels of education. 130 BLS is introducing a new indicator of labor quality. It is particularly difficult to adjust for the quality of education delivered. Declines in standardized test scores during the 1970s and early 1980s seemed to indicate that an adjustment should be made. Estimates show that growth in quality adjusted labor inputs may be 0.1 to 0.25 percent per year lower than unadjusted labor inputs during the 1980s. 131 It is important to measure changes in the real education assets available in the American workforce since competitiveness of U.S. industry appears to depend critically on changes in labor quality. Adjustments for labor quality depend on techniques linking wages to education, age, sex, work experience, and other factors. Few, if any of the measures take into account training received on the job (except to the extent that training is measured by years of work experience). It is difficult to determine whether higher wages are paid for education because the education was needed on the job or simply because education was needed to get the job in the first place. Planned improvements to the Survey of Income and Program participation (SIPP) survey (discussed later) may help. E. How Productively Do Domestic Producers Use Inputs? Given accurate information about outputs and inputs, it is possible to compute changes in the productivity growth rate (output per unit of labor input) of the economy. In the absence of productivity IZAH~~n, op. cit., footnote 118. p. 50. l~ptir Ko~~~, rhe CWIW stock Smey proj~t, ~u~ey ~fcu~re~t B~iness, May 1985, VO1. 64, No. s, p. 31, 126A s~lw ~~t co~d ~W be m~ about expenditures on health care. See Eisner, op. cit., fOOtnOte 112, p. 10. lzTRo~r J. wu~ and Sue E. Berryman, Employer-Sponsored Trti g: Current Status, Future Responsibilities, prepared for the Conference on Employer-Sponsored Training, Alexandri& VA, Dec. 1-2, 1988, P 3. lzg~ve~=ment, The Atlantic, May 1989, p. 105. 129A.p. cmev~e ~d H. Gol&ein, ~p]oy~ Tr~ing: [ts ch~ging Role @ An Analysis of New Data, American Society fOr Training ~d Development Press, Washington, DC, 1983 and A.P. Carnevale, testimony before the Subcommittee on Taxation and Debt Management of the U.S. Senate Fiiance Committee, Washington, DC, Nov. 30, 1987. 13~w~ F. ~wn, Tre~S in A~r~c~n Ecowmc Growth, 1929-1982 (Washington, DC: The Brookings lnstltuthm), 1985; D. Jor&nsont F. Gollop, and B. Fraurneni, Productivity and U.S. Economic Growrh (Harvard University Press, 1987), E. Dean, K. Kunze, and L. Rosenbhun, Troduetivity Change and the Measurement of Heterogeneous Labor Inputs, paper presented at the Conference on New Measurement procedures for U.S. Agricultural Productivity (Mar. 31-Apr. 1, 1988, Washington DC). 131~~ NB~y, $~~uctivi~ ad tie semices of Capital ~d Labor, Brookings P~ers on ECOnOmC Activiry, VOI. 1, 1981, p. 49; J. Bishop, S the IkQ Score Decline Responsible for the Productivity Growth Decline? working paper 87-05 (Cornell University, Jan. 6, 1988).
PAGE 36
29 growth, economic output per person and therefore living standards can only increase over the long term by increasing inputse.g. by having each person work more hours or by putting more people to work. Productivity is a key measure of progress in the economy. 132 The output per hour worked in the business sector of the U.S. economy in 1986 was 60 percent higher than it was in 1960. 133 Each hour worked therefore produced 60 percent more to be paid as wages or profits. Tracking changes in productivity, and helping explain changes in productivity, is clearly one of the central goals of the national statistical effort. It has been very difficult to explain the striking changes in rates of growth in labor productivity that have occurred during the past few years. Labor productivity grew an average of 2.9 percent per year between 1948 and 1973, fell to 0.6 percent per year between 1973 and 1979, and increased to 1.4 percent per year from 1979 to 1986. 134 The challenge of explaining the changes is made more difficult by the fact that the post 1979 increases in productivity show a very peculiar feature. Unlike previous periods of productivity growth, the post 1979 growth was dominated by increases in manufacturing productivity (it averaged 3.5 percent per year between 1979 and 1986) while productivity growth in other business sector activities (primarily services) was very slow (0.6 percent per year). 135 A variety of explanations have been offered to explain the change. Martin Baily and Robert Gordon suggest that as much as 30 percent of the productivity decline after 1973 is the result of errors i n measuring both outputs and inputs. 136 Even after granting some errors in measurement, however, a real decline has clearly occurred. Explanations for this decline include a surge of less experienced workers from the baby boom, a decline in spending for research and development, dramatic shifts in oil prices that required increased inputs of capital and labor, and new government regulations that increased labor inputs without increasing sales. 137 The Bureau of Labor Statistics (BLS) provides three kinds of productivity statistics: l l l Business Sector Productivity that measures the value-added in the business sector per hour worked. 138 Industry Productivity Series that measures gross output (or sales) produced per unit of labor for each major industry. 139 Multi-factor Productivity Series that separates changes in Business Sector output due to changes in capital and labor inputs from changes that result from new technologies or other practices that can increase output without increased use of capital and labor. A new series (the KLEMS) provides a more detailed analysis of the effect of different types of inputs (capital, labor, energy, materials, and services) on productivity at an industry (2-digit SIC) detail for the manufacturing sector. In addition to these productivity series, the BLS produces a number of unpublished series that cover industries such as construction which are not IJZRo&~ E. Lit~, Ro&~ Z. ~wrence, and Charles L. Schultze (eds. ), American Living Smnalzrds, (Washington, W: The Brookings hstitution, 1988), p. 178. ~BBSta~tlcal Abstract of the United States, 1989, op. cit., table 659, P. 403. IJq~niWn, op. cit., foomote 29, p. 3. 135~niWn, op. cit., fOOt.nOte 29, p. 3$ 13bB~Iy md Gordon, op. cit., footnote 11, p. 418. 13TS= ~wwd F. ~fimn, The ~temptlon of Productivity Gro~ in tie Unitd States, The Economic Journd, VO1. 93, Mach 1983; Herbert Giersch and Frank Welter, ~wards an Explanation of the Productivity Slowdown: An Acceleration-Deceleration Hypotheses, The Economic Journal, vol. 93, March 1983; Wayne B. Gray, The Impact of OSHA and EPA Regulation on productivity, working paper, National Bureau of Economic Research, Cambridge, MA, July 1984; Assar Lindbeck, The Recent Slowdown of Productivity Growth, The Economc Journa/, vol. 93, March 1983; Zvi Griliches, *R&D and the Roductivity Slowdown, The American Economic Review, vol. 70, No. 2, May 1980; Martin N. Baily and Alok K. Chakrabarti, Innovation and Productivity in U.S. Industry, Brookings Papers on Economic Activity, No. 2, 1985; Richard J. Murnarne, Education and the Productivity of the Workforce: Imoking Ahead, in American Living Standara% Robert E. Litan, et. at., (eds. ) (Washington, DC: The Brookings Institution, 1988), pp. 215-245. 138The bufiness ~{or ~clu&s roughly ~~.q~ers of GNP, 1t excludes output from (,he re~-of-~e-wor]d, general government, output from ptid employees of household help, nonprofit institutions, the rental values of owner-occupied buildings, and the statistical discrepancy in computing NIPA. See U.S. Department of Labor, Bureau of Labor Statistics, Handbook of Labor Smtistics, Butletin 2217, June 1985, p. 226, 139us. ~pment of L~r, B~eau of Labor Smtlstics, prod~rivi~ Measures for Se/eC(ed fruf~trles M cove~m~t Services, Bulletin 2296, February 1988.
