Productivity: Will the Faster Growth Rate Continue?

Productivity: Will the Faster Growth
Rate Continue?
Updated October 5, 2007
Brian W. Cashell
Specialist in Macroeconomic Policy
Government and Finance Division

Productivity: Will the Faster Growth Rate Continue?
Summary
While policymakers have at least some direct or indirect influence over many
economic variables, productivity growth may be among those that remain relatively
removed from the influence of deliberate economic policy. Although many policy
proposals are advocated on the grounds that they will help boost productivity, it may
be that productivity growth rates have a greater influence on policy than policy does
on the growth of productivity. It seems that variations in the productivity growth rate
are so incompletely understood that there is no clear consensus among economists
about the best way for policymakers to promote it.
Since the mid 1990s productivity growth appears to have accelerated. This is
unusual in a mature economic expansion, which has suggested to more than a few
observers that it was not just a short-term phenomenon, but rather a sign that there
was an increase in the long-term economic growth rate.
An important question for policymakers is how long this surge in productivity
growth will continue. Higher productivity growth means higher real incomes, which
in combination with progressive income tax rates yields higher federal revenues. As
long as Social Security operates on a pay-as-you-go basis, productivity growth also
extends the date of reckoning as far as the trust fund balances are concerned, because
the incomes of those paying Social Security taxes will grow more rapidly than the
benefits. Whether productivity growth continues at the rate it did in the late 1990s
is a critical concern for those making and using long-term economic forecasts.
Productivity is a highly cyclical variable, so that assessing its rate of growth at
any single point in time requires knowing the particular stage of the business cycle.
However, it is the long run trend rate of growth of productivity that is of particular
interest because that is the source of rising standards of living. Following the
business cycle peak in the fourth quarter of 1973, productivity growth slowed
substantially. Unfortunately, that slowdown remains poorly understood, which
makes it difficult to design policies that might promote productivity growth.
The recent pickup in productivity is at least in part attributable to the rapid rate
of decline in the prices of computers and other IT equipment. An important factor
in those price declines has been innovation in the manufacture of microprocessors.
As computer prices have fallen, their use has become much more widespread.
Because of falling prices it has become profitable to put computers to uses with
smaller and smaller returns. There is also the prospect that it may take firms a
considerable amount of time to adapt the way they do business to take advantage of
their investments in IT equipment. As was the case with other historic technological
advances, the productivity gains attributable to investments in IT equipment may
ripple through the economy for some time. This report will be updated as economic
developments warrant.

Contents
In troduction ......................................................1
What is Productivity?...............................................2
The Difficulty of Projecting Productivity...............................2
The Cyclical Nature of Productivity...................................3
Measurement Issues................................................4
Problems in Measuring Real Output...............................4
Growth in Labor Productivity........................................6
The Post-1973 Slowdown in Productivity Growth....................7
Accounting for Economic Growth ....................................8
Computers and Productivity Growth..................................10
Accounting for the Role of Computers in the Post-1995 Acceleration
in Productivity Growth....................................12
Is IT the Whole Story?.............................................15
Will the Faster Productivity Growth Continue?.........................17
List of Figures
Figure 1. Chain-Weighted Price Index for Computers and Peripheral
Equipment ..................................................11
Figure 2. Real Investment in Computers and Peripheral Equipment.........12
Figure 3. Investment in Computers and Peripheral Equipment as a
Percentage of Gross Domestic Product............................16
List of Tables
Table 1. Growth in Output per Labor Hour, Nonfarm Business Sector........6
Table 2. Productivity Growth Rates, Nonfarm Business....................9
Table 3. Contributions to Productivity Growth..........................13
Table 4. Sources of Productivity Growth...............................14

Productivity: Will the Faster Growth
Rate Continue?
Introduction
Of all the economic variables policymakers track, productivity growth may be
one of the most important, at least over the long run, because it determines the rate
of improvement in our national standard of living. Economy-wide increases in
productivity indicate increases in real production and incomes, which have been
achieved without an increase in work. Consumers can buy more of those goods and
services (or leisure) that make their lives easier or more enjoyable. Even when
productivity growth is limited to certain industries, everyone benefits from the lower
prices (or the improved quality) for those goods and services.
While policymakers have at least some direct or indirect influence over many
economic variables, productivity growth may be among those that remain relatively
removed from the influence of deliberate economic policy. Although policy
proposals may be represented as promoting productivity growth, productivity growth
may have a greater influence on policy than policy does on the growth of
productivity. It seems that variations in the productivity growth rate are so
incompletely understood that there is no clear consensus among economists how
policymakers might go about promoting it.
Nonetheless, productivity growth rates do have important consequences for
policymakers. The budget process, for example, typically looks at least five, if not

10, years ahead in setting spending and tax policies. With respect to Social Security,

the time frame is even longer. Over such an extended period of time, some insight
into the outlook for productivity growth is critical to projecting other economic
variables and establishing an economic baseline on which to base budget decisions.
Without at least some understanding of underlying factors, projections of
productivity may simply reduce to quantifying forecaster optimism.
Between 1973 and 1995, productivity growth was sluggish compared to what
it had been during the 1960s. In 1995, productivity growth picked up. That it
happened well into an economic expansion was unusual, and it raised hopes that it
was more than temporary and promised a durable increase in the rate of growth. This
report examines both the cyclical and long-run characteristics of productivity growth,
discusses continuing efforts to explain the acceleration in productivity growth, and
considers whether the faster growth rate can be expected to continue.

