February

Mid-Decade Strategic Reviewof ’s Economic Accounts:Maintaining and Improving Their Performance

Contents

Part . Economic Accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Economic accounting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ’s economic accounts and their uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Characteristics required of ’s accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Part . Source Data and Methods Used to Prepare the Accounts . . . . . . . . . The interaction of source data and estimating methods . . . . . . . . . . . . . . . Source data as determinants of initial release and revision schedules Source data and methods: Examples, special requirements, and

. publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The effect of source data and methods on the performance of the

. accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Part . A Comprehensive Review of ’s Economic Accounts . . . . . . . . . A comprehensive review: Scope and approaches . . . . . . . . . . . . . . . . . . . . . . . . . Background papers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Economic change and the economic accounts . . . . . . . . . . . . . . . . . . . . . . Recommendations from outside experts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Changes in data sources and estimating methods as they affect the

. economic accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The newly revised international guidelines for economic

. accounting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Revisions in the economic accounts: Implications for

. improvements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The priority areas of concern: Three cross-cutting issues . . . . . . . . . . . . . .

Change in the nature of output and the organization of production:. The need for new and improved output measures . . . . . . . . . . . . .

Investment: The need for better measures of investment, saving,. and wealth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Internationalization: The need for measures to fill gaps in the. coverage of international transactions . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Part . ’s Strategic Plan for Maintaining and Improving the. Economic Accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

The need for new and improved output measures . . . . . . . . . . . . . . . . . . . . . . Difficult-to-measure components of real . . . . . . . . . . . . . . . . . . . . . . . Reliability of key components of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Substitution bias in real . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outdated and inconsistent view of the structure and organization

. of production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The need for better measures of investment, saving, and wealth . . . . . .

Issues about the scope of existing measures . . . . . . . . . . . . . . . . . . . . . . . . . . Issues in the measurement of investment, saving, and wealth . . . .

The need for measures to fill gaps in the coverage of international. transactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Milestones in implementing the proposed actions . . . . . . . . . . . . . . . . . . . . . . An environment for change in the accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . .

A comprehensive review of ’s economic accounts—national, international, and regional—is underway.The purpose of this Review is similar to that of reviewsin earlier decades, but some other important featuresdiffer. Rather than being conducted by a “blue-ribbon”panel of outside experts, this Review is being conductedby , and the outside perspective that is vitally im-portant to such a review is being obtained in a differentway and at a different step in the process.

For this Review, the outside perspective is being ob-tained, as the third and final step, by comment anddiscussion of the draft strategic plan that is pre-senting in this article. Did correctly identify thepriority issues for maintaining and improving the ac-counts? Are the actions that proposes to address theproblems the best ones? Your answers to these questionsand your comments on other aspects of the draft planare invited. Please send your comments, by April , bymail to Bureau of Economic Analysis, U.S. Departmentof Commerce, Washington, ; by fax to ()–; or to carol.carson@bea.doc.gov on Internet.

Carol S. CarsonDirector, Bureau of Economic Analysis

A of the U.S. eco-nomic accounts produced by the Bureau of

Economic Analysis () was put in motion inmid-. Its purpose is to evaluate the perfor-mance of the economic accounts and to developa plan to maintain and improve that perfor-mance. The Review is one of three initiatives thatwill be used to guide ’s work over the nextdecade. The other two are a benchmarking of’s information technology system to guide thereengineering of ’s data collection, process-ing, and dissemination and a customer survey toguide a program of improved customer service at. The resulting overall plan is meant to help achieve its goal: Providing its customers withthe right numbers at the right time in the rightway.

The Review consists of three steps. The firststep was the preparation of a series of back-ground papers to evaluate the state of the eco-nomic accounts—their strengths, problems, and

February •

. This Review of the economic accounts does not include ’s systemof cyclical indicators, best known for its composite index of leading indicators.

. The Board of Governors of the Federal Reserve System maintains an-other set of economic accounts, the flow of funds accounts. The flow offunds accounts show the acquisition of physical and financial assets through-out the U.S. economy and the sources of funds used to acquire the assets.An associated set of balance sheets shows holdings of physical and financialassets.

prospects for improvement. These papers, whichwere made available in January , identifyneeds and conclude with a “menu” of recommen-dations for addressing these needs. The secondstep is to transform this menu into a priori-tized agenda, or strategic plan, to maintain andimprove the economic accounts over the nextdecade. This article presents a draft plan to usersof the accounts and others interested in their fu-ture. As a third step, the plan will be discussedand refined in a process that includes public com-ment, comment from the Federal agencies whoseassistance will be needed to implement some ofthe improvements, and a meeting in March of users of the accounts. With the refined planin hand, will be in a position to work withothers to develop detailed implementation plans.

As indicated in the table of contents, the draftplan is presented in part , structured aroundthe three priority issues it is designed to ad-dress. The first three parts provide summarybackground and may be read at several levelsof detail. Executive summaries, shown in ital-ics at the beginning of each part, allow readerswho wish to do so to proceed to the plan withminimal detail. The main text in parts – intro-duce economic accounting and ’s economicaccounts, describe the source data and estimatingmethods used to prepare them, and summarizethe background papers that identified issues to beaddressed in the draft plan. Boxes in each partsupport the main text by providing examples andfurther explanations.

Part . Economic Accounts

Economic accounts provide statistical picturesof the economy. ’s economic accounts—national, regional, and international—serve astools for tracking and projecting economic activity,for macroeconomic analysis of the economy’s work-ings, and for operational decisions in which theeconomy plays a major role. For these uses, accu-racy, reliability, and relevance are the interrelatedcharacteristics that are required of the accounts.

Economic accounting

Economic accounting—now in its sixth decadeas a specialty at the intersection of economicsand statistics—organizes economic information

about transactions and stocks, or holdings, toprovide complete and consistent statistical pic-tures of the economy. These pictures in-clude summary measures, component detail,and a framework that defines the interrelation-ships among the summary measures and theircomponents.

Economic accounting, and thus the economicaccounts themselves, are guided by theoreticallybased or empirically useful concepts. Theseinclude—to name just a few—income, consump-tion, and investment. The accounts resemble, ata macroeconomic level, the income and balancesheet accounts that describe the operations andgauge the performance of a business enterprise.

As statistical pictures, the accounts are designedto be complete in the sense that they count alleconomic transactions or stocks, but they do notdouble count. The estimates in the accounts aredesigned to be consistent with respect to whenthe transactions and stocks are recorded and howthey are valued.

’s economic accounts and their uses

The best known of ’s accounts are the na-tional income and product accounts (’s), thebalance of payments accounts (also called theinternational transactions accounts), and the re-gional accounts. The ’s show the Nation’sproduction, distribution, consumption, and sav-ing. The keystone of the ’s is their summarymeasure—gross domestic product (). Thebalance of payments accounts are known by theirsummary measures—the goods and services bal-ance and the broader current-account balance.The regional accounts provide estimates of per-sonal income for States and local areas and ofgross state product, the counterpart of forStates. In addition to these systems of accounts, maintains input-output accounts, which de-tail the interaction of industries, and wealthaccounts, which provide estimates of the Nation’sreproducible capital stock.

The ’s, the balance of payments accounts,the regional accounts, and the other accounts andstatistics prepared by are used by a widerange of individuals and organizations, both pub-lic and private. Business economists and Federal

• February

Questions That ’s Economic Accounts Help To Answer

National income and product accounts (’s)

The keystone of the ’s—gross domestic product ()—measures the market value of the goods and services produced inthe United States. From the product side, is the total of finalsales plus the change in business inventories (goods that have beenproduced but not yet sold). From the income side, it is the sumof costs, including compensation of employees and profits associ-ated with producing . The usefulness of stems in part fromproviding answers about the output of the economy—its size, itscomposition, and its use.

This system of accounts also traces the principal economic flowsamong the major sectors of the economy. In this way, the systemhelps to answer questions about the process by which output isproduced and distributed.

The ’s, sometimes described as the mainstay of macroeco-nomic analysis, are presented in a set of quarterly and annualtables. They help to answer questions such as these:

• How much has the Nation’s production, as measured by ,grown in the last quarter? the last year? the last decade?

• How much have prices—of goods and services produced inthe United States, or of goods and services purchased in theUnited States irrespective of where produced—increased overthe same periods?

• How much have standards of living, as measured by realpersonal income per capita, grown in the last decade?

• How much of goes for investment? How does the U.S.investment rate compare with that of other nations?

• How much of personal income goes to purchase goods andservices? to pay taxes? for saving?

Balance of payments accounts

The balance of payments accounts provide information on inter-national flows of goods, services, investment income, internationalassistance, and capital. This system of accounts also providesintegrated balance-sheet information on the U.S. international in-vestment position. The accounts help to answer questions such asthese:

• How large is the U.S. deficit in trade in goods? the surplus intrade in services? the surplus or deficit in investment income?

• How do the rates of return to foreign-owned companies in theUnited States compare with the rates of return to U.S.-ownedcompanies abroad?

• How much have foreigners invested in the United States?How much has the United States invested abroad?

Regional economic accounts

’s regional accounts provide estimates of gross state product andof total and per capita personal income by region, State, metropoli-tan area, and county. They help to answer questions such asthese:

• Which regions had the fastest growth in per capita income inthe last decade? The slowest growth?

• In which States is the share of residents’ income from wagesand salaries the highest? from dividends, interest, and rent?from transfer payments such as social security?

• Is gross state product from manufacturing becoming less con-centrated geographically? In which States are “high-tech”industries growing the fastest?

Other accounts and data

also produces the following sets of statistics that are related toits best-known accounts.

Input-output accounts: These national accounts detail the interac-tion of industries. They allow users to track the effects of changesin resource costs, or changes in final demand, on specific industries,on the users of these industries’ products, and on suppliers of laborand other products to these industries.

• How much of manufacturing industries’ inputs are fromother manufacturing industries? are from domestic indust-ries? are labor inputs?

• What are the effects on industry output of a general increasein exports? an increase in industrial chemical exports? adecrease in agricultural exports?

Wealth accounts: produces estimates of the Nation’s repro-ducible tangible wealth in the form of nonresidential structures andequipment, residential structures, consumer durable goods, and in-ventories. These accounts detail the U.S. capital stock by industryand by legal form of ownership.

• Over the last decade, how much has the Nation’s infrastruc-ture, as measured by its stock of fixed capital, grown? Howdoes this compare with past rates of capital formation?

• What portion of the capital stock is owned by persons? bycorporations? by others?

U.S. direct investment abroad and foreign direct investment in theUnited States: This detailed data set on the operations of foreign-owned companies is used to estimate investment income and capitalflows for the balance of payments accounts and holdings for theinternational investment position. By itself, it helps to answerquestions such as these:

• What percentage of the U.S. workforce is employed in foreign-owned companies?

• Which countries account for the largest share of foreign directinvestment in the United States?

• In what countries do U.S. companies invest? What share ofU.S. exports and imports are accounted for by trade betweenU.S. companies and their foreign subsidiaries?

February •

economic policy analysts use estimates from the’s to track business cycles, trends in over-all economic activity, and the contributions ofthe various sectors. (See the box, “QuestionsThat ’s Economic Accounts Help To An-swer,” on page .) Economic forecasting islargely directed toward explaining and predict-ing measures, and the behavioral equationsin econometric models—consumption function,investment function, profits equation—use vari-ables defined by the ’s. For example, the“Economic Assumptions” section of the Budgetof the United States Government is framed largelyin terms of measures. Academic and otherresearchers use various combinations of es-timates in econometric and other studies of suchtopics as economic growth, productivity, andsaving.

Estimates from the balance of payments ac-counts are used by policy analysts to track theflows of goods and services needed to assess theU.S. competitive position in world trade andby the Federal Reserve, the Treasury Depart-ment, and financial market analysts to monitorcapital flows and exchange-rate developments.The U.S. Government reports its balance ofpayments accounts to the International Mon-etary Fund in support of the Fund’s roles inmonitoring countries’ economic policies and inproviding financial assistance to help correctbalance-of-payments disequilibria.

Estimates from the regional accounts are usedby State Government offices to project tax rev-enue and to prepare State economic developmentpolicies, by Federal agencies in allocation formu-las for over billion in Federal funds annually,and by private industry for market analysis andplant location studies.

Public interest in ’s estimates is widespread,and its monthly and quarterly news releases arereported in the general, as well as the business,press. This wide coverage indicates both the de-gree of public interest in the state of the economyas depicted in the accounts and the trust thepublic places in the objectivity of ’s estimates.

Characteristics required of ’s accounts

The uses made of the estimates in the accountsdetermine the characteristics required of them.Some characteristics that may be required, suchas for comparability with the estimates pre-pared in other countries or for long time series,reflect specific uses. More generally, users re-quire three interrelated characteristics that may

be summarized in terms of accuracy, reliability,and relevance.

Accuracy.—In an abstract sense, accuracy may bethought of as referring to the level of an esti-mate. For example, an estimate of , which isa measure of the goods and services produced, isaccurate when it captures all production but doesnot double count. Thus, would be inaccu-rate to the extent that it did not fully capture theproduction of, for example, a new service suchas the “online” services now available to house-holds. is also inaccurate to the extent thatproduction is not correctly allocated between fi-nal purchases, which can be added up as one ofthe ways to measure , and intermediate pur-chases, which should be excluded from this total.For example, one difficult allocation that is madeby is for spending at restaurants; spendingby households should enter , but spending bybusiness should not.

However, users often focus on the change inlevels—for example, the change from quarter toquarter in real , or the change over a decadein State per capita personal income. Thus, mis-counted flows of goods and services, of income,or of other components that are subject to cycli-cal fluctuations or that grow or decline over timeare of particular concern because they affect thechange in levels.

Reliability.—Reliability refers to the size and fre-quency of revisions to ’s estimates. It isa characteristic that arises because additionalinformation used in preparing an estimate—information that is more complete, more de-tailed, or otherwise better—is incorporated intothe estimate as it becomes available over a periodof time. (See also the discussion of timeliness aspart of relevance.) Revisions, then, are indicativeof measurement error in estimates that do not in-corporate all the information that will eventuallybecome available.

Users express continuing concern about re-liability and the impact of revisions on theiranalysis, but on occasion they have been partic-ularly vocal. For example, for and , thefirst annual revised estimates of real showeda different picture than did the earlier estimates.According to the annual revision released in July, the cyclical peak in real occurred onequarter earlier than was initially estimated, andthe contraction was deeper than previously es-timated. In their annual report for , theCouncil of Economic Advisers noted that the reli-ability of such estimates is critical to policymakers

• February

and the public and made the comment, “Policymight have been conducted in a different fash-ion if the true severity of the recession had beenknown earlier.”

