Primary Outcomes (end points)
The primary outcomes are measured for the tenant.
Measured at the beginning of the season:
Experimental Investment Decisions for Tenants:
1) Crop maturity preference: A binary choice between receiving cassava stalks or corn seeds. Both serve as staple crops and close substitutes; however, cassava has a longer maturity period and is generally preferred for consumption by the local population.
2) Investment duration choice: The selection among different investment durations, where the return is conditional on continued tenancy (e.g., an investment of 1,000 USh yields 2,000 USh in 2 months, 4,000 USh in 4 months, 8,000 USh in 8 months, or 10,000 USh in 10 months)
Observed Outcomes:
3a) Perceived tenancy security (index): An index constructed from self-reported responses of tenants on a Likert scale that captures:
- The perceived likelihood that the contract terminates earlier than agreed;
- The risk of eviction before harvest;
- The perceived likelihood that the landlord restricts access to the agreed-upon land area;
- Concerns regarding food access for the respondent and their household.
Measured at the end of the season:
3b) Tenancy security
i. Perceived tenancy security: reassessment using the same measure as at baseline.
ii. Actual tenancy security: dummies denoting occurrences of early eviction, whether before the agreement’s end or harvest.
iii. Contractual changes: incidences where modifications to the rental agreement (in terms of payment, land area, or conditions on land use) occurred against the tenant’s preference during the season.
4) Actual investment
a. Labor investment: the self-reported number of hours per week spent on the land on pre-harvest activities by the tenant, household members, and any hired labor during the past season.
5) Landlord-tenant interaction index
6) Food security index: Measured for both the tenant and their household
For all index variables, we aggregate self-reported Likert scale responses by computing a standardized, inverse covariance–weighted average of the indicators.
Analysis
We will estimate the impact of the treatment on the different outcomes using linear regression models. Our main specification will control for the baseline value of the outcome and include strata dummies while clustering the standard errors at the village level.
Other important control variables include the duration of the tenancy, the rented land area, and the languages spoken by the tenant. Ultimately, we will use the Post-double selection Lasso (PDS Lasso) method to select which control variables to include in our final regression.
We will estimate the Treatment Effect on the Treated (TOT) by using the treatment assignment dummy as an instrument for receiving a formal contract.
We will winsorize the continuous variables, including investment measures, tenant harvest value, and income for both landlords and tenants, to mitigate the influence of outliers.
We will conduct both multiple and single hypothesis testing in our analysis.
We will explore heterogeneous effects by gender, ethnicity, total rented land area, and the duration of the relationship between the landlord and tenant.