Working papers

Information affecting a candidate's reputation might have significant electoral consequences. Do candidates respond to the release of information? Using Brazilian elections and audits as an exogenous source of information, I show that both incumbent and challenger increase their campaign spending when detrimental information affects the incumbent's reputation. Conversely, beneficial information decreases candidates' spending. The main channel is that information affects the expected competitiveness of elections and, therefore, candidates' spending. Only information disclosed before electoral campaigns impacts campaign spending. Furthermore, incumbents also adapt a conditional cash transfers program by increasing (decreasing) the beneficiaries when detrimental (beneficial) reputation shocks occur.

This paper studies crime dynamics and police performance by analyzing an unprecedented policy shifting how individuals interact within the society: a nationwide lockdown. The paper disentangles the underlying mechanisms on how lockdowns affect crime and how diverse types of crimes evolve across different lockdown stages. To do so we use novel criminal case-level data for Bihar, India. First, using a regression discontinuity design in time, we estimate an immediate reduction in aggregate crime of nearly 60 per cent due to the lockdown. However, most types of crimes returned to prelockdown levels after a month and a half. Second, the crime reduction seems to be driven by the higher difficulties faced by criminals in finding potential victims. We observer a larger reduction in crime in districts where citizens had higher compliance with the lockdown (proxied by the number of crimes against public health) compared to districts with low compliance, meanwhile no differential impacts are observed across districts with distinct police strength. While lockdowns reduce the number of new crimes, they also decrease the police’s capacity for solving pending criminal cases. We find that the number of arrests associated with open criminal cases decreased by 86 per cent due to the lockdown. The impact is more severe in districts with lower police staff.

Are elected politicians treated more leniently when facing criminal charges? I present evidence of judicial discretion in the world's largest democracy, India. I analyze whether pending criminal cases of politicians marginally winning the election are more likely to be closed without a conviction compared to cases from politicians marginally losing the election. I find that winning office increases the chances of a favorable outcome only for politicians from the ruling party. Evidence suggests that the misuse of executive powers and witnesses turning hostile are among the main explanations for this result.

This paper introduces a two-stage contest model with reference-dependent preferences to study the determinants of conflict and its intensity. I show the existence of a Subgame Perfect Nash equilibrium in pure strategies, and characterize the properties of the equilibrium. The model shows that reference points play a crucial role in the decision of waging war, and in the level of intensity of the conflict. The model delivers predictions in line with the evidence, and explains empirical regularities that previous models cannot account for. The model encompasses two of the most common empirical patterns found in the conflict literature. Conflicts are more likely to occur after negative income shocks due to the current situation being perceived as a loss compared to agents' reference points. Additionally, income reduces the odds of conflict if agents are more risk-averse for gains than risk-seeker for losses.


A global analysis of the impact of COVID-19 stay at home restrictions on crime, with A. Nivette and others, Nature Human Behavior, forthcoming.

The Great Lockdown and criminal activity - Evidence from Bihar, India , CEPR COVID Economics, 2020, 1(29): 141-163.

Equilibrium with limited-recourse collateralized loans (2013) with J.P. Torres-Martínez, Economic Theory, 53: 181.