’s most accurate estimates presumably arethe ones for which there will be no furtherrevisions—that is, the ones in which all the in-formation that will become available has beenincorporated. Even for these estimates, thereremains some unknown amount of inaccuracy.Measures of this error cannot be determined foreconomic accounting estimates even to the extentthat they can for statistical surveys for several rea-sons: Not all information used in the accounts isfrom surveys, the relevant measures of error fromdata based on surveys are not always available,and even if they are, they cannot meaningfullybe added together.

In practical terms, estimates that undergo littlerevision as additional information is incorpo-rated may accordingly be viewed as reliable;however, they may or may not be accurate. Forexample, if the information from which estimatesare prepared has no gaps in its intended coverage,the estimates are both reliable and accurate. If,however, the information has gaps in coverage,the estimates may be viewed—strange as it mayseem—as reliable but inaccurate.

Relevance.—Relevance has at least two dimen-sions. One, timeliness, refers to the length oftime between the close of the period to which theestimates refer and the release of the estimatesfor that period. Timeliness can be viewed as adimension of relevance because estimates that arenot available soon enough for a particular in-tended use are, in fact, irrelevant for that use. Inthis sense, relevance is often thought of in thecontext of a tradeoff with accuracy and reliabil-ity, and the preferred tradeoff depends on the useto be made of the estimates. For estimates usedin tracking cyclical and other short-term devel-opments, there is a premium on timeliness; forestimates that show detailed interrelationships,such as input-output accounts, there is more ofa premium on accuracy.

The second dimension of relevance refers tothe ability of the accounts to provide summarymeasures, the kind and amount of detail, and theanalytical frameworks that answer the questionsabout the economy that are being asked. Issuesof relevance change as the economy changes, aspolicy concerns evolve, and as economic theory

. Economic Report of the President (Washington, : U.S. GovernmentPrinting Office, January ): .

advances. For example, in the mid-’s, rele-vance was the attribute at issue when users soughtthe development of constant-dollar quarterly es-timates of to supplement the current-dollarmeasures, and it was the attribute at issue when,as international services grew in importance inworld trade, users sought to focus on a tradebalance that included both goods and services.

Part . Source Data and Methods Used toPrepare the Accounts

Source data are the information uses to pre-pare estimates, and estimating methods are thesteps takes to transform these data into esti-mates. The accounts are built up as a mosaic froma wide range of source data using a variety of esti-mating methods to adjust the available source datato the concepts needed, to fill in gaps in coverage,and to obtain the time of recording and valuationneeded for the accounts. The interaction of sourcedata and estimating methods determines the ac-curacy and reliability of the accounts and sets thestatistical limits for estimating relevant measures.Source data also determine the initial release andrevision schedules for the estimates.

The interaction of source data and estimatingmethods

In an ideal world, source data for each de-tailed component of the accounts would mapexactly to the concept specified by the accountingframework. Further, the source data would beaccurate, would have the needed coverage, wouldhave the needed time of recording and valuation,and would be available quickly.

Of course, source data do not fit this descrip-tion, nor are they ever likely to do so. Individualsand businesses do not keep their records in away that would make it easy to provide what needs. Most individuals do not keep recordsof what they spend at the level of detail neededfor the accounts. Many businesses—for example,restaurants and gasoline service stations—do notkeep records that distinguish sales to a businessfrom sales to a household, and many businessesdo not keep records on the commodity compo-sition of their inventories or on the geographicbreakdown of their profits. Many businesses—especially small businesses—do their accounting

February •

annually and cannot provide monthly, or evenquarterly, data. As a result, to provide witheverything needed to prepare the accounts wouldimpose an unacceptably high reporting burdenon respondents and would greatly increase thecost to the government of collecting statistics.Thus, the source data that uses are collectedfrom a variety of sources and, with few excep-tions, for purposes other than the preparation of’s estimates.

Data collected by Federal Government agen-cies provide the backbone of the estimates; thesedata are supplemented by data from trade asso-ciations, businesses, international organizations,and other sources. The Government data arefrom a number of agencies, mainly the Com-merce Department’s Bureau of the Census, theLabor Department’s Bureau of Labor Statistics,the Internal Revenue Service and other agen-cies of the Treasury Department, the Office ofManagement and Budget, and the AgricultureDepartment. Some of the Government-collecteddata, referred to as “administrative” data, arebyproducts of government functions such as ed-ucation programs, tax collection, defense, andregulation. Nonadministrative data, sometimesreferred to as “general purpose” or “statistical”data, include the periodic economic and popula-tion censuses and a wide range of sample surveys,such as those that collect data on manufacturingand farm activity, prices, and corporate profits.Of the relatively few data items that collects,most refer to international transactions. Theseinclude international trade in services and di-rect investment (both by foreign residents in theUnited States and by U.S. residents in foreigncountries).

Because the source data available are not idealfrom the point of view of preparing the economicaccounts, must “make do.” To continue themosaic metaphor for the accounts, the sourcedata are the small pieces of colored stone, and must develop estimating methods to shapethe stones to fit and hold them together accordingto the artist’s design. The estimating methods ad-just the best available data to the concepts neededfor the accounts, fill gaps in coverage of thesource data, and make adjustments to the sourcedata to obtain the needed time of recording andvaluation. Some examples of these estimatingmethods follow.

Adjustments to needed concepts:

• Internal Revenue Service Corporate Returnsdata include gains (net of losses) from the

sale of property in corporate profits. subtracts these gains to arrive at a meas-ure of profits consistent with the concept ofincome from current production underlyingthe income side of .

• Wages and salaries reported in the State em-ployment security agencies’ tabulations ofpayroll data are on a place-of-work basis. makes a residence adjustment to putthem on the needed place-of-residence basisfor personal income by State.

Filling gaps in coverage:

• Census Bureau censuses and surveys of tradedo not include inventories of nonmerchantwholesalers. estimates the change in theinventories of these wholesalers—of petro-leum bulk stations, using physical quantitiesand a price index, and of manufacturing salesbranches, using changes in the correspondingmanufacturing industries—to provide fullcoverage of wholesalers for the change inbusiness inventories component of .

• The monthly survey of establishments con-ducted for the Bureau of Labor Statisticsincludes data on employment and on averageweekly hours and average hourly earningsof production and nonsupervisory workers,which are used by to estimate wages andsalaries. The data do not include bonus pay-ments, such as are common in the securitiesindustry. makes estimates of these bonuspayments, often on the basis of fragmen-tary information, to provide fuller coverageof wage and salary disbursements for theinitial monthly and quarterly estimates of personal income and for the prelim-inary quarterly estimates of State personalincome.

Adjustments to needed time of recording andvaluation:

• Financial statements for State and local gov-ernments, compiled and published by theCensus Bureau in Census of Governments andGovernment Finances, report receipts and ex-penditures data on a fiscal year basis that isnot uniform for all governments. For exam-ple, in estimating the government purchasescomponent of , makes adjustmentsto put the data on a calendar year basis.

• Depreciation charges used by business in taxaccounting—which are compiled and pub-lished by the Internal Revenue Service in

• February

Corporate Returns, Partnership Returns, andSole Proprietorship Returns—reflect severaldepreciation patterns and service lives. adjusts profits and other business incomesto put the depreciation charges on a consis-tent accounting basis (based on straight-linedepreciation and uniform service lives).

• Data on imports from Canada, as received bythe Census Bureau in a data exchange withCanada, value certain goods at the point ofmanufacture. adds the cost of inlandtransport to provide the valuation at pointof foreign export to be consistent with othertrade data and approximate the market priceused throughout the accounts.

The source data uses represent a vari-ety of different economic statistics—wages andsalaries, premiums, interest rates, mortgage andother debt outstanding, tax collections, employ-ment, unit sales, and average prices, to name afew. In most cases, the source data are “valuedata”; that is, they embody both the quan-tity and price dimensions that are required forcurrent-dollar estimates. (Most of the estimat-ing methods just sampled are adjustments tovalue data.) When value data are not used, datawith separate quantity and price dimensions arecombined to derive the required value estimates(along with any needed adjustments). For ex-ample, the estimate for purchases of new autosis prepared by multiplying unit sales by averageprice.

For estimates of , gross state product, tradein the balance of payments accounts, and otherseries that measure goods and services, usesadditional source data for the preparation ofprice-adjusted, or real, estimates. The estimatingmethod used for most of these series is defla-tion. In deflation, real estimates are obtained bydividing the most detailed current-dollar compo-nents by appropriate prices indexes, with the baseperiod—at present, the year —equal to .Components of the Consumer Price Index, theProducer Price Index, and the International PriceIndex prepared by the Bureau of Labor Statis-tics are the source data used to deflate manycomponents.

Source data as determinants of initial releaseand revision schedules

Because source data are the colored stones—theessential material—in ’s mosaics, they largelydetermine the schedules for the initial release ofthe estimates and the schedules on which they

are revised. One factor is the speed with whichthe source data first become available. A secondis whether or not the source data are part of aprogram that, over time, provides more completeor otherwise better coverage—for example, if thesample is larger for annual than quarterly surveysor if the amount of detail is larger for annualsurveys.

For the first, or advance, quarterly estimateof , the availability of the monthly series onsales, shipments, and trade in goods from theCensus Bureau (along with the time it takes to process it) determines the release date. Oncethese data become available, the initial estimateof each major component of can be basedon at least months of source data or on reliable projections. For the quarterly balance of pay-ments estimates, the availability of monthly serieson exports and imports of goods determines therelease schedule. For the quarterly State personalincome estimates, the availability of wage andsalary data from the States plays a similar role.

The estimating schedule for illustrates thelink with source data that are part of a pro-gram that, over time, provides better coverage.In general, the most comprehensive source datafor the product components of are avail-able at the -year intervals associated with theeconomic censuses conducted by the Census Bu-reau. The economic census data are used to“benchmark” ’s estimates for the quinquen-nial census years—for example, , , and. The related annual surveys are drawn froma sample of establishments covered in the censusand provide less detailed data than the census. Asmaller sample provides monthly data for most ofthe annual surveys. These monthly data are usedto produce the monthly and quarterly estimatesof several product components. These estimatesare revised when more reports become availablefrom the monthly samples, when data from theannual surveys become available, and when datafrom the census become available; thus, a givenestimate of a product component may berevised as many as six times over a -year period.

Source data and methods: Examples, specialrequirements, and publications

An accompanying box, “Examples of Source Dataand Methods Used To Prepare the EconomicAccounts,” demonstrates several of these points:The variety of source data uses, the methods has developed to provide estimates that fitthe concepts, coverage, and other needs of theaccounts, and the incorporation by of more

February •

Examples of Source Data and Methods Used To Prepare the Economic Accounts

Example estimate from the ’s: Personal consumption ex-penditures on goods

Personal consumption expenditures () makes up about two-thirds of , and durable and nondurable goods account for overtwo-fifths of that (services account for the remainder). The tablebelow shows the methodology for the advance quarterly and annualcurrent-dollar estimates for goods (with estimates for ).

In the context of the discussion of sources and methods, there areseveral points of interest:• Diverse source data: Among the source data are the Census

Bureau’s sequence of a monthly survey, an annual survey, and a cen-sus for retail trade; the Bureau of Labor Statistics Consumer PriceIndexes; data from several trade sources; administrative data fromseveral Federal agencies; and data from ’s balance of paymentsaccounts.• Sequence of revisions: The “most goods” category undergoes a

sequence of revisions as the source data become more complete.

Subcomponent (billions of dollars) ofPCE, which was $4,378.2 billion

Annual estimates: Source data and methods used to determine level for bench-mark and other final years or, for other years, used to prepare an extrapolator or

interpolator

Advance quarterly estimates: Sourcedata and methods used to prepare an

extrapolator

Durable and nondurable goods:($1,877.2) 1

Most goods (goods exceptsubcomponents listed separately)($1,562.0).

Benchmark years—Commodity-flow method, starting with manufacturers’ shipmentsfrom Census Bureau quinquennial census and including an adjustment forexports and imports from Census Bureau merchandise trade.

Other years—Retail-control method, using retail trade sales from Census Bureauannual survey or, for most recent year, monthly survey of retail trade.

Same as annual for most recent year.

New autos ($93.4) .................................... Physical quantity purchased times average retail price: Unit sales, information withwhich to allocate sales among consumers and other purchasers, and averagelist prices, all from trade sources.

Same as annual.

Net purchases of used autos ($45.9) ...... Benchmark years—For net transactions, change in the consumer stock of autosfrom trade sources. For dealers’ margin, retail sales from Census Bureauquinquennial census and margin rate from Census Bureau annual survey ofretail trade.

Other years except most recent—For net transactions, same as benchmark. Fordealers’ margin, franchised dealers’ unit sales times sales price, both from tradesources, times margin rate for independent dealers from Census Bureau annualsurvey; independent dealers’ margin from Census Bureau annual survey.

Most recent year—For net transactions, same as benchmark. For dealers’ margin,for franchised dealers, unit sales and sales price from trade sources; forindependent dealers, sales from Census Bureau monthly survey of retail trade.

For net transactions, residual based onnet sales by other sectors. Fordealers’ margin, unit sales offranchised dealers from trade sourceand sales price from Bureau of LaborStatistics consumer price index forused cars.

New trucks ($52.3) ................................... Benchmark years—Commodity-flow method, starting with manufacturers’ shipmentsfrom Census Bureau quinquennial census and including an adjustment forexports and imports from Census Bureau merchandise trade.

Other years except most recent—Abbreviated commodity-flow method, startingwith manufacturers’ shipments from Census Bureau annual survey andincluding an adjustment for exports and imports from Census Bureaumerchandise trade.

Most recent year—Physical quantity purchased times average retail price: Unitsales and information with which to allocate sales among consumers and otherpurchasers from trade sources and average price based on Bureau of LaborStatistics consumer price index for new trucks.

Same as annual for most recent year.

Gasoline and oil 2 ($105.6) ....................... Benchmark years—Physical quantity purchased times average retail price: Gallonsconsumed from the Department of Transportation, information with which toallocate that total among consumers and other purchasers from Federalagencies and trade sources, and average retail price from Census Bureauquinquennial census.

Years except most recent—Same as benchmark years, except average retail pricefrom the Energy Information Administration.

Most recent year—Physical quantity purchased times average retail price: Gallonsconsumed and average price both from the Energy Information Administration.

Same as annual for most recent year.

Food furnished to employees (includingmilitary) ($12.0).

