View original document

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

Productivity Puzzles
University of Tennessee
Martin, Tennessee
October 26, 1999


lmost everyone is aware by now of
the fact—and I believe it is a fact—
that U.S. productivity growth has risen
substantially. The painful period of
unusually slow productivity growth in the 1970s
and 1980s is behind us. The increase in output
per hour of labor input has been high enough
over the last few years that it is increasingly
reasonable to believe that the United States has
indeed turned the corner on productivity growth.
This picture is reinforced by extensive anecdotal
reports from across the country.
I have been struck, however, by the extraordinary degree of uncertainty we face on this subject.
My purpose today is to outline the puzzles of the
1970s slowdown in productivity growth slowdown and to discuss more extensively the puzzles
surrounding the rising productivity growth of the
1990s. In thinking about these puzzles, common
sense and basic economic reasoning are a good
deal more helpful than hype about the “new
Before I get into these issues, it is important
that I issue a disclaimer. The views I express here
are my own and do not necessarily reflect official
positions of the Federal Reserve System. I’ve had
a lot of help with these remarks from colleagues—
especially Joe Ritter—in the research department
of the St. Louis Fed; they deserve credit for the
strengths of my argument. I retain responsibility
for the errors.

The importance of productivity growth is
easy to demonstrate. The simplest productivity

statistic is output per hour of labor input, usually
called “labor productivity.” During the 1950s
and 1960s, labor productivity in the nonfarm
business sector grew by about 2.8 percent per
year. At that rate, productivity doubles in about
25 years. From 1973 to 1990, labor productivity
grew at a rate of only 1.04 percent per year. At
that rate, it takes 67 years for output per hour to
double. Currently, it appears that output per hour
is growing at a rate of about 2 percent per year,
which doubles in 35 years. Even a small amount
of extra growth yields astonishing gains for the
United States. With an extra quarter percentage
point of labor productivity growth, GDP would
be about $300 billion higher after a little more
than 10 years. The impact on the federal budget
alone would be a shift toward surplus on the
order of $60 billion.
Because the growth in real wages and, therefore, the standard of living depends on productivity growth, the time it takes for productivity
to double at various growth rates translates quite
easily into per capita income. It makes an enormous difference to our society whether income
is doubling every 25 years or every 67 years. Individuals, and society as a whole, are much better
off when the median-income family can enjoy a
standard of living that the upper-income family
enjoyed a generation or two earlier.

Let me explain why this topic concerns me
as a monetary policymaker. A proposition universally accepted by monetary economists is that


monetary policy has relatively little to do with
long-term economic growth, as long as the inflation rate remains modest. I believe that low inflation is better than not-so-low inflation, but I am
not one who makes the extravagant claim that zero
inflation yields enormous benefits over some
modest rate of inflation. Monetary policy can
contribute to general economic stability; and a
stable, less cyclical economy probably raises longterm growth somewhat. Central banks also make
valuable contributions to the efficiency and safety
of the payments system, which is an essential
piece of infrastructure for a modern economy.
Still, as important as these central bank
responsibilities are, it is clear that the central
government’s activities have far more to do with
growth than anything the central bank does. The
soundness and efficiency of the legal system, the
degree of safety of citizens, tax policy, government
spending, and regulation—all these affect productivity growth to a vastly greater degree than central
bank policy.
Productivity growth, however, is terribly
important to monetary policy in a different way.
Here is the monetary policy issue as I see it. If we
knew how to set the rate of inflation directly, then
we should just choose a zero rate and be done
with it. (My guess is that zero inflation, properly
measured, translates to something like a 1 to 11/2
percent annual increase in the consumer price
index as the Bureau of Labor Statistics constructs
that index today.) But the Fed can’t set the rate of
inflation directly; that is not possible in a marketbased economy.
So, the Fed has to work indirectly. In a broad
sense, we alter the inflation rate by controlling
growth in the stock of money, though, from
meeting to meeting, the Federal Open Market
Committee focuses on the federal funds rate. But
when is the federal funds rate too low, leading to
too much money growth and, in turn, to inflation?
If the U.S. economy were static, we could just
experiment a bit and find the right number (as
many macroeconomics textbooks imply). But in
the real world, figuring out the right level for the
federal funds rate is a tough issue. Among other
things, we need to track actual GDP growth against

