Probability Weighting and the Newsvendor Problem: Theory and Evidence

Abstract

Behavioral studies have consistently reported the pull-to-center (PTC) effect observed in laboratory studies of the newsvendor problem. We examine whether this and some other observed effects can be reconciled under the general framework of prospect theory without special assumption on reference point. Specifically, we allow decision makers to value each potential outcome with decision weight rather than the actual probability and further validate through some experiments if the PTC effect is correlated with the shape of the weighting function. The results confirm that the general framework of prospect theory can explain the PTC effect without special assumption on reference point. The only stipulation is a general set of probability weighting functions that that admit underweighting of small probabilities. The proposed model is inclusive, robust, and can explain a number of prominent newsvendor behavioural observations with reasonable benchmark predictions. Our finding suggests that the systematic, suboptimal decisions in behavioural newsvendors could be due to over-focusing on the big picture and neglecting rare events. Accordingly, both academicians and practitioners should revisit the design of decision support systems as well as competitive strategies that involve human decision making against large number of uncertain future outcomes.