This paper examines the relationship between mortgage default decisions and relevant observable variables under the light of a random utility model. The focus of the study is the Colombian mortgage market between 1997 and 2004 using two separate data sets that are matched using simulation techniques. The estimation allows the computation of mortgage default probabilities that are directly related to an underlying model of optimal default. Results are sharp and indicate that variation in current income has little effect on mortgage default, compared to housing prices and mortgage balances.