The Speed Limit

Say it ain't so, Joe," a young boy is reputed to have implored Shoeless Joe Jackson, the tarnished star of the infamous Chicago "Black Sox" in 1919. That same cry went up in February 1997 when the Council of Economic Advisers, then headed by Joseph Stiglitz, pegged the U.S. economy's long-term growth rate at only 2.3 percent per annum. "Say it ain't so, Joe."

Like the young baseball fan, growth optimists from both the left and the right of the political spectrum do not want harsh realities intruding upon and ruining their dreams. But Joe Jackson couldn't say it wasn't so, because it was. And neither could Joe Stiglitz.

Based on some simple arithmetic that I will display shortly, mainstream economists are exceptionally united right now around the proposition that the trend growth rate of real gross domestic product (GDP) in the United States—the rate at which the unemployment rate neither rises nor falls—is in the 2 percent to 2.5 percent range. In fact, a central aspect of the allegedly sharp "controversy" between the Clinton administration and the Congressional Budget Office (CBO) over the appropriate "economic assumptions" to use in budget projections is whether that growth rate is 2.3 percent or 2.1 percent. (The CBO favors the latter.) That's some controversy.

Nevertheless, a number of leading politicians, influential business executives, and popular writers refuse to accept this "pessimistic" conclusion. We could grow much faster, they insist, if only the government would pursue more growth-oriented policies. Actually, the argument comes in two distinct variants, which often get confused but ought to be kept separate.

One variant makes a cyclical claim: that overly tight monetary policies are holding back the economy. According to critics from both the left and the right, a niggardly Federal Reserve keeps pulling the U.S. economy up short of its potential by hitting the monetary brakes prematurely, perhaps because it is seeing inflationary ghosts. The implication is that there is considerable slack left in the system, so we could grow much faster for a while by bringing unused resources into production.

The second variant makes a claim about long-run growth instead: that trend growth either is or could be much faster than the conventionally estimated 2 to 2.5 percent. We could exceed this false "speed limit" forever, the argument goes, by adopting more capitalist-friendly tax and regulatory policies. Normally, this critique comes from the right. In the 1996 presidential campaign, for example, Bob Dole claimed that his economic plan, which featured a large income tax cut, would boost trend growth to 3.5 percent per year. His running mate, Jack Kemp, went even further, asserting that those policies would double the growth rate. Double? To 5 percent? Now there's a real Shoeless Joe Jackson fan.

Mainstream economists dismiss both of these arguments as mostly poppycock. This article explains why. But let me first state unequivocally that no one wishes more than I that both arguments were true. I also wish the Dodgers would return to Brooklyn and that the world were free of want. Unfortunately, none of these lovely things is in the offing.

Subscribe to The American Prospect


First, the cyclical argument. On this view, lower interest rates could boost growth for a while without raising inflation, so the Fed should ease up and give the economy more running room. Is it true? Let's examine some pertinent facts.

The unemployment rate is by no means a perfect indicator of aggregate pressure on capacity, but it is about as good as we have. (Essentially the same story could be told using the Federal Reserve's capacity utilization index instead.) It declined from 7.7 percent in June 1992 to 5.8 percent by September 1994, and then fluctuated within a narrow range between 5.8 percent and 5.4 percent until June 1996, when it dropped another notch. Unemployment averaged just 5.3 percent between then and March 1997. The April and May 1997 readings came in below 5 percent, the lowest levels since 1973.

Let's consider what happened to inflation during each of those three episodes. From June 1992 through August 1994, the unemployment rate was continuously above 6 percent, averaging 6.8 percent, and inflation drifted lower. (See "Inflation and Unemployment Rates Over the Past Decade".) Then, from the summer of 1994 to the summer of 1996, unemployment averaged 5.6 percent, and inflation remained remarkably stable. That experience nominated 5.6 percent as an excellent estimate of the so-called NAIRU (an awful acronym for the "Non-Accelerating Inflation Rate of Unemployment"—in plain English, the rate at which inflation neither rises nor falls).

