had been forty years in the making. It began after World War II with the
development of the transistor, which provoked a surge of innovation. The
computer, satellites, the microprocessor, and the joining of laser and fiberoptic
technologies for communications all helped set the stage for the Internet's
seemingly sudden and rapid emergence.
Business now had an enormous capacity to gather and disseminate
information. This accelerated the creative-destruction process as capital
shifted from stagnant or mediocre companies and industries to those at the
cutting edge. Silicon Valley venture capital firms with names like Kleiner
Perkins and Sequoia and investment banks like Hambrecht & Quist suddenly
achieved great wealth and prominence by facilitating this money
shift. But the financing actually involved, and continues to involve, all of
Wall Street.
To take a more recent example, compare Google and General Motors.
In November 2005, GM announced plans to terminate up to thirty thousand
employees and close twelve plants by 2008. If you looked at the company's
flows of cash, you could see GM was directing billions of dollars
it historically might have used to create products or build factories into
funds to cover future pensions and health benefits for workers and retirees.
These funds, in turn, were investing the capital where returns were most
promising—in areas like high tech. At the same time Google, of course, was
growing at a tremendous rate. The company's capital expenditures increased
nearly threefold in 2005 to more than $800 million. And in the expectation
that the growth would continue, investors bid up the total market
value of Google stock to eleven times that of GM's. In fact, the General
Motors pension fund owned Google shares—a textbook example of capital
shifting as a result of creative destruction.
Why should information technology have such a vast transforming effect?
Much of corporate activity is directed at reducing uncertainty. For
most of the twentieth century, corporate leaders lacked timely knowledge
of customers' needs. This has always been costly to the bottom line. Decisions
were made based on information that was days or even weeks old.
168
IRRATIONAL EXUBERANCE
Most companies hedged: they maintained extra inventory and backup
teams of employees ready to respond to the unanticipated and the misjudged.
This insurance usually worked, but its price was always high.
Standby inventories and workers are all costs, and standby "work" hours
produce no output. They produce no revenue or added productivity. The
real-time information supplied by the newer technologies has markedly
reduced the uncertainties associated with day-to-day business. Real-time
communication between the retail checkout counter and the factory floor
and between shippers and truckers hauling freight has led to shorter delivery
times and fewer hours of work required to provide everything from books
to factory gear, from stock quotes to software. Information technology has
released much of the extra inventory and the ranks of backup workers to
productive and profitable uses.
Also new for the consumer was the convenience of being able to call
up information online, track packages in shipment, and order virtually anything
for delivery overnight. Overall, the tech boom also had a major positive
effect on employment. Many more jobs were being created than were
being lost. Indeed, our unemployment rates fell, from over 6 percent in 1994
to less than 4 percent in 2000, and in the process the economy spawned
sixteen million new jobs. Yet, much as happened with the nineteenth-
century telegraph operators I'd idealized in my youth, technology began
in a major way to upend white-collar occupations. Suddenly millions of
Americans found themselves exposed to the dark side of creative destruction.
Secretarial and clerical functions got absorbed into computer software,
as did drafting jobs in architecture and in automotive and industrial design.
Job insecurity, historically a problem mainly of blue-collar workers, became
an issue starting in the 1990s for more highly educated, affluent people.
This came through dramatically in survey data: In 1991, at the bottom of
the business cycle, a survey of employees of large corporations showed 25
percent were afraid of being laid off. In 1995 and 1996, despite a sharp intervening
decline in the unemployment rate, 46 percent were afraid. That
trend, of course, put job worries squarely in the public eye.
Important, but not as obvious, was the increase in job mobility. Today
Americans change employers on a truly stupendous scale. Out of nearly 150
million people employed in the workforce, 1 million leave their jobs each
169
THE AGE OF TURBULENCE
week. Some 600,000 quit voluntarily, while roughly 400,000 get laid off,
often when their companies are acquired or downsized. At the same time,
a million workers are hired or return from layoffs each week as new industries
expand and new companies come onstream.
The swifter the spread of technological innovation, and the broader its
impact, the more we economists had to scramble to figure out which fundamentals
had changed and which hadn't. Experts in the mid-1990s spent
endless hours debating the so-called natural level of unemployment, for
instance (technically, the Non-Accelerating Inflation Rate of Unemployment,
or NAIRU for short). This is a neo-Keynesian concept that was used
in the early 1990s to argue that if unemployment fell below 6.5 percent,
then workers' wage demands would accelerate, causing inflation to heat up.
So as unemployment trended down, to 6 percent in 1994, 5.6 percent
in 1995, on its way to 4 percent and lower, many economists contended
that the Fed should put the brakes on growth. I argued against this way of
thinking within the Fed and in public testimony. The "natural rate," while
unambiguous in a model, and useful for historical analyses, has always
proved elusive when estimated in real time. The number was continually
revised and did not offer a stable platform for inflation forecasting or monetary
policy, in my judgment. No matter what was supposed to happen,
during the first half of the 1990s wage rate growth held to a low and narrow
range, and there was no sign of mounting inflation. Ultimately it was
the conventional wisdom itself that gave way—economists began revising
the natural level of unemployment downward.
Years later, Gene Sperling told a story of how this controversy played
out in the Oval Office. In 1995 President Clinton's top economic advisers—
Sperling, Bob Rubin, and Laura Tyson—worried that the president
was getting carried away with his hopes for the high-tech boom. So they
enlisted Larry Summers to administer a reality check. As I knew from our
lively breakfast debates, Larry was a technology skeptic. And he normally
weighed in with the president only on international issues, so Clinton
would realize this was an unusual event.
