Neural network models for initial public offerings

Abstract In this article, we construct models that predict the first-day return of an initial public offering. Our data set consists of the first-day returns for 1075 firms that went public between 1989 and 1994 and information that we gathered on 16 predictor variables. We segment the data set into technology and nontechnology offerings and construct three types of models for each segment — a regression model and two neural network models. Factorial experiments are used to construct the neural network models. We find that the neural network models perform well on both types of offerings.