Estimating the piecewise-linear budget constraints model with non-normal error distributions

摘要

The maximum likelihood estimation methodology of the piecewise-linear budget constraints model has been used in a wide variety of empirical approaches. However, the sensitivity of the estimates to the normality assumption of the error terms has not received much attention in the past two decades. This paper investigates the maximum likelihood estimates1 performance of the piecewise-linear budget constraints model with non-normal error distributions. Using the Monte Carlo technique, the simulation results show that, except for the price coefficient, the estimates do not induce a significant bias from the non-normal error distributions. The estimating bias of the price coefficient, however, can be substantial. Since the price coefficient is particularly important in empirical and policy analysis, there is a requirement to correct this non-normality problem. This paper subsequently introduces an inverse hyperbolic sine transformation to correct this problem. The estimation results indicate that this transformation methodology can significantly reduce the estimating bias of the price coefficient.