An Exploration of Behavior in Dynamic Games

University Of California-Santa Barbara, Santa Barbara CA

Investigators

Abstract

Economic models of dynamic environments are highly important for forming policy, and applied theory for these settings has been influential across the entire economic spectrum: public finance, environmental economics, macroeconomics, labor economics, and industrial organization. However, while economic theory for dynamic environments is well-developed, in general it does not provide a precise prediction. Quite the opposite. General theorems instead indicate that most anything can happen in a dynamic environments, dependent on how the participants react to both contemporaneous features of the environment (prices, the number of other competitors in an industry, atmospheric pollutants) and observed features of the past (in particular, other actors behavior). To deal with indeterminacy in the predictions applied economic theory has instead relied on strong assumption - for the most part untested. These assumptions serve to examine both theorized effects from a policy change, but also to produce estimates for use in policy. While these assumptions can make complex environments tractable, failure of the assumption has the potential to produce very different policy outcomes than those posited. Where a societal gain was predicted, a loss may instead be observed. This project's aim is to use controlled observations of human behavior in dynamic environments to examine when these theoretical assumptions are likely to hold true. Greater insight into what features of the environment drive will allow for more-robust policy discussions, and a subsequent benefit to society from better economic policy. The project addresses a topic with growing applied interest in economics, but where theory does not (generically) make precise predictions. It will provide evidence from human behavior, with the overarching aim being the construction of predictive selection criteria to indicate in which settings the standard assumptions are likely to hold true. Further, in those settings where the standard assumptions fail, the project aims to provide evidence for alternatives. Such alternatives can provide greater power when specifying alternative hypotheses. In this project, the PIs propose a series of experiments that will examine behavior in dynamic strategic environments. Four sub-projects are proposed: i) an examination of how uncertainty over other participants' strategic choices affects selection; ii) an examination of the extent to which teams and individuals behave differently in these settings; iii) an examination of the information revealed by self-interested experts across a long-run relationship; and iv) an examination of the effects on behavior from the number of active participants in an environment. In all four sub-projects, the PIs construct a simple baseline environment, and several modifications of it, each of which is chosen to isolate and measure the effects from a relevant feature of interest. Broader impacts from the study follow from a greater understanding of which facets of a strategic environment lead humans to focus on observations from the recent past (historical outputs, prices, extraction levels, etc.) to determine their present actions, as opposed to current conditions (input prices, demand, number of competitors, etc.). In terms of policy, the results aim to produce evidence-based criteria for the most-common assumptions in the applied literature on dynamic games. Greater insight into what drives selection will allow for more-robust policy discussions, and subsequent benefits to society.