Delong on productivity miracle..In the second half of the 1990s American productivity picked itself up, and more-or-less
resumed its pre-1973 pace. Between the beginning of 1995 and the semi-official NBER
business cycle peak in March 2001, U.S. nonfarm-business output per person-hour
worked grew at an annual rate of 2.80 percent per year. (Extending the sample through
the 2001 recession to the likely trough point of 2001:4, the late-1990s growth rate is 2.69
2 See Krugman (1994) for one interpretation of how the productivity slowdown made a big difference.
4
percent per year.) Between the beginning of 1995 and the semi-official NBER business
cycle peak in March 2001, U.S. real GDP grew at a pace of 4.21 percent per year.
(Extending the sample through the 2001 recession to the likely trough point of 2001:4,
the late-1990s growth rate is 3.85 percent per year.) Non-economists tended to attribute a
large chunk of fast late-1990s growth to “cyclical” factors,3 but economists had a much
harder time attributing more than a few tenths of a percentage point per year of late-
1990s growth to the business cycle.4 Moreover, as Susanto Basu, John Fernald, and
Matthew Shapiro have powerfully argued, there are stronger reasons for thinking that the
adjustment costs associated with moving to a more information technology capitalintensive
growth path led actual growth to understate trend growth than for thinking that
cyclical factors led actual growth to overstate trend growth in the second half of the
1990s.5 And the extremely rapid run-up of stock prices indicated that at least the marginal
investor in equities anticipated that the acceleration of economic growth that started in
the mid-1990s would last for decades or longer.6
The causes of the productivity slowdown of the 1973-1995 or so period remain
disappointingly mysterious. Baily (2002) calls the growth-accounting literature on the
slowdown “large but inconclusive.” No single factor provides a convincing and coherent
3 See, for example, Kosterlitz (2002).
4 See Gordon (2002) and Gordon (2000).
5 See Basu, Fernald, and Shapiro (2001).
6 See Greenwood and Jovanovic (1999).
5
explanation, and the residual position that a large number of growth-retarding factors
suddenly happened to hit at once is unlikely.7
By contrast, nearly all agree on the causes of the productivity speed-up of 1995-2001: it
is the result of the extraordinary wave of technological innovation in computer and
communications equipment—solid-state electronics and photonics.8 Robert Gordon
(2002) writes that cyclical factors account for “0.40” percentage points of the growth
acceleration, and that the rest is fully accounted for by information technology—an “0.30
[percentage] point acceleration [from] MFP growth in computer and computer-related
semiconductor manufacturing” and a “capital-deepening effect of faster growth in
computer capital… in the aggregate economy accounts [for] 0.60 percentage points of the
acceleration.” Kevin Stiroh (2001) writes that “all of the direct contribution to the post-
1995 productivity acceleration can be traced to the industries that either produce
[information technology capital goods] or use [information technology capital goods]
most intensively, with no net contribution from other industries… relatively isolated from
the [information technology] revolution.” Oliner and Sichel (2000) write that “the rapid
capital deepening related to information technology capital accounted for nearly half of
7 See Fischer (1988), Griliches (1988), Jorgenson (1988), and Gordon (2000b) and (2002). Jorgenson
(1988) convincingly demonstrates that the oil price shocks can account for slow growth in potential output
in the 1970s, but why does potential output growth remain slow after 1986 after real oil prices have fallen
again? Griliches (1988) finds that an explanation in terms of a slowdown in innovation is unattractive, but
Gordon (2000b) and (2002) finds such an explanation attractive.
8 The only major study taking a stand against this explanation is the McKinsey Global Institute (2001),
which presents a regression of the growth in value added per worker and the increase in computer capital
by industry. When industry observations are counted equally, they find next to no correlation between
computer capital and labor productivity. When they weight industries by employment, they find a
statistically significant and substantively important connection.
