Digital Economy
CHAPTER IV
CONTRIBUTION OF INFORMATION TECHNOLOGY TO U.S. PRODUCTIVITY GROWTH (62)
This chapter examines recent studies of the impact of information technology
(IT) on labor productivity in the United States. Our analysis of these studies
concludes that, based on macroeconomic and firm-level evidence, IT does
contribute significantly to productivity growth. However, studies at the
industry level continue to produce mixed results.
MACROECONOMIC ASSESSMENTS
The current U.S. productivity pattern, in which productivity gains have
strengthened as the expansion has matured, is unprecedented for the postwar
period. In previous postwar expansions, productivity growth has slowed as the
expansion enters its mature phase. (Figure 4.1)
One reason for the extraordinary pattern of productivity in the current
expansion appears to be the rapid growth in the real net stock of IT capital
per labor hour, especially computer hardware (including peripheral equipment).
This rapid growth of real net IT capital created significant IT "capital
deepening," beginning in 1991 and accelerating sharply after 1995. (63) The
ratio of the capital stock of computer hardware to hours worked increased, on
average, by 16.3 percent per year over the period 1991-95, and 33.7 percent per
year during 1996-99. (Figure 4.2) Capital deepening in computer software also
grew at double-digit rates during both periods, while the growth rate in
communications equipment increased from 2.4 to 5.0 percent. By contrast, over
the 1990s, the rate of capital deepening for all other forms of
capital--covering over 95 percent of the total U.S. capital stock--averaged
only about one-half of one percent per year. (64)
A major factor behind IT capital deepening has been the falling prices of IT,
especially computer hardware, reflecting rapid and continuous improvements in
quality. (65) The quality-adjusted price deflator for computer hardware fell 14
percent per year during the first half of the 1990s and 29 percent per year
during 1996-98. (Figure 1.3, Chapter I.)
Figure 4.3, comparing the 1991-95 and 1996-99 periods, shows that IT capital
deepening accounts for a large and increasing share of the economy's rising
productivity gains. The figure also shows that the acceleration of labor
productivity growth has been accompanied by an acceleration in "multifactor
productivity" (MFP) growth within the IT-producing sector itself. Multifactor
productivity growth reflects the impact of factors in addition to
quality-adjusted capital and labor inputs--for example, technical changes not
directly incorporated in capital and labor (such as new production processes),
organizational improvements, and economies of scale. As discussed below, growth
in multifactor productivity has been especially strong in the computer and
semiconductor industries.
All categories of IT capital contribute disproportionately to labor
productivity growth, compared to their shares of the total private nonfarm
business sector and their shares of the total net capital stock. However, the
contribution of computer hardware to productivity growth has been
extraordinarily large. Recent research indicates that during the years 1996-99,
computer hardware deepening accounted for 24 percent of all labor productivity
growth. (Figure 4.4) The size of this contribution is especially
remarkable because computers constitute just 1.8 percent of the private nonfarm
business sector and less than 1 percent of overall capital stock (1998).
The reason that IT, with such a small share of the economy and of the total
capital stock, has contributed so powerfully to productivity growth is that the
rapidly growing IT investments have been unusually productive. Market
conditions dictate that business investments in computer hardware must earn
very high rates of return. For one thing, the rapid and continuous improvements
in IT quality mean that existing computer hardware becomes obsolete and hence
depreciates very quickly. In addition, sharply falling hardware prices mean
that businesses investing in IT equipment face rapid capital losses as
purchased equipment quickly loses market value. Oliner and Sichel estimate that
investment in computer hardware must produce gross rates of return of about 68
percent in order to cover an estimated depreciation rate of 30 percent and
capital loss of 34 percent per year, and a competitive net rate of return of 4
percent per year. By their estimates, the payback period for computer hardware
investments is less than two years.
The Emerging Consensus on Resolving the "Computer Productivity Paradox"
Economists who held until recently that the impact of computers on U.S.
productivity could be a transitory effect of unusually favorable economic
circumstances have begun to credit IT for dramatic increases in the trend
growth rates of U.S. output and productivity since 1995. One reason for this
change in view has been the increasing attention paid by researchers to the
productivity effects of software and communications equipment, in addition to
computer hardware. This shift in attention follows the 1998 and 1999 editions
of this report that introduced a broader definition of IT (68) and the
reclassification by the Bureau of Economic Analysis (BEA) of software spending
from current expenditures to investments.
Thus, Oliner and Sichel conclude that surging use of IT (including computer
hardware, software, and communications equipment) in the second half of the
1990s, together with advances in the production of computers and
semiconductors, contributed about two-thirds of an estimated 1.06-percentage
point acceleration in productivity growth between the first and second halves
of the decade.
