View original document

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

FRBSF

WEEKLY LETTER

Number 92-07, February 14, 1992

Services: A Future of
Low Productivity Growth?
The question is so common lately, it has become
trite: "Is America turning into a nation of hamburger flippers?" The cause for concern is a
potentially disturbing pair of trends. First, employment in the United States increasingly has
shifted away from manufacturing toward services.
In 1963 manufacturing accounted for 30 percent
of all jobs, but in 1991 that share had fallen to
less than 17 percent.
Second, "real" productivity increases (that is, adjusted for inflation) have been significantly lower
in services than in manufacturing. Between 1963
and 1986, real output per worker in manufacturing rose at an annual rate of 2.6 percent, while
real output per worker in the services sector rose
by only 0.2 percent. Taken together, these trends
suggest dire consequences: a sector with negligible productivity gains is rapidly becoming the
most important source of new jobs, while the
more productive manufacturing sector is losing
employment. The inference many observers draw
is that America faces stagnation, with most of its
labor force engaged in employment that exhibits
little productivity growth, and hence, provides
little increase in its standard of living.
This Letter takes a closer look at the data. After
combining the results of other researchers with
evidence from data on compensation, it appears
that the fundamental data on which these dire
predictions are made may be misleading. While
the evidence is not conclusive one way or the
other, data on compensation at least raise the
possibility that the opposite is true: that labor
productivity is growing faster in services than it
is in manufacturing.
The source of the "real" evidence
Concern about lack of productivity growth in the
service sector emerges from the data released by
the u.s. Department of Commerce's Bureau of
Economic Analysis (BEA) on nominal output, real
output (that is, adjusted for inflation), and employment for each industry. Using these data, it is
possible to compute "real" output per employee.

The difficulty facing BEA in determining changes
in real output per worker is daunting. In many
cases, outputs are not directly priced, and often
it is difficult or impossible to measure the quantity of the output. This is particularly true in the
service sector. For example, bank customers re- .
ceive services from tellers and loan officers, but
they do not pay directly for those services; instead they pay indirectly with interest rate spreads
and fixed service charges. Similarly, service industries producing information (such as research)
generate a product that is consumed by the pubiic, but the product is often not priced and there
is no way to count how many people benefit
from that information.
Adding to the complexity is the problem of quality changes. With the exception of a few agricultural products, nearly all products change over
time, with most embedding improved features.
For example, a 1992 car cannot be compared
directly with a 1972 car. We can count the number
of cars sold and observe the prices charged, but
counting the change in output-the stream of benefits derived from the car-will not be correct
unless it is possible to standardize the car with
earlier models, to distinguish and count improvements in dimensions of quality, safety, and emissions as well as the number of units produced.
BEA's methodology
Because of limitations on data collection, BEA
must extrapolate from available information on
prices and production to calculate price or quantity indexes for each industry. In the case of most
manufacturing industries, BEA has information
available about input costs, prices of some final
products, and output indexes. Although the problem of quality remains, and not all final product
prices are recorded, most manufacturing industries produce a physical output that is countable,
which makes the process less abstract.

BEA faces the same conceptual problems in services that it does in manufacturing, but fewer
data typically are available for services. Quantity

FABSF
indexes usually are not available-in fact, there
often is no physical "good" to count-and
prices also are not observed directly. To derive
"real output;' therefore, BEA must extrapolate
using indirect measures of input costs as proxies
for price changes. In other industries, such as
banking and recreation, measures of inputs are
used-often the number of employees-to proxy
for changes in the level of production.
While this approach is perhaps the only available
strategy in some industries, the effects on productivity measurements are predictably biased.
When real output is calculated for industries using labor quantities as an important measure of
output, then by definition the industry will show
no productivity growth. Moreover, price indexes
based on input costs ignore potential quality improvements-such as higher skilled labor and
better capital-and hence, may attribute quality
improvements inaccurately to price increases
rather than output increases.

wage differentials is related to observable labor
quality factors-experience and education, for
example-and other differentials are attributed
to unobservable quality differences.
Compensation growth per worker in an industry,
therefore, should be related to the growth in the
productivity of its labor force. If workers are paid
the value of their marginal product, the growth in
compensation per worker in an industrv reflects
the growth in the average marginal product of
that industry's labor force.
The accompanying chart compares real compensation growth per worker (the BEA's compensation
series deflated by the consumer price index) to
BEA's real output measure for manufacturing and
several service-producing industries. In many
cases, there are wide disparities in the relative
growth rates for compensation and output.

