Dissertation

Although conflict has been studied by economics, our empirical knowledge has tended come from either aggregate cross-country studies or post-conflict outcomes. As a result, our knowledge of how households adapt during conflict remains very limited. My dissertation contributes to the literature by disentangling the effects of conflict risk and realization thereby allowing me to study how households respond to insecurity.

Essay I: “The Cost of Fear: The Welfare Effects of the Risk of Violence in Northern Uganda”

The micro-conflict literature focuses almost exclusively on direct exposure to violence and post-conflict outcomes. By focusing only on directly exposed households, the literature ignores the effects of risk on households in surrounding areas. This paper presents the first estimates of the economic costs of the risk of violence separate from the costs of the actual experience of violence, and finds that it is a significant mechanism by which conflict influences development. Using representative community and household data from Northern Uganda, I estimate measures of objective and subjective risk using geo-spatial variation in the distribution of violence over time. On average, the risk of violence lowers per capita household expenditure by 2 to 6 percent. Even within households that are attacked, risk alone accounts for a significant share, between 17 and 38 percent, of their losses. On aggregate, half of conflict-related losses are due to risk as opposed to direct exposure to violence, with much of these risk-related losses in households that are not directly attacked. Compounding these losses over the duration of the conflict, the risk of violence has reduced per capita expenditure in the affected region by roughly 70 percent and national GDP by 4.6 to 8.2 percent. Lastly, I find that food aid reduces risk-related losses by 17 to 30 percent.

Essay II: “Living Within Conflicts: Risk of Violence, Livelihoods, Portfolio Choice and Returns” (Please contact me before citing)

Economic theory suggests that households will respond strongly to the risk of violence during conflicts, however, the effect of this risk on livelihoods has not been well studied empirically. Using spatial variation in the placement of violence in Northern Uganda and data from over 690,000 households, this paper examines the effects of the risk of violence on livelihoods and disaggregates its possible effects along extensive and intensive margins. I find that households alter both the value of the livestock holdings (-65% at the mean level of risk) and its composition; livestock portfolios shift away from more profitable but riskier large livestock to smaller livestock. This risk also appears to strongly influence cropping decisions. Whereas risk has little effect on female labor force participation, men shift from unpaid family labor to becoming employees. In contrast, risk seems to have only limited effects on sources of income. Similarly, there is no real effect on the returns to assets. Moreover, the practical effect of risk disappears once asset portfolio allocation is accounted for. This suggests that the effect of the risk of violence may operate primarily through changes in assets composition and in the activities in which these are employed and not through changing returns to these activities

Essay III: “Insecurity in Pastoralist Setting; Evolution and Effects” (currently writing with Chris Barrett)

While studies highlight the importance of household responses to insecurity, our empirical knowledge of insecurity and its effects remains very limited. Due to the absence of direct measures of perceptions of insecurity, research has relied on responses consistent with qualitative priors or with risk mitigation. Even when measures of risk are constructed, these do not directly measure risk but rather rely on the spatial-temporal variation in the placement of violence. Using a unique ten period panel of household perceptions of insecurity, this paper focuses on three important and previously unstudied issues regarding perceptions of insecurity. First, how do individual perceptions of insecurity evolve across time? The analysis examines the effects of previous beliefs while accounting for weather, an important factor in the literature on conflict. Second, we study the relative contribution of current and lagged expectations of insecurity on incomes. Since levels of insecurity may persist across time, lower current outcomes may simply reflect the result of previous perceptions and responses. Third, the effects of the risk of violence have been studied in isolation of other risks even though it is just one of many risks to which households might respond. Consequently, using similar data on the household perceptions of a variety of other common risks, including health, weather and prices, we examine the robustness and relative effect of insecurity. Moreover, prior studies have typically relied on cross-sectional data and therefore could on control for observed characteristics. The panel nature of the data allows us to control for the likely unobserved time invariant characteristics that are correlated with outcomes examined.