Benchmark years—For commercial employees, number of employees ofappropriate industries from Bureau of Labor Statistics tabulations times BEAestimate of per capita expenditures for food; for military personnel, outlays fromthe Budget of the United States prepared by the Office of Management andBudget.

Years other than benchmark years—Same as benchmark years, except per capitaexpenditures for food based on Bureau of Labor Statistics consumer price indexfor food.

For commercial employees, same asannual for years other than benchmarkyears; for military personnel,judgmental trend.

Expenditures abroad by U.S. residents($3.2) less personal remittances inkind to nonresidents ($0.8).

Estimated as part of the balance of payments; see the entry for service exportsand imports, net, under net exports of goods and services.

Judgmental trend.

1. Includes $3.6 billion for food produced and consumed on farms, standard clothing issued to military per-sonnel, and used trucks.

2. The retail-control method cited under PCE for most goods is based on retail trade sales data that include

sales of gasoline service stations. Estimates of PCE for gasoline and oil are derived separately and are deduct-ed from the retail-control totals (that include goods sold by gasoline service stations) to derive the estimatesfor PCE for most goods.

Census Bureau retail sales data are available for all months ofthe quarter at the time of the advance quarterly estimate; data forthe third month are preliminary and subject to revision. The retailsales data are further revised with the release of the annual survey,and these are incorporated into the estimates in the secondannual revision. The data are further revised with the release of theeconomic census, and these are incorporated into the estimatesin the comprehensive revision.

• Role of estimating methods: The retail control method mentionedin the “most goods” description provides both an indicator seriesused in interpolating and extrapolating and a total to which thecategories in the group must sum. This method makes it possibleto use retail sales by type of business to obtain type-of-goods detail(assuming that the types of goods purchased at various kinds ofstores do not change rapidly).

Source: S C B, July .

• February

complete and more consistent source data as theybecome available. (The box begins on page and continues on pages and .)

All of ’s accounts are built up from sourcedata in a similar way, but each system of accountshas special requirements according to the dimen-sions of the economy on which it focuses. For theregional accounts, data must be found that eitherare available by State or county and add up to areliable national total, as do the wage and salarydata, or that can be used to allocate a nationaltotal to the States and counties. Of particularconcern is the distinction between data that areon a place-of-residence basis (such as receipts ofdividends, interest, and rental income) and datathat are on a place-of-work basis (such as wagesand salaries and other labor income). For theinternational accounts, data must be found thatdistinguish between transactions and holdingsof residents and nonresidents, with a particularfocus on the geography of the nonresidents.

publishes papers that describe in detail themethodologies it uses—that is, its source dataand methods. Changes in methodologies—forexample, when a source is discontinued or whennew source data are introduced—are typically de-scribed in the S C B.For example, the source data for each compo-nent of are published in the S articleson the annual revisions of the accounts. Pub-lishing these methodologies provides users withinformation to evaluate the estimates and theirsuitability for actual and intended applications.

The effect of source data and methods on theperformance of the accounts

The interaction of source data and estimatingmethods determines the accuracy and reliabilityof the accounts and sets the statistical limits forimplementing relevant measures. This sectionuses examples to make this point.

Accuracy.—As explained in part , accuracy dealswith the possibility of error in the level of, and

. A complete list of ’s methodologies is in User’s Guide to Infor-mation, which is updated annually (most recently in the January S C B). The methodology is being described in a se-ries of papers; see the listing in User’s Guide to Information. See alsoU.S. Department of Commerce, Bureau of Economic Analysis, BenchmarkInput-Output Accounts of the United States, (Washington, : U.S. Gov-ernment Printing Office, November ). The methodologies for personalincome in the regional accounts are in U.S. Department of Commerce, Bu-reau of Economic Analysis, State Personal Income, – (Washington, :U.S. Government Printing Office, forthcoming) and in U.S. Department ofCommerce, Bureau of Economic Analysis, Local Area Personal Income, – (Washington, : U.S. Government Printing Office, September ). Themethodology for the balance of payments is in U.S. Department of Com-merce, Bureau of Economic Analysis, The Balance of Payments of the UnitedStates: Concepts, Data Sources, and Estimating Procedures (Washington, :U.S. Government Printing Office, May ).

change in, the estimates. One of the two exam-ples that follow focuses on gaps in coverage of thesource data that affected both the level and rateof change in the estimates. The other focuses onthe effect of an estimating method on the rate ofchange in the estimates.

International portfolio capital flows.—During the’s, major advances in computer and com-munications technology, combined with dereg-ulation in financial services industries, sparkeddynamic changes in global financial markets.These changes opened new financing channelsand inspired the introduction of new financialinstruments. These developments, in turn, led togaps in ’s coverage of international transac-tions, particularly international flows of portfoliocapital. Inaccurate portfolio capital measuresalso had implications for the current account be-cause the capital positions are used in estimatingincome flows.

To meet the need for improved coverage ofthese transactions, launched a multiyear ef-fort starting early in the ’s. First, greatlyexpanded the use of counterparty data—that is,data from the country in which the foreign trans-actor is resident—in the June revision ofthe balance of payments accounts, which covered–. Use of these data added nearly bil-lion in capital outflows and nearly billion incapital inflows that were previously not recorded.

Second, has supported expansion in thecoverage of the Treasury Department’s surveysof portfolio investment to capture direct trans-actions between large U.S. pension funds andinvestment managers, on the one hand, andforeign residents, on the other. Previously,these transactions had bypassed the survey sys-tem, which was based on data provided by U.S.financial intermediaries. Third, the Treasury De-partment is currently processing a benchmarksurvey of U.S. portfolio investment abroad for, the first in more than years.

Although the use of counterparty data hashelped fill gaps in coverage of international cap-ital flows, increased use of such data to capturedirect financial transactions will require furtherwork in establishing standardized definitions anddata collection systems across countries. Inaddition, counterparty data are unlikely to beavailable to provide measures of new finan-cial instruments, one of the largest and mostrapidly widening gaps in coverage. Accordingto data collected by the Bank for InternationalSettlements from its reporting banks alone, thenotional principal value on interest rate swaps—

February •

one of the new derivative instruments—grewfrom . trillion in to . trillion in ,and the value of currency swaps grew from .trillion in to . trillion in . Consis-tent data on U.S. transactions in derivatives, or

Examples of Source Data and Methods

Example estimate from the balance of payments accounts:actions in securities other than U.S. Treasury securities

Portfolio investment—that is, securities transactionsU.S. and unaffiliated foreign residents other than foreignagencies—is measured in two capital accounts: “U.S. securitforeign purchases,” which was . billion in , andsecurities—net U.S. purchases,” which was . billion.

The securities covered are stocks and bonds with an origiturity of more than year. The accounts cover new issueUnited States and abroad as well as trading in, and redemptoutstanding U.S. and foreign securities.

The primary source data are from the U.S. DepartmenTreasury’s International Capital reporting system—specificamonthly S form, “Purchases and Sales of Long-Term SecurForeigners.” Filing of S forms is required for all securitiesdealers, and other persons in the United States who, on thbehalf or on behalf of customers, engage in transactions iterm securities with foreigners.

The methodology described in the table below (which follorder of the standard quarterly balance of payments tables“foreign securities—net U.S. purchases”; the methodology fosecurities—net foreign purchases” is similar.

In the context of the discussion of sources and methods, tseveral points of interest:• Source data coverage: The S form covers mainly interm

transactions; however, direct transactions for large pensionvestment managers are covered.• Role of estimating methods: makes numerous adjus

some from fragmentary data, to arrive at the needed coverexample, with respect to coverage of commissions, taxes, ancharges) and timing (for example, transactions not yet recoS-form data).

• Relations among the estimates: Positions—that is, outsholdings—that correspond to these capital flows are shownnet international investment position; the positions are estimcumulating the capital flows from periodic benchmark survelevels and adjusting for price change. (The positions, alongestimated representative yield, are used to estimate income flthe current account of the balance of payments.)

Line Foreign securities, net U.S. purchases (-)

1 Total ...........................................................................................................2 Stocks .................................................................................................................3 New issues in the United States ..................................................................4 Transactions in outstanding stocks, net .......................................................5 Bonds .................................................................................................................6 New issues in the United States ..................................................................7 Redemptions of U.S.-held foreign bonds .....................................................8 Other transactions in outstanding bonds, net ..............................................

the market value of U.S. exposure to foreign riskthrough derivative instruments, are not available.

Fixed-weighted real .—Rapid change in thecomposition of output and in relative prices has

Used To Prepare the Economic Accounts—Continued

Trans-

betweenofficial

ies—net“foreign

nal ma-s in theions of,

t of thelly, theities by

brokers,eir ownn long-

ows the) is forr “U.S.

here are

ediatedand in-

tments,age (ford otherrded in

tandingin the

ated byy-basedwith anows for

1993(Billionsof dol-lars)

−120.0−60.6−13.8−46.8−59.4−46.8

8.9−21.5

Net U.S. purchases of foreign stocks (line ): estimates nettransactions in outstanding stocks (line ) as follows:

() Data on gross sales and gross purchases of foreign stocks inthe United States by foreigners are based on the monthly Sreports. adjusts the data to exclude estimates of commis-sions, taxes, and other charges from reported gross foreignpurchases; to include estimates of commissions, taxes, andother charges from reported gross foreign purchases; and toinclude estimates of charges in reported gross foreign sales.

() The value of stocks representing U.S. direct investment abroadis deducted from the net figure.

() The value of stocks exchanged as part of a foreign directinvestment in the United States is added.

() Other adjustments include additions or subtractions fortransactions that have not yet been incorporated into theTreasury data and additions for transactions that have beenomitted, but that have been verified from other sources.

estimates new foreign issues (line ) on the basis of financialmarket information.

Net U.S. purchases of foreign bonds (line ): estimates nettransactions in outstanding bonds other than redemptions (line )as follows:

() Data on gross sales and gross purchases in the United Statesby foreigners of foreign corporate bonds are based on themonthly S reports. adjusts the data to include estimatesof commissions, taxes, and other charges from reported grossforeign purchases and to include estimates of underwritingfees on new issues, other fees, taxes, and other charges inreported gross foreign sales.

() Adjustments are made to the data covering U.S. purchasesof Canadian bonds to account for additional redemptions ofCanadian-issued bonds that are held by U.S. residents.

() Other adjustments include additions or subtractions fortransactions that have not been incorporated into the Treas-ury data; additions for transactions that have been omittedfrom the S form, but that have been verified from othersources; and additions for acquisitions of foreign debt securi-ties by U.S. residents through the exchange of securities withforeign residents, including those resulting in foreign directinvestment in the United States.

estimates new foreign issues (line ) on the basis of financialmarket information, with separate estimates by type of issuer.

estimates bond redemptions (line ) from information onscheduled retirements.

Source: U.S. Department of Commerce, Bureau of EconomicAnalysis, The Balance of Payments of the United States: Concepts,Data Sources, and Estimating Procedures (Washington, : U.S. Gov-ernment Printing Office, May ) as updated in the June issues ofthe S C B, –.

• February

. Both of the alternatives are what are known in the economic literatureas “Fisher” or “superlative” price indexes.

brought into question the longstanding methodsthat underlie constant-dollar and other esti-mates. Constant-dollar estimates are the familiarkind of “real” estimates, the ones that are cur-rently denominated in dollars because theyvalue each component in the price of a base yearthat is currently . This method is the equiv-alent of output measures that have fixed priceweights—those of the base year. (Similarly, thefixed-weighted price indexes have fixed outputweights—those of the base year.) Use of thesame (fixed) price weights over all time periodsprovides a set of estimates in which the compo-nents add up to the totals. has featured suchmeasures partly because many users consider thisadditive property to be useful—for example, itfacilitates analysis of contributions to growth andchanges in shares of economic activity.

The professional literature has long recog-nized that output measures that have fixed priceweights tend to overstate current-period growthas one moves further from the base period. Thistendency, often called “substitution bias,” reflectsthe fact that the commodities for which out-put grows rapidly tend to be those that registerdeclines, or the smallest increases, in prices.

Two developments contributed to the need toinvestigate alternatives that did not exhibit thesubstitution bias found in the fixed-weighted in-dex. First, beginning in the ’s, changes inthe prices and quantities of the energy and foodcomponents of were large enough in certainperiods for the choice of price weights to affectthe measurement of change in real . Second,computers provided a classic, but extreme, case ofthe source of the bias: Computer prices have de-clined rapidly, while computer output has grownrapidly.

Recognizing the important effect of changes inrelative prices on real growth rates, initi-ated a research program to investigate alternativemeasures. In April , published two alter-native measures of annual change in real forthe period –, and in March , be-gan publishing them for quarterly changes. Thetwo alternative measures are not based on theprice weights of a single base year. Rather, theyare indexes that account for changes in relativeprices over the periods for which growth rates arecomputed. In the chain-type annual-weightedquantity index, the weights are from adjacentyears. In the benchmark-years-weighted quantity

index, the weights are from adjacent benchmarkyears—about -year intervals.

Further work in this area will involve examina-tion of other methods that attempt to combinethe advantages of fixed-weighted measures withthe more up-to-date weights embodied in thealternative measures.

Reliability.—As explained in part , reliability isgauged by the size and frequency of revisions. Itshould be noted that filling gaps, such as thosejust described for international capital flows, andmaking other improvements give rise to revisions.Thus, in these cases, revisions are a symptom thatimprovements have been made. In other cases,revisions are a symptom that there is potentialfor improvement.

Among the major sources of revision are thefollowing: Incorporation of source data withmore complete reporting, replacement of judg-mental projections with source data, incorpo-ration of source data that more closely matcheconomic accounting concepts and other needs,and incorporation of updated seasonal adjust-ments. These sources of revision, reflecting theinteraction of source data and estimating meth-ods, are explained by referring to the results of studies of revisions.

Incorporation of source data with more completereporting.—Some revisions are due to the in-corporation of revisions in monthly source datathat embody more complete reporting. Amongthe larger sources of revision from the advanceto the preliminary current quarterly estimates of are the incorporation of revised Census Bu-reau data on retail sales, manufacturing and tradeinventories, manufacturers’ shipments, and newconstruction put in place. Revisions from thepreliminary to the final current estimates, andfrom the final current estimates to the annualand comprehensive revision estimates, are oftendue to the introduction of annual and benchmarksurveys that are progressively more comprehen-sive in coverage than the quarterly and monthlydata. For example, large revisions in personalconsumption expenditures for goods in the annual revision were mainly due to revisions thatreflected the incorporation in the data of the largeand growing number of discount “clubs,” whichhad not been fully covered in the monthly andannual retail trade surveys.

