Research
One question runs through my work: why does a party that rests on the same communities hold together in one place and come apart in another? I study population structure, the spatially arranged communities that political parties rest on, and how it shapes party systems. Communal groups are largely made rather than inherited; they take shape through geography, colonial institutions, and state-led settlement, and they shift again as development and migration rearrange where people live. As that structure changes, the coalitions parties depend on tend to change with it, on two fronts at once: voter blocs realign from below, and elites defect from within.
My work centers on Malaysia, where I trace how the settlement geography of Malay and Chinese communities built, and later unsettled, a long-dominant coalition, and I extend the argument to Taiwan, Japan, and Indonesia. It is now reaching the United States: in Florida, I trace how Latino communities have moved and re-sorted to understand the conservative turn among Latino voters; in New York City, I am beginning fieldwork on an overseas-Chinese community that also figures in Malaysian politics.
Because these questions outrun existing tools, I also build new ones: ecological inference that recovers group behavior from aggregate returns, spatial models for electorates whose boundaries move, and corrections for measurement error in demographic panels. The flagship is BACH, a simulation-based ecological-inference estimator that recovers subgroup voting behavior where survey nonresponse, partisan polarization, and cross-cutting cleavages defeat conventional tools.
Peer-Reviewed Publications
“Myth and Measurement: Spatial Heterogeneity, Ecological Fallacy, and the Urban Narrative in Peninsular Malaysia’s Electoral Politics.” 2026. Journal of East Asian Studies 26(1): 87–112. [Open Access] [Replication]
Does Malaysia vote along urban–rural lines or along ethnic ones? Combining aggregate election returns from Peninsular Malaysia with survey evidence and modeling spatial heterogeneity to guard against ecological fallacy, the paper shows that the celebrated urban–rural divide is largely an artifact of aggregation: ethnicity, not urbanization, remains the dominant cleavage structuring party competition. The result demonstrates that aggregate election data, analyzed with attention to spatial heterogeneity, can substitute for surveys in settings where reliable polling is scarce.
“Social Transformation in Malaysia and the Fragmentation of the Malay-Dominant Political Parties: Based on a Social Cleavage Structure Theory Perspective” (with Congcong Fu). Southeast Asian Studies (东南亚研究) 2021(3): 1–29. In Chinese. [Chinese Full Text] [English Translation]
After Malaysia’s 2018 general election, five Malay-dominant parties held federal seats from Peninsular Malaysia alone, all tracing back to successive splits from UMNO. Drawing on social cleavage structure theory, the article argues that the fragmentation of Malay-dominant parties is the joint product of elite contradictions within the political society and shifting cleavage structures within the civil society. When intra-elite conflicts cannot be reconciled, or when a new social cleavage emerges, a new party is born; its survival depends on whether it can anchor itself to an existing or newly formed cleavage.
“Social Transformation, Dispute over Routes and the Split of Malaysian Islamic Party (PAS)” (with Congcong Fu). Southeast Asian Affairs (南洋问题研究) 2020(1): 79–95. In Chinese. [Chinese Full Text] [English Translation]
The expansion of the Malay ethnic middle class and the imbalance of regional development weakened the Ulama faction and differentiated the professional group within the Malaysian Islamic Party (PAS), triggering a dispute over how to realize an Islamic state. The internal struggle culminated in the 2015 split that produced the National Trust Party (AMANAH). The article traces this process through the changing factional composition of PAS and argues that AMANAH’s long-term survival requires differentiated positioning relative to other Malay-based parties and a unified ideological foundation.
Under Review
“The Front Page and the Back Room: Demand Narrowing, Elite Coordination, and the Survival of Authoritarian Elites.”
When an authoritarian leader faces a crisis they may not survive, how does a divided elite coordinate on abandoning them before the moment passes? The paper argues that two features of the crisis resolve the elite’s coordination problem: a public signal that narrows the opposition’s demand onto the leader by name, making their removal a survivable exit, and a communication environment open enough for divided elites to read that narrowing as common knowledge. Where both hold, the elite sheds the leader and keeps its places; where the demand stays systemic, or never settles on a target, the elite closes ranks and the leader endures. A global-game model carries the mechanism, tested at two scales: a day-by-day reconstruction of the week of Soeharto’s fall in May 1998, coded daily across twenty-seven elites in two dailies, and a seven-crisis comparison that separates leader-removal from elite survival on the same two features. The leader falls; the elite survives.
“Colonial Institutionalization and Unborn Civil Society in Post-Colonial States: Malaysia and Indonesia in Comparative Perspective.”
