TECHNOLOGICAL PROGRESS AND PRODUCTIVITY GROWTH IN THE U.S. SEMICONDUCTOR INDUSTRY

How significant a role has learning by doing played in increasing labor productivity in the U.S. semiconductor industry? Do learning economies remain important if other potential influences on labor productivity are taken into account? (Such potential influences might include increases in the capital/labor ratio, economies of scale, exogenous improvements in technology, and changes in the composition of labor.) How significant a role have these other influences played?; To answer these questions, a log linear production function incorporating learning variables and the other major factors was used. Two different data sets provided a cross-check on the results. First, an aggregate production function was estimated using industry-wide data from published sources, mainly the Census of Manufactures. A disaggregated production function also was estimated using data from one fabrication line producing one family of integrated circuits, using proprietary data obtained from the manufacturer. The estimation of the functions showed that while learning economies played a much smaller role in productivity growth in full models than in naive models, experience remained correlated positively with productivity. The learning parameters were much smaller than usually found in naive models, and the capital coefficient was much larger than usually found in production functions. Measures for scale economies, changes in labor composition, and technical change failed to produce credible results. Collinearity and some measurement problems contributed to estimation difficulties. The results were consistent across data sets. The importance of learning by doing has been overstated in this industry. While still present, the major factor explaining the rapid rate of productivity growth has been the increased use of capital. However, since the capital equipment has been improving in quality as well as in quantity, the capital coefficent probably overstates the pure contribution of capital to productivity growth because it incorporates advances in technology as well. While the statistical evidence did not corroborate the presence of scale economies, the collinearity and the nonstatistical evidence suggesting its presence leave this an open question.