Can We Boost Demand for Rainfall Insurance in Developing Countries?


This page in:

Ask small farmers in semiarid areas of Africa or India about the most important risk they face and they will tell you that it is drought. In 2003 an Indian insurance company and World Bank experts designed a potential hedging instrument for this type of risk—an insurance contract that pays off on the basis of the rainfall recorded at a local weather station.

The idea of using an index (in this case rainfall) to proxy for losses is not new. In the 1940s Harold Halcrow, then a PhD student at the University of Chicago, wrote his thesis on the use of area yield to insure against crop yield losses. In the past two decades the market to hedge against weather risk has grown, especially in developed economies: citrus farmers can insure against frost, gas companies against warm winters, ski resorts against lack of snow, and couples against rain on their wedding day.

Rainfall insurance in developing countries is typically sold commercially before the start of the growing season in unit sizes as small as $1. To qualify for a payout, there is no need to file a claim: policyholders automatically qualify if the accumulated rainfall by a certain date is below a certain threshold. Figure 1 shows an example of a payout schedule for an insurance policy against drought, with accumulated rainfall on the x-axis and payouts on the y-axis. If rainfall is above the first trigger, the crop has received enough rain; if it is between the first and second triggers, the policyholder receives a payout, the size of which increases with the deficit in rainfall; and if it is below the second trigger, which corresponds to crop failure, the policyholder gets the maximum payout. This product has inspired development agencies around the world, and today at least 36 pilot projects are introducing index insurance in developing countries.

Figure 1. Example of a Payout Schedule for an Insurance Policy against Drought

DTheredespite the potentially large welfare benefits, take-up of the product has been disappointingly low. Explanations for this low demand abound. The first and obvious reason is that the product is too expensive relative to the risk coping strategies now used by the farmers. After all, when it is not heavily subsidized (as it is in several states in India), average payouts, which are based on historical rainfall data, amount to about 30–40 percent of the premiums. In a recent paper several coauthors and I estimate that if insurance could be offered with payout ratios similar to those of U.S. insurance contracts, demand would increase by 25–50 percent. But even if prices were close to actuarially fair, demand would not come close to universal participation. So the price cannot be the whole story.

Another explanation is based on liquidity constraints: farmers purchase insurance at the start of the growing season, when there are many competing uses for the limited cash available. In the same paper we randomly assign certain households enough cash to buy one policy and find that this increases take-up by 150 percent of the baseline take-up rate. This effect is several times as large as the effect of cutting the price of the product by half and is concentrated among poor households, which are likely to have less access to the financial system.

In addition, potential buyers may not fully trust the product. Unlike credit, which requires that the lender trust the borrower to repay the loan, insurance requires that the client trust the provider to honor its promise in case of a payout. We measure the importance of trust by varying whether or not the insurance educator visiting households is endorsed by a trusted local agent during the visit. Demand is 36 percent higher when the insurance is offered by a source the household trusts. Trust may be particularly important because many households have only limited numeracy and financial literacy, which is likely to reduce their ability to independently evaluate the insurance.

These results point to several possible improvements in contract design. For example, the trust issue might be overcome by designing a product that pays often initially, since it is easier to sell insurance where a past payout has occurred. Liquidity constraints might be eased by ensuring that payouts are disbursed quickly or by offering loans to pay the premium. Finally, agricultural loans could be bundled with insurance, creating what is in effect a contingent loan, with the amount to be repaid depending on the amount of rainfall. This product was tested in a pilot in Malawi, and to our surprise demand for the bundled loan (17.6 percent uptake) was lower than that for a regular loan (33 percent). The reason may have been that the lender’s inability to penalize defaulting borrowers (in part, because of lack of collateral) was already providing implicit insurance and so farmers did not value the insurance policy.

What is remarkable about the Malawi experience is that after the pilot the lenders decided to bundle all agricultural loans with insurance. In their view, rainfall insurance had proved to be an attractive way to reduce the risk of credit default and had the potential to increase access to agricultural credit at lower prices.

