April 1997
Jeremy Greenwood, professor of economics at the University of Rochester, led the tenth of AEI's Understanding Income Inequality seminars. His study of how the computer revolution has (and has not) affected productivity and wage inequality is summarized below.
To the surprise of many observers, the computer revolution of the past twenty years has failed to accelerate productivity growth. This puzzling phenomenon, along with an increase in wage inequality, is a normal--but temporary--attribute of major technological innovation and change.
Examination of historical data from the past two centuries reveals that similar patterns in equipment prices, productivity growth, and wage inequality characterized several different eras of profound technological innovation, such as the introduction of steam power and the beginning of the use of electricity. Prices of the relevant capital equipment decline; productivity growth actually slows for an initial period; and, in the case of wages, skill premiums quickly develop because of the difficulty of learning to use new technology. This pattern is now repeating itself in the United States as a consequence of the revolution in information technology.
The current computer revolution will likely follow the pattern of previous industrial revolutions. In the long run, everybody will gain. Technological change implies that eventually more output can be produced by a unit of labor. Hence, a unit of labor will become more valuable. Given time, this process translates into higher wages and standards of living for everyone. Everybody today is better off due to the British industrial revolution. That was not true, however, in 1760.
The underlying phenomenon of initially retarded productivity growth is due to the learning curve--the sometimes extended period of time over which managers and workers learn to use new technology. Data from a cotton mill in Lowell, Massachusetts, in the period between 1836 and 1856 illustrate the point. These data show that output per man-hour rose substantially and steadily over those twenty years, despite the fact that, after the installation of new machinery at the beginning of the period, no additional machinery was purchased. The entire average annual productivity growth of 2.3 percent per year was attributable to progress up the learning curve.
A related factor in delaying positive productivity benefits from new technology is the slow pace of diffusion of the new technologies. Studies of innovations ranging from phonograph records to lasers show that the initial incarnations of new ideas are often expensive and plagued with bugs. This situation too delays the improvement in productivity, but, in the end, the expected effect occurs.
While new technologies initially lead to a sizable skill premium and, consequently, widen the wage gap, eventually those effects reverse themselves for two reasons. First, as technology becomes established, the production process substitutes cheaper unskilled labor for expensive skilled labor, shifting demand toward the unskilled. Second, on the supply side, more young persons acquire the skills associated with the new technologies, and the wage premium begins to subside.
This thesis differs from the view that there has been and will continue to be a gradual upward shift in the skill requirements of work, which implies that those with lower "innate" abilities will be increasingly left behind. Based on experience, we can be confident that skill differentials in wages will narrow again after the passage of time.
The time lags between the introduction of major technological change and the payoff in higher productivity--and hence general well-being--can be long. It takes about twenty years before productivity growth surpasses its old level and forty years for the level of productivity to cross its old trend line--the path that productivity would have traveled along if it had continued its old growth rate.
