Abstract

Previous studies on laboratory newsvendor problem observed the pull-to-center (PTC) effect, that decision makers tend to concentrate their orders around average demand and fail to optimize the expected profit. However, existing experiments in this context have been limited to random demand with uniform distributions. This study examines the robustness of the PTC effect under a boarder range of demand distributions and estimates whether some reference-point based models may explain the observed behavior of the decision makers.

This study is based on three sessions of newsvendor experiments with performance-based compensation. In each session, subjects are provided with a uniform, normal, or fat-tail demand distribution and two profit scenarios. For each profit scenario, subjects place their order quantities before running a computer simulation that determines the actual demand and their realized profits. Parallelly, simulation behavioral data can be generated based on five different reference-point based models with a wide spectrum of risk and loss aversion parameters.

The study provides statistical comparisons between the experimental data and simulation data from reference-point based models which concludes that in terms of measure of errors, the stochastic mean model performs the best in the low-profit scenario across all three demand distributions, whereas the status quo and participation payout models perform the best in the high-profit scenario when demand follows a uniform distribution. Moreover, the statistics show that the security-based model performs the best in the high-profit scenario when demand follows a normal distribution whereas the participation payout model performs best in the high-profit scenario when demand follows a fat-tail distribution. This study also discusses how demand distribution can affect the newsvendor decision behavior.