PAGE 37
30 covered in the published series because of weaknesses associated with the data. Business Sector Productivity The broadest and most widely followed measure of productivity is called business sector productivity. 140 The business sector is chosen to include only the parts of the economy for which BLS has reasonably good price deflators (e.g., it does not include any government activities) and is highly aggregated. Business sector productivity is published only for a few not mutually exclusive sectors: private non-farm business, manufacturing, durable and nondurable manufacturing, and nonfinancial corporations. This high level of aggregation avoids many of the difficulties encountered when attempts are made to assign economic output to specific businesses. Business sector labor productivity is calculated using constant dollar estimates of value-added generated by the business sector and estimates of the total value-added in manufacturing provided by BEA in its Gross Product Originating series. 141 But there are problems even at this high-level of aggregation. The difficulty BEA encounters in estimating value-added in constant dollars was discussed at length earlier in this paper. A number of researchers including Baily and Gordon, 142 Denison, 143 John Kendrick, 144 Lawrence Mishel, 145 and OTA 146 think that manufacturings productivity rate has been overestimated because manufacturings constant dollar value-added is improperly measured. 147 If this is true, non-manufacturings productivity has been underestimated. Industry Productivity The problem of computing value-added can be avoided if labor productivity is defined to be gross business output (or sales) divided by hours worked. In this case it is only necessary to develop a deflator for the products sold by an industry. The difficulty with using gross output as a measure of a business contribution to economic activity has already been discussed in the section on Computing Business Output by Sector. A potential problem with gross output is that it includes the value of intermediate inputs produced by suppliers and the value-added by the firm. Thus artificial boosts in labor productivity could occur if a firm kept gross output steady while increasing suppliers products and decreasing labor that used to be used to make those products internally. BLS claims that it does not publish series for industries where the intermediate inputs have changed so significantly that they would distort the results. The industry labor productivity series covers mining and manufacturing industries in extensive 3 and 4 digit SIC (Standard Industrial Classification) detail, but does not have any published productivity data on any segment of the construction industry and the published data covers less than half of the jobs in private sector service industries. 148 The services that are included tend to be industries whose output can be quantified, such as ton miles in the railroad transportation industry. 149 This means that impor1~ b~~ ~tor iul~es ~@y ~qu~m of GNP. II cxcl~es o~put f~rn tie ~St-of-tie-world, general government, output from paid employees of housebold help, nonprofit institutiotm the rental values of owner-occupied buildings, and the statishcal discrepancy in computing NIPA. see Us. Departmen t of Labor, Bureau of Labor Statistics, Handbook ofl.abor Statistics, Bulletin 2217, June 1985, p. 226. 141u.s+ ~=ntof ~r, B~au of ~bor s~tistics, Handb~~k of Met~&, Bul]etin 2134-1, (w&in@on, DC: U.S. Government Printing Offii% 1982), p. 94. 14zBaily and Gordon, op. cit., footnote 11, p. 420. lqs~w~ Fe ~Wn, Es-es of pr~~tiv@ Ctinge @ /~~ (w~hgton, DC: The Brookings Institution, 1989), p. 58. 144Jo~ w. K~&& sfi~ s~~ ~IIctiMty, B~iness Economics, April 1987, p, 23. 145~~nW R. -1, *C~uf~_ Numb: How ~~~ate Statistics ~~e~ U.S. ~dusrn~ ~li~ (Washington, ~: &OnOIItiC pOliCy Institute, April 1988), p. 57. l~u.s. congress, Ofnce of lkchnology Assessment, Paying the Bill: Man@cturing and Americas Trade Deficit, OTAITE-390 (Washington DC: U.S. Government Rinting Office, June 1988), p. 46. 147* & ~~~ on 6*C~pU@ BUS&SS -t by s~mr for a mom complete discussion of this mismeasurement. 14S~riv~~m~le 1 ofPr~~dvi~Me~~esforSe/ectedlnd~t~es ~Gover~~Se~ices, f3~letin 2*%, February 1988 ad U.S. ~-CIM of Commerce, Bureau of Economic Analysis, National Income and Roduct Accounts, op. cit., footnote 15, table 6.1OB. l@Je~e A. h4ark, Measuring Producu vity in Service Industries: llte BLS Experience, Resented at the Symposium on lkchnology and the Services Mustries, Hanover, NH, August 1987.
