Labor productivity and firm-level TFP with technology-specific production functions

Abstract

We investigate the technological dimension of productivity, presenting an empirical methodology based on mixture models to disentangle the labor productivity differences associated with the firm's choice of technology (BTFP) and those related to the firm's ability to exploit the adopted technology (WTFP). The estimation endogenously determines the number of technologies (in the sector) and degree of technology sharing across firms (i.e., for each firm, the probability of using a given technology). By using comparable data for about 35,000 firms worldwide distributed across 22 (two-digit) sectors, we show BTFP to be at least as important as WTFP in explaining the labor productivity gaps across firms. Intra-sectoral and inter-sectoral heterogeneity is substantial and, even in sectors in which BTFP dominates on average, we find a considerable number of firms for which labor productivity is mostly determined by the ability to use the adopted technology. Hence, dissecting the labor productivity gaps is crucial to achieving more targeted innovation policies. The estimated number of technologies ranges from one (in only three industries) to five, being three in most cases. The suggested estimation strategy takes simultaneity into account. The BTFP measure is unaffected by omitted price bias. The presence of BTFP dispersion can be associated with the action of frictions preventing firms from switching to superior and more productive technologies. Eliminating BTFP does not eliminate misallocation.

Keywords: TFP, technology adoption, production function estimation, mixture models

JEL Classification: D24, 033, C29