NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Experimental games to teach farmers about weather index insurance
Last registered on September 05, 2017


Trial Information
General Information
Experimental games to teach farmers about weather index insurance
Initial registration date
August 29, 2017
Last updated
September 05, 2017 1:52 PM EDT
Primary Investigator
University of Georgia
Other Primary Investigator(s)
PI Affiliation
World Agroforestry Center
PI Affiliation
World Agroforestry Center
PI Affiliation
University of Florida
PI Affiliation
Montana State University
Additional Trial Information
On going
Start date
End date
Secondary IDs
In this study we evaluate demand for weather index insurance (WII) among smallholder producers in Kenya’s arid and semi-arid lands (ASALs), focusing on the role of basis risk. We estimate demand for two different products exhibiting differences in basis risk (i.e. mismatch between WII payouts and actual losses) and test an innovative “experiential” approach to teach producers about basis risk and WII as a risk management tool. Our novel extension approach consists of an insurance game based on the one developed by Cai and Song (2017) and modified for weather index insurance, clearly illustrating basis risk to participants.

Unlike many agricultural innovations, learning about insurance products and other risk reducing technologies can take a long time. If shocks that result in payouts are infrequent, individuals who purchase insurance may not see that it pays out in bad years until a bad year occurs. Further- more, for risks that are highly covariate, such as drought, one might not be able to readily learn from their peers. Thus, providing the opportunity for farmers to rapidly experience different out- comes with and without insurance could be an effective way to educate and increase demand (Cai and Song, 2017). Learning about index insurance (as opposed to indemnity insurance) is further complicated because there is a wider range of outcomes. With indemnity insurance, a farmer will be compensated an amount corresponding to their loss.

An aggregator working in Tharaka South sub-county of Tharaka Nithi county currently sup- plies farmers (“program farmers”) with a bundle of inputs to grow sorghum and/or green gram with deferred payment. Included in the input package is mandatory WII for the value of inputs. This “status quo” insurance contract was developed by Acre Africa using ARC2 data in 2013. In our study, 457 program farmers and 30 additional farmers are given the chance to buy additional insurance. More specifically, all farmers in our study will have the option to buy top-up WII for the value of sorghum and/or green gram production. Green gram insurance is only offered to farmers who indicate that they will not be planting sorghum during the upcoming season. For program farmers, WII is purchased in addition to mandatory input insurance. This top-up WII product was developed by Acre Africa specifically for this study using the CHIRPS dataset (Funk et al., 2015).

During the experiment, demand for two unique insurance products will be compared. These products differ in the amount of basis risk they present, holding all else constant. CHIRPS data can be downloaded at a resolution of 5x5 km; the true product uses this level of resolution for the index area. We call this high resolution (HR) index insurance. A second product developed by Acre Africa averages the high resolution CHIRPS data to create a broader index area (10x10 km). We call this low resolution (LR) index insurance.

We employ a 2x2 randomized control trial in which farmers are randomly assigned a contract type (HR or LR), and are randomly exposed to either a basic information treatment or the basic treatment plus the insurance game. There are two versions of the game, one calibrated for each insurance product. The primary outcome of interest is demand for WII, which we measure in two ways: (1) willingness to pay (WTP) across a variety of prices elicited using a multiple price list auction (Anderson et al., 2007), which is a modified version of a Becker-DeGroot-Marschack (BDM) auction (Becker, DeGroot, and Marschak, 1964), and (2) actual purchases at the offered price. We also test farmer understanding of and attitude towards weather index insurance and basis risk at the time insurance is offered. Following treatment, all farmers will have the opportunity to purchase the HR insurance product, and receive randomized discounts through the auction to do so.

We make two primary contributions. First, we examine whether farmers are sensitive to basis risk by comparing demand for two insurance products that differ only by resolution. Basis risk is considered to be one of the greatest barriers to index insurance adoption (Carter et al., 2014), yet relatively little is known about how sensitive farmer demand is to it.3 We believe this is an important question given continual efforts to improve the resolution of index insurance products. If such improvements do not increase farmer demand it would suggest that commercially-viable improvements in basis risk might not induce higher uptake of WII on their own. Previous studies have used distance from weather stations as a proxy for basis risk, and in some cases found that demand is quite sensitive it, particularly when the price is high (Hill, Hoddinott, and Kumar, 2013). However, it is possible that distance from weather stations is correlated with unobservables that affect demand. Our design offers a transparent way to measure sensitivity to basis risk, albeit at only two different levels.

Our second contribution is to evaluate the use of insurance games as an extension tool in the promotion of WII, and in particular, to help farmers understand basis risk. We will analyze how a game focused on basis risk alters demand for insurance, and how individual or peer experiences in the game affect demand for insurance. Games have been used to study insurance demand in other contexts, including projects in Kenya (Janzen and Carter, 2013), Ethiopia (Norton et al., 2014), and Peru (Boucher and Mullally, 2010). Elabed and Carter (2015) use a compound lottery to estimate the effect of basis risk on insurance demand, but in the hypothetical. Cai and Song (2017) found that playing insurance games in China led to a 48% increase in yield insurance uptake (an increase of 9 percentage points), and that random shocks experienced within the games also increased insurance demand. Like Cai and Song (2017), we incorporate games into extension rather than using them solely as a research tool. Our games differ from Cai and Song (2017) in several important ways. First, our games are incentive compatible (payouts are based on game results) in an attempt to increase salience. Second, our games simulate WII, as opposed to standard indemnity insurance, and highlight the role of basis risk.

External Link(s)
Registration Citation
Hughes, Karl et al. 2017. "Experimental games to teach farmers about weather index insurance." AEA RCT Registry. September 05. https://doi.org/10.1257/rct.2401-2.0.
Former Citation
Hughes, Karl et al. 2017. "Experimental games to teach farmers about weather index insurance." AEA RCT Registry. September 05. http://www.socialscienceregistry.org/trials/2401/history/21197.
Sponsors & Partners

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

Request Information
Experimental Details
Please see attached PDF.
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
Please see attached PDF.
Primary Outcomes (explanation)
Please see attached PDF.
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
Please see attached PDF.
Experimental Design Details
Please see attached PDF.
Randomization Method
Please see attached PDF.
Randomization Unit
Please see attached PDF. One treatment was clustered at individual level, one at group level.
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
Please see attached PDF.
Sample size: planned number of observations
Please see attached PDF.
Sample size (or number of clusters) by treatment arms
Please see attached PDF.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Please see attached PDF.
IRB Name
Montana State University
IRB Approval Date
IRB Approval Number
Analysis Plan

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

Request Information
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports, Papers & Other Materials
Relevant Paper(s)