Data mining for higher education advancement: A study of eight North American colleges and universities

The purpose of the study was to examine alumni advancement databases at eight public and private higher-education institutions in North America to identify variables that predict the overall likelihood of alumni giving. A multiple regression analysis was completed for each institution. The analyses tested the predictive power of all potential candidate predictor variables created through examination of data fields provided by the advancement offices of the eight institutions.;Although this study identified a number of single-predictor variables strongly correlated with cumulative lifetime giving through fiscal year 2008, it was the linear combination of these variables (multiple linear regression) that revealed their collective power to predict alumni giving. Further, the researcher developed predictive scores (1-20) for each alumnus record where these records could be used to focus advancement resources on alumni most likely to give.;The model for each institution was tested by evaluating the percentage of alumni who gave anything in each of these 20 categories during FY 2009. A chi-square test of these data revealed a highly significant relationship between score level and giving in FY 2009. The research provided detailed information about how advancement staff for higher-education institutions can develop unique statistical models from alumni databases to help predict those most likely to provide financial support.