Generalized Maximum Entropy estimation of discrete sequential move games of perfect information

We propose a data-constrained generalized maximum entropy estimator for discrete sequential move games of perfect information. Unlike most other work on the estimation of complete information games, the method we proposed is data constrained and requires no simulation or assumptions about the distribution of random preference shocks. We formulate the GME estimation as a (convex) mixed-integer nonlinear optimization problem which can be easily implemented on optimization software with high-level interfaces such as GAMS. The model is identified with only weak scale and location normalizations. Monte Carlo evidence demonstrates that the estimator can perform well in moderately size samples. As an application we study the location choice of German siblings using the German Ageing Survey.