Economic Statistics: Measurement Problems Can Affect the Budget and
Economic Policy-Making (Letter Report, 05/02/95, GAO/GGD-95-99).
Pursuant to a congressional request, GAO reviewed problems with economic
statistics, focusing on how they can affect the federal budget as well
as the formulation of economic policy.
GAO found that: (1) many economists believe that several technical
problems limit the Consumer Price Index's (CPI) accuracy in measuring
inflation; (2) these technical problems could cause CPI to overstate
inflation, although complete information on measurement errors is
inconclusive; (3) the extent and direction of inflation measurement is
important, since slight misstatements could significantly affect
government costs; and (4) statistical limitations can affect the
policymaking process by increasing policymakers' uncertainty about
current economic conditions and the potential effectiveness of economic
policies.
--------------------------- Indexing Terms -----------------------------
REPORTNUM: GGD-95-99
TITLE: Economic Statistics: Measurement Problems Can Affect the
Budget and Economic Policy-Making
DATE: 05/02/95
SUBJECT: Statistical data
Demographic data
Economic policies
Economic indicators
Consumer Price Index
Inflation
Economic analysis
Budgeting
Statistical methods
Errors
IDENTIFIER: Consumer Price Index
Supplemental Security Income Program
North American Free Trade Agreement
NAFTA
**************************************************************************
* This file contains an ASCII representation of the text of a GAO *
* report. Delineations within the text indicating chapter titles, *
* headings, and bullets are preserved. Major divisions and subdivisions *
* of the text, such as Chapters, Sections, and Appendixes, are *
* identified by double and single lines. The numbers on the right end *
* of these lines indicate the position of each of the subsections in the *
* document outline. These numbers do NOT correspond with the page *
* numbers of the printed product. *
* *
* No attempt has been made to display graphic images, although figure *
* captions are reproduced. Tables are included, but may not resemble *
* those in the printed version. *
* *
* A printed copy of this report may be obtained from the GAO Document *
* Distribution Facility by calling (202) 512-6000, by faxing your *
* request to (301) 258-4066, or by writing to P.O. Box 6015, *
* Gaithersburg, MD 20884-6015. We are unable to accept electronic orders *
* for printed documents at this time. *
**************************************************************************
Cover
================================================================ COVER
Report to Ranking Minority Member, Committee on Banking and Financial
Services, House of Representatives
May 1995
ECONOMIC STATISTICS - MEASUREMENT
PROBLEMS CAN AFFECT THE BUDGET AND
ECONOMIC POLICYMAKING
GAO/GGD-95-99
Economic Statistics
Abbreviations
=============================================================== ABBREV
BEA - Bureau of Economic Analysis
BLS - Bureau of Labor Statistics
CBO - Congressional Budget Office
CPI - Consumer Price Index
GDP - gross domestic product
OMB - Office of Management and Budget
Letter
=============================================================== LETTER
B-260039
May 2, 1995
The Honorable Henry B. Gonzalez
Ranking Minority Member
Committee on Banking and Financial Services
House of Representatives
Dear Mr. Gonzalez:
This report responds to your request, which cited press reports that
problems with economic statistics were hindering the formulation of
economic policy. In the past several years, these reports have
identified problems with certain economic statistics. Specifically,
you requested that we discuss how the reported problems can affect
the federal budget as well as the formulation of economic policy.
After consulting with your office, we decided to use (1) the Consumer
Price Index (CPI) to illustrate how federal statistics can affect the
federal budget and (2) international transaction and economic output
and productivity statistics to show how federal statistics can affect
the economic policymaking process.
As agreed with your office, we are separately reporting on the
Economic Statistics Initiative, a multiyear effort by statistical
agencies, which sought to address acknowledged problems with a broad
range of economic statistics, including several addressed in this
report.
RESULTS IN BRIEF
------------------------------------------------------------ Letter :1
Statistics on changes in prices are closely linked to billions of
dollars in federal spending and tax receipts. Changes in prices are
commonly measured by the CPI, to which many federal benefits and
individual income taxes are tied in order to offset the effects of
inflation. However, officials of the Bureau of Labor Statistics
(BLS)--the agency responsible for producing the CPI--and many
economists agree that several technical problems limit the CPI's
accuracy in measuring inflation.
Most of the recent studies we reviewed indicated that these technical
problems could cause the CPI to overstate inflation. However, the
studies do not provide definitive support for a specific estimate of
the CPI's misstatement of inflation. According to BLS, the studies
are inconclusive because they do not determine the effects of all
measurement problems on the CPI, particularly problems relating to
product quality changes. To illustrate, research indicated that
price increases are sometimes the result of improvements to the
quality of a product, such as upgrades to an automobile that make it
safer or last longer. BLS uses a number of methods to adjust the CPI
to account for quality changes in products. However, BLS officials
note that these methods do not adequately capture the effect of
quality changes on the price of some products. BLS officials are
uncertain whether the overall measurement error associated with
quality changes results in price increases being overstated or
understated in the CPI.
