Job Market Paper:
Reelection Incentives and Political Coercion: Evidence from Indonesian Villages - under review
Reelection incentives are intended to hold politicians accountable to voters. This paper tests whether central governments can exploit these incentives to control local politicians. I study this phenomenon during Indonesia's transition to democracy, testing whether village head reelection concerns induce greater support for the incumbent administration. Leveraging staggered village election cycles for identification, I find that villages with leaders facing more proximate elections are more likely to support the party of the autocratic regime. Results are generated entirely by village heads who are eligible for reelection, suggesting that effects are driven specifically by approaching reelection campaigns. Finally, results disappear when villages receive large development grants through non-governmental actors, pointing towards government transfers as a likely mechanism influencing local politicians.
Working Papers:
Democratization and Political Responsiveness: Evidence from Indonesian Villages - under review
Press: World Bank Development Impact Blog
This paper studies changes in political responsiveness following Indonesia's transition to democracy. Leveraging the staggered introduction of elected district heads, event study estimates compare the allocation of resources with and without an elected politician in office, at the same point in time, within the same broader institutional context. When elected district heads take office, night light growth is 2.6 percent greater across villages supporting the winning political party - an effect driven by districts with stronger media presence and political competition in the baseline. These effects, however, are not associated with improvements in local public goods. Disparities in night light growth are more pronounced during village head election years, suggesting that village officials - in addition to voters - are targeted with preferential favors by elected district administrations. Taken together, results suggest that democratization reshapes political responsiveness across new voter constituencies, but does so in part through new clientelist systems rather than broader investment in public goods.
Decentralization and Political Accountability: Evidence from Indonesia's 2014 Village Law
Theory suggests that decentralization can increase the efficiency of public service delivery - but only if elections are able to select accountable leaders. This paper tests this relationship between elections, accountability, and development outcomes following a large decentralization reform in Indonesia. By interacting simultaneous revenue windfalls with staggered election cycles at the village level, this reform generates plausibly exogenous differences in the exposure of villages to newly-appointed village heads following the reform. Findings suggest that newly-elected politicians generate increases in public service provision and night light intensity. Meanwhile, elections following the reform are associated with turnover in under-performing village heads, the appointment of better educated village heads, and heightened implementation of accountability measures mandated by the reform.
Selected Work in Progress:
Incumbent Advantage and Community-Driven Development
Existing literature has shown that development funding may exacerbate incumbent advantage when distributed directly to voters. This project tests whether community-driven development programs avoid these political distortions by putting funding directly in the hands of local communities, while explicitly removing politician control over the allocation of funding. To do so, I use data from Indonesia's Kecamatan Development Program, which ultimately distributed more than $700 million across more than 30,000 villages. Importantly, funding was distributed based on arbitrary population thresholds across local polities. Using a regression discontinuity design based on these population thresholds, I find that villages receiving larger development grants were more likely to support the incumbent political party in each district. Results are more pronounced in villages with greater voter turnout, and are reversed entirely - generating decreases in incumbent support - within villages that do not manage to secure program funding. Taken together, findings suggest that international aid can distort political attribution and accountability, even when politicians do not control funding allocations.
Mandatory Minimum Sentences and Plea Bargaining Outcomes
More than 90% of criminal cases in the U.S. do not go to trial. Instead, the vast majority of cases are decided through negotiated plea bargaining. This project tests whether the threat of severe criminal sentences at trial are leveraged by prosecutors to secure both guilty pleas and cooperation from defendants. Rigorous testing of this hypothesis has been made difficult by (a) the relative dearth of sentencing reform in the U.S., and (b) the shift in criminal behavior that would likely result from the anticipation of such modified sentences. To avoid these challenges, this project uses variation in sentencing guidelines generated by the First Step Act of 2018, which expanded the set of criminal defendants who were eligible to be sentenced without the use of mandatory minimums. Importantly, the act adjusted these guidelines abruptly and unexpectedly - thereby limiting potential anticipation effects - and allowed these changes to apply retroactively, so that they affected the full universe of defendants already awaiting trial. Following the act's passage, newly eligible defendants experienced a 85% drop in the application of mandatory minimums, and a 20% drop in total prison sentence lengths. Despite this sharp improvement in bargaining power, eligible defendants are no less likely to plead guilty. Instead, newly eligible defendants are less likely to assist the government in prosecuting other defendants, and to formally "accept responsibility" for their crimes. Results imply that the threat of severe sentences at trial is in fact leveraged by prosecutors to coerce defendant cooperation in the baseline.
