Behavioral Inventory Management: Supply Uncertainty, Policy Choice, and Stockout Responses

Abstract

This dissertation combines three papers that examine behavioral factors influencing inventory decisions. Chapter 2 shows that two nearly equivalent inventory policies, a reorder point policy and a Kanban policy, can lead to different decision behavior due to differences in framing and decision space. A laboratory experiment compares performance under the two policies in a repeated decision environment with feedback. The results show statistically significant profit differences in the first decisions, before learning effects reduce the performance gap. A second study uses a survey to examine whether the observed decision patterns also occur under different demand settings. The main findings suggest that participants interpret the two policy parameters differently across policies. Under the reorder point policy, the parameters appear to be interpreted additively, while under the Kanban policy they appear to be interpreted multiplicatively. In addition, the restricted decision space under Kanban excludes several poorly performing parameter combinations that remain possible under the reorder point policy. These findings are relevant for the selection and implementation of inventory policies, as initial performance differences may justify investment in a Kanban setup. However, because these differences diminish quickly, such investment appears relevant primarily for high-value goods. Chapter 3 examines how exogenous stochastic lead times influence ordering decisions in the downstream firm. A laboratory experiment compares ordering behavior under constant lead times and stochastic lead times. The results show that stochastic lead times reduce the aggressiveness of inventory corrections compared to constant lead times. This weaker adjustment behavior contributes to higher inventories, backlogs, or both. In addition, participants report higher stress under stochastic lead times than under constant lead times. A further comparison examines deterministic but dynamic lead times against stochastic lead times. Providing information about future lead times reduces uncertainty and lowers reported stress, but does not significantly reduce total costs. Participants use the additional information to anticipate longer lead times, which reduces backlogs but increases inventory holdings. The results therefore indicate that stochastic lead times lead to higher costs and that lead time information alone may reduce stress without necessarily improving cost performance. Chapter 4 addresses how consumers respond to stockouts in fast moving consumer goods. This is relevant because inventory models often assume pure backlogging, whereas empirical evidence from the early 2000s on consumer responses to stockouts challenges this assumption. Given the age of this evidence, this chapter reinvestigates whether this claim still holds. Chapter 4 uses contemporary survey data and compares the results with selected earlier studies. The findings show that stockout response distributions remain stable for some product categories but differ significantly for others. The results also indicate that the effect of certain product characteristics, such as whether a good is hedonic, differs from earlier evidence. In line with prior studies, substitution within the same store remains the dominant response. Postponement, which is most closely related to backlogging assumptions, is the smallest response category. By contrast, store switching and no purchase account for a sizable share of responses and point to the importance of lost sales in inventory model specification.