PAGE 38
31 tant and quickly growing service industries whose output is difficult to measure, such as business services, health care, and private education, do not have published productivity estimates. Left with this gap, researchers either calculate their own productivity estimates using BEAs National Income and Product Accounts data (value-added divided by persons or hours) or unofficial, unpublished BLS estimates. In sectors like business services and banking, inflation adjusted gross product originating (value-added by industry) is estimated using changes in employee hours as an indicator of how the quantity of output in this sector has changed. 150 Because labor productivity is calculated by dividing an industrys total output by the total number of hours needed to generate the output, this process of estimating output through the use of hours essentially puts hours in both the numerator and the denominator of the ratio, canceling each other out. The result is that labor productivity increases are practically eliminated by definition algebraically. *s* This might help explain why this sectors productivity has been flat even though the service sector has made substantial investments in productivity enhancing capital equipment, like computers. It is because of these limitations that the BLS publishes detailed productivity indexes only for selected industries. Nevertheless, the widely followed business sector productivity series does indirectly make use of this limited, extrapolated data because many of the constant-productivity services like business services are large and growing inputs into other sectors, particularly manufacturing. 152 Because of this and other factors, Denisen estimates that the productivity of farms and manufacturing has not only been overstated, but has been increasingly overstated over time 153 Multifactor Productivity The third measure of productivity produced by the Bureau of Labor Statistics is the multifactor series. Favored by most economists, the multifactor series does not ascribe all changes in output to one factor such as labor, but rather breaks it down into broad categories of inputs. The series published on a regular schedule traces changes in business sector productivity to changes in multifactor, labor, and capital productivity. A new series (KLEMS) add energy, materials, and purchased services productivity for all manufacturing industries. Whatever growth that can not be attributed to these factors reflects a more qualitative change such as a shift in technology or different management. Although, this series is conceptually a large improvement, it suffers from some of the same problems that plague the other productivity and output measures; particularly the lack of detailed data on purchased services. F. How Has the Corporate Structure of the U.S. Economy Changed? Sophisticated technology, regulatory changes, an active merger wave, a severe recession and the need to participate in intricate marketing and financial networks, have led to a realignment of the structure of Americas enterprises. While it is difficult to provide precise documentation, there is clear evidence that the growth of large firms is increasingly built around the aggregation of large numbers of comparatively specialized small establishments. 154 Many large firms, l55 including AT&T, DuPont, General Motors, Hewlett-Packard, IBM, Martin Marietta, NCR, 3M, and Xerox claim that they are reorganizing operations to encourage more entrepreneurial behavior on the part of individuals and establishments. 156 15M10 ~tem,~ product by Indusq, Survey of Current B~iness, April 1987, p. 27. ~slK+ck op. cit., f~tnote 144, p. 20 and Bady and Gordon, op. cit., footnote 11, pp. 394 and 397. lS2T=he~er, op. cit., fOOtnOtC 3A. P. 34. 153&jwad F. ~~n, ESt~eS @p~o&tiv@ Chnge by /ndus~ (w~n~on, DC: The Brookings Institution), 1989, p. 55. 15~~ ~~j~at is ~e sm~lest tit in which b~iness ~tivi~ is conduc~. It CUI rcpxnt the entirety of the business activity within a business, making that business a sole-establishment enterprise, or it can be one of hundreds of establishments (branches, subsidiaries, and corporate headquarters) that make up a muki-establishrnent enterprise. 155w tm fm Rfem to tie entfi c~ra~ f~i]y or enterprise which can be made up of numerous establishments. 156jmeS Brim ~m, ~$~w~g ~matlon: c~moll~ ~am, ~anar~ Business Revi~, ~y/June, 1985, p. 80 ~d R.M. KMter, The Alt~k on Pay, The Harvard Bainess Review, vol. 65, No. 2, March-April 1987, pp. 60-67; and Michael Piore and Charles Sable, The Second industrial Divide (New York, NY: Basic Books), 1984.
PAGE 39
32 Small firms traditionally flourish during periods of rapid transition, since the bureaucratic inertia of large firms may blind them to opportunities in places where none were expected. Computers and communications technology are providing opportunities for small enterprises to hook into larger networks of production. Sophisticated production technologies capable of tailoring products to specialized markets without a significant sacrifice in productivity or increase in cost can vastly diminish the values of economies of scale-benefiting small businesses. When the trends are combined there appears to be a convergence in structure as firms in sectors that are traditionally fragmented (farming, physicians, and home construction firms) become amalgamated into larger firms while sectors that are traditionally highly concentrated, like automobile production, bear less resemblance to Fords enormous Rouge River Plant (where wood and iron went in one end and cars rolled out the other) and more like the craft-based production that preceded heavy automation. This observation parallels the popular notion that advances in production technologies have reduced the need for large production units, but have simultaneously made it possible to connect and manage many different units under one corporate structure. 157 The data needed to track these changes and their impacts are spotty at best. Even answering more mundane industrial organization questions, such as how mergers and acquisitions have changed the makeup of the U.S. economy, whether small firms are the source of a disproportionate amount of new jobs, or what the levels of new business startups and closures are, requires elaborate methodological contortions. A complete data base of U.S. business establishments organized into their corporate families that researchers can study to observe some of the dynamics discussed above does not exist. The constant churning of the economy makes this a difficult task. Currently, there are two primary data sets that show the corporate structure of the U.S. economy over time: the Small Business Administrations (SBA) Small Business Data Base (SBDB) 158 and the Enterprise Statistics file compiled by the Bureau of the Census. 159 Both of these data sources suffer from limitations that restrict their usefulness. The Enterprise Statistics data series is a file built from establishment level data collected from the economic censuses (e.g., Census of Manufacturing). The data are converted from the establishment level to the enterprise level through use of another Census data file, (The Report of Organizations), which takes a snap-shot of the enterprise at a specific time. 160 Because the data are cross-sectional and not longitudinal, it is impossible to understand the dynamics underlying changes in employment or sales. Questions such as whether the employment growth was due to the birth of new firms or the expansion of old firms cannot be answered. Collected in five-year increments, the Enterprise Statistics currently available dates back to 1982. In addition, the data do not provide links that allow longitudinal tracking and lack coverage in some of the fastest growing segments of the economy, such as transportation, communication, finance, insurance, and real estate. 