What is Productivity?
Productivity is a ratio. It is a measure of the quantity of output produced relative
to the amount of work required to produce it. Most often it is expressed as the ratio
of some measure of inflation-adjusted output to the number of labor hours involved.
Mathematically, it looks something like this:
output
productivity=
hours
Rearranging the terms can help illustrate the significance of productivity:
output producti vity hours=×
This shows that total output is a function of both work and productivity. Any
increase in output must therefore come about as the result of increases in either hours
worked or productivity.
In the short run, hours worked may vary over the business cycle as the
unemployment rate rises and falls. Beyond that, hours worked may vary somewhat
over time as the proportion of the population in the labor force changes. But, in the
long run, hours is primarily determined by population growth. Output growth is an
important policy goal, but if it only comes by increasing hours worked then living
standards are unlikely to improve.
When the economy is at full employment, the combined growth rates of labor
and its productivity represent a sort of speed limit. Sustained economic growth
above that limit is considered likely to result in an accelerating rate of inflation. That
is another reason why it is important for policymakers to be aware of productivity
growth trends.
The Difficulty of Projecting Productivity
To make long-run projections of output, forecasters must estimate what
productivity growth will likely be over the forecasted period. However, the study of
productivity has not advanced to the point where it can be projected based on what
is known now about economic conditions. Most forecasts project productivity
growth to continue at its current trend rate of growth.
The current trend rate of productivity growth can be difficult to discern.
Usually, trend rates of growth in productivity are measured over the entire course of
a business cycle to control for cyclical variability in the rate of productivity growth.
The trend rate is thus determined by comparing productivity at successive business
cycle peaks. But, depending on the length of those cycles, that may mean estimates

of the trend rate of growth of productivity are based on somewhat dated information.
It can take a considerable length of time before a change in the trend rate of growth
is fully appreciated.¹
Although it is now clear that productivity growth picked up beginning in 1995,
it was some time before it was thought to be part of a potentially durable shift in the
long-run rate of productivity growth. Most economic forecasters, including the
Congressional Budget Office and the Office of Management and Budget,
underpredicted real economic growth by an average of about 2 percentage points in
each of the fours years after 1995.²
Now that the acceleration in productivity has outlived the last expansion and
seems to be continuing into the present one, there is growing confidence it will
persist. Whether that confidence is misplaced, only time will tell.
The Cyclical Nature of Productivity
In the short run it can be difficult to tell whether a change in the rate of
productivity growth is temporary or indicative of a change in the long-run trend. To
economists, productivity’s significance has more to do with the long run, and so
variations in its long-run trend and the factors that influence it are the focus of
considerable research. Productivity is a cyclical variable, and tends to fluctuate in
somewhat predictable ways over the course of the business cycle. Understanding
those cyclical patterns is necessary to any analysis of productivity data.
Although each business cycle has unique aspects, there are certain tendencies
that characterize them. One of those tendencies is for productivity growth to be
procyclical. In a recession, productivity tends to decline, or grow less rapidly. As
the recession ends, and the economy begins to expand, productivity growth usually
picks up.
At the beginning of an economic contraction, demand for goods and services
declines but firms may be slow to lay off workers both because they may have
invested a considerable amount of time and money in their recruitment and training,
and because there are costs associated with laying those workers off and then rehiring
them when business recovers. The other input to production, physical capital, is
relatively fixed in the short run. So, at the beginning of an economic downturn,
output tends to fall more rapidly than either labor or capital, and so measured
productivity declines.
If the contraction continues and production falls enough, firms will begin to lay
off workers. At first, they will tend to be those most recently hired with the least
amount of training and who were relatively less productive than those hired before.

¹ The last two complete business cycles exceeded 100 months measured from peak to peak.
² CRS Report RL30239, Economic Forecasts and the Budget, by Brian W. Cashell.

Reducing the quantity of labor employed tends to moderate any initial deceleration
in measured productivity growth.
As the contraction comes to an end and the economy begins to expand again,
firms can increase their output initially by putting idle capital back to work and
taking advantage of any under-utilized labor already on hand. This increase in output
with little or no increase in hours worked is reflected in relatively rapid productivity
growth. Once increasing demand can no longer be satisfied with existing capacity,
additional labor will be added. Those hired first will tend to be those relatively more
experienced. As more and more labor is hired, the contribution to output of each
additional hire tends to drop. As the expansion ages, productivity growth slows from
the rates earlier on in the expansion.
The general tendency is thus for productivity growth to decline during
contractions and increase in expansions. Other things being equal, productivity
growth is likely to be faster in the early stages of an expansion than it is after the
economy has regained full employment.
Measurement Issues
Productivity is a ratio of the quantity of output produced to the quantity of
inputs used in its production. An increase in the quantity of output with no increase
in hours worked would be an increase in productivity. As long as output is rising
faster than the contribution of labor and capital, measured productivity will rise.
The most commonly cited measure of productivity published by the federal
government is average labor productivity. It is published quarterly by the Bureau of
Labor Statistics of the Department of Labor (BLS). The measure of output used by
BLS in its calculation is based on data from the national income and product
accounts published by the Bureau of Economic Analysis (BEA) of the Department
of Commerce. Labor productivity is measured in terms of average output per hour.
It is a ratio of the quantity of output to the hours of work done. If the quantity of
output rises by the same proportion as the amount of work, then the economy is only
producing more either because there are more workers, or workers are putting in
longer hours, and there is no productivity growth. If output rises faster than hours
worked, labor productivity is also increasing.
Problems in Measuring Real Output
Production of goods and services is necessarily measured in terms of dollar
values because that is the only unit of measure common to all of the goods and
services produced. The dollar value of output, however, reflects not only the quantity
of goods and services produced, but also their prices. The dollar value of output will
rise with in increase in the quantity of goods and services produced, but it will also
rise with an increase in their prices. Distinguishing between changes in output that
are “real” (i.e., indicative of changes in quantity), and changes that are due only to
variations in the general price level is a difficult problem.

