'Seemingly-Beneficial' Interventions

Problem Definition and Academic/Practical Relevance: Organizations routinely introduce hitherto-unexplored interventions aimed at improving their supply chains. Consider a principal (e.g., a firm or a social planner) that implements a “seemingly-helpful” intervention, broadly interpreted as follows: For any fixed actions of the principal and the agents (e.g., consumers), the payoff of the principal is higher in the presence of the intervention relative to that in its absence. While one would expect such well-intentioned interventions to benefit the principal, several papers within the operations management (OM) literature show that the principal's equilibrium payoff can be hurt, even when the implementation costs of the intervention are ignored. While such a conclusion is often based on the analysis of a single-shot, simultaneous-move game, a repeated-interaction setting can also serve as an appropriate environment in many cases. A fundamental question then arises: Does this conclusion hold even under repeated interactions over the long run?Methodology: We study this question using the framework of infinitely-repeated games and the notion of a precommitment equilibrium from the literature on reputation in repeated games.Results and Managerial Implications: We identify two key characteristics that determine whether a seemingly-beneficial intervention will help, or can possibly hurt the firm: (i) the nature of the intervention (ceteris paribus, does the intervention induce the agents to react in a manner favorable to the principal?), and (ii) the extent of interaction (single-shot at one extreme and infinitely-repeated at the other). Interestingly, we demonstrate the following two possibilities using settings analyzed in the recent OM literature: (a) seemingly-beneficial interventions can hurt the firm in a single-shot analysis but benefit the firm under repeated interactions, and (b) seemingly-beneficial interventions can continue to hurt the firm even under repeated interactions. We also obtain a set of easy-to-interpret sufficient conditions under which the benefit of such interventions can be confirmed under repeated interactions. We illustrate our findings on the settings analyzed in two recent papers