The Econometrics of Piecewise-Linear Budget Constraints
用蒙特卡洛方法评估了分段线性凸预算约束下需求模型的几种估计量，发现最大似然估计远优于普通最小二乘，且OLS的偏差随误差方差减小而减小。
This article presents a Monte Carlo evaluation of some alternative estimators for a demand model when the budget constraint is piecewise-linear and the budget set is convex. We examine the performance of two maximum likelihood (ML) estimators and an ordinary least squares (OLS) estimator under varying sample sizes and error variances. A simple log-linear demand function, with income and price as the explanatory variables, is specified. Although I find that the OLS bias decreases as the error variance decreases, the ML results are far superior. Furthermore, statistical tests based on the OLS results lead to erroneous conclusions regarding the structure.