Productivity measures are based on inflation-adjusted measures of output. The
way in which price change is measured can thus affect measures of productivity
growth. If existing price indexes understate the rate of inflation, that will cause
estimates of productivity growth to be overstated.
Few goods or services stay the same from year to year. Over time, most
products acquire new characteristics that make it difficult to compare them with
earlier models. An increase in the price of a car, for example, may reflect rising
prices throughout the economy, but it may also reflect new features such as catalytic
converters or airbags. Ideally, those price increases due to the addition of these new
characteristics would not affect the price index for cars. Even though the same
number of cars might be sold in successive periods, the newer model car might
provide a better (e.g., safer or less polluting) service over its useful life. A more
difficult problem is the introduction of an entirely new product because there is no
price from an earlier period with which to compare the introductory price.
Of all the goods and services produced, computers may be changing the most
rapidly from year to year. The prices of computers have been falling, and their
performance has been improving dramatically. Rather than simply track change in
the price of a “computer” from one year to the next, the BEA attempts to track
changes in the price of “computing power.”³ That means that it tries to take into
account changes in memory, processing speed, and other features when estimating
price change in successive models of computers.
Some sectors of the economy may be easier to measure than others. In the case
of manufactured goods there is at least a tangible product that can be counted even
though there may be difficulties in assessing changes in its quality or other
characteristics. In the case of services, it can be difficult even to define what is being
produced. Take medical care, for example. In the case of physician services, what
should be measured as production, the number of office visits per hour? Should
success at treating various ailments be taken into account?
Some have argued that because the service sector accounts for a growing share
of total national output, and because it is more difficult to measure productivity in the
service sector, that overall measures of productivity have become more prone to
error. At one time it was suggested that at least part of the slowdown in productivity
growth that began, by most accounts, in 1973 may have been due to measurement
problems associated with the increased size of the service sector.⁴ More recent
evidence suggests that is unlikely to have been the case. Those sectors considered
to be harder to measure account for some of the larger productivity gains since 1995.⁵

³ J. Steven Landefeld and Bruce T. Grimm, “A Note on the Impact of Hedonics and
Computers on Real GDP,” Survey of Current Business, Dec. 2000, pp. 17-22.
⁴ Zvi Griliches, “Productivity, R&D and the Data Constraint,” American Economic Review,
vol. 84, issue 1, Mar. 1994, pp. 1-23.
⁵ Robert J. Gordon, Recent Productivity Puzzles in the Context of Zvi Griliches’ Research,
paper presented to meetings of the American Economic Association, Jan. 5, 2002, 17 pp.

Growth in Labor Productivity
How does productivity growth in the economic expansion of the 1990s compare
with those of the past? Table 1 presents data for annual rates of growth in labor
productivity between successive business cycle reference dates for the post-war era.⁶
Two observations are apparent from the data in the table. First, there is clearly
a tendency for productivity growth to be procyclical. That is to say that productivity
growth is higher during expansions (trough to peak) than during contractions (peak
to trough). Moreover, productivity growth was faster in the first year of each
expansion (with the exception of the one-year expansion of 1980-1981) than it was
in the last year. Second, it is clear from the data that, following the business cycle
peak in the fourth quarter of 1973, productivity growth was quite a bit slower than
it had been in prior years.
Table 1. Growth in Output per Labor Hour,
Nonfarm Business Sector
Business Cycle Reference DatesAverage Annual Rate of Growth from:
(year and quarter)(percent)
peaktroughpeakpeak totroughtrough topeakpeak topeak
1948:4 1949:4 1953:3 3.7 3.4 3.5
1953:3 1954:2 1957:3 0.6 2.3 2.0
1957:3 1958:2 1960:2 0.9 2.9 2.3
1960:2 1961:1 1969:4 0.2 2.9 2.7
1969:4 1970:4 1973:4 1.5 3.0 2.9
1973:4 1975:1 1980:1 2.6 1.5 1.1
1980:1 1980:3 1981:3 -1.5 2.2 1.0
1981:3 1982:4 1990:3 -0.6 2.0 1.6
1990:3 1991:1 2001:1 -1.0 2.1 2.0
^Sources^:^N^a^tⁱ^oⁿ^a^l^Bu^reau^of^Econ^omⁱ^c^R^e^s^earch^;^D^e^part^m^eⁿ^t^of^L^a^{bor, Bu}^reau^of^L^a^{bor S}^t^atⁱ^s^tⁱ^c^s^.
To separate changes in long-term trends from those changes that reflect more
variable short run economic conditions, analysts often compare data from similar
points in successive business cycles. A look at the peak-to-peak rates of growth in

