Agricultural insurance and contract farming

Last registered on April 15, 2018


Trial Information

General Information

Agricultural insurance and contract farming
Initial registration date
August 24, 2014

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
August 24, 2014, 8:54 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
April 15, 2018, 5:07 PM EDT

Last updated is the most recent time when changes to the trial's registration were published.



Primary Investigator

Columbia University

Other Primary Investigator(s)

PI Affiliation
Stanford University

Additional Trial Information

Start date
End date
Secondary IDs
Agricultural production entails large risks from crop failure which farmers living at subsistence levels are ill-suited to bear. Attempts to reduce these risks through insurance contracts have typically been unsuccessful due to low take-up rates. In this project, we work with one of the largest private sector contract farming schemes in East Africa, investigating what constrains the take-up rate and testing innovative insurance schemes for contract farming settings. This project will consider the role of inter-temporal distortions in constraining take-up.

The theoretical justification for insurance is as a transfer across states, not across time: the welfare gain arises through shifting consumption from good states, where the marginal utility of consumption is low, to bad states, where it is high. However, standard insurance contracts work by the payment of a premium ex-ante for an ex-post state dependent payout. That is, real world insurance contracts combine state dependent payouts with saving, rather than with other state dependent payouts. Insurance shifts risk from those from whom it is more costly to those for whom it is less costly, but in its standard formulation these welfare improvements are partly offset by shifting saving from those for whom it is likely to be cheap to those for whom it is expensive.

The most likely reason for the standard ex-ante payment is enforcement. An insurance contract which is just a transfer across states with no ex-ante payment is unlikely to be incentive compatible: in the absence of enforceable contracts, as may be the case in many rural agricultural settings, in good states the insured would not pay the insurer. Contract farming offers a potential solution to this, thus allowing insurance to be offered without a transfer across time. Since the farmer and buyer enter a contract (typically enforced on the side of the farmer through the buyer having a monopsony) the buyer can provide inputs on credit to the farmer, with the cost reclaimed with interest from the harvest revenue of the farmer. Currently, the partner company offers several inputs (e.g., land preparation, fertilizer) to farmers in this way, enabling them to solve the challenge of increasing levels of fertilizer use in Kenya. Similarly the buyer can offer insurance with the premium paid on credit (or, equivalently, in the form of a flatter payment schedule).

The proposed research will investigate whether demand for insurance with ex-post premium is higher than traditional insurance and if so, which of the possible mechanisms / inter-temporal distortions are important in constraining demand in this setting. It will do so by randomly offering either ex-post premium or standard insurance to farmers and then observing the difference in take up rate.
External Link(s)

Registration Citation

Casaburi, Lorenzo and Jack Willis. 2018. "Agricultural insurance and contract farming." AEA RCT Registry. April 15.
Former Citation
Casaburi, Lorenzo and Jack Willis. 2018. "Agricultural insurance and contract farming." AEA RCT Registry. April 15.
Experimental Details


The main aspect of the treatment to be randomized is whether the insurance offered by the partner contract farming company is paid for up front or on credit.. The insurance on credit is identical to the standard insurance, except the cost (plus interest at the standard rate for other inputs) is agreed to be borne as a deduction from the farmer's revenue after harvest. We will study the impact of offering the insurance through deduction and compare this effect to the effect of a large reduction (about 30%) in premium when paying upfront.

The insurance offered has a double trigger design. Payment occurs only if both farmer's individual yield is low and the local average yield is low. The triggers are based upon an econometric model using historical data.

[POST-TRIAL AMENDMENT, 4/17/2018:] In addition to the main experiment, we also run two smaller experiments, with 120 farmers each. In the first, we cross cut the insurance on credit treatment with a cash drop. In the second, farmers chose between a cash payment and free enrollment in the insurance. One half of farmers received their choice immediately, whereas the other half of farmers received their choice in one month's time.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Our primary analysis will test the difference in take-up of insurance with upfront payment (at actuariall fair and discounted premium) vs. payment through deduction.

We will also look at heterogeneity of take-up by each of the following covariates.

- Rate of intertemporal substitution of income and time inconsistency of this rate.
- Wealth (land ownership, number of cows)
- Index of trust toward the company (index 1-4)
- Percentage difference between subjective expected yield and yield predicted by the model
- Yield of last cycle
- Subjective likelihood that the plot will be in debt at the end of the cycle.

If any of these variables presents a particularly skewed distribution, we will report the heterogeneity by the trimmed variable or by deciles

In addition, if the take-up in the upfront insurance with actuarially fair price and that with discounted premium are similar, we will bundle these two treatments to maximize power on the heterogeneity analysis.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
We study take-up of agricultural insurance in a sample of approximately 600 sugarcane contract farmers.

Plots are randomized in three treatment groups
1. Insurance with actuarially fair premium paid upfront
2. Insurance with premium equal to 70% of the actuarially fair value paid upfront
3. Insurance with actuarially fair premium (+interest) paid through deduction at the end of the cycle

[POST-TRIAL AMENDMENT, 4/17/2018:] In addition to the main experiment, we also ran two smaller experiments, with 120 farmers each. In the first small experiment (``cash drop experiment’’), we cross cut the insurance on credit treatment with a cash drop, worth approximately the value of the insurance premium. Each cross-cut treatment group comprised about 30 farmers. In the second small experiment, 120 farmers chose between a cash payment, equal to the insurance payment, and free enrollment in the insurance. One half of farmers were told they would receive their choice immediately, whereas the other half of farmers were told they would receive their choice in one month's time. Both small experiments had the same randomization method as the main experiment (described in the Experimental Design Details below).

Experimental Design Details
Randomization Method
Farmers invited to the meeting are assigned random numbers. Attended farmers are entered into an attendance list in order of this random number. A randomization sheet, pre-generated for each field, is used to assign treatment based on the attendance list.
Randomization Unit
Individual farmer
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
600 sugarcane plots
Sample size: planned number of observations
600 plots
Sample size (or number of clusters) by treatment arms
Approximately 200 plots in each of the three groups. If the take up rates for the upfront 100% and 70% of the actuarially fair treatments are similar in the beginning of the trial we may reduce their sample sizes slightly in order to increase the sample size of the deduction group.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
In pilot data, in the treatment group where the insurance is offered with upfront premium payment at an actuarially fair value, the take-up is around 10 percent (s.d. 0.35). With sample size of 200 plots per group (upfront actuarially fair vs. deduction payment vs. upfront discounted), we have a minimum detectable effect of around 10 percentage points. Note that these calculations do not take into account field fixed effects and therefore, as long as field have predictive power, are likely to be conservative At the time of registering the first version of this trial, we do not have reliable data on the power of the regressions by baseline covariates. This will depend on i) the distribution of the covariates in the population of interest; ii) the impact of the covariate in the cash group. We will report power calculations for these heterogeneous effects in an appendix and will only report the heterogeneities for which we have power in the main paper. The rest will be reported in the appendix.

Institutional Review Boards (IRBs)

IRB Name
Harvard University-Area Committee on the Use of Human Subjects
IRB Approval Date
IRB Approval Number
IRB Name
Stanford University Panel on Non-Medical Human Subjects
IRB Approval Date
IRB Approval Number
349 (Panel: 2)


Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials