The impact of weather index insurance on agricultural technology adoption evidence from field economic experiment in China

Yingmei Tang (College of Finance, Nanjing Agricultural University, Nanjing, China)
Yue Yang (College of Finance, Nanjing Agricultural University, Nanjing, China)
Jihong Ge (China Center for Food Security Studies, College of Economics and Management, Nanjing Agricultural University, Nanjing, China)
Jian Chen (The Ohio State University, Columbus, Ohio, USA)

China Agricultural Economic Review

ISSN: 1756-137X

Article publication date: 26 June 2019

Issue publication date: 18 October 2019

Abstract

Purpose

The purpose of this paper is to empirically investigate the impact of weather index insurance on agricultural technology adoption in rural China.

Design/methodology/approach

A field experiment was conducted with 344 rural households/farmers in Heilongjiang and Jiangsu Provinces, China. DID model was used to evaluate farmers’ technology adoption with and without index insurance.

Findings

The results show that weather index insurance has a significant effect on the technology adoption of rural households; there is a regional difference in this effect between Heilongjiang and Jiangsu. Weather index insurance promotes technology adoption of rural households in Heilongjiang, while has limited impact on those in Jiangsu. Weather, planting scale and risk preference are also important factors influencing the technology adoption of rural households.

Research limitations/implications

This research is subject to some limitations. First, the experimental parameters are designed according to the actual situation to simulate reality, but the willingness in the experiment does not mean it will be put into action in reality. Second, due to the diversity of China’s climate, geography and economic environment, rural households are heterogeneous in rural China. Whether the conclusion can be generalized beyond the study area is naturally questionable. A study with more diverse samples is needed to gain a fuller understanding of index insurance’s effects on farmers in China.

Originality/value

This research provides a rigorous empirical analysis on the impact of weather index insurance on farmers’ agricultural technology adoption through a carefully designed field experiment.

Keywords

Citation

Tang, Y., Yang, Y., Ge, J. and Chen, J. (2019), "The impact of weather index insurance on agricultural technology adoption evidence from field economic experiment in China", China Agricultural Economic Review, Vol. 11 No. 4, pp. 622-641. https://doi.org/10.1108/CAER-05-2018-0107

Publisher

:

Emerald Publishing Limited

Copyright © 2019, Emerald Publishing Limited


1. Introduction

For many developing countries, agriculture provides the leading source of employment and contributes large proportion of the national income. Numerous studies have reported that adoption of improved agricultural technologies enhances household well-being in most less developed countries (Bourdillon et al., 2002 in Zimbabwe; Mendola, 2007 in Bangladesh and Kassie et al., 2011 in Uganda).

Because of China’s geographic and climatic diversity, extreme weather events occur almost every year. Rural households in China are exposed to the risk of natural disasters. From 2006 to 2015, about 35m hectares of crops were affected by natural disasters each year, accounting for 29 percent of the arable land, with an average annual food loss of nearly 17m tons (China Statistical Yearbook, 2015).

Due to the high cost of or inaccessibility to formal risk management, most of the natural risks remain uninsured in China. Shocks of natural disasters can not only reduce current returns, but also destroy assets accumulated over years. Small households are liable to adopt safer, but lower-return agricultural technologies, and may keep low consumption because they are always exposed to natural risks (Dercon, 1996; Zimmerman and Carter, 2003; Barnett and Mahul, 2007; Hill, 2008; Carter et al., 2012; Lybbert and McPeak, 2012; Hill et al., 2013). Agricultural technological improvements are crucial to raising agricultural productivity, reducing poverty and solving the problem of food security. Less adoption of new risky agricultural technologies and lack of technological change locked small householders into low productivity and subsistence production, then make them fall into trap of poverty (Barnett et al., 2008; Asfaw et al., 2012). Agricultural insurance has been believed as an effective way of helping small householders diversify production risks. Fadhliani et al. (2019) revealed that an increase in the insurance coverage level will lead the farmer to apply more inputs to increase expected yield in Indonesia.

Chinese government promotes multi-peril crop insurance (MPCI) as a way to deal with risks in agricultural sector. The coverage of MPCI is too low and franchise deductibles are too high, and the largest indemnity cannot offset the variable material costs of production, which makes that the insurance product cannot significantly improve farmers’ ability to bear risks (Zhao et al., 2016; Ye et al., 2017). Simultaneously, MPCI possesses some well-known structural problems such as moral hazard, adverse selection, blurry boundaries between government and market, disordered market competition and improper management, which makes it extremely expensive. Its operation heavily depends on government subsidies (Hazell, 1992; Miranda and Glauber, 1997; Skees et al.1999; Tuo, 2016). Given the high costs of MPCI, a growing number of academic researchers and governments have exhibited interest in the use of weather index insurance to manage the risks faced by poor agricultural producers (Miranda and Vedenov, 2001). Weather index insurance indemnifies the insured based on the observed value that is highly correlated with the losses, but cannot be influenced by the insured (Miranda and Farrin, 2012). Weather index insurance may be the best alternative for achieving Chinese government policy objectives of providing agricultural insurance to as many farmers as possible at reasonable costs.