⁶ These data reflect productivity in the nonfarm business sector. The business cycle
reference dates use here are those established by the National Bureau of Economic Research
(NBER), and indicate the beginnings and endings of periods of economy-wide expansions
and contractions.

productivity for past business cycles reveals that productivity growth fell by about
half after 1973.
The Post-1973 Slowdown in Productivity Growth
For much of the post-World War II era, the United States experienced relatively
rapid rates of productivity growth. Between the fourth quarter of 1948 and the fourth
quarter of 1973, output per labor hour in the nonfarm business sector grew at an
annual rate of 2.7%. Between the fourth quarter of 1973 and the fourth quarter of
1995, that rate fell to 1.5% per year. Most economists point to 1973 as the beginning
of an extended period of slower productivity growth. That drop in the rate of
productivity growth has been the focus of much economic research. Thus far, the
slowdown remains poorly understood.
The prime suspect in that slowdown, at least initially, was the OPEC oil price
hike, which in 1973 roughly doubled the price of crude oil. The mere coincidence
of the productivity slowdown and the rise in price of a major input to the production
of goods and services motivated research into the connection. The theory was that,
because of higher energy costs, much of the existing capital stock which relied on
energy to contribute to output became obsolete.⁷
Subsequent experience, however, cast doubts on the significance of the
coincidence. Between 1979 and 1981, oil prices doubled again, and then in 1986 the
price of oil fell by nearly half. That only one of these large oil price changes was
associated with a shift in the trend rate of growth in productivity suggested that if
there was a single factor to blame for the 1973 slowdown, it was likely to be found
elsewhere.
Another potential cause for the post-1973 slowdown was spending on research
and development (R&D). Griliches found that R&D spending, as a percentage of
GDP, declined beginning in the mid-1960s.⁸ The timing of the decline would seem
to implicate it, but it is hard to make a strong case out of a single instance.
Moreover, other countries also experienced a decline in productivity growth without
R&D spending having dropped.
Griliches also pointed out that, in the 1970s, the number of patents granted in
the United States declined. That resulted in a drop in the number of patents per
dollar of R&D spending. Griliches suggested that decline in the number of patents
per dollar of R&D spending may have been evidence of diminishing returns to R&D
spending, and he wondered if there might be a sort of technological frontier near
which opportunities for invention become relatively scarcer.

⁷ Martin Neil Baily, “Productivity and the Services of Capital and Labor,” Brookings Papers
on Economic Activity 1, 1981, pp. 1-50.
⁸ Zvi Griliches, “Productivity, R&D, and the Data Constraint,” American Economic Review,
vol. 84, no. 1 (Mar. 1994), pp. 1-23.

Maddison examined a number of factors in an effort to account for the role each
one may have played in the slowdown.⁹ Maddison was able to “explain” only 41%
of the total deceleration in output growth. Of 14 separate factors, the most important
was found to have accounted for less than one-seventh of the slowdown.
Because the slowdown in productivity has been dated to a single point in time,
it was suspected that a single cause for the slowdown might be found. Identifying a
single cause would have been more satisfying in that policy measures might have
been designed to reverse it. If the slowdown had affected different industries more
or less equally, that might have favored arguments that a single cause was
responsible. However, when productivity trends were examined for individual
industries, some were found to have fared well in comparison with others.¹⁰
Gordon offered a slightly different perspective on the post-1973 slowdown. He
maintained that, rather than trying to explain why productivity growth slowed in the

1970s and 1980s, the focus should instead be on why it was so rapid earlier.¹¹

Gordon argued that, in the 1970s, productivity growth simply fell back to its long-run
trend rate, and that the growth experienced earlier in the century was unusually rapid
due to a number of technological advances. Among the advances he cites are the
spread of electric motors, the internal combustion engine, and the telephone and its
derivatives. Gordon argues that these innovations had a much greater economic
effect than the electronic computer. He goes on to suggest that the economy may be
on a long-run curve of diminishing returns to technological advancement, and that
the various modifications and improvements to devices such as computers are less
important than their introduction.
Accounting for Economic Growth
Thus far, economic growth has been explained as the sum of the growth rates
of labor and labor productivity. But, labor is not the only factor of production, so that
explanation remains incomplete. Most economic textbooks present a basic overview
of the theory of economic growth that is known as the “neoclassical” model. In this
model, output is explained as the result of a combination of not just labor, but also
capital and “technology.”
In this model, the size of the capital stock is determined by the saving rate and
technology is treated as “exogenous,” which is economic jargon for a variable which
does not react to the internal influences of the variables in the model but rather is

⁹ Angus Maddison, “Growth and Slowdown in Advanced Capitalist Economies,” Journal
of Economic Literature, vol. 25, no. 2 (June 1987), pp. 649-681.
¹⁰ Robert J. Gordon, Problems in the Measurement and Performance of Service-Sector
Productivity in the United States, National Bureau of Economic Research, Inc., Working
Paper no. 5519, Mar. 1996.
¹¹ Robert J. Gordon, “Comments,” in Ben S. Bernanke and Julio J. Rotemberg, eds., NBER
Macroeconomics Annual 1996, MIT Press, pp. 259-267.