The insurance covers only the loans. But informal discussions with borrowers suggest that they remain largely unaware that the loans are insured. Banks may not be telling borrowers about the insurance, however—because if they did, borrowers would need to know the exact amount of the payout (if any) to compute what they need to repay to the bank. In other words, uncertainty about the payout can undermine the culture of repayment. This happened in the Malawi pilot. One region of the pilot experienced a mild drought that triggered only a small payout. But because farmers were told that there had been a payout, they assumed that it covered the entire repayment amount and thus defaulted on their loans.

This example suggests that where financial literacy and understanding of the product are limited, insurance policies could instead be targeted to a group—such as an entire village, a producer group, or a cooperative—rather than to individuals. The decision to purchase insurance would be made by the group’s managers, who are likely to be more educated and more familiar with financial products than other group members and may also be less financially constrained. The group could then decide ahead of time how best to allocate funds among its members in case of a payout.

Further reading

Giné, X., R. M. Townsend, and J. Vickery. 2007. “Statistical Analysis of Rainfall Insurance Payouts in Southern India.” American Journal of Agricultural Economics 89 (5): 1248–54.

Giné, X., R. Townsend, and J. Vickery. 2008. “Patterns of Rainfall Insurance Participation in Rural India.” World Bank Economic Review 22 (3): 539–66.

Giné, X., and D. Yang. 2009. “Insurance, Credit, and Technology Adoption: Field Experimental Evidence from Malawi.” Journal of Development Economics 89 (1): 1–11.

Cole, S., X. Giné, J. Tobacman, P. Topalova, R. Townsend, and J. Vickery. 2010. “Barriers to Household Risk Management: Evidence from India.” Policy Research Working Paper 5504, World Bank, Washington, DC.


Join the Conversation

Nachiket Mor
January 15, 2011

Dear Xavier,

Our calculations show that for a moderately risk averse small farmer engaged in the production of rice or wheat rainfall insurance is not superior to her own existing risk coping strategy and that payoffs with insurance are inferior to payouts without insurance under all but extreme states of the world. Do you agree with this? Are there detailed tools that you could point us to that could help us do this accurately for Indian farmers?

We have built a very trusted financial services channel in Tamil Nadu, Orissa and Uttarakhand ( and plan eventually to take it pan India. We are just not convinced that routine rainfall insurance is in the interests of the small farmer.

Would be eager to hear your views.


Nachiket Mor

January 18, 2011

Dear Nachiket,
Thanks a lot for the comment.

I agree that there are instances when existing risk coping mechanisms can do a good job. For example, participation in NREGA program can provide some form of insurance for eligible households. Also, the state and central governments offer financial relief measures for borrowers from commercial banks and credit cooperatives. But rainfall insurance can actually be very valuable precisely because it provides a payout in situations where informal risk coping mechanisms will not work well. Take the 2009 Monsoon, for example. Rainfall insurance in the two districts in AP that we've been studying paid handsomely (up to Rs 1,000 for a policy that cost Rs 80), yet most farmers agricultural revenue was poor, so that their ability to help others was limited.

At any rate, I'd be curious to see how you simulate farmer welfare under different weather patterns and your assumptions about available risk coping mechanisms they have access to.

A key problem with rainfall insurance that we do not have a good handle on is basis risk, that is, the lack of correlation between what the farmer cares about, ie agricultural yields and the event being insured, ie rainfall at a nearby rainfall gauge. Basis risk can be high because the rains at the rainfall gauge differ from the rains at the farmer' plot or because the correlation between rain and yield is poor. Either way, the insurance product will be less desirable. We are starting some work to try to quantify the magnitude of basis risk, so any feedback on this will be greatly appreciated.

As for tools, I know that colleagues at the bank and at IRI (U. Columbia) are developing a toolkit, but the focus is on insurance pricing, more than welfare calculations for farmers. I'll ask the details and report back.

Desire Balazire
January 20, 2011

The topic is interesting.

I have two major points:
(i) I don't agree with the last statement that the decision to purchase insurance would be made by the group's managers...

With the market economy we should not take options that bring people to socialist or communist theories.

(ii) Is the amount of the insurance calculated on the same basis regardless the place are located: tropical, equatorial. The rainfall is random depending on the area.