Why do some communal institutions survive decades of hostile policy while comparable ones next door are extinguished? Through a comparative-historical analysis of Malaysia and Indonesia, the paper traces the divergence to colonial-era decisions: whether the colonial state chose to register, codify, or spatially contain civil-sphere domains. These decisions set the institutional terrain on which postcolonial communal politics still runs, even as independent governments pursued divergent strategies toward communal organization.
“Symbolic Power and the Functional Divide: Campaigns Across the Party-State Spectrum.”
Across the party-state spectrum, recurrent mobilizational campaigns are the instrument by which a party keeps the state it built subordinate to political ends. Drawing on seven decades of the Chinese party press, firm-level event studies that estimate campaigns’ economic impact, and daily social-media discourse, the paper documents campaigns’ functional logic: they discipline the state apparatus and renew the party’s symbolic authority over it.
Working Papers
“Written on the Soil: Why Urbanization Broke Only Half of a Dominant-Party Regime.” [Latest Draft]
Whether urbanization threatens an electoral authoritarian regime depends on institutions the regime itself built: the state-made geography of ethnic groups, and the form of each group’s distributive incorporation. A group incorporated through brokered exchange is tied to place, because benefits flow through settlements and their brokers, and cities dissolve both; a group held by an entitlement attached by law to membership loses nothing when its members move. Using ecological inference, a settlement-level event study, and survey evidence across 1974–2018 in Malaysia, the paper shows that Chinese support collapsed where cities grew, while the constitutionally entitled Malay base held until an intra-Malay elite rupture. Exclusion did not doom the regime; it decided which coalition could bring it down.
“Crisis from Within: Elite Defection, Intraparty Competition, and the Breakdown of Dominant Party Systems.” [Latest Draft]
When and how do dominant party systems break? The cracks they do not recover from come from within, after years of electoral dominance, led by the party’s most successful politicians rather than its marginal ones. Across 4,638 candidate-elections in Malaysia’s UMNO (1959–2018), Japan’s LDP (1958–1993), and Taiwan’s KMT (1986–2001), a monopoly over office forces losing elites to choose between marginalization and exit, producing bloc exits, when cohesive factions leave together after losing leadership struggles, and marginal exits, when weakened candidates depart alone. Whether a split actually ends the party’s dominance is downstream: breakaway parties endure only when they attach to a social cleavage the parent cannot absorb. Breakdown is an elite phenomenon in its onset and a cleavage phenomenon in its consequences.
“What Race and Ethnicity Hide: Bayesian Approximate Computation with Hierarchical Framework (BACH) for Subgroup Voting Behavior under Partisan Polarization, Cross-Cutting Cleavages, and Survey Nonresponse.” [SSRN] [Poster at PolMeth 2026]
A simulation-based ecological-inference framework that recovers subgroup voting behavior from aggregate election returns under the three conditions that defeat existing tools simultaneously: partisan polarization, cross-cutting cleavages, and survey nonresponse. The framework is validated through large-scale simulation experiments and benchmarked against statewide voter-file ground truth that reveals individual-level vote choice. Substantively, national origin rather than education organizes Hispanic vote choice, a within-race cleavage that conventional race-level ecological estimators collapse.
“Research Note: A Higher-Order Method of Moments Approach to Ecological Inference under Extreme Survey Nonresponse.” [SSRN]
When surveys are unreliable or absent, as in historical elections and many democracies without polling infrastructure, how can we recover who votes for whom? This research note derives a method-of-moments estimator that exploits higher-order moments of the cross-unit distribution of aggregate election outcomes to identify group voting behavior without individual-level data, offering a closed-form alternative to simulation-based approaches.
“What the Slope Loses, the Intercepts Gain: Measurement Error and Misattributed Effects in Fixed- and Random-Effects Models” (with Robert Franzese). [Latest Draft]
Fixed- and random-effects models compare each unit to itself over time. When a treatment moves slowly but its measure carries fast-moving error, the within transformation strips out the signal and keeps the noise, and the effect it loses does not simply attenuate; it migrates into the unit intercepts, masquerading as stable “unit characteristics.” Larger samples make this worse, tightening confidence intervals around the wrong value and driving the Hausman test toward the most biased specification. The paper proves the result, establishes a scope condition (unit effects worsen attenuation only when the signal is more persistent than the error) and supplies a diagnose–correct–redesign protocol validated on three datasets.
“Plant Closures and Trump-Related Twitter Discussion: A Spatial-Temporal Description” (with Robert Franzese). [Latest Draft]
How do local economic shocks shape online political extremism? Using a national geocoded tweet corpus matched to plant-closure events across the United States, the paper traces how local plant-closing shocks shaped Trump-related and extremist engagement on social media, mapping the spatial and temporal diffusion of economic-grievance discourse.