The allocation of decision rights—and the decision hierarchies it creates—shape incentives in organizations and thus individual motivation. We study whether the motivational effects of decision hierarchies go beyond pecuniary incentives and how they relate to the perceived legitimacy of the decision hierarchy. In a laboratory experiment, we exogenously manipulate the payoff structure of an organization in a way that leaves the pecuniary incentives of all involved parties unaffected but shifts subordinates’ perceived legitimacy of the decision hierarchy. Our data show that subordinates’ motivation to provide effort is causally affected by the organizational payoff structure, and significantly associated with the perceived legitimacy of the decision hierarchy.
We show that greater memory overload results in a disproportionately greater underreaction to belief-challenging information compared to belief-confirming information. In an experiment, we keep constant the information subjects receive but vary the difficulty of memorizing the signals. In the treatment condition, the preceding signal disappears as the next signal appears; in the control condition, the preceding signal remains visible. Memory load is therefore greater in the treatment than in the control condition. We find that in the treatment condition, subjects underreact more to belief-challenging information than in the control condition, but updating to belief-confirming information is not significantly affected. Our results caution against information campaigns that ignore the roles of memory load and attention.
Information in repeated real-world interactions is rarely fixed. Instead, over time, players typically learn more about the game structure, understand more about their own payoff correspondences, and find out what others did and earned in the past. However, how this dynamic information path influences learning and which aggregate outcomes are reached as a result have not been investigated to date. To study this, we conducted a series of laboratory experimental games where we provided more information over time in different orders and along different paths. These games span two strategically distinct classes: a Cournot game with a unique symmetric Nash equilibrium, and step-level public-goods (coordination) games with multiple Nash equilibria of differing Pareto efficiency. Our evidence confirms a natural mapping from information to predominant learning rule: information about own payoffs triggers payoff-based learning, feedback concerning others’ realized payoffs triggers imitation, and structural information about the game triggers best response. In addition, when multiple learning rules are feasible, which one dominates depends on information paths. In particular, feedback about others’ actions and payoffs, especially when supplied marginally, may trigger persistent imitation that locks into Pareto-inferior Nash equilibria, even when more information later becomes available.
A principal wants to rank and reward teams in an organization. Agents perfectly observe the local ranking of their teams, and strictly prefer being on higher-ranking teams. I show there exists an ex-post incentive compatible and efficient team ranking mechanism if (i) there is an agent who belongs to every team (supervisor condition), or (ii) whenever two teams share an agent in common, they share at least two agents in common (connectivity condition). I identify a class of organizations for which these conditions are both necessary and sufficient. In the special case where the principal wants to rank every group of co-workers in the organization, there exists a team ranking mechanism if and only if the organizational network is a star.
Reinforcement learning algorithms play an increasingly important role in economic situations. These situations are often strategic, and the artificial intelligence may or may not be cooperative. We compare human and algorithmic cooperation rates in the infinitely repeated two-player prisoner's dilemma and study which strategies they choose to cooperate and punish deviations. Through a sequence of computational Q-learning and human-player experiments, we find that our Q-learning algorithms tend to cooperate less than humans, particularly when cooperation is risky or not incentive-compatible. Algorithms often use different strategies than humans, leading to distinct on-and off-path behavior.
We study optimal mechanisms for a data broker selling market segmentation to a producer. The producer engages in discriminatory pricing and is privately informed about a payoff-relevant parameter. We first characterize a class of markets from which the segmentations in optimal mechanisms can be constructed. If the profit-maximizing prices of different producer types are ordered and only downward or only upward incentive constraints bind, this analysis shows that in any created market either all consumer types are served or just a single one. We then characterize optimal mechanisms for a valuation structure that captures differing product quality. In particular, the consumers are segmented into markets in which everyone is served. Lastly, we give a sufficient condition under which optimal mechanisms are efficient regardless of the binding constraints.
In evolutionary game theory, when a population of players with the same preferences may freely choose a location in which to interact with other players, previous literature has shown that long-run evolutionary forces select in favor of payoff-dominant equilibria. In this paper, I extend this location choice aspect of evolutionary game theory to the case of two different payoff types of players in 2 & times; 2 games. I show that whether or not payoff-dominant equilibrium selection occurs depends on the precise interaction of types, especially whether they exhibit mutual strategic homophily or a mix of homophily and heterophily. The former case guarantees payoff-dominant equilibria, and the latter case may not, because coordination becomes imperfect and may settle on an inefficient strategy.
We develop a dynamic model of autocracy in which repression helps quell acute threats, increasing an autocrat’s survival probability or deterring challenges altogether, but at the expense of policies that reduce popular discontent and future opposition mobilization. Our main contribution is characterizing a unique stationary Markov Perfect equilibrium, which takes one of three forms: deterrence, where the autocrat is never challenged; instability, where challenges occur every period; or a deterrence/contention mixed-strategy equilibrium with periods of deterrence and periods with challenges. The mixed-strategy equilibrium arises because the opposition is easier to deter today when they expect opportunities to challenge in the future. We then examine indirect foreign interference–such as sanctions or withholding aid–aimed at protecting the opposition or promoting democratization. Even under ideal conditions, such interference cannot achieve both goals simultaneously.
We propose a new dominance notion in strategic games—weak* dominance—that reflects the idea of cautious* rationality, whereby a rational player believes that each opponent’s individual actions are played with positive probabilities. In this paper, we show that an action is weakly* dominated if, and only if, it is not a best response to any marginally full-support conjecture. We characterize Gul’s (1996) notion of perfect τ-rationalizability through an iterative procedure (S∞W*): removing all weakly* dominated actions in the first round, and then removing only strictly dominated actions in the subsequent rounds. Moreover, we provide epistemic foundations for the iterative procedure in a lexicographic type structure model.