The extent and direction of inflation misstatement is important. For
example, according to the Office of Management and Budget (OMB), if
the CPI overstated inflation by 0.2 percentage point annually--the
lower end of the range of estimates of the overstatement in the
studies--the result would be increased entitlement payments and
decreased tax receipts that could cost the government about $20
billion over the next 5 years.
Shortcomings of economic statistics can also affect the policymaking
process. Policymakers use economic statistics to improve their
understanding of the potential effect of policy alternatives on the
economy. However, many of the statistics series used by policymakers
do not adequately depict the economic conditions they are intended to
measure. For example, statistics on international trade and
investment have not kept pace with changing patterns of international
economic relations. Similarly, measures of economic output and
productivity fail to account for the increasing importance of the
service sector to the nation's economy. While it is not possible to
demonstrate how the shortcomings of particular statistics affect
policy decisions, such shortcomings do increase policymakers'
uncertainty about current economic conditions or long-term trends and
the potential effectiveness of policies.
BACKGROUND
------------------------------------------------------------ Letter :2
Economic statistics are intended to summarize various characteristics
of the nation's economy in an accurate and understandable way. Over
the last decade, there has been growing concern among many experts
that economic statistics have not kept pace with changes in the
economy. Structural changes in the economy--such as new
technologies, emerging industries, and demographic shifts--require
statistical agencies to continually reassess old methodologies and
data collection techniques and devise new ones. This process can be
analytically complex and costly.
Responding to concerns about problems with federal economic
statistics, President Bush, as a supplement to his first State of the
Union message in February 1989, proposed an increase in federal
spending to improve the quality and accuracy of business and economic
statistics. As a result, a working group was established in April
1989 to assess problems within the federal economic statistical
system and recommend needed improvements. The working group focused
on recommending improvements that could be met in a timely manner
with limited budget increases. The recommendations issued by the
group were known as the Economics Statistics Initiative and were to
be implemented in stages from fiscal years 1990 to 1996. Despite
this initiative, many problems with economic statistics remain
unresolved.
OBJECTIVE, SCOPE, AND
METHODOLOGY
------------------------------------------------------------ Letter :3
Our objective was to discuss how reported problems with selected
statistics produced by the federal government can affect the federal
budget and the formulation of the nation's economic policy. To meet
this objective, we first identified several federal statistical
series that have direct effects on the federal budget and/or are used
by government officials in the formulation and implementation of
economic policy. Relying on our previous work, the work of experts
on economic statistics, and interviews with officials from federal
statistical agencies, we determined that many of these statistical
series are subject to measurement problems to some degree. After
consulting with your office, we decided to focus our work on selected
statistical series that are used extensively in either the budget or
policymaking process and which also have measurement problems that
could be readily demonstrated. We chose (1) the CPI to illustrate
how federal statistics can affect the federal budget and (2)
international transaction and output and productivity statistics to
show how federal statistics can affect the economic policymaking
process. The problems of other statistics and their effects,
however, may differ from those of the statistics we selected for our
illustrations.
To develop these illustrations, we first discussed the problems with
the selected statistics with officials from the agencies responsible
for producing them.\1 In addition, to obtain information on the
agencies' plans for addressing these problems, we reviewed the
agencies' budget and other submissions to OMB, prior studies and
reports on economic statistics, and federal budget documents and
analyses. We also judgmentally selected and reviewed several of the
most recent academic studies and reports, congressional testimony,
and other records documenting concerns with these statistics. We
obtained views on the quality of the selected statistics from
recognized experts outside of government.
Our second step in developing our illustrations was determining how
identified problems with the selected statistics could affect the
budget and economic policy. To do so, we first discussed potential
effects of the identified problems with officials from the agencies
that produce these statistics. We then examined budget and other
documents to determine what effect, if any, problems with CPI could
have on the budget. To analyze the sensitivity of federal spending
and taxes to variations in consumer prices, we obtained assistance
from officials at OMB and the Congressional Budget Office (CBO). To
illustrate how problems with international transaction and output and
productivity statistics affect economic policy, we obtained the views
of agency officials and congressional staff involved in shaping
policy on the economic activities measured by these statistics. We
also reviewed our prior products that discussed CPI, international
transaction statistics, and other topics related to this report.
(They are listed at the end of this report.)
As agreed with your office, we did not attempt to identify the
underlying causes or possible corrective actions for the problems
with the selected federal statistics.
We obtained written comments on our report from the Department of
Commerce where BEA and Census are located. We also obtained oral
comments from the Office of Federal Statistical Policy within OMB,
which is responsible for managing the federal statistical system.
The Department of Labor, where BLS is located, was not able to meet
our deadline for providing comments. However, BLS staff informally
provided suggestions for technical clarifications to the report. We
did our work primarily in Washington, D.C., from December 1993 to
March 1995 in accordance with generally accepted government auditing
standards.
