The World Bank Policy, Research Working Paper No. 5504, December
Shawn Cole, Xavier Gine, Jeremy Tobacman, Petia Topalova, Robert Townsend and James Vickey (2010)

The paper examines a financial contract designed to insure rural Indian households against an exogenous income risk: rainfall variation during the monsoon season. 

The study was conducted in the rural areas of the states of Gujarat and Andhra Pradesh and it reveals the importance of liquidity constraints faced by households, especially during the monsoon season. Other determinants of demand for rainfall insurance were also identified. Finally the study finds that the degree of trust that households display towards the insurance seller is important in assessing whether the households buy insurance or not.

Rainfall has been deemed an important factor for income variability in the semi-arid areas of India. Previous research has shown that farmers use a range of measures to cope with rainfall risk by borrowing and saving, relying on remittances and asset sales and also engaging in costly income-smoothing measures like shifting towards safer but less profitable production activities.   This paper studies a financial contract intending to insure rural households against this exogenous income risk. Rainfall insurance is sold commercially before the start of the monsoon season and then pays off based on rainfall recorded at the local weather station.

Despite the potentially large welfare benefits of taking up rainfall insurance, its acquisition is rather low – there might be other factors that prevent the adoption of such financial instruments.

The authors studied the rainfall insurance policies offered by companies like ICICI Lombard in the state of Andhra Pradesh and IFFCO-Tokio in Gujarat. Payoffs were calculated based on rainfall measured at a nearby government rainfall station.

Separate policies were sold for each phase corresponding to sowing, flowering and harvesting between Rs. 80 and Rs. 120 (US $2-3). In Andhra Pradesh, the study was based on a survey of 1047 households in 37 villages while in Gujarat the survey consisted of 50 villages. The non-government organizations (NGOs) sold the policies on behalf of the insurance companies in these two states. In Andhra Pradesh, the insurance was sold by an NGO with an extensive rural network of local agents known as Livelihood Services Agents (LSAs). In Gujarat the product was sold by a large NGO known as SEWA.

The researchers designed the field experiments to identify the responsiveness of insurance demand to changes in price and to determine non-price factors like liquidity constraints and trust in the product and in the agents selling those products. Prior to the start of the monsoon season in May 2006, 700 households in Andhra Pradesh were randomly assigned a) whether the household was visited by an insurance educator, b) whether the educator was endorsed by an LSA and declared trustworthy, and c) whether the household received a high compensation for its time.  In Gujarat, 20 villages were assigned videos which depicted a) whether the video featured a strong brand emphasis of SEWA, b) whether a peer or an authority figure endorsed the insurance, c) whether the payout from the policy is depicted positively as paying in 2 out of 10 years or not paying in 8 out of 10 years, and d) whether the insurance is framed as being a “safety” measure for households or “vulnerable” for households without insurance. Additionally the Gujrati households were randomly assigned flyers of two types: a) the religious content of the flyer, whether it was Hindu, Muslim or neutral, b) whether the flyer emphasized the benefits of insurance accruing to the group or the individual. The rationale behind this design was that the authors wanted to find out how religion affects trust and influences the opposite group.

A number of key findings have emerged from this study. In Andhra Pradesh, the insurance take-up rate was 28%. Households visited by an educator, when not combined with the other two treatments, increased take-up by 11.5 to 17.2 percentage points, while a high compensation increased insurance purchase by 39.4 to 40.8 percentage points.  For households familiar with the workings of the NGO BASIX in Andhra Pradesh, the LSA endorsements increased the acquisition of the product by 10.1 percentage points. The results also indicate that the high cash reward had a greater impact on the product purchase among the poorer households. In Gujarat the overall acquisition of the product was 29%, around the same level as the households in Andhra Pradesh. Among the households receiving a flyer emphasizing the religion, Muslims households were less likely to take up insurance when the flyers included Hindu symbols. Similarly for Hindu households witnessing Muslim symbols, the purchase was lower but not as low as the Muslim homes. This suggests that the collective nature is an important determinant for insurance demand when group members of a specific religion are emphasized.

When estimating the sensitivity of insurance demand to price, the researchers found out that demand would increase significantly by 25-50% if the insurance was offered with the same payout-to-premium ratio as insurance contracts in the US. However, even if the ratio was increased along with a reduction of the policy price, only a small fraction of households will purchase the insurance. This again highlights the importance of non-price factors in influencing demand. The demand is very low among the poorest households which have very low access to financial services and face severe liquidity constraints; whereas wealthier homes, especially in Gujarat, are more likely to buy insurance. The reason could be that for such households there are benefits of accumulating limited liquid assets or using those assets to invest in agriculture, instead of buying insurance. The Andhra Pradesh findings suggest that the farmer’s trust in the insurance educators considerably increases demand. This implies that trust is likely to be important for financial services in rural environments, where formal legal protections are very weak and where household financial literacy and education is low.

Finally, some qualitative self-reports reveal that the most common reason stated by households in Andhra Pradesh was the lack of funds to buy insurance, with 81% of households citing it as the most important reason for not obtaining insurance. These responses are consistent with the observation that liquidity constraints matter a lot for households when they are deciding to purchase a policy.

The authors conclude that the importance of liquidity constraints dictate that polices should be designed to provide payouts quickly, especially during the monsoon season when households are faced with a credit shortage. They further recommend that the government should subsidize rainfall insurance, which will increase participation. The findings regarding inadequate individual financial literacy and education, suggest that insurance policies should instead be targeted to groups, for example, an entire village.