external to it and independent.¹² Labor productivity growth is thus explained by the
increasing amount of capital per worker and technological progress.
In an effort to match data to theory, BLS produces a set of productivity measures
that account for the contribution of both labor and capital in production. Multi-factor
productivity measures attempt to account for both physical and human capital
accumulation.¹³ Accounting for more of the inputs to production reduces the amount
of growth that remains unexplained. Table 2 compares multi-factor productivity and
average labor productivity trends for selected periods.
Table 2. Productivity Growth Rates, Nonfarm Business
Annual Rate of Change in:
Average Labor Multi-Factor
Productivity Productivity
1948 to 19732.81.9
1973 to 19951.40.4
1995 to 20012.51.0
1995 to 20022.81.1
1995 to 20032.91.3
1995 to 20042.91.4
1995 to 20052.81.5
1995 to 20062.61.4
^Source:^D^e^part^m^eⁿ^t^of^L^a^{bor, Bu}^reau^of^L^a^{bor S}^t^atⁱ^s^tⁱ^c^s^.
Both measures show that after 1973 productivity growth slowed, and that after
1995 it accelerated. In the case of multi-factor productivity, growth is faster than it
was between 1973 and 1995 but still below what it was before 1973.¹⁴ The growth
rate of average labor productivity is now close to what it was prior to 1973. Those
two years of faster growth may be related to the usual cyclical pattern.

¹² Robert Solow, “Technological Change and the Aggregate Production Function,” Review
of Economics and Statistics, vol. 39 (Aug. 1957), pp. 312-320.
¹³ Because of the requirement for data on the capital stock, multi-factor productivity data
are only available annually.
¹⁴ Multi-factor productivity data are only available through 2001. Multi-factor productivity
data are not usually as up-to-date as labor productivity because of the need to collect data
on the capital stock.

Computers and Productivity Growth
Robert Solow, a major contributor to the theory of economic growth, is often
quoted for his remark that the effect of computers can be seen everywhere but in the
productivity statistics.¹⁵ Through the 1980s and early 1990s, there seemed to be no
big payoff from the growing stock of computers. That presented a puzzle to those
who expected significant returns.
It is now believed that computers have had much to do with the acceleration in
productivity growth since the mid-1990s. Whether that is the case and how
computers have affected productivity growth are important in trying to assess how
durable the acceleration will prove to be.
It is getting to the point where consumers expect the rapid pace of innovation
in the manufacture of computers to continue. It is also widely assumed that the speed
and memory capacity of those computers will continue to improve at a steady pace.
This rapid rate of technological advance in the development and manufacture of
computers was predicted in 1965 by Gordon E. Moore, one of the co-founders of
Intel Corporation.¹⁶ Specifically “Moore’s Law” predicted that the number of
transistors that could be put on a computer chip would double every 18 months.
Whether that prediction was a self-fulfilling prophecy may be open to question, but
the fact is that the pace of technological advance in the manufacture of computers has
vindicated Moore’s Law over time.
Because of the rapid innovation in the production of computer chips, the prices
of computers, as well as other goods related to information processing and
communications, sometimes referred to collectively as information technology (IT),
have been falling steadily for some time. Figure 1 shows the chain-weighted price
index, published by BEA, for computers and peripheral equipment from 1959
through the middle of 2007. Because the changes are so large the chart is plotted on
a logarithmic scale. Using a logarithmic scale has the added advantage in that the
slope of the line indicates the rate of change in the variable. In this case, the rate of
change in computer prices has been fairly steady for a long time. Between 1959 and
1995, computer prices fell at an average annual rate of 17.2%, and between 1995 and

2007 prices fell at an annual rate of 16.4%.

¹⁵ Robert Solow, “We’d Better Watch Out,” New York Times Book Review, July 12, 1987,
p. 36.
¹⁶ Gordon E. Moore, “Cramming more components onto integrated circuits,” Electronics,
vol. 38, no. 8, Apr. 19, 1965. See also the Intel website at [http://www.intel.com/technology
/mooreslaw/index.htm] .

Figure 1. Chain-Weighted Price Index for Computers
and Peripheral Equipment

^S^o^u^rce:^D^e^pa^r^t^m^e^{nt of}^C^o^m^m^e^r^c^e^,^B^u^r^e^a^u^of^Ec^onom^ic^Aⁿ^a^l^y^s^is^.
These price declines reflect substantial improvements in the quality of
computers. BLS has developed a procedure for estimating price indexes for goods
whose characteristics are changing rapidly. These are referred to as “hedonic” price
indexes. Hedonic price indexes attempt to estimate a statistical relationship between
prices and a set of characteristics, such as memory and processor speed.
These price indexes are important to the measurement of productivity, because
estimating price change is necessary to estimating change in real output and thus
productivity. If the rate of price decline in computers is overestimated, then
measures of productivity will be overstated. Most studies estimate that, in the late
1990s, prices for personal computers alone fell at an annual rate of somewhere
between 30% and 40%.¹⁷
Rapid declines in computer prices have, not surprisingly, stimulated a surge in
investment. Figure 2 shows the chain-linked quantity index for investment in
computers and related equipment from the national income and product accounts
(NIPA). Although data date back to 1959, production of computers was negligible
until the 1980s. Thus, even though real output of IT equipment was increasing
rapidly, it did not account for a very large share of total output until recently.
¹⁷ J. Steven Landefeld and Bruce T. Grimm, “A Note on the Impact of Hedonics and
Computers on Real GDP,” Survey of Current Business, Dec. 2000, pp. 17-22.