161 157s= Russcll Johnston and Paul R, Lawrence, op. cit., footnote 109, pp. W101. 15EM SBDB C~SCS of ~vti~ ~temlat~ files. TWO are implicitly referred to here: USEEM (U.S. Establishment and Enterprise Microdata) ~d USELM (United States Establishment Longitudinal Microdata). See U.S Small Business Administration, Handbook of Small Business Data, 1988, (Washington, DC: U.S. Governmental Printing Office), 1988. 15gSeve~ o~= @ ~Ws fa business level &a ~ exist, such u unemployment ~sur~ (UI) ~ords, ~gitudin~ Employee-Employer Data (LEED), Statistics of income, the hmgitudinal Research Database (LRD), and the Standard Statistical Establishment list (SSEL). These data sounxs are not included as primary data sets because they have either been discontinued, are subject to disclosure provisions, cover only one sector such as manufacturing, or lack fiudamental data of interest to researchers such as employment or the enterprise structure. See Handbook of Srnd Business Data, 1988, op. cit., footnote 158, pp. 10-17. 1~.s. ~pmatof Co-erW, B~~of~ ~us, ]982 Enferpr~e s~tistics: Gener~ReportonI~~trial organization, OCtOkr 1986, p. 31 ]. 161fA1tito cover ~me ~ims of ~ ~miw indus~ should & ~~eviat~ when we Census B~au exp~ds i~ industria~ wnsws in 1992. h prior years, the Census Bureau could not include these sectors because they were in regulated industries. For a comparison of the two data sources, see U.S. Small Business Administration, Handbook of SmaJIB~iness Data, 1988, (Washington, DC: U.S. Government Printing Office), op. cit., foomote 158, pp. 26-79, and Candee S. Harris, Handbook of Small Business Data, report to the Small Business Administration, January 1983, p. 8.
PAGE 40
33 Under a mandate from Congress, 162 the Small Business Administration started in 1979 to develop a database built from private Dun & Bradstreet Duns Market Identifier (DMI) data that roughly cover all private businesses with employees. l63 A private data source was used so that individual firms could be analyzed without violating confidentiality restrictions associated with public data. Organized into their respective corporate enterprises, the firms are linked over time for the even years from 1976 to 1986, exposing the evolution of the business as it gains and loses jobs and establishments, shifts lines of business, and relocates geographically. The births of new businesses appear alongside deaths. The problem is that the Dun & Bradstreet (D&B) data were collected for the purpose of assigning credit ratings and creating other business products, not for academic economic analysis. Nor was it ever intended to be organized on a longitudinal enterprise basis. As a result, the data underreport branches of multiestablishment firms since most credit decisions are made at the corporate headquarters or subsidiary level. 164 Thus the true corporate structure from an establishment viewpoint is obscured. The database does not include firms without employees, such as the self-employed, who represent about 9 percent of the labor force and roughly 60 percent of all businesses. 165 Other problems include the fact that it can take several years for D&B to include new firms (especially smaller ones) in its filepossibly missing a large number of firms that go in and out of business during that time. 166 Likewise, firms, especially smaller ones, are not consistently updated, creating possible distortions in calculations of growth 167 When these problems are added to clerical errors that were found, nearly half of the establishments in the full SBA USEEM database were deemed inadequate for tracking employment over time. To correct for this, editing, imputing, and weighting schemes had to be devised. 168 Depending on how one adjusts for these deficiencies, researchers using Dun & Bradstreet data can get significantly different results. One of the early users of D&B data, David Birch, states in his new book, Job Creation in America, that firms with less than 20 employees created 88 percent of all the new jobs between 1981 and 1985. 169 But Mr. Birch, whose work has attracted a lot of criticism, asserts that: Anybody can make it come out any way they want. 170 SBA analyses using a database built on Dun & Bradstreet data conclude that only 36,5 percent of new jobs came from firms with less than 20 employees between 1982 and 1986. 171 The lack of a comprehensive longitudinal database on corporate structure means that questions concerning the roles of small versus large business in generating economic growth are answered with great imprecision-if they are answered at all. Studies about the source of job growth, as discussed above, are controversial. Research that looks at the correlation between productivity and firm size yield ambiguous results. 172 The amount of business startups and closures a seemingly basic indicator of the direction of the economy-is unavailable with any 162~ Stil Bs Economic Policy Act of 1980, Public Law %302. lbs~~l~ti~ity RXtiCtiOIIS prewmxl the SBA from using detailed Census data. 164~ 1978, *R Wm a 15 million employ= ~ffme~ ~tw~n tie tot~ employment re~~ by tie corporate tops of the fimls ~d the Stlm of the firms establishment employment. Dun & Bradstrcet says that it has recently improved its coverage of branch establishments. See A.L. Walton, How Small Businesses Contribute to Job GenerationThe Pitfalls of a Sesmingly Simple Question, National Science Foundation, prepared for the 1983 Conference on Industrial Science and lkchnologicat Innovation, Evanston, IL, May 2-4, 1983 and The State of Small Business, March 1983, p. 290. ~~Htiook Of small Business, 1988, op. cit., footnote 158, P. 7. 166Bmce D. ~~lps ~ BA. Kfihhoff, F~~ion, Grofi ad s~iv~: Small Firm Dynamics in the U.S. EConomy, Srnaff Business Economics, vol. 1, 1989, pp. 66-67; Sue Birley, Finding the New Firm, Academy of Management Proceedings 44th Amual Nkxting, Boston, MA., Aug. 12-15, 1984, p. 67; Douglas P. Handler, Business Demographics, a study by Dun & Bradstrcet, p. 7. 167A.L. w~tm, How f$m~] Busj~~s con~bute to Job Generatjon_The pitf~ls of a s~mtigly simple Question, National Science Foundation, paper prepared for the 1983 Conference on Industrial Science and Ikdnological Innovation, Evanston, IL, May 2-4, 1983, pp. 10-14. 168s=H~mkofs~lBwiN5s Da~, 1988, ~. cite, foo~ote 158, p. 2 ~d The S~te OfSnaUBu.siraem (Washington, DC: U.S. Government printing Office) May 1985, p. 425. I@David Bjrch, Job Creation in America (New York, NY: The Free Press, 1987), p. 16. ITODavid w-l ~d Buck Brown, me Hyping of Sm~]-Firm Job GrOW~, ~~e Wui/ street ~Ourw/, NOV, 8, 1988, SCCtk)n B, p. 1. l?l~e Hyp~g of Sm~-Firm Job Growth, The Wail Street JoIund, N OV 8, 1988, p. B7. ITzSteva L~g~en, Fi~ Repo~ t. t~ s~//B~i~ss A&irusmmon on Firm Size and Productivity, (W~inlYon, DC: U.S. Government printing Office), Sqtcmbcr 1982 and Stahrl W. Edrnunds, Organizational Size and Efficiency in the U.S., The Antitrust Bulletin, Fall 1981, pp. 507-519.