the economy’s underlying growth potential. Three
percent GDP growth could be sustainable, or it
could be a harbinger of accelerating inflation, and
the answer might change from quarter to quarter.
The answer depends, to a large degree, on the
underlying growth of productivity.
High-frequency economic data are not accurate enough, comprehensive enough, or timely
enough to answer the question with any degree
of reliability. Thus, there is far less science behind
our decisions than I would like. I think I can safely
say that every member of the FOMC would like
to feel more certain about when and how much
to adjust the intended federal funds rate. The
bottom line, though, is that in the course of fulfilling our FOMC responsibilities, we have to
judge the probable strength or weakness of the
economy. We want the economy to grow as fast
as its resources and productivity permit; thus,
ongoing evaluation of underlying productivity
growth trends is an important part of the art of
making monetary policy.

Adam Smith, in his Wealth of Nations published more than 200 years ago, was the first to
argue with clarity that a nation’s wealth was in
the output of its people, not the gold in its vaults.
And Smith certainly understood the tremendous
importance of productivity growth; he sought to
convince his readers that competitive markets
generated wealth and that restrictive government
policies made England poorer.
Since Smith’s day, we’ve filled in some of
the details on how the economy grows and have
amassed a huge amount of empirical information.
We have not, however, improved upon Smith’s
fundamental framework for understanding economic growth. As with so many things, Smith had
it right.
Let me put Adam Smith’s analysis into modern language: There is broad agreement among
economists that the main factors that enable an
economy to grow are

Productivity Puzzles

• the growth of the quantity of labor input;
• the growth of the quantity of capital input;
• the rate of improvement in the processes
that turn inputs into outputs.
The only amendment flowing from advances
in economic knowledge this century—and it is an
important amendment—is our greatly increased
understanding of the importance of human capital.
It’s not hard to understand that the total value
of what an economy produces will increase if the
number of people working increases or if some
people acquire human capital through education
or on-the-job learning. Similarly, providing workers with more physical capital will increase their
output; a worker can dig a longer ditch in a day
with a backhoe than with a shovel. Economists
have a pretty good handle on these things, both
conceptually and quantitatively.
The mystery lies in that third category,
“improvements in processes.” One might call it
“technological progress” or “innovation,” but that
does little more than rename it. It‘s not the same
thing as labor productivity. In fact, the outputper-hour data combine the effects of all of these
factors except labor hours.
It’s helpful to keep this framework in mind,
because people often talk about productivity
growth as though it’s just technological progress.
That’s partly because we have no direct way to
measure the contribution of that third category
other than by subtracting the contributions of
increased quantities of labor and capital from output. That exercise gives us the residual category
that economists call “total factor productivity.”
What ends up in that residual category? Well,
it’s a little like art—we know it when we see it.
Total factor productivity soaks up the effects of
everything from rearranging a warehouse so that
popular items are near the loading dock to sweeping changes introduced by innovations like electricity or computers. It shouldn’t surprise us that
it is difficult to measure the contents of the pigeonhole where we dump the effects of fuzzy but profound concepts like creativity and innovation.
So where has U.S. growth come from? First
the big picture—the past 50 years. By 1997, out-