Figure 2. Real Investment in Computers and Peripheral Equipment

^Source:^D^e^part^m^eⁿ^t^of^C^o^m^m^e^{rce, Bu}^reau^of^Econ^omⁱ^c^Aⁿ^al^y^sⁱ^s^.
Computers have affected growth in productivity in at least two ways. First,
there has been rapid productivity growth in the production of computers which, as
computers accounted for an increasing share of total production, tended to raise the
overall measure of productivity growth. Second, the sharp drop in computer prices
has stimulated increased investment in computers, which has contributed to an
increase in the overall amount of capital available to the workforce. This is often
referred to as “capital deepening.” Increases in the capital stock generally tend to
raise worker productivity.
Accounting for the Role of Computers in the
Post-1995 Acceleration in Productivity Growth
Prior to the recent acceleration in productivity growth, most analyses found that
computers had not yielded much benefit. One reason for that is that, until recently,
computers accounted for a relatively small share of the total capital stock.¹⁸
¹⁸ Stephen D. Oliner and Daniel E. Sichel, “Computers and Output Growth Revisited: How
Big is the Puzzle?” Brookings Papers on Economic Activity, 2:1994, pp. 273-334.

But, that view has now changed. Although perhaps not yet embracing all of the
claims of proponents of a “new economy,” economists are encouraged that the
acceleration in productivity growth of the late 1990s may mean that the economy is
on a higher growth path and that computers have had a lot to do with it.
Two widely cited studies, discussed below, found considerable evidence that the
computer, or more generally IT equipment, is behind most of the recent acceleration
in productivity growth. There is also evidence of a modest “spillover” into other
sectors of the economy. In other words, investment in computers can raise the
productivity of the workers who use them, but it may also lead firms to change the
way they operate leading to further productivity gains.
The first study, by Oliner and Sichel at the Federal Reserve Board, found that
of a 0.9 percentage point increase in the growth rate of total factor productivity from
the first half of the 1990s to the second half, all of it could be accounted for by
advances in the production of computers themselves and the also by the use of those
computers.¹⁹ Table 3 presents a breakdown of Oliner and Sichel’s accounting for
productivity growth for selected periods since 1974.
Table 3. Contributions to Productivity Growth
^1974-1990^1991-1995^1996-2001^C^h^an^ge
⁽¹⁾⁽²⁾⁽³⁾^{(3) - (2)}
^G^r^ow^th^{rate of l}^a^b^o^{r p}^r^od^u^c^ti^vi^ty^1.36^1.54^2.43^0.89
^C^ont^rⁱ^butⁱ^ons^f^r^om^:
^{Capital Deepening}⁰^.77⁰^.52¹^.19⁰^.67
^In^f^o^rm^atⁱ^oⁿ^t^echⁿ^o^l^o^g^y^capi^t^a^l^0.41^0.46^1.02^0.56
^Oth^e^{r capital}⁰^.37⁰^.06⁰^.17⁰^.11
^L^a^{bor quality}⁰^.22⁰^.45⁰^.25^-^0.20
^M^u^l^tⁱ^{-factor p}^r^od^u^c^ti^vi^ty^0.37^0.58^0.99^0.41
^S^e^mⁱ^con^d^u^c^t^o^rs^0.08^0.13^0.42^0.29
^C^o^m^p^u^t^{er h}^a^rdw^a^re^0.11^0.13^0.19^0.06
^S^o^f^t^w^a^re^0.04^0.09^0.11^0.02
^Co^m^m^uni^c^a^tⁱ^o^{n e}^q^ui^p^m^eⁿ^t⁰^.⁰⁴⁰^.⁰⁶⁰^.⁰⁵^-⁰^.⁰¹
^O^t^h^e^{r n}^oⁿ^f^arm^bu^sⁱⁿ^e^s^s^0.11^0.17^0.23^0.06
^S^o^u^rce:^Olⁱⁿ^{er an}^d^Sⁱ^ch^el^.

¹⁹ Stephen D. Oliner and Daniel E. Sichel, Information Technology and Productivity:
Where Are We Now and Where Are We Going? Board of Governors of the Federal Reserve
System, May 2002, 78 pp.