With the outpouring of index insurance pilot program in Africa, South and Southeast Asia, a number of studies have examined farmer’s demand for index insurance and its impact. Theoretical models, framed field experiment and randomized controlled trial have been used to assess the impact of insurance on production decisions. Many researchers have found that improved technologies can increase production, but it is risky and require up-front investment. After index insurance becomes available, farmers can get more credit and invest it in profitable production. Insurance leads to more adoption of innovative agricultural technology (Stein, 2011; Cole et al., 2013; De Nicola and Hill, 2013; Elabed et al., 2013; De Nicola, 2015; Farrin and Miranda, 2015). Jensen et al. (2014) claimed that insured households reduce the precautionary savings, and then have more cash to invest in production activities.

But some researchers have stated that the insurance improves household welfare individually, but not for a cumulative household welfare. Insurance has different effects on the households with different wealth. With insurance, low-income households avoid cutting down consumption and families with high income do not sell their productive assets to withstand the shocks (Janzen and Carter, 2013; Karlan et al., 2014). McIntosh et al. (2013) advised that index insurance does not boost investments, but help those who have already invest in high agricultural technology spread the risk. Farrin and Murray (2014) argued that the premium of insurance in good years increases farmers’ costs and had a negative effect on wealth.

Taken together, the conclusions of existing studies are not consistent and samples are limited to Africa, South and Southeast Asia; no empirical evidence about the impact of index insurance has been investigated in China so far. Further research is needed. The main purpose of this research is to empirically investigate the impact of weather index insurance on farmers’ improved agricultural technology adoption. A framed field experiment was conducted with farmers to detect their technology adoption with and without index insurance in Heilongjiang and Jiangsu Provinces in China. To the best of our knowledge, no prior literature has used field experiment to study weather index insurance and its impact in China. Our experimental design was based on some research that have been successfully implemented in other developing countries (Giné and Yang 2009; Lybbert et al., 2010; Patt et al., 2010; Hill and Viceisza, 2012; Norton et al., 2014; Carter et al., 2014). As field experiment allows farmers to make technology choices in a continuous and dynamic context; this increases the external validity of our findings and the policy implications will be more credible.

The rest of this paper is organized as follows. Section two explains how weather index insurance affects farmers’ agricultural technology adoption through theoretical analysis. The third section describes data set and field experiment design. Empirical strategies are presented in section four. Section five introduces the robust test. Section six is the conclusion and discussion.

2. Conceptual framework and hypotheses

Farmers are faced with two technological options when making production decisions: traditional technologies with low risk and low return; innovative technologies with high risk and high return. Natural hazards (e.g. droughts, floods and diseases) are major risks in agricultural production. Households are assumed to maximize their utility function subject to family endowment and natural risks. If the utilities of adopting innovative technologies are larger than that of traditional technologies households will choose the innovative technologies. Unlike other sectors, uncertainty caused by natural risks has a significant effect on decision making in agricultural sector (Ahsan et al., 1982). Most farmers are risk averse and will often be hesitant to adopt innovative technologies with high profitability and high risks, and may choose to maintain existing low productivity technologies with low risks, due to the lack of both ex ante and ex post risk management strategies such as formal credit and insurance (Mukasa, 2018; Dercon and Christiaensen, 2011; Rosenzweig and Binswanger, 1993). Insurance has two functions: it promotes farmers’ ex ante adoption of advanced production technology with higher expected return and variance, through reduced credit rationing. It helps rural household smooth consumption when they experience large seasonal fluctuations due to unusual weather events (Ghosh et al., 2000). By allowing farmers shifting risks, insurance enables rural households to undertake risky investment which they would not engage in if there were no insurance, and thus could lead to Pareto-preferred states (Ahsan et al., 1982). In countries where the insurance market is growing rapidly, better designed insurance contracts can mitigate the weather shocks and boost investments in fertilizer, hired labor, irrigation, and pesticides as well, which helps farmers’ production choices to achieve a potential Pareto improvement in aggregated social welfare (Ye et al., 2017; Hill et al., 2019). By spreading individual risks, insurance serves as an effective hedge against natural disasters by reducing the variance of output and increasing farmers’ risk-taking ability, which then could result in improved allocation of resources in innovative technologies.

In general, by offering the possibility of transferring risks, weather index insurance could improve farmers’ ability to withstand risks and enable them to allocate more resources into risky investments such as adopting new technologies. Thus, our hypothesis is:

H1.

Weather index insurance will promote farmers’ innovative technology adoption.

