12.2 Experimental game theory and empirical findings
5 min read•july 30, 2024
Experimental game theory bridges the gap between theoretical predictions and real-world behavior. It uses controlled lab experiments to test game-theoretic concepts and uncover systematic deviations from rational self-interest. This approach reveals fascinating insights into human decision-making in strategic situations.
These experiments have uncovered key findings like cooperation in social dilemmas, concerns in bargaining, and limited strategic reasoning. They challenge standard assumptions and inspire new models that incorporate , , and psychological factors to better explain observed behavior.
Principles and methods of experimental game theory
Experimental methods and design
Top images from around the web for Experimental methods and design
Frontiers | Trust Game Database: Behavioral and EEG Data From Two Trust Games View original
Is this image relevant?
Frontiers | Cooperative Behavior in the Ultimatum Game and Prisoner’s Dilemma Depends on Players ... View original
Is this image relevant?
Frontiers | Trust Game Database: Behavioral and EEG Data From Two Trust Games View original
Is this image relevant?
Frontiers | Cooperative Behavior in the Ultimatum Game and Prisoner’s Dilemma Depends on Players ... View original
Is this image relevant?
1 of 2
Top images from around the web for Experimental methods and design
Frontiers | Trust Game Database: Behavioral and EEG Data From Two Trust Games View original
Is this image relevant?
Frontiers | Cooperative Behavior in the Ultimatum Game and Prisoner’s Dilemma Depends on Players ... View original
Is this image relevant?
Frontiers | Trust Game Database: Behavioral and EEG Data From Two Trust Games View original
Is this image relevant?
Frontiers | Cooperative Behavior in the Ultimatum Game and Prisoner’s Dilemma Depends on Players ... View original
Is this image relevant?
1 of 2
Experimental game theory studies human behavior in strategic situations using controlled laboratory experiments to test game-theoretic predictions and identify systematic deviations
Key principles include inducing valuations through monetary incentives, controlling the information structure, and randomly matching participants to implement one-shot interactions
The strategy method elicits contingent responses for every possible decision node, enabling the study of counterfactual and off-equilibrium behavior
Experiments often implement simplified versions of games like the , , , and to isolate specific strategic considerations
Types of experimental designs
Within-subject designs expose the same participants to multiple treatment variations
Allows for direct comparisons of behavior across different conditions
Enables the study of individual-level heterogeneity in responses to game parameters
Between-subject designs compare behavior across different participant groups
Each group is exposed to a single treatment condition
Mitigates potential order effects or spillovers that may arise in within-subject designs
Statistical analyses test hypotheses about treatment effects, often employing non-parametric tests robust to distributional assumptions
Examples include Wilcoxon rank-sum tests for between-subject comparisons and Wilcoxon signed-rank tests for within-subject comparisons
Seminal experiments in behavioral game theory
Experiments on social dilemmas and cooperation
In the Prisoner's Dilemma, a non-negligible fraction of participants cooperate despite defection being the dominant strategy, suggesting other-regarding preferences or bounded rationality
Public Goods game experiments find positive contributions that decline over time in finitely-repeated play
Introducing punishment opportunities can sustain cooperation
Conditional cooperation strategies are common, where individuals contribute more when others are expected to contribute
Experiments on bargaining and fairness
The Ultimatum game reveals that responders frequently reject low offers, contradicting subgame perfect equilibrium predictions, indicating fairness concerns or negative
Proposers often make equal splits, anticipating the possibility of rejection
Cultural differences in rejection rates have been observed, with higher acceptance of unequal offers in some societies
Trust game studies show that many players exhibit trust and trustworthiness, with the degree varying based on the social context and framing of the interaction
Higher trust and trustworthiness are observed in repeated interactions or when reputation is at stake
Communication and face-to-face interactions can enhance trust
Experiments on market entry and auctions
Market entry experiments find excess entry relative to , especially with a small number of potential entrants, consistent with overconfidence bias
Providing feedback and experience can reduce excess entry over time
Introducing market selection can drive behavior closer to equilibrium predictions
