Measuring the Economic Impact of Civil War

Civil wars impose substantial costs on the domestic economy.1 These wars are destructive of human lives and economic infrastructure. They also undermine the legitimacy of the state, threatening its institutions, the security of property rights, and the rule of law. Moreover, internal wars introduce tremendous uncertainty into the economic environment, making both public and private investment riskier. While it might be readily apparent that war will impact economic production, there is little understanding about the di*erentchannels through which civil war a*ectstheaggregate economy. How costly are these internal wars? What is the mechanism through which civil war negatively impacts the domestic economy? Building on the emerging literature on the economics of civil war, this paper measures the economic costs of war andidentiÞesthewaysinwhichcivilwarstripsa country of its growth potential. This paper contributes to the existing analysis both theoretically and methodologically. In particular, we explore the economic channels through which civil war a*ectsgrowthby investigating two theoretical arguments. First, civil war impacts the domestic economy by reducing the level and growth of the capital stock. The occurrence of civil war initiates capitalßightandthus dramatically reduces private investment. Second, political economy models suggest that internal conßicts a*ecttheaggregate domestic economy by worsening the government'sÞscalbalance. Economically, governments shift expenditure from output enhancing activities into the conduct of war. Politically, they face weaker incentives to maintain Þscalbalanceowingtoashortertime horizon and weaker accountability to an electoral constituency. We directly test these two potential explanations. Our results indicate that the driving force behind the negative e*ectsofcivilwaron economic growth isadecreasein domestic investment, and in particular, private investment. The second goal of this paper is to provide abetter measurement of the economic impacts of civil war. To do so, we investigate the characteristics of civil war that are the most damaging for the domestic economy. Civil wars vary tremendously in their scope. Some wars are fought entirely in one region of the country, while in others the Þghtingextendsthroughoutthecountrysideand into urban areas. Some civil wars may involve high levels of civilian fatalities as in Sierra Leone,Mozambique, and Guatemala, while others maybe fought largely between competing militaries. In our empirical analysis, we ÞndthatthescopeofthecivilwarsigniÞcantlyinßuencesthemagnitude of the economic e*ects. Wars that spill out across the entire country, require the highest level of military recruitment, and result in the greatest number of fatalities are the most damaging to the domestic economy. This paper also makes methodological improvements. First, we improve the data. The data set used in the existing literature on the economic consequences of civil war contains a number of missing observations despite the fact that it relies on the decade-average of economic indicators. AsigniÞcantproblemisthattheseobservationsarenotmissingcompletelyatrandom. In our sample, thelistwise deletion of missing observations leads tosigniÞcant over-representation of war- decades. Such non-random missing data is likely to bias the result of analysis. We utilize a multiple imputation program developed by King et al. (2000) in order to exploit all available information. The empirical analysis based on this improved data set gives a more accurate assessment of the economic e*ectsofcivilwar. Second, we improve the model. The existing empirical literature exploring politicalinßuences on growth has often applied a pooled Ordinary Least Squares to panel data. However, this approach is problematic in that it ignores countryspeciÞce*ects, possibly leading to biased results. In this paper, we test the robustness of our pooled OLS results by applying appropriate panel data models to the analysis of economic growth and politicalconßicts.download