February •

Examples of Source Data and Methods Used To Prepare the Economic Accounts—Continued

Example estimate from the regional accounts: Wage and salarydisbursements

Wage and salary disbursements makes up about percent of per-sonal income. The estimates are prepared by industry. As shownin the table that follows, there are two major sources of data: ()The monthly survey of more than , nonagricultural estab-lishments conducted by the State employment security agencies andcoordinated by the Bureau of Labor Statistics (the - series)and () the tabulations (sometimes called the - series) of thewages and salaries reported by employers on their quarterly un-employment insurance () tax returns to the State employmentsecurity agencies.

The State-level - series is used as the extrapolator for thepreliminary quarterly estimates. The State-level series is used asthe extrapolator for the second quarterly estimates, as the princi-pal basis for the revised annual State estimates (which incorporatemore detailed and more reliable data), and as the interpolator ofthe annual estimates used to prepare the revised quarterly State es-timates.

The preliminary, second, and revised quarterly State estimatesand the revised annual State estimates of wages and salaries are allcontrolled to the totals in the ’s. The quarterly and monthlynational estimates of wages and salaries are based mainly on the

Industry: 1 (Billions of dollars in 1993; U.S.total: $3,072.3)

Preliminary quarterly estimates (4months after the close of quarter)

Second quarterly estimates (7 monthsafter the close of quarter)

Revised annual estimates (8 monthsafter the close of the year)

Farm ($11.9) ............................................... Trend extrapolation 2 .............................. Trend extrapolation 2 .............................. USDA estimates of farm labor expensesAgricultural services, forestry, fisheries,

and other ($16.2).Trend extrapolation ................................. Quarterly UI data for wages and

salaries.Annual UI data for wages and salaries

and USDA estimates of farm laborexpenses

Mining ($25.8) ............................................. BLS–790 monthly employment data ...... Quarterly UI wage data .......................... Annual UI wage dataConstruction ($132.8) ................................. BLS–790 monthly employment data ...... Quarterly UI wage data .......................... Annual UI wage dataManufacturing:

Nondurable goods ($235.0) ................... BLS–790 monthly data for employmentand average weekly hours andaverage hourly earnings forproduction and nonsupervisoryworkers.

Quarterly UI wage data .......................... Annual UI wage data

Durable goods ($353.5) ......................... BLS–790 monthly data for employmentand average weekly hours andaverage hourly earnings forproduction and nonsupervisoryworkers.

Quarterly UI wage data .......................... Annual UI wage data

Transportation and public utilities:Excluding railroads ($188.6) .................. BLS–790 monthly employment data ...... Quarterly UI wage data .......................... Annual UI wage dataRailroads ($12.9) .................................... Monthly ICC payroll data and AAR

State employment data for Class Irailroads.

Monthly ICC payroll data and AARState employment data.

Annual ICC payroll data and AAR Stateemployment data

Wholesale trade ($204.8); retail trade($295.4); finance, insurance, and realestate ($250.7).

BLS–790 monthly employment data ...... Quarterly UI wage data .......................... Annual UI wage data

Services (770.8) ......................................... BLS–790 monthly employment data ...... Quarterly UI wage data .......................... Annual UI wage data, data fromCensus Bureau County BusinessPatterns and Census Bureaupopulation data

Federal civilian ($115.0) ............................. BLS–790 monthly employment data ...... BLS–790 monthly employment data ...... Annual UI wage dataFederal military:

Active duty ($42.0) ................................. DOD number of personnel and averagepay by service and Coast Guardpayroll data.

DOD number of personnel and averagepay and Coast Guard payroll data.

DOD and Coast Guard data

Reserves ($7.5) ...................................... Trend extrapolation ................................. Trend extrapolation ................................. DOD payroll outlay dataState and local government ($409.5) ........ BLS–790 monthly employment data ...... Quarterly UI wage data .......................... Annual UI wage data

1. The quarterly estimates of wages and salaries are prepared at the standard industrial classification (SIC)division level, and the annual estimates are prepared at the SIC two-digit level.

2. The trend extrapolation is based on a regression that estimates a State’s share of the Nation historicallyby industry.

AAR Asssociation of American RailroadsDOD Department of DefenseICC Interstate Commerce CommissionUI Unemployment insuranceUSDA U.S. Department of Agriculture

- series; the annual estimates are based mainly on the se-ries.

In the context of the discussion of sources and methods, there areseveral points of interest.• Sequence of revisions: The data, which become available only

in time for the second quarterly estimates ( months after the quar-ter to which the data refer), include the wages and salaries of allemployees, whereas the - wage data for States are confined tothe wages of production and nonsupervisory workers in the man-ufacturing industries. The wage data include bonus payments,which are not reflected in the - data.• Relation of State and national estimates: Until the preliminary

annual State estimates for (published in April ), hadautomatically controlled the annual estimates for the most currentyear to the --based U.S. totals for wages and salaries in the’s. Now, if the national total of the wage data for the firstthree quarters of a year differs substantially from the --basedestimates for those quarters, the U.S. total for wages and salariesthat is used for the preliminary annual State estimates is based onthe wage data. This change reduces the revisions between thepreliminary and final annual State estimates of wages and salariesand personal income.

Source: Table H in the methodology text in State Personal Income,– (forthcoming).

• February

Replacement of projections with source data.—Some other revisions are due to the replacementof ’s judgmental projections for the quartersand the year with newly available annual and eco-nomic census data. Some of the largest revisionsto have come from annual services surveydata and from annual State and local govern-ment finances data. Data such as State and localgovernment purchases other than structures andemployee compensation are available only annu-ally; quarterly estimates must be extrapolated andinterpolated.

Incorporation of source data that more closelymatch the needs of the accounts.—Revisions tothe quarterly and annual estimates can comefrom the introduction of more comprehensivesource data or alternative source data that aremore consistent with the concepts, coverage,time of recording, and valuation needed for theeconomic accounts. For example, revisions tocorporate profits are often sizable when tabu-lations of tax-return data replace tabulations ofpublicly available shareholder reports. The tax-return data provide coverage of all firms andall industries, and they are based on accountingguidelines that are closer to concepts thanare the shareholder reports.

Seasonal adjustment.—Another of the sourcesof revision to the quarterly estimates, and onethat affects many components, is seasonal ad-justment. For some series, particularly volatileseries (such as merchandise trade, inventories,and structures), the effect of revisions in seasonaladjustments may account for a large part of therevision between the final current estimate andthe first annual revision estimate. In most ofthese cases, however, the revisions shift growthbetween adjacent quarters and have little effecton the general picture of economic growth.

Relevance.—In the mid-’s, there is probablymore concern about the ability of the accountsto provide answers about the economy to poli-cymakers, analysts, and others than about theirtimeliness, the other aspect of relevance. (Themajor exceptions to this statement are concernsabout the timeliness of ’s input-output tablesand about some of the regional estimates.)

Within the existing structure of the accounts,the source data and estimating methods set thestatistical limits for implementing some relevantmeasures. For example, there is considerableinterest in identifying purchases of computersoftware and in making more consistent use of

quality-adjusted prices for preparing real esti-mates of “high-tech” goods and services, but bothimprovements would require new source dataand estimating methods.

Looking further, changes in the economy, theevolution of policy concerns, and advances intheory suggest changes in the structure of theaccounts and the definitions embodied in themthat would enhance the relevance of the es-timates. One longstanding set of questionsinvolves the definition of investment. Investmentin the ’s consists of purchases of struc-tures, durable equipment, and inventories by theprivate business sector. Should other sectors—notably government—be viewed as making in-vestment when structures and durable equipmentare purchased? Should some purchases of in-tangibles, such as research and development, betreated as investment? Once economic theory,experience in economic accounting, interest ininternational comparability, and other elementsare brought to bear in deciding on a change indefinition to enhance relevance, the implemen-tation would again require new source data andestimating methods.

In addition, satellite accounts may be devel-oped, as described in the newly revised inter-national guidelines in the System of NationalAccounts , to expand the analytical capacityof the accounts. has prepared prototypesof two satellite accounts—one to show the in-teraction of the economy and the environmentand the other to identify research and develop-ment expenditures within the ’s and to showthe capital stock that results from treating themas investment. Both efforts pointed to sourcedata and methodological improvements neededto strengthen and extend the estimates. Addi-tional satellite accounts, such as those pioneeredin other countries for health and education, arelikely to point in the same direction.

Part . A Comprehensive Review of ’sEconomic Accounts

For the Mid-Decade Strategic Review, un-dertook to replicate the scope and approaches ofthe “blue ribbon” panels and other comprehensiveevaluations that have helped shape the economicaccounts in the past. As part of this process, prepared background papers that looked atthe accounts from five perspectives: Past experiencein adapting the economic accounts to changes in

February •

the economy, outside evaluations of the accounts,changes in source data available to (resulting,for example, from deregulation) and in estimatingmethods, the newly revised international guide-lines for an integrated system of national accountsand for the balance of payments, and the sizeand sources of revisions in key components of theaccounts as suggesting an agenda for future work.

identified three priority issues for maintain-ing and improving the accounts:

• Change in the nature of output and the organ-ization of production: The need for new andimproved output measures,

• Investment: The need for better measures ofinvestment, saving, and wealth, and

• Internationalization: The need for measuresto fill gaps in the coverage in internationaltransactions.

Earlier Reviews of th

In the ’s, there were two major reviews of theaccounts. The first, National Economic Accounts of theUnited States: Review, Appraisal, and Prospects, wasprepared in by the National Accounts Review Com-mittee of the National Bureau of Economic Researchat the request of the Bureau of the Budget (predeces-sor of the Office of Management and Budget). TheCommittee was chaired by Raymond Goldsmith, and themembers included Richard Easterlin, Joseph Pechman,and Richard Ruggles. The report was presented at a hear-ing of the Subcommittee on Economic Statistics of theJoint Economic Committee. The second review, Critiqueof the United States Income and Product Accounts, from, was the result of a symposium on the accounts heldby the Conference on Research in Income and Wealth,which functions as a research forum for work on eco-nomic measurement. These reviews dealt with emergingissues of the time, many of which related to the expan-sion in the complexity and scope of the accounts that wasnecessary to accurately portray the U.S. economy. Theyalso dealt with conceptual issues, such as the treatmentof capital gains and the coverage of nonmarket produc-tion and consumption, and they discussed the need forbetter integration of the income and product, flow offunds, and other components of the existing accounts.

In , published The Economic Accounts of theUnited States: Retrospect and Prospect. This volume, onthe occasion of the th anniversary of the S

C B, was a series of papers contributedby some of the country’s most prominent economists,including past and future Federal Reserve Board chair-persons (Arthur Burns and Alan Greenspan), Nobellaureates (Wassily Leontief, Simon Kuznets, LawrenceKlein, and Paul Samuelson), Council of Economic Advis-ers chairpersons (Arthur Okun and Raymond Saulnier),and American Economic Association presidents (Robert

A comprehensive review: Scope and approaches

In past decades, a series of reviews have evalu-ated the performance of the economic accountsin terms of their ability to provide a relevantpicture of the changing economy and to adaptto changes in source data, estimating methods,and economic accounting. These reviews alsoevaluated the accuracy of the accounts throughsome combination of revision and other statisti-cal studies. (See the box, “Earlier Reviews of theEconomic Accounts,” below.) The reviews pro-vided guidance, in most cases confirming ’sown directions, for expanding and updating theaccounts. The last review was in the early ’s.

Given the pace of change in the economy, acomprehensive review seemed overdue. Earlierexperience suggests that such a review shouldcover methodological and statistical issues as wellas concepts and accounting structure. Further,

e Economic Accounts

Eisner, Robert Gordon, and Charles Kindleberger).

catalogued and prioritized the suggestions from thesepapers, and ’s Director at that time, George Jaszi,responded.

In , the Gross National Product Data ImprovementProject Report was released. This report was preparedby the Advisory Committee on Gross National ProductData Improvement under the auspices of the Office ofManagement and Budget. The committee was chaired byDaniel Creamer, and the members were Rosanne Cole,Edward Denison, Raymond Goldsmith, Alan Greenspan,and John Kendrick. The report, referred to as theCreamer Report, was undertaken as a result of concernsover relatively large revisions in the accounts in theearly ’s and focused on needed improvements inthe source data, rather than on needed extensions andconceptual modifications.

In , the Conference on Research in Income andWealth addressed several aspects of the national incomeand product accounts pertaining to their role as a sys-tem of information about the behavior of the economy.Topics included the concepts and structure of the ac-counts, the issues involved in deflation and the treatmentof quality change in price indexes, and source data. Thelast topic included an evaluation of major parts of theCreamer Report.

In , the General Accounting Office published TheBureau of Economic Analysis Should Lead Efforts to Im-prove Estimates. This study was intended to evaluaterevisions to estimates and to assign different pri-orities to the Creamer Report’s recommendations,made to Federal agencies, for improving the accounts.It, too, focused more on statistical than on conceptualissues and—as the title indicates—urged to take amore proactive role in obtaining the source data neededto improve the accounts.

• February

Economi

Structural change:Constant-dollar gross pr

Input-output accounts ...

Gross state product .....

Other structural develop

Price change ...................

Internationalization:Trade in services .........

International investment

International capital tran

Other international deve

Other developments .......

1. Years in parentheses are

it should look at the accounts from at least fiveperspectives:

• The changes in the economy in the ’s and’s and ’s responses to these changes,to serve both as an agenda for future workand as a guide to success in the future.

• Recommendations by outside experts.• Updated international guidelines in eco-

nomic accounting, mainly the System ofNational Accounts and the Balance ofPayments Manual.

• Changes in source data available to (reflecting changes such as those due to

Table 1.—Selected Changes in the Economy and Chang

c development Issue

oduct originating ......................... Estimates were not capturing changes; neepicture of growth in services and importsof information processing equipment.

..................................................... Incomplete picture of industries; needed modetail on industry interaction.

Incomplete picture of industries; needed moinformation on changing industry structur

..................................................... Incomplete picture of regional activity; needState contributions to GDP to understandeconomy’s regional structure.

ments .......................................... Changing nature of Federal deposit insuranreflect the exposure due to savings and

Changing nature of health care programs; nconsistent picture.

..................................................... Substitution bias of fixed-weighted index ....

Changing relative prices and/or unique prod

..................................................... Gap in coverage; as trade (especially in finservices) increased, gap became larger.

Incomplete monthly picture of trade ............

.................................................... Position: Valuation at historical cost and theunderstated and inconsistent.