put in the private business sector was five times
its 1948 level. Increasing quantities of labor and
capital each accounted for roughly 30 percent of
that increase, leaving about 40 percent of postwar
growth “explained” by growth of this mysterious
residual category we call total factor productivity.
Growth of the labor force washes out of the
labor productivity statistics—it affects both the
numerator and denominator of output per hour
by the same percentage. So something like 70
percent of postwar growth has “come from” the
growth of labor productivity. But keep in mind
that labor productivity is not an independent
economic force: It measures the combined effects
of investment, learning, and innovation.
Before I dig into the puzzles, I’ll digress briefly
to make an observation that illustrates very nicely
why productivity is a complicated and difficult
subject. In one of his recent books, Stephen Jay
Gould takes up the question of why the .400 batting average has disappeared—nobody has accomplished this feat in the major leagues since 1941.
Mark McGwire, Sammy Sosa, and home runs
aside, are batters less productive than they used
to be? “The problem,” Gould writes, “seems so
obvious in outline: something terrific, the apogee
of batting performance, was once reasonably
common and has now disappeared. Therefore,
something profoundly negative has happened in
baseball.” But Gould doesn’t really believe that;
he spends the next 50 pages expanding on the
following idea: The .400 average is not a part of
the game itself, but a very simple statistic produced by a complex and dynamic system, major
league baseball. Since 1941, rules, pay scales,
training regimens, schedules, and stadiums have
all changed. Imagine the difficulty of trying to pin
down exactly how, and to what extent, each of
these factors contributed to the decline in batters’
averages. And baseball is surely simple compared
to the entire U.S. economy!

The fact that labor productivity accounted
for about 70 percent of growth over the last half


century hides one of the most important and
longest-running stories in macroeconomics, the
productivity slowdown that started around 1970.
Starting in the early 1970s, the trend rate of productivity growth fell by almost 60 percent in the
nonfarm business sector. After 1975, it gradually
became apparent that productivity growth had
slowed to a crawl, compared with the rapid pace
of the 1950s and 60s. Some experts flagged the
changed trend quickly; others insisted for several
years that lower productivity growth was likely
a temporary phenomenon. Not until 1978 or so
did most economists agree that something serious
had in fact taken place.
Economists have debated the causes of the
slowdown for years, and the issue is still unresolved. Part of the slowdown was clearly due to
slower capital accumulation, but that only pushes
the question back a step. In any case, slower capital accumulation was certainly not the whole
story; total factor productivity—that residual
category—slowed dramatically as well.
Some people are convinced that the explanation lies in the energy crises of the 1970s; some
believe that a policy environment unfriendly to
business bears much of the blame. Others point
to the higher inflation rate of the 1970s, and still
others to environmental controls. We have more
theories than data points. To this day, the decline
in productivity growth that occurred after 1972
is a puzzle.

The second set of productivity puzzles has
been unfolding for the past several years. They
are summarized in the famous quip by Robert
Solow: “You can see the computer age everywhere
but in the productivity statistics.”
Did the productivity slowdown truly end in
the 1990s? At first glance, the answer appears to
be “partly.” The last few years have certainly seen
stronger labor productivity growth. But if we look
at the data in more detail, things don’t look as
reassuring. It appears that growth of total factor

productivity has not increased. That means the
higher growth of labor productivity in recent
years primarily reflects the investment boom of
recent years—more capital—but not a higher
growth rate of total factor productivity. The slowdown in the 1970s showed up on both dimensions—capital accumulation and total factor
productivity—but our productivity recovery
seems to be reversing only one dimension of the
But wait. It gets worse. The deeper we dig,
the more puzzles we find, and computers are at
the center of these puzzles. In an effort to understand what is going on, productivity experts drill
down into the data. The data I’ve been discussing
so far refer to the entire economy except for the
government sector, for which no overall productivity measures are available.
For the manufacturing part of the total business
sector, it is apparent that most of the productivity
slowdown has evaporated. In manufacturing, both
labor productivity and total factor productivity
have been growing rapidly for several years. But,
and this is truly astonishing, Robert Gordon in
his recent research argues that within manufacturing, after allowing for normal cyclical effects,
almost all of the labor productivity growth has
been in the sector that produces computers. When
you take out the durables manufacturing sector,
where computers come from, and look at what’s
left, productivity growth looks downright tepid.
In other words, there is productivity growth where
computers are made, but not where they are used.
The first part of this is entirely credible—we don’t
need government statisticians to tell us that productivity gains in the electronics industry have
been astonishing.
To understand just how amazing this puzzle is,
I’ll put the point this way: All, or most—depending on your choice of expert—of the increase in
productivity growth for the entire U.S. economy
can be attributed to a single industry—computer
manufacturing—that amounts to about 11/4 percent of the economy! That is a truly remarkable
finding, and it really doesn’t matter much whether
the truth is “all” or “most.”
The idea that the recovery of productivity
growth has been lackluster among computer-using