--------------------
\1 BLS, in the Department of Labor, is responsible for the CPI. The
Bureau of Economic Analysis (BEA) and the Bureau of the Census, both
in the Department of Commerce, are responsible for most international
transaction statistics. BEA and BLS are responsible for statistics
on output and productivity, respectively.
THE CPI SIGNIFICANTLY AFFECTS
FEDERAL SPENDING AND TAXES
------------------------------------------------------------ Letter :4
The CPI measures average price changes over time for consumer goods
and services. BLS has reported that several measurement issues
affect the accuracy of the CPI, and several recent studies and a
number of prominent economists have said that the CPI overstates
inflation. However, the studies do not provide firm support for a
specific estimate of the overstatement or even a definitive
demonstration that the CPI, in fact, overstates the true rate of
inflation. The studies quantified the effect on CPI of some
measurement problems but did not measure the effect of quality
changes on prices, which according to BLS is the most serious CPI
measurement issue. BLS officials said that without more information
on the effect of quality on prices, a definitive estimate of the
CPI's misstatement of inflation is not possible.
Any misstatement of the CPI would have serious consequences. OMB and
CBO estimated that even a small overstatement of the rate of
inflation by the CPI could add billions of dollars to the federal
deficit as a result of higher federal benefits payments and lower tax
revenues. Some entitlement programs and income tax brackets\2 are
tied to the CPI to protect entitlement recipients and taxpayers from
the effects of inflation.
--------------------
\2 The Internal Revenue Code requires that the personal exemption,
the standard deduction, the minimum and maximum dollar amounts of
each tax rate bracket, the thresholds for the phaseout of the
personal exemptions deduction, and the earned income credit, among
other provisions, be indexed to the CPI. Such indexation is designed
to prevent taxpayers from edging into higher tax brackets as their
income rises to compensate for inflation.
THE CPI IS USED AS A MEASURE
OF INFLATION
---------------------------------------------------------- Letter :4.1
The CPI is a statistical measure of changes in prices of a fixed
market basket of goods and services purchased by urban consumers. It
is the only index compiled by the federal government that is designed
to measure the changes in the purchasing power of the urban
consumer's dollar.
The CPI is used as a measure of the rate of inflation in the economy.
The administration, Congress, and the Federal Reserve use trends in
the CPI as an aid in formulating fiscal and monetary policies.
Business and labor leaders as well as private citizens use the CPI as
a guide to making economic decisions. The CPI is also used to adjust
other measures for price changes, such as the income thresholds used
by the government to determine the number of people living in
poverty.
In addition, a version of the CPI\3 is used to adjust some federal
benefit payments for inflation. For example, in 1993, 42 million
Social Security beneficiaries, 6 million Supplemental Security Income
recipients, about 27 million food stamp recipients, and more than 3
million military and federal civilian retirees and survivors had
their benefits adjusted for inflation, as a result of changes in the
CPI. Finally, the CPI is used to adjust the federal individual
income tax structure to prevent bracket creep, i.e., increases in
real tax rates due solely to inflation. Benefit payments and tax
brackets are adjusted automatically by the CPI, rather than on the
basis of discretionary policy decisions.
Because the CPI is used to adjust benefit programs and tax brackets,
it is often thought of as a cost-of-living index. However, according
to BLS, while the CPI measures price changes, it is not intended to
be a measure of changes in living costs. Like the CPI, a
cost-of-living index would measure price changes between time
periods. The two would differ, however, in what they would seek to
measure. The CPI is designed to measure the price changes for a
fixed market basket--the same amounts of the same goods and
services--over time. A cost-of-living index, however, would allow
changes in the mix of goods and services whose prices are measured.
A cost-of-living index would thus recognize that prices for all goods
and services do not change by a uniform amount, and that consumers
are able to substitute one product for another as their relative
prices change. An ideal cost of living index would measure how much
more it would cost to maintain a constant standard of living or level
of satisfaction even after such product substitutions. It could also
measure how changes in areas such as environmental quality or public
services would induce changes in consumer purchases of goods and
services that would be included in the index. While such an index
could be conceptually superior to the CPI as an inflation index for
adjusting federal benefits and tax brackets, there are no generally
accepted methodologies for calculating one.
BLS bases the CPI on a sample of prices of food and beverages,
apparel, housing, transportation, medical services, entertainment,
and other goods and services that consumers buy. These items are
known collectively as the market basket. BLS conducts a continuous
Consumer Expenditure Survey to determine which particular goods and
services consumers are purchasing. This information is to
subsequently be used when BLS revises the market basket. BLS
generally does a revision every 10 years. BLS last did a revision in
1987, which reflected consumer spending patterns for 1982 through
1984.
