Three Essays on the Interrelationships Among Climate, Conflict and Economics

Abstract

Conflict in a country is socially expensive and many are trying to understand what factors stimulate it in an effort to figure out ways to lessen its incidence. In this work three essays are presented on factors that drive conflict. The factors examined are: 1) the interrelationship between climate and conflict, 2) the causality between commodity prices and conflict, 3) the ways cereal demand affects and is affected by terrorism.

In the first essay, we use a global dataset to econometrically explore whether the probability of conflict is affected by climate. We find that precipitation variation does have a statistically significant effect. That is, the less precipitation this year relative to the last, the more likely the country is to suffer from civil conflict. Methodologically the best predictions are obtained from a semiparametric estimation technique.

In the second essay, we econometrically investigate the dynamic relationship between commodity prices and the onset of conflict in Sudan. Applying Structure Vector Autoregression (SVAR) and Linear Non-Gaussian Acyclic Model (LiNGAM), we find that wheat price is a cause of conflict events in Sudan. However, we find no feedback from conflict to commodity prices.

In the third essay, we examine the extent that demand for three main cereals in Sudan (sorghum, millet, and wheat) is altered by the incidence of terrorism plus the effect of terrorism events on cereal demand. This is done by using an Almost Ideal Demand System (AIDS) and a Directed Acyclic Graph (DAG) approach. The results show terrorist attacks do cause changes in commodity demand for wheat. The DAG analysis also tentatively suggests that wheat demand is both marginally affected by and directly affecting the incidence of terrorism (conflict) in Sudan. Subsequently, we generate forecasts for the three commodities shares with the AIDS and DAG models, incorporating the effects of terrorist attacks. Examining those results independently and jointly, we find that a composite forecast of the two generates better forecasts.