The Pursuit of Knowledge: Speeches and Papers of Richard C. Atkinson

The Numbers Game and Graduate Education

77 The Numbers Game and Graduate Education October 1996 The National Science Board report Science and Engineering Indicators , 1996 has a new section this year, entitled “Science and Engineering Labor Market.”1 It begins with the following statement: “The performance of the U.S. economy is the major determinant of current and future demand for scientists and engineers.” I would argue that this statement represents a short-term perspective on the science and engineering labor market. Clearly, the currenteconomydeterminesthe flowoftaxes,companyrevenues,and the number of individuals who will be hired at any given time. A long-term perspective, however, would focus on the importance of science and engineering as a driver of the future economy; the investments made in R&D today will be a dominating factor in the level of economic growth experienced in the future. In your packet for this conference on graduate education in the biological sciences, you have an article on the supply and demand for scientists and engineers that I published in Science in 1990, based on work done in 1988.2 This article reported on a National Science Foundation study that I was involved in, much like the study that Bill Bowen and a colleague at Princeton were doing at about the same time.3 Bill and I were both projecting a significant future shortfall of Ph.D.’s. Bowen was looking at the humanities and the social sciences as well as the natural sciences and engineering. My paper was concerned only with the natural sciences and engineering and excluded the social sciences. The study began with the year 1988 and projected the supply of Ph.D.’s that would be trained in future years. That projection was made on the assumption that a certain percent of undergraduate students would go on for Ph.D.’s, and thus was based on the demographics of the twenty-two-year-old population . If you look back over the past twenty years, the proportion of twenty-two-year-olds who eventually earned a Ph.D. in science and engineering is remarkably stable. Added to that was the assumption that the number of foreign students taking Ph.D.’s in the United States in future years would remain at the 1988 level. In 1988 we had a large number of foreign students taking Ph.D.’s, and the assumption was that this number wouldn’t increase significantly. A further assumption was that 50 percent of the foreign students who earned Ph.D.’s in the United States would stay in the United States. And these assumptions led to the wiggly curve on the chart labeled “Supply of Ph.D.’s.” You can see on the far left [of the chart] the actual number of Ph.D.’s produced in 1988. Supply was projected through the year 2010. The D0 curve was based on the assumption that the future demand for Ph.D.’s would remain constant. That is, whatever 78 / The Research University The Numbers Game and Graduate Education / 79 the demand was in 1988, that demand would stay the same out to the year 2010. Note that for the D0 demand curve, it is not until the late 1990s that an undersupply of Ph.D.’s begins to occur. But the constant D0 scenario seemed highly unlikely for at least three reasons. First, yearly replacements due to retirements and deaths were expected to increase over the next two decades. Second, we considered it almost certain that college and university enrollments would increase in the late 1990s with the expanding college-age population, necessitating an increase in the number of faculty hired. Third, we assumed that if federal and private investments in R&D continued to grow at even moderate rates, the number of new Ph.D.’s required by industry would be well above the 1988 level. These 1990 1988 25 20 15 Ph.D. Degrees (in thousands) 10 0 1995 2000 Year 2005 2010 D3 D2 D1 D0 Supply Supply and demand projected to the year 2010 for Ph.D.’s in the natural sciences and engineering. Four demand scenarios are indicated by the D0 , D1 , D2 , and D3 curves. three factors generate four cumulative demand scenarios, labeled D0 , D1 , D2 , and D3 . We knew the demographics of the workforce in 1988 and the age distribution of Ph.D.’s in that workforce, and therefore it was fairly easy to predict the expected increase in the...