Analyzing learning effects in the newsvendor model by probabilistic methods

Abstract

In this paper, we use probabilistic methods to analyze learning effects in a behavioral experiment on the newsvendor model. We argue why we should believe that suggested orders follow a multinomial logit distribution, and use the single parameter in that model to extract information on learning effects. We revisit the data, analyzed previously by Bolton et al. (2012), and show that our model predicts the pull-to-center effect in these experimental data very well.