View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

FEDERAL RESERVE BANK OF NEW YORK

I N E C O N O M I C S

A N D F I N A N C E

June 1999

Volume 5 Number 9

The Impact of Reduced Inflation Estimates on Real Output
and Productivity Growth
Charles Steindel

Despite posting their strongest sustained performance in many years, recent measures of
output and productivity growth have still fallen short of their 1960-73 averages. Could
data-measurement problems affecting the pricing of some services account for the inability
of these widely tracked U.S. growth indexes to match their earlier rates?
Two key measures of U.S. economic growth—the
annual rate of change in real output and in productivity—have improved considerably in recent years. 1
Between 1996 and 1998, growth in productivity in the
largest segment of the economy, the private nonfarm
business sector, averaged nearly 2 percent. This growth
rate, the strongest sustained pace since 1983-85,
appears to signal some fundamental improvement in
the economy. Nevertheless, the growth trend for productivity has remained well below its 1960-73 average
of nearly 3 percent (Chart 1).
Measurement difficulties provide one possible
explanation for the failure of current growth rates to
match those of the past. For instance, the aggregate
data used to calculate GDP and productivity rates are
based on samples drawn from surveys whose representation of the economy may, of necessity, be somewhat
out of date. In today’s rapidly changing economy,
reliance on such data may lead to overly conservative
estimates of growth.
The possibility that measurement problems are contributing to the shortfall in the growth numbers appears
stronger when we consider the ongoing concerns about
the measurement of inflation. Many economists and
policymakers believe that the data used in the construction of the consumer price index (CPI) overstate prices.
These concerns carry over to economic growth because,
all else equal, an increase in reported inflation would

decrease the reported rate of real output and productivity
growth. Furthermore, an overstatement of inflation
may be especially important in the services sector,
whose contribution to the economy has grown steadily
over the years. Not only are many services—such as
medical care and financial services—inherently hard
to price, but the nature of some of them has changed
radically with the introduction of new technologies.
These factors could be making it more difficult for
government data to capture the true prices of services.

Chart 1

Productivity Growth in the Nonfarm Business Sector
Annual percentage change
6
1960-73 average:
5
2.9 percent
1974-98 average:
1.1 percent

4
3
2
1
0
-1
-2
1960

65

70

75

80

85

Source: U.S. Department of Labor, Bureau of Labor Statistics.

90

95

98

CURRENT ISSUES IN ECONOMICS AND FINANCE

Another key factor to consider when examining
inflation and growth indexes is the change in the bias in
inflation. Even if the current data on real output and
productivity are being substantially underestimated
because of overstatements of price increases, one cannot
assume that the true growth rates of GDP and productivity today are closer to the rates of the past. For such
an assumption to be correct, the current overstatement
of inflation would have to be larger than it was in the
earlier period. For instance, if inflation throughout the
economy (not just inflation reported in the CPI) had
always been overstated by 1 percentage point, we could
reasonably say that recent productivity growth has been
around 3 percent, rather than the reported average of a
little less than 2 percent. Based on this example, however, past productivity growth would actually have been
higher than has been reported—around 4 percent in the
1960s. Thus, an analysis of the effect of inflation mismeasurement on the historical record of output and
productivity growth rates should focus less on the level
of the bias in inflation at some period in time and more
on the change in the bias over time.

This edition of Current Issues examines whether
measurement problems can account for the inability of
real GDP and productivity growth rates to rival the
growth rates of 1960-73. Specifically, we consider technical problems associated with the pricing of certain
hard-to-measure services. An overstatement of inflation
in these services may be resulting in an understatement
of real GDP and productivity growth.
We find that hard-to-measure services have accounted for a greater share of GDP over the years and that
inflation has indeed been higher in these services than
in other sectors of the economy. Moreover, reduced
estimates of the inflation rate in these hard-to-measure
services are found to raise recent output and productivity growth rates. However, despite these findings, the
estimates also reveal that these growth rates would still
have been much lower in recent years than in the 196073 period—even if an especially strong upward bias in
service-sector inflation estimates had begun only after
1973. Therefore, it is difficult to attribute the inability
of recent output and productivity growth rates to regain
their 1960-73 pace solely to the inexact pricing of hardto-measure services.

