Abstract

Hopenhayn (1992) considers a competitive environment of industry dynamics with heterogeneous firms. There exists the time-invariant stationary equilibrium distribution of firm size, and we can define the unique cross-sectional distribution, given the parameter values. Existing literature calibrate the structural parameters by using the related panel data estimates and the moment-matching method. There are three flaws in the calibration process. First, the relation between reduced-form estimates and structural parameters is unknown. Second, the panel data estimates can differ depending on which econometric methods researchers employ. Third, the observed firm size distribution is so fat-tailed that only the first moment exists, which can be not sufficient to specify the distribution. In order to structurally estimate the underlying parameters that can appropriately reflect the property of the distribution, I employ the Bayesian likelihood-free inference method, named Approximate Bayesian Computation (ABC). The estimation method is effective where the likelihood is intractable but easy to simulate from the model conditional on typical parameters, can therefore be used to calculate the posterior distributions of structural parameters.