Direct investment in the United States: Neeestablishment detail for analysis of specif

sactions ...................................... Gap in coverage; needed to capture new chnew financial instruments.

lopments ..................................... Increased interest in multinational firms; needetail on ownership.

Other .............................................................

..................................................... Increased concern about the impact of econon the use of natural resources and enviquality.

Growth of the underground economy .........

the years in which the change was introduced.

deregulation and those in data collec-tion technology) and changes in estimatingmethods.

• Revisions in key components of andthe other accounts, examining the size andsources of revisions as an agenda for futurework.

Background papers

As the first step in the Mid-Decade Review, prepared a set of background papers that coverthe five perspectives just described. Papers II–VI

es in the Economic Accounts

Action 1

ded better and the role

Revised the gross product by industry series (1991)

re timely Improved the benchmark input-output accounts (1994)

re frequente.

Published annual input-output tables (1987)

ed detail on changes in

Introduced gross state product by industry (1988)

ce; needed toloan crisis.

Developed a new classification of federal depositinsurance (1991)

eeded Reclassified medicaid outlays (to conform its treatmentwith medicare) (1985)

....................... Published new alternative measures of real GDP changethat reflect changing relative prices (1992)

ucts .............. Introduced new index for deflation of computers basedon hedonic techniques (1985)

Introduced a new price index for deflation of multifamilystructures based on hedonic techniques (1991)

Used BLS import price indexes for deflation of importedproducers’ equipment (1988)

Improved the deflation of Federal defense purchases(1980)

ancial Conducted new surveys of 30 services (1989)

Improved the foreign travel survey (1989)Developed new benchmark survey of financial

services(1994)....................... Presented new monthly estimates of international

services (1994)refore Revalued direct investment and used market values for

U.S. gold reserves (1991)dedic industries.

Linked BEA database of foreign-owned companies withCensus database of U.S. establishments (1992)

annels and Expanded use of partner-country data (1994)Supported improvements in Treasury surveys of portfolio

investmentded more Developed supplemental balance-of-payments accounting

frameworks and estimates (1994)....................... Improved end-use classifications and deflation for trade

(1988)

omic growthronmental

Developed integrated economic and environmentalsatellite accounts (1994)

....................... Improved the adjustments for the underground economy(1984)

February •

are summarized in this section. (Paper I is anintroduction and executive summary to the set.)

Economic change and the economic accounts.—Paper II considers ’s experience of the last years in adapting the accounts to the chang-ing economy. The bulk of the paper is aboutthe period beginning in the late ’s. A ta-ble in the paper that highlighted many of thesechanges is summarized in table in this arti-cle. The first group of changes in table dealswith the structure of the economy, ranging frombroad changes such as the growth of imports andservices to changes in regional patterns and tochanges in specific government programs. Thesecond group deals with changes in price struc-ture. The third group deals with the internationalaspects of the economy, focusing on trade inservices, capital flows, and international invest-ment. The actions listed as ’s responses to thechanges include some that deal with methodol-ogy, some that modify and extend surveys, andtwo that deal with new surveys. In some cases,even though an action is indicated, the issue hasnot been completely resolved.

Recommendations from outside experts.—PaperIII summarizes a number of external reviewsof the accounts during the last two decades.This sample of wide-ranging reviews and of that are more subject-specific includes thefollowing: Several administration-sponsored re-views, from the comprehensive Gross NationalProduct Data Improvement Report of the ’sthrough the Economic Statistics (“Boskin”) Ini-tiative of the early ’s; several volumes by theConference on Research in Income and Wealth;two studies by the International Monetary Fund;two presidential addresses at the American Eco-nomic Association; papers by other individualresearchers; and a continuing effort by a tradegroup.

The recommendations of outside experts havea “living” nature. As the economy changes, newconcerns arise and generate new recommenda-tions so that the accounts are under evolvingpressures. As well, users are persistent in their ef-forts to influence economic accounting; concernsare expressed repeatedly if left unaddressed. Forexample, the adequacy of services estimates hasbeen questioned by many groups over the lasttwo decades. Similarly, the coverage of govern-ment purchases, particularly of State and localgovernment purchases, has been a concern formany years.

National accounts.—The reviews provided a sub-stantial number of recommendations about con-cepts (such as the scope of investment), sourcedata, and estimating methods for the majorcomponents of . (See the next section forrecommendations about net exports.)

• Within consumer spending, by far thelargest number of recommendations focus onimprovements in services.

• Among the components of investment, thelargest number of recommendations dealwith improvements in inventories, followedby nonresidential fixed investment (mainlyconstruction).

• Among the components of government pur-chases, by far the largest number of rec-ommendations deal with State and localgovernment purchases.

A number of comments call directly for betterconstant-dollar estimates or do so indirectly bycalling for better price indexes to use in deflation.Many of these recommendations address the de-mands that a rapidly changing economy puts onprice measures. Rapid changes in the nature ofgoods and services require that price indexes beable to separate price change from quality change,and shifts in the pattern of demand require thatthe product samples underlying price indexes bekept up to date.

Other recommendations classified as “nationalaccounts recommendations” include those oninput-output accounts, by industry, andsaving and wealth.

International accounts.—The largest number ofrecommendations about the balance of paymentsare for improvements to the capital account, fol-lowed by the current account and by prices usedfor deflation of exports and imports. Withinthe capital account, the largest number of rec-ommendations are for improved estimates ofportfolio investment, followed by direct invest-ment. Within the current account, most of therecommendations are for improved estimates oftrade in services.

Other recommendations.—A number of the rec-ommendations address generic issues. Sev-eral recommendations, especially those from the’s, focus on the need for better documenta-tion of ’s methodology, and several suggestways, such as an advisory committee, to formal-ize contact with users of the accounts. A numberof recommendations focus on increased funding,

• February

both for and for source data agencies. Oth-ers deal with improving coordination and tappingpossible efficiencies across the Federal statisti-cal agencies; these include recommendations fordata sharing and revised industry classificationsystems.

Changes in data sources and estimating methodsas they affect the economic accounts.—Paper IVfocuses on the developments that have caused, orthat call for, changes in source data and methods.It also notes ’s assessment of the additionalsteps that should be taken to deal with thesedevelopments.

The supply of administrative data used inthe accounts has been adversely affected by fac-tors that are not primarily statistical in origin.Perhaps the most important of these is thederegulation of the transportation, telecommuni-cations, and finance industries in the late ’sand early ’s, which has resulted in the dis-appearance of some of the data once used tomonitor and regulate them. is currentlyworking with the Census Bureau to expand thecoverage of the services annual survey to allservices, including transportation and commu-nications, to recover the lost data. In addition,changes in tax laws have caused discontinuitiesin the tax-based data used in the accounts. Inthe ’s, these included changes in the pass-through of income to shareholders, limitationson business entertainment expenses, capitaliza-tion of interest, and the uniform capitalizationof inventories. More generally, reductions ingovernment budgets have caused nonstatisticalagencies to eliminate, or place a lower priority on,those statistics not critical to their central mis-sion. Continued reductions in the availability oftax and other administrative data raise questionsabout reliance on these data sources. It is clearthat there are large efficiencies in the dual use ofsuch data, but certain generic questions—for ex-ample, about the funding of such data—need tobe resolved.

Many of the changes in the economy noted inpaper II appear in paper IV as reasons to changedata collection, data processing, and estimatingmethodology. For example, rapid changes inthe nature of retail outlets call for more fre-quent updating of sample frames, and sharpchanges in relative prices call for less reliance onfixed-weighted measures. For the nationalestimates, progress includes the introduction ofalternative output and price indexes that usemore frequently updated weights, use of hedonicprice indexes to capture increased quality, and

improved adjustments to tax-based data to ac-count for the underground economy. For theinternational estimates, progress in adapting tothe effects of changes in the economy include theuse of counterparty data to fill gaps in coverageand the use of data sharing. ’s recent link ofits data on foreign-owned companies with Cen-sus Bureau and Bureau of Labor Statistics data onU.S. establishments resulted in a significant in-crease in the amount of information available onforeign-owned companies—without an increasein respondent burden. However, further workis required in improving real output and priceindexes, in updating sample frames, in measur-ing new services, in revamping data collectionsystems for international capital flows, and inupdating industry classification systems.

The adverse effects on the accounts of thesechanges in data sources have been partly offsetby changes in data collection technology and sta-tistical methods. Electronic data collection andtransfer methods are beginning to improve thecollection and editing of data. One of the mostimportant improvements that can make toaddress the problems discussed in this paper isthe reengineering of its information technologysystem.

The newly revised international guidelines foreconomic accounting.—International standards,or guidelines, in statistics are designed to guidecountry statistical offices in the development oftheir own statistics and, in the interests of inter-national comparability, to serve as a frameworkin which countries report their statistics to in-ternational organizations. Globalization of tradeand financial markets has made internationalpolicy coordination, supported by consistencyin the underlying statistics that guide policy,increasingly important.

In the last decade, the United States took a leadrole in the development of newly revised interna-tional economic accounting guidelines found inthe System of National Accounts ( )and the Balance of Payments Manual (). Therevised guidelines reflect efforts to bring themup to date with changes in economies aroundthe world—including many of the same changesaffecting the U.S. economy described in otherbackground papers—and with developments ineconomic accounting, such as those that reflectlonger experience in compiling balance sheets.

As discussed in paper V, modernizing and ex-tending the existing U.S. accounts to make themcompatible with the and the willrequire work in several areas. Compatibility with

February •

the would mean developing accountsfor sectors as well as for the entire economy;improving the statistical picture of government’srole in the economy; improving the treatment offinancial and insurance services; and better inte-grating the ’s, financial accounts, and balancesheets. Further, the encourages the de-velopment of satellite accounts to increase theflexibility of the accounting system.

Compatibility with the revised wouldmean changes in the treatment of goods, par-ticularly intellectual property, investment goods,and goods acquired in ports. Changes would alsobe necessary to the treatment of services, suchas affiliated services, rental services, and financialservices, to name a few. Other changes would berequired in the treatment of income, unilateraltransfers, the coverage of the capital accounts,and the coverage of the international investmentposition.

Revisions in the economic accounts: Implica-tions for improvements.—Paper VI focuses onrevisions and other statistical studies of nationalestimates—specifically, the estimates. Manyof the implications of these studies hold, how-ever, for the international and regional estimatesas well. First, a large component of nationalestimates—net exports—comes from the balanceof payments accounts. Second, regional esti-mates are, in many cases, extensions of nationalestimates and thus exhibit similar characteristics.

Studies of revisions are often used as a basis forrecommendations for improvements, althoughrevisions can be quite misleading for such pur-poses. Revisions in series do not necessarilyreflect errors, nor does a lack of revisions reflectthe absence of errors. Revisions due to the re-placement of initial estimates with those basedon more complete source data can reflect errorsassociated with projections or with the smallsample frame or other features of the source dataused for the early estimates. However, revisionscan also reflect changes in accounting structure,such as definitional changes to provide moreup-to-date or otherwise more relevant measures,that should not be regarded as the correction oferrors.

Revisions in .— has made a number ofstudies of revisions in the ’s. Many of thesestudies featured estimates of dispersion and biasof estimates of (or of ) and its majorcomponents. (Dispersion is the average, or mean,of the absolute values of the revisions, typicallycalculated using percent change from quarter to

quarter at annual rates. Bias is the average, ormean, of the values of the revisions.)

The studies indicate that the current estimatesof —advance, preliminary, and final—aregenerally able to tell us whether the economyis expanding or contracting, whether growth isaccelerating or decelerating, and whether thegrowth rate is high or low relative to trend. How-ever, their ability to do so is least when economicgrowth is hovering near zero and—although theevidence is less clear on this point—at turningpoints in the economy. The following paragraphssummarize results of the studies for the period–:

• The current estimates correctly indicated thedirection of change in real between and percent of the time. If quar-ters in which growth is percent or lessare excluded, the percentage correct rises tobetween and percent.

• The current estimates correctly indicatedwhether real was accelerating or deceler-ating between and percent of the time.If quarters in which growth was percent orless are excluded, the percentage correct risesto between and percent.

• The estimates also correctly indicated highgrowth (above percent) versus low growth(below percent) between and percentof the time.

• Dispersion in the current estimates of real was . percentage points, or between and percent of average real growth of. percent during this period. Bias was smalland averaged between and . percentagepoint, or between and percent of averagereal growth.

The revisions of for the United States com-pare quite favorably with those of other countries.Using estimates from the mid-’s to the mid-’s, one study found that revisions in quarterly growth in Canada, Japan, Australia, WestGermany, and the United Kingdom were all dis-tinctly larger than those in the United States, eventhough U.S. estimates are released more quicklythan their counterparts. Later, but less compre-hensive, studies suggest some convergence, butthat the U.S. estimates are still among the mostreliable and timely.

Revisions in components.—The revisionstudies also provided information about the sizeand sources of revisions in the components.Among the relevant points about the revisions for

• February

– for the product components and selectedincome components are the following:

• Personal consumption expenditures ():Dispersion in the quarterly estimates isroughly equal to that in ; revisions, meas-ured in dollars, are about three-quarters ofthe size of the revisions in . Among the components, the largest revisions in per-centage points are to durable goods; thelargest dollar revisions are to services. Forservices, revisions to the initial current es-timates are mainly due to the replacementof judgmental projections with partial sourcedata. Revisions from the final current esti-mates to the first annual revision were largelydue to the incorporation of a large and di-verse volume of annual source data froma variety of government agencies and tradesources, quarterly survey data, and monthlydata on residential energy use.

• Investment: The largest single source of re-visions between the successive current quar-terly estimates of growth is from inven-tory investment. The first revisions occurbetween the advance and preliminary esti-mates of inventories, as judgmental estimatesfor the third month of the quarter are re-placed with preliminary estimates based onsource data from surveys of manufacturingand trade inventories and as the source dataare revised. Revisions are also large betweenthe final current quarterly estimates of in-ventories and the latest estimates. Theserevisions reflect (in addition to a concep-tual change for –) the replacement ofmonthly survey data with annual survey dataand revisions of seasonal adjustments. Theyalso reflect the replacement of projec-tions for nonfarm inventories other thanmanufacturing, merchant wholesale trade,and retail trade.Most of the revisions to fixed investment aredue to a combination of past improvementsin source data and methods that affected re-visions for – and of revisions in sourcedata. Substantial revisions to the currentestimates of nonresidential investment aredue to revisions in the monthly data onmanufacturers’ shipments and in the dataon the value of construction put in placethat are mainly related to incomplete re-porting. Large annual revisions reflect theincorporation of source data that providemore complete and up-to-date coverage.