Changes in the Bias in Inflation
Several studies have investigated the level of the bias in
the price data, yet relatively few have examined
changes in the bias over time—perhaps because of the
difficulty of making a good estimate of the bias in any
period. However, it is plausible to think that, for many
products, the inflation bias has not increased systematically over time. Procedures for estimating the price of
goods have undergone many changes, in part as a result
of efforts to address problems identified in price studies.3 In general, when corrections to the price data are
made and the real output data are adjusted, there is
an attempt to correct as much of the historical data as
possible. On the whole, then, it seems unlikely that the
change in the bias in goods pricing in recent years is
appreciably higher than it was in the past.4

Inflation Measurement Problems and the Calculation
of Economic Growth
Inaccurate inflation measures could affect the calculation of economic growth indexes because many index
computations rely on the rate of inflation. For example,
real output growth is defined as nominal spending
growth less inflation; productivity growth is computed
by the Bureau of Labor Statistics (BLS) as nonfarm business real output growth less the growth of worker-hours.
Concern over the accuracy of inflation measurements came to the fore in 1996. In its report to the
Senate Finance Committee late that year, the Boskin
Commission—an advisory group of scholars formed to
examine the CPI—estimated that the index overstated
growth in the cost of living by about 1 percentage point
a year (Boskin et al. 1996). In the period since the
report’s release, the BLS has made extensive changes to
its construction of the CPI.

For some services, changes in the bias in inflation
have probably been minor. In transportation and communications, for instance, one can measure output and
prices by observing the cost of standardized products,
such as passenger miles or messages, respectively.
Technological changes in recent years may have complicated the job of measuring real output and prices for
these products, but corresponding changes have been
made to some of the procedures used to compute the
prices. Overall, there is little evidence that the bias in
pricing these services has increased.5

However, the Boskin Commission’s findings shed
little light on the impact of overstated inflation on
reported economic growth. For example, the commission’s estimate of a 1-percentage-point upward bias in
the CPI does not necessarily suggest that growth in
either real output or productivity is being understated
by that amount. The reason is that the price data used to
gauge real output can be quite different from the data
used to measure the CPI.2 Most notably, the price data
used to calculate real output and productivity are not
subject to a number of the problems that have affected
the computation of individual prices in the CPI.
FRBNY

However, the pricing of other services—such as
medical care and financial services—is a more complicated matter. It is very hard to standardize, and therefore price, products in these sectors. Moreover, the
2

extraordinary pace of technological change in these
sectors may have exacerbated pricing problems,
despite the best efforts of government statisticians to
address the changes. This argument is supported by the
observation that the published productivity data for
much of the services sector continue to be disappointingly low.6 It is these hard-to-measure services that we
now examine.

Chart 2

Hard-to-Measure Services as a Share of GDP
Percent
25

20

15

The Output Share of Hard-to-Measure Services
Could high and increasing inflation bias in some rapidly
changing services sectors help explain why the reported
growth rates for real output and productivity remain
below the pre-1973 rates? That is, has this growth
actually been higher in recent years than has been
officially reported? To answer these questions, we need
to determine what share of the economy is attributable
to services that may have major pricing problems.

10

5
1960

65

70

75

80

85

90

95 97

Source: Author’s calculations, based on data from the U.S. Department of
Commerce, Bureau of Economic Analysis.

