Crop insurance can be an important tool to help reduce smallholder farmers’ risk from weather-related shocks. However, many smallholders are not able to access crop insurance because their farms are too small and too remote for insurers to affordably verify crop damage. A new project, conducted by IFPRI in partnership with the Borlaug Institute for South Asia (BISA), and HDFC Ergo General Insurance, Ltd and funded by the CGIAR Program on Policies, Institutions and Markets and International Initiative for Impact Evaluation (3ie), examines whether these constraints can be overcome using pictures taken on farmers’ own smartphones.
The Picture-Based Insurance (PBI) project aims to deliver affordable, easy-to-use crop insurance. A formative evaluation of the project was conducted during the Rabi (winter) season of 2016 and 2017 in six districts in Haryana and Punjab, India. A total of 592 randomly selected farmers participated in the study; all participating farmers met the following criteria: (1) having less than 15 acres of operational farmland and (2) planning to grow at least two acres of wheat during the 2016-2017 Rabi season. Farmers received crop insurance covering one acre of their wheat crop, and insurance was conditional on taking regular pictures of their plots from sowing to harvest using WheatCam, a smartphone app developed for the project. The insurance included coverage against damage visible in the WheatCam pictures (PBI) and coverage against excess rainfall and above-normal temperatures between February and April (weather-based insurance, or WBI). At the end of the season, experienced local agronomists examine each farmer’s picture series to verify crop damage and losses.
The project team has produced two project notes discussing the results of this formative evaluation, one discussing the feasibility of the program and the other discussing the sustainability of the program.
In terms of feasibility, the first question the research team examined was whether farmers were willing and able to take regular, high-quality pictures of their plots. Of the 592 farmers surveyed, 80 percent uploaded at least one valid picture during the Rabi season; of the farmers who uploaded pictures, more than 83 percent took more than six pictures during the season. Overall, the amount and quality of the pictures uploaded suggests that farmers are generally willing and able to take pictures for loss assessment purposes.
The second question that the team looked at in terms of feasibility was whether damage could be accurately assessed using farmers’ pictures. Both farmers and experts agreed that important causes of crop damage – including lodging (bending of the plant due to winds or wet, loose soil), hail, and common wheat diseases like yellow rust – would indeed be visible in photographs. These perceptions were supported by the fact that the majority of local agronomists agreed on the level of crop damage actually seen in the pictures and that yields for farmers assessed as having lower levels of damage were in fact higher than yields for farmers assessed with high levels of damage (20 quintals per acre compared to 10 quintals).
Finally, the feasibility study compared the PBI program to weather-based index insurance programs and found that the payouts for PBI are better correlated with farmers’ actual yields than the payouts for the WBI programs. According to the authors, this indicates that PBI can reduce the basis risk that is characteristic in many standard index-based insurance programs.
The authors highlighted the fact that automating loss verification through image processing algorithms will be more cost-effective in the long run and that bundling the PBI product with agro-advisory or pest management services could further improve farmers’ participation in the program.
In the second project note, the research team examined the sustainability of the PBI program. Unlike weather index-based insurance, many indemnity or yield-based insurance programs face challenges of adverse selection (farmers enrolling plots more prone to damage) and moral hazard (farmers reducing effort on their plots once insured). These challenges can increase insurance premiums, making them unaffordable and significantly reducing demand.
To test for moral hazard, more than 700 farmers from 50 villages were randomly assigned to one of two treatment arms: the control group, in which farmers who took crop pictures received weather index-based insurance that covered against excess rainfall and above-normal temperatures, and the treatment group, in which farmers who took pictures received both the WBI product and the PBI product.
The results show that use of fertilizers, pesticides, fungicides, herbicides, and farm labor was virtually the same for the two groups, indicating that farmers did not put forth less effort into crop management when covered by the PBI product. In fact, during focus group discussions, farmers reported that receiving the product actually helped them improve their management practices - since they had to visit their plots more often to take photos, they were better able to monitor the crops. In addition, the study found that PBI coverage did not appear to result in lower yields or higher damage, as would be the case if moral hazard were present.
To test for adverse selection, the team elicited willingness-to-pay from a sub-sample of 100 farmers for four products: WBI only, WBI plus pictures (which paid out only if farmers took and uploaded regular pictures), WBI plus PBI, and PBI only. The farmers were able to choose which plot they insured, which allowed the research team to determine whether farmers selectively enrolled plots that were more prone to damage.
The willingness-to-pay results show that demand for PBI was stronger than demand for WBI and that having to take pictures regularly did not significantly reduce farmers’ willingness to pay for the PBI product. Farmers also expected to receive higher payouts and more complete insurance from a product that includes PBI. The results regarding adverse selection show that farmers’ willingness-to-pay for the extra PBI coverage did not decrease based on yields and also was not significantly higher for farmers with visible crop damage in the previous season. This suggests that there was no adverse selection present in farmers’ willingness-to-pay. In terms of whether farmers chose to enroll plots that were more susceptible to damage, the study found no quality differences between the plots selected for PBI insurance and those not selected for PBI insurance. Plot quality was measured through a variety of indicators, including distance to an irrigation source, previous year’s yields, the plot’s sales and rental value, drainage, and soil fertility.
The authors do point out, however, that while adverse selection and moral hazard do not currently appear to be a factor in farmers’ decisions to utilize the PBI program, increased knowledge of the program’s benefits could cause these factors to arise in the long term. Thus, continued monitoring and evaluation of the program is needed.
Overall, the results show that Picture-Based Insurance is both a feasible and a sustainable way to provide affordable crop insurance to poor smallholder farmers. Moving forward, the project team plans to automate loss assessments using crop modeling and machine learning, as well as to use the data collected from farmers’ pictures to provide agro-advisory and weather monitoring services. The project will also be expanded to additional crops, specifically those with higher production risk than wheat.
By: Berber Kramer and Sara Gustafson, IFPRI