Historical Schemas, Diagnosticity, and Multi-Attribute Political Evaluation: Evidence from 49 Waves of Survey in Taiwan (2012-2024) (under review, with Sanho Chung)
East Asia’s late-democratizing societies experienced rapid economic expansion before political liberalization linked to western intervention, creating durable cognitive linkages between former authoritarian ruling parties and the era of high growth. Meanwhile, newly democratic parties became associated with human rights, national identity, and foreign policy. This article argues that these historical legacies produce a schema-driven multi-attribute evaluation process that departs from the classic assumption in economic voting that economic performance equally shapes approval of all incumbents. The argument is examined using 49 waves of nationally representative Taiwanese surveys (TEDS) from 2012-2024 (N = 58,344) merged with quarterly GDP growth rates and stock market indices. The results reveal systematic asymmetries aligned with the legacy of Taiwan’s democratization: presidential approval correlates positively with macroeconomic performance only during administrations of the Kuomintang (KMT), Taiwan’s authoritarian successor party, whereas such correlations are absent under governments of the Democratic Progressive Party (DPP), which emerged from democratic and identity-based movements. Individual-level analyses likewise show that economic evaluations carry greater diagnostic weight during KMT administrations, while foreign-policy and cross-Strait evaluations carry greater weight during DPP administrations. These differences persist after accounting for partisanship and sociodemographics in the regression models, indicating that citizens rely on party-specific evaluative weights shaped by historically activated schemas. The findings show that the historical legacy generates systematic variation in how voters assess different governing parties after democratization. Since authoritarian successor parties remain electorally competitive across contemporary democracies, this schema-based multi-attribute framework provides a general psychological mechanism for understanding voter accountability across post-authoritarian contexts.
Open Ecosystem, Fleeting Traffic? Durable Usage versus Transient Attention in US and Chinese LLMs (under review)
In the global contest between the United States and China over artificial intelligence (AI), winning the users around the world is treated as a central battleground. OpenRouter, an API routing marketplace that reports weekly token volumes, has become a closely watched gauge that governments and firms use to read this competition. This article argues, however, that a brief traffic win on OpenRouter is not the same as keeping those users. Instead, we propose that the two countries’ competing AI policies leave a visible mark on how people actually use these models globally. China has pursued an open-source approach, releasing a large number of models and derivatives, while the United States has concentrated on a smaller set of frontier models improved steadily over time. We argue that this strategic difference shapes user behavior, and that the difference could be found through OpenRouter. Analysing fifty weeks of OpenRouter’s public usage data (May 2025 to May 2026, 242 trillion tokens), we find exactly this pattern. After the start of 2026, Chinese models more often draw a brief surge of token usage and attention that drains away within weeks, without holding users over the longer run. US models, by contrast, more often hold OpenRouter users who keep choosing and routing to them over months. The result suggests that the two countries’ differing AI-ecosystem policies are reflected in the everyday choices of developers and users, and it invites reflection on how AI usage leaderboards like OpenRouter come to shape AI policy itself.
Too Many Social Media Sites, Political Segregation, and Political Efficacy (under review, with Shaka Y. J. Li)
How does social media usage influence political participation? This article introduces a three-stage framework linking the expansion of social media usage to political participation, reconciling previously mixed findings in the literature. In the Network Expansion stage, the initial adoption of social media, which is driven by the diffusion of the Internet and smartphones, encouraged online mobilization and political discussion. In the Platform Convergence, as users across-generations concentrated on a few major platforms, social media became a site of frequent exposure to opposing views and confrontation, lowering political efficacy. In the Political Segregation stage, the continued expansion and oversupply of social media platforms led users to strategically allocate their time toward platforms with clear political leanings, driven by the clustering of like-minded partisans. We argue that political segregation restores a positive relationship between social media usage and political efficacy. We test this three-stage framework using two large-scale surveys in the United States (ANES 2024, n = 4,931) and Taiwan (PollcracyLab 2025, n = 1,300). The results show that political segregation is evident on newer platforms in both Taiwan (TikTok, Xiaohongshu, and Threads) and the United States (Instagram, Reddit, YouTube, and TikTok), but not on established platforms such as Facebook and Twitter (X). Moreover, partisans who use the platforms dominated by their preferred political camp report higher levels of political efficacy. These findings carry implications for survey methodology, polarization, mobilization, and fact-checking.
When Information Backfires: Nonpartisans and Attitudes on Military Intervention (R&R, With Fang-Yu Chen, Charles Wu, and Yao-Yuan Yeh)
How do nonpartisans form foreign policy opinions regarding the use of force overseas? This question has become important given the increase in the number of nonpartisans in recent years. The literature suggests two main hypotheses: (1) dispositional cynicism: nonpartisans, compared to partisans, will register the lowest level of support for any interventions, and (2) informational persuadability: their stance may be moderate, but would be malleable after receiving new information. To test the two competing hypotheses, we fielded two survey experiments on Amazon’s MTurk in 2022 and 2023, using treatments related to Taiwan’s democratic system, public policies, and culinary culture. Results show that both hypotheses only explain part of the story: nonpartisans were moderate on Cross-Strait intervention, but their support reduced after receiving any additional information about Taiwan. To enhance the generalizability of our findings, we conducted additional analyses using the 2020 ANES data and observed a similar negative correlation between political knowledge and support for foreign intervention among nonpartisans. For policymakers, this implies that mobilizing the persuadable middle of the American electorate for overseas military commitments may require different strategies than those used to rally partisan bases.
Economic Patronage and Issue Linkage (under review)
Economic patronage is when a powerful donor country renders additional economic benefits beyond regular trades to attract another weak recipient country to alter its political position. Why do some donors offer patronages (China to Taiwan after 2008) while others do not (US to Cuba after 1996)? Why do some patronages succeed (EU to Rwanda after 2012) while others fail (China to Taiwan in 2014)? This article provides a new incomplete information model formalizing economic patronages and issue linkages. The powerful donor may not offer patronages if it perceives its weakness or if the recipient is resilient. The weak recipient may accept patronages but does not alter its position if it perceives that the donor may not take revenge. The powerful donor’s “paper tiger” mechanism and the weak recipient’s “opportunism” mechanism are further supported by the empirical analysis of ANES2008 and a representative survey in Taiwan in 2012.
