Behavioral Aspects of Newsvendors: a study of their variability

For this dissertation, we propose an analysis of the variability in the answering behavior over time of respondents participating in the Newsvendor Model tasks. When posed in front of the Newsvendor task participants seems to be prone almost to every kind of mistake made (O’Keefe T. Carlson, 1969), with deep oscillation of the quantity purchased during each repetition of the model. In other words, the orders placed by the newsvendors not only differ from the optimum at the aggregate level, as already indagated by many prominent authors, but they also differ from each other in terms of how far they are from it. These points directly translate into a different level of economic inefficiency or efficiency presented by the same single participant. As put in the apt word of Daniel Kahneman, our ability to look at the past and to infer the right thing from it is sorely lacking. Humans seem to be, as a matter of fact, not prone to “rational decision-making” and, even more, unlikely to decide in a way that could be defined as “economically” efficient or rational. The idea to focus on this aspect of a vast and rich topic such as the Newsvendor Model emerged from the presence of a gap in the literature on this research field, in which it seems that close to no contributions studying this characteristic are present. The investigation was focused on not only studying variability present in the orders placed by participants but also on their forecast of the expected demand level for each period. Indeed, the two variables present enough differences to be studied separately. Starting from this assumption we tried to investigate the main drivers of the variability in the answering patterns by focusing on the levels of demand forecasted and exploring, through the usage of Analysis of Variance (ANOVA), the effects of different treatments on the respondents’ behavior when forecasting. Some particular combinations of various factors held some interesting insight: specific combinations of framing and product margins can lead respondents to have more optimistic forecasts than what would have been forecasted by an ideal optimal newsvendor participant.