Inflation and Unemployent Rates Over the Past Decade

Graph--inflation and unemployment rates
Source: Bureau of Labor Statistics

Since inflation fell when unemployment averaged 6.8 percent and stabilized when unemployment averaged 5.6 percent, it stands to reason that unemployment much below 5.6 percent should make inflation rise. The last time we experienced unemployment that low for a protracted period of time was 1988 to 1990. The unemployment rate first dipped below 6 percent in the fall of 1987, fell all the way to 5 percent by March 1989, and did not climb back to 6 percent again until November 1990. During that period, core inflation increased from about 4 percent to about 5.5 percent. Thus the theory seemed to work pretty well.

Very recent experience has, however, been somewhat more favorable than history would suggest—as it has been remarked by many. Specifically, inflation appears to have inched downward even as unemployment fell below 5.4 percent. But this period is less than a year long—far too short to draw any conclusions, given the sluggish response of inflation to tight markets. For example, a standard rule of thumb suggests that if the NAIRU is 5.6 percent, one year of 5.2 percent unemployment should boost the inflation rate by 0.2 percent. Such a minuscule rise in inflation is simply too small to pick up in the data. And furthermore, improvements in the Consumer Price Index (CPI) have reduced measured inflation at precisely the same time. To be sure, recent experience does offer some hope that the NAIRU might be below 5.5 percent. But to conclude so at this point would be premature.

Finally, let us ask what the miserly Fed was doing while all this was going on. During the roughly two-year period of 5.6 percent unemployment—a number that most observers in 1994 thought to be below the NAIRU—the Fed first raised its overnight interest rate by 1.25 percentage points (in November 1994 and February 1995) to slow the economy, and then lowered it by 0.75 percentage points (in July 1995, December 1995, and January 1996) to give the economy a little boost. Thereafter, until March 1997, it sat on its hands while the unemployment rate not only remained low but actually drifted down. That does not sound like Scrooge-like behavior to me. Rather, the Fed was cautiously (and perhaps accidentally) probing how low the NAIRU might be.

Was the Fed really holding growth below its potential? Remember, trend growth is defined as the rate that is just enough to absorb the normal year-to-year increase in the labor force, leaving the unemployment rate neither rising nor falling. Let's look at the numbers. From the third quarter of 1994 through the first quarter of 1997, the annual growth rate of real GDP averaged 2.6 percent. During those two and a half years, the unemployment rate dropped by about 0.7 percent—which means that 2.6 percent growth is somewhat above trend.

To claim that we could have grown much faster without higher inflation is to claim that unemployment could have fallen much more than it did without rising inflation. Again, let's look at the numbers. Had growth averaged 3.5 percent instead of 2.6 percent, the unemployment rate would now be below 4 percent—a rate last seen during the Vietnam War inflation. No serious student of the U.S. labor market believes we have this much unutilized labor. Indeed, anecdotal evidence from around the country supports the conventional view that labor markets are now very tight.

In sum, it is highly unlikely that our economy has much running room before it reaches its normal, full-employment capacity. In fact, it is more likely that we have already passed this benchmark. Do economists know that with 100 percent certainty? Certainly not; this is not physics. But, if you are planning to bet against it, insist on long odds.



Now let's turn to the bigger issue: estimates of the economy's long-run growth trend. To begin with a "top-down" estimate, recall that unemployment fell while GDP growth averaged 2.6 percent. That means that the trend must be below 2.6 percent per annum. How much below? The simplest way to estimate the trend is to calculate real GDP growth between any two years with equal unemployment rates. (Quarters will also do, but using years rather than quarters helps smooth over blips in the data.) The most recent such example is the period from 1990 to 1995. Over that five-year period, real growth averaged 1.9 percent per annum.

A "bottom-up" calculation yields a similar estimate. The laws of arithmetic dictate that you get the trend growth rate of output by adding up the growth rates of labor input and labor's productivity—that is, output per hour of work. Conventional estimates place each number around 1.1 percent per annum. So the estimated growth trend is 2.2 percent, give or take a tenth or two.

