Observation Bias: The Impact of Demand Censoring on Newsvendor Level and Adjustment Behavior
Executive Summary — As the fundamental model for managing inventory under demand uncertainty, the newsvendor model has received significant research attention, but behavioral issues—the focus of this paper—have been less well studied. Nils Rudi and David Drake demonstrate how different aspects of the newsvendor model, a rather complex managerial decision setting, result in a combination of behavioral deviations from the normative solution prescribed within existing literature. The results can help managers prioritize order quantity improvements based on product margins and the degree of demand feedback available in the setting that they operate in. Key concepts include:
- In general, changing how order quantity decisions are made preferably comes through training by building awareness of level and adjustment costs and their sources.
- This research provides managers with insight into how adjusting order quantity policy over time and ordering at a suboptimal level combine to erode profits.
- These results also provide guidance on which of sources of behavioral cost (adjustment cost and level cost) managers are likely to be most exposed to given a product's unit cost and margin and the degree of demand visibility.
In an experimental newsvendor setting we investigate three phenomena: Level behavior—the decision-maker's average ordering tendency; adjustment behavior—the tendency to adjust period-to-period order quantities; and observation bias—the tendency to let the degree of demand feedback influence order quantities. We find that the portion of mismatch cost due to adjustment behavior exceeds the portion of mismatch cost due to level behavior in three out of four conditions. Observation bias is studied through censored demand feedback, a situation which arguably represents the majority of newsvendor settings. When demands are uncensored, subjects tend to order below the normative quantity when facing high margin and above the normative quantity when facing low margin, but in neither case beyond mean demand (a.k.a. the pull-to-center effect). Censoring in general leads to lower quantities, magnifying the below-normative level behavior when facing high margin but partially counterbalancing the above-normative level behavior when facing low margin, violating the pull-to-center effect in both cases.