Estimation of Random-Coefficient Demand Models: Two Empiricists' Perspective
使用两个知名数据集和全面的优化设计，记录了估计Berry、Levinsohn和Pakes（1995）式随机系数需求模型时遇到的数值挑战，发现优化算法常在最优性条件不满足的点收敛，且参数估计的波动会导致经济预测（如价格弹性、福利变化）出现2到5倍的差异。
Abstract We document the numerical challenges we experienced estimating random-coefficient demand models as in Berry, Levinsohn, and Pakes (1995) using two well-known data sets and a thorough optimization design. The optimization algorithms often converge at points where the first- and second-order optimality conditions fail. There are also cases of convergence at local optima. On convergence, the variation in the values of the parameter estimates translates into variation in the models' economic predictions. Price elasticities and changes in consumer and producer welfare following hypothetical merger exercises vary at least by a factor of 2 and up to a factor of 5.