Estimation of the Cobb-Douglas with zero input levels: bootstrapping and substitution

Abstract

New informational technologies have led to a rebirth of interest in primal production functions. This study revisits the vexing question of the estimation of the Cobb-Douglas production function given zero input levels for some observations. A resampling technique is proposed that produces a sample with nonzero input levels. The results of the resampling technique are then compared with the substitution results using informational measures of the implied cost shares.