Table 1 provides a breakdown of current-dollar
aggregate spending in 1997 (indented items sum to the
heading above them). That year, services spending
accounted for about 54 percent of GDP. A large portion
of this spending fell into three categories: government
compensation of employees, depreciation of government capital, and space rent (the imputed services provided by the housing stock). However, we can discount
the services in these categories because they are not
produced by the private nonfarm business sector; errors

in their pricing therefore will not affect the productivity
data for that sector. This leaves only 36 percent of 1997
GDP accounted for by actual purchases of services
from the private sector, identified in the table as
“other.” A significant portion of these services (“all
other services”) consisted of items such as utilities and
transportation; we can discount these services too
because, as suggested earlier, their prices are probably
not being measured any less accurately today than they
were in the past.
This leaves us with a “hard-to-measure components”
category, which contains services that account for
roughly 23 percent of GDP. 7 As noted earlier, one could
raise significant questions about the reliability of the
price data for these services, such as financial and
business services, medical care, and educational and
charitable expenses.8 These hard-to-measure services
have accounted for a steadily increasing share of currentdollar GDP over the years (Chart 2). Significantly, the
reported price data suggest that inflation has typically
been much higher in these services than in the rest of
the economy (Chart 3). Of course, the higher inflation
rates could simply reflect a rapidly growing demand for
these services. However, it is also possible to interpret
the higher rates as evidence that a stronger upward bias
exists in the reported prices of these services than in
prices elsewhere in the economy.

Table 1

1997 Composition of GDP
Amount
Percentage
Sector
(Billions of Dollars) of GDP
Goods
2,978.5
36.7
Structures
718.3
8.9
Services
4,414.1
54.4
Government compensation of employees
780.2
9.6
Depreciation of government capital
128.3
1.6
Space rent
590.3
7.3
Other
2,915.3
35.9
Hard-to-measure components
1,875.2
23.1
Consumer
1,589.5
19.6
Medical care
843.4
10.4
Personal business
459.1
5.7
Educational
129.4
1.6
Religious and welfare
157.6
1.9
Government
179.0
2.2
Net exports
106.7
1.3
All other services
1,040.1
12.8
Total
8,110.9
100.0

Alternative Growth Estimates
Although the higher inflation rates for hard-to-measure
services are suggestive of a pricing bias, we cannot
attribute the decline in average output and productivity
rates to problems in pricing unless we can establish
either that the problems are large relative to those of the
past, or that these services are now such a large part of

Source: U.S. Department of Commerce, Bureau of Economic Analysis.
Note: Indented items sum to the heading above them.

3

CURRENT ISSUES IN ECONOMICS AND FINANCE

tions about the start date of additional inflation bias
in the price data for these selected services. We use
1974 because it corresponds to the start of the slower
productivity era in the official data; 1983 and 1992 are
starting dates for the last two economic expansions; for
1960-97, data availability determined the start date and
the very preliminary nature of the 1998 data dictated
the end date.

Chart 3

Sectoral Inflation Rates
Percent
12

Hard-to-measure services

10
8
6

Our estimates reveal that, in general, a reduction in
the rate of inflation in the rapidly growing, hard-tomeasure services categories does raise recent growth
rates. We find that both GDP and productivity growth
since 1974 would have been, on average, about 0.5 percentage point higher if the inflation rates in these categories had been scaled back in line with our alternative
estimates.10 These are nontrivial adjustments.

4
2
0
1960

Other components of GDP
65

70

75

80

85

90

95 97

Source: Author’s calculations, based on data from the U.S. Department of
Commerce, Bureau of Economic Analysis.

Consider in particular how productivity growth
would be affected if we assume that prices in the hardto-measure services component grew at the same rate
as other elements of nonfarm business output. In this
case, annual productivity growth since 1983 would
have averaged more than 2 percent, instead of less than
1½ percent. However, even this rather substantial change
does not radically alter our view of the long-term
dynamics of the economy. Even after the major upward
revision to the post-1983 data, we would still find that
the recent pace of productivity growth was considerably
slower than the nearly 3 percent pace reported before
1974 in the published data.

