Algorithmic Representations of Managerial Search Behavior

作者：Tracy William M^*; Markovitch Dmitri G; Peters Lois S; Phani B V; Philip Deepu

来源：Computational Economics, 2017, 49(3): 343-361.

DOI：10.1007/s10614-015-9559-7

• Greedy algorithms
• Search behavior
• NK models
• Simulation methods
• Behavioral experiments

We use targeted behavioral experiments to test the extent to which greedy algorithms replicate search behavior. Many simulation models use greedy algorithms to represent a firm's trial-and-error based exploration (i.e., backward-looking search). This implies that managers always reject changes that decrease performance relative to the status quo. Although we observe significant heterogeneity in backward-looking search behavior, over 50 % of our subjects deviate from greedy search behavior by occasionally preserving performance-decreasing changes. The likelihood of such preservation was inversely related to the magnitude of the performance decrease. While search behavior is likely context specific, our analysis suggests that non-greedy firm search cannot be dismissed outright. Substituting non-greedy algorithms for greedy ones will alter the behavior of some simulation models used in economic research. We recommend that future work in this area report whether key findings are dependent on the use of greedy or non-greedy search algorithms. We also suggest that researchers explicitly discuss which algorithm best represents backward-looking search in the phenomenon under study.

• 出版日期2017-3