Ridge regression and direct marketing scoring models

Abstract

The objective of a direct marketing scoring model is to pick a specified number of people to receive a particular offer so that the response to the mailing is maximized. This paper shows how ridge regression can be used to improve the performance of direct marketing scoring models. It reviews the key property of ridge regression—it can produce estimates of the slope coefficients having smaller mean squared error than ordinary least squares models. Next, it shows that ridge regression can be used to reduce the effective number of parameters in a regression model. Thus, ridge regression can be used as an alternative to variable subset selection methods such as stepwise regression to control the bias-variance tradeoff of the estimated values. This means that direct marketers can include more variables in a scoring model without danger of overfitting the data. Ridge regression estimates are compared with stepwise regression on direct marketing data. The empirical results suggest that ridge regression provides a more stable way of moderating the model degrees of freedom than dropping variables.