Productivity Puzzles

firms doesn’t seem right, though, does it? We see
productivity improvements all around us. And
we see innovation, not just the capital deepening
implied by the investment boom. There is a
more technical way to express this discomfort:
Remember that labor productivity combines the
effects of capital investment and innovation.
Since we’ve been in an investment boom, slow
labor productivity growth implies that total factor
productivity is completely stalled or going backwards! This just does not make sense. Furthermore, why would businesses invest billions of
dollars in computers that don’t increase productivity? That is, why should businesses now be
investing so heavily in information technology
if the rate of return to such investments were no
higher than in the 1970s and 1980s? If the rate of
return is higher today, that higher return should
show up in total factor productivity.
So here is the crux of our second productivity
puzzle: Outside manufacturing, productivity
growth, as measured by government statistics,
looks slow. Within manufacturing, productivity
growth is concentrated in the manufacturing of
computers. Although the workers who produce
computers have become immensely productive,
the overall productivity data don’t seem to support
the idea that computers enhance productivity
growth. But this conclusion doesn’t seem consistent with businesses choosing to invest in computers. And the overall picture doesn’t seem
consistent with a thriving, healthy economy.
There are two ways to interpret this discrepancy. A hard-core data hound might conclude
that the numbers are correct; manufacturing, in
general, and durables manufacturing, in particular,
really has been more innovative—streamlining
production processes and so forth. There is probably some truth to that, but if it’s the whole story,
the rest of the private sector is doing very badly,
indeed. As I said, my observations suggest that
innovation and improved productivity are all
around us—in manufacturing and elsewhere. It
is useful to remember that it took economists
almost a decade to recognize the productivity
slowdown of the 1970s; the data are very noisy
and very cyclical, making it difficult to extract

trends from small numbers of observations. It’s
not inconceivable that it will take us a long time
to be sure of a turnaround in the 1990s.
A second angle on these numbers is to think
about whether the measurement of productivity
is distorted. Zvi Griliches, who is one of the leading researchers in this area, argues that the part
of the economy he calls “reasonably measurable”
has declined from about half to less than 30 percent since World War II. The problem is that much
of the economy produces things that are extremely
difficult to measure, and the share of this sector—
services, broadly speaking—keeps growing. Moreover, the productivity slowdown appears to be
persisting in these difficult-to-measure industries.
Griliches’ bottom line is that outside of sectors
like agriculture and manufacturing, where it’s
more or less possible to count things in order to
measure output, we should be extremely suspicious of productivity numbers.
Beyond this conceptual problem, it’s no
secret that statistical agencies have a hard time
keeping up with innovations in the economy. In
particular, it takes a while to figure out how to
measure new kinds of output. Just this month, in
fact, the Commerce Department has significantly
updated how it measures the contribution of software to the U.S. economy. So, in a broad sense, it
shouldn’t surprise us very much that output and
productivity statistics are a bit slow to capture
rapid innovation.
As you can see, we policymakers have a real
problem. We have to decide how much weight to
give to our eyes and ears and how much to formal
statistics. We know the statistics may be misleading, but we also know that the discipline imposed
by statistics is our main bulwark against eyes that
look through rose-colored glasses and ears that
turn to tin when they hear what we do not want
to hear—wishful thinking of all sorts.

I’ve talked a bit about the statistics, and so
let me now turn to our eyes and ears and how we
might best think about what we see and hear. We


