Evaluating the impact of recipient choice in aid provision

Programs included in the study are:

1. Agricultural extension: we hired a team of 11 agriculture experts, with a combined experience of 66 years in the agricultural sector, to deliver in-person group training to randomly selected farming households. The training sessions ran from September to October 2016 - leading up to the “short rains” agricultural seasons in Kenya. The training includes education on: land preparation, planting, soil fertility, crop selection, soil and water management, field management (fertilization, pest and disease management, weeding), record keeping and financial management, farmer group dynamics and con conflict resolution, harvesting, post-harvest management, value addition and marketing.

2. Agricultural inputs: based on the advice of agricultural experts, we provided recipients with enough inputs to plant approximately 0.5 acres of cabbages or maize. The type of inputs to be provided were recommended by our agricultural consultants who determined the requirements based on terrain and crops grown in the study areas. Specifically, for the cabbage-growing region of Mbooni, we provided 50 grams of Baraka F1 seeds and 75 kilograms of planting fertilizer. For the maize-growing region of Kilungu, we provided 4 kg of Duma 43 seeds, 25 kg of planting fertilizer and 25 kg of top-dressing fertilizer. These inputs are roughly modeled after the Government of Kenya's National Accelerated Agriculture Inputs Access Program. The program includes a voucher, valued at USD 60 - 80, to cover the cost of 10 kg of hybrid maize seed, 50 kg of basal fertilizer, and 50 kg of top-dressing fertilizer, inputs sufficient for approximately 1 acre (0.4 ha) of maize. This study provides similar, though not precisely the same, package. Our agricultural inputs package also included a one-time information session on proper input usage provided by our extension agents.

3. Livestock transfers: recipients received 25 one-month old chicks vaccinated for common diseases as well as a starter pack of feed (~10 kg). Recipients were also provided with basic information about taking care of their chicks by our team of agricultural experts and were visited occasionally by the agriculture team over the following 4 months.

4. Cash transfers: some households were randomly selected to receive direct cash transfers. The size of these transfers match the per-recipient cost of one of the above programs $15 for agricultural extension, $75 for the 319 agricultural input recipients in Mbooni and $35 for the 181 agricultural input recipients in Kilungu, and $120 for livestock transfers. Cash transfers were delivered using the MPesa mobile money platform.

2016-09-01

2016-12-15

1. Consumption
2. Food security
3. Assets
4. Psychological well-being
5. Autonomy, dignity, trust

**Note: prior to launch, most questions from the GHQ scale were removed (to reduce survey length). The remaining questions include:

1. Have you recently felt that you were playing a useful part in things?
2. Have you recently felt capable of making decisions about things?
3. Have you recently felt that you couldn't overcome your difficulties?

These will be used in lieu of the full scale in analysis.

1. Consumption
2. Food security
3. Assets
4. Psychological well-being
5. Autonomy, dignity, trust

We randomized 1,000 respondents into the “extension or cash” group, 1,000 into the “inputs or cash” group and 1,000 into the “livestock or cash” group. Within each group, we randomized individuals to receive either the program or a cash transfer equal to the cost of the program.

Randomization is conducted at the individual level. Though the informational components may have spillover effects, individual randomization is a deliberate choice: the primary goal of this study is to compare across cash or program arms in order to isolate the effect of the choice mechanism. In equilibrium, were recipients to be given a choice between programs and cash transfers, we expect some would choose the program, thus having a mix of those receiving cash and the program in the same village provides the most relevant comparison. Further, we chose a valuation based approach to estimating whether a respondent prefers cash or the program, as opposed to a direct choice between the two. This choice was made as a valuation approach could potentially be extended to multiple interventions of various costs (if preferences matter, it may be wise to provide the program with the highest ratio of valuation to cost).

Computer

Individual

No

3,000 individuals

3,000

1,000

Based on a sample of 3,000 (those receiving either a program or cash) and 4, we can detect an effect size of 0.09 standard deviations with 80% power. For reference, based on data from (Haushofer & Shapiro, 2016), a 0.09 SD change corresponds to a 7% change in the value of assets and a 5% change in consumption.

Impact_of_recipient_choice_20170208-2.pdf