Consistent with Oliner and Sichel's findings, the Congressional Budget Office,
the Economic Report of the President, Jorgenson and Stiroh, Whelan, and
Macroeconomic Advisers, LLC find strong evidence that the mid-1990s
acceleration in productivity growth was due largely to IT capital deepening
among IT users and also to technical advances and innovations made by IT
producers. These analysts' recent estimates of IT's contribution both in
computer use (capital deepening) and computer and semiconductor production
(technical advance) are summarized in Table 4.1. (69)
When one takes into account the differences in the periods studied and in the
coverage of economic activity, idiosyncratic adjustments for limitations in the
available data, and other factors, these estimates appear to be remarkably
consistent.
Recent studies also suggest that robust productivity growth is likely to
continue. For example, Macroeconomic Advisers found that: "Given the large gap
between discovery and application in the computer industry, it is reasonable to
conclude that real computer prices, which on average have declined 20 percent
per year since 1996, will continue falling rapidly. As long as they do, the
special contribution to productivity growth coming from the technology sector
will persist." (70)
_______________________________________________________________________________
Table 4.1
Contribution of IT Capital to the Acceleration of Labor Productivity Growth
in the U.S. Private Nonfarm Business Sector
Studies* Capital Technical Total IT Productivity IT Share of
Deepening Advance Contribution Acceleration
Acceleration
(Percentage (Percentage (Percentage (Percentage (Percent)
Point) Point) Point) Point)
Oliner and Sichel IT Other IT Other (a) (b) (a/b)x100
1996-99
over 1991-95 0.45 0.03 0.26 0.41 0.71 1.04 68.3
Congressional Budget Office
1996-99
over 1974-99 0.40 - 0.20 - 0.60 1.10 54.5
Economic Report of the President
1995-99
over 1973-95 0.47 - 0.23 0.70 0.70 1.47 47.6
Jorgenson and Stiroh
1995-98
over 1990-95 0.31 0.18 0.19 0.44 0.50 1.00 50.0
Whelan
1996-98
over 1974-95 0.46 - 0.27 - 0.73 0.99 73.7
*The studies summarized are not strictly comparable because they use different
definitions of IT capital and examine different time periods. Oliner and
Sichel define IT capital to include "computer hardware, software, and
communication equipment." The Congressional Budget Office talks about
"computers," distinguishing between computer "use" (capital deepening) and
computer "production" (technical advance), while the Economic Report of the
President refers to "computers and software." Jorgenson and Stiroh include in
IT "capital services" those from computer, software, and communications
capital. Finally, Whelan's "computing equipment" includes mainframes,
terminals, storage devices, printers, and personal computers.
In the above table, "IT capital deepening" means increase in IT capital per
labor hour and "other capital deepening" means increase in other types of
capital per labor hour. "Technical advance" covers capital quality
improvements and multifactor productivity growth from IT and other sources.
Finally, there are factors contributing to labor productivity growth
acceleration other than capital deepening and technical advance that are not
identified in the table (e.g., improvements in labor quality). These other
factors are omitted since the table is intended to highlight IT's contribution
to the acceleration of labor productivity growth.
_______________________________________________________________________________
SECTORAL AND INDUSTRY-LEVEL ASSESSMENTS
Since IT investments improve productivity, those industries making the most
intensive use of IT should show higher productivity growth than industries that
use IT less intensively (all other factors held constant). Evidence of such a
pattern at the industry level, however, remains mixed. IT-producing industries
have recorded astonishingly high productivity gains and have been a dominant
force in aggregate U.S. productivity growth. (71) Furthermore, outside the
IT-producing sector itself, goods-producing industries that are IT intensive
have achieved higher productivity gains than their counterparts that have not
invested heavily in IT. However, official output measures for IT-intensive
service industries do not indicate significant productivity gains. Indeed,
between 1990 and 1997, despite heavy investments in IT and a three-decade
buildup of the real net IT capital stock, IT-using service industries as a
group recorded declining productivity.
The following sections review analyses that show significant multifactor
productivity growth in IT-producing industries, improved labor productivity
growth in both IT-producing and IT-using goods industries, and alternative
views of IT's effect on productivity in service industries.