Experiments on common-value auctions document the winner's curse, where bidders systematically overbid and earn negative profits, suggesting a failure to account for adverse selection
The severity of overbidding is reduced with experience or when bidders receive additional information about the value of the object
Experiments on strategic reasoning and learning
Centipede game experiments show that players often fail to backward induct, continuing to play even in the final rounds, implying limited strategic reasoning
Increasing the number of rounds or the stakes can lead to more unraveling
Introducing intermediate payoffs or incomplete information can sustain cooperation
Experiments on learning in games find that players adjust their strategies over time based on feedback and experience
Simple reinforcement learning models capture some aspects of observed behavior
Belief learning models, where players form beliefs about others' strategies and best respond, provide a better fit in some cases
Standard vs. empirical game theory
Assumptions of standard game theory
Nash equilibrium assumes rational self-interested agents with perfect information
Experimental evidence suggests that human players have bounded rationality and social preferences
Deviations from Nash predictions are common, particularly in one-shot or finitely repeated games
Subgame perfect equilibrium requires credible threats at every decision node
Experiments reveal that incredible threats are sometimes carried out due to emotional reactions or fairness concerns
Forward induction reasoning is not always applied, as players may not perceive the full strategic implications of past actions
Mixed strategy equilibrium predicts serially uncorrelated choices that are best responses to the opponent's equilibrium mixtures
Empirical studies find excessive switching and payoff-dependent mixtures
Players often exhibit a "gambler's fallacy" or hot hand beliefs, expecting positive or negative autocorrelation in random sequences
Empirical regularities and extensions
Evolutionary game theory models the dynamics of strategy adoption in populations
Experimental tests identify the importance of noise, mutations, and the speed of the replicator dynamics
Behavioral strategies like tit-for-tat or win-stay-lose-shift can emerge and sustain cooperation in repeated interactions
Standard theories abstract from the framing of the strategic interaction
Experiments manipulating framing find substantial effects on behavior, holding the game structure constant
Presenting a Prisoner's Dilemma as a "Community Game" rather than a "Wall Street Game" can increase cooperation rates
Psychological game theory incorporates belief-dependent preferences like reciprocity, guilt, and surprise
Empirical evidence supports the relevance of these motivations in shaping behavior
Models of inequity aversion, where players dislike unequal payoff distributions, can explain deviations from standard predictions
Strengths and limitations of experimental approaches
Advantages of experimental methods
Experiments allow for the controlled testing of game-theoretic predictions by manipulating specific variables of interest while holding other factors constant, strengthening internal validity
The use of randomization and treatment-control comparisons enables causal inference about the effects of game parameters on behavior
Experimental designs can implement novel institutional arrangements that are difficult to study using naturally-occurring data, expanding the range of empirical investigations
Inducing valuations through monetary payments aligns incentives and encourages participants to make considered decisions, enhancing the reliability of the data
Limitations and methodological challenges
Experiments face challenges in terms of external validity due to the abstract, simplified settings and the use of convenience samples like university students
The generalizability of findings to real-world contexts should be carefully considered
Field experiments or natural experiments can provide complementary evidence in more realistic settings
The strategy method may elicit different behavior compared to "hot" decision-making in direct-response designs
The interpretation of results should consider these elicitation effects
Comparing outcomes under different elicitation methods can shed light on the robustness of the findings
Experimenter demand effects can arise if participants infer the research hypotheses and modify their behavior
Careful experimental instructions and protocols aim to mitigate these concerns
Using neutral framing, double-blind procedures, or deception can help reduce demand effects
The exclusive focus on behavioral outcomes limits insights into the underlying decision processes
Combining experiments with other methods like surveys, eye-tracking, or neuroeconomics can provide a more complete understanding
Eliciting beliefs, emotions, or strategic reasoning can shed light on the mechanisms driving behavior