Economic Shocks and Civil Conflict: An Instrumental Variables Approach EDWARD MIGUEL, SHANKER SATYANATH, ERNEST SERGENTI P R E S E N TAT I O N B Y: F AT O U M ATA C I S S É & C A N D I C E YA N D A M

Outline I. Presentation of the research question II. Endogeneity issue III. The estimation strategy

IV. The dataset V. The results VI. Conclusion

VII. Critical assessment of the empirical approach

I. Presentation of the research question Civil wars provoke immense human suffering, with famine and deaths. War causes people to move to other regions and also causes disease and disability. In Sub-Saharan Africa, 29 of 43 countries suffered from civil conflict in the 1980-1990. El Niño in Somalia, caused an important drought which in turn caused famine and civil war. The drought itself caused 250 000 deaths in 2012 alone. Droughts are a recurring plight in Somalia and have been at the heart of civil conflicts. Therefore, it is an important development matter which has gained increasing attention from researchers.

What is the impact of economic conditions on the likelihood of civil conflict ? The literature has shown that economic growth may affect civil conflict through many channels. - Collier and Hoeffler (2002): channel of the labor market, education, lower income. - Elbadawi and Sambanis (2002): ethnic fractionalization, democracy. - Fearon and Laitin (2003): lower per capita GDP associated with onset of civil war. BUT, there is a significant endogeneity problem between economic conditions and the onset of civil wars (simultaneity bias). They have tried to resolve it with lagged variables but it is not a convincing approach. There is also a problem of omitted variables.

II. Solving for endogeneity The relationship between economic conditions and civil conflict is not necessarily causal. The authors needed to find a good instrumental variable. What makes a good instrument?

The authors use the exogenous variations in rainfall as an instrumental variable for growth - they show that variations in rainfall are associated with variations in growth but not directly associated with variation on the onset of civil conflict. - variations in rainfall are uncorrelated with the error term

III. The estimation strategy 3 major estimation equations First stage: 𝐺𝑟𝑜𝑤𝑡ℎ𝑖𝑡 = 𝑎1𝑖 + 𝑋′𝑖𝑡 𝑏1 + 𝑐1,0 Δ𝑅𝑖𝑡 + 𝑐1,1 Δ𝑅𝑖,𝑡−1 + 𝑑1𝑖 𝑦𝑒𝑎𝑟𝑡 + 𝑒1𝑖𝑡 Reduced form: 𝑐𝑜𝑛𝑓𝑙𝑖𝑐𝑡𝑖𝑡 = 𝑎1𝑖 + 𝑋′𝑖𝑡 𝑏1 + 𝑐1,0 Δ𝑅𝑖𝑡 + 𝑐1,1 Δ𝑅𝑖,𝑡−1 + 𝑑1𝑖 𝑦𝑒𝑎𝑟𝑡 + 𝑒1𝑖𝑡 Second stage:

𝑐𝑜𝑛𝑓𝑙𝑖𝑐𝑡𝑖𝑡 = 𝛼2𝑖 + 𝑋′𝑖𝑡 𝛽2 + 𝛾2,0 𝑔𝑟𝑜𝑤𝑡ℎ𝑖𝑡 + 𝛾2,1 𝑔𝑟𝑜𝑤𝑡ℎ𝑖,𝑡−1 + 𝛿2𝑖 𝑦𝑒𝑎𝑟𝑡 + 𝜖2𝑖𝑡 where they instrument for growth using predicted relationship from the first stage

IV. The dataset Panel and cross-country data 41 African countries over the 1981-1999 period. A. Data on civil conflict

Armed Conflict Data: records conflicts with 25 battle deaths per year and 1000 battle deaths per year. Conflicts between states and organized groups and conflicts between countries. B. Data on rainfall Global Precipitation Climatology Project Data: monthly rainfall estimates. They measure the rainfall shock as the proportional change in rainfall from the previous year (𝑅𝑖𝑡 −𝑅𝑖,𝑡−1 )/𝑅𝑖,𝑡−1 . C. Other data Fearon and Laitin (2003) and World Bank: ethnolinguistic fractionalization, religious fractionalization, measures of democracy, log per capita income, proportion of country that is mountainous.

V. The results

VI. Conclusion - Solving for the endogeneity problem with use of rainfall shocks - Negative correlation between growth shocks and the likelihood of civil war

VII. Critical assessment of the empirical approach - Rainfall might not be the best instrument (cf appendix A) - The existence of other channels than economic growth (cf Rainfall and Conflict, Sarsons 2011) - High levels of rainfalls might directly lead to civil conflict

- Potential violation of the exclusion restriction