Research

Abstract: Historically, authoritarian governments have commonly adopted a patriotic education system during political repression, aimed at legitimizing their authority and mitigating citizens' dissent against political violence. We investigate the long-term impact of the patriotic educational curriculum during the nationwide turmoil of the Cultural Revolution in China. Our analysis leverages two concurrent natural experiments that occurred in the late Cultural Revolution: (i) the sudden initiation of state-led violence against civilians and (ii) the educational reform promoting patriotism in primary education. Employing a cohort difference-in-differences design, we find that early-life exposure to such patriotic indoctrination amidst state-led violence decreases the probability of having mental illnesses in adulthood. The underlying mechanism suggests that patriotic education nurtures political loyalty that legitimizes the authoritarian government and it reduces psychological burdens caused by state-led violence. However, the post-Cultural Revolution regime reversed the legitimacy of the Mao Zedong's Cultural Revolution regime and implemented a new educational curriculum. Our evidence shows that those who were exposed to conflicting education programs are more likely to suffer from mental health problems. This highlights that unsteady patriotic indoctrination due to unstable political regimes might deteriorate the young generation's long-term mental health.


Abstract: Extremism and ideological polarization in online communities are global issues. While previous research attributes these phenomena to online search algorithms, known as filter bubbles, and closed communication structures, termed echo chambers, this paper introduces an alternative source: a mutual rating system through a "like-button" feature, which fuels reputation-building among users. We focus on a popular Japanese Q&A platform that experienced two sequential natural experiments: (i) the introduction of a "like-button" system and (ii) monetary rewards tied to the number of "likes" received. By leveraging complete text data, we measure the degree of extremism in xenophobic sentiment using a recently developed machine learning approach called latent semantic scaling. A simple before-after comparison, controlling for question thread fixed effects, reveals that concerns over reputation intensify extremism and polarization. We further apply a difference-in-differences approach to demonstrate that non-anonymous users who are sensitive to online reputation are more likely to radicalize their statements relative to anonymous users after the interventions. Further evidence shows that the platform has no incentive to mitigate reputational concerns as it boosts user engagement, suggesting the persistence of polarization in online communities.



(joint with Kazushi Takahashi, Abu S Shonchoy, Takeshi Aida, Yu Ri Kim, Hisaki Kono, and Takashi Kurosaki)

Selected for the ADB-IEA Innovative Policy Award 2023, Honorable Mention

Abstract: Can commitment-saving (CS) ahead of a lean season alter consumption downfalls among the ultra-poor? We collected 36 rounds of bi-weekly household panel data over two-years in Bangladesh and conducted a savings experiment in the second year by randomly allocating commitment-saving (CS) products with either temporary savings subsidy with “premium”, or prevailing “market” interest rate. Premium group doubles the formal savings, resulted in increased food and non-food expenditure by 8.6-12.6% during the lean season, with no lasting post-lean season impact. Market group shows no discernable impacts. Our results suggest that, while imperfect, a better-designed savings product could potentially mitigate seasonal deprivation.


(joint with Yasuyuki Sawada and Takeshi Aida)

Abstract: Social preferences shape behavioral principles in various aspects of society. Therefore, it is important to understand the determinants of social preferences to ensure the overall efficacy of any organizations. However, little is known about how repeated daily interactions foster social preferences. This study investigates whether social interactions in the workplace affect the conformity of social preferences among workers, and whether social coherence underlies the development of fairness norms in organizations. A two-week field experiment was conducted with agricultural manual laborers in the Philippines, where team compositions and payment contracts were randomized. Additionally, lab-in-the-field experiments were conducted to measure social preferences. The results show that workers who belonged to the same group more frequently during the experiment have similar social preferences compared to those who did not belong to the same group. Moreover, workers who worked together are more likely to comply with the fifty-fifty norm in the dictator game than those who did not. Finally, we find that group-based financial incentives improve social preference conformity and the emergence of the fairness norm. These findings suggest that social interactions and incentive schemes play an important role in influencing an organization's performance by shaping their social preferences and norms.


Abstract: Historically, the debate on judicial independence has been centered on whether the participation of external branches in the appointment process of judges hampers the impartial administration of justice. However, less attention has been given to how the internal judicial organization shapes the judge's decision. This paper, focusing on the Indian lower judiciary, explores whether the prospects of promotion to higher courts result in currying favor with judicial leaders who possess discretionary power over personal matters within the judicial organization. For this purpose, I exploit natural experiments: female judicial leader's appointments replacing male incumbents, causing the unpredictable shift of gender preferences in personnel evaluation of local judges. Then, I use the triple difference strategy exploiting the high-court level variation on whether it has at least one female judicial leader ever, the individual judge level variation on the eligibility as the high court judge, and the temporal variation pre- and post-female judicial leader's appointment. Estimation results show that eligible judges pander to female leaders and hand down lenient sentences for female defendants. Evidence suggests that discretionary appointments of judges through the non-transparent "judges-selecting-judges" system can be detrimental to the independent administration of justice. Therefore, careful institutional design is necessary to balance judicial accountability and independence.