Real Compensation and Output per Worker, Selected Industries
(Average Annual Percent Growth: 1963-86)

Evidence from compensation data
Several researchers have looked at the validity
of BEA's inter-industry productivity comparisons.
Denison (1989) argues that BEA's approach could
not accurately distribute output gains to industries when intermediate products (that is, products developed by one manufacturer and used by
another for producing final consumer products)
are involved. Moreover, Baily and Gordon (1988)
used a case study analysis and questioned BEA's
measurement of output gains attributed to each
industry, suggesting a bias toward manufacturing.
This bias also was noted by Smith (1972), who
found that the data consistently pushed measured
productivity gains toward the manufacturing sector and away from the service sector.
Unfortunately, these criticisms typically are qualitative rather than quantitative, making it difficult
to reject BEA's conclusions. It is possible, however, to provide a quantitative check of the BEA
data byusing data on compensation growth by
industry.
One of the central conclusions of economic theory is that wages are determined by the marginal
value of labor's contribution to that output. Thus,
changes in compensation should be related to
changes in labor's marginal productivity. Research on inter-industrywage differentials offers
some support to this proposition. Empirical research has found that a large portion of these

• Compensation/Worker
iiiiiiI Output/Worker

-4

-3

-2

] All Industries
~ Manufacturing
Services
Hotels
Personal Services
Business Services
Health
Legal
F.I.R.E.
Banking

-1

Consider the comparison between services and
manufacturing. In contrast to the BEA real output
data, compensation data indicate higher gains in
services than in manufacturing. Service industry
compensation per worker rose at a 1.5 percent
annual rate, compared to 1.1 percent for manufacturing. In fact, manufacturing reported compensation gains just slightly above the average
for all industries.
Other important differences are worth noting.
Banking is found in the output data to have virtually no productivity growth, yet compensation
growth matches the average for all industries.
Moreover, output growth per worker in the health
care and legal industries is reported to have been
negative according to the real output series while
recording higher increases in compensation
growth than are found in any other industries.

Which measure is right?
Compensation is not a direct measure of productivity either, however. Economic theory shows
that changes in compensation over time can be

split roughly into two parts: changes in the relative output price of the final good and changes
in the marginal productivity of the labor used to
make the good.
If output prices rise for a particular product, that
may feed into compensation to workers in the
form of raises and bonuses. Over time, though,
the ability of labor to capture increases in output
prices is determined by labor's ability to restrict
the entry of new laborers. If higher wages are not
the result of higher marginal productivity by the
firm's labor force, those higher wages will attract
new workers and bid down those wages unless
the existing labor force can effectively restrict
entry through unions or Iicensing restrictions.

measured real productivity gains may signal
measurement errors. Particularly in the case of
service industries, where output and prices are
nearly impossible to observe directly, compensation may offer an alternative measure of productivity gains. While it is clear that compensation
growth is not a perfect measure of productivity
growth, trends in compensation raise an important question: If productivity growth is so low in
some sectors, why are employers willing to pay
the workers so much? If market participants are
relatively rational, the payment workers receive
may be a better indicator of the market's evaluation of their productivity than is the traditional
BEA measure.

Conclusions
The second source of compensation increases,
linked to productivity gains, would appear where
the firm has raised the quality of its labor force
over time. Hiring larger proportions of skilled
labor would raise the relative compensation
growth of that industry.
Determining whether compensation increases
are due to productivity gains or to output price
gains, therefore, can be problematic. Nevertheless, one factor seems to be useful in distinguishing between the two effects: labor mobility. In
some of the more heavily unionized and licensed
industries (manufacturing, transportation, legal,
and health care), compensation increases may
be less useful as a measure of productivity gains.
In other industries, however, where employment
growth has been rapid, it is likely that compensation gains do reflect the market's evaluation of
the relative productivity of those workers.
Moreover, in industries where capital/labor ratios
have changed dramatically, compensation may
provide a superior measurement of changes in
labor productivity. BEA's simple measure of labor
productivity growth-output per worker-is not
adjusted for total productivity gains that are the
result of increased productivity of non-labor inputs. Thus, in industries that have become more
capital intensive, gains in average output per
worker may reflect increasing productivity of
otherfactors, not just labor. Compensation, which
is a payment to labor for its contribution to total
productivity gains, provides the market's assessment of the value produced by those workers.
For many industries, therefore, differences
between the compensation growth rates and