• Exports and imports: Despite the relativelysmall size of net exports, large revisions toexports and imports cause the dispersion intheir current estimates to be large relative todispersion in the current estimates of .In dollars, exports and imports are secondonly to inventories as a source of revisions;in percentage points, their dispersions are – times the size of the dispersion in .The sources of these revisions include thereplacement of judgmental projections forthe last month of each quarter with sourcedata and improvements in the estimates ofinternational services.

• Government purchases: The major source ofrevisions (aside from a conceptual changethat had an offset in inventories) is the re-placement of judgmental projections withsource data. For the Federal Government,detailed data from the Monthly TreasuryStatement for the third month of each quar-ter and monthly data on civilian wagesand salaries replace judgmental projections.For State and local governments, data ongovernment purchases (except compensationand structures) become available and replacequarterly extrapolations of as many as quarters.

• Compensation of employees: The largest dol-lar revisions in national income are to com-pensation of employees. A large share ofthe revisions to the final current estimatesof wages and salaries occurs as those esti-mates are replaced with estimates based ondata providing fuller coverage.

In terms of dispersion and, in certain years,in terms of dollars, revisions to other la-bor income are larger than those to wagesand salaries. These revisions mainly reflectthe paucity of current quarterly data and thelong lag before annual data are available.

• Proprietors’ income: The dispersion for thiscomponent, especially for farm proprietors’income, is the largest of all the major com-ponents of national income. The volatility offarm output and inventories results in largerevisions in source data and in seasonal ad-justment factors. For nonfarm proprietors’income, current quarterly and first-annual-revision estimates are based on indicators ofindustry activity and judgmental trends. Inthe second annual revision, tabulations ofnoncorporate business tax returns becomeavailable.

February •

• Corporate profits: In terms of dispersion, re-visions to corporate profits are second insize only to proprietors’ income among thecomponents of national income. Revisionsreflect more complete coverage of the sourcedata on domestic profits and more completereporting on ’s surveys of multinationalcompanies.

Other studies.—A number of studies of revisionsin the ’s have been prepared outside . Theprincipal implication of these studies is that someimprovements could be made in the early esti-mates. Despite several limitations, these studieshave provided with tools to further evaluateits revisions.

Although the evidence is mixed, revisions to appear to account for a substantial shareof the errors in short-term econometric forecastsused for business and macroeconomic policypurposes.

The priority areas of concern: Threecross-cutting issues

has identified three priority issues that cutacross the subjects of the background papers. Forexample, certain changes in the nature of theeconomy were referred to in all papers becausethey have led to gaps in coverage and to prob-lems with existing source data and methods, aswell as to significant revisions. These three is-sues are also likely to be the focus of currentpolicy interest and the subject of outside experts’recommendations.

Change in the nature of output and the organ-ization of production: The need for new andimproved output measures.—Measuring outputis fundamental to economic accounting, but anincreasing share of output is in sectors in whichoutput is difficult to define and measure. Howdoes one define, for example, the output of alegal firm or of an economic consulting firm?If hourly billing rates for such firms rise overtime, how does one determine whether there areoffsetting improvements to the value of the serv-ices performed for the firms’ clients? There areno easy answers to such questions, and answersare necessary in order to separate current-dollarmeasures into price and quantity, or real output,terms.

Similar problems also arise in some portions—the more dynamic portions—of the goods-producing sectors. For example, new productstypically enter the market at very high prices;

subsequently, prices fall very rapidly, with result-ing expansions in the size of the market. Thepricing histories of these new products must in-clude the periods of their sharpest price declines,even though the initial market shares of thesenew products are small, because otherwise overallprice measures may be biased. In times of rapidintroduction of new products, however, it maybe very difficult to keep product samples suffi-ciently up to date to track the price declines ofnew products.

More broadly, changes in the composition ofoutput have caused problems as newly emerg-ing services and goods have opened gaps in thecoverage of existing surveys, methods, and clas-sification systems. For example, sample framesfor retail trade become outdated as distributionchannels shift, as when “wholesale” outlets beganto account for an increasing share of direct salesto consumers. New and rapidly growing prod-ucts, such as computer software and exports ofinternational financial services may—for any ofseveral reasons—escape the statistical nets, whileproducts on their way out, such as records,are still fully covered.

Rapid changes in output and prices also causeproblems for statistical methods. Notable amongthese is the use of fixed weights for output andprice indexes, a method introduced when updat-ing such indexes every years was thought to besufficient.

Finally, changes in the structure and organ-ization of the economy threaten the relevanceof some estimates in the economic accounts.For example, accounting for the changing roleof government, for the interaction between theeconomy and the environment, and for researchand development is limited.

Investment: The need for better measures of in-vestment, saving, and wealth.—Changes in theeconomy have also increased the need for broaderand more reliable measures of investment, sav-ing, and capital stocks. For example, changes intechnology call for broader coverage of products,such as computer software, whose investment-like properties are increasingly apparent. At thesame time, integration in world financial marketsand the effect of changes in wealth on con-sumer spending, investment, and internationalcapital flows have increased the importance ofdeveloping integrated accounts for real and finan-cial transactions and stocks for macroeconomicanalysis.

Longstanding problems affect the measures ofdepreciation, capital stocks, and inventories. Al-

• February

though ’s straight-line depreciation and cap-ital stocks at replacement cost were a significantimprovement over tax-based estimates at histori-cal cost when they were introduced in the ’s,an overhaul of these estimates is overdue.

Internationalization: The need for measures tofill gaps in the coverage of international trans-actions.—One of the most obvious aspects ofeconomic change affecting the accounts is theinternational integration of markets. The inte-gration of world capital markets has significantlyincreased monetary and regulatory authorities’interest in accurate and complete informationon capital flows. At the same time, the inte-gration of markets for goods and services hasincreased business and trade officials’ interest innewly emerging goods and services.

Unfortunately, while increased integration hasincreased the importance of such data, it has alsoincreased the difficulty of measuring such trans-actions. Gaps have developed in the coverage ofnewly emerging categories of international tradein services and goods, in the coverage of deriva-tives and other new financial instruments, andin the coverage of security and other portfoliotransactions that bypass U.S. brokers, banks, andother financial institutions. Changes in the struc-ture of international markets have also resulted inoutdated and incomplete source data for existingmeasures of goods and services.

Part . ’s Strategic Plan for Maintainingand Improving the Economic Accounts

’s draft strategic plan is framed around thethree priority issues identified in the backgroundpapers. Within each issue, first identified themost important measurement problems in terms ofthe accuracy, reliability, and relevance of the ac-counts. then screened the recommendationsin the consolidated “menu” that addressed theseproblems to identify the most cost-effective actionsin terms of both direct cost and respondent bur-den. The actions proposes include changes inmethodology and accounting structure, source datamodifications and extensions, and, in a few cases,new source data. Milestones in the implementationof the actions are proposed through .

’s draft plan for maintaining and improv-ing the economic accounts is framed around the

three issues identified in the background papers:The need for new and improved output measures;the need for better measures of investment, sav-ing, and wealth; and the need for measures to fillgaps in the coverage of international transactions.Within each issue, assessed the practicalimportance and relative size of the various meas-urement problems and then identified the onesthat were most important in terms of the accu-racy, reliability, and relevance of the accounts. then focused on the recommendations in theconsolidated “menu” that addressed these prob-lems. These recommendations were screened toidentify the most cost-effective actions in termsof both direct cost and respondent burden. Theresults of this process are summarized in table .

Table shows, in column two, some quan-titative indicators of the importance of variousmeasurement problems in the accounts. All ofthese indicators should be used with caution. Forgaps in coverage, the indicators are no betterthan educated guesses; if a precise estimate wereknown, it would be possible to simply plug it inand fill the gap. For revisions, the quantitativeindicators are based on past experience, and thatexperience may not hold for the future. In somecases, the indicator may only be suggestive of im-portance; in others, no quantitative indicator ofthe importance of the problem can be given.

In screening the recommendations to iden-tify the most cost-effective actions, the choicewas clear where the measurement problem isamenable to resolution by improvements inmethodology and accounting structure using ex-isting, but heretofore unexploited, source data.(See the box “A Three-Way Grouping of ActionsTo Improve the Accounts,” on page .) How-ever, many of the measurement problems existlargely because there are no reliable source dataupon which to base estimates. In these cases, theleast burdensome and most efficient action hasbeen suggested. Wherever possible, the modifi-cation or extension of an existing survey has beensuggested; a new survey has been suggested onlyas a last resort. In either case, would workwith the Census Bureau, the Bureau of LaborStatistics, other source data agencies, data users,and the respondent community to obtain thenecessary information with the lowest possiblerespondent burden.

The need for new and improved outputmeasures

Difficult-to-measure components of real .—Difficulties in measuring and defining real output

February •

and prices may have a significant effect on keycomponents of , as well as on and itsgrowth rate. Because price measures are usedto deflate current-dollar expenditures to obtainreal output measures, any errors in existing pricemeasures will impart an equivalent error—but ofopposite sign—to the growth rate of real andof productivity. For example, if the net effect ofprice measurement difficulties is to overstate the

Table 2.—

Issues, problems, and uses affectedQuantitative indicators (e.g., p

of gap, size of revision, sizenent treated differently or

Need for n

Difficulties in measuring and defining certaincomponents of real GDP.Uses affected: Analysis of economic

growth, especially of economy’s long-term, noninflationary growth potential;macroeconomic policy; forecasting;business, budget, and investmentplanning.

Possible understatement of grespecially in fixed investmefor understatement in real Ggrowth 1.

No quantitative indicator of theof defining output.

Revisions to key components of GDP andnational income.Uses affected: Macroeconomic policy;

forecasting; business, budget, andinvestment planning.

1.4–9.4 percentage point (±) rquarterly changes (SAAR) fcomponents of current-dollaChange in business invento

billion.Trade in goods and serviceGovernment purchases, $8Consumer expenditures for

$6 billion.Consumer expenditures for

$5 billion.1.4–8.5 percentage point (±)

quarterly changes (SAAR) components of national inc

Overstatement of real GDP growth in recentyears (and understatement in earlieryears).Uses affected: Analysis of economic

growth, especially of current growthrelative to long-term, noninflationarygrowth; macroeconomic policy;forecasting; business, budget, andinvestment planning.

0–1.2 percentage point overstquarterly rates of change in(average 1991:I–1994:III, 0.percentage point) 3.

Outdated and inconsistent view of thestructure and organization of production.Uses affected: Federal and State and

local tax analysis and budget planning;business location and marketingstudies; regional analysis; industrialorganization studies; and cross-borderanalysis of trade effects.

For industry classifications,inconsistencies across U.S.and incompatibilities amongAmerican countries, with spattention needed for new aindustries, service industrietech industries (for a discusquantitative indicators, see

See footnotes at end of table.

rate of inflation, then the effect is to understatethe growth rate of real .

Determining the magnitude of the net error,if any, imparted to real growth is difficult.Some of the broader-based studies are now datedin that they do not cover the ’s, when signif-icant methodological improvements were madeby the Bureau of Labor Statistics in the Pro-ducer Price Index and the Consumer Price Index,

Issues and Proposed Actions

otential size of compo- added)

Statistical source(s) of the problem Proposed actions

ew and improved output measures

owth,nt; potentialDP

Difficulties in measuring quality changes,especially in investment goods.

Methodology and structure: Extension ofquality adjustment of prices used inreal GDP, including hedonic work ongoods amenable to such measurement:High-tech goods and nonresidentialstructures.

difficulties Difficulties in defining output ...................... Methodology and structure: Furtherconceptual work on more difficult-to-measure goods and services.

evisions toor keyr GDP: 2

ries, $13

s, $9 billion. billion. services,

goods, $4–

revisions tofor keyome.

Inability of existing source data used inthe quarterly estimates to capturechange in the economy.

Data modification and extension:More frequent updating of sample

frames for existing surveys, includingtrade and manufacturing.

Data extension: More frequent surveysfor certain growing sectors such asinternational trade in services,medical care, and State and localgovernment purchases.

Data extension: Extension of existingsurveys such as those forinventories, services, and employeecompensation (including bonuspayments) to fill gaps in coverage.

Difficulties in seasonal adjustment ............ Methodology and structure: Improvementsin seasonal adjustment for volatilecomponents such as inventories andtrade in goods and services.

Errors in projections for missing sourcedata.

Methodology and structure: Improvementsin projections for components such asinventories, trade in goods andservices, and bonus payments.

atement of real GDP4

Substitution bias, specifically the use offixed weights (1987) inappropriate forthe current period.

Methodology and structure: Introduction ofmore current weights for real GDP forcurrent estimates and more appropriateweights for historical estimates.

industries Northecialnd emergings, and high-sion oftext).

Outdated and inconsistent industryclassification system, source data, andindustry accounts.

Methodology and structure: Develop anew industrial classification system.

Data modification and extension:Implement a new industrialclassification system, starting with arestructuring of surveys.

Methodology and structure: Update andbetter integrate the input-output,industry, gross state product, and GDPestimates within the context ofmodernizing the accounts along thelines of the new internationalguidelines.

• February

Issues, problems, and

Extend the concept and upmeasurement of investmcapital stock.Uses affected: Analysis

economic growth, proto public and private and expenditures poli

Need for better integration measurement of investmwealth/capital stocks.Uses affected: Analysis

economic growth, proto public and private saving; tax and expen

Gaps in the coverage of inin certain goods and serand capital.Uses affected: Analysis

monetary, and regulaforecasting; business planning.

1. For a discussion of quant2. Based on BEA revision st3. Based on BEA alternative4. Based on historical differe

flow-of-funds estimates; most o

which provide most of the detailed componentsfor deflation of real . Nevertheless, there issubstantial evidence to suggest that price meas-urement difficulties may be severe for sectors inwhich rapid technological change has led to asubstantial drop in the quality-adjusted price ofoutput. For example, ’s hedonic computerequipment price indexes, which were introducedinto the accounts in , have declined at an an-nual rate of – percent for –. Studies onsome other products—including semiconductorsand medical scanners—also suggest large rates of

Table 2.—Issues and Proposed Action

uses affectedQuantitative indicators (e.g., potential size

of gap, size of revision, size of compo-nent treated differently or added)

Statistica

Need for better measures of investment, sav

date theent and wealth/

of sources ofductivity, returnsinvestment; taxcies.

Treating government spending onstructures and equipment andgovernment and business spending oncomputer software as fixed investmentwould raise investment andreproducible capital stock in nationalwealth by about 20 percent.Government capital, $2,863 billion.Computer software, $20–$40 billion.