Abstract: Gender quotas are intended to address the underrepresentation of women in government. However, their effectiveness is still debated. We explore the random assignment of gender quotas across villages in India to investigate whether women members of local councils affect public good allocation in their constituencies. We conducted a field survey in 100 villages with populations of approximately 400 each, constituency size for which only one council member is elected for representation. We find that the performance of women council members is generally the same as men members in public good allocation, except for a lower performance by women in delivering water facilities. We suggest that the influence of non-member actors such as husbands, lower levels of educational achievement, and lower electoral incentive as obstacles that women members face. We also find that the gender and castes of the council president affects the performance of women council members.


Abstract: While recent empirical evidence reveals some effective interventions in preventing corruption among bureaucrats and politicians, there has been little discussion on how to prevent the bribe-giving behavior of ordinary citizens. This paper investigates the role of social media information in influencing the supply of bribes by citizens instead of the demand side. We therefore developed and published an original news application in India and implemented a three-month experiment. In this application, we randomly circulate live news related to corruption to users and incorporate a lab experiment into the app system to elicit users' bribery behavior every week. We find that corruption news involving politicians within a close geographical proximity lowers users' moral costs against the anti-social bribery act, leading to an increase in the amount of bribes. However, news of accused citizens and officials within the geographic proximity increases their moral cost against bribes and decreases the amount. This suggests that individually-tailored local information on corruption may be an effective tool to reduce citizens' supply of bribes.


Abstract: This paper investigates the impact of receiving extreme information through social media on the formation of normative perceptions of corruption. For this purpose, we developed an original mobile application in which two types of randomizations were built: First, we randomly circulated news and blogs related to corruption to users; second, we randomly deployed other users' comments on each article. We quantify the degree of extremism in individual perceptions of social norms against corruption by employing a lab-in-the-app experiment and supervised machine learning algorithms. Our estimation results show that users respond to both counter- and pro-attitudinal extreme information by shifting their norms to even more radical directions. Further analysis implies that socialized motivation to conform to other people's opinion is the main driver of polarized social norms relative to social-psychological explanations. Finally, we show that polarized social norms coincide with the segregation of online communities. These results suggest that to curb severe social fragmentation and eliminate corruption, we need to reconsider the design of information circulation and communication structures in social media.


Abstract: This paper provides field-experimental evidence on how repeated exposure to biased news through social media affects societal discrimination and political polarization in India. I developed and released an original mobile application in which social media news is delivered to users' mobile phones on the daily basis during the experimental period (202 days). In this application, treatment and control arms are randomly built-in. In the treatment arm, a machine learning algorithm customizes news circulation based on individual user’s past behavior on the web. In the control arm, news circulation is fully randomized. Three main findings are obtained by econometric analyses. First, the customized news circulation significantly makes an individual flow of daily news more religiously biased and politically extreme. Second, users' political beliefs and preferences toward religious discrimination are unlikely to be affected by repeated exposure to such slant news. However, if users can recognize that a majority of other users express positive opinions on the slant news, they get to be less against religious discrimination and more strongly favor political extremism. Finally, the cognitive bias rather than concerns for social reputation is a main driver of emergence of societal discrimination and political polarization.


Abstract: This paper examines historical origins and persistence of parochial altruism by exploiting natural experiments during the early modern and the colonial period (1801-1932) of South India. Theory suggests that parochial altruism composed of in-group favoritism and out-group hostility can be nurtured by repeated exposure to intergroup conflicts among socially-defined groups. Natural experiments in South India provides the suitable setting to test this theory: some communities in this region had historically adopted a randomized formation of fishing groups for each available fishing lot every year and intergroup conflicts had frequently occurred among these randomized fishing groups. All fishing groups are obligated to report any type of conflicts to religious judicial systems. I digitized detailed conflict information from 1801 to 1932 and combined it with the data from lab-in-the-field experiments for descendants. I find that an individual is more altruistic to other community members if their ancestors had more experiences of sharing a fishing group membership and more frequently sued other groups together (i.e., they shared common enemies). Furthermore, I demonstrate mechanisms through which such parochial altruism has persisted over generations.


Abstract: Rice planting contracts in the Philippines are predominantly fixed-wage (FW) although individual work–efforts are largely observable. We compare different theories which can account for the use of this seemingly inefficient contractual arrangement. We “perturbate” the existing contract and observe its consequences by adopting a hybrid experimental method of framed and lab-in-the-field experiments. While we found positive incentive effects in individual piece rate (IPR) and moral hazard problems in FW, non-monetary incentives play a significant role under FW.


Abstract: This study aims to provide empirical evidence on the economic and social impacts of access to microloans disbursed through self-help group (SHG) programs. For this purpose, primary data were collected from households in Kerala, South India, and combined with detailed financial transactions of SHG members and information collected through laboratory experiments. The estimation results show that wealthier group members are significantly more likely to reap economic benefits, probably from productive investments. For poor member, asset accumulation and consumption smoothing are the two main pathways out of poverty through SHG-modeled microfinance initiatives. Finally, we find that reciprocal cooperation and trust among group members are developed by repeated social interactions, which are facilitated by weekly meetings of SHGs.


Work in Progress

Publications (Japanese)