Nonparametric and Semiparametric Estimation of Car Production Function with an Emphasis on Energy Inputs: Introduction of Three-Step Olley-Pakes Approach in Estimation of Panel Data

Unobservable productivity shocks cause selection and simultaneity problems in firm’s decisions and these problems cause estimators such as ordinary least squares, have biased estimation for coefficients of production function inputs. In this study, data of five automaker companies in the period of 1383-1387 have been used and production function of car industry have been estimated by ordinary least squares, fixed effects, random effects, Olly and Pakes (1996) and Levinsohn and Petrin (2003a) approaches. The results show that fixed effects and Levinsohn and Petrin (2003a) approaches can’t be appropriate for the production function estimation of car industry. In other words, reaction of automaker companies to productivity shocks will not be done through adjustment in labor, capital and energy demands and there is no significant correlation between inputs adjustment and productivity shocks in car industry. But estimated coefficients of energy and capital in semiparametric, random effects and ordinary least squares approaches show that estimated coefficients of energy and capital in random effects and ordinary least squares approaches are upwardly and downwardly biased, respectively. These results are perfectly consistent with the viewpoint of Olly and Pakes (1996) about bias of traditional estimators and show that automaker companies, in response to the productivity shock, adjust their investment level. In addition based on estimation of semiparametric approach, output elasticity of capital and energy will be respectively 0.82 and 0.64.