PAGE 41
34 sort of precision. 173 More abstract, but crucial questions, such as the role of organizational structure in promoting innovation, cannot be definitively answered. 174 This uncertainty hinders and possibly misleads the development of economic growth policies. The difficulty associated with developing a complete list of businesses cannot be understated. Dun& Bradstreet estimates that several hundred thousand businesses see a change in their status through starting or discontinuing operations or through M&A [mergers and acquisitions] activity. 175 A potential solution to this problem exists through use of the Standard Statistical Establishment List (SSEL) that is used by the Census Bureau or the unemployment insurance data collected by the BLS. Currently both databases provide a sampling frame for the agencies surveys which could possibly be used for other purposes such as examining changes in corporate structure. Consisting of the address, SIC code, sales, and employment of nearly every business establishment in the country, the SSEL provides fairly complete data on the way business organize to form the economy; although efforts would be required to organize the file into a corporate structure. But the use of detailed Census data trigger confidentiality limitations that restrict their use by outsiders and even by other statistical agencies like BLS. 176 Because of these restrictions BLS has developed its own business list from unemployment-insurance administrative lists that have extensive coverage of small businesses. 177 External use might be possible if the data are aggregated in a way that protects the confidentiality of respondents. G. How Does Growth Affect Incomes and Income Distribution? All the discussion thus far has been directed at national averages. Aggregate data on growth of national income and wealth say nothing about who benefits from this growth. There are clear indications that the benefits of growth have been enjoyed primarily by the wealthiest families in the United States since 1979. 178 The gap separating those able to benefit from the transformation of the American economy (primarily people with good educations) and those left behind (primarily those with poor educations and families headed by single women) is growing. 179 This trend has a direct effect on theU.S. labor force because it means that about a fifth of all children now live in poverty and may enter the work force disadvantaged by this experience. Information about the effects of economic change on income distribution are important for policies ranging from changes in the personal income tax to changes in welfare policies to identifying who are the poor. The bipartisan welfare reform program recently enacted drew heavily on data showing how many people remained in welfare programs for long periods of time. Data linking incomes to individuals and households come primarily from four sources: the Statistics of Income, the Current Population Survey, the Survey of Income and Program Participation, and the Consumer Expenditure Survey. Of the four, the most widely used for studying patterns of income distribution is the Census Bureaus Current Population Survey (CPS) because of its rich demographic detail on households. While most wage income is reported in the CPS, income from dividends, interlT3H@~k of s~l B~i~ss Dw, 19~, op. cit., foomote 158,pp. 21-22. lT4~ c. ~a, j. cl~k, ~d LOG. ~~, UMW~yw~ ~ Techmc~ ]~va~n (WeSp~ m: Greenwood FWXS, 1982); W.M. Cohen, R.C. bvin, and D.C. Mowery, Firm Size and R&D Intensity: A Re-examination, National Bureau of Economic Research (Cmbridge. MA). w~ing p-r No. 2205; and L. Rmuuzky, et. al., The Process of Technofogicaf Innovation: Reviewing che Literature (Washington, DC: U.S. Government Printing Office, 1983). 175 Hmdl=, op. cit., foomote 166, p. 10. 176Eff~~ ~ le~l~ve]y ~ ~ ~fi&nti~~ limitations have met witi opposition. See National Associatmn of Business EcOnOmi sts, Report of the Statistics Ccxmm ttce of the National Associatkm of Business Economists, February 1988, p. 18. 177 Sl*r, op. cit., foomote 5S ~ 1* P 178s. ~iger, p, G@~h~, ~ Eugene Smolensky, How & Rich Have F-, IWS-8T, The American ECOtIOmiC Review, VO1. 79, No.2, hhy 1989, PP. 310-314. US. COngrCSS, Congrcasioti BIJ@t Off@, The C~n8ing Dis~~~n Of Feder~ T~S: 1~75-~~v %* 1987 179s= U.S. *P, conmlm~ B-t offi~, Trends in FamL/y [~ome; ]970-1986 (Washington, D, C,: U.S. Government Printing office), February 1988, p. xviii; U.S. House of Representatives, Committee on ways and M=w ChM~en ~nfovero! May Z 1985) P. 601; ad U.S. COn81WISS Congressional Rcsearc h Service, Economic Benefits of Education, Dec. 13, 1988.
