We will partner with a cotton outgrown farming company in Zambia to test a program that allows farmers to take an advance on their harvest earnings during the lean season. We will measure impacts on company profits, on recipient labor allocation and income, and on spillovers to ineligible households via prices and wages. The study is designed to test the scaling potential of findings from Fink et al. (2018) and investigate the general equilibrium effects of lowering lean season interest rates for selected households in a labor market.
Intervention
Our study partner, Alliance Ginneries, provides farming inputs loans (cotton seeds and chemicals) at the beginning of the growing season in October through December, and purchases crops back from farmers starting in June, deducting the value of inputs from harvest earnings. The intervention offers farmers the opportunity to take an advance of 200 Kwacha on their harvest earnings in January, at the start of the lean season. Repayment, plus an additional transaction fee of 30 Kwacha, will be deducted from the farmer’s cotton sales earnings.
Experimental design
For the 2018-2019 season, Alliance is operating out of 62 area sheds. Alliance determines area boundaries or designations every season based on yields in the previous season or logistical drivers. A sample of 42 areas across the 6 regions were eligible for the program. 24 were randomly selected for the program, using equal strata size (N=4) across Alliance's six operational regions. Within each of the 24 selected area sheds (referred to as “areas” in the remainder of this text), either 2 or 6 distributors (out of an average of about 20 distributors per area) were randomly chosen for the advance payment program.
The design is thus nested with areas assigned to treatment and control, and distributors within treated areas assigned to treatment or control. All farmers associated with a treated distributor are eligible for treatment.
Design details
Randomization method
Area selection was based on random draws using equal strata (N=4) for each of Alliance's six regions of operation.
Cluster selection (randomization at the distributor level) occurred in real-time at the distributor trainings in September and October 2018. Clear and colored marbles (2 or 6 colored marbles for low and high intensity areas, respectively) were put into a bag and distributors that drew a colored marble were assigned to treatment.
All farmers associated with a treated distributor are eligible for the advance payment. To aid distributors’ compliance and understanding of the program, all distributors (including control group distributors) in treatment areas will get the advanced payment option (but will be excluded from the analysis, which will focus on “regular” farmers).
Randomization unit
The study design is randomized at 2 levels.
1) Areas (small regions) are randomized into treatment and control. Treatment areas are either low intensity (2 treated distributors) or high intensity (6 treated distributors).
2) Within treatment areas, distributors (a lead farmer responsible for signing contracts with farmers, delivering training and inputs) are further randomized into another layer of treatment and control, with the number of selected distributors determined by the treatment intensity.
Clusters
A total of 96 distributors were assigned to treatment. Each distributor works with 15-20 farmers, though the final number of farmers associated with the treated distributors will not be known until Alliance's books are updated for the year.
Control distributors for the analysis using administrative data can be drawn from any of the roughly 750 distributors in eligible treatment areas not assigned to treatment (either because they are in a control area or they are control distributors in treatment areas). See the pre-analysis plan for further details.
Observations
The target population for this study are small-scale farmers in Zambia. A total of at least 2,500 farmers in Alliance’s network will be offered the advance payment option during the 2018/19 agricultural season.
The lean season surveys will collect data on around 1000 households, drawn from both the network of eligible (Alliance) farmers and ineligible (non-Alliance) farmers, across treatment and control.
Arms
24 areas assigned to treatment, 18 to control
6 or 2 distributors per treatment area assigned to treatment (high and low intensity arms)
15-20 farmers per distributor, with farmer treatment status matching distributor treatment status
Primary outcomes
Our primary outcomes are impacts on the partner company's profits, based on administrative data:
(i) harvest sold to Alliance in kilograms (valued at the price Alliance pays its farmers),
(ii) total liabilities (i.e. value of all loans including APP),
(iii) payments to farmers (harvest value minus farmer liabilities),
(iv) the percentage of all loans (including the APP) repaid,
(v) the percentage of input loans (excluding the APP) repaid,
(vi) a measure of Alliance's profit that includes operational costs of the APP such as foregone interest and an estimated margin per kilogram of cotton (conditional on receiving the necessary information from the partner).
Secondary outcomes
Secondary outcomes will measure impacts on untreated households and market equilibria, using survey data:
(a) hungry season consumption,
(b) hungry season labor allocation,
(c) food prices,
(d) wages
Cited by
No citation data available for this registration yet.