A Logit Model with Interactions for Predicting Major Gift Donors

Abstract

We provide a new statistical model developed from the alumni database at Northwestern University for identifying potential major gift donors. Our logit model with interactions predicts which individuals will give $10,000 or more over three years using CHAID (Chi-square Automatic Interaction Detection) combined with logit analysis. Our work goes beyond our earlier research (Lindahl and Winship, 1992) by providing a more effective means of selecting fund-raising prospects. The critical new component in our model is the inclusion of interactions between past giving and other variables. Specifically, we find that for individuals with different past giving records different variables are important. Our results show that prospects with a low past giving level will rarely give gifts over $10,000.