Semiparametric and Nonparametric Estimation off Structural Models

该项目将开发使用半参数或非参数方法估计非线性计量经济学模型参数的新方法。将开展三大方面的研究工作。首先，在弱非参数约束下，考虑样本选择下线性模型的参数估计问题。具体地说，选择效应将被假定为仅依赖于要使用非参数回归方法估计的某个可观测选择变量的条件平均值；该估计将出现在感兴趣参数的第二步估计器中。其次，将研究使用辅助变量的结构方程的非参数估计。这将涉及积分方程解的非参数估计，以及具有回归变量的初步非参数估计的加性回归模型的非参数估计。第三，当误差分布满足单个条件分位数限制时，得到了单调回归模型估计量的可达效率的界。此外，还将开发一种可行的估计程序，以达到这一界限。这一研究将继续推动计量经济学中半参数和非参数估计的前沿。这些方法将对经济模型的发展极为有益，因为允许更灵活的假设，并为传统估计技术提供基准。

The project will develop new methods for estimating the parameters of nonlinear econometric models using semiparametric or nonparametric approaches. Work on three broad lines of research will be undertaken. First, estimation of parameters of linear models subject to sample selection will be considered under a weak nonparametric restriction on the form of the selection correction. Specifically, the selection effect will be assumed to depend only on the conditional mean of some observable selection variable, which is to be estimated using a nonparametric regression method; this estimate will appear in the second-step estimator for the parameters of interest. Second, nonparametric estimation of structural equations using instrumental variables will be investigated. This will involve nonparametric estimation of solutions of integral equations, as well as nonparametric estimation of additive regression models with preliminary nonparametric estimates of regressors. Third, a bound for the attainable efficiency of estimators for monotonic regression models will be derived when the error distribution satisfies a single conditional quantile restriction. Also, a feasible estimation procedure which attains this bound will be developed. This line of research will continue to push the frontier of semiparametric and nonparametric estimation in econometrics. These methods will be extremely beneficial in the development of economic modelling, as more flexible assumptions are allowed and provide benchmarks for traditional estimation techniques.