the range of overlapping consecutive fifteen-year averages of the annual increase
in output per hour stays consistently between 1 and 3 percent.* I
suspect that a good deal of even that modest volatility is statistical "noise,"
random aberrations resulting from the uncertain quality of the data, especially
for the years preceding World War II.
There is little doubt, however, that the burst of U.S. nonfarm productivity
growth from 1995 to 2002 has given way to a lessened pace of
growth. Output per hour, for example, after the large surge in growth
peaked at 4 percent (a four-quarter rate of change) and above in 2002 and
2003, slipped to a 1 percent rate by the first quarter of 2007. Profitable
opportunities for further advance appear to have temporarily dwindled, as
*Growth rates slowed following the sharp increase in energy costs in the 1970s and presumably
because of it. The technology boom of the past decade has accelerated growth in output
per hour, restoring it to its long-term trend.
471
THE AGE OF TURBULENCE
has often occurred in the past. Innovative expansion seems to come in
waves. New products and new companies were major factors in the surge
of new issues of stock between 1997 and 2000, and the apparent fall-off
in applications of innovation since then has been reflected in the decline
in stock issuance. The slowdown in innovation is particularly evident in
the dramatic swing in corporations' use of their internal cash flow (the result
of earlier gains in the application of new technologies) from fixed investment
to buybacks of company common stock and cash disbursed to
shareholders in the process of implementing mergers and acquisitions.
Such return of cash to shareholders of nonfinancial corporations rose from
$ 180 billion in 2003 to more than $700 billion in 2006. Fixed investment,
on the other hand, rose only from $748 billion in 2003 to $967 billion in
2006. A corporation returns equity capital to shareholders when it cannot
find opportunities for prospective risk-adjusted rates of return superior to
the rate of return that the corporation obtains from existing assets. Large
cash disbursements to shareholders are usually a signal of lowered prospective
rates of return on fixed investments available to the corporation,
the likely result of a slowed pace of profitable new applications of
innovation.*
Similar signals are reflected in price trends in high-tech equipment,
which had been the driving force in rising overall nonfarm productivity
growth between 1998 and 2002. At the Federal Reserve, we monitored
those price trends as one proxy of the rate of growth of productivity in the
high-tech equipment sector itself, a significant part of recent overall productivity
gains. Falling prices are generally possible over protracted periods
only if unit labor costs are falling in tandem, a trend that is not likely unless
productivity is rising fast. And thus the rate of productivity advance should
be reflected, and readily observable, in the rate of decline in price. Prices of
information-processing equipment and software, for example, fell by more
than 4 percent in 2002, but by less than 1 percent at an annual rate by
the first quarter of 2007. Prices of information-processing equipment (and
software) have fallen every quarter since 1991. But declines were especially
Th e withdrawal of shareholder capital from corporations with less promising investment opportunities
for investment in companies with cutting-edge technologies is an important example
of the financing of creative destruction.
472
TH E DE LPH IC FUTU RE
rapid during periods of surging innovation, like 1998, when prospective PC
buyers often hesitated because prices were dropping rapidly; waiting afforded
the opportunity to obtain a cheaper and better PC. The recent slowing
of high-tech price decline is thus further confirmation of a decline in
the availability of new, cutting-edge technological applications that can be
exploited to increase the rate of growth of overall productivity.
As this book goes to press (June 2007), evidence of a rebound in measured
productivity growth or the rate of price decline for high-tech equipment
is lacking. But history tells us that such a turn will take place. It
always has.
Our historical experience strongly suggests that as long as the United
States remains at technology's cutting edge, annual productivity growth
over the long run should range between 0 and 3 percent. As I've noted, since
1870, growth in nonfarm business output per hour has averaged slightly
more than 2 percent per year, which implies that real GDP per hour has
risen slightly less.* The near century and a half of data encompasses periods
of war, crises, protectionism, inflation, and unemployment. I do not believe
it is too great a stretch to assume that the same fundamental forces that
governed the United States over the past two centuries will govern this
country between now and 2030. That 2 percent is probably not a bad approximation
of how fast, on average, humans can advance the frontier of innovation,
and it seems our best forecast for the next quarter century.
But why not higher—say, 4 percent per year or more? After all, in much
of the developing world, annual output per hour has been averaging growth
of far more than 2 percent. But those nations have been able to "borrow" the
proved technologies of the developed world and have not themselves had to
undertake the slow step-by-step effort to advance cutting-edge technologies.
U.S. productivity in 2005 was 2.8 times higher than in 1955. That is because
we knew so much more in 2005 than a half century earlier about how
our physical world operates. Every year, millions of innovations incrementally
improved overall productivity. This process has become particularly
*Much of the GDP excluding nonfarm business is measured by input, not output, and therefore
is implicitly assumed to have no productivity growth.
It is conceivable that by 2030 economists will have devised a new means of measuring an
economy's productivity directly, rather than through its proxy, output per hour.
473
THE AGE OF TURBULENCE
evident since the discovery of the exceptional electrical properties of silicon
semiconductors following World War II. Gordon Moore, a founder of Intel,
suggested in 1965 that the complexity of an integrated circuit, with respect
to cost, doubled every year.* He proved prophetic. The persistent downsizing
of all electronic applications has enabled the large, bulky walkie-talkies
of World War II to morph into today's tiny cell phones, and the boxy original
television tubes and computer display screens to go flat. All machinery output,
from textile looms and motor vehicles to the routers and servers of the
Internet, embodies progressively smaller microprocessors. We have turned
light waves into lasers that, when joined with digital technology, dramatically
improved data and voice communication and helped create a whole
new world of information. It enabled business to adopt just-in-time inventorying,
to lower scrap rates, and to reduce the need for backup employment
to ensure against production and supply snafus.
