The estimation of nonstationary dynamic discrete choice models typically requires making assumptions far beyond the length of the data. We extend the class of dynamic discrete choice models that require only a few-period-ahead conditional choice probabilities, and develop algorithms to calculate...