Communal Violence, Bargaining Power, and Reproductive Autonomy: Evidence from the 2002 Gujarat Riots
This paper provides evidence that targeted communal violence persistently erodes the intra-household bargaining power and reproductive autonomy of already-married women in an affected minority community. I estimate the long-run effects of anti-Muslim riots in 2002 on Muslim women's bargaining power and reproductive autonomy using four rounds of India's National Family Health Survey (1998–2021) and a triple difference-in-differences design, comparing Muslim to non-Muslim women in Gujarat relative to other Indian states. Decision-making power over household purchases fell by 5.8 percentage points and over decisions about visiting family and friends by 5.3 percentage points, effects that persisted for nearly two decades, while the shift from concealable to male-controlled contraception indicates women lost private control over fertility. Gujarati Hindu women experienced comparable declines, and the effect is not replicated by Muslim women outside Gujarat under similar political or geographic conditions. The impact of communal violence spread far beyond the immediate deaths, diminishing the independence of women in Gujarat for decades afterward.
Submitted · Updated August 25, 2026
Racial Match in One-Shot State Encounters: Evidence from Traffic Stops
This paper studies how racial match between police officers and drivers affects enforcement outcomes during traffic stops. Traffic stops are brief, anonymous, one-shot interactions with substantial discretion and no expectation of repeated contact, making them uniquely informative for distinguishing between competing theories of discrimination. Using administrative stop records with officer identifiers, we exploit within-officer variation to estimate the causal effect of officer--driver racial match on warnings, citations, searches, and arrests.
We find that racial match systematically alters enforcement severity, with effects concentrated in discretionary margins. The pattern of results is consistent with ingroup favoritism in low-severity contexts and a ``black sheep'' effect in high-severity or norm-salient contexts. We further show that racial match does not improve predictive accuracy (e.g., hit rates), helping rule out statistical discrimination as the primary channel. Our findings highlight that interpersonal identity dynamics can shape state enforcement decisions even in one-shot encounters.