Comments and Discussion

Robert J. Gordon:

The economic miracle of the "Goldilocks" U.S. economy of the late 1990s was the source of pride at home and envy abroad. The miracle was unusual not just for its combination of rapid output and productivity growth, low inflation, and low unemployment, but also for the absence of academic controversy as the news unfolded of one broken record after another. Compared with earlier postwar macroeconomic issues such as the acceleration of inflation in the 1960s, the natural rate hypothesis, the monetarists versus the fiscalists, the supply shocks, and the Reagan-era tax cuts and budget deficits, there has been surprisingly little debate about the sources of the 1990s economic boom.

By the standards of those past debates, there is little disagreement between the paper under discussion and the papers by Stephen Oliner and Daniel Sichel and myself.¹ All these papers concur that productivity growth ended its long 1972-95 torpor, when it averaged a mere 1.4 percent per year, and suddenly doubled to a rate approaching 2.8 percent after 1995. They also agree that this sudden spurt resulted from an acceleration of technical change in that segment of U.S. manufacturing, often called the information technology sector, that produces computer hardware and peripherals, semiconductors, and software. Almost all the other ingredients of the miracle can be traced, directly or indirectly, to this one critical event of accelerating technical change in IT. Accelerating productivity held down inflation despite accelerating wages; low inflation allowed the Federal Reserve (at least until early in 2000) to maintain the Federal funds rate no higher than it had been five years earlier, while long-term government bond rates remained considerably lower; easy monetary policy and rapid economic growth spurred unprecedented growth in profits; and booming profits, together with a wave of optimism about the technological acceleration, propelled stock market valuations to grow far faster than profits. The average American household was showered in stock market wealth, which spurred growth in consumption expenditure greatly in excess of growth in disposable income, with the corollary of a disappearing household saving rate, at least as conventionally measured.

These papers are also complementary. Dale Jorgenson and Kevin Stiroh, like Oliner and Sichel, have produced an imaginative and convincing decomposition of the role of computers, semiconductors, software, and other elements of the "new economy" in the post-1995 acceleration in capital deepening and in total factor productivity. Because they use different concepts and definitions of that portion of the new economy that they cover, their results differ modestly, and I defer to Sichel's comment, which follows, to enlighten us on the sources of those differences. My paper differs from these two papers not in its analysis of the contribution of IT to the productivity acceleration, but rather in its attempt to decompose that acceleration into a permanent, "structural" acceleration in the underlying productivity trend and the remaining, cyclical component.²

Before turning to my analysis of the cyclical element in the post-1995 productivity acceleration, a few more general comments on the present paper are in order. It is always a privilege to discuss any paper on economic growth and productivity by Dale Jorgenson and his many former students and other coauthors, for his work since the early 1960s is virtually unique in the continuity and consistency of its research program. The integrity of that program has persuaded the U.S. Bureau of Labor Statistics to adopt his methodology, developed originally in several of his own papers and with the late Zvi Griliches,³ for the nation's official measures of growth in output, inputs, and TFP.

A critical element of this research is the emphasis on the role in economic growth of changes in the composition of capital and labor inputs. Because short-lived equipment involves far more depreciation expense per dollar of capital cost than do long-lived structures, such equipment must yield a higher marginal product of capital in order to warrant its purchase by firms. As Jorgenson and Griliches showed long ago, equipment and structures must be weighted by their marginal products (or...