There are two interventions. First, the NGO GiveDirectly is providing cash transfers to households. The value of this transfer is experimentally selected (at the village level) to be $250, $500, or $750 per household. These sums are paid out in monthly $250 increments, so that villages receive 1, 2, or 3 payments. There are several additional sub-randomizations. In Liberia, respondents are randomly chosen to be paid out quarterly, while others are paid over consecutive months. In Malawi, monogamously married households were randomly selected for either the male or female head of households to receive the transfer. Transfers are made via mobile money; since pre-existing mobile money usage is low, beneficiaries are given the option to buy cell phones (with the cost deducted from later transfers) and are given help in signing up for mobile money.

Second, a market access intervention is cross-randomized. Respondents are given a voucher to defray transportation costs to access inputs. In Liberia, agricultural input dealers set up stalls at local market centers where inputs were made available. In Malawi, a major agricultural input dealer set up events at prominent locations (such as primary schools). Inputs were not subsidized.

2019-01-02

2020-12-31

Food Security, Expenditures, Agriculture, Income, Labor Supply, Assets, Health Behavior, Child Anthropometrics, Household Resilience, Interpersonal Transfers, Intimate Partner Violence

The full list of outcomes will be examined to estimate the impact of cash transfers. For the market access evaluation, attention will be on agricultural outcomes, particularly the adoption of inputs such as fertilizer and improved seeds. We plan to write two separate pre-analysis plans for the two analyses.

Prices of food items

We are collecting data on the prices of food items in approximately 100 markets in Liberia and 95 markets in Malawi. In Liberia, approximately half of these are in areas that received cash transfers, and the other half are in areas that did not. In Malawi, about one quarter are in areas that received the transfer. We plan to conduct a difference-in-difference analysis to evaluate the effect of cash transfers on prices.

The experiment includes 600 total villages (300 in each country). The experiment will be cross-cut so that in each country 100 villages receive cash transfers, 100 serve as control, 50 receive market access only, and 50 receive both market access and cash. We plan to enroll a sample of 10 households per village, so that our target sample is 6,000 total households.

There are several sub-treatments within the cash transfer treatment arm. First, the value of the transfers was randomly selected to be $250, $500, or $750 per household. These sums are paid out in monthly $250 increments, so that villages receive 1, 2, or 3 payments. In addition, in Liberia, respondents are randomized into being paid this amount quarterly (i.e. spread out every 3 months for a year), or in 3 consecutive months. Finally, in Malawi, GiveDirectly randomized whether the male or female head of the household received the cash transfer (conditional on both spouses being home at the time of enrollment).

The randomization is stratified by administrative units known as “clan” in Liberia and “group village” in Malawi. Treatment intensity is randomly varied across group villages in Malawi and across geographic clusters in Liberia.

Randomization by a computer, in Stata.

Village

Yes

600 villages (300 in Liberia and 300 in Malawi)

6,000 households (3,000 in Liberia and 3,000 in Malawi).

(1) 200 villages receive cash transfers only (100 in each country)
(2) 200 villages receive nothing (pure control - 100 in each country)
(3) 100 villages receive market access only (50 in each country)
(4) 100 villages receive cash and market access (50 in each country)

Pre analysis plan October 2020