Comments and Discussion

Martin Neil Baily:

The three authors of this paper have made some of the strongest contributions to the productivity literature in recent years, and it is terrific to see them team up to provide an important new analysis of the productivity acceleration that started in the mid-1990s. In particular, I liked the creative way they have adapted the growth accounting framework to take account of intangible capital, and I welcome the new insights provided by their industry-level regression analysis, particularly those highlighting the role for competitive pressure.

Labor productivity accelerated in the United States starting in 1996, after over twenty years of slow growth. That acceleration has been widely attributed to the revolution in information technology, a natural enough inference given that the acceleration coincided with a wave of capital investment in IT hardware and software. Indeed, some fraction of the productivity acceleration can certainly be attributed directly to an acceleration within the IT hardware-producing sector. Around 2000-01, however, the IT bubble burst, and IT investment slumped as the economy went into a mild recession. Yet, surprisingly, productivity growth did not slow down but actually grew even faster over 2002-04. This meant that the simple correlation between IT investment and productivity broke down after 2000.

There are three possible responses to what happened. The first is to conclude that perhaps IT was not as important to the post-1995 productivity acceleration as had been thought. Second, one could argue that IT investment has a lagged effect on productivity, so that the high-technology investment boom in the late 1990s had an impact that spilled over into the post-2000 period. A third hypothesis is that IT investment creates intangible capital that should be counted as part of total output. This last option is the approach taken in the growth accounting section of this paper, and it shifts some of the productivity acceleration from the post-2000 period backward in time to 1995-2000, where it coincides with the surge in IT investment.

Working only with aggregate productivity data, one has very limited information available to identify which (if any) of these three options is correct. Indeed, on the productivity side, there are really only three observations to work with: slow growth until 1995, faster growth after 1995, and even faster growth after 2000. Meanwhile much of the accumulation in intangible capital is very difficult to observe. The paper by Corrado, Hulten, and Sichel discussed by the authors develops measures of intangible capital based on a variety of data sources and includes software investment, company training, consulting, and the labor input of employees in job categories that contribute to organizational capital.¹ The estimates from this work were not available beyond 2003, and so the present authors do a quick update through 2005. They report, in their table 3, that intangible capital accumulation by this measure turned down sharply after 2000.

In this paper the authors do not use the Corrado, Hulten, and Sichel estimates directly but turn instead to the paper by Basu and others to develop their new approach to growth accounting.² Basu and others is an interesting and helpful paper, but I am not persuaded that their approach is a real substitute for direct observation of intangibles. The basic assumption is that intangible capital investment is tied very closely to investment in IT hardware, so that the time-series pattern of the former is derived from that of the latter. It is entirely plausible that high investment in IT demands an increase in intangible investment, but whether or not this is the dynamic driving the observed pattern of productivity growth remains unknown. I note also that the Basu and others paper has a mixed record in tracking productivity trends. They do find regression coefficients for the United States that suggest that heavy IT investment can depress measured productivity contemporaneously. But as they themselves note, "For the United Kingdom, the same regression shows little. Almost nothing is statistically significant, and the signs are reversed from what theory suggested."³

The growth accounting section of the present paper takes a perfectly...