Efficient Estimation of Neoclassical Parameters of Substitution and Biased Technological Progress

Abstract

In an earlier issue of this Journal [1], we reported estimates of the elasticity of substitution and of biased or neutral technical progress for 20 two-digit American manufacturing industries over the period 19491962. The method of estimation consisted of applying ordinary least squares (OLS) to a constrained cost minimization equation for each industry. This method yields best linear unbiased estimates of the regression coefficients under three standard assumptions: (i) the regressors are stochastically independent of the disturbances in each industry; (ii) the disturbances are homoskedastic and are not autoregressive; and (iii) the disturbances in the i-th industry are independent of those in the j-th industry for all i # j. Given the time-series data available, there are some objections to these standard OLS assumptions. In regard to (i), we reasoned that the regression coefficients are not likely to be affected by a simultaneous equation bias [1, 315-16]. Nonetheless, a likely error of measurement in the factor-price-ratio regressor remained a problem.1 Concerning assumption (ii), we tested the regression