TL;DR
This paper uses laboratory experiments to compare compulsory and voluntary voting systems in a simplified "Condorcet Jury" model. They find that voters adapt their behavior to the voting rules, with strategic voting occurring under compulsory voting and sincere voting dominating under voluntary systems. This adaptation helps equalize group decision accuracy across the systems, potentially explaining why both exist in the real world.
What Problem Does This Paper Address?
The paper tackles a fundamental question in political science and economics: how do different voting rules affect voter behavior and the quality of collective decisions? Specifically, it investigates the theoretical predictions that compulsory voting (where abstention is not allowed) should lead to strategic voting (voting against one's private information), while voluntary voting (where abstention is allowed) should encourage sincere voting (voting according to one's private information). While rational choice theory predicts these differences, empirical evidence, especially from controlled settings, is needed to confirm them and understand their implications for group decision-making. The authors aim to bridge this gap by experimentally comparing these two institutional settings.
Methods
The authors employ a laboratory experiment designed to mimic a simplified Condorcet Jury Theorem environment. Participants, acting as members of a jury or committee, receive noisy, private signals about the true state of the world (e.g., "guilty" or "innocent"). They then vote on an outcome that maximizes their payoff if it aligns with the true state. The experiment compares three voting mechanisms:
- Compulsory Voting (C): Participants must vote, and there are no costs associated with voting.
- Voluntary and Costless Voting (VN): Participants can choose to vote or abstain, and there are no costs for voting.
- Voluntary and Costly Voting (VC): Participants can choose to vote or abstain, and there is a small, randomly assigned cost for voting.
The experiment uses a between-subjects design, meaning each participant experienced only one of the three treatments. They controlled for group size (9 participants), the number of rounds (20), payoff functions, and signal precisions to isolate the effect of the voting mechanism.
Datasets
Not discussed in the paper. The study uses data generated from laboratory experiments.
Key Insights and Ideas
- Voters Adapt to Rules: The most striking finding is that individuals don't just vote randomly; they strategically adjust their voting behavior based on whether voting is compulsory or voluntary. This supports the core predictions of rational choice theory in this context.
- Strategic Voting Under Compulsion: When forced to vote, participants with less informative signals sometimes vote against their signal to try and influence the outcome, a behavior not observed when voting is voluntary.
- Sincere Voting Dominates Voluntarism: When participants have the option to abstain, they overwhelmingly vote sincerely, aligning their vote with their private information. This suggests that the ability to abstain removes the incentive for strategic voting.
- Participation as a Strategic Tool: In voluntary systems, participation becomes the strategic lever. Voters with more informative signals tend to participate more, and the introduction of voting costs (VC) significantly reduces participation compared to costless voluntary voting (VN).
- Accuracy Convergence: Despite differences in individual voting strategies, the overall accuracy of group decisions across the three mechanisms is surprisingly similar. This suggests that the adaptive behavior of voters helps to equalize the effectiveness of these different voting systems.
Key Results
- Sincerity: Under compulsory voting, approximately 15.6% of "red signal" types (those with less informative signals) were predicted to vote insincerely, and the experiment found a significant deviation from 100% sincerity for this group. In contrast, under both voluntary mechanisms, nearly all voters (both signal types) voted sincerely.
- Participation: In voluntary voting, participants with "blue signals" (more informative) participated significantly more than those with "red signals." Participation rates were also higher in the costless voluntary (VN) treatment than in the costly voluntary (VC) treatment.
- Group Accuracy: While theoretical predictions suggested a ranking of accuracy (VN > C > VC), the experimental data showed that group decision accuracy was remarkably similar across all three mechanisms, with differences not being statistically significant in most comparisons. The observed accuracy was generally high (around 85-90%).
Limitations and Future Work
- Small Group Size: The study used small groups (N=9). The authors note that in the limit, as group size increases, information aggregation becomes more efficient and differences between mechanisms diminish. Future work could explore larger group sizes.
- Abstract Task: The experiment used an abstract task to avoid real-world voting motivations (like civic duty). While this controls for extraneous factors, it might limit the direct applicability to complex political elections.
- Equilibrium Selection: In the voluntary and costless (VN) treatment, two equilibria exist: a sincere voting equilibrium and an insincere voting equilibrium (similar to the compulsory case). The experiment found evidence leaning towards the sincere voting equilibrium, but further investigation into equilibrium selection is suggested.
- Bounded Rationality: While the study found strong support for rational choice predictions, some deviations like over-participation in voluntary systems were observed. The authors suggest exploring models of bounded rationality (like Quantal Response Equilibrium) to better explain these anomalies, though their initial analysis indicated these models didn't significantly improve upon Nash equilibrium predictions.
- Welfare Analysis: The paper focuses on accuracy and behavior, not welfare. Future work could analyze the welfare implications of voting costs under different mechanisms.
Summary updated