see new electronic technology all around us. News
stories about the Internet are incessant. One would
surely be justified in suspecting that all of this
represents a productivity revolution of sorts. And
I am partly sympathetic to this view. Do I think
then, that the productivity data have nothing to
tell us? Do I think that, despite the evidence, we
are seeing the birth of some sort of “new economy,”
beyond the bounds of historical experience and
the laws of economics? Hardly. In fact, I think
that history and sensible economics can tell us a
lot about the future of productivity.
First, it’s my belief that the policy environment is considerably more hospitable to the economic activities that generate productivity growth
than it was in the 1970s. There is wider understanding than in the past that bad policy—tax
policy, financial policy, environmental policy,
trade policy—can profoundly affect a firm’s incentive to invest in productive assets. Too often in the
past, conflicting policy goals have been resolved
without regard for economic efficiency. Although
I think we still have a long way to go in this regard,
today we are more likely to see innovative policies like tradable pollution permits. This kind of
market-based approach is far less damaging to
productivity than the style of regulation that says
simply, “Thou shalt not pollute more than 3 parts
per billion.” In short, we have learned a lot about
how to protect the environment at less cost than
in the past.
Turning to the innovation process itself, we
should start by recognizing that, in the macroeconomic sphere, revolutions take decades. Most
people call that evolution. I believe that information technology will be a genuine engine of growth
for decades, but there hasn’t been and won’t be a
sudden swerve toward some sort of “new economy.” It is important to understand that some of
the new technologies reduce the productivity of
older technologies. For example, for the economy
as a whole, productivity growth will reflect the
high productivity of Internet commerce and the
declining productivity of the huge existing investment in bricks and mortar and people engaged
in traditional retailing.

The history of “general-purpose technologies,”
as economists call them, tells us why evolution is
a better word than revolution. One such generalpurpose technology, electricity, has been studied
extensively by economic historian Paul David.
According to David, less than 5 percent of
mechanical power in the nation’s factories was
provided by electric motors in 1899. It took about
20 years for that number to rise to 50 percent.
David addressed two interrelated questions about
the spread of electricity use in general and the
use of electric motors in particular. First, why
was the adoption rate so low? Second, what is
special about the spread of a general-purpose
technology like electricity?
Think about what a factory was like before
the electric motor. There was typically a single
source of mechanical energy: a water wheel or,
later, a steam engine. This energy had to be distributed around the plant by mechanical means—
gears, drive shafts, belts, and pulleys. Because of
the number of interconnected moving parts, this
system was expensive to build, inflexible, and
dangerous. But the initial expense was a sunk cost,
and once in place the system didn’t cost much to
run. So in most cases it didn’t make sense to scrap
an old plant until it physically wore out, even
though the new technology was markedly superior. Electric motors spread rapidly in industries
that were expanding, but elsewhere the old technology continued to prevail. There is a tremendous
amount of inertia in this sort of thing. That is the
first lesson about the spread of technology: It’s
simply too expensive for an industrial economy
to rearrange its production and scrap a large part
of its capital stock overnight, no matter how
exciting the new technology is.
But the ramifications of the electric motor for
manufacturing, when it was finally adopted, were
immense. Mechanical energy didn’t have to be
distributed from a central source; you just put a
motor where you needed it, so you could easily
reconfigure the production process. The production process could be physically stretched out,
allowing the development of a true assembly line.
New factory structures needed only to keep the
rain off; they no longer needed bracing for heavy,

Productivity Puzzles

rapidly moving power-distribution machinery.
Maintenance could be performed on a single
machine, without shutting down the entire factory.
All of these advantages were clear in principle
at the turn of the century, but each business had
to figure out how to adapt the technology to its
needs (as well as needing to amortize older investments). Thus, though the impact of electricity on
manufacturing and daily life was profound, it
was spread over many years.
A more recent example illustrates a slightly
different theme. When the laser was invented in
1957, no one recognized it as a general-purpose
technology. Indeed, Bell Labs didn’t even patent
its invention. For some years, the laser was
regarded as an extremely specialized tool. In fact,
it was biding its time, waiting for complementary
developments in the semiconductor industry.
When inexpensive semiconductor lasers became
available, the laser became ubiquitous. Although
we tend to connect the laser with gee-whiz inventions and weapons, probably its most profound
effect on the U.S. economy has been via the
humble bar-code reader, which was not practical
before cheap lasers. This innovation has altered
the economic landscape in retail stores, libraries,
the post office, even Red Cross blood collection.
Virtually anywhere we need to keep track of the
movement of physical objects, you’ll see bar codes
of one sort or another.
Of course, cheaper and cheaper computing
power enables wider spread of bar-code scanners,
just as bar-code scanners allow businesses to bring
computing power to bear on inventory control,
marketing, and sales. Who’d have imagined that
their combined power would be most visible in
the grocery checkout lane? That’s the second big
lesson about technological change that I take from
economic history: It’s hard to predict the biggest
effects until you’re right in the middle of them.
Let me give you a final example, directly
related to the Internet. A few months ago the
New York Times ran a long story about difficulties
faced by the online sales operations of Recreational
Equipment Inc., an outdoor equipment retailer.
REI’s online operation has been profitable, though
not wildly so. An Internet retailer that makes