PAGE 42
35 est, rents and other types of property-type income are very poorly reported. Estimates indicate that even after some imputations are made, only half of some categories of income (e.g., interest and dividend) may be reported. The survey also does not fully capture the income of the very wealthy. In an effort to improve the accuracy and the response rates, the survey does not ask the wealthy to report their exact income, but instead simply to state whether their income exceeds a certain threshold such as $300 thousand or above giving no information about how much is earned in excess of the threshold. 180 The data available to the public has an even more restrictive threshold of $100 thousand. This not only leads to a significant underestimate of the income of high income households which appear to be a growing segment of the population, but also frustrates accurate comparisons of the income distribution over time since the threshold level has been increased erratically over the past 15 years. 181 Because this level changes over time, researchers trying to analyze expenditure patterns on the basis of income and estimates of household income growth and inequality, particularly for this upper group, are constrained. To get around the problem researchers usually throw out the top group, leading to an underestimate of the inequality in income. 182 Some of the confusion surrounding the so-called Missing Middle income distribution question, where it is alleged that the middle class is shrinking as the lower and upper classes grow, involves limitations with the CPS. A report from the Congressional Budget Offie, which made creative use of IRS records to correct for this upper-income problem, found that all of the inflation-adjusted, after-tax income growth between 1977 to 1988 occurred in the upper 10 percent of all families (see figure 8).183 Efforts to track growth that exclude this upper income group are obviously severely hampered. A potential solution would be for the Census Bureau to assign a mean or median income level for this top group rather than the threshold level. This estimate could be derived through matching the CPS to another data source (e.g., IRS Statistics of Income Figure 8Mergers and Takeovers, 1 2 3 4 5 6 7 8 9 10 Top 5% Top 1% (constant 1982 dollars) 1972-1986 O 5 10 15 20 25 30 35 40 Percent of All Family Income NOTE: Assumes Corporate Tax Allocated to Captal Income, SOURCE: Julius Allen, Corporate Takeovers: A Survey of Recent Developments,U.S. Congressional Research .%rwca report No 87-726E, Washington, DC, Aug. 6, 1987. data) that does not suffer the same limitations. The characteristics of the data used to accomplish this task and necessary confidentiality restrictions require that it be done at a very aggregated level. The other sources for income data include: The Statistics of Income (SOI) published by the Internal Revenue Service consists primarily of income data, providing good detail on the source of income (wages, interest, dividends, etc.), but lacks explanatory demographic data and information on people who do not file tax returns. The Survey of Income and Program participation (SIPP) is designed to provide more detailed information about the income households receive from all sourcesparticularly noncash transfer benefits. Unlike the CPS, the SIPP monitors changes in the income of individual families over a number of years and data are available on a monthly basis. It also attempts to provide a measure of the wealth of households, not just current income. I%s setting of an iwome threshold is commonly referred to as top-coding. 181u.s. HOU~ of Rep~~n~vm, Cornmitt= on Ways and Means, Ckfdren in POVer~, op. Cit., fOOtnOte 179, p. 599. 1821bid., p. 599. 183u.s. Cmms, ~c cm=s~m~ B~get office, The c~g~ng Dislribu~n Of Fe&raJ T~es. )975-)990, &tO&r 1987, table A-2, p. 85,
PAGE 43
36 l l The Consumer Expenditure Survey provides information about household purchases and income for a small sample of families each year (see the next section for more details). A small sample of households has been followed by the-University of Michigan since the mid 1970s. Known as the Panel Study of Income Dynamics, the series is a rich longitudinal database but suffers from a small sample size and data inconsistencies. While the data available from these sources can be used to outline basic changes in income distribution, the limitations noted greatly limit the analysis possible. Changes in the wealth and income of wealthy households is not well reported. Sample sizes in SIPP and other longitudinal studies are so small that many small but important demographic groups are not represented in large enough numbers to be statistically significant (e.g., nonwhite welfare mothers, or the very old). Information about employer benefits is not well reported either by employers or by individuals. What is reported is of uneven quality. Efforts to link total employer contributions to benefits enjoyed by different kinds of households, are only in the experimental stages. 184 As mentioned before, a potential solution to some of these shortcomings involves matching these data files to other data sources. This process could compensate for some data deficiencies, but suffers from the fact that much of this data was collected from respondents with an assurance of confidentiality. Conforming to these confidentiality restrictions can limit the usefulness of the matching technique. 185 Other Problem Areas This report has only touched on data designed to address gross measures of economic change. A number of other areas clearly deserve attention. These include: The redundancy of two business directory lists for sampling purposes should be eliminated. Currently, two nonfarm business lists exist as sampling frames for establishment censuses and surveys. The Bureau of the Census uses the Standard Statistical List (SSEL) and BLS uses its business directory list compiled from Unemployment Insurance (UI) data. The existence of two lists means that data is not strictly comparable and that an expensive duplication of effort occurs. Although the Presidents Economic Policy Group recommended that BLS serve as the agency in charge of compiling a common list, 186 efforts have been stymied by a lack of legislation that amends the confidentiality law allowing BLS access to the Census data. Data on scientific and engineering manpower also suffer from weaknesses. BLS surveys report only half of the scientists and engineers reported by NSF while NSF reports only 60 percent of the engineers counted by the Bureau of the Census. 187 While extensive data are available on regional economic activity, comparatively little work is done to analyze or present the data in ways that would allow states or regions to reproduce national analyses. Regional input/output data are constructed with painful deliberation. 188 BEA has long published personal income estimates by state and county and has recently introduced an experimental data series on Gross State Product. As businesses focus on niche markets and as economic development issues increasingly fall into the hands of the states, there is a need to add to databases of regional economic statistics. Significant economies seem to exist by having the Federal government undertake this task rather than having the individual states do their own data collection. The two standard measures of net savings BEAs NIPA measure and the Federal Reserve Iud~ ~mW B-u h= exP~a~ wi~ coll~~g ~ta from emp]oyers of respondents of SIPP and hopes IO do so on a larger scale in the future. lass= RW= H-it, ~~r Bowie, Daniel Kasprzyk, and Sheldon Ha~r) Enhanced Demographic-Economic Data Sets, Survey of Current Business, November 1988, pp.44+8 and Walter Y. Oi, How Valuable Are Matched Data Files, Survey of Current Business, November 1988, pp. 49-50. 18&R~ of tie Wofing ~up on he @~ity of &onomic s~istics to the Economic pOliCy COllIICil, April 1987 and Implementation of the Working Group on the Quality of Economic Statistics Rexomrnendations, March 1987. 107s= N~~~ ~~my of Science, Sumeylng the Nations Scietiists ~ Engi~ers: A Data System for tk 1990s, 1989 ~d U.S. Congress, Office of Ikchnology Asseasmen t, Preparing for Science and Engineering Careers: Field-Level Profiles, Staff Paper, Jan. 21, 1987. lsspole~e, op. cit., foomote 43, p. 3.