Where could such a calculation go wrong? There cannot be much dispute about the trend growth rate of the labor force, which is known within a small margin of error. Notice, by the way, that the labor force expanded about 1.7 percent annually in the 1980s and 2.7 percent in the 1970s—which is one reason why we can no longer grow as fast as we did then. (The huge expansion of female labor force participation, for example, cannot be repeated.)

So the argument between a 2.2 percent growth trend and a 3.5 percent growth trend cannot be about labor force. It must be about productivity growth. And so it is. The evidence for the mainstream view is neatly summarized in "The Slowdown in Productivity Growth" (below), which is adapted from the 1997 Economic Report of the President. It shows that output per hour in the U.S business sector grew much more slowly after 1973 than before—just 1.1 percent per annum for 23 years. A prudent forecaster would extrapolate that behavior, not presume that a sharp upsurge in productivity growth is imminent.

The Slowdown in Productivity Growth

Output per hour of work in the U.S. business sector.
Graph--slowdown in productivity growth
Source: U.S. Department of Labor

Here the growth optimists rise to object. "Wait a minute," they say. "With all the evident technological miracles in computers and telecommunications, and with all the industrial restructuring, how can you expect us to believe that productivity growth in the U.S. is just 1.1 percent per annum? The measurements must be wrong." I'd like to answer this objection at two levels, for in one sense I agree and in another sense I disagree.

The disagreement is more important. Suppose our trend productivity growth rate—under current official measurements—was really much higher than 1.1 percent. For concreteness, let's use the Dole-Kemp estimate of 2.4 percent. In that case, GDP growth of 1.9 percent per year from 1990 to 1995 should have been accomplished with decreasing use of labor input because output per hour grew faster than output. In fact, payroll employment expanded by almost eight million jobs. That's quite an error! Another way to make the same point is this: If actual growth over these five years really fell short of trend growth by 1.6 percent per year (1.9 percent instead of 3.5 percent), the unemployment rate should have risen by about 4 percentage points over this period. In fact, it was unchanged.

But there is a sense in which the growth optimists may be right: The official data may badly underestimate productivity growth. Government statistics want us to believe that what economists call total factor productivity—the increment to output you get from skills, technology, and managerial efficiency, without the need to apply any additional inputs—has not grown at all since about 1977. That is simply not believable.

Readers of The American Prospect will be aware of the recent debate over the so-called bias in the Consumer Price Index. Arguments that the CPI overstates inflation are compelling, although the magnitude of the bias is hotly disputed. Fewer people seem to have noticed, however, that any upward bias in measuring inflation implies a corresponding downward bias in measuring real growth. And since labor input is measured fairly precisely, the entire measurement error shows up in productivity. For example, if inflation is overestimated by 1 percent per year, then real growth may actually be a full percentage point higher than recorded in the official statistics—say, 3 percent instead of 2 percent.

In this case, however, the growth optimists who rail at the Fed for restraining the economy are still wrong. Critics are confused on two points.

First, errors in measuring productivity affect data on actual and potential GDP equally. So they carry no implication that the economy has more spare capacity than we think. It is not that we should be growing faster than we are; it is that we are in fact growing faster than the data say.

Second, those who argue that the U.S. economy has experienced a recent upsurge in productivity growth must explain why the measurement error has grown much worse in the last year or two, for the official data show no such upsurge. No one has yet offered such an explanation.



The evidence I have just presented is not secret; it is available to anyone who looks. Why, then, do intelligent people ranging from Bob Dole and Jack Kemp on the right through Jerry Jasinowski in the center to Lester Thurow and Felix Rohatyn on the left argue otherwise? I have been able to think of six reasons. Three of them are measurement issues.

First, as just noted, the likely overestimation of inflation means that real growth has probably been underestimated. We may well have been growing at 3 percent during the 1990s.