Criminal Records and Legal Employment: Evidence from the Massachusetts Drug Lab Scandal
Do criminal records undermine legal employment opportunities for convicted individuals? If so, could permanent criminal records worsen recidivism outcomes? These questions are difficult to answer in a way that holds constant defendants' criminal history, place of residence, and (potentially unobserved) individual characteristics, while introducing exogenous shocks to criminal records. This project will answer these questions by leveraging variation in cleared criminal records generated by the Massachusetts drug lab scandal. Following the criminal convictions of two drug lab chemists in 2012 and 2014, the MA Supreme Judicial Court dismissed convictions for 37,000 defendants whose drug samples had been processed by these two state employees - generating by far the largest mass dismissal of criminal convictions in U.S. history. Since the assignment of drug samples to chemists was plausibly exogenous prior to the discovery of chemist malfeasance, these dismissals should be unrelated to the individual characteristics and criminal histories of affected defendants. By linking MA court data to employment data from the U.S. Census Bureau, this project will test whether the removal of criminal records increased legal employment following dismissals. Additional work will test whether this shock generates meaningful reductions in defendant recidivism, and whether these effects vary based on the elapsed time between initial convictions and subsequent dismissals. (Data acquisition and IRB approval in progress.)
Judicial Bias and Dynamic Sentencing Behavior
Does biased behavior update in response to external shocks? This project investigates this question within the U.S. federal court system, studying the response of district judges to appeals court decisions. Leveraging the random assignment of cases to district judges, this analysis first generates a judge-specific measure of sentencing discrepancies based on defendant race. Results show that this measure is strongly predictive of sentencing decisions (i) out of sample across subsequent cases, and (ii) across (minority) defendant races. Future work will exploit the precise timing of appeals court decisions to test whether these racial discrepancies adjust in response to overturned rulings. Specifically, work will test whether subsequent sentencing decisions adjust based on (a) the defendant's race in the overturned case ruling, (b) defendant race in subsequent cases, (c) the extent of racial discrepancies in baseline sentencing behavior, and (d) the racial composition of the judicial panel overturning the original decision. Identification for this latter analysis will benefit from the random composition of judge panels at the appellate court level. Taken together, results will shed light on the dynamics of biased behavior, and the ability of external feedback to moderate this behavior within a high-stakes setting. (Data acquisition complete, preliminary analysis ongoing.)
Reelection Opportunities and Political Malfeasance
This project tests for electoral cycles in political corruption. While prior empirical work has identified reductions in politician malfeasance driven by reelection incentives, data limitations often make it difficult to measure the persistence of these effects. This work overcomes these challenges by observing annual measures of corruption across Indonesian districts, while leveraging staggered electoral cycles for Indonesian district heads for identification. Results show that observed corruption decreases just prior to mayoral elections, but recovers to baseline levels shortly thereafter. Results suggest that elections may induce short-term reductions in corrupt behavior, without generating sustained changes in corrupt behavior.
Lame Duck Politicians and Deforestation across Brazilian Municipalities
Do reelection incentives affect enforcement effort exerted to control illicit activities? This project seeks to answer this question across Brazilian municipalities, testing whether deforestation rates are higher under "lame duck" mayors immediately following failed reelection bids. Estimates are derived using monthly deforestation rates within each municipality, and make use of a two-month window where lame duck mayors continue to hold office prior to relinquishing power. Results show higher rates of deforestation under lame duck mayors, and are particularly pronounced when preceding elections were closer.