1. Misinformation in Representative Democracy: The Role of Heterogeneous Confirmation Bias. (Joint with Satoshi Kasamatsu)
[First version: November 2022, Latest version: November 2024]Previous title: Strategic Misinformation: The Role of Heterogeneous Confirmation Bias
Abstract. This study investigates politicians' incentive to promote false claims supported by only a minority of the electorate and its impact on representative democracy. For this purpose, we construct a two-period electoral accountability model where an incumbent politician decides whether to support or deny the truth about a particular issue and exerts effort to produce public goods. We find that a low-competent incumbent has an electoral incentive to support misinformation, even if only a minority of voters believe in it. The key to this incentive is the heterogeneity of confirmation bias that leads people to persistently accept misinformation. When the degree of confirmation bias is heterogeneous among voters, promoting misinformation enables the incumbent to weaken the role of elections as a device for punishing the incumbent's low performance. Hence, to avoid electoral punishment, low-competent politicians may strategically deny the truth, which also reduces their effort level as a side effect. Moreover, we argue that social media platforms amplify the heterogeneity of confirmation bias, thereby inducing politicians' misinformation.
2. Public Perceptions of False Positive and Negative Errors in News Reports. (Joint with Greg Chih-Hsin Sheen)
[First version: May 2024, Latest version: November 2024]Abstract. Due to the problem of misinformation, the public is advised to obtain information from reputable sources to avoid inaccurate news coverage (false positive errors). However, reputable sources might engage in selective reporting (false negative errors). We developed a concise model to explain why both inaccurate coverage and limited coverage should be considered if electors are to make an appropriate political decision. Then, we conducted a survey in Japan (𝑁 = 2020) to examine the hypotheses derived from the model. The results suggest that respondents perceive different types of media to vary in their likelihood of committing the two errors; these perceptions correlate with media consumption, and media literacy interventions do not change attitudes toward media consumption.
3. Trust in Politicians and Preferences over False Positive and Negative Errors in News Reports. (Joint with Greg Chih-Hsin Sheen)
[First version: August 2025, Latest version: January 2026]Abstract. This study investigates how trust in politicians shapes citizens' news consumption preferences. We focus on two dimensions of media accuracy-false-positive and false-negative errors in the reporting on political scandals-and hypothesize that political trust alters citizens' preferences for avoiding these two errors in a different way. To test this, we conducted a survey experiment combining information provision treatment with a conjoint design in Japan. We found that citizens value both types of accuracy, and that increased political trust reduces the preference for avoiding false-negative errors, but has no significant effect on the preference for avoiding false-positive errors. The treatment reduces even the preference for avoiding false-positive errors among the supporters of governing parties. Our findings suggest that political trust affects the type of media citizens prefer, which has important implications for electoral accountability.
4. Population Dynamics of Conspiracy Thinking. (Joint with R. Emre Aytimur)
[First version: November 2025]Abstract. Conspiracy theories undermine the credibility of information sources and decision-making; and their prevalence varies dramatically across countries and political groups. We develop a model in which conspiracy thinking and information quality co-evolve through two-way feedback. When more citizens hold conspiratorial beliefs, experts face greater reputational pressure to pander to preferred beliefs, degrading information quality. Poor information quality, in turn, sustains conspiracy thinking across generations through evolutionary dynamics. This feedback generates two main findings. First, improving expert knowledge can paradoxically increase conspiracy thinking, creating a backfire effect. Second, multiple steady states coexist and generate historical path dependence, which may explain persistent crosscountry differences. Finally, fact-checking mitigates the spread of conspiracy thinking.
5. Affluence and Influence Under Tax Competition: Income Bias in Political Attention. (Joint with Satoshi Kasamatsu and Taiki Susa) (under review)
[First version: September 2023, Latest version: September 2025]Abstract. This study reveals that tax competition magnifies the political overrepresentation of the rich in democracy, and thus prevents redistribution both economically and politically. We develop a capital tax competition model between countries, each comprising two distinct classes: rich and poor. An income bias in political attention creates an overrepresentation of the rich in each country. First, we show that tax competition diminishes the political attention of the poor, amplifying the rich’s political influence. Therefore, tax competition reduces capital taxation not only through conventional economic channels but also by altering political power in favor of the rich. From a global perspective, the poor's political attention is underprovided for their benefit. Second, rising inequality encourages the poor to pay attention to politics, thereby increasing capital taxation. However, we show that tax competition weakens this mechanism, and increasing inequality in tax competition is more likely to lead to reduced capital taxation than in a closed economy.
