A Randomized Evaluation of the Effects of an Agricultural Insurance Program on Rural Households’ Behavior: Evidence from China

Publication Details

Cai, J, de Janvry, A and Sadoulet, E, 2014. A randomised evaluation of the effects of an agricultural insurance programme on rural households’ behaviour: evidence from China, 3ie Impact Evaluation Report 19. New Delhi: International Initiative for Impact Evaluation (3ie)

Link to Source
Jing Cai, Alain de Janvry, Elisabeth Sadoulet
Institutional affiliations
None specified
Grant-holding institution
None specified
China, People's Republic of
East Asia and Pacific (includes South East Asia)
Agriculture and Rural Development
Weather Insurance
Gender analysis
Weather Insurance
Gender analysis
Equity Focus
None specified
Evaluation design
Randomised Control Trials (RCT), Mixed Methods
3ie Series Report
3ie Funding Window
Open Window Round 2


The study sites include over 200 randomly selected villages in the Jiangxi province of China, with around 8000 households. Treatment villages were visited twice for a first and a second round of education session. At each round different farmers participated. In order to examine the causal effect of financial education and social learning on insurance take-up, households were randomly assigned to eight different treatment and control groups. During the first round session there was a control group (1079 households) and a treatment group that obtained intensive insurance education (1096 households); At the second round session the simple education group was divided in three different treatments: a control (660 households), a group 2 that obtained information on the insurance take-up after the first round session (355 households) and a group 3 that obtained the list of farmers that took up insurance after the first round (362 households). 

The intensive education group in the second session was also split up in these three treatments, i.e. no additional information (657 households), information on overall take-up (350 households) and a group that received a list showing who took up insurance (343 households).The study sites included over 200 randomly selected villages and approximately 8,000 households in Jiangxi Province in China. In order to examine the causal effect of financial education and social learning on insurance take-up, households were randomly assigned to different treatment and control groups.

Main findings

The study has the following main findings:

1. It finds that financial education about insurance and its benefits improves take-up by 15 percentage points (from 35 to 50 per cent) or 43 per cent of its base value.

2. Large positive spillover effects of financial education on adoption by others were observed: having one additional close friend attending training is equivalent to 50 per cent of the direct training effect, equivalent to the impact of reducing the average insurance premium by 15 per cent.

3. This spillover (social network) effect is driven by the diffusion of knowledge (farmers who have more friends exposed to financial training perform significantly better on an insurance knowledge test), rather than by trust, imitation (the take-up decisions of other villagers, even close friends, do not significantly influence behaviour because this information is not shared through the social network), or informal risk sharing.

4. In the long run, farmers who have more friends receiving an insurance payout in the previous year are 23.6 per cent more likely to buy insurance in the second year. That effect is equivalent to 50 per cent of the effect of directly receiving a payout.

5. Cost-sharing subsidy policies are better than free distribution policies for the long run development of such insurance programmes.

6. Offering a menu of insurance contracts rather than a single contract increases take-up dramatically.

Additionally, the study takes advantage of exogenous variation in baseline insurance take-up generated by the financial education treatment to identify the causal effect of insurance provision on households’ production, saving and borrowing decisions. The authors found that households increased production of insured crops (rice), but the effect was not statistically significant. Moreover, households borrowed more and were more likely to repay a loan after they purchased insurance. However, these effects were not statistically significant.

There was no significant effect of insurance provision on saving.

It is very early to derive policy recommendations based on a two-year study of the introduction of a brand new financial product that is difficult to understand. However, three points can be made. First, it is clear that social networks are important in helping people understand and adopt this new product. Hence the insurance company should continue to share the information on people’s decisions to insure and on payouts distributed. Second, making sure that people understand the product is critical and worth several points of subsidies. It is a very cost-effective marketing tool. Third, in so far as the government is willing to subsidise the insurance, this is definitely welfare enhancing for the farmers. Two years is, however, too short to see how farmers will adjust their production, saving and borrowing behaviour. Nevertheless, there are some indications that they indeed respond by increasing investment in production, a very encouraging result.

Additional publications

About this impact evaluation

This study analyses the determinants of adopting a new weather insurance product in China and the impact of insurance provision on farmer behaviour. It assesses the first rice production insurance policy for rural households, which was designed and offered by the People’s Insurance Company of China (PICC). In particular, the study analyses the role of social learning and contract design in improving the diffusion of a new insurance product.

Website : http://www.piccnet.com.cn/

Scroll to Top