Yet why hasn't productivity growth been even faster? Couldn't what
we knew in 2005 have been figured out by, say, 1980, thereby doubling the
rate of productivity gains (and increases in standard of living) between
1955 and 1980? The simple answer is that human beings are not smart
enough. Our history suggests that the ceiling on the productivity growth of
an economy over the long term at the cutting edge of technology is at the
most 3 percent per year. It takes time to apply new ideas and often decades
before those ideas show up in productivity levels. Paul David, a professor
of economic history at Stanford, wrote a seminal article in 1989 that addressed
the puzzle of why, in the famous words of Nobel laureate economist
and then-MIT professor Robert Solow, computers were "everywhere
but in the productivity statistics."
It was David's article that heightened my interest in long-term productivity
trends. He pointed out that it often took decades for a new invention
to be diffused sufficiently widely to affect the levels of productivity. As an
*Ten years later, in 1975, Moore revisited his analysis and reported, "I had no idea this was going
to be an accurate prediction, but amazingly enough instead often doublingfs], we got nine
over the ten years." He added that he thought the rate of doubling from then on would slow to
a still-amazing once every two years. Moore's basic insight has now held true for more than
four decades.
474
TH E DE LPH IC FUTU RE
example, he offered the U.S. experience of the gradual displacement of the
steam engine with the electric motor.
Following Thomas Edison's spectacular illumination of lower Manhattan
in 1882, it took some four decades for even half of the nation's factories to
be electrified. Electric power did not fully exhibit its superiority over steam
power until a whole generation of multistory factories was displaced after
World War I. David explains vividly what caused the delay. The best factories
of the day were poorly designed to take advantage of the new technology.
They ran on so-called group drives, elaborate arrangements of pulleys and
shafts that transferred power from a central source—a steam engine or water
turbine—to machines throughout the plant. To avoid power losses and breakdowns,
the lengths of the shared drive shafts had to be limited. This was best
achieved when factories rose vertically, with one or more shafts per floor,
each driving a group of machines.*
Simply substituting large electric motors to power the existing drive
shafts, even when feasible, did not improve productivity very much. Factory
owners realized that electricity's revolutionary potential would require far
more dramatic change: power delivered by wire made central power sources,
group drives, and the very buildings that housed them obsolete. Because electricity
opened the way to equipping each production machine with its own
small, efficient motor, sprawling single-story plants came into vogue. In them,
machinery could readily be arranged and rearranged for greatest efficiency,
and materials could be moved about with ease. But abandoning city factories
and moving to the wider spaces of the countryside was a slow, capital-intensive
process. That was why, David explains, electrifying America's factories
took dozens of years. But eventually millions of acres of one-story plants embedding
electric-motor-driven power dotted America's Midwest industrial
belt, and growth in output per hour finally began to accelerate.
The low-inflation, low-interest-rate period of the early 1960s, as best I
can judge, was owing to the application for commercial use of World War II
military technology, as well as the large backlog of invention built up during
*I recall visiting in the 1960s a tall and narrow stamping plant built at the turn of the century.
I was struck by its unusual shape. But it was only decades later that I learned that I had entered
one of the last surviving relics of a certain aspect of America's industrial history.
475
THE AGE OF TURBULENCE
the 1930s.* Decades later, the delayed emergence of accelerating produc
tivity repeated itself: computers (and the Internet) are now everywhere, in
cluding the productivity statistics.*
Which brings us to our bottom line. Coupled with the projected
0.5 percent annual increase in hours worked between 2005 and 2030 that
follows from the demographic assumptions cited earlier, a slightly less than
2 percent annual average growth in GDP per hour implies a real GDP
growth rate of slightly less than 2.5 percent per year, on average, between
now and 2030. That compares with 3.1 percent per year, on average, over
the past quarter century, when labor force growth was considerably faster.
A
A
rriving at a credible forecast for the level of real GDP for 2030 is a start,
but it doesn't tell us much about the nature of the dynamic that will be
driving U.S. economic activity a quarter century in the future, or about the
quality of our lives. For superimposed on these powerful trends will be the
consequences of an inevitable completion of major aspects of globalization.
At some point, globalization's vast economic migration—the epoch-
making shift of fully half of the world's three-billion-person labor force
from behind the walls of economies that were centrally planned, in part or
in whole, to competitive world markets—will be complete, or as complete
as it can possibly get.
*Low inflation reflected flat nonfarm business unit labor cost, the result of solid growth in pro
ductivity, which in turn was the result of increased investment in, but especially the delayed
application of, the earlier technologies. Professor David demonstrated the extraordinary lag
from technological advance to its consequence in rapidly rising total factor productivity, a
measure of applied technology and other insights. That disinflationary episode lasted only a
few years, coming to an end with the Vietnam military buildup. A much larger continuing dis
inflation was to come as a consequence of the end of the cold war.
tRecent decades' productivity growth derives largely from the continuous improvement and
filling out of networks of interrelated technologies. Innovation renders parts of existing net
works obsolete, as new technologies sprout up to replace them. Efficiency and productivity
improve. But at any point in the process, only part of what is technologically known has had
time to be applied. Purchasing managers year after year consistently identify only half their fa
cilities as embodying state-of-the-art technology. There is always a lot of existing network con
struction in progress, implying that a higher level of productivity will emerge upon its
