No abstract

Policymakers can take actions to prevent local conflict before it begins, if such violence can be accurately predicted. We examine the two countries with the richest available sub-national data: Colombia and Indonesia. We assemble two decades one fine- grained violence data by type, alongside hundreds of annual risk factors. We predict violence one year ahead with a range of machine learning techniques. Models reliably identify persistent, high-violence hot spots. Violence is not simply autoregressive, as detailed histories of disaggregated violence perform best. Rich socio-economic data also substitute well for these histories. Even with such unusually rich data, however, the models poorly predict new outbreaks or escalations of violence. \Best case" scenarios with panel data fall short of workable early-warning systems.

We estimate the effects of one of the largest anti-vote-buying campaigns ever studied—with half a million voters exposed across 1427 villages—in Uganda’s 2016 elections. Working with civil society organizations, we designed the study to estimate how voters and candidates responded to their campaign in treatment and spillover villages, and how impacts varied with campaign intensity. Despite its heavy footprint, the campaign did not reduce politician offers of gifts in exchange for votes. However, it had sizable effects on people’s votes. Votes swung from well-funded incumbents (who buy most votes) towards their poorly-financed challengers. We argue the swing arose from changes in village social norms plus the tactical response of candidates. While the campaign struggled to instill norms of refusing gifts, it leveled the electoral playing field by convincing some voters to abandon norms of reciprocity—thus accepting gifts from politicians but voting for their preferred candidate.

We study two interventions for underemployed youth across five Ethiopian sites: a $300 grant to spur self-employment, and a job offer to an industrial firm. Despite significant impacts on occupational choice, income, and health in the first year, after five years we see nearly complete convergence across all groups and outcomes. Shortrun increases in productivity and earnings from the grant dissipate as recipients exit their micro-enterprises. Adverse effects of factory work on health found after one year also appear to be temporary. These results suggest that one-time and one-dimensional interventions may struggle to overcome barriers to wage-or self-employment.JEL codes: J24, O14, F16, J81, O17