the economy that continuing price measurement problems will have very substantial effects on the aggregate
data. As we keep these points in mind, our next step is
to determine the effect of reducing the inflation biases
(assuming they exist) on the time profile of real growth.
To do this, we calculate the effects of alternative,
reduced estimates of inflation in the hard-to-measure
services sectors on the history of real GDP and productivity growth (Table 2). Our calculations rely on two
alternative assumptions: (1) inflation in these sectors
has in reality been equal to inflation in the rest of the
economy; (2) inflation in the hard-to-measure sectors
has been uniformly overestimated by 2 percentage
points a year. The assumptions are roughly comparable,
because the reported price increases for these hard-tomeasure services have been, on average, about 2 percentage points higher than price increases for other
goods and services.9

Overall, our estimates reveal that both real GDP and
productivity growth would have remained much lower in
recent years than in 1960-73—even if the differential
overstatement of inflation in these services had started

Table 2

How Reduced Inflation Rates in Hard-to-Measure
Services Would Affect GDP and Productivity Growth

We should note that ours are not formal estimates of
the amount of bias in the price data for these services.
We are merely using alternative assumptions about
inflation rates to approximate the additional real growth
that would be reported by reducing (but not necessarily
eliminating) the bias in these prices by a large, but perhaps plausible, amount. Certainly, there may be systematic upward bias in prices throughout the economy—
growth may always have been significantly greater than
the published data indicate. Nevertheless, our estimates
aim to gauge the impact on reported real growth of an
assumed amount of additional upward bias in a particular set of prices.

Inflation Rate
Reported rate
Assumption 1a
Assumption 2b
Reported rate
Assumption 1a
Assumption 2b

1960-97 1960-73 1974-97 1983-97 1992-97
GDP Growth (Percent)
3.2
4.2
2.6
3.1
3.0
3.5
4.3
3.0
3.5
3.4
3.5
4.4
3.0
3.5
3.5
1.8
2.2
2.2

Productivity Growth (Percent)
2.9
1.1
1.3
3.2
1.7
2.1
3.3
1.6
1.7

1.3
2.1
1.4

Sources: U.S. Department of Commerce, Bureau of Economic Analysis;
U.S. Department of Labor, Bureau of Labor Statistics; author’s calculations.

We offer alternative estimates of growth for several
periods: 1960-97, 1960-73, 1974-97, 1983-97, and
1992-97. This variety allows us to measure the effect on
the historical real growth record of different assump-

a Prices

of hard-to-measure services grow at the same rate as other products.

bPrices

of hard-to-measure services grow at a rate that is 2 percentage points
slower than the published rate.

4

FRBNY

method of pricing is designed to capture the value of various product components to the user. More recently, techniques for pricing
individual food items were enhanced. For an overview of changes in
pricing procedures and their impact on the historic record of GDP
growth, see Council of Economic Advisers (1998).

after 1973. (If the overstatement had begun before 1973,
the earlier growth rates would have been higher, with the
result that the inflation adjustments would produce even
less narrowing of the gap.) As we suggested earlier, one
cannot assume that continued low aggregate output and
productivity growth relative to 1960-73 is an artifact of
problems in pricing certain services unless the problems
are believed to be very large relative to those of the
past.11 Of course, this conclusion applies only up to the
present: As the share of these services in nominal GDP
increases, any mismeasurement of their price inflation
will have an increased impact on the aggregates.

4. A number of authors, including Nakamura (1998) and Reinsdorf
(1998), have pointed out that revisions made in the late 1970s to the
construction of the CPI appear to have resulted in an upward bias of
about 1.5 percentage points a year in the food at home component
from 1978 to the mid-1990s. However, one could argue that the
longstanding practice in CPI computation of removing the cost of
mandated pollution control equipment in motor vehicles and seasonal pollution control additives in gasoline from the prices of these
products has, all else equal, helped introduce a downward bias in
the corresponding CPI components. Since January 1999, these
costs have been included in the CPI.

Conclusion
Growth rates for real output and productivity have
improved in recent years, and it is possible that the
improvements will continue. However, we find it hard
to attribute the inability of this growth through most of
the 1990s to regain its 1960-73 pace solely to difficulties
in the pricing of various hard-to-measure services—
unless the pricing problems have widened to such a
magnitude that the relative price of these services is
now, in reality, declining. If problems in the price data
in the rest of the economy are no worse than they were
in the past, we conclude that it is difficult to blame
technical problems in the computation of inflation for
the continued failure of real output and productivity to
display growth figures like those of the 1960s.