BLS measures price changes each month by checking the prices of the
sample of goods and services and then comparing the aggregate cost of
the sample to the cost in the previous month. BLS obtains prices for
most goods and services in the market basket through personal visits
by its field representatives to approximately 21,000 retail
establishments. BLS staff also sample about 60,000 housing units to
obtain information on housing costs. In addition, BLS sponsors a
survey of consumers, conducted by the Bureau of the Census, to
determine which retail outlets its representatives should visit to
obtain its monthly price quotations. This survey is conducted in
approximately 20 percent of a sample of urban areas each year and, as
a result, the entire sample is updated every 5 years.
--------------------
\3 BLS calculates two CPIs. The first, CPI-W, represents the buying
habits of wage earners and clerical workers and is used to adjust
benefit programs. The second, CPI-U, represents the buying habits of
all urban consumers and is the measure generally used by
policymakers.
BLS RESEARCHERS HAVE
REPORTED THAT MEASUREMENT
ISSUES AFFECT CPI ACCURACY
---------------------------------------------------------- Letter :4.2
In December 1993, BLS researchers reported that several measurement
problems affect the accuracy of the CPI.\4 According to BLS
researchers, these measurement problems could cause the CPI to
inaccurately gauge price changes for some products. One measurement
problem is that the CPI may not adequately reflect current consumer
spending patterns because BLS only does a revision of the market
basket to reflect new consumer spending patterns about every 10
years. As a result, the current CPI may not fully reflect some
current consumer spending patterns, such as the substitution of
lower-priced goods for higher-priced ones.
Another measurement problem, according to BLS, stems from the
procedures BLS uses to update the sample of retail outlets BLS field
staff visit to obtain monthly prices for the items in the market
basket. In recent testimony before the Committee on Finance, U.S.
Senate,\5 the BLS Commissioner noted that the procedures for
systematically introducing new outlets and items into the CPI
inadvertently tend to misstate the effect of temporary price
increases. The procedures give higher weight than is justified to
prices that are temporarily low in the month the new samples are
introduced and lower weight than is justified to prices that are
temporarily high. Thus, these procedures can cause an overstatement
of price change in the period immediately following the sample
update.
A third measurement problem is the difficulty in separating price
changes resulting from quality changes from those due to inflation.
Research indicated that price increases are sometimes the result of
improvements to the quality of a product, such as upgrades to an
automobile that make it safer or last longer. In these situations,
consumers are paying more for a product, but they are getting a
better product. BLS uses a number of methods to adjust the CPI to
account for quality changes in products. However, BLS officials note
that these methods may not adequately capture the effect of quality
changes on the price of some products. Recent research suggests that
BLS underadjusts the CPI for quality improvements to some products
and overadjusts the index for quality improvements to other products.
An underadjustment for quality would result in an overstatement of
inflation while an overadjustment would result in an understatement.
Some research also indicates that the CPI does not account for
quality deterioration in some products and services, which would
result in an understatement of inflation. BLS officials indicated
that it is not known whether the overall measurement error associated
with quality changes results in price increases being overstated or
understated in the CPI. They said that more research is needed to
improve the methodologies for separating the effect of quality
modification price changes from inflation-related price changes.
--------------------
\4 Monthly Labor Review, U.S. Department of Labor, Bureau of Labor
Statistics (Dec. 1993), pp. 3-46, 59-62. Articles appearing in the
Monthly Labor Review reflect the views of the authors and not
necessarily those of BLS.
\5 Statement of Katherine G. Abraham, Commissioner, Bureau of Labor
Statistics, before the Senate Finance Committee, U.S. Congress, Mar.
13, 1995.
STUDIES INDICATE THAT THE
CPI DOES NOT ACCURATELY
REFLECT CURRENT PRICE
CHANGES
---------------------------------------------------------- Letter :4.3
BLS researchers and other price measurement experts disagree on the
extent of the measurement problems surrounding the CPI. Several
recent studies on the accuracy of the CPI indicated that it may
overstate inflation. We reviewed four of the most recent of these
studies. Each made a different estimate of the overstatement. A
study by BLS researchers estimated that problems in measuring
consumer spending patterns caused CPI to overstate inflation by 0.1
to 0.2 percent annually.\6 A study by staff of the Federal Reserve
Bank of Dallas concluded that the CPI overstated inflation by less
than 1.0 percent annually,\7 while CBO staff and Federal Reserve
staff studies estimated annual overstatement ranges of 0.2 to 0.8
percent \8 and 0.4 to 1.5 percent,\9 respectively.
BLS officials emphasized that the authors of each of the four studies
consider their estimates to be subject to great uncertainty. The
estimates are based, in part, on projecting measurement problems that
have been identified for specific components of the CPI to other
components that may face similar problems. BLS officials noted that
in making their estimate, the authors of the Federal Reserve Bank of
Dallas study warned that "at present we simply do not know" the
magnitude of CPI mismeasurement.
--------------------
\6 Monthly Labor Review, U.S. Department of Labor, Bureau of Labor
Statistics (Dec. 1993). BLS researchers note that this is not an
estimate of the CPI's overall overstatement of inflation.