PAGE 44
37 l l l Boards flow-of-funds saving measure-lead to conflicting estimates. 189 The NIPA incomeless-outlays method faces the problem of subtracting two large numbers. 190 Thus errors in either end up as errors in the residual which is the estimate of savings. Flow-of-funds savings estimates are actually built up from data on assets and liabilities which face new problems because of the many new savings instruments made possible by a deregulated financial indusQ Nagging problems remain in treatment of the underground economy and possible under count of inner-city residents in the decennial census. 191 Time-Use Accounting needs to be improved. The way Americans spend their time is becoming a critical part of economic analysis since major changes are occurring in the things formally counted as a part of the economy. Many activities formerly done with unpaid household time (child care, cooking, care for the elderly) are now purchased while capital equipment in the home (microwaves, VCRs) substitutes for services that might otherwise have been purchased. Changes in time use provide a sensitive way to measure changes in the overall performance of the national economy. It is clear, for example, that Americans have purchased economic growth in recent years by sacrificing free time. The National Science Foundation funded small time-use surveys in 1965 and 1975, but budget cuts made a 1985 survey impossible. Fortunately, a private corporation, AT&T, provided the bulk of funding for a similar time-use study that year based on many of the same categories as the earlier research, and has recently agreed to make most of the data available to the public. 192 Timely data are needed on the bridge used to convert consumer spending to business output. 193 A critical link between the world familiar to consumers (where things like pizza and automobiles are bought in retail stores) and the world of economic statistics comes in the form of an obscure table called the bridge table in the jargon of input/output accounting. These tables are used to translate a dollar spent (say on a pizza) into economic output in standard business categories (e.g., $10 in pizza purchases might result in $2 for the grocery store, $0.50 for the trucking and warehousing company, $0.50 for insurance, $5 for food manufacturing companies and the rest for farmers). Since it was assumed that the ratios in these tables would change only slowly (retailing was never considered a progressive industry) there was little point in devoting a lot of attention to updating them. Our work convinces us, however, that the new production networks necessarily alter the chain of production that includes transportation and retail operations. This is precisely where the inventory reductions occur when information equipment is used to improve the flow of products. Research and Development data by industry and product type have not been reported in much detail. The National Science Foundation has attempted to correct many of the deficiencies (e.g. no summary data are published for construction, services, and some manufacturing) and will shortly publish much more complete data. Many problems remain, however, because of incomplete reporting and because research and development conducted at the corporate level is difficult to assign to activities by individual establishments that may be owned by the corporation. One of the weakest links in the chain of statistics occurs between the highly detailed demographic statistics on household spending patterns contained in the Consumer Expenditure Survey (CSX) and consumer spending information available in other data series like those that appear in NIPA (the first item in the right-hand column of table 1). Administered to IWB~~, opt Cit., fm~o~ I I 1, pp. 15-20, and F. & heUW Conflicting Measures of mva~ savings, Survey of Current Business, November 1984. ~~t]ays include consumer expenditures, taxes, and interest payments made to b~inwws. 191JWI F. Hou~~, The Un&rground Economy: A Troubling kue fOr policymakers, Business Review, September-October 1987, pp. 3-12; and Barbara A. Bailar, Finding Those the Census Missed, Technology Review, May/June 1988, pp. 22-24. 19~ Natio~ sci~~ Foundation provided some funding for the 1985 time-use smey. 193s=, U.S. ~ment of Commerce, Bureau of Economic Andysls, The hput-Output Structure of the U.S, Economy, 1977, Survey of Current Business, May 1984, table B.
PAGE 45
38 5,000 household (consumer units) a year, the CSX has since 1980 been conducted on an ongoing (annual and quarterly) basis-prior to that time it was done about every 10 years. The primary purpose of the CSX is to provide weights for the Consumer Price Index. The relatively small sample size limits any detailed attempt to figure out how or why different type of consumers buy what they do; by the time you slice the data by expenditure category, type of household, and income level, the divisions have become so fine that further attempts to analyze the data by variables such as age, sex, or region are impossible because the results quickly become unreliable or statistically insignificant. This is frustrating when advances in information technology have allowed a much finer targeting of products to consumers and have in-turn allowed consumers much more flexibility in tailoring products to their needs. This so-called niching phenomenon is difficult to track given the limits to public data. In a broader sense, many surveys linking spending to consumer amenity are not well coordinated with the national accounts. It is all but impossible to link spending patterns to measures of the quality of the amenities that are the ultimate result of the spending. A number of surveys are taken that could, in principle, be coordinated with the consumer expenditure surveys to help make some of these connections. They could also be designed with a more comprehensive view to documenting changes in the quality of American life and understanding the nature of economic change instead of focusing narrowly on the programmatic issues for which they were initially designed.