We extend Kuhn’s Theorem to games of the extensive form with unawareness. We prove that if a game of the extensive form with unawareness has perfect recall, then for each mixed strategy there is an equivalent behavior strategy. We show that the converse does not hold under unawareness without restricting the evolution of the player’s awareness to constant awareness along paths of play. Both directions of Kuhn’s Theorem for games of the extensive form with unawareness require a condition complementary to perfect recall that rules out falsely believing in some events when the player is unaware of the actual past events.
Each agent in our model occupies a position in a hierarchy (a directed tree) and generates returns through collaborating with his superiors. Thus his superiors can also claim their ownership rights over his returns. Our main axiom for the allocation of these returns is the standard monotonicity with regard to the collective ownership. We establish axiomatic characterizations of monotonic allocation rules. They are represented by hierarchical transfers of the returns at each position to the superiors. When the rate of transfer is symmetric, the rules coincide with the geometric rules (ownership rates constitute a geometric sequence, moving up the hierarchy). Other fair allocation rules are also available, when the rate of transfer is asymmetric; a focal example is the hierarchical equal sharing rule, which is shown to be the unique one with an equal treatment axiom.
We propose an experimental design and data-analytic method for eliciting and distinguishing the strict-preference, indifference, and indecisiveness components of individual preferences in general choice environments. The design combines a forced-choice treatment with a free-choice treatment. In both treatments, subjects may select multiple alternatives from a menu. In the free-choice treatment, subjects may also avoid or delay choice at a small expected cost. To analyze such data, we extend a standard non-parametric goodness-of-fit criterion to accommodate multi-valued choices. We apply it to evaluate the consistency of subjects' 50 decisions with utility maximization and two models of incomplete-preference maximization. Around 55% of subjects are well explained by one of these models, with 33% and 22% best explained by utility and incomplete-preference maximization, respectively. Revealed preferences typically feature non-trivial indifferences, and those that are incomplete often exhibit the predicted theoretical distinctions between indifference and indecisiveness, which are documented empirically for the first time.
This paper examines effort-maximizing move orders in two-player Tullock contests across varying levels of contest accuracy and player asymmetries. Unlike prior work that favors strong-lead sequential formats in lottery contests, our analysis demonstrates that all three formats-stronglead, weak-lead, and simultaneous-can be optimal in the general Tullock contests. Strong-lead contests are most effective in noisy environments, but lose their advantage as contest accuracy increases. Weak-lead contests may become optimal when the weaker player successfully preempts. Simultaneous-move contests tend to perform better as the players' ability differences become less severe. Overall, the optimal contest design jointly depends on the contest accuracy levels and relative strengths of the contestants.
We study a class of social learning games in which players choose whether and when to make an irreversible investment. We ask whether the option value of learning induces delay. Surprisingly, under natural and standard assumptions, delay does not arise in equilibrium. Although waiting even momentarily would yield informative observations, we show that this learning is payoff-irrelevant. As a result, players either invest immediately or not at all. Lab rat 1: "Did you get the COVID vaccine yet?" Lab rat 2: "Nah, I'm waiting for the results of the human trials."
We theoretically and experimentally study centralized college admissions in which colleges evaluate students under a 'translucent' admission system and students can learn each college's suitability through costly information acquisition. In centralized matching via Gale and Shapley's deferred acceptance algorithm, students decide whether to acquire information before submitting their rank-order lists. However, uncertainty about the final assignment lowers the expected gain from learning, thereby reducing social welfare, compared to a scenario without such uncertainty. Our experiments demonstrate that the welfare loss is greater with more opaque admission systems. The empirical social welfare obtained in our experimental treatments is consistently lower than the theoretical welfare, and we identify non-equilibrium learning as a main contributor.
This article examines, both theoretically and experimentally, the welfare implications of implementing symmetric bid caps in lobbying competitions. We model lobbying games as lottery contests with two radicals and one centrist, where all players strictly prefer to win. However, if they lose, radicals prefer to lose to a centrist rather than to the opposing radical. We establish the existence and uniqueness of equilibrium and find that, in the absence of a bid cap, radicals may bid aggressively enough to drive the centrist out of the competition, thereby preventing the social welfare-maximizing outcome. Consequently, implementing a bid cap enhances social welfare in two ways: it reduces aggregate expenditure and increases the likelihood of the centrist winning. Our experimental results qualitatively align with the theoretical predictions. However, we also identify and address notable behavioral patterns: in contrast to the vast experimental literature (Sheremeta, 2013) on contests, aggregate bids do not exceed theoretical predictions in all treatments.
We propose a global games approach to the standard two-stage entry game. The entry decisions in the first stage are strategic substitutes. The second-stage game of product market competition reflects a fundamental common value "market attractiveness" parameter, about which firms get private noisy signals. The main result establishes the selection of a unique equilibrium in the entry game, as noise vanishes, in cut-off strategies implying efficient entry. This provides a theoretical foundation for the equilibrium selection commonly used in entry models in the empirical literature. In addition, using supermodularity techniques, we provide novel conditions of independent interest on the primitives (demand and cost functions) of market competition to justify our assumptions for Bertrand and Cournot competition. These include results on the effects of entry and demand shifts and highlight the critical relevance of the property of log-supermodularity of demand.
