Research

Abstract: The theory of continuous time games (Simon and Stinchcombe (1989), Bergin and Macleod (1993)) shows that  continuous time interactions can generate very different equilibrium behavior than conventional discrete time interactions.  We introduce new laboratory methods that allow us to eliminate natural inertia in subjects' decisions in continuous time experiments, thereby satisfying critical premises of the theory and enabling a first-time direct test.  Applying these new methods to a simple diagnostic timing game we find strikingly large gaps in behavior between discrete and continuous time as the theory suggests.  Reintroducing natural inertia into these games causes continuous time behavior to collapse to discrete-time like levels in some settings as predicted by Nash equilibrium.  However, contra Nash equilibrium, the strength of this effect is fundamentally shaped by the severity of inertia:  behavior tends towards discrete time benchmarks as inertia grows large and perfectly continuous time benchmarks as it falls towards zero. We provide evidence that these results are due to changes in the nature of strategic uncertainty as inertia approaches the continuous limit.

Abstract: We evaluate data on choices made from Convex Time Budgets (CTB) in Andreoni et al. (2012) and Augenblick et al. (2015), two influential studies that proposed and applied this experimental technique. We use the Weak Axiom of Revealed Preference (WARP) to test for external consistency relative to pairwise choice, and demand, wealth and impatience monotonicity to test for internal consistency. We find that choices made by subjects in the original Andreoni et al. (2012) paper violate WARP frequently; violations of all three internal measures of monotonicity are concentrated in subjects who take advantage of the novel feature of CTB by making interior choices. Wealth monotonicity violations are more prevalent and pronounced than either demand or impatience monotonicity violations. We substantiate the importance of our desiderata of choice consistency in examining effort allocation choices made in Augenblick et al. (2015), where we find considerably more demand monotonicity violations, as well as many classical monotonicty violations which are associated with time inconsistent behavior. We believe that the frequency and magnitude of WARP and monotonicity violations found in the two studies pose important confounds for interpreting and structurally estimating choice patterns elicited through CTB. We encourage researchers employing CTB in present and future experiments to include consistency tests in their design and pre-estimation analysis.

Abstract: This paper studies, using a laboratory experiment, the effects of uncertainty aversion (the union of risk aversion and ambiguity aversion) on behavior in a normal form game. We isolate and identify two components of uncertainty aversion in games: the first-order effect of an agent's own preferences and the second-order effect of the agent's beliefs regarding their opponent's preferences. In this environment, theory is consistent with both a large effect of uncertainty preferences on behavior and no effect of uncertainty preferences on behavior. We find evidence of both first-order and second-order effects of uncertainty aversion on strategic behavior. The second-order effect of risk aversion is larger than the second-order effect of ambiguity aversion. The results may be helpful in guiding future theoretical advances.

Abstract: We study the use of mixed strategies in games by ambiguity averse agents with a preference for randomization. Applying the decision theoretic model of Saito (2015) to games, we establish that the set of rationalizable strategies grows larger as preference for randomization weakens. An agent's preference for randomization is partially observable: given the behavior of an agent in a game, we can determine an upper bound on the strength of randomization preference for that agent. Notably, data in previous experiments on ambiguity aversion in games is not consistent with a maximal preference for randomization for approximately 30% of subjects. 

Law enforcement officials face numerous decisions regarding their enforcement choices. One important decision, that is often controversial, is the amount of knowledge that law enforcement distributes to the community regarding their policing strategies. Assuming the goal is to minimize criminal activity (maximize citation rates), our theoretical analysis suggests that agencies should reveal (shroud) their resource allocation if criminals are uncertainty seeking, and shroud (reveal) their allocation if criminals are uncertainty averse. We run a laboratory experiment to test our theoretical framework, and find that enforcement behavior is approximately optimal given the observed non-expected utility uncertainty preferences of criminals.

Individuals often possess private information about the common value of a public good. Their contributions toward funding the public good can therefore reveal information that is useful to others who are considering their own contributions. This experiment compares static and dynamic contribution decisions to determine how hypothetical contingent reasoning differs in dynamic decisions. The timing of individuals' sequential contribution decisions is endogenous. Funding the public good is more efficient with dynamic than static decisions in equilibrium, but this requires decision-makers to understand that in the future they can learn from past events. Our results indicate that a substantial fraction of subjects appreciate the benefits of deferring choice to learn about and condition their behavior on the contribution decisions of others. Many subjects, however, exhibit a bias away from rational choices in the direction of Cursed equilibrium, and some appear to extract information only from prior, and not concurrent, behavior.

Do experimental subjects have consistent first and higher-order beliefs about other's preferences? How does any inconsistency affect strategic decisions? We introduce a simple four-player sequential social dilemma where actions reveal first and higher-order beliefs. The unique sub-game perfect Nash equilibrium (SPNE) is observed less than 5% of the time, even though our diagnostic treatments show that a majority of our subjects are self-interested, higher-order rational and have accurate first-order beliefs. In our data, strategic play vastly deviates from Nash predictions because first-order and higher-order beliefs are inconsistent for most subjects.

We study the value of and the demand for instrumentally-valuable information in a simple decision environment where signals are transparently biased. We observe remarkable sophistication in information aggregation and acquisition. A majority of our subjects (63%) made unbiased reports even when faced with biased signals and the few subjects who made biased reports were split between under- and over-correcting for the signal bias. When allowed to buy pairs of opposite or similarly biased information sources, subjects actively shopped for diverse information at personal costs, and their demand for diverse information reacted rationally to its value and cost. Subjects who were worse at aggregating information, were more likely to purchase diverse signals, perhaps in an attempt to make their inference problem easier. Our results advocate for greater transparency in media bias, so that individuals can choose the right portfolio of information to make better choices. 

This paper introduces and studies the Network Dilemma, which embeds a social dilemma within a network formation game. The game models the voluntary provision of a costly public good that is locally non-excludable within an endogenously formed network of agents. In this environment, all equilibrium networks are star networks that exhibit ex-post payoff inequality between the central public good provider and peripheral players. We observe high levels of equilibrium behavior, despite the coordination problem faced by the ex-ante homogeneous subjects. In our treatments with a moderate level of ex-post payoff inequality, coordination on a star network appears to be facilitated by subjects who signal a willingness to volunteer as the center persistently. We use a sequential version of the Network Dilemma, and a diagnostic treatment, to find that aversion to behavioral uncertainty is the main motivating factor behind voluntary public good provision in the Network Dilemma. 

In normal form games, when agents exhibit ambiguity aversion the exclusion of mixed strategies from agents' choice sets can enlarge the set of equilibria. While it is possible, in a game theoretic experiment, to enforce pure strategy reporting it is not possible to prevent subjects from mixing before reporting a pure strategy. This short paper establishes conditions under which the set of equilibrium in a game with ambiguity averse agents and pure strategy reporting is invariant to the existence of pre-play mixing devices. This result is crucial for the interpretation of recent experimental work on the role of ambiguity aversion in normal form games.

We estimate the effects of introducing an emissions offsets scheme (i.e. where producers are paid to reduce their greenhouse gas emissions) in the Australian agriculture industries. We find that an emissions offsets scheme has the potential to induce substantial reductions in greenhouse gas emissions in the long run (8 million tonnes of carbon dioxide equivalent per annum by 2030). We also find, however, that the implicit subsidy for (what are currently) highly emissions intensive industries induces a substitution effect away from low emissions intensity grains production into high emissions intensity beef and sheep meat production.