Coordination is critical to the social interactions of everyday life, as well as solving important problems such as how to avoid bank runs or how to get a group of people to work together as an effective team. In this Doctoral Dissertation Research Improvement Grant, the co-PI will conduct research to better understand coordination problems like these where strategic uncertainty makes the Pareto efficient equilibrium risky. That is, when an ideal outcome clearly exists, how does uncertainty about the actions of others undermine success and how can such failures be avoided?
By using controlled laboratory experiments, the researchers will isolate key features of these real world problems in order to understand what leads to success or failure at a fundamental level. The first experiment uses simple messages and explicit costs in a coordination game with only two players and two strategies. Even in this highly idealized environment, the results show that very small communication costs dramatically decrease the use of messages but increase the frequency of tacit coordination. The second experiment follows the same approach but uses many players and many strategies in order to determine whether the previous findings apply to more complex interactions. The third experiment explores behavior in an environment that more closely models that of real firms, including free text chat and opportunity costs (as opposed to out-of-pocket costs) of communication. Each of these experiments builds directly from existing research and together will form the groundwork for a better understanding of costly communication and its effects on strategic uncertainty and coordination.
On a strategic level, it is tempting to think that when people have closely aligned interests, coordination is easy. But in a world where communication requires time and other resources, people often fail to achieve good outcomes, even when they have the same goals. One underlying reason is that uncertainty about what actions other people are going to take (strategic uncertainty) can undermine confidence and, in a self-fulfilling way, lead to outcomes that are worse for all involved. The research supported by this award uses economic laboratory experiments to explore this phenomenon and considers how various features of organizations help overcome the coordination problem. One way in which organizations may circumvent some of the problems that arise when communication is costly is through the use of hierarchy. A focus of this research is the extent to which the ability of organizational members to send messages to a manager impacts the credibility of communication from that manager. The experiment shows that such ``open door'' policies do indeed enhance the mutual confidence that other players will follow the recommendation. That is, being able to hear input from subordinates enhances a manager's credibility and can help improve coordination within an organization. However, we also find that there is a hidden cost to such communication structures. Namely, after receiving discouraging messages from subordinates, the manager often passes on this (self-fulfilling) pessimism about the group's ability to coordinate. We conclude that a better understanding of how communication channels impact the credibility of leaders, along with the wisdom of managers to ignore counterproductive feedback, has significant potential to improve coordination within organizations. This work demonstrates how simple laboratory experiments can help us understand how people with common interests can more effectively coordinate their actions. By isolating key features of such problems, we identify underlying drivers of success or failure in real organizations. Though many institutions for solving collective action problems emerge organically without the aid of such research, controlled laboratory experiments can give insight into precisely why they succeed or fail. In doing so, this line of work seeks to identify how such institutions can be improved or applied in new ways for the benefit of those concerned.