Treating other candidates as investmentin satellite accounts would raise wealth/capital stock, for example:.R&D capital, $1,050–$1,380 billion.Natural resources, $950–$1,600 billion.

Exclusion of private exthe nationcapacity.

No quantitative indicator of the need toupdate measurement.

Use of straig

andent, saving, and

of sources ofductivity, returnsinvestment, andditure policies.

3–9 percentage point differences—conceptual and statistical—betweenNIPA and flow of funds measures ofpersonal saving rates 4.

Treating government ‘‘investment’’ in GDPconsistently with internationalguidelines would eliminate more thanhalf of the apparent 5.8 percentagepoint shortfall in U.S. versus Europeaninvestment rates 5. For 1993,investment as a percent of GDP:NIPA rate is 13.7 percent; adding

government ‘‘investment’’ raises therate to 16.7 percent.

Average rate for Europe is 19.5percent.

Lack of comfinancial a

Need to fill gaps in the coverage of internatio

ternational tradevices, income,

of trade,tory policy;and investment

Gaps in key components:Trade in goods and services, as much

as $10–$20 billion 6.Capital flows, as much as $100, and

stocks as much as $200 billion 6.

Inability of exsystems totypes of ginstrument

itative indicators, see text.udies; see text for details. output and price indexes; see text for details.nce between BEA’s NIPA measures and the Federal Reserve Board’sf the difference between the two series are conceptual, with statistical

differences rang5. Calculated

operation and D6. Based on

price decline. Other studies, including studies ofprescription drugs, have pointed to measurementproblems that arise when the turnover of newproducts is quite rapid; in the case of prescriptiondrugs, these studies have led to methodologicalimprovements in the Producer Price Index.

On the other hand, quality-adjusted measureshave been criticized for overstating the increase inoutput. For example, prominent macroeconomicmodeling firms have reservations about the he-donic indexes for computer equipment and do

s—Continued

l source(s) of the problem Proposed actions

ings, and wealth

certain types of public andpenditures that contribute to’s wealth and productive

Methodology and structure: Expand theaccounting for investment and wealth/capital stock by (1) inclusion ofgovernment spending on structures andequipment and government andbusiness spending on computersoftware and other intangibles ininvestment in the national accountsand (2) inclusion of research anddevelopment and natural resources insatellite accounts, in the context ofmodernizing the accounts along thelines of the new internationalguidelines.

ht-line depreciation ................ Methodology and structure: Use ofconceptually and empirically baseddepreciation patterns and valuationmethods

plete integration betweennd real accounts.

Methodology and structure: Betterintegration of real and financialaccounts in the context of modernizingthe accounts along the lines of thenew international guidelines.

nal transactions

isting data collection capture new markets and

oods, services, and financials and intermediaries.

Data modification and extension:Extension of existing surveys to covernew products, services, and markets.

Methodology and structure: Extension ofdata exchanges with other countriesand central banks.

New data: Development of new surveyssuch as for financial services andportfolio investment.

ing between 0 and 2.9 percentage points over the last 10 years. from Quarterly National Accounts, compiled by the Organisation for Economic Co-evelopment. ‘‘Europe’’ includes the 13 countries for which data were published.

indicator series and past revisions for similar components.

February •

not use them in certain aspects of their analysesand forecasts.

The potential importance of price measure-ment problems in measuring investment may beillustrated as follows: For every -percentage-point overstatement (understatement) in the rateof quality-adjusted price change for “high-tech”equipment other than computers, the annualgrowth rate of real nonresidential fixed invest-ment will be lowered (raised) by about . per-centage point. For example, if quality-adjustedprices for this “high-tech” equipment over thedecade – fell at an annual rate of per-cent rather than increased at the .-percent ratemeasured by deflators (that is, a differenceof about percentage points), the growth rate ofreal nonresidential fixed investment would havebeen understated by . percentage points and thegrowth rate of real by . percentage point.

Of particular interest is the potential for mis-statement of real growth. In recent years,concerns have been raised that if problems inmeasuring output have become worse over thelast several decades and have resulted in aprogressively larger understatement in real growth, measurement problems may be partly re-sponsible for the reduction in the growth trendfor real and productivity since the early’s. Understatement of the economy’s growthtrend can be quite important because of its role in

A Three-Way Grouping of Ac

Methodology and structure: These actions include ad-vances in economic accounting structures and conceptsand changes in estimating methods that improve the ac-curacy, reliability, and relevance of the accounts. Theactions reflect empirical and conceptual research—eitherdirectly or indirectly—through reference to interna-tional guidelines in economic accounting. They buildon existing—albeit heretofore un- or underexploited—source data. Such improvements have several practicaladvantages: They often can be undertaken within

and require few additional resources in comparison withthe resources required to pursue a new survey; the leadand start-up times are relatively short; and perhaps mostimportant, because they build on existing source data,they do not increase respondent burden.

Source data modification and extension: These actionsfeature changes that can be made to existing sourcesof data: Adding new questions and detail to a survey,broadening the coverage of a survey, increasing the fre-quency of a survey, or speeding up the processing of asurvey or administrative source. In revising existing sur-veys, an effort is always made to see if outdated questionscan be dropped or exemption levels raised by wider use of

analyzing the economy’s noninflationary growthpotential.

However, resolution of the question of theoverall impact of measurement problems on real growth will have to await the develop-ment of improved price indexes for a numberof key products; it is not now clear whetherimprovements in the deflators of the many af-fected products would, on balance, increase ordecrease the growth rate of real , nor is it clearwhether measurement problems have increasedor decreased over time.

Improvement of the measures of real outputwill require a long-term, coordinated programinvolving both and the Bureau of LaborStatistics, because improved real output meas-ures depend heavily upon improved price indexesfor deflating current-dollar . ’s proposedshort-term priorities include working closely withthe Bureau of Labor Statistics to lay the groundfor improvements for components that appearto be amenable to hedonic methods and forwhich adequate commercial data are availableto implement these methods. These compo-nents include high-tech products—such as med-ical scanners and semiconductor manufacturingequipment—and nonresidential construction.

Moreover, because some of these problemsarise from the difficulty in defining output, should undertake a long-term program to de-

tions To Improve the Accounts

sampling and statistical estimation. Such efforts usuallyinvolve another statistical agency and, hence, tend to bemore complicated and have longer start-up periods. Re-vising an existing survey also involves consultation withdata users and respondents and clearance of the revisedform through the Office of Management and Budget;these steps—along with time for notification, collection,and processing—add significantly to the lead time be-fore the improved source data can be incorporated inthe accounts. Data extensions also usually involve someincrease in respondent burden and processing cost, butthese costs are still significantly lower than those involvedin conducting a new survey.

New source data: New surveys are pursued only whenmethodological solutions are not adequate and there isno existing survey that can be modified to fill a sta-tistical gap. As in the cases of data extension, newsurveys normally involve another statistical agency. Newsurveys also require a more extensive development pro-cess because both the benefits to the data users andthe costs to respondents and statistical agencies arehigher. As a result, the start-up and lead times aresignificantly longer.

• February

velop new concepts and methods for measuringoutput in areas such as banking, insurance,financial and legal services, and managementconsulting activities.

Reliability of key components of .—Revisionsto key components of , national income, andpersonal income—especially the occasional largerevisions—cause significant difficulties for fore-casting, for budget, business, and investmentplanning, and for macroeconomic policy. Forexample, in as part of the mid-session re-view of the budget, the Office of Managementand Budget revised upward the projected FederalGovernment deficit by over billion as a re-sult of the impact of ’s billion downwardrevision to wages and salaries on the TreasuryDepartment’s projections of tax receipts.

The average revisions (without regard to sign)to the quarterly changes for key components ofcurrent-dollar range from . to . percent-age points for –. The average revisions(without regard to sign) in the quarterly changeswere billion for change in business invento-ries, billion for trade in goods and services, billion for government purchases, billion forconsumer spending for services, and - billionin consumer spending for goods. Within nationalincome, the largest dollar revisions to quar-terly changes were in compensation of employees.(These revisions also affect personal income, boththe national and regional estimates.)

These revisions reflect both methodologicalproblems and gaps in source data. Among themethodological problems, the most significantrelate to seasonal adjustment and to projectionsfor missing source data. Although past researchsuggests that the scope for improvements in sea-sonal adjustments is limited, will continueto explore improvements for volatile componentssuch as change in business inventories and tradein goods and services, because of the importanceof this problem and the large potential “bang forthe buck” associated with these improvements.In addition to these improvements, there maybe some room, albeit limited, for improvementsin ’s methods of projecting missing monthsand quarters of source data for components suchas inventories, trade in goods and services, andbonus payments.

These methodological improvements notwith-standing, the largest improvements in the re-liability of key components are likely to comefrom improvements in source data. These im-provements include more frequent updating ofsample frames, improving response rates, and

modifying and extending existing surveys to fillgaps in coverage. One of the most important ofthese is more frequent updating of sample framesfor the annual surveys that serve as benchmarksfor the monthly surveys. The Census Bureauhas already initiated several programs relevantto key source data used by . For example,the annual and monthly trade surveys were re-cently put on a schedule that would update themmore frequently using improved procedures thatmake more timely use of administrative recordsdata. Similar work is underway at the Bureauof Labor Statistics to address recommendationsby an American Statistical Association panel toimprove reporting for the monthly survey of em-ployment, hours, and earnings through use of afrequently updated probability sample at the Stateand national levels.

In addition to these and other projects to up-date sample frames, efforts are needed to improvereporting on existing surveys—either throughrestoration of cutbacks in sample size, improvedfollowup procedures, or mandatory reporting forkey indicators.

Improvements in other components of will require extensions of existing surveys. Forthese components, revisions are due to gaps inmonthly and annual survey coverage. Reducingrevisions in these areas will thus largely dependon extending coverage: In construction, trans-portation, finance, insurance, and real estate, byextending the annual services survey to cover allservice activities; in inventories, by extending theannual wholesale trade survey to cover nonmer-chant wholesalers; in nonresidential construction,by extending the monthly value-put-in-place sur-vey to better cover nonresidential reconstruction;in wages and salaries, by extending the monthlyestablishment survey to cover hours and earn-ings for all workers and to provide data onbonus payments; in profits, by extending theQuarterly Financial Report to cover construction,communications, utilities, and insurance.

In other components, revisions are due tothe lag with which survey data become avail-able and the difficulty in making projections formissing data. Many of these components wereonce sufficiently small that they could be rea-sonably estimated for quarters by projecting pasttrends from annual surveys. However, as thesecomponents have grown in size and volatility,large revisions have occurred when annual andbenchmark data are finally incorporated into theestimates to replace projections. Reducing re-visions of this kind will require increasing the

February •

frequency of existing surveys: In State and lo-cal purchases by extending the existing annualsurvey to cover State and local government pur-chases on a quarterly basis; in international tradein services, by extending the existing annual se-lected services surveys to cover key categories ona quarterly basis.

In summary, ’s proposed priority improve-ments in this area are as follows:

• Improvements in seasonal adjustments andin ’s projections for certain key compo-nents such as inventories, merchandise trade,and bonus payments.

• Improvements of the coverage of existingsurveys by more frequent and more completeupdating of samples.

• Extension of existing surveys to fill gapsin coverage by more frequent surveys forvolatile and growing categories and byextending the scope of surveys.

Substitution bias in real .—Rapid change inthe composition of output (and prices) has notonly opened coverage gaps in source data, buthas worsened pre-existing difficulties with someof the statistical methods used in the accounts.A bias occurs in fixed-weighted measures, suchas ’s featured measure of real , becausethese measures do not reflect substitution by con-sumers and producers in response to changingrelative prices. This “substitution bias” reflectsthe fact that the commodities for which outputgrows rapidly tend to be those that register de-clines, or the smallest increases, in prices. Thus,when real is recalculated using more re-cent prices, the commodities with strong outputgrowth generally receive less weight, and real growth is reduced. Changes in the economy inthe ’s and ’s exacerbated this bias, andin response developed two alternative meas-ures of real that allow for changes in therelative structure of prices. Comparisons of thesealternative measures with ’s featured meas-ure based on weights indicate the degree ofsubstitution bias. In general, fixed weights arean accurate measure for time periods relativelyclose to the base year, but they begin to overstategrowth the further one moves forward in timefrom the base period (and understate growth thefurther one moves back in time):

• For –, there was no evidence of sig-nificant substitution bias in ’s fixed--weighted measure of real growth; the

fixed-weighted and alternative indexes in-creased at the same rate. However, by and , real growth as measured byfixed weights was overstated; the fixed-weighted index increased an average of .percentage point more than the alternativeindexes.

• Since the recession trough in the first quarterof , the average annual growth rate of thefixed--weighted measure has been over-stated by an average of . percentage point.In contrast, for the economic expansions be-tween and , the average annualgrowth rate of the fixed--weighted meas-ure is understated by an average of . or. percentage point, depending upon thealternative index used.

Bias of this magnitude causes significant prob-lems for budget, investment, and business plan-ning and for forecasting. It may pose particularproblems for business cycle and policy analyststrying to assess current growth relative to long-term growth potential because the bias distortsboth estimates and hence exacerbates the prob-lems associated with quality change in measuringlong-term growth potential.

The challenge for will be to develop a pro-gram that provides users with featured measuresof output and prices that reflect weights appro-priate to each period and with information thatcan be used to assess the underlying sources ofchanges in real . Later this year, willpresent an article in the S CB that lays out the plan for this program.

Outdated and inconsistent view of the structureand organization of production.—Change in theU.S. economy has affected not only the com-position of output, but also the way output isproduced and distributed. The rapid pace ofchange has highlighted the need for improvedmeasurement systems.

The existing standard industrial classification() system presents an outdated and concep-tually inconsistent picture of economic activity.It is outdated especially in that it does notadequately detail the range of services in theeconomy, and it is inconsistent in that it distin-guishes among some industries on the basis ofhow they produce and among others on the basisof to whom they market.

The system is focused on manufacturedgoods, and although for the foreseeable future,users will probably continue to want finer de-tail for manufacturing than for other sectors,

• February

the focus in the existing system seems somewhatskewed. Manufacturing industries account for percent of the industry codes in the , yet in to-day’s economy, manufacturing accounts for only percent of by industry and percent ofemployment.

The United States, Canada, and Mexico areworking together to develop a new industry clas-sification system. This new system, the NorthAmerican Industry Classification System (),will use a production-oriented economic conceptthat will provide the following benefits: Moreconsistent classification of industries for use inanalyses of industry performance, productivity,and employment; consistent data across the threecountries; and more consistent data for new andemerging industries, for service industries, andfor high-technology industries.

