Loss Aversion and Agricultural Investment: Experimental Evidence from Smallholder Farmers in Kenya

How to cite

Levine, Madison; Janzen, Sarah; Kramer, Berber; Shikuku, Kelvin. 2025. "Loss Aversion and Agricultural Investment: Experimental Evidence from Smallholder Farmers in Kenya." AEA RCT Registry. AEARCTR-0017074. https://doi.org/10.1257/rct.17074-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Madison Levine* , Sarah Janzen , Berber Kramer , Kelvin Shikuku

* Primary investigator

Registry ID
AEARCTR-0017074
DOI
https://doi.org/10.1257/rct.17074-1.0
First registered
Last updated
Status
On going
Trial dates
2025-05-19 – 2026-05-19
Countries
Kenya (Africa)
Registry record
http://www.socialscienceregistry.org/trials/17074

Abstract

This study examines how loss aversion influences smallholder farmers' investment decisions in the presence of innovative tools such as loans, index insurance, and emergency credit. Using a lab-in-the-field experiment with 210 farmers in two of Kenya’s semi-arid counties, we investigate how behavioral biases influence investment in productivity-enhancing inputs with and without the availability of risk-mitigating products. We hypothesize that farmers perceive upfront costs as certain losses, discouraging adoption even when long-term benefits outweigh costs. Additionally, we assess whether emergency credit, which removes upfront payment requirements, mitigates this reluctance and encourages investment. Participants engage in experimental games under risk and uncertainty, and make repeated agricultural investment choices across varying financial and climatic conditions. We also elicit behavioral parameters of risk aversion, loss aversion, and inverse probability weighting. This approach allows us to observe how different innovative tools and preferences influence decision-making.

Intervention

The intervention consists of a lab-in-the-field experiment designed to study how behavioral preferences, particularly loss aversion, shape smallholder farmers’ agricultural investment decisions under risk.
Participants play two digital, incentive-compatible games:

1. Financial Investment Game (FIG) – a multi-round simulation that mimics real-world farming. Farmers decide how much to invest in fertilizer and whether to adopt three risk-management tools: drought-tolerant (DT) seeds, index insurance, and emergency credit (EC). Early rounds offer these tools for free; later rounds use a Becker–DeGroot–Marschak (BDM) auction to elicit willingness-to-pay.

2. Multiple Price Listing (MPL) Risk and Loss Aversion Game – a series of incentivized lotteries used to estimate each farmer’s risk- and loss-aversion parameters recovered through maximum likelihood estimations.

The intervention combines experimental exposure to financial tools with elicitation of behavioral parameters in a controlled but realistic environment.

Experimental design

This lab-in-the-field experiment examines how exposure to different risk-management tools affects farmers’ fertilizer investment decisions and how these effects interact with individual loss aversion. Each participant plays both the Financial Investment Game and the MPL lotteries task in a tablet-based setting. Modules introducing DT seeds, insurance, and EC are randomly ordered within sessions to prevent learning biases. Each session includes 12–16 farmers across 17 villages drawn from Machakos and Makueni counties in Kenya.

Design details

Randomization method
From this study population, eligible smallholder farmers were randomly selected using a proportional sampling method across Machakos and Makueni counties. Within the experiment, randomization occurs entirely within the digital game application. Each participant receives: 1. A randomized order of financial-tool modules (drought-tolerant seeds, index insurance, emergency credit) to prevent ordering effects. 2. A randomized strike price in the BDM auctions (either a 25% or 50% subsidy). 3. Randomized weather outcomes drawn by the software according to fixed probabilities explained to participants in advance. All randomization is performed automatically by computer using the built-in random number generator in the experimental software, ensuring transparency and independence from enumerators.
Randomization unit
Individual
Clusters
Not applicable (no clustering). However, the study sample consists of approximately 250 individual farmers, organized into sessions of 12–16 participants across ~17 sessions. Each farmer makes repeated decisions across multiple rounds.
Observations
The design is within-subject, meaning every participant encounters all three tool treatments (DT seeds, index insurance, and EC) in randomized order. Target sample size is 250.
Arms
Because treatment assignment occurs within the game rather than across groups, there are no separate arms in the traditional sense. All 250 participants experience all three financial tools in randomized sequence, and therefore contributes multiple observations across all treatment conditions.
Primary outcomes
1. Fertilizer investment per round – total amount invested (continuous outcome) 2. Tool uptake – binary indicator of whether the participant adopted drought tolerant seeds (DT), insurance, or emergency credit (EC). 3. Willingness to pay (WTP) – stated maximum bid for each tool via BDM mechanism. 4. Loss aversion and risk aversion parameters – elicited through the MPL lotteries.

Cited by

No citation data available for this registration yet.

Keywords