Oliner and Sichel found that of a 0.89 percentage point increase in average labor
productivity between the early and late 1990s, 0.56 was due to increased investment
in IT related capital (an increase from 0.46 to 1.02), and 0.35 was due to increased
productivity in the production of IT equipment (an increase from 0.26 to 0.61 in the
combined computer and semiconductor sectors). Thus, the contribution of IT
equipment to the increase in productivity was greater than the overall increase.
Oliner and Sichel also found that labor quality’s contribution to productivity growth
declined during the 1990s. That is likely related to cyclical factors as the
unemployment rate fell and the available pool of skilled workers shrank.
Oliner and Sichel, using an economic model, attempted to assess the
implications of recent developments in the technology sector for prospects for
continued rapid productivity growth. They conclude that productivity growth is
likely to fall somewhere in the range of 2% - 2¾% over the next 10 years.
A second study, by Jorgenson, Ho, and Stiroh, came to similar conclusions.²⁰
Table 4 presents the results of their analysis.
Table 4. Sources of Productivity Growth
^1959-1973^1973-1995^1995-2001^C^h^an^ge
⁽¹⁾⁽²⁾⁽³⁾^{(3) - (2)}
^G^r^o^w^t^h^ra^t^e^o^f^la^bo^{r pro}^ductⁱ^vⁱ^t^y^2.63^1.33^2.02^0.69
^C^ont^rⁱ^butⁱ^ons^f^r^om^:
^{Capital Deepening}^1.13^0.80^1.39^0.59
^I^T^Cap^{ital Deep}^enⁱⁿ^g⁰^.19⁰^.37⁰^.85⁰^.48
^Oth^e^r^Cap^{ital Deep}^enⁱⁿ^g⁰^.95⁰^.43⁰^.54⁰^.11
^La^bo^{r Q}^u^a^lit^y^0.33^0.27^0.22^-⁰^.05
^T^o^tal^{Factor Prod}^u^c^ti^vi^ty^1.16^0.26^0.40^0.14
^In^f^o^rm^atⁱ^oⁿ^T^echⁿ^o^l^o^g^y^0.09^0.21^0.41^0.20
^N^oⁿ^-ⁱⁿ^f^o^rm^atⁱ^oⁿ^T^echⁿ^o^l^o^g^y^1.07^0.05^-⁰^.01^-^0.06
^S^o^u^rce:^J^o^r^g^eⁿ^s^{on, H}^o^{, a}^{nd Stir}^oh.
According to Jorgenson, Ho, and Stiroh’s estimates, of a 0.69 percentage point
rise in average labor productivity growth during the 1990s, increased investment
(capital deepening) accounted for 0.59 percentage point, and improved productivity
in the IT sector itself contributed another 0.20 percentage point of the acceleration.

²⁰ Dale W. Jorgenson, Mun S. Ho, and Kevin J. Stiroh, Lessons From the U.S. Growth
Resurgence, paper prepared for the First International Conference on the Economic and
Social Implications of Information Technology, held at the U.S. Department of Commerce,
Washington, DC, on Jan. 27-28, 2003, 28 pp.

The evidence suggests that increased productivity in the sector producing IT
equipment has had a modest direct effect on total factor productivity. By far the
more important factor has been the declining price of IT equipment stimulating a
surge in investment and increasing the size of the capital stock.
Remember that total factor productivity measures changes in output that are not
accounted for by changes in economic inputs such as labor and capital. There is no
doubt that computers are raising productivity of many firms, but, as long as economic
statistics measure them correctly, the increased share of work computers do will not
show up in increased multi-factor productivity because that measure of productivity
tracks the increase in output not associated with the increase in investment in
computers.²¹ It is unclear whether computers have had any “spillover” effects on
multi-factor productivity beyond their direct contribution to growth in output.
Some evidence suggests that those spillover effects — of computers on total
factor productivity — are fairly small. Jorgenson and Stiroh found that those sectors
of the economy that invest most heavily in computers and IT equipment, such as
financial services, had among the lowest rates of productivity growth measured.²²
Jorgenson, Ho, and Stiroh also estimated projections of growth in average labor
productivity. They projected that productivity growth would range between 1.14%
and 2.38% over the next decade, with a base case of 1.78%, just below the 1995 -

2001 rate of growth.

In another study, Robert Gordon found that most of the acceleration in labor
productivity was attributable to capital deepening and faster productivity growth in
the production of computers and IT equipment.²³ Of the roughly 0.2 percentage point
increase in total factor productivity, most was accounted for by faster productivity
growth in the manufacture of durable goods. That suggests that any spillover effects
of computers on the overall economy were limited.
Is IT the Whole Story?
Since the publication of those studies attributing much of the acceleration in
productivity growth to IT investments, there was a significant decline in the share of
GDP allocated to IT investments. Figure 3 shows investment spending as a
percentage of GDP since 1995.

²¹ Barry P. Bosworth and Jack E. Triplett, What’s New About the New Economy? IT,
Economic Growth and Productivity, Brookings Institution, Dec. 2000, 35 pp.
²² Dale W. Jorgenson and Kevin J. Stiroh, “Raising the Speed Limit: U.S. Economic Growth
in the Information Age,” Brookings Papers on Economic Activity, 2000(1), vol. 2, pp. 125-

212.

²³ Robert J. Gordon, Technology and Economic Performance in the American Economy,
National Bureau of Economic Research, Working Paper no. 8771, Feb. 2002.