An updated industry classification system willprovide the platform for modernizing ’sindustry-based accounting systems. It will pro-vide more consistent source data and structurefor ’s estimates of and gross state prod-uct by industry, for the input-output accounts,for estimates of pollution abatement and controlby industry, and for estimates of foreign directinvestment and international services by industry.

Introduction of data collected under will complement ’s efforts to improve thetimeliness and degree of integration in ’sindustry-based accounting systems. Although aquantitative indicator of these problems is notavailable, there is ample evidence of the needfor updating and better integrating these sys-tems. Problems with these accounts and the needfor better industry data have been a continuingtheme of outside experts. Affected uses includeanalyses of the interindustry and regional impactof Federal, State, and local policies, business loca-tion and marketing analyses, and various studiesin the field of industrial organization. Indeed, thegaps in coverage, lack of integration, and lack oftimeliness have contributed to a sharp decline inthe use of the input-output and related accounts.

An integrated and modernized structure foreconomic accounts is available at the interna-tional level in the newly revised and harmonizedinternational guidelines for national accounts—the and the Balance of Payments Manual.These guidelines can provide the broad out-lines for ’s efforts to update concepts and toprovide a more integrated accounting structure.Modernization along these lines will allow theUnited States to take advantage of the experienceand expertise embodied in the revisions and to

make U.S. estimates more comparable with thoseprepared by other countries.

Completion of the several projects to updateand better integrate the structure of the economicaccounts is a high priority. These projects includethe following:

• Completion and implementation of the new in conjunction with other statisticalagencies.

• Modernization and better integration of’s industry accounts, especially the input-output accounts, and gross state productby industry, and pollution abatement andcontrol expenditures by industry.

• Modernization of ’s ’s and balanceof payments accounts in line with the newlyrevised international guidelines.

The need for better measures of investment,saving, and wealth

Issues about the scope of existing measures.—In the ’s, investment is limited to businessinvestment in inventories, structures, and equip-ment, but a broader view of the Nation’s wealthwould include highways, dams, schools, andother public infrastructure, natural resources,and intangible assets, such as computer softwareand, even more broadly, training and education.

A broader definition of investment may bequite helpful in understanding the sources ofeconomic growth and the returns to, and ad-equacy of, various types of public and privateinvestment. Treatment of government capitaloutlays as fixed investment would add , bil-lion, and the inclusion of computer software ininvestment, between billion and billion—together about percent—to the net stock ofreproducible capital in the national wealth. Theexpansion of the definition of investment throughsatellite accounts can also aid in developing amore comprehensive picture of the stock of pro-ductive capital and wealth. The inclusion ofresearch and development capital would add be-tween , and , billion, or – percent,to the net stock of wealth; mineral and othermarketed natural resources would add between and , billion, or – percent.

’s proposed projects to address these issueslargely rely on new methods and existing sourcedata. In order of importance, they are as follows:

• Expansion of investment to in-clude government expenditures on struc-tures and durable equipment and pur-

February •

chases of computer software and some otherintangibles.

• Extension of ’s satellite accounts for re-search and development and for naturalresources.

Issues in the measurement of investment, saving,and wealth.—Differences in source data, meth-ods, and concepts result in inconsistencies among’s and capital stock estimates, the Fed-eral Reserve Board’s flow of funds accounts andbalance sheets, and the Bureau of Labor Statisticscapital services estimates. These inconsistenciespresent significant difficulties for integrated anal-ysis of such issues as productivity, domestic andinternational capital flows, returns to investment,the adequacy of domestic saving and capital for-mation, and the effect of changes in financialholdings on consumer and business spending.

The importance of such integrated analysis wasunderscored by the depressing effect on house-hold saving of the appreciation in the net worthof households as housing and securities pricesrose in the ’s. To researchers trying to un-derstand the sources of the decline in the U.S.household saving rate—and the potential effectof tax incentives on raising it—the - to -percentage-point difference between the personalsaving rates derived from ’s ’s and thosederived from the Federal Reserve Board’s flow-offunds accounts posed significant problems.

In addition to these inconsistencies between thevarious accounts, researchers have long been con-cerned with conceptual and empirical problemsassociated with the use of straight-line deprecia-tion in ’s estimates of depreciation and capitalstocks.

’s proposed priorities in this area deallargely with methodology and structure:

• Development of measures of depreciationand capital stocks that are consistent witheconomic theory and with existing empiricalmeasures of the decline in used-asset prices.

• Better integration of and Federal ReserveBoard real and financial accounts along thelines of the System of National Accounts .

The need for measures to fill gaps in thecoverage of international transactions

Increased integration in world markets for goods,services, and capital, in combination with ma-jor advances in computer and communications

. Government purchases of capital goods is included in ; treatingthem as investment would reallocate them from government purchases tofixed investment.

technology, have resulted in large gaps in ’scoverage of international transactions. Thesegaps pose difficulties for the analysis of trade,monetary, and regulatory policy. For example,in , with the economy in the midst of whatwas described as a credit-crunch-induced reces-sion, a billion statistical discrepancy in thebalance of payments accounts made it difficultto tell whether or not foreign capital flows weredrying up: Based on the financing implied by thecurrent account, there was only a modest dropoff,whereas the direct measure from the capital ac-count suggested a sharp reduction. At the sametime, on the regulatory policy side, gaps in thecoverage of foreign lending to U.S. nonfinancialcorporations by banks in the Caribbean causedthe official statistics to understate by roughly one-third the indebtedness of U.S. companies and thepenetration of foreign banks into U.S. markets.

Since then, has been able to close manyof these gaps in coverage by data exchanges withother countries and by improvements in surveycoverage through its work with the Treasury De-partment and the Federal Reserve. Nevertheless,gaps remain in some areas, and new gaps areemerging in others.

Two of the largest remaining gaps in the cover-age of goods and services are in financial servicesand computer software. The proliferation of newfinancial services traded in international marketshas left a hole in ’s coverage of services, whilethe rapid growth in trade in computer softwarehas caused a growing gap in what was once amiscellaneous category in the harmonized tariffcode that was valued by the cost of the media(for example, the value of a blank diskette) ratherthan the market value of the software itself. Thesize of these and other gaps in trade in servicesmay be as large as - billion.

In the capital accounts, large gaps remain inthe coverage of U.S. portfolio investments abroadand foreign portfolio investments in the UnitedStates. These gaps result from direct transactionsbetween U.S. and foreign residents that bypassthe U.S. brokers, dealers, banks, and other finan-cial institutions that form the U.S. data collectionsystem. In addition to these existing gaps, newgaps are emerging through growth in new finan-cial instruments that are not separately identifiedor fully covered by the existing data collectionsystem. A rough guess of the size of the omissionsin U.S. international capital flows and stocks dueto remaining and newly emerging gaps might beas large as billion for the flows and billion for the stocks.

• February

Table 3.—Proposed Actions and Milestones in Implementation

Proposed actionsMilestones in implementation

1995 1996 1997 1998 1999

Need for new and improved output measures

Extension of qualityadjustment of pricesused in real GDP.

............................................. Incorporate quality-adjustedmeasures for selectedhigh-tech products.

Develop new estimates fornonresidentialconstruction using modelpricing or hedonicmethods.

Continue work onextension of qualityadjustments.

Further conceptual work ondifficult-to-measure goodsand services.

............................................. Develop new concepts andmethods for measuringbanking and otherfinancial services.

Publish new estimates forbanking and otherfinancial services.

Identify and develop newconcepts and methodsfor measuring otherdifficult-to-measuregoods and services.

Publish new estimates forother difficult-to-measuregoods and services

More frequent updating ofsample frames forexisting surveys, morefrequent surveys forcertain sectors, andextension of existingsurveys to fill gaps incoverage.

Begin work with otheragencies to identifycost-effective actions.

Continue work, extendingconsultations tobusiness community.

Conduct new and revisedsurveys using newsample frames andmethods.

Incorporate data from newand revised surveys.

Improvements in seasonaladjustments.

............................................. Begin work with otheragencies on improvingseasonal adjustment.

Begin work on revisedseasonal adjustment aspart of reengineering.

Incorporate revisedseasonal adjustmentmethods for categoriessuch as inventories andmerchandise trade.

Incorporate revisedseasonal adjustmentmethods for other keycomponents.

Improvements in projectionsfor missing source data.

............................................. Begin design work for built-in analytics andprojections methods aspart of reengineering.

Incorporate improvedmethods into currentquarter projections forcategories such asinventories andmerchandise trade.

Incorporate improvedmethods into currentquarter projections forother key components.

Introduction of newweighting schemes.

Introduce more currentweights for real GDP.

............................................. ............................................. ............................................. Update base-year weightsfor current estimates

Develop a new industryclassification system.

Present NAICS prioritycategories to industrygroups and data usersfor comment andrevision.

Continue work withindustry groups anddata users and finalizeNAICS.

Work with Census toredesign forms on thebasis of NAICS.

Update and better integrateBEA’s accounts withinthe context of newinternational guidelines.

............................................. ............................................. Complete the 1992benchmark input-outputaccounts 5 years afterthe reference year.

Need for better measures of investment, saving, and wealth

Expand the coverage ofinvestment along theoutlines of the newinternational guidelines.

............................................. Publish new estimates ofcomputer software.

Update and improveestimates of researchand development.

Extend integratedeconomic andenvironmental satelliteaccounts (IEESA’s) toinclude selectedrenewable resourceestimates and revisenonrenewable resourceestimates.

Continue long-termimprovements incomputer software andwork on otherintangibles.

Expand disaggregation ofexisting NIPA items inthe IEESA’s.

Use of improveddepreciation andvaluation methods.

Introduce new depreciationand capital estimates.

Extend empirical work onused asset prices toother assets.

Extend empirical work onused asset prices toother assets.

Integration of real andfinancial accounts.

Begin work with theFederal Reserve Boardto develop multiyearplan for betterintegration.

February •

’s proposed priorities in these areas are asfollows:

• Modification and extension of existing sur-veys and customs documents to better cap-ture exports and imports of computer soft-ware and of courier and other rapidlygrowing services.

• Completion of ’s benchmark financialservices survey and establishment of a newannual survey of financial services.

• Completion, and institutionalization, ofwork with the Treasury Department ontheir benchmark survey of U.S. portfolioinvestment abroad.

• Extension of data exchanges with other coun-tries and central banks through the adoptionof standard definitions and coverage andthe international coordination of benchmarksurveys.

Table 3.—Proposed Action

Proposed actions1995 1996

Need to fill gaps in

Extension and revision ofexisting surveys andforms to cover newproducts, services, andfinancial instruments.

Revise product code anddata collection to includea separate category forcomputer software.

Develop initial ethe full markecomputer sofbased on revmethods anddata.

............................................. Work with the bcommunity torevised and e‘‘selected sersurvey to covgrowing, and services cate

Begin work with Treasuryand the FederalReserve on cost-effective means ofcollecting data onderivatives and newfinancial instruments.

Work with the bcommunity torevised survederivatives annew financialinstruments.

Extension of dataexchanges with othercountries and centralbanks.

Continue work oninternationallycoordinated definitionsfor data exchanges.

Incorporate dataappropriate, fcountries adostandardized

Development of newsurveys.

............................................. Incorporate resubenchmark finservices survconduct annuservices surv

Process new benchmarksurvey of U.S. portfolioinvestment abroad.

Continue procesbenchmark suU.S. portfolioabroad.

Continue internationalcoordination indevelopment ofinternational benchmarksurvey.

Continue internacoordination idevelopment international bsurvey.

• Extension of existing portfolio surveys toseparately identify and better cover deriva-tives and other new financial instruments.

Milestones in implementing the proposed actions

Although some of the methodological improve-ments can be implemented in the next year orso, many of the proposed projects are multiyearefforts. The timing of these projects is naturallydependent on the availability of resources. Themilestones for each of the projects included intable are based on the assumption that fundingis commensurate with that of similar projects inthe past.

These projects have been chosen so as to focuson those with the highest relative importance atthe lowest possible cost. Despite this focus ongetting the largest “bang for the buck,” many ofthe projects cannot be accomplished with exist-

s and Milestones in Implementation—Continued

Milestones in implementation

1997 1998 1999

the coverage of international transactions

stimates oft value of

twareised source

Update and improveestimates of the fullmarket value ofcomputer software

usiness developxtended

vices’’er new,volatile

gories.

Conduct extended servicessurveys.

Process and incorporateresults from extendedsurveys into nationaland internationalaccounts

usiness develop ay to coverd other

Conduct revised survey tocapture derivatives andother new financialinstruments.

Process and incorporateresults from revisedsurvey into national andinternational accounts

, asrompting newdefinitions.

Incorporate data, asappropriate, fromcountries adopting newstandardized definitions.

Incorporate data, asappropriate, fromcountries adopting newstandardized definitions.

Incorporate data, asappropriate, fromcountries adopting newstandardized definitions

lts fromancial

ey andal financialey.

Incorporate results fromannual financial servicessurvey

sing newrvey of

investment

Incorporate results fromnew benchmark surveyof U.S. portfolioinvestment abroad

tionalnofenchmark

Finalize design andcollection ofinternationallycoordinated benchmarksurvey.

Collect and process datafrom internationallycoordinated benchmarksurvey.

Continue processing resultsof internationallycoordinated benchmarksurvey

• February

ing resources; this time schedule and project listwill clearly have to be adjusted in light of futureresources. The milestones in table , however,do provide a notion of the sequencing and tim-ing of the returns from an integrated strategicplan to maintain and improve the accounts. Im-plementation of such a plan would update theaccounts in a timely manner to reflect changes inthe economy. Thus, the accounts could continueto provide the degrees of accuracy, reliability, andrelevance that their users have come to expect.

An environment for change in the accounts

Outside experts, in commenting on ’s work,have made at least two other recommendationsthat should be noted. One is that should

extend and update its documentation of method-ology, and the other is that should form anadvisory committee. Both recommendations takeon increased importance in an environment ofchange in the accounts. takes seriously itsobligations to provide documentation and willbe working toward a revamped system of doc-umentation that is as up to date and readilyavailable as its estimates. Over the years, ’saccounts have benefited from critique and com-ment, ranging from formal but intermittent “blueribbon” panels to informal day-to-day contactwith users. Just as is seeking advice as it putstogether this plan, would welcome advice—especially regularized advice from a group ofusers of the accounts—as it seeks to carry out theplan.