\7 Mark A. Wynne and Fiona Sigalla, "The Consumer Price Index,"
Economic Review - Second Quarter 1994, Federal Reserve Bank of Dallas
(Feb. 1994).
\8 Is The Growth of the CPI a Biased Measure of Changes in the Cost
of Living? Congressional Budget Office (Oct. 1994), p. 4. The
Director of CBO included this estimate in testimony before the
Committee on Finance, U.S. Senate, on March 13, 1995.
\9 David E. Lebow, John M. Roberts, and David J. Stockton,
Division of Research and Statistics, Board of Governors of the
Federal Reserve System, Monetary Policy and "The Price Level" (Aug.
1994), table 5. In testimony at the March 13, 1995, hearing, the
Chairman of the Board of Governors of the Federal Reserve System said
that he believed the CPI overstated inflation.
BLS HAS ADDRESSED SOME
ISSUES AFFECTING CPI
ACCURACY
---------------------------------------------------------- Letter :4.4
Because of its concern for the accuracy of the CPI, over the years
BLS has made changes to the CPI's methodology. For example, in the
1980s, BLS changed its methodology for measuring changes in cost of
shelter. It switched from measuring the costs associated with
purchasing a house to measuring the cost of consuming the services
provided by housing. Since the CPI is intended to measure
consumption, this methodology allows BLS to remove the investment
aspect of homeownership from the CPI and thus more accurately
reflects the cost of shelter in the CPI. More recently, in January
1995, BLS incorporated new estimation techniques intended to further
increase the accuracy of the CPI's shelter, food-at-home, and
prescription drug components.
BLS also has taken steps designed to improve its methodologies for
measuring the effect of quality on price changes. Research had
pointed to the possibility that quality-related problems may have
resulted in the CPI understating inflation.\10 For example, a 1988
study indicated that measurement problems associated with quality
caused the CPI to understate inflation for several large CPI
components including housing, new automobiles, and apparel. Since
this study, BLS has changed its methodologies for measuring quality
changes for the housing and apparel components of the CPI.\11
In addition, BLS has initiated an effort intended to improve the
CPI's ability to reflect changes in consumer spending patterns. In
late 1994, BLS began a 6-year, $61 million project, part of which is
to update the information on the percentage of consumption spending
on particular items in the market basket. BLS officials told us that
when the market basket update is finished, the CPI will be revised to
better reflect current consumer spending patterns, which would
address one of the measurement problems that tend to cause CPI to
overstate inflation. The update is scheduled to be completed by
January 1998.
--------------------
\10 Jack Triplett, "The Measurement of Inflation: A Survey of
Research on the Accuracy of Price Indexes," in Paul H. Earl (ed.),
Analysis of Inflation (Lexington, Mass.: Lexington Books, 1975) and
Jack Triplett, "Price Index Research and Its Influence on Data: A
Historical Review" (Paper presented to the 50th Anniversary
Conference of the Conference on Research on Income and Wealth,
Washington, D.C., May 12-14, 1988) as cited in Wynne and Sigalla,
1994.
\11 In 1988, BLS began adjusting the housing component of CPI to
account for the effect of aging and depreciation on housing stock
quality, which it believes improves the accuracy of its measurement
of changes in housing prices. Since 1991, BLS has been making direct
adjustments for quality differences in the price observations used to
calculate the apparel component of the CPI. Previously, BLS imputed
the effects of quality changes on apparel prices.
ESTIMATES OF THE EFFECTS OF
CPI OVERSTATEMENT ON THE
FEDERAL BUDGET
---------------------------------------------------------- Letter :4.5
The CPI directly affects a significant portion of the budget because
many federal benefits as well as individual income taxes are tied, or
indexed, to the CPI in an effort to negate the effects of
inflation.\12 For example, during fiscal year 1994, 31 cents of every
federal dollar spent, or $460 billion, and 44 cents of every dollar
in tax revenue collected, or $550 billion, were subject to indexing
to the CPI.
If the CPI overstated the rate of inflation, this overstatement would
result in increased benefit payments and reductions in the amount of
tax revenues the government collects from individuals' taxes.
Estimates by OMB and CBO indicated that even a small overstatement of
inflation by the CPI may add billions of dollars to the federal
deficit.\13 For example, if the CPI overstated inflation by as little
as 0.2 percentage point annually from 1995 through 1999, an estimated
$19.1 billion would be added to the deficit over that 5-year period,
according to OMB estimates and an estimated $21.8 billion, according
to CBO estimates. If the CPI overstated inflation by 0.5 percentage
point annually over the same 5-year period, the budget deficit would
increase by about $47.8 billion, according to OMB estimates (See fig.
1.).
Figure 1: The Estimated Effect
on the Federal Deficit of a 0.5
Percentage Point Annual
Overstatement of Inflation by
the CPI for Fiscal Years
1995-1999
(See figure in printed
edition.)
Source: OMB estimates.