Concern about the observed shift of employment
toward service industries often is based on the
assumption that services have lower productivity
growth than manufacturing. The BEA real output
data are used to support this proposition. As
shown in this Letter, however, data on real compensation per worker provide conflicting evidence, at least suggesti ng the possi bil ity that
services have had faster productivity increases
than manufacturing.
The compensation data are not a perfect measure
of productivity growth either, but the fact that the
data differ so dramatically from the output data
is disturbing. At a minimum, the compensation
data, along with results from other researchers
looking at particular industries, suggest that the
inter-industry productivity comparisons made
with the "real" output data are biased and potentially misleading.

Ronald H. Schmidt
Senior Economist

References
Baily, Martin N., and Robert J. Gordon. 1988. "The
Productivity Slowdown, Measurement Issues, and
the Explosion of Computer Power:' Brookings
Papers on Economic Activity, vol. 2, pp. 347-420.
Denison, Edward F. 1989. Estimates of Productivity
Change by Industry: An Evaluation and an Alternative. Washington, DC: The Brookings Institute.
Smith, A.D. 1972. The Measurement and Interpretation of Service Output Changes. Washington, DC:
National Economic Development Office.

Opinions expressed in this newsletter do not necessarily reflect the views of the management of the Federal Reserve Bank of
San Francisco, or of the Board of Governors of the Federal Reserve System.
Editorial comments may be addressed to the editor or to the author.... Free copies of Federal Reserve publications can be
obtained from the Public Information Department, Federal Reserve Bank of San Francisco, P.O. Box 7702, San Francisco 94120.
Phone (415) 974-2246, Fax (415) 974-3341.
Printed on recycled paper Q
~.
with soybean inks.

\:tI ~

OU176 V:l 'oJspueJJ ueS
WLL xog ·O·d

O)SI)UOJ~

JO

UOS

~uo8

aAJaSa~ IOJapa~

~uew~Jodea 4)Joese~

Index to Recent Issues of FRBSF Weekly Letter

DATE NUMBER

TITLE

AUTHOR

8130
9/6

91-29
91-30

Public Preferences and Inflation
Bank Branching and Portfolio Diversification

9/13
9/20
9/27
10/4
10111
10118
10/25
1111
11 /8
11115
11 /22
11/29
12113
12/20
1/3
1110
1117
1/24
1/31
2/7

91-31
91-32
91-33
91-34
91-35
91-36
91-37
91-38
91-39
91-40
91-41
91-42
91-43
91-44
92-01
92-02
92-03
92-04
92-05
92-06

The Gulf War and the
Economy
The Negative Effects of Lender Liability
M2 and the Business Cycle
International Output Comparisons
Is Banking Really Prone to Panics?
Deposit Insurance: Recapitalize or Reform?
Earnings Plummet at Western Banks
Bank Stock Risk and Return
The False Hope of the Narrow Bank
The Regional Concentration of Recessions
Real Wages in the 1980s
Solving the Mystery of High Credit Card Rates
The Independence of Central Banks
Taxpayer Risk in Mortgage Policy
The Problem of Weak Credit Markets
Risk-Based Capital Standards and Bank Portfolios
Investment Decisions in a Water Market
Red Ink
Presidential Popularity, Presidential Policies
Progress in Retail Payments

u.s.

Walsh
Laderman/Schmidt/
Zimmerman
Throop
Hermalin
Furlong/Judd
Glick
Pozdena
Levonian
Zimmerman
Neuberger
Pozdena
Cromwell
Trehan
Pozdena
Kim
Martin/Pozdena
Parry
Neuberger
Schmidt/Cannon
Zimmerman
Walsh/Newman
Laderman

The' FRBSF Weekly Letter appears on an abbreviated schedule in June, July, August, and December.