Evaluating Demand for Multi-Season Crop Insurance: Experimental Evidence from Uganda

We partner with the Uganda Agricultural Insurance Consortium (AIC), to study the demand for multi-season insurance contracts among smallholder farmers. AIC will offer Pay-at-Harvest Insurance—farmers subscribe at planting time but pay the premium at the end of each insured season as a deduction from the harvest payments the off-taker makes to the farmers. The crop insurance product will be an area yield index insurance: the insurance payout is disbursed if the average yield across farmers in a certain area (e.g. an Agro-Ecological Zone, which the insurance consortium maps on the basis of similarity in weather patterns) is below a certain threshold, as measured in a crop-cutting exercise on a sample of farmers.

Farmers can choose to purchase insurance for two seasons and can decide whether they want an option to cancel the contract after the first season. They can also choose to have a cancellation option before the second season only if they receive/don’t receive payouts in the first season (i.e., a state-dependent option).

In addition, we test the effect of an information intervention on farmers' demand. The treatment group (T1) will receive an information treatment aimed at making farmers aware of the potential state dependency of the crop activities and of their insurance demand---farmers will be guided to think through farming activities and insurance demand in different scenarios when the current season is good or bad and the associated pros and cons of the cancellation option in their insurance plan. The control group (T0) will not receive such an intervention.

2024-04-15

2025-04-30

Insurance demand outcome: no purchase, purchase of two-season insurance with/without cancellation option.

• Purchase of a two-season insurance product with a state-dependent cancellation option
• Follow up on whether the chosen cancellation option is used
• Manipulation checks:
o farmers' awareness of state dependence
o farmers' recalled experience with state dependence in other contexts
o farmers’ normative attitude towards state dependence
o farmers’ trust in insurance
• Robustness checks and exploration of heterogeneity with respect to the following covariates:
o Comprehension of insurance terms
o Yield of last cycle
o Trust in offtaker
o Other variables concerning household characteristics and farming activities
o Behavioral measurements
o Survey wave

The intervention targets a sample of farmers who sell their product to a local offtaker. The offtaker provided a list of farmers at the beginning of the study.

Farmers complete a baseline survey and receive a two-season insurance offer that entails paying the insurance premium via deduction from harvest time payments in each season. Specifically, farmers are given the choice to buy insurance (or not) for two seasons, and whether they would like to have the option to cancel the insurance after the first season. They are further asked whether they would like to have the cancellation option regardless of the outcome of the first season, or whether they would like to retain the cancellation option if they received/didn’t receive an insurance payout in the first season.

An information treatment is randomized at the farmer level. The treatment group (T1) will receive an information treatment that guides farmers to think through farming activities and insurance demand in different scenarios when the current season is good or bad and discusses the associated pros and cons of the cancellation option in the insurance plan. The control group (T0) will not receive such an intervention. We also randomly assign sub-counties into 3 waves of data collection.

Not available

The randomization of the information treatment is done on SurveyCTO, only with farmers who are reached for the baseline survey.
The randomization of sub-counties into waves is done in Stata.

The randomization of T0/T1 is at the farmer level. In addition, the order of some survey sections is randomized at the farmer level. The randomization determines whether scenarios presented to the farmers are first about a positive or a negative shock and whether a manipulation check is presented before or after the information treatment. These order-randomizations only occur for those farmers in the information treatment groups.

Survey waves are randomized at the subcounty level stratifying by two district groups such that each district group has seven farmer groups.

No

Approximately 650 farmers targeted for the baseline survey (some may not be reached).

Approximately 650 farmers targeted for the baseline survey (some may not be reached).

50% of the farmers reached in the baseline will be in T1 and 50% in T0.