Publications

Politicians have long served as intermediaries for citizens navigating the state. Can elected members of village councils facilitate citizens' access to the formal justice system? Justice seekers face formidable social and administrative costs that cause most grievances to remain outside the state's purview. Council members lower these barriers through a distinct form of assistance: by acting as intermediaries. I test this theory in 28,000 village councils in Maharashtra, India, leveraging population-based discontinuities in council size, paired with an original dataset of millions of court cases. I find that additional council members increase court filings and use qualitative data to document these direct interactions. Surveys of citizens and council elections data show that this phenomenon is explained by greater capacity and descriptive representation in larger councils. These findings reveal how political institutions act as a catalyst for legal mobilization.

Does the quality of representatives shape how they respond to institutional incentives? The duration of a political term is a powerful incentive that can change representatives' priorities and shape their performance. We argue that positive selection of representatives can override the negative incentives created by shorter terms. We demonstrate this by leveraging natural experiments in the indirectly-elected *Rajya Sabha* of the Indian Parliament. In 1952, a public lottery randomly assigned term lengths to a founding cohort of high-quality representatives. We combine this setup with detailed data from their parliamentary records to show that shorter terms did not worsen parliamentary performance of these representatives. Second, we leverage quasi-random variation in term lengths following the unexpected deaths of sitting representatives. We find that a shorter term significantly worsens parliamentary performance once positive selection is much harder. Taken together, these findings indicate that the effects of perverse incentives are conditional, and depend on the type of actors they operate upon.

While criminal legal reforms are enacted to protect vulnerable populations, they can impose burdens on the judicial infrastructure due to increased caseloads. When unfunded, these reforms risk overwhelmimg existing judicial capacities. This paper evaluates the effect of one such reform-a state-wide prohibition on alcohol in Bihar, India-on judicial performance. Employing a difference-in-differences research design that leverages quasi-random variations in case assignments across courts, I analyze micro-level data from over 650,000 case filings between 2015 and 2017. I find that the reform significantly increased demands on the justice system and deteriorated case resolution rates, as judges reallocated time toward urgent, resource-intensive litigation. The findings highlight how unfunded criminal legal reforms can inadvertently degrade the quality of justice.

We argue that a major and underappreciated cause of judicial delays is the frequent transfer of judges, which reduces productivity by leaving courtrooms vacant and prompting new judges to rehear inherited cases. In India, judge transfers are frequent partly because the rules that insulate assignments from politics generate transfers. A seniority rule converts each judge's departure into a chain of cascading reassignments. Drawing on new administrative data on 6.4 million cases from India's largest state, where judges are transferred every 7 months, we estimate that each transfer causes a courtroom to issue 8-10 fewer decisions and reduces the probability of a case being decided by 12%. Estimates based on mandatory retirements at age 60 are similar. Staggering transfers and time-bound disposal rules can attenuate these costs. We extend research on the bureaucracy to the judiciary and show how rules designed to depoliticize personnel decisions can degrade state capacity.

Understanding treatment effect heterogeneity is important, but many standard approaches to estimate heterogeneous treatment effects rely on strong assumptions. We present a novel approach to estimate heterogeneous treatment effects in a variety of experimental and observational set-ups commonly used in economics and political science such as multi-arm RCTs and difference-in-differences designs. We set up the estimation and inference problem in the form of a linear varying coefficient model and use Bayesian Additive Regression Trees (BART) to estimate the coefficients. We demonstrate how the approach can be leveraged to understand heterogeneity in multi-arm experiments and difference-in-differences designs. Our approach makes no parametric assumptions about the treatment effect function and estimates it using a flexible regression tree ensemble. We illustrate the benefits of using the approach to estimate heterogeneous treatment effects for studies across various sub-fields of political science. In doing so, our study adds to the growing literature on using flexible machine-learning methods for causal inference.

The COVID-19 pandemic severely disrupted health landscapes, yet evaluating its full toll across the diverse U.S. rural-urban spectrum remains incomplete. Using restricted county-month cause-specific mortality data from the National Vital Statistics System from January 2020 through December 2023, we implemented Bayesian hierarchical spatio-temporal models to estimate baseline expected deaths relative to pre-pandemic trends (2015–2019). We evaluated both total excess deaths and net non-COVID excess deaths across the nine-tier USDA Rural-Urban Continuum Codes to observe subnational geographic and temporal shifts. Over this four-year window, the U.S. experienced over 1.47 million excess deaths, with 32% (473,452 deaths) not formally attributed to COVID-19. Relative excess mortality scaled progressively with rurality: urban centers experienced 12.4% to 12.7% higher-than-expected mortality, while remote, small-population rural counties peaked at 14.1% to 15.1% excess mortality. Excess mortality that was not attributed to COVID-19 mirrored this gradient, tracking at disproportionately higher rates in remote rural landscapes. Temporally, spikes in unexplained non-COVID excess deaths closely matched pandemic waves. This temporal alignment suggests a pervasive, place-based undercounting of direct COVID-19 deaths, possibly driven by limited resources and varying death investigation frameworks in rural counties. These findings reveal a stark, persistent rural mortality penalty long after initial pandemic policy interventions waned, emphasizing the urgent need for targeted healthcare infrastructure, resource equalization, and strategic rural public health investments.

The COVID-19 pandemic significantly altered mortality landscapes in the U.S., and included substantial excess mortality that was not attributed to COVID-19 itself. Yet, the specific drivers of non-COVID-19 excess deaths remain largely unexplored. This study examines the role of "deaths of despair" (drug overdoses, alcohol-related deaths, and suicides) as a component of excess mortality across the urban-rural continuum in the first four years following the onset of the pandemic. We used county-month level restricted mortality data and a Bayesian hierarchical modeling framework to project expected deaths based on 2015–2019 trends, and then compared to observed DOD in 2020-2023 to estimate excess DOD. Our findings indicate that the pre-pandemic upward trend in DOD intensified substantially during the pandemic period. Nationally, increases in DOD explained nearly 13% of the estimated 458,000 total excess deaths not attributed to COVID-19. While urban counties saw high absolute numbers of excess DOD, the most rural counties experienced the highest relative increases, with DOD rates up to 15.2% above expected levels. Local spatial analysis identified significant hot spots in the Mississippi Delta and portions of Appalachia. Temporal trends show sustained increases in DOD from late 2020 through mid-2023 across most rurality categories. These results suggest that the pandemic intensified pre-existing rural health disadvantages and structural disintegration. Addressing this widening mortality gap requires place-attentive, cross-sector strategies that combine evidence-based health interventions with investments in rural economic and social infrastructure.