The role of information on preferences and demand for bundled index-based livestock insurance

We will use a choice experiment to assess preferences for bundled index-based livestock insurance. The experiment will entail giving pastoralists in northern Kenya a discount coupon to spend on services or products pertaining to livestock production. The amount of the coupon will be varied randomly across the pastoralists. The experiment will also entail providing a training session to a random sub-sample of the communities. In the training session, we will provide information on the pros and cons of bundled products, and information on the condition of the vegetation in the area (proxy for pasture condition and availability, and availability of water).

Pastoralists will be invited to a central place for the experiment. In the first session, the pastoralists will receive instructions about the experiment in groups of 8. The second session will entail a short exercise where pastoralists engage in an activity to earn some money to later use in the experiment. This will help reduce the “house money effect”, likely to arise when individuals receive an “unearned” endowments (Thaler & Johnson, 1990). The fourth session will entail a short demographic survey, where we collect demographic information about the pastoralists. The pastoralists in the no information intervention will then proceed to the experiment.

Pastoralists in the information intervention group will receive the intervention before the experiment. Two sets of information will be given. First, the pastoralists will receive basic training on bundling (the rationale for bundling, the pros and cons of bundled products, and what to consider when selecting bundled products) in groups of 8. Second, pastoralists will receive forecast information on vegetation conditions (proxy for pasture condition and feed availability). After the interventions, the pastoralists will proceed to the experiment individually. The experiment will be conducted as follows:

We will start with a practice session to enhance understanding of the bidding (switching) process hence reducing the so called game form recognition failure characteristic of the BDM mechanism (Cason & Plott, 2014). The practice session will entail three types of soap with easily differentiable attributes. Each pastoralist will be randomly assigned to one of the three types of soaps and asked whether she/he will keep the soap or prefer to switch to another soap after paying some amount of money (more details in the appendix). To determine whether the pastoralist switches from one soap to another, we will follow the standard BDM procedure (Becker et al., 1964) as follows.

First, we will inform the pastoralists about their “default” products. We will then ask them to state the maximum amount of money they are willing to pay to switch to their preferred bundles if they do not prefer the default bundle. We will inform the pastoralist that actual switching (not switching) will depend on the amount they are willing to pay, and luck. We will then offer the pastoralist a bag with numbered 10 chips. An individual drawing number 1-5 will get a discount of KES1,000 and individuals drawing 6-10 will get a discount of KES2,000. We will open a sealed envelope and compare the maximum willingness to pay with the amount indicated in the sealed envelope. Those whose maximum willingness to pay beats or exceeds the binding price will be allowed to switch. After the practice round, the pastoralists will then bid for the IBLI bundles.

2024-08-13

2024-09-15

1. Knowledge of bundling before and after the training. This will be a composite consisting of 6 questions with equal weights.
2. Pastoralists preference for bundled IBLI products (maximum willingness to pay).

Knowledge will be a composite index consisting of 6 questions with equal weights.

1. Whether or not pastoralists redeemed their coupons.
2. The type of product/service redeemed at the points of sale. This will represent actual demand for the bundled products/services.
3. The total expenditures on the bundled products/services.

The experiment will target 260 pastoral groups (130 men groups, and 130 women groups), each with approximately 8 members. We will have three levels of exogeneous variation in the experiment. The first level will entail randomly assigning pastoralists to either the information intervention group (n = 130) or the no information intervention group (n = 130). The second level will entail randomly assigning pastoralists in each of the two groups to a coupon amount of either Ksh. 1000 (n = 65) or Ksh. 2000 (n = 65). This will be done at the field to enhance transparency—the pastoralists will select a piece of paper from a bag to determine whether they receive the Ksh. 1000 coupon or the Ksh. 2000. Lastly pastoralists in each of the 4 groups will be assigned randomly to one of the four products, namely IBLI+animal health, IBLI+animal nutrition, IBLI+breeding, and IBLI+a flexible package.

Though CEs are usually used to elicit preferences for products or product attributes (De-Magistris et al., 2013), they suffer from major pitfalls namely hypothetical bias (Johnston et al., 2017), and lack of incentive compatibility (Lusk & Schroeder, 2004). Hypothetical bias is not a challenge in our study because the products are well known to the study participants. In addition, our experiment is incentive compatible because choices in the experiment will have real life consequences. The experiment will target 260 pastoral groups (130 men groups, and 130 women groups), each with approximately 8 members. We will have three levels of exogeneous variation in the experiment. The first level will entail randomly assigning pastoralists to either the information intervention group (n = 130) or the no information intervention group (n = 130). The second level will entail randomly assigning pastoralists in each of the two groups to a coupon amount of either Ksh. 1000 (n = 65) or Ksh. 2000 (n = 65) . This will be done at the field to enhance transparency—the pastoralists will select a piece of paper from a bag to determine whether they receive the Ksh. 1000 coupon or the Ksh. 2000. Lastly pastoralists in each of the 4 groups will be assigned randomly to one of the 4 products, namely IBLI+animal health, IBLI+animal nutrition, IBLI+breeding, and IBLI+a flexible package.

Description of the bundles
1) IBLI + Health related products (B1)
This bundle will include IBLI and health related products namely veterinary medicines (for example antibiotics), vaccines, and clinical services. The pastoralist will have the liberty to use the coupon on any of the three health related items, the only condition being that any expenditure above the amount of the coupon will be incurred by the pastoralist. The products purchased can be used on livestock other than cattle.
2) IBLI + Nutrition related products (B2)
The bundle will include IBLI and products relating to the nutrition of livestock, for example feeds and fodder. Pastoralists will have the liberty of using the coupon on any nutrition related products (for example fodder, pellets, or other commercial feeds), the only caveat being that the expenditure should not exceed the amount of the coupon. The pastoralist will incur any extra cost above the amount of the coupon. The products purchased can be used on livestock other than cattle.
3) IBLI + Breeding related products (B3)
This bundle will consist of IBLI and products relating to livestock breeding such as artificial insemination (AI). The coupon can be used on breeding material for other livestock in addition to cattle. Any expenditures above the amount of the coupon will be incurred by the pastoralist.
4) IBLI + health or nutrition or breeding related products (B4)
Bundle 4 will be a flexible bundle, where pastoralists can use the coupon on a product of their choice, either health, nutrition, or breeding related. Like all the other products, any extra expenditure above the amount of the coupon will be incurred by the pastoralist.

Randomization will be done in the office by a computer.

Groups of men and women will be randomized into the information treatment. Exogeneous variation in coupon amount will be at individual level.

Yes

260 men and women groups.

2,080 pastoralists

130 men groups and 130 women groups. Each group comprises 8 individuals.

Assuming an inter-cluster correlation of 0.2, an effect size of 0.2, and 8 pastoralists in each cluster, gives us a power of 0.903 (after 1000 simulations). At the second level of randomizations, holding the same assumptions with 125 clusters, we obtain a power of 0.6. However, important to note is that the power estimates are conservative—every pastoralist will evaluate all the 4 bundles meaning that the number of observations will increase by a factor of 4.