money is a peculiar animal, but the real reason
I found this story interesting is that very little of
it had anything to do with the Internet, per se.
Instead, the things that REI has found difficult
are problems any retailer would understand.
Already, the problems of fast servers and broadband connections are looking similar to the problem of renting a store that has electrical service.
The real problem is figuring out what to do with
them. REI, for example, found that pictures taken
for its print catalog didn’t work well on the
Internet. To me this sounds more like older problems such as learning how to sell things in malls
than like a “new economy.”
I don’t feel as curmudgeonly about new technology as that sounds. Today we are in the middle
of the adoption cycle for a remarkable set of technological innovations in microprocessors and
communications. It is difficult to believe that
these things will not spur economic growth. But
let’s not kid ourselves: We have yet to figure out
what to do with all of this computer power and
the Internet, and it’s going to take time to figure
out what works and what doesn’t. In effect, we
must write the economic software for this technology. That will take a long time, and we won’t
understand how it has shaped our economy until
it has already happened. That seems to be the
way these things have always been.
What is the bottom line for U.S. growth
prospects? Optimists and pessimists among serious students of economic growth are not as far
apart as the popular press would have you believe.
Pessimists believe that the recent higher productivity growth reflects a transitory cyclical phenomenon and that the underlying or trend growth of
labor productivity remains bogged down at about
the level of the 1970s and 1980s. That rate is in
the range of 1 to 11/2 percent per year and yields
likely trend GDP growth of about 21/2 percent per
year. Optimists believe that the growth rate of
productivity has risen to the 21/2 to 3 percent
range, which translates into average GDP growth
in the 3 to 4 percent range over the next few
years. Although I am an optimist on growth, my
instincts as a policymaker compel me to concentrate on the midrange of informed opinion—pro7


ductivity growth of about 2 percent per year and
trend GDP growth of about 3 percent per year.
That to me is the appropriate basis for monetary
policy decisions. But I do want to emphasize the
importance of the word “about” in these judgments. With all I’ve said, it should be clear that
analysts should not lock in their view of these

Productivity growth is a terribly important
subject for the United States, indeed for every
country. But this is not an easy subject. The puzzles of the 1970s slowdown in productivity growth
have not been resolved. The puzzles of the 1990s
increase in productivity growth seem only to
deepen with further research into what is going on.
It is amazing, but still a fact at this time, that most
of the reported increase in productivity growth
can be attributed to the computer-producing
industry and little to the computer-using industries—that is, the whole rest of the economy.
The central bank is really a bit player in the
growth process, provided inflation is kept rela-


tively low. Nevertheless, the Federal Reserve, in
setting the intended federal funds rate, cannot
avoid making some judgments about the productivity growth.
The crude state of our knowledge about current and future productivity trends is uncomfortable for me as a policymaker. I see little reason for
pessimism, however. My best judgment is that the
productivity slowdown of the 1970s and 1980s
is over. However, we have to be realistic about
the magnitude of the improvement. With all of
the optimism that so marks U.S. culture, and with
our satisfaction about the fine performance of
the economy in recent years, we must not allow
ourselves to be lulled into wishful thinking about
productivity and economic growth. Even modest
improvements in productivity growth are important, particularly if they can be sustained for the
long run. And we certainly do not want our eyes
and ears and heads to be closed to the possibility
that trend productivity growth might now be high
enough that economic historians may in time
refer to the 1990s as the beginning of a new age.
What a wonderful outcome that would be.