5. For example, new procedures for the pricing of cellular phone
services were incorporated into the National Income and Product
Accounts in 1998; the new price series has been used to modify the
real output of the communications industry back to 1995 (Seskin
1998). A case can therefore be made that the bias in communications price data is smaller from 1995 on than it was before 1995.
6. Corrado and Slifman (1999), for example, point to the oddly slow
growth of productivity in the noncorporate sector of the economy.
We should note that although there are many services in this sector,
the sector is by no means representative of services as a whole, nor
is it clear that all of its data problems are related to pricing.
7. This list of of hard-to-measure services is, of course, arbitrary,
although Griliches (1994, note 15) produces a list that is virtually
the same as ours. However, Griliches includes housing, which does
not enter into the productivity data. Kroch (1991), who also discusses the relative difficulty of measuring the prices of different
types of services, provides a similar list of spending categories with
more significant measurement problems.

Our results suggest that improvements in the U.S.
statistical system—with an emphasis on improved pricing of services—may not involve a radically different
profile for aggregate growth, unless we believe that the
improvements will raise the recent growth of nominal
output relative to the past. 12 Rather than resulting in a
fundamentally changed path for the real growth aggregates, an improved statistical system could well pay
dividends in the form of a new understanding of the
industrial sources of U.S. growth—a recognition, perhaps, that the share of real output and productivity
growth produced by the services sector and other rapidly
growing industries is greater than previously thought. It
seems much less likely that we would see a major reconfiguration of the time path of the growth aggregates.

8. Financial services are not specifically listed, but they are
included in the personal business, government, and net export categories. Note that spending on financial services enters directly into
GDP when it is made by households or governments or as part of
foreign trade.
9. The Boskin Commission provides some evidence to support our
assumptions. For example, it estimated that medical care inflation in
the CPI was overstated by 3 percentage points a year and personal
financial services inflation was overstated by 2 percentage points a
year. The commission did not supply a formal estimate of the bias in
the other hard-to-measure categories in Table 1. Although the commission’s point estimates are taken from the middle region of
ranges, the assumption that 2 percentage points is an upper bound
on the differential bias in the hard-to-measure services in the
National Income and Product Accounts data is plausible, given the
commission’s finding that the probable bias in the CPI from individual pricing problems was less than 1 percent (other parts of the CPI
bias come from the index’s aggregation procedures, which are not at
issue here), and recent changes in the pricing of medical care in the
National Income and Product Accounts data. Berndt et al. (1998)
provide a recent comprehensive review of the measurement of medical prices, but they do not estimate the overall bias in this sector.

Notes
1. Real output, or GDP, is measured by the U.S. Commerce
Department’s Bureau of Economic Analysis; productivity, or output
per worker-hour, is measured by the U.S. Labor Department’s
Bureau of Labor Statistics.
2. Steindel (1997a) compares the CPI and output price indexes.
3. One of the most noteworthy changes, in the 1980s, was the
introduction of hedonic pricing of computers in the GDP data. This

5

FRBNY

CURRENT ISSUES IN ECONOMICS AND FINANCE

10. The availability of 1997 data and revisions to 1995-96 data
resulted in slightly higher estimates of the impact of alternative
inflation assumptions in recent years than in Steindel (1997b).
The productivity adjustment is almost surely exaggerated,
since it was made assuming that all outlays in the hard-to-measure
categories were produced in the nonfarm business sector. Many
medical, educational, religious, and welfare services are produced
in the nonprofit sector of the economy.
11. Sichel (1997), Carlson and Schweitzer (1998), and Triplett
(1999) reach similar conclusions.
The fairly substantial difference between the two alternative
assumptions in the last two columns of the productivity growth half
of Table 2 reflects the combined effect of two factors: the very large
share of nonfarm output credited to hard-to-measure services (now
more than 40 percent) in recent years and an inflation divergence
between hard-to-measure services and the rest of nonfarm output
averaging more than 2.5 percentage points a year in the recent
periods.
12. It is possible that corrections to measured inflation—and the
subsequent changes to reported real growth—cannot be made to the
entire historical record. If such changes affect only part of the
record, the more recent growth figures could look substantially
stronger relative to those in the more distant past—but this would in
part reflect a break in the consistency of the data over time.
The discussions in the Boskin report (Boskin et al. 1996) and
Griliches (1994) of the possibility of including quality-of-life measures in price and output measures may be viewed as outlines of
ways to redefine nominal and real output, and could potentially
change the output measures’ growth paths.

