The Impact of Village Insurance Savings Accounts on Women's Assets and Savings

We will implement a randomized controlled trial measuring the impact of a two-pronged intervention that:

• Provides insurance education to women and promotes the use of their savings group as a way to save money in advance for insurance purchases through what we call a village insurance savings (VISA) account. These accounts will be essentially an accounting device in which members of a savings group publicly contribute monthly sums of money into a common account with the explicit objective of saving enough funds to purchase insurance at a future time period. As with all contributions held by savings groups, all funds are held in a locked box by the group’s treasurer in a publicly known location. Withdrawal of these funds at any point of time will be allowed and the decision to purchase insurance with the accumulated savings will always remain voluntary.

• Reframes the current IBLI contract as a product that can be used to protect women’s assets and income in the event of a drought, even if the insurance holder does not directly manage livestock. This ‘Family Framing’ of the insurance product contrasts with the standard approach of focusing on the livestock-protection aspect of this kind of insurance.

Half of the 410 savings groups in our sample will be randomly selected to receive this intervention.

All of the savings groups in the sample were initially conformed in the context of a previous poverty-graduation program (the REAP initiative carried out by BOMA, a local NGO), and the intervention will be delivered by community-based ‘facilitators’ who have previously worked in this prior BOMA Project intervention and speak the local language. Facilitators will make monthly visits to savings groups in the treatment arm between March and September when the IBLI sales window opens, and the product is available for purchase. On the first visit to treatment-arm groups facilitators will introduce the insurance product in its ‘Family Framing’ version, and will invite group members to start a new savings fund (operated within the structure of their savings group) meant to accumulate enough money to buy insurance on the next sales window. On each subsequent visit facilitators will provide general information on risk and risk-mitigation strategies, and specify the details of the insurance product being sold by Takaful Insurance of Africa (TIA), our partner insurance company. Groups in the control arm will be visited only twice (at the beginning and at the end of the intervention) and will be presented with the insurance product in its traditional ‘Livestock Framing’ version. Facilitators will not invite group members in the control arm to start an insurance-dedicated savings fund.

Additionally, all group members in both treatment and control arms will receive a subsidy covering a fraction of the price of the insurance product. On average 25% of women within each group will receive a ‘high’ subsidy covering 50% of the product’s total price, while the remaining 75% of group members will receive a ‘low’ subsidy covering 5% of the product’s price. The distribution of the subsidy coupons will be made at random during the facilitator’s first visit by drawing colored balls from a bag, so the actual share of high- and low-subsidy recipients in each group will vary randomly.

We expect this intervention to increase insurance uptake rates by women belonging to savings groups in the savings arm. We plan to further investigate the effect of the intervention on women’s business assets, income level, total savings, and food security status one season after the purchase of insurance, comparing these outcomes both across insured and uninsured individuals and across households in regions affected and unaffected by drought. Specifically, food security status will be measured using, i) the number of days in a week the household’s children skipped dinner, and ii) the number of times in a day during the past week where children ate a meal. Information on all of these outcomes will be collected during a baseline survey carried out between March and April 2023 and during an endline survey collected between January and February 2024. Both surveys will be collected from 5 women in each savings group, yielding a sample of 2,050 individuals.

2023-03-06

2023-09-30

- Insurance Uptake
- Business assets
- Total Savings
- Food insecurity

- Household Cash Income

While all of the women in our sample have already ‘graduated’ from the BOMA program, they continue to meet recurrently as savings group and keep saving and lending their savings as a group. The number of members within each savings group varies between 6 and 21, with the average being 15. We thus expect to have insurance purchase information for nearly 6,500 individuals. Information on business assets, income, total savings, and food security will be collected from 5 women in each business group using a survey instrument based on the one previously used by BOMA to track the progress of their program. We expect to collect information on these outcomes from 2,050 individuals.

Half of the savings groups will be randomly assigned into either the treatment or the control arm. This randomization will be stratified at the facilitator level, ensuring that each facilitator is responsible for roughly the same number of groups in each arm. There will be a total of 25 facilitators across 5 counties (Garissa; Isiolo; Marsabit; Samburu; Turkana). Each facilitator will be responsible for a varying number of groups according to geographical proximity and local language requirements. Facilitators will make 5 visits to treatment groups between March and September providing information on risk-mitigation practices. Control groups will be visited only twice: once in March at the beginning of the intervention and once in September once the sales window opens. For both treatment and control groups the baseline survey will be collected between March and April 2023, and an endline survey will be collected between January and February 2024.

Randomization done in office by a computer

Treatment will be randomized at the Savings Group level, stratified by facilitator.

Yes

410 Savings Groups

Insurance purchase decisions - 6,500 observations: all members of 410 savings groups Assets, Income, Savings, Food Security - 2,050 observations: 5 surveyed members for each of the 410 groups

Treatment arm: Around 3,250 group members out of which1,025 will be surveyed, distributed across 205 clusters (savings groups).

Control arm: Around 3,250 group members out of which1,025 will be surveyed, distributed across 205 clusters (savings groups).

We conduct power calculations using parameters estimated from data collected during a previous research intervention studying pastoralist households in Samburu county during the 2018-2021 period. We compute the mean and variance of all of our primary main outcomes (insurance purchases; assets; income; savings; food security) from this dataset, as well as the intra-cluster correlation coefficients calculated at the manyatta level (in our current study, there is almost always a single savings group per manyatta). Based on historical data on drought occurrences across the 5 pastoralist counties of our study, we spatially distributed our sample in such a way to maximize the probability of observing a drought shock in at least some of our clusters. We assume that roughly one third of the clusters in our sample will have a drought shock during the September-January 2023 rainy season, and therefore adjusted our sample size to have adequate statistical power so that we can explore the impact of the treatment on mitigating the impact of shocks. Based on the parameters obtained from the prior 2018-2021 Samburu-county intervention, and under the assumption that each treatment cluster (i.e. each savings group) is composed of 15 members, our study is powered to detect a 44% increase in insurance uptake (a 2 percentage-point increase above the observed 4.5% baseline uptake rate) with power of 0.8. For our remaining primary outcomes, we are assuming that out of the 410 treatment clusters in our sample, roughly 136 will have a drought shock during the September-January 2023 rainy season. Assuming as well that insurance uptake rates reach 6.5%, and taking this figure as our first-stage treatment compliance rate, our study is powered to detect (with a power of 0.8), a 3.5% increase in business income; a 4% increase in business assets; a 6.2% increase in total savings; a 2.2% increase in the household’s dietary diversity score, and a 6% impact on the number of days in which the family’s children skipped dinner.