Press coverage: Marginal Revolution
Abstract: Are peaceful or violent protests more effective at achieving policy change? I study the effect of protests during the Civil Rights Era on legislator votes in the US House. Using a fixed-effects specification, my identifying variation is changes within the congressional district over time. I find that peaceful protests made legislators vote more liberally, consistent with the goals of the Civil Rights Movement. By contrast, violent protests backfired and made legislators vote more conservatively. The effect of peaceful protests was limited to civil rights-related votes. The effect of violent protests extended to welfare-related votes. I explore alternative explanations for these results and show that the results are robust to them. Congressional districts where incumbents were replaced responded more strongly. Furthermore, congressional districts with a larger population share of whites responded more strongly. This is consistent with a signaling model of protests where protests transmitted new information to white voters but not to black voters.
Works in Progress
“Network and Spillover Effects with Endogeneous Multidimensional Networks” (with Rob Garlick and Kate Orkin)
Abstract: Social networks are crucial for a range of economic activities, e.g., for sharing information, pooling capital, or insuring against economic shocks. Consequently, development policies, by strengthening or weakening connections, can significantly change the well-being of the most vulnerable households. Our paper asks two questions. First, how do economic and behavioral interventions affect economic and psychosocial networks? Second, how do economic and behavioral interventions affect non-recipients through economic and psychosocial networks? To answer these, we run a randomized trial in 415 villages in Western Kenya. We cross-randomize respondents into receiving a large unconditional cash transfer or a behavioral intervention designed to increase long-term planning. We measure network connections in both the economic (borrowing and lending) and the psychosocial domain (planning and emotional support). Therefore we are able to pick up intervention effects both within and across domains. We also measure connections both before and after the intervention, allowing us to study the dynamics of network formation. In this way, the data we collect allow us to estimate the broader impact of economic and behavioral interventions, especially on economically vulnerable non-recipient households.