Ambiguity attitudes and demand for weather index insurance with and without a credit bundle: experimental evidence from Kenya

We investigate the impact of ambiguity attitudes on the willingness-to-pay (WTP) for index insurance among female smallholders in Kenya. We gauge incentive-compatible measures of ambiguity aversion and insensitivity in the domain of gains and losses, as well as loss aversion. Next, we setup a framed experiment to measure WTP for insurance with basis risk. For a random subsample we introduce an alternative ‘rebate’ insurance, comparable to an insurance purchased through a loan – repaid in good years and deducted from payout in bad ones – that is expectedly more palatable for the loss averse. We find that ambiguity aversion significantly increases WTP for the standalone insurance, while loss aversion reduces it as expected. The former result is seemingly at odds with previous evidence from the field, but is consistent with a setting in which insurance ambiguity engenders relatively less disutility compared to the vagaries of weather. We show that this apparent divergence is not caused by differences in the method used to estimate ambiguity aversion compared to existing field studies. Rather, we exploit exogenous variation in the familiarity with insurance within our sample to show that it is explained away by the role of experience with the novel technology—a previously underestimated mediator. Ambiguity aversion hinders adoption at early stages but increases when the insurance is better understood. The rebate scenario, instead, all but cancels the effect of loss aversion on WTP, but the increased contractual ambiguity results in significantly lower bids by the ambiguity averse. In the lab, the WTP for rebate-type insurance-credit bundles is not different from that of the actuarially equivalent standalone insurance, implying that evidence from the field on greater uptake for the former may be attributable to liquidity constraints and time discounting effects, rather than to behavioural traits.