On using the beta-logistic model to update response probabilities given nonresponse

Abstract

Rao and Steckel (1995) propose that direct marketers use the beta-logistic model to update estimated individual response probabilities after observing one or more nonresponses. Ehrman and Funk (1997), however, pointed out that the beta-logistic model does not explicitly account for nonreaders—individuals who never read junk (unsolicited, third-class) mail. To account for these nonreaders, in this paper we propose a mixed beta-logistic model. The procedure for updating estimated response probabilities using this mixed model is derived. The paper goes on to clarify an independence assumption which underlies the use of both of these models. The implications of this independence assumption will restrict the use of the models to situations where resolicited prospects behave as if they have no memory of previous solicitations. © 1998 John Wiley & Sons, Inc. and Direct Marketing Educational Foundation, Inc.