An understatement of inflation by the CPI would have the opposite
effect of an overstatement on the budget.
--------------------
\12 Indexation is the process of relating federal benefits and taxes
to an indicator of some kind, usually inflation.
\13 OMB and CBO each estimated the budgetary effect over 5 years,
1995 through 1999. Their estimates cover programs indexed directly
to the CPI, including Social Security, Supplemental Security Income,
military retirement, and civil service retirement as well as personal
income taxes. Programs whose outlays are partially related to
subcategories of the CPI, such as food stamps and child nutrition,
are not included in either estimate because a correction in
calculating the CPI would not necessarily affect them.
STATISTICAL LIMITATIONS
INCREASE THE UNCERTAINTIES THAT
FACE POLICYMAKERS
------------------------------------------------------------ Letter :5
Economic statistics can also affect the policymaking process. While
it is not possible to demonstrate that if the shortcomings of
particular statistics were resolved, policymakers would have made
different decisions, such shortcomings increase policymakers'
uncertainty about current economic conditions and long-term trends.
In some cases, these shortcomings, which include limited coverage or
outdated measures of some economic activities, preclude a full and
up-to-date understanding of current economic conditions. Such
shortcomings may also hinder analytical efforts to assess whether
policies are working as intended. In the following sections, we
discuss how outdated and incomplete measures of international trade
and investment and economic output and productivity increase economic
policymakers uncertainty.
INCOMPLETE COVERAGE OF TRADE
AND INVESTMENT FLOWS
AFFECTED POLICYMAKERS'
ABILITY TO EVALUATE TRADE
POLICY
---------------------------------------------------------- Letter :5.1
International economic statistics are compiled primarily by BEA and
Census. BEA compiles the balance of payments, a framework for
estimating the flow of goods, services, capital, and other transfers
between the United States and other countries. BEA uses data from
Census, the Department of the Treasury, and other sources to produce
the balance of payments. Census compiles statistics on merchandise
trade, which are drawn from import and export documents collected by
the U.S. Customs Service. These statistics are used in a variety of
ways, ranging from guiding policymakers in trade negotiations and
assessments of international macroeconomic conditions to the
administration of trade agreements, trade programs, and trade damage
determinations.
Statistics on international trade and investment have not fully kept
pace with the changing patterns of international economic relations.
As a consequence, the possible undercounting of U.S. exports has
long been suspected. In addition, problems with import statistics
have also been reported by us and others.\14 While a number of steps
have been taken in recent years to improve the accuracy of
merchandise trade statistics,\15 problems remain. For example, in
1993, as Congress considered the North American Free Trade Agreement,
we testified that Census officials believed U.S. exports to Mexico
were undercounted.\16 Further, we subsequently reported\17 that
available data made it difficult to get a complete picture of
U.S.-Mexico trade, including, for example, how much of U.S. imports
from Mexico came from maquiladora manufacturing plants.\18
Statistics on services transactions lack adequate detail and
coverage, according to statistical agency officials and users of
trade statistics. While some improvements have been made in the
collection of such statistics, the surveys that produce these
statistics are still limited in frequency, level of detail, and
coverage. Statistics are collected for about 30 different categories
of services on the basis of surveys that vary in frequency, extent of
coverage, and exemption levels. In comparison, merchandise trade
statistics are based on a monthly count of roughly 14,000 separate
categories of imports and nearly 8,000 export categories. We
reported in 1994 that evaluating the economic impact of the services
portion of the General Agreement on Tariffs and Trade was hampered by
the lack of available and adequate data.\19 The Industry Sector
Advisory Committee on Services strongly recommended earlier in 1994
that U.S. services transactions statistics needed to be improved in
order to monitor the agreement.
Officials at BEA and the Department of the Treasury readily
acknowledged that similar to statistics on services transactions,
measures of overseas private investment assets do not adequately
reflect growth in this area. A 1992 National Research Council report
concluded that the measurement of international investment flows was
more likely prone to error than any other category of trade and
investment.\20 One factor contributing to this error was that BEA's
last benchmark survey of U.S. private investment in other countries,
from which annual investment flows are calculated, was done in
1943.\21 As a result, annual earnings accrued from overseas
investment may have been underestimated by as much as $6 to $14
billion in 1988.\22 Another factor contributing to measurement
problems is that BEA does not track holdings of U.S. currency in
other countries or new types of financial transactions, such as
financial swaps, which have grown substantially.
Problems in the measurement of international trade and investment
contribute to a sizable error, or discrepancy, in the U.S. overall
balance of payments that limits the accuracy of information available
to policymakers. In some years, the discrepancy was so large that it
obscured the nature and direction of international flows. For
example, in 1990, a large decrease in capital flowing to the United
States, including a net outflow of $5 billion in private capital,
contributed to concerns about the availability of capital in the
United States. However, the size of the 1990 statistical
discrepancy, an estimated $73 billion, made it difficult to determine
whether the availability of capital was actually diminished.