6
this increase” in labor productivity growth, with a powerful “additional growth
contribution… com[ing] through efficiency improvement in the production of computing
equipment.” Jorgenson, Ho, and Stiroh (2001) reach the same conclusions about the
importance of information technology capital-deepening and increased efficiency in the
production of computing and communications equipment as major drivers of the
productivity growth acceleration, and they go on to forecast that labor productivity
growth will be as high in the next decade as it has been in the past half-decade.9
The failure of economists to reach consensus in their explanations of the productivity
slowdown has to leave one wary of the reliability of the consensus about the causes of the
productivity speed-up. This paper, however, will assume that this consensus is correct:
that the productivity growth speed-up in the second half of the 1990s was the result of the
technological revolutions in data processing and data communications. It will then go on
to ask what the boom of the past seven years means for productivity growth in the next
decade or so. Will the decade of the 2000s be more like the late 1990s, or more like the
1980s as far as growth in productivity and living standards is concerned?
The answer is that the smart way to bet is that the 2000s will be much more like the fastgrowing
late 1990s than like the 1980s. The extraordinary pace of invention and
innovation in the information technology sector has generated real price declines of
between ten and twenty percent per year in information processing and communications
9 However, Jorgenson, Ho, and Stiroh expect total real GDP growth to slow because of slower growth in
hours worked—they forecast 1.1 percent per year growth in hours over the next decade, compared to 2.3
percent per year growth in hours from 1995 to 2000.
7
equipment for nearly forty years so far. There are no technological reasons for this pace
of productivity increase in these leading sectors to decline over the next decade or so. In
the consensus analysis, creased total factor productivity in the information technology
capital goods-producing sector coupled with extraordinary real capital deepening as the
quantity of real investment in information technology capital bought by a dollar of
nominal savings grows have together driven the productivity growth acceleration of the
later 1990s. It may indeed be the case that a unit of real investment in computer or
communications equipment “earned the same rate of return” as any other unit of real
investment, as Robert Gordon (2002) puts it. But the extraordinary cost declines had
made a unit of real investment in computer or communications equipment absurdly
cheap, hence the quantity of real investment and thus capital deepening in informationtechnology
capital absurdly large.
Thus the most standard of simple applicable growth accounting approaches predicts a
bright future for the American economy over the next decade or so. Continued declines in
the prices of information technology capital mean that a constant nominal flow of savings
channeled to such investments will bring more and more real investment. As long as
information technology capital earns the same rate of return as other capital, then labor
productivity growth should continue very high. The social return to information
technology investment would have to suddenly and discontinuously drop to nearly zero,
or the share of nominal investment spending devoted to information technology capital
would have to collapse, or both, for labor productivity growth in the next decade to
reverse itself and return to its late 1970s or 1980s levels.
8
Moreover, additional considerations tend to strengthen, not weaken, forecasts of
productivity growth over the next decade. It is very difficult to argue that the speculative
excesses of the 1990s boom produced substantial upward distortions in the measured
growth of potential output. The natural approach that economists to model investment
spending in detail—the approach used by Basu, Fernald, and Shapiro (2001)—tells us
that times of rapid increase in real investment are times when “adjustment costs” are
unusually high, and thus times when actual productivity growth undershoots the long-run
sustainable trend. Both a look back at past economic revolutions driven by generalpurpose
technologies10 that were in their day analogous to the computer in their effects11
and a more deeper look forward into the likely determinants of productivity growth
suggest a bright future.
The pace of technological progress in the leading sectors driving the "new economy" is
very rapid indeed, and will continue to be very rapid for the foreseeable future. Second,
the computers, switches, cables, and programs that are the products of today's leading
sectors are what Bresnehan and Trajtenberg (1985) call “general-purpose technologies,”
hence demand for them is likely to be extremely elastic. Rapid technological progress
brings rapidly falling prices. Rapidly falling prices in the contest of extremely elastic
demand will produce rapidly-growing expenditure shares. And the economic salience of a
leading sector--its contribution to productivity growth--is the product of the rate at which
the cost of its output declines and the share of the products it makes in total demand.
10 See Bresnahan and Trajtenberg (1995).
9
Thus unless Moore's Law ceases to hold or the marginal usefulness of computers and
communications equipment rapidly declines, the economic salience of the data
processing and data communications sectors will not shrink, but grow. |