PAGE 46
Part III Conclusion Accurate, complete, and timely statistics are critical tools for effective public policy and business management. The perplexing nature of the economic changes underway in todays economy have, however, increased the difficulty of providing precise measures of change and growth. This background paper evaluates problems confronted in answering some of the most basic economic questions. Some of the difficulties are longstanding. For example, there has always been a need to get a more accurate estimate of the output of the construction industry. Some are old problems that have become much more important as the economy changes. Growth in service businesses makes it impossible to ignore defects in the way productivity in services is estimated. And economic change has created some entirely new problems-such as the emergence of complex international production networks, and the need to understand the way information is used as an input in the economy. The statistical agencies must continue to reevaluate their methods in view of new needs of public and private data consumers. Although this background paper has focused on the shortcomings, a number of improvements have occurred. Some of the most important recent developments include: the newly revised Standard Industrial Classification, BEAs development of a computer deflator, BEAs Gross State Product series, BEAs revision of the Gross Product Originating series, the Bureau of the Census Survey of Income and Program Participation, Census Bureaus Longitudinal Research Database, the BLS International Price Program, and the new BLS Multifactor Productivity series. These improvements indicate that the system can react to new policy-oriented needs, especially if the resources to do so are available. These and the many other innovations underway are commendable, but to be effective they and other planned improvements need to be part of a coordinated program that sets priorities, develops a coordinated response, and evaluates how well the needs are being met. This function was given to the Office of Management and Budget. But for many years OMB has elected to take a very narrow view of that responsibility and has not fulfilled its larger mission in this area. This paper makes no attempt to provide a comprehensive critique of national statistics nor does it attempt to offer comprehensive solutions. This important task needs to be undertaken on a regular basis by the Federal statistical agencies themselves acting under the guidance of OMB. We have identified a number of areas where a coordinated response is clearly needed. These include: l l l l Develop better techniques for evaluating real (i.e., constant dollar) growth in areas where most growth involves changes in quality or capability. This typically includes areas where technology is redefining the nature of the product in fundamental ways. Improvements would include expanded efforts in accounting for growth in manufacturing areas like computers, semiconductors, communication equipment, advanced machine tools, etc. It would also find a way to measure improvements in quality that occur as firms make more timely deliveries to suppliers or offer consumers a wider variety of products, Without accurate measures of changes in quality, policy makers have a distorted view of where real growth is occurring in the economy. Improve techniques for evaluating real growth in services. This means developing better methods for recognizing that the quality of education, health care, financial services, software development, and other services can change. Lacking such methods, real growth and productivity change for many services is underestimated, obscuring the innovations that are occurring in these sectors. Strengthen methods used to show the way purchased services are used as an input in the economy, particularly by manufacturing. By underestimating the real value of purchased services, the contribution of manufacturing operations may be overestimated and thereby mask problems in some industries and lead to a misunderstanding of the value of service industries. Improve the methods used to track imported products through the U.S. economy. Errors that may underestimate the value of the foreign inputs purchased by domestic businesses may overestimate the contribution of domestic fins, especially in the manufacturing sector. A n
PAGE 47
40 l l l l l inability to trace imported products through the economy makes it difficult to estimate the vulnerabilities or the competitive strength of different industries. Develop better methods to monitor the construction industry. The health of this important business, particularly the nonresidential segment is difficult to track using current methods of measuring outputs, inputs, and productivity. Improve methods of measuring investment in education and training as well as the quality of these services. Since worker education and training has become a critical input to virtually every business it is important to measure its role as precisely as possible. Virtually nothing is known about corporate investment in training and only crude estimates exist of the practical knowledge of people in the work force. Establish better methods for monitoring changes in the size and scope of firms and establishments. Major changes are taking place in the size of individual establishments and in the number and kinds of products produced in an individual establishment. Likewise, major changes have occurred in the number, size, and type of establishments owned by a single firm. Policymakers concerned about trends in the sources of job generation, the effect of mergers and acquisitions, or the regional shifts of industry, need to understand the nature of such changes with greater precision. Improve methods for measuring changes in the distribution of income. It appears that income distribution in the United States has changed substantially in the recent past. A significant amount of change has occurred in households with very high and very low incomes. Neither group is well tracked by existing data series. Policy designed to affect such changes needs to be informed by better data--particularly data that shows changes in the income history of individual households. Welfare policy, tax policy, and a variety of Federal expenditure programs are strongly influenced by data in this area. Develop methods for tracking the effect of new technologies. This could involve more timely input/output series and a capital-flows table with an improved set of business categories. Such data is critical to tracing linkages between l industries and understanding the connection between management of technology and economic growth. Improve methods for tracking standard national economic accounts and other measures of economic well-being in an integrated way. The national system of accounts is designed primarily to provide support to Federal macroeconomic policy makers. They were not designed to provide a macroeconomic view of changes affecting particular industries or a complete perspective of changes in the welfare of Americans. But by necessity, the accounts serve as a crucial resource for informing policy decisions in areas ranging from energy policy to social welfare policy. Policy analysis inmost areas requires combining this data with statistics in areas like environmental quality, resource depletion, income distribution, and health. These links could be much more clearly understood given a more integrated way of reporting economic progress in the United States that uses both the national accounts and other measures of social change. Such reports could also provide a systematic view of the quality and completeness of information not easily reported in economic accounts. Few of these problems have easy solutions. They all require a commitment to a long-term process and management committed to making the system work well as a whole. In some cases it will require additional investments in computational equipment few of the statistical agencies have adequate computational facilities. It may also involve a concerted effort to ensure that enough young people are trained to take jobs in the statistical agencies and that the agencies are in a position to attract a continuing flow of new talent. An adequate response to these challenges also requires coordinated approaches to budgeting and undoubtedly more money. The need for resources, however, cannot be established without a clearer view of the needs and priorities of the system taken as a whole. Such a perspective is not now available from any source. It is clear, however, that the price paid for public policy mistakes that stem from defects in national statistics can be many times higher than the entire national statistical budget.
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E92U4HDQG_BA2TY2 INGEST_TIME 2017-05-18T17:26:20Z PACKAGE AA00055066_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
|