An Econometric Analysis of Firm Specific Productivities: Evidence from Japanese plant level data

In estimating the production function of firms, problems of endogeneity and self selection exist as a result of firm-specific productivity shocks and entry/exit decisions. Several methods have been proposed to handle these problems, such as those by Olley and Pakes (1996) and Levinsohn and Petrin (1999, 2003). However, the endogeneity of labor input does not seem to be completely solved by these methods. We therefore propose an alternative semiparametric IV estimator. We suppose that firm-specific productivity influences labor input as well as capital input. We adopt the lagged variables of inputs as their instruments instead of investment inputs, unlike Olley and Pakes. Moreover, our econometric model should automatically adapt to the effect of the exit decision of each firm. We applied the model to Japanese plant-level panel data from 1982 to 2004 on the Census of Manufactures provided by the Ministry of Economy, Trade and Industry. We found that our estimator works well in an empirical study in terms of sign and magnitude of the technological parameters. Using the estimation residuals, we decomposed the TFP into firm-specific productivity and other exogenous shocks. We also aggregated the productivity shocks to industry-level productivities to determine the transition. We examined whether negative technological shocks were the main cause of poor economic performance in Japan during the "lost decade", and found that productivity did not decline in most Japanese industries since the 1980s. This implies that the recession might have been caused by demand-side factors rather than supply-side issues.