A discrete choice model with misclassification and multiple recall periods

Self-reported data collected via surveys are often subject to measurement error caused by recall errors. While methods to minimize such problems are to be encouraged we argue that such errors are often unavoidable and need to be accommodated in estimation. Such methods have been developed to facilitate estimation in the case of a binary choice model with a misclassified dependent variable. These methods are extended to situations where the survey contains questions with multiple recall windows. The estimation procedure is illustrated using an analysis of Australian data on visits to a GP.