Second, the government altered its measurement system at the start of 1996, adopting what the green-eyeshade crowd calls "chain-weighted" GDP. Through the end of 1995, the government calculated real GDP by valuing all goods and services at 1987 prices—which vastly overpriced computers, for example. Chain-weighted GDP uses more recent market prices, thereby putting less weight on computers. That naturally produces slower measured growth even with no change in the real economy. The new measurement system probably reduced the 1996 growth rate by about three-quarters of a percentage point.

Third, our antiquated statistical system lavishes far too much attention on the manufacturing sector—which accounts for only about 20 percent of GDP. Productivity performance in manufacturing has indeed been excellent in recent years. No argument there. The problems reside in the other 80 percent of the economy, which is where most of us work.

Fourth, biases in reporting lead to what I call "the tyranny of the selective anecdote." The business press—like businesses themselves—tends to trumpet success stories and bury failures. When leaders of particular successful companies tell me that they have achieved dramatic productivity gains, I believe them. But the U.S. economy is not BusinessWeek cover stories writ large. Some companies downsize and fail. Some companies automate with catastrophic results. That's why we need economy-wide data.

Fifth, people tend to forget about the reallocation of labor that accompanies downsizing. When some large corporation restructures to produce the same output with far less labor, its productivity rises dramatically. But the displaced workers then must seek jobs elsewhere. If they find employment in firms with much lower value-added per worker, then economy-wide productivity may not rise much despite the vaunted productivity miracles.

Finally, I come to what may be the biggest puzzle of them all—and perhaps the biggest reason for the "Say it ain't so, Joe" attitude: advances in computers, and in information technology (IT) more generally. As Robert Solow has put it, "The computer is everywhere except in the productivity statistics." Why not?

The pace of technological advance in electronics has indeed been mind-boggling. Businesses now communicate with lightning speed. Some serve their customers via automated devices rather than human beings—ATMs, voice mail, and sales over the Internet are common examples. Computerization has also revolutionized some factory floors and the inventory management practices of many companies. And so on. All this is beyond dispute.

But is this new computerized world a vastly more productive world—in the narrow sense of producing more GDP per hour of labor? The official statistics say no. In fact, the timing of the productivity slowdown shown in "The Slowdown in Productivity Growth" corresponds roughly to the invention of the personal computer. Furthermore, when you examine the industry-by-industry data, some of the worst productivity performances have been turned in where you might expect innovations in IT to have paid the richest dividends. What's going on here?

No one knows for sure, but my tentative answer is: Don't be taken in by the hyper-hype. Sure, I now can surf the Net, send and receive e-mail in seconds, and have more computing power on my desk than ever before. But has any of this made me produce more GDP per hour of work? Don't forget that rapid turnover of hardware and software keeps us perpetually in the learning mode, that people spend countless hours mindlessly exploring the Internet and playing amusing computer games, and that most of us suffer more from information overload than from information shortage. Perhaps most important, the human brain has not advanced apace with the microprocessor.

A productivity miracle based on the computer may be just around the corner. Perhaps. But, if so, it is around the next corner, not the last one.



One final thought. This article is intended as a reality check, not as a counsel of despair. Our economy's long-run growth trend, be it 2.1 percent, 2.3 percent, or 2.5 percent per year, is not a constant of nature. Growth can be enhanced by intelligent economic policies and damaged by foolish ones.

For example, the principal rationale for reducing the government budget deficit is not to pay homage to our Puritan ancestors, but to spur capital formation and thus accelerate economic growth. Similarly, I have long urged greater investments in education and training as a pro-growth policy. And the basic case for government-supported research and development is that R&D is the mainspring of total factor productivity growth. Even well-designed tax, regulatory, and trade policies can make modest and transitory contributions to growth.

Many such policies are well worth doing. An artfully chosen combination might conceivably add one-quarter or even one-half of a percentage point to the growth rate for a time, which would certainly be a notable achievement. But with economic growth, as with all things, you should be wary of fast-talking purveyors of miracle cures. Nothing—I repeat, nothing—that economists know about growth gives us a recipe for adding a percentage point or more to the nation's growth rate on a sustained basis. Much as we might wish otherwise, it just ain't so.

You may also like