Figure 3. Investment in Computers and Peripheral Equipment as
a Percentage of Gross Domestic Product

^S^o^u^rce:^D^e^pa^r^t^m^e^{nt of}^C^o^m^m^e^r^c^e^,^B^u^r^e^a^u^of^Ec^onom^ic^Aⁿ^a^l^y^s^is^.
Gordon suggests that the continuation of rapid productivity gains following the
decline in spending on computers may mean that the earlier studies exaggerated the
contribution of IT spending on the post-1995 acceleration in productivity growth.²⁴
Gordon argues that it is implausible that increased investment in IT equipment would
have an immediate effect on productivity growth. Gordon points out that it took forty
years for the benefits of electrification to be fully realized. Factories had to be
reorganized to take advantage of electric power. In the same way, the introduction
of computers has driven firms to change business practices. Those changes take
time. But the reorganization, as well as worker training and other novel business
practices made possible by the introduction of computers, may have contributed to
productivity in ways that are difficult to measure.
One of the sectors of the economy that experienced relatively rapid productivity
growth is retail trade. Fernald and Ramnath cite the particular example of so-called
big-box stores that have become pervasive in retailing.²⁵ The success of those large
scale retailers may have been made possible by the introduction of computers.
Computers allowed those stores to manage sophisticated distribution networks and
²⁴ Robert J. Gordon, “Exploding Productivity Growth: Context, Causes, and Implications,”
Brookings Papers on Economic Activity, 2:2003, pp. 207-298.
²⁵ John G. Fernald and Shanthi Ramnath, “The Acceleration in U.S. Total Factor
Productivity after 1995: The Role of Information Technology,” Federal Reserve Bank of
Chicago Economic Perspectives, 1Q/2004, pp. 52-67.

improve inventory control, and by growing in size also benefit from economies of
scale.
Baily also suspects that the contribution of IT investments to faster productivity
growth might have been overstated.²⁶ He points out that prior to 1995, growth
accounting was not especially helpful in explaining variations in productivity growth.
Given the relatively small number of observations on which the studies linking IT
investments to productivity are based they might still be treated with some
skepticism. He argues that it might also be the case that the increase in IT spending
was motivated by the acceleration in growth.
Nordhaus analyzed the contribution of what he called the “new economy” to the
post-1995 acceleration in productivity growth.²⁷ The four industries included in his
definition of the new economy are: industrial machinery and equipment, electronic
and other electric equipment, telephone and telegraph, and software. He found that
labor productivity in total non-farm business accelerated by 1.61 percentage points
between the 1977-1989 period and the 1995-2000 period. Of that, the new economy
industries contributed only 0.29 percentage point, or about a sixth of the total
acceleration.
Nordhaus also calculated productivity growth for those industries whose
production was “well measured.” Well-measured industries included, agriculture
forestry and mining; manufacturing; transportation and public utilities; wholesale
trade; retail trade; and a small number of services. Nordhaus found that the
productivity growth in the well-measured economy accelerated by 1.31 percentage
points, less than the increase in total labor productivity. Nordhaus concludes that,
while the new economy contributed, the rise in productivity growth was not limited
to those firms but was widespread.
Will the Faster Productivity Growth Continue?
Perhaps the most important question for policymakers is whether or not the
surge in productivity growth of the late 1990s will continue. Higher productivity
growth means higher real incomes, which in combination with progressive income
tax rates yields higher federal revenues. As long as Social Security operates on a
pay-as-you-go basis, it also extends the date of reckoning as far as the trust fund
balances are concerned because the incomes of those paying Social Security taxes
will grow more rapidly than the benefits. Whether or not productivity growth

²⁶ Martin N. Baily, “Recent Productivity Growth: The Role of Information Technology and
Other Innovations,” Federal Reserve Bank of San Francisco Economic Review, 2004, pp.

35-41.

²⁷ William D. Nordhaus, “Productivity Growth and the New Economy,” Brookings Papers
on Economic Activity, 2:2002, pp. 211-265.

continues at the rate it did in the late 1990s is a critical concern for those making and
using long-term economic forecasts.²⁸
It now seems likely that most of the increase in productivity growth of the late
1990s was not just a cyclical variation, but rather faster productivity growth seems
likely to persist. But there is still no guarantee.
The major difficulty in projecting productivity growth remains an imperfect
understanding of past variations. Some of the sources of productivity growth are
clearly understood. Increased investment and a growing capital stock raise labor
productivity. Increased education and training also contribute. But aside from the
contributions of human and physical capital, much less is certain. To a great extent,
projections of productivity still reflect the optimism or pessimism of the forecaster.
It is clear however that the recent pickup in productivity is at least in part
attributable to the rapid rate of decline in the prices of computers and other IT
equipment. An important factor in those price declines has been innovation in the
manufacture of microprocessors. Whether or not that rapid pace of innovation keeps
up, and prices continue to fall will be important factors in future rates of productivity
growth. However, ultimately there may be limits to the number of transistors that can
be put on a single computer chip.
As computer prices have fallen, their use has become much more widespread.
Because of falling prices it has become profitable to put computers to uses with
smaller and smaller returns. But, as long as recent rates of innovation in the
production of computers and IT equipment continue, productivity might at least be
expected to continue growing more rapidly than it did between 1973 and 1995.
There is also the prospect that it may take firms a considerable amount of time
to adapt the way they do business to take advantage of their investments in IT
equipment. As was the case with other historic technological advances, the
productivity gains attributable to investments in IT equipment may ripple through the
economy for some time.

²⁸ Paul W. Blauer, Jeffrey L. Jensen, and Mark E. Schweitzer, Productivity Gains, How
Permanent? Federal Reserve Bank of Cleveland Economic Commentary, Sept. 1, 2001.