6. Political Trust and Preferences for Redistribution: Wasteful Spending and Plutocratic Influence. (Joint with Tomoko Matsumoto) (under review)
[First version: October 2024, Latest version: March 2026]Abstract. Higher political trust has been argued to stimulate support for redistribution. We revisit this argument by distinguishing between two types of trust: trust in efficient government spending and trust in limited plutocratic influence. By developing a political economy model of income taxation and public spending, we found that, while higher trust in efficient spending increases the ideal tax rate, higher trust in limited plutocratic influence may decrease it. To test this theory, we conducted an online survey experiment in Japan with approximately 2,000 respondents, where we measured and manipulated these two types of trust separately. The results showed that increasing trust in efficient spending had no effect on preferences for income taxation. However, increasing trust in limited plutocratic influence reduced support for progressive income taxation among supporters of governing parties. The additional experiment using a conjoint design to measure policy preferences also confirmed the result on the effect of trust in limited plutocratic influence.
7. Partisan Misperception in Knowledge Gap and Biased Information Processing. (Joint with Shinnosuke Kikuchi, Yesola Kweon, and Yuko Kasuya) (under review)
[First version: October 2025, Latest version: February 2026]Abstract. This paper introduces and tests a new concept - partisan disbelief in knowledge - the tendency for partisans to believe that their in-group is more knowledgeable than the opposing party, even about basic non-partisan facts. Using large-scale surveys and experiments in South Korea and the United States, we document that partisans perceive their in-group's accuracy rate in judging non-partisan facts to exceed that of the out-group by about 15 percentage points. This partisan disblief distorts information processing: When identical information is attributed to outgroup sources, individuals are less likely to update their opinions, revealing an in-group bias that extends beyond explicitly partisan issues. Providing corrective evidence that both sides are equally knowledgeable reduces partisan disbelief and weakens this in-group bias. It also temporarily lowers affective polarization, though the effect fades over time. Together, these results show that polarization extends beyond politics to perceptions of competence, identifying a novel cognitive mechanism through which social identity undermines accurate information processing and mutual understanding in societies.
8. When State Concessions Backfire: Information Cascades, Protest Mobilization, and Authoritarian Survival. (Joint with Yuki Takagi and Hans H. Tung)
[First version: October 2025, Latest version: March 2026]Abstract. Do a dictator's concessions demobilize protests or instead fuel mobilization that threatens regime survival? To address this question, we develop a two-period model of protest cascades in which vanguards first decide whether to initiate protests, followed by ordinary citizens who choose whether to join. Concessions directly reduce incentives to protest, but they also make a vanguard's protest a stronger signal of regime weakness. As a result, concessions lower the likelihood of vanguard protests while paradoxically increasing ordinary citizens' propensity to follow once protests occur. We show that concessions raise the probability of regime change when (i) vanguards enjoy a sufficiently large advantage over ordinary citizens in participation costs and information, and (ii) when the regime is relatively resilient, that is, it has low vulnerability to protest and a mild commitment problem. Beyond theory, we also derive testable predictions and discuss empirical implications for research on contentious politics and comparative authoritarianism.
9. Fighting Fake News with Peer Feedback: Theory and Experiment. (Joint with Yasushi Asako, Yoshio Kamijo, and Masayuki Odora) (under review)
[First version: February 2026]Abstract. The diffusion of fake news on social media poses a growing challenge to society. This study develops a theoretical model and laboratory experiment to examine whether userto-user feedback such as likes or negative comments can reduce fake news sharing. We construct a sender-receiver game in which the sender receives a signal from potentially unreliable sources and decides whether to share it, while feedback from the receiver enables learning about information quality over time. The model predicts that (i) peer feedback induces self-selection: individuals with unreliable sources learn to stop sharing, and (ii) fake news spreads more when senders are motivated by reputation rather than accuracy, though this effect is modest. We test these predictions by developing a novel lab experiment based on the ball-and-urn experimental design. Consistent with the theory, feedback reduces fake news sharing, but effects are weaker due to underreaction in belief updating and noisy decisions. Differences across motivational incentives are minimal. These findings highlight both the potential and limits of peer feedback in preventing fake news sharing, offering implications for platforms seeking to curb misinformation through user-to-user feefback.