The economists trooped into the Oval Office, and Summers did a short
presentation on why tightness in the labor market meant growth would
have to slow. Then the others chimed in. Clinton listened for a while, then
170
IRRATIONAL EXUBERANCE
finally interrupted. "You're wrong/' he said. "I understand the theory, but
with the Internet, with technology, I can feel the change. I can see growth
everywhere." The fact was, Clinton wasn't relying solely on instinct. He'd
been out talking to CEOs and entrepreneurs, as he always did. Politicians
never want to believe that there are limits to growth, of course. But at that
moment the president probably had a better hands-on feel for the economy
than his economists.
Both the economy and the stock market continued to boom. Output
as measured by GDP grew at a superhot rate of over 6 percent in the spring
of 1996—calling into question another chunk of conventional wisdom,
namely that 2.5 percent was the maximum growth the U.S. economy could
healthily sustain. We were doing a lot of rethinking at the Fed. It's easy to
forget the speed with which innovations like the Internet and e-mail went
from exotic to ubiquitous. Something extraordinary was happening, and the
challenge in trying to figure it out as it was happening, in real time, was
considerable.
By the time I convened the FOMC on September 24, 1996, eight
months and seven meetings had passed since we'd last lowered the interest
rate. Many committee members were now leaning the other way, toward
an increase so as to preempt inflation. They wanted to take away the punch
bowl again. Corporate profits were very strong, unemployment had dropped
to well under 5.5 percent, and one big factor had changed: wages were finally
rising. Under boom conditions like these, inflation was the obvious
risk. If companies were having to pay more to keep or attract workers, they
might soon pass along that added cost by raising prices. The textbook strategy
would be to tighten rates, thereby slowing economic growth and nipping
inflation in the bud.
But what if this wasn't a normal business cycle? What if the technology
revolution had, temporarily at least, increased the economy's ability to expand?
If that was the case, raising interest rates would be a mistake.
I was always wary of inflation, of course. Yet I felt certain that the risk
was much lower than many of my colleagues thought. This time, it wasn't
a case of upsetting conventional wisdom. I didn't think the textbooks were
wrong; I thought our numbers were. I'd zeroed in on what I believed to be
the primary riddle of the technology boom: the question of productivity.
171
THE AGE OF TURBULENCE
The data we were getting from the Commerce and Labor departments
showed that productivity (measured as output per hour worked) was virtually
flat in spite of the long-running trend toward computerization. I could
not imagine how that could be. Year in and year out, business had been
pouring vast amounts of money into desktop computers, servers, networks,
software, and other high-tech gear. I had worked with a sufficient number
of plant managers on capital investment projects over the years to know
how such purchasing decisions were made. They would order expensive
equipment only if they believed the investment would expand their production
capacity, or enable their employees to produce more per hour. If
the equipment failed to do at least one of these, the managers would stop
buying. Yet they'd kept pumping money into high tech. This became evident
as early as 1993 when new orders for high-tech capital began to accelerate
after a protracted period of sluggish growth. The surge continued into
1994, suggesting that the early profit experience with the new equipment
had been positive.
There were other, even more persuasive indications that the official productivity
figures were awry. Most companies were reporting rising operating
profit margins. Yet few had raised prices. That meant that their costs per unit
of output were contained or even falling. Most consolidated costs (that is, for
business considered as a whole) are labor costs. So if labor costs per unit of
output were flat or declining, and the rate of growth of average hourly labor
compensation was rising, it was an arithmetical certainty that if these data
were accurate, the growth of output per hour must be on the rise; productivity
was truly accelerating. And if so, then rising inflation would be unlikely.
Even though I was sure my analysis was right, I knew better than to try
to convince my colleagues on the basis of back-of-the-envelope figuring. I
needed something more persuasive. With a few weeks to go before the
September 24, 1996, meeting, I asked the Fed staff to pull apart the federal
productivity statistics and study the underlying data, industry by industry,
for dozens of industries. I had been bothered by an apparent discrepancy
between the Bureau of Labor Statistics data on output per hour for all
nonfarm industries and a separate estimate for corporations. Matching the
two implied that there was no productivity growth in noncorporate America,
an unlikely conclusion.
172
IRRATIONAL EXUBERANCE
When I asked for the detailed industry breakdown, typically the staff
would joke that the chairman wanted "embellishments and enhancements."
This time they said it was more like I was asking them to undertake the
Manhattan Project. Nevertheless, they burrowed into the data and rendered
their report just in time for the FOMC meeting.
On that Tuesday, opinion in the committee was divided. Half a dozen
members wanted to raise rates right away—as Tom Melzer, the hawkish
president of the Federal Reserve Bank of St. Louis, put it, "to take out an
insurance policy" against inflation. Others were on the fence. Alice Rivlin,
who was now in her third month as vice chairman of the Board of Governors,
sized up the situation in her usual droll manner: "The worried faces
around this table, I think we should remind ourselves, are worrying about
the best set of problems we could think of having. Central bankers all
around the world would wish for this set of statistics." While she agreed
that we were in an inflationary "danger zone," she also pointed out that "we
have not seen higher inflation yet."
When it was my turn to speak, I came on strong using the staff's report.
It seemed that the government had been underestimating productivity
growth for years. It had not, for example, found any efficiency gains in the
service economy—in fact, the government's calculations made it appear
that productivity there was shrinking. Every committee member knew that
was absurd on its face: law firms, business services, medical practices, and social
service organizations had been automating and streamlining themselves
