In fragile states, regimes must cultivate military forces strong enough to ward off external threats, but loyal enough to resist launching a coup. This requires that leaders distinguish the loyal from the untrustworthy, a particularly challenging exercise in post-conflict settings with weak institutions. In this study, I explore how Congolese soldiers operating in North Kivu, the largest operational theater in the Democratic Republic of Congo and the epicenter of one of the most violent conflicts in Africa, solve this crucial task. I argue that leaders use non-payment as a screening strategy that reveals commitment by driving non-loyal soldiers to defect and loyal soldiers to weather challenging times. This process fuels unpaid soldiers engage in civilian abuse, which military units use as a further test of loyalty and as an additional strategy to generate internal cohesion. To develop and test this argument, I couple thick description based on 100 open-ended qualitative interviews with a fine-grained quantitative analysis of 350 surveys of soldiers from the Congolese Armed Forces. This analysis provides a novel explanation for how leaders use financial constraints to overcome classic organizational dilemmas in ways that ultimately cause violence against civilians.
Does establishing evidence of atrocities through the systematic monitoring of war affect the strategic use of violence during conflict? Does monitoring conflict ultimately deter crimes against humanity before they take place? To answer this question, I analyze the impact of “Eyes on Darfur”, the first-ever satellite intervention implemented by Amnesty International USA amidst a brutal genocide with the objective of reducing violence. To examine whether this project changed the behavior of the Sudanese Government in Darfur, I estimate the impact of exposure to monitoring using a novel dataset I construct of high-frequency sub-national data on genocidal attacks. Using difference- in-difference and matching estimators, results suggest that monitoring had pernicious and persistent effects: Amnesty’s intervention increased violence in monitored villages during the program and in subsequent years. The program did not increase or decrease violence in neighboring villages, which assuages fears that monitoring may simply dis- place violence but tempers hopes that it may generate geographically protective en- claves. Coupled with qualitative data, these results suggest that the Government of Sudan leveraged the monitoring intervention to signal their resolve to stay in power.
“Civilian Abuse by Armed Groups: Evidence from Aceh,” with Macartan Humphreys and Jeremy Weinstein. Abstract +/-
We examine strategic and opportunistic logics of civilian abuse in the context of the secessionist conflict in Aceh from 1998 to 2005. We develop a simple model that integrates logics from existing accounts of abuse strategies by warring groups, high- lighting the ways in which, rather than producing distinct logics of abuse, strategic and opportunistic motivations interact. The model also highlights the contingent nature of marginal effects and conditions under which control, support, and discipline may have interactive effects on abusive behavior. We assess hypotheses using data on violence against civilians during the Aceh conflict, with measures drawn from surveys of populations, local elites, ex-combatants, and third parties. In our analysis we seek to examine general hypotheses using a particular case, rather than to tailor an explanation for the Aceh case. Broadly, this exercise yields little or no support for the hypotheses generated. While civilian support for GAM is associated with greater, untargetted, violence by the Indonesian government; it is not associated with reduced violence by GAM. Control is related to abuse, we find, especially when it is imperfect, but it operates in different ways for different groups. Discipline, fighter motivation, and civilian ties are not related to violence in expected ways.
The cooperation of local populations, particularly through the provision of information, is critical to peacekeeping effectiveness. Under what conditions do local populations decide to cooperate with peacekeepers? How do civilians form beliefs about peacekeepers and do these beliefs influence behavior towards them? How does exposure to peacekeeping activities in security or relief, or peacekeeping abuse – three of the primary ways that local communities experience peacekeepers – affect the likelihood that local populations will provide information to peacekeepers? We develop and test a theory that individuals decide to cooperate with peacekeepers based on three key parameters – efficacy, benevolence, and abusiveness – that affect the expected costs and benefits of cooperation. We argue that in low-information post-conflict environments, individual experiences with peacekeepers provide individuals the opportunity to update their beliefs about these key parameters. Using an original survey of a random sample of residents living in metropolitan Port-au-Prince, Haiti, we demonstrate that beliefs about peacekeepers’ efficacy, benevolence and abusiveness are strongly related to willingness to cooperate with them by providing information. Civilians exposed to peacekeeper security and relief services are significantly more likely to be willing to provide them with general and crime-specific information, while those exposed to abuse are less likely to provide critical information on violent events; the magnitude of these effects are substantial. These findings are an opportunity and a challenge for peacekeepers: if public opinion and cooperation are strongly responsive to peacekeeper programming and policy, then peacekeepers must deliver services and prevent abuse in order to solicit the cooperation that they need to achieve their mandates.
“Transparency and Spin: How Political Communication Reinforces (or Undermines) Political Accountability,” with Macartan Humphreys and Jeremy Weinstein. Abstract +/-
Political transparency is often advocated as a tool for improving the performance of public officials. Transparency gives citizens the information they need to assess the activities of politicians and to police their behavior. This simple story is complicated, however, by the fact that new information is not always credible and is often contested by sitting officials with powerful incentives to dispute it. When does political communication affect how voters respond to factual information about the performance of their representatives? Under what conditions do voters take seriously new information about their politicians, and when are they swayed by others to reject it? We address these questions using an experiment in Uganda in which political endorsements were coupled with a broader accountability intervention that tracked and reported to citizens the performance of their Members of Parliament. By providing citizens with information on their representative’s performance, as well as endorsements and rejections of this new information by independent third parties, co-partisans, and non-copartisans, we analyze the way in which endorsements shape citizens’ views. In assessing the effect of information from this angle, we reassess previous findings on transparency interventions in light of the scope for political reinterpretation by actors that seek to endorse, reject, and ultimately mediate the impact of these interventions.
There is increasing support for the use of research registries in social sciences. One possible advantage of the use of a registry is that it would limit the scope for publication or analysis biases that result from selecting statistically significant results. However, to date, there is surprisingly little evidence for the claim that registration will reduce these biases. We look to historical data from medical publishing for evidence, comparing the distribution of p-values before and after the introduction of registration in prominent journals. We couple this analysis with a pre-analysis survey of medical experts and social scientists to assess their prior expectations of the impact of registration on medical publishing and to assess their perceptions on the specificity and sensitivity of our test of effects. Although there is evidence of publication bias in medical studies, our registered analyses uncovered no evidence that registration affected that bias, leading us to moderately downgrade our confidence in the curative effects of registration.