Science & Technology

No Thanks For Sharing

Has Sematech helped the semiconductor industry?


Has the public-private venture Sematech accomplished its goal of reinvigorating the U.S. semiconductor industry? In 1992, the government's own General Accounting Office said yes: "Sematech has demonstrated that a government-industry R&D consortium on manufacturing technology can help improve a U.S. industry's technological position."

That's the wrong conclusion, contend Douglas A. Irwin and Peter J. Klenow, professors at the University of Chicago's Graduate School of Business. Their recent National Bureau of Economic Research working paper, "High Tech R&D Subsidies: Estimating the Effects of Sematech," concludes that far from generating the kind of productive research and development intended, Sematech subsidizes research that would be undertaken anyway. And, the authors suggest, the "slight rebound" in market share by U.S. semiconductor makers is more attributable to a falling dollar, trade agreements with Japan, and the declining importance of memory chips in the semiconductor market.

The government created Sematech in 1987. Member firms are required to contribute personnel and financial resources: 1 percent of their semiconductor sales revenue, with $1 million as the donation floor and $15 million as the donation ceiling. The consortium receives about $100 million a year in matching government funds until 1997. Members have a "right of first acceptance" on Sematech-funded equipment, which means they can acquire new equipment six to nine months before non-members. Of the 14 original member firms, 11 remain.

To evaluate Sematech's effectiveness, Irwin and Klenow develop two hypotheses. The "commitment" hypothesis holds that competing firms will shy away from "high-spillover" research–research that will tend to benefit the industry in general along with the particular firm doing the research. The commitment hypothesis provides a rationale for government subsidies, since without them basic R&D may never get done. If that hypothesis is true, then, Sematech should have boosted participating firms' research spending.

The "sharing" hypothesis, by contrast, suggests that participating in Sematech would allow firms to reduce duplicative R&D efforts. If true, it does not provide a rationale for government subsidies, since "[f]irms should have every private incentive to form joint ventures to raise their R&D efficiency," write Irwin and Klenow.

In a comparison of R&D spending by member and non-member semiconductor firms, Irwin and Klenow found no evidence that Sematech increased research productivity or spurred more R&D spending. In fact, member firms cut their overall R&D by about $300 million a year, consistent with the sharing hypothesis. "Sematech," says Klenow, "is not worth the government money put into it."