Experimental Design
The goal of this study is to understand the impact of providing coupons for fruit and vegetable on the consumption behavior of the household, both at the individual level and overall. To do this we will conduct a randomized evaluation of the coupon program. Participants will be randomized, at the individual level, into three groups:
1. Control: Participants in the control group will not receive any intervention.
2. Small coupons: Participants in this treatment group will receive weekly coupons in a “small” amount. This amount will be determined to be below the average weekly consumption at baseline for the designated fruits and vegetables.
3. Large coupons: Participants in this treatment group will receive weekly coupons in a “large” amount. This amount will be determined to be above the average weekly consumption at baseline for the designated fruits and vegetables.
The small and large coupon sizes are designed to test the difference between an inframarginal coupon (one that is below their existing consumption level), and an extramarginal coupon (one that is above their existing consumption level). A voucher for an amount below what is normally spent means that an individual could use that voucher and purchase their normal amount of that item, and use the savings on other things. If they increase their consumption, it is because they were previously consuming less than they desired due to their budget constraint. If however they do not increase their consumption that indicates they are already consumed their preferred amount. In that case, the large extramarginal coupon would be needed to incentivize increased consumption. As that coupon is larger than their current consumption levels, they would need to increase consumption in order to not waste the voucher.
We are concerned that the non-divisibility of the coupons will limit their usefulness, people may shop for different fruits at different stands, or make multiple trips per week given that these are perishable items. As such we will pilot offering three smaller coupons or one larger coupon, and determine which we should include in the project implementation.
We seek to answer the following research questions:
• Does providing coupons (vouchers) for fruits and/or vegetables increase the consumption of the specific products on the voucher? Of fruits and/or vegetables in general?
• Does providing coupons crowd in additional expenditures on fruits and vegetables?
• Does providing coupons affect other consumption patterns?
• How much does the size of the voucher matter?
We will examine these questions both during the intervention itself, and 2-3 months following the conclusion of the intervention to assess whether the intervention has lasting impacts on the following outcomes.
Outcomes
We hypothesize that by providing coupons for specific fruits and vegetables, households will purchase and consume more of those specific fruits and vegetables. Consumption of targeted items will be our primary outcome. We further hypothesize that the coupons will further increase consumption of fruits and vegetables generally; in other words, substitution would occur only from other types of food, or any substitution that would occur away from other fruits and vegetables would be dominated by the positive effect on demand of the coupons.
To address multiple hypothesis testing we will create an index for each of the three consumption categories (targeted outcomes, fruits, and vegetables), and also calculate sharpened q-values within each family (Anderson 2008). We will additionally calculate sharpened q-values across the dietary diversity outcomes.
We will also examine several exploratory outcomes. These are intended to track understanding of how the vouchers affected other types of behavior related to consumption. Because these are exploratory we will not apply any multiple hypothesis testing.
Randomization into the treatment groups will be conducted at the household level in each country. Randomization will be stratified by the neighborhood and the median level of baseline fruit consumption (ie households will be split into two groups based on sample median baseline fruit consumption). We will randomize the households into the control group, the small voucher group, and the large voucher group, with equal numbers of households per treatment arm. That randomization will be stratified on neighborhood and median level of baseline fruit consumption.
Potential competing studies are the other intervention elements of the overall project. These will be implemented in the same neighborhoods are our study. We will rely on the randomization to ensure that we are able to calculate an unbiased estimate of the impact of the coupons. If possible, we will control for exposure to the other interventions.
Data collection
The evaluation will be conducted using several sources of data. Baseline data will come from the household survey which has already been completed. Evaluation of the program’s impact will use three sources of follow-up data.
Monitoring surveys
The project plans to conduct short monitoring surveys with a rotating cross section of the baseline sample. While details are still being finalized, current plans are to interview approximately 50 households per month for 12 months. The 12 month period will coincide with the 12 month implementation period of WP2a and WP2c. Ideally the 6 month implementation of the coupon intervention will be timed so that the monitoring data covers a couple of months prior to the intervention beginning, the period of the intervention, and several months after. These monitoring surveys are designed to proxy 24 hour recall data, collecting data on which food types were consumed in the previous 24 hours, without a full accounting of all values and quantities. It will also contain some limited data regarding consumption in the previous 7 days. Given small sample sizes, we anticipate using this as a complementary (but not primary) data source for our evaluation.
Endline survey 1
The first endline survey, designed as a monitoring survey for WP2b, will take place in the final month of the coupon intervention implementation, while households are still receiving coupons. It will allow us to measure the impacts of the intervention during implementation.
Endline survey 2
The second endline survey, is the project endline survey. Implementation of the coupon intervention will be designed so that this survey is conducted 2 to 3 months after the conclusion of the intervention, and will allow us to measure the persistence of impacts.