References
Berndt, Ernst R., David M. Cutler, Richard G. Frank, Zvi Griliches,
Joseph P. Newhouse, and Jack E. Triplett. 1998. “Price Indexes
for Medical Care Goods and Services: An Overview of
Measurement Issues.” NBER Working Paper no. 6817,
November.
Boskin, Michael J., Ellen R. Dulberger, Robert J. Gordon,
Zvi Griliches, and Dale Jorgenson. 1996. Toward a More
Accurate Measure of the Cost of Living. Final Report to the
Senate Finance Committee from the Advisory Commission to
Study the Consumer Price Index. December.

Carlson, John B., and Mark E. Schweitzer. 1998. “Productivity
Measures and the ‘New Economy.’” Federal Reserve Bank of
Cleveland Economic Commentary, June.
Corrado, Carol, and Lawrence Slifman. 1999. “Decomposition of
Productivity and Unit Costs.” American Economic Review 89,
no. 2 (May): 328-32. Papers and Proceedings of the 111th
Annual Meeting of the American Economic Association,
January 1999.
Council of Economic Advisers. 1998. Economic Report of the
President. Washington, D.C., February.
Griliches, Zvi. 1994. “Productivity, R&D, and the Data Constraint.”
American Economic Review 84, no. 1: 1-23.
Kroch, Eugene. 1991. “In Brief: Tracking Inflation in the Service
Sector.” Federal Reserve Bank of New York Quarterly Review
16, no. 2: 30-5.
Motley, Brian. 1992. “Index Numbers and the Measurement of Real
GDP.” Federal Reserve Bank of San Francisco Economic
Review, no. 1: 3-13.
Nakamura, Leonard I. 1998. “The Measurement of Retail Output
and the Retail Revolution.” Federal Reserve Bank of
Philadelphia Working Paper no. 98-5, March.
Reinsdorf, Marshall B. 1998. “Formula Bias and Within-Stratum
Substitution Bias in the U.S. CPI.” Review of Economics and
Statistics 80, no. 2: 175-87.
Seskin, Eugene P. 1998. “Annual Revision of the National Income
and Product Accounts.” Survey of Current Business 78
(August): 7-32.
Sichel, Daniel E. 1997. “The Productivity Slowdown: Is a Growing
Unmeasurable Sector the Culprit?” Review of Economics and
Statistics 79 (August): 367-70.
Steindel, Charles. 1997a. “Are There Good Alternatives to the
CPI?” Federal Reserve Bank of New York Current Issues in
Economics and Finance 3, no. 6.
———. 1997b. “Measuring Economic Activity and Economic
Welfare: What Are We Missing?” Federal Reserve Bank of
New York Research Paper no. 9732, October.
Triplett, Jack E. 1999. “Economic Statistics, the New Economy,
and the Productivity Slowdown.” Journal of the National
Association for Business Economics 34, no. 2 (April): 13-7.

About the Author
Charles Steindel is a senior vice president in the Business Conditions Function of the Research and Market
Analysis Group.

The views expressed in this article are those of the author and do not necessarily reflect the position of
the Federal Reserve Bank of New York or the Federal Reserve System.

Current Issues in Economics and Finance is published by the Research and Market Analysis Group of the Federal
Reserve Bank of New York. Dorothy Meadow Sobol is the editor.