Risk, Complexity, and the Demand for Agricultural Insurance: Evidence from a Lab Experiment in Ghana
How to cite
Shenoy, Ashish; Chakraborty, Anujit; Gallenstein, Richard; Flatnes, Jon Einar; Korb, Mira. 2024. "Risk, Complexity, and the Demand for Agricultural Insurance: Evidence from a Lab Experiment in Ghana." AEA RCT Registry. AEARCTR-0012876. https://doi.org/10.1257/rct.12876-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
In this experiment, we present participants with a lottery framed as revenue derived from the sale of maize production on an agricultural plot. We then offer participants the opportunity to purchase various forms of insurance against negative lottery outcomes. The insurance choice is presented under various conditions that alter the transparency of the contract and the participant's return from the lottery. We make real payouts to participants based on one such decision to incentivize accurate reporting. Participant willingness-to-pay for insurance constitutes the primary outcome data from this trial.
Intervention
In this experiment, we bring participants into a lab for a 3-4 hour session where they make choices framed as insurance decisions on a hypothetical maize plot. The insurance choice is presented under various conditions that alter the transparency of the contract or the participant's return from the lottery. We make real payouts to participants based on one such decision to incentivize accurate reporting.
Agricultural Revenue Lottery
Revenue lottery outcomes consist of the product of a random draw of crop yield and a random draw of sale price. Yield can take a value of 2, 4, or 6 bags; and price can take a value of 400, 500, or 800 Ghanaian cedis (GHS) per bag; leading to nine possible revenue lottery outcomes. The lottery is implemented by drawing one of nine chips from a bag, with each chip representing a unique combination of yield and price, which generates a uniform probability distribution.
Agricultural Insurance Contracts
We offer participants three forms of insurance against the lottery framed as yield insurance, a hybrid yield-or-price insurance, and a comprehensive revenue insurance. Each contract provides a fixed supplemental payment in case of some lottery realizations, scaled so that all contracts have the same actuarially fair value. Yield insurance offers a payment of 500 GHS in any of the six cases where yield takes on a value of 2 or 4. The yield-or-price contract offers a payment of 600 GHS in any of the five cases where either yield takes a value of 2 or price takes a value of 400. Revenue insurance offers a payment of 600 GHS in any of the five cases where revenue (i.e. yield times price) takes a value of 2,000 GHS or lower. We elicit participants' willingness-to-pay separately for each form of insurance in comparison to the uninsured lottery.
Treatment Conditions
Elicitation of Willingness-to-Pay
Study participants are endowed with 500 GHS in experimental earnings. For each insurance contract under each experimental condition, we elicit willingness-to-pay by offering a list of prices and asking if the participant would prefer to spend some of their 500 GHS to buy insurance at each price on the list, in ascending order, or to play the uninsured lottery. After all elicitations are complete, we select one round at random to implement the participant's choice for incentive compatibility. We then select a price at random for the implemented round, and a lottery realization if a non-mirror round is selected. Participants' farm revenue and insurance costs/payouts are determined according to these draws and added to their participation fee. Actual participant compensation is calculated as the value of experimental earnings divided by 25, made clear in advance.
Experimental design
Insurance is offered under four different experimental conditions that vary either how the lottery is described or how participant lottery outcomes are determined. First, the "standard" condition describes the lottery as independent draws of yield and price, and describes insurance contracts according to the circumstances that trigger payment as explained above. Each participant's outcome in the standard condition is the result of a single lottery draw, plus the insurance payout if insurance is purchased, plus the remainder of a 500 GHS endowment that is not used to purchase the insurance.
Second, the "window" condition simplifies the lottery and insurance setup by providing participants with a list of possible revenue outcomes, ordered from low to high. Insurance contracts in the window condition are presented by explicitly identifying which realizations in the list trigger payment. Notably, this condition makes clear that both the yield and yield-or-price contracts have cases where they make a payment for a higher revenue realization while not paying for a lower realization. We do not explicitly name the contracts in the window condition, but simply present each as a set of contingent payments on the revenue list, to limit contamination from window rounds onto subsequent standard rounds where participants must match payments to revenue realizations on their own. Participant outcomes are again defined by a single lottery draw in the window condition, plus the insurance payout if insurance is purchased, plus the remainder of a 500 GHS endowment that is not used to purchase the insurance.
Third, the "mirror" condition presents the lottery and insurance as in the standard condition, but alters how outcomes are determined. In this condition, a participant's outcome is defined to be the expected value of the lottery or lottery-plus-insurance rather than the result of a single draw. This condition effectively induces expected-value preferences over a stochastic outcome while preserving the complexity of how the uncertainty is presented, following Oprea (2024, "Simplicity Equivalents," unpublished manuscript). In practice, we explain that we will add up the payment from every chip in the bag and divide by nine so that participants do not require statistical literacy about concepts such as mean or expected value to understand their payout. Participant outcomes are equal to the expected value of the lottery, plus the expected value of the insurance policy if they purchase insurance, plus the remainder of the 500 GHS endowment.
Fourth, the "mirror-window" condition presents the lottery and insurance with an ordered revenue list as in the window condition, and determines participant outcomes according to expected value as in the mirror condition.
Randomization Deisgn
Every participant expresses willingness-to-pay for all three contract types under all four experimental conditions. We randomize the order of elicitation to control for order effects in participant response following a three-tier randomization process. At the highest level, we randomize the order of mirror vs. non-mirror elicitations so that the method of determining the lottery outcome only changes once for each participant. Second, within the mirror and non-mirror elicitations, we randomize the order of the window and non-window presentation of contracts. Finally, within each experimental condition, we randomize the order in which the three contract types are presented.
Elicitations take place in sessions of sixteen participants with four enumerators per session. Participants in a session are trained and practice the general structure together, and then break up into groups of four participants per enumerator for actual demand elicitation. Elicitation order is randomized at this enumerator-session level, leading to order assignment in clusters of four participants each.
Within a cluster, the enumerator introduces a treatment condition to all four participants, and then brings each participant into an isolated privacy station to elicit demand. Each participant expresses willingness-to-pay for all three contract types before moving on to the next participant, and participants are discouraged from discussing responses outside the privacy station. Once all four participants in the cluster have expressed demand, the cluster moves on to the next treatment condition.
After all four treatment conditions are complete, participants express their certainty equivalent for three additional lotteries as thirteenth through fifteenth incentivized rounds using the same multiple price list format.
Design details
Cited by
No citation data available for this registration yet.