--------------------
\14 Among these, Measuring U.S.-Canada Trade: Shifting Trade Winds
May Threaten Recent Progress (GAO/GGD-94-4, Jan. 19, 1994); Customs
Service: Trade Enforcement Activities Impaired by Management
Problems (GAO/GGD-92-123, Sept. 24, 1992); Federal Statistics:
Merchandise Trade Statistics-Some Observations (GAO/OCE-89-1BR, Apr.
21, 1989); "Quality Issues Affecting the Compilation of the U.S.
Merchandise Trade Statistics," Bruce Walter (1989); and the National
Research Council's 1993 report, Behind the Numbers: U.S. Trade in
the World Economy.
\15 For example, Census now reconciles U.S. export data with some
other countries' import data, and Customs has automated the reporting
of U.S. import data in order to speed their collection.
\16 U.S. Trade Data: Limitations of U.S. Statistics on Trade With
Mexico (GAO/T-GGD-93-25, Apr. 28, 1993).
\17 U.S.-Mexico Trade: The Maquiladora Industry and U.S. Employment
(GAO/GGD-93-129, July 20, 1993).
\18 Mexico's maquiladora program allows Mexican and foreign investors
to establish manufacturing plants in selected areas of Mexico to
produce goods for export, exempting their imports from certain
customs duties.
\19 The General Agreement on Tariffs and Trade: Uruguay Round Final
Act Should Produce Overall U.S. Economic Gains (GAO/GGD-94-83b, July
1994), p. 118.
\20 Behind the Numbers: U.S. Trade in the World Economy, Committee
on National Statistics, Commission on Behavioral and Social Sciences
and Education, National Research Council. National Academy Press,
1992.
\21 According to OMB, an update of the foreign private portfolio
investment benchmark survey was approved on October 3, 1994.
According to BEA, it has received approval from OMB for a new survey
of financial services. Funding for both projects is uncertain.
\22 This estimate was made in 1990 by Federal Reserve economist Lois
Stekler.
INCOMPLETE OUTPUT AND
PRODUCTIVITY STATISTICS
HINDER POLICYMAKERS' ABILITY
TO UNDERSTAND ECONOMIC
CONDITIONS
---------------------------------------------------------- Letter :5.2
One long-term goal of monetary and fiscal policy is to raise the
living standards of the American people. In the long run, higher
standards of living largely depend on rising economic output\23 and
productivity.\24 However, according to statistical agency officials
and many users of economic statistics, such as economists, key
measures of economic output have not kept pace with changes in the
economy. The primary problem they cited was the limited measurement
of service sector output. This problem is particularly important
because service sector output is estimated by BEA to have grown from
38 percent of the gross domestic product (GDP) in 1960 to more than
half of the GDP today.
The growing importance of the service sector presents greater
measurement and data collection challenges to statistical agencies.
For example, steel production is easier to measure than the provision
of services. Steel output is measured in tons produced, while the
provision of services is measured less concretely, in terms of
improved quality and convenience. A bank that extends lobby hours or
provides automated teller services may not process more transactions.
However, by expanding these services, the bank improves the quality
and convenience of its banking services for its customers. In
addition, certain service industries, such as the restaurant
industry, have a large and changing universe of establishments
compared to the small, well- known number of car manufacturers, thus
making data collection more difficult.
Researchers also cited problems in accurately measuring productivity,
particularly for the service sector. Zvi Griliches, a Harvard
University professor and Director of Productivity Research for the
National Bureau of Economic Research, has noted that sectors of the
economy that are easier to measure, such as manufacturing and mining,
have shown greater productivity gains than service sectors such as
finance.\25 However, he concluded that the difference may stem as
much from difficulties in collecting appropriate statistics for
measuring service sector productivity as it does from actual
differences in productivity. For example, financial services output
is an "imputed value" of labor input (i.e., the growth in output is
not directly measured, and so the growth in labor inputs, such as
hours worked, is used as a substitute measure). Accordingly,
productivity measures for this sector should be interpreted with
caution.
Users of output and productivity data said that more reliable data on
output and productivity would be helpful in assessing policies to
enhance future economic growth and higher living standards. For
example, Federal Reserve Board staff said that more accurate measures
of service sector output data would improve their ability to assess
productivity gains and the reasons for them. Other users also noted
that improved data would aid researchers in determining the effect of
fiscal actions on output. This determination would provide a better
basis for evaluating policy decisions in areas such as government
sponsored research and development, training, and tax policy. For
example, in commenting on the administration's fiscal year 1995
budget, the House Committee on Science, Space and Technology noted
that "it is extremely difficult to measure the economic and social
benefits of (research and development) investments in dollar terms
with any precision." According to the Chief Economist for the
Committee, "better data would help policymakers in making choices
between different policy alternatives."
