Supplement to “underinvestment in a Profitable Technology: the Case of Seasonal Migration in Bangladesh”

THIS APPENDIX PROVIDES a simple model of basis risk based on Clarke (2011) and uses it to argue that our 2011 insurance experiment can be used to test whether migration is risky and migrants are risk averse.1 There are two payoff-relevant states {L H} which lead to income at the destination yL ≤ yH . We assume pr(H) = πy . There are two rainfall states {RL RH} and rainfall insurance makes a payment of p in state RH and costs c in state RL. We denote pr(RL)= πR. This setup leads to four possible states of the world: {LRL LRH HRL HRH}. Following Clarke, we parameterize basis risk with a variable r = pr(LRL)—that is, the probability that income is low but that the insurance contract does not pay out and is in fact costly.2 This implies that the remaining probabilities are { pr(LRH)= 1 −πR − r;pr(HRL)= πR − r; pr(HRH)= πy −πR + r }