Research

We consider various mechanisms for validating and verifying data through human labor, a problem frequently encountered in managerial applications such as AI data labeling. Under the traditional audit mechanism, an auditor randomly inspects the output of a worker. Alternatively, under the agreement mechanism, a principal hires multiple workers and only investigates conflicts in their work. Combining these features, we introduce the random agreement mechanism, and show that it is less costly than either mechanism. Using experiments to test these predictions, we discover that subjects substantially deviate from the equilibrium prediction of full output, and interestingly, produce 20 percentage points more output under the agreement mechanism. However, because subjects do not complete all tasks, the agreement mechanism is substantially more costly than theory would predict. As a result, the random agreement mechanism is still the most efficient in terms of cost per task completed.

This paper demonstrates how social norms for meritocratic evaluations can be exploited to incentivize good service, high effort, and contributions to the public good. It develops a theoretical model showing how peer evaluation norms (e.g., meritocratic, truth-telling, or collusive) affect equilibrium effort and welfare. Our experimental results demonstrate subjects tend to follow meritocratic norms, and that appropriately structured peer evaluations are an effective means of increasing contributions to the public good and welfare. We find that groups that follow more meritocratic norms achieve higher welfare than groups that follow less meritocratic norms.

The two most prominent strategy-proof school choice mechanisms, Deferred Acceptance and Top Trading Cycles, offer a trade-off between eliminating justified-envy and Pareto efficiency, respectively. We introduce a novel, biometric approach to investigate how much individuals care about and physiologically react to each property. Using an experimental design that randomly varies whether subjects learn about others’ assignments in a school choice game, we measure arousal through galvanic skin response and valence through facial expressions when subjects experience envy with and without justification. We find increased arousal associated with receiving successively lower-ranked allocations, i.e., the (unjustified) envy of the allocation(s) of others. We note additional arousal when that envy is justified, but that result requires an explicit message to accentuate the justification. Eye-tracking data confirm that subjects do not notice markers of justified-envy until receiving this message. Facial-expression valence yields less conclusive results, but this may be due to the limitations of that method for anger detection.

Oil field unitization is a unique, high-stakes economic problem involving multilateral bargaining among many agents with certain positive surplus from agreement. We theoretically characterize and experimentally test the problem under three information structures: complete information, symmetric incomplete information, and asymmetric information in the form of private information about holdings. Our cooperative bargaining theoretical framework suggests that agreement is always possible under both symmetric information structures but may fail under asymmetric information. Conditional on agreement, the returns to private holdings should be highest under asymmetric information, and weakly higher under complete information than under symmetric incomplete information. Experiments confirm that agreement rates are lowest under asymmetric information and similar under both symmetric information structures. However, returns to private holdings are discounted similarly under asymmetric and symmetric information, with both falling below those under complete information. It appears that subjects consistently impose norms that discount holdings that are not commonly known with certainty. We re-examine evidence from failed unitization attempts in the field and find they are consistent with this explanation.

To ease consumer choice, regulators often require producers to make standardized and comparable disclosures over key product attributes. At the same time, product advertising may emphasize similar comparisons, albeit over strategically chosen attributes. We show that individuals may conflate these two cases; largely failing to understand that self-serving disclosures are uninformative, they are consequently exploited by producers. The fallacy is not due to consumer naiveté regarding producer behavior; rather, subjects incorrectly adjust prior beliefs in response to strategically disclosed information. Thus, for the boundedly rational to benefit from reduced information loads, limited disclosures must be independent of other parties' motives.

Experimental economics research has consistently documented systematic deviations from weakly dominant strategies. Economic theory has introduced behavioral refinements informed by these findings. This paper tests three such refinements in a standardized environment: initial play with value elicitation mechanisms. We utilize a unique design that juxtaposes each refinement with a closely related, “unrefined,” baseline elicitation mechanism. Specifically, using a Becker-DeGroot-Marschak (BDM) mechanism, a dynamic descending-price list, and a descending-price clock mechanism, we test behavioral refinements that simplify the payoff function (contingency-by-contingency), the game structure (game-structure obviousness), and the dominant strategy (obvious dominance), respectively. While no refinement significantly outperforms its baseline, the price list format—regardless of whether it satisfies game-structure obviousness—substantially outperforms both the BDM and clock mechanisms in terms of the rate of subjects' dominant strategy play.

We test whether lying aversion can steer equilibrium selection in mechanism design. In a principal–worker environment, the direct mechanism admits two dominant-strategy equilibria: the designer’s target and a worker-optimal outcome. We show this limitation persists for all robust mechanisms, then ask whether framing misreports as explicit lies helps. We develop a 2 × 2 experiment that varies direct vs. extended mechanisms with implicit vs. explicit messages. We find that framing misreporting of type as an explicit lie shifts play away from the worker-optimal outcome toward truthful reporting, raising designer payoffs with minimal efficiency loss. These findings indicate that lying aversion is an effective lever for aligning behavior with social objectives.

We experimentally evaluate the comparative performance of the winner-bid, average-bid, and loser-bid auctions for the dissolution of a partnership. The analysis of these auctions based on the empirical equilibrium refinement of Velez and Brown (2020) arXiv:1907.12408 reveals that as long as behavior satisfies weak payoff monotonicity, winner-bid and loser-bid auctions necessarily exhibit a form of bias when empirical distributions of play approximate best responses (Velez and Brown, 2020 arXiv:1905.08234). We find support for both weak payoff monotonicity and the form of bias predicted by the theory for these two auctions. Consistently with the theory, the average-bid auction does not exhibit this form of bias. It has lower efficiency that the winner-bid auction, however.

We advance empirical equilibrium analysis (Velez and Brown, 2020, arXiv:1907.12408) of the winner-bid and loser-bid auctions for the dissolution of a partnership. We show, in a complete information environment, that even though these auctions are essentially equivalent for the Nash equilibrium prediction, they can be expected to differ in fundamental ways when they are operated. Besides the direct policy implications, two general consequences follow. First, a mechanism designer who accounts for the empirical plausibility of equilibria may not be constrained by Maskin invariance. Second, a mechanism designer who does not account for the empirical plausibility of equilibria may inadvertently design biased mechanisms.

In some strategic environments, the long-run behavior of boundedly rational agents converges to Nash equilibrium, but in others, behavior may be persistently non-convergent. In this paper, we derive and experimentally test theoretical predictions regarding long-run population-level behavior in all-pay auctions where all but the lowest bidder win a prize. We compare two-prize to one-prize auctions. Consistent with evolutionary stability criteria, we find bidding behavior is more stable in one-prize auctions. Consistent with adaptive models, we observe persistent disequilibrium cycles. These results suggest that stability criteria can help predict whether Nash equilibrium provides a reliable characterization of long-run behavior and that adaptive models can help characterize long-run behavior in unstable strategic environments. As an orthogonal treatment, we also vary the presentation of information given to subjects by focusing on individual payoffs or others’ performance. We note subtle changes in disequilibrium dynamics across informational treatments.

Empirical tests of strategy-proof mechanisms often demonstrate that agents persistently report non-truthful messages. One possibility is that this behavior is consistent with equilibrium play, albeit an equilibrium not intended by the mechanism’s designer. While these undesirable equilibria may exist under many mechanisms, we determine that the actual, empirical observation of such equilibria is only likely when the underlying social choice function violates a non-bossiness condition and information is not interior. Our analysis introduces and relies upon an empirically-based approach to the refinement of Nash equilibrium. A survey of experimental and empirical results on games of this type supports our findings.