This paper shows that the frequency of migration can be best modeled by a zero-inflated Poisson process, because it takes into account the overwhelming presence of zeros (nonmigrants) in the data. A failure to do so can cause the coefficients to be biased and also result in poor prediction. The major finding is that by using a zero-inflated process, the performance of the model in predicting migration behavior is substantially improved. In addition, frequent movers tend to be white, nonunionized, and tend to have fewer children, less stable marriages, and more frequent occupational changes.