We are going to replicate an instrumental variables analysis of civil conflict in Africa, using rainfall as an instrument for economic growth. Consider skimming the article, by Edward Miguel, Shanker Satyanath, and Ernest Sergenti. Download replication data, which comes with variable definitions.
Please replicate model 5 of table 4. Do you get the same coefficient estimates? How about standard errors? If there is a difference, can you explain why?
How strong is the instrument? Are there any obvious difficulties with the estimate?
Assume that the instrument is unproblematic. What is the causal meaning of the estimate?
Weak instruments: Compute the first‑stage F‑statistic manually. Is it above 10?
LATE interpretation: Who are the compliers in this context? Explain in plain language.
Suppose the exclusion restriction is violated because rainfall affects conflict through channels other than growth (e.g., agricultural livelihoods directly affect grievances). Use sensemakr to assess how strong a direct effect of rainfall on conflict would need to be to render your IV estimate insignificant. (Hint: you’ll need to run a reduced‑form regression.)