Computer Production
A study conducted by Kevin Stiroh examined the relationship between computers
and economic growth, at both the aggregate and sectoral levels, over the period
1947 to 1991. This study found strong labor productivity, as well as
multifactor productivity gains in the computer-producing sector, implying that
this sector positively contributes to overall productivity growth. (72)
Estimates by the Bureau of Labor Statistics (BLS) confirm Stiroh's finding that
IT-producing industries make an outstanding contribution to multifactor
productivity growth. (73) The BLS estimates show that industrial machinery and
equipment (SIC 35) and electronic and electric equipment (SIC 36)--the
categories that include the computer and semiconductor industries--ranked
highest in multifactor productivity growth among all manufacturing industries
between 1990 and 1996. Similarly, a May 2000 analysis by Dale Jorgenson and
Stiroh concluded that IT production is a major force behind the current
resurgence in multifactor productivity growth. (74)
Computer Use
Analyses of computer-using industries outside the IT sector, however, continue
to show mixed results. For the period prior to 1991, Stiroh found that in
computer-using sectors, rapidly falling computer prices led firms to substitute
capital for labor and other inputs. The result was that remaining workers had
more capital to work with, and labor productivity rose. However, Stiroh found
little evidence that investments in computers affected multifactor productivity
growth in these sectors. These findings have been broadly confirmed by
Jorgenson and Stiroh himself in studies in 1999 and 2000. Both researchers note
that price declines in IT have led to capital deepening in IT-using industries,
but they still see "no corresponding eruption of industry-level [multifactor]
productivity growth in these sectors." (75)
By contrast, evidence of multifactor productivity growth in some IT-using
industries has been documented in a study by Jack Triplett and Barry Bosworth.
They estimate that from 1987 to 1997, multifactor productivity grew 9.0 percent
per year among security and commodity brokers, 2.1 percent a year among
insurance carriers, and 2.2 percent among holding and investment offices. (76)
These estimates of productivity growth for the period 1987 to 1997
significantly exceeded productivity gains for the same industries in the years
1960 to 1973.
ESA's own industry-level analysis covering the 1990-97 period supports Stiroh's
conclusion. (77) We found that gross product originating per worker (GPO/W), an
approximate measure of labor productivity, was stronger in IT-using goods
industries than in non-IT-intensive goods industries--2.4 percent per year
compared to 1.3 percent. No similar pattern emerged, however, among service
industries. IT-using service industries actually showed a negative growth rate
of 0.3 percent a year, compared to 1.3 percent annual productivity gains by
non-IT intensive service industries. These results largely reflect the
difficult problems in conceptualizing and measuring output in many service
industries.
In view of these measurement problems, we compared growth in GPO/W of IT-using
service industries with that of the non-IT intensive service industries, but
excluding 10 hard-to-measure service industries. (78) We found that when the
hard-to-measure industries are excluded, IT-using service industries show
slightly greater GPO/W growth than non-IT intensive service industries, and the
overall annual average GPO/W growth for 1990-97 rises from 1.38 percent to 2.34
percent. (Figure 4.5) Because hard-to-measure service industries together
account for 44 percent of the total GPO by IT-using service industries, the
effect of IT on service industry productivity will remain clouded until
development of better output measures.
BEA, BLS, and other statistical agencies are currently grappling with the
challenge of improving output measurements for service industries. BEA plans to
release new GPO-by-industry data this summer. We expect that, based on this new
data, estimates for some hard-to-measure service industries will show
productivity gains. (79)
FIRM-LEVEL EVIDENCE
Like the macroeconomic studies, recent firm-level analyses show that IT
contributes substantially to productivity growth. This contribution is
especially strong where businesses undertake organizational and other changes
that complement the adoption of IT.
In one study, Erik Brynjolfsson and Lorin Hitt analyzed firms in four groups
based on their levels of IT investment and degree of decentralization. While
they found that average productivity was highest among firms that were high in
both IT investment and decentralized organization, they also found that
productivity was lowest among those firms that were high in IT investment but
low in decentralization. In fact, the productivity of firms that invested
heavily in IT but remained highly centralized declined relative to firms which
were low on both measures. (80)
Another study by the same researchers similarly concluded that while computers
make a positive contribution to productivity growth at the firm level, "the
greatest benefit of computers appears to be realized when computer investment
is coupled with other complementary investments; new strategies, new business
processes, and new organizations all appear to be important." (81)
Another analysis, by Brynjolfsson and Shinkyu Yang, found that a one-dollar
increase in computer capital is associated with a $10 increase in the valuation
of the firm by the stock market, based on eight years of data for 820
non-financial U.S. firms. (82) This finding does not imply that the market
values a dollar of computers at $10, but rather that "the firm that has a
dollar of computers typically has another $9 of related intangibles." (83) In
order to make effective use of computers, firms have to make expensive
investments in software, training, and organizational changes, which together
create intangible assets. The researchers estimate that when the costs of these
intangible assets and other adjustment costs are added to the direct
expenditures on computers, the firms had normal returns on investment. No other
category of capital investment shows such high valuations relative to tangible
investments.
In conclusion, based on both macroeconomic and firm-level analyses, IT makes a
substantial contribution to overall productivity growth. The firm-level studies
show that firms that have made the organizational and other changes necessary
to effectively use IT become more productive over time than those that have
not. However, analyses of the impact of IT on productivity at the industry
level have produced mixed results, largely reflecting the limitations of
measuring the output of many service industries. Until these measures are
improved, the full effect of IT on service industry productivity will remain
clouded.