Miguel, Satyanath and Sergenti (2004)

Question 1

why can’t we just run an OLS regression between economic growth and the likelihood of civil war to establish the causal effect?

Answer 1

endogeneity: war or its future threat affects economic outcomes
- civil war can affect income growth but income growth can also affect civil war

omitted variables
- unmeasured X variable like weak governments/political institutions correlated with economic growth could also affect war risk

Question 2

starting point of the paper

Answer 2

evidence that lower rainfall leads to slower economic growth in Africa, since the economy relies heavily on rain-fed agriculture

taking advantage of this relationship to understand the link between economic growth and civil war
- low rainfall/drought leads to falling incomes which leads to violence

Question 3

data

Answer 3

focus on African countries during 1981-1999

using GPCP data from satellites to estimate local rainfall patterns in Africa
- combines measures from ground weather stations and the density of cold cloud cover to assess precipitation

Question 4

first-stage regression

Answer 4

economic growth on rainfall growth

instrument isn’t very weak but also not very strong

Question 5

reduced form analysis

Answer 5

war on rainfall growth

when rainfall growth is faster/higher, conflict decreases
- significant at 95%

when rainfall is better and economic conditions improve as a result, there’s less conflict in SSA

Question 6

second-stage regression

Answer 6

war on economic growth
- original analysis of interest

Question 7

how do first-stage and reduced form regressions help us estimate the second-stage regression?

Answer 7

if rainfall has a large impact on economic growth but doesn’t increase war risk, then it’s unlikely that economic growth has a causal impact on war

Question 8

IV coefficient estimate on lagged GDP growth (b2)

Answer 8

-2.55 significant at 95%

large in magnitude
- drop in growth of 5% is associated with an increased civil conflict risk of 12%

mean civil conflict in the dataset is 27% so a large economic shock of -5% of GDP increases civil war risk by 40%

Question 9

concern with IV estimation approach

Answer 9

possible failure of the exclusion restriction

variation in precipitation and temperature can affect other factors beyond economic growth that can directly impact the risk of violence
- e.g. effects on roads from flooding, impacts on other economic outcomes (poverty and inequality) and psychological impacts