Similarly, in 1993, we reported that an investment focus within the
budget would provide a valuable supplement to the unified budget's
concentration on macroeconomic issues by directing attention to the
consequences of choices within the budget for long-term economic
growth.\26 However, without accurate statistics on output and
productivity, deciding among investment and consumption priorities
within the budget becomes much more difficult.
--------------------
\23 BEA's National Income and Product Accounts detail the level of
economic output in the United States. Calculation of economic output
is summarized by the gross domestic product (GDP), which is comprised
of consumer purchases of goods and services, investment spending,
exports and imports of goods and services, and government purchases.
\24 BLS issues several sets of productivity measures for sectors of
the U.S. economy, individual industries, and the federal government.
Labor productivity measures provide statistics on output, such as
automobiles, produced per hour. Multifactor productivity measures
output per combined unit of inputs, including labor and capital and,
for some measures, energy, materials, and purchased services.
\25 Productivity, R&D, and the Data Constraint, Presidential address
delivered at the 106th meeting of the American Economic Association,
Jan. 4, 1994, by Zvi Griliches, Professor, Department of Economics,
Harvard University.
\26 Budget Policy: Federal Capital Budgeting (GAO/T-AFMD-93-7, May
26, 1993).
AGENCY COMMENTS
------------------------------------------------------------ Letter :6
The Department of Commerce provided written comments on a draft of
this report. Commerce said it had no substantive comments, but
offered technical corrections, which we have made. Commerce
recognized that measurement problems can affect economic policy-
making, and said that the problems are being addressed by BEA and the
Census Bureau. On March 30, 1995, we met with the Chief Statistician
and an economist in OMB's Office of Information and Regulatory
Affairs, who generally agreed with the information presented in this
report. They suggested some technical corrections and clarifications
that we have incorporated. The Department of Labor offered oral
comments from BLS. On several occasions, senior BLS officials
suggested technical changes to the report concerning the CPI and
productivity statistics that we have incorporated where appropriate.
BLS officials noted that the BLS Commissioner's March 13, 1995,
testimony before the Senate Finance Committee (see p. 7 of this
report) provides an extensive description of the measurement issues
surrounding the CPI. They also said that the October 1992 Monthly
Labor Review includes a discussion on the limited impact of service
sector measurement procedures on the accuracy of the widely watched
business sector productivity measures.
---------------------------------------------------------- Letter :6.1
We are sending copies of this report to the Secretaries of Labor and
Commerce, the Director of the Office of Management and Budget, and
interested congressional committees. Copies also will be made
available to others upon request.
If you have any questions regarding this report, please call me on
(202) 512-8676. This report was prepared under the direction of
James McDermott, Assistant Director, Federal Management Issues.
Other major contributors to this report are listed in appendix I.
Sincerely yours,
L. Nye Stevens
Acting Director
Federal Management Issues
MAJOR CONTRIBUTORS TO THIS REPORT
=========================================================== Appendix I
GENERAL GOVERNMENT DIVISION
Anthony Hill, Senior Evaluator
Edward Laughlin, Senior Evaluator
SEATTLE FIELD OFFICE
Neil Asaba, Issue Area Manager
Paul Aussendorf, Evaluator-in-Charge
OFFICE OF THE CHIEF ECONOMIST
Loren Yager, Assistant Director
RELATED GAO PRODUCTS
=========================================================== Appendix 0
The General Agreement on Tariffs and Trade: Uruguay Round Final Act
Should Produce Overall U.S. Economic Gains (GAO/GGD-94-83b, July
1994).
Measuring U.S.-Canada Trade: Shifting Trade Winds May Threaten
Recent Progress (GAO/GGD-94-4, Jan. 19, 1994).
U.S.-Mexico Trade: The Maquiladora Industry and U.S. Employment
(GAO/GGD-93-129, July 20, 1993).
Budget Policy: Federal Capital Budgeting (GAO/T-AFMD-93-7, May 26,
1993).
U.S. Trade Data: Limitations of U.S. Statistics on Trade With
Mexico (GAO/T-GGD-93-25, Apr. 28, 1993).
Customs Service: Trade Enforcement Activities Impaired by Management
Problems (GAO/GGD-92-123, Sept. 24, 1992).
Federal Statistics: Merchandise Trade Statistics-Some Observations
(GAO/OCE-89-1BR, Apr. 21, 1989).
Developing a Consumer Price Index for the Elderly (GAO/T-GGD-87-22,
June 29, 1987).
Funds Needed To Develop CPI Quality Control System (GAO/GGD-83-32,
Apr. 1, 1983).
A CPI for Retirees Is Not Needed Now but Could Be in the Future
(GAO/GGD-82-41, June 1, 1982).
A Consumer Price Index for Retirees and Alternatives for Controlling
Indexing (Testimony, Apr. 20, 1982).
Measurement of Homeowner Costs in the Consumer Price Index Should Be
Changed (PAD-81-12, Apr. 16, 1981).
Alternatives for Modifying the Indexation of Federal Programs
(Testimony, Mar. 10, 1981).