My research asks why multiracial democracy has proven so difficult to achieve and how those on the margins respond politically. One line of work examines the American political system itself, arguing that racial exclusion is not a defect in its democratic design but one of its constitutive features. A second turns to the people navigating that system, investigating how Latinos make political choices under conditions of persistent exclusion. Both inquiries have required building new computational tools, and those methodological innovations now constitute a third, independent line of contribution to the discipline.

This line of research examines how racial exclusion and democratic order are co-constituted in American politics.
 
Herrenvolk Democracy: Race, Immigration, & the American Political Order | Book project.

The dominant frameworks in American politics assume that liberalism is the baseline and racial exclusion is a deviation. This book argues the reverse. The United States was founded as a herrenvolk democracy, a regime that extends democratic rights to a dominant racial group while systematically excluding others. The book develops this argument through a four-era reinterpretation of American political development. The foundational period established a herrenvolk order through laws like the 1790 Naturalization Act, which legally defined the citizenry as white. Reconstruction created the first genuine possibility for multiracial democracy, a project that was violently overthrown to restore herrenvolk order. A long period of bounded democracy from 1877 to 1965 maintained racial exclusion through legal and de facto measures. The contemporary period, beginning with the landmark reforms of 1965, dismantled much of the legal architecture of exclusion but also triggered a powerful backlash that continues to intensify. Across all four eras, immigration and citizenship law served as the primary mechanism for defining the racial boundaries of the demos and enforcing exclusion.

"Democracy for Whom? Herrenvolk Origins and the Design of American Exclusion" | Revise and resubmit, PS: Political Science & Politics

APSA Preprint

Using an originalist historical approach, this paper reconstructs four critical junctures in the founding era: the early colonial period, the Declaration of Independence, the Constitutional Convention, and the Naturalization Act of 1790, which restricted citizenship to "free white persons" and stood for 162 years. At every juncture where the founders could have moved toward inclusion, they chose exclusion. On the nation's 250th anniversary, with the project of multiracial democracy still unrealized, understanding this founding design is critical to confronting its legacy.

"Power-Sharing & Exclusion: Toward a Theory of Herrenvolk Democracy" | V-Dem Working Paper (in preparation) | Target: Perspectives on Politics

V-Dem’s Electoral Democracy Index scores the United States at 0.48 during slavery and Jim Crow, when the majority of the population was excluded from political power by law. The score is not wrong. It accurately measures how well democratic institutions functioned for the people they were designed to serve. This paper introduces the Power-Sharing Index (PSI), a measure of cross-group power distribution that captures what existing indices were not built to ask: whether political power is accessible across group lines. Applied to the full arc of American political development, the PSI reveals that institutional quality and group-based access to power moved independently for most of the nation’s history, converging only after the major democratic reforms of 1965, including the Voting Rights Act and the Immigration and Nationality Act, and diverging again in the contemporary period.

This line of research focuses on Latino political behavior, partisan incorporation, and heterogeneity within the Latino electorate.

"Latino Support for Trump in 2024: Trends and Insights from an Empirical Analysis" | Published, Political Science Now, APSA's 2024 Post-Election Reflection Series. (with Luis Fraga)

"The Myth of the Zero-Sum: Rethinking Acculturation in American Politics" | Revise and resubmit, Politics, Groups, and Identities

Political science has assumed for decades that immigrants face a zero-sum choice between adopting American culture and preserving their heritage. This assumption has never been empirically tested. Using the 2006 Latino National Survey and comparative cluster analysis across three algorithms with extensive validation, I tested it. The binary model fails: it misclassifies over 75% of the Latino electorate. Four distinct acculturation orientations emerge, with the vast majority occupying hybrid space that the dominant framework structurally cannot detect. The people the models erase turn out to be the majority.

APSA Preprint ⏐ GitHub Code ⏐ Harvard Dataverse Replication Data 

"Acculturation, Identity, and Latino Politics" | Under review, Political Psychology

If the binary model is wrong about how Latinos experience acculturation, it is likely wrong about how acculturation shapes political behavior. This paper demonstrates that the four orientations recovered through the bidirectional framework produce systematically different political profiles. The strongest effects concentrate on immigration policy, the domain most directly tied to questions of belonging and democratic membership, and in several domains exceed the explanatory power of ideology and partisanship. Latino political diversity is not puzzling variation but the predictable consequence of navigating American society as members of a racialized immigrant-origin population.

APSA Preprint ⏐ GitHub Code ⏐ Harvard Dataverse Replication Data

"Stable Drivers or Campaign Effects: Mapping Latino Support for Trump, 2016–2024" | Paper in preparation. Accepted for presentation, MPSA 2026.

What explains Latino support for Donald Trump, and has the answer changed across elections? Using interpretable machine learning on data from the Collaborative Multiracial Post-Election Survey, I model the drivers of Latino vote choice across three presidential elections. The project asks whether Latino support for Trump reflects stable partisan incorporation, shifting coalition dynamics, or election-specific campaign effects. The method captures non-linear relationships and complex interactions that conventional regression cannot detect. The 2016 and 2020 analyses are complete. In 2016, opposition to Black Lives Matter emerged as the single most powerful non-partisan predictor for Latino men, while opposition to the Affordable Care Act was the leading predictor for Latina women. The project’s broader goal is to trace how the attitudinal structure of Latino Trump support evolves across the Trump era and whether that structure differs systematically by gender.

GitHub Code (work in progress)

This line of research develops computational and inferential tools for identifying stable structure in high-dimensional political data. 

"Inferential Cluster Analysis: From Discovery to Measurement" | Under review, Political Science Research and Methods

Cluster analysis is the most widely used unsupervised method in political science, yet the discipline treats it as merely exploratory. This paper develops a systematic framework for using cluster analysis to test theoretical expectations, moving beyond description to inference. The core insight is that when algorithms with different mathematical foundations and independent failure modes converge on the same solution, that convergence constitutes evidence that the recovered structure reflects the data-generating process rather than methodological artifacts. The framework introduces a multi-method validation process combining theoretical pre-specification, multi-algorithmic convergence, structured model selection, and bootstrap stability assessment.

APSA Preprint ⏐ GitHub Code ⏐Harvard Dataverse Replication Data

"Learning From Survey Data: A Nonparametric Framework for Explanatory Outcome Modeling" |  In preparation, target: Political Analysis. Accepted for presentation, AAPOR 2026.

Survey researchers routinely rely on regression models that impose linearity, additivity, and correct specification on relationships they have not tested. This paper proposes a portable framework combining Random Forest estimation with SHAP interpretation and a bootstrap rank stability criterion. The framework makes no assumptions about the functional form of the relationship between a feature space and an outcome, learns that structure from the data, and produces findings that are both interpretable and validated for robustness. A step-by-step reproducible pipeline, demonstrated on Latino vote choice using the 2016 CMPS, provides a template that other researchers can adapt to their own survey outcomes of interest.

GitHub Code (work in progress) 

Power-Sharing Index (PSI): Cross-group power distribution measure built from V-Dem indicators.
Bidimensional Acculturation Model (BAM): Multi-algorithm cluster analysis and scale construction.
ML Latino Vote: Random forest + SHAP analysis of Latino vote choice (2016, 2020, 2024).
Acculturation & Politics: Cross-pressures analysis linking acculturation orientations to political behavior.