Decision Bias in the Newsvendor Problem: Real-world Evidence From Airline Flight Scheduling

The newsvendor model and its several variants have been analyzed extensively in the OM literature. However, the behavioral operations domain suggests that managerial decisions may deviate from the newsvendor optimal solution because of judgmental biases, such as "mean anchoring" and "demand chasing." It is noteworthy that these studies were primarily conducted in laboratory settings and mostly used students or working professionals as test subjects. There is scant empirical evidence suggesting such biases exist in real-world managerial decisions, particularly in a newsvendor setting. Also, while these biases have been established at an individual level, it is unknown if they carry over to the firm (or industry) level. We test whether behavioral biases exist in real-world decisions in this paper.

We borrow the framework provided in Deshpande and Arikan, 2012, to make the connection between the classical newsvendor problem and flight scheduling decisions. We use panel data on observed flight scheduling decisions over a ten-year period in the US domestic airline industry to test for behavioral biases. Our results show evidence for the mean anchoring and demand-chasing biases in airline flight scheduling decisions. This highlights that these biases are observed not only in experimental studies but also exist in real-world decision-making with advanced decision support systems, thus establishing the robustness of the results found in lab settings. In contrast to earlier research, we observe that these biases are persistent with no learning over time. Our data and setting also show that these biases are widely prevalent, i.e., not confined to specific settings. Essentially, the biased decision-making at the individual managerial levels translates to the firm level and, eventually, at the industry level. We believe this is the first study that establishes the existence of these biases at the firm and industry level in a real-world setting. We also show that these biases are influenced by profit margins and demand variability. Finally, our work contributes to the behavioral OM literature by identifying a new bias that arises in dynamic settings - the trend-chasing bias - which adds to the behavioral biases that can cause deviations from optimality.