A Comparative Study of Instrumental Variables Estimators for Nonlinear Simultaneous Models
研究了非线性联立模型单方程估计中的三类工具变量估计量，包括使用非线性函数拟合值、朴素工具和条件期望估计，证明了朴素方法的一致性和渐近分布，并通过模拟比较了效率。
Abstract In the context of single equation approaches to the estimation of nonlinear simultaneous models, three classes of estimators can be established, using different fitted values established by a stage-one regression. The first uses the Fitted values of the nonlinear function in the second-stage regression. The second, called here the naive instrument, substitutes a fitted value inside the nonlinear function. The third uses an estimate of the conditional expectation of the nonlinearity. Consistency and asymptotic distribution results are established for the naive method. Simulation studies compare the efficiency of the three techniques.