10. Fragility of Losers' Consent: Political Institutions and Polarization. (Joint with Satoshi Kasamatsu)
[First version: February 2026]Abstract. Losers' consent, defined as the willingness of electoral losers to accept election outcomes, is central to democratic legitimacy, yet it is fragile. This paper explores how political institutions shape losers' consent and policy polarization by modeling losers' consent as a choice of belief distortion by the electoral loser. We develop a two-period model of electoral competition with partially policy-motivated parties, where the implemented policy is a convex combination of the winner's and loser's platforms. The weight placed on the winner captures the degree of institutional disproportionality. While electoral outcomes provide information about voter preferences, losers may strategically reject this information to preserve anticipatory utility. We show that institutional disproportionality has a non-monotonic effect on electoral denial. As long as losers' consent is interior, greater disproportionality increases denial. However, once institutions become sufficiently proportional, denial discontinuously jumps to full rejection. This denial, in turn, fuels policy polarization. Consequently, although disproportional institutions have a direct moderating effect on platforms, they may indirectly increase polarization by encouraging denial. These findings highlight the importance of accounting for psychological biases when evaluating the design of political institutions.
11. Search and Knightian Uncertainty Revisited: Roles of Optimism and Pessimism. (Joint with Kiyohiko G. Nishimura and Hiroyuki Ozaki) (under review)
[First version: February 2026]Abstract. Optimal stopping under ambiguity is central to dynamic decision-making, yet standard models impose uniformly ambiguity-averse preferences that conflict with experimental evidence showing heterogeneous and context-dependent ambiguity attitudes. This paper develops a job search model without recall in which the magnitude of ambiguity is separated from attitudes toward it. Using Choquet expected utility with JP-capacities, we show that the classical prediction that greater ambiguity shortens search holds only when the agent is sufficiently pessimistic; when optimism is high, ambiguity instead lengthens search. Focusing on neo-additive capacities, we further identify how the economic environment shapes this reversal. A key insight is that the discount factor plays a decisive role: even slight contamination of optimism induces ambiguity-seeking stopping behavior when the agent is sufficiently patient. This reveals that patience can fundamentally alter the effect of ambiguity in optimal stopping.
12. Damned if You Do, Damned if You Don’t: The Dilemma of Consistency and Revision in Expert Communication. (Joint with Yoshio Kamijo and Satoru Shimokawa) (under review)
[First version: March 2026]Abstract. This paper identifies a fundamental expert communication dilemma: citizens distrust consistent advice as a signal of bias, yet distrust revision as indicating limited knowledge. We formalize this in a repeated cheap-talk model with bias uncertainty and gradually accumulating evidence. Theoretically, perfect compliance obtains without private information but collapses otherwise. Experiments reveal the dilemma is more severe than predicted, emerging even without private information. Importantly, private signals do not reduce welfare but instead filter incorrect advice. Finally, compliance substantially recovers with algorithmic advisors, suggesting automated advice mitigates communication failures in controversial policy environments.
Abstract. This paper studies an electoral competition model in which parties maximize legislative power rather than vote shares. Voters are uniformly distributed on the unit interval and vote for the party proposing the closest policy platform. After the election, parties form coalitions through a communication network arising from ideological proximity: two parties are directly linked if their policy distance is at most d. A party's objective is its Myerson value in the resulting graph-restricted voting game. I characterize symmetric pure-strategy equilibria in two-, three-, and four-party systems. The two-party case yields convergence to the median. The three-party case admits a continuum of symmetric equilibria in which the two extreme parties are directly linked. In the four-party case, the unique symmetric equilibrium places two parties at (1-d)/2 and two parties at (1+d)/2. In both three-and four-party systems, more restrictive coalition communication, represented by a smaller d, generates a centripetal force, and the median voter theorem holds in the limit despite the multiparty setting.
14. Why Leaders Build Institutions They Will Ignore: Information and Coordination. (Joint with Greg Chih-Hsin Sheen)
15. Pregnancy-based Discrimination and Women's Social Policy Preferences. (Joint with John Ahlquist, Tomoko Matsumoto, and Megumi Naoi)