1: \begin{abstract}

2: This paper develops maximum score estimation of preference parameters in the

3: binary choice model under uncertainty in which the decision rule is affected

4: by conditional expectations. The preference parameters are estimated in two

5: stages: we estimate conditional expectations nonparametrically in the first

6: stage and then the preference parameters in the second stage based on Manski

7: (1975,\thinspace 1985)'s maximum score estimator using the choice data and

8: first stage estimates. The paper establishes consistency and derives rate of

9: convergence of the two-stage maximum score estimator. Moreover, the paper

10: also provides sufficient conditions under which the two-stage estimator is

11: asymptotically equivalent in distribution to the corresponding single-stage

12: estimator that assumes the first stage input is known. These results are of

13: independent interest for maximum score estimation with nonparametrically

14: generated regressors. The paper also presents some Monte Carlo simulation

15: results for finite-sample behavior of the two-stage estimator.%

16: %\vspace{0.2cm}

17: 

18: \noindent \textbf{Keywords:} \textit{discrete choice, maximum score

19: estimation, generated regressor, preference parameters, M-estimation, cube

20: root asymptotics} %\medskip

21: 

22: \noindent \textbf{JEL Codes:} C12, C13, C14.

23: \end{abstract}