3. Data collection and experiment design

3.1 study area

The field experiment was conducted in three counties in Heilongjiang Province and two counties in Jiangsu Province in July and August of 2017 (see Table I). Totally, 344 participants were recruited from 32 villages and 1,376 observations were obtained with four-round panel data. The counties in Heilongjiang Province are all located within the main rice growing areas which suffered a catastrophic flood in 2013. Most households in this region already took up multiple-peril crop insurance to help manage agricultural risks. As a result, it is expected that households from the study region in Heilongjiang Province are more likely to be receptive to weather index insurance. Another important province in the reform and innovation of agricultural insurance is Jiangsu Province which is densely populated with less per capita arable land (0.85 mu[1]). The cultivated land conditions are similar and all in the plains with developed irrigation system. The natural risks are less than that in Heilongjiang Province (see Table VIII). On the contrary, Heilongjiang is vast and sparsely populated and the per capita arable land is 6.27 mu. Some of cultivated land is located in hills and is vulnerable to drought disasters; the other part of cultivated land is located in the plains which is vulnerable to floods due to their proximity to the Songhua River. In order to avoid the difference in cultivated land conditions affects the farmers’ response in the experiment, the research team ensure that the samples from villages in the hills equal to those in the plains.

These differences in geographic location, modes of agricultural production and frequency of natural disasters may lead to farmers’ different demand for weather index insurance and technology adoption. So these two provinces are selected as study area.

3.2 Sample selection

To ensure subjects could understand the experiment context about agricultural technology adoption and weather index insurance terms, criteria for selecting the subjects were as follows: Subjects should have farming experience and participate in family decision making. They should have received primary or higher educations. We got the villager list from local officials and chose farmers who meet the above criteria to form a sub list, and then randomly recruited subjects from it[2].

3.3 Field experiment design

Field experiment recruits farmers as subjects in the field instead of recruiting students in the lab. Farmers are in the context of agricultural production which they are very familiar with, rather than be introduced of abstract terminologies. It allows researchers to analyze the effects of exogenous treatments on farmers’ behaviors (Harrison and List, 2004). The questionnaire survey obtains information by asking farmers some subjective questions. It can only get the static data. The field experiment simulates agricultural production under different weather conditions and lasts for a few rounds. It helps farmers understand insurance terms under different climate in this experiment. Farmers’ technology choices are made in a continuous and dynamic context which is close to the real life of famers, and then the results would be more reliable.

3.3.1 Division of the control and treatment group

To test the effect of index insurance on farmers’ technology adoption, all subjects were divided into control group and treatment group randomly. With other conditions being equal, treatment group was provided with index insurance and control group without it. Two groups of farmers would choose agricultural technology between traditional technologies with low risk and low return and innovative technologies with high risk and high return. The influence of index insurance could be valued through comparing the data of the two groups.

Experiment was conducted for four rounds with two groups, respectively. Each round represented one planting cycle. To analyze the impact of index insurance on the treatment group, there is no insurance for the first two rounds and with insurance for the next two rounds in this group.

According to preliminary research, it was found that improved seeds, new chemical fertilizers and pesticides are the main technologies affecting rural households’ farming income. Among them, seeds had the greatest impact on yields, and then were selected as representative of improved agricultural technologies (Giné and Yang, 2009; Lybbert et al., 2010; Stein, 2011; Ward and Singh, 2015; Freudenreich and Mußhoff, 2018).

3.3.2 External validity

To ensure the external validity of the experimental design, the experimental process need to be controlled, prerequisites need to be introduced and the experimental parameters should be set carefully before the formal experiment.

3.3.2.1 Liquidity constraints

In the experiment, the subjects who suffered bad weather may get net income equal to or less than zero, and have no funds to purchase agricultural production materials next year. To continue agricultural production, they must borrow money from relativities or rural financial institution in real life. The purpose of this research is to test the impact of insurance and other factors were excluded. It is assumed that there are no credit or liquidity constraints.

3.3.2.2 Status quo bias

Status quo bias means that people tend to maintain choices that have been made in the past. To avoid this preference, the farmers draw lots to determine the weather at the end of each round. Different weather leads to changes of income, which made farmers rethink and choose technology, not just repeating his choice of former rounds. After the experiment, if the farmer’s choice remains unchanged in four rounds the experimenter would ask him the reason to judge whether his choices were based on status quo bias or serious thinking. In the total samples, about 30 percent of the farmers’ choices remained unchanged, which was close to the reality[3].

3.3.2.3 Peer effects

To minimize the effects of subject’s peers and social networks on individual’s technology adoption, the subjects sat in a class room of local primary school and were separated to ensure that they cannot see each other’s choice and talk with each other from the beginning to the end of the experiment. If they have any questions, they were instructed to ask the experimenter for help. Each experimenter was responsible for four subjects in every experiment. They were trained before the experiment to ensure they express all instruction in neutral and unified language. All experiment information and tasks were put into a unified laboratory manual and were expressed in neutral language. The manual was handed out to famers for reference.

3.3.3 Experiment scenario

To simplify the experiment, each farmer is provided with the same initial endowment (10 mu of land and ¥11,200) as startup funds. In the preliminary experiment, we consulted local farmers and technicians about the cost and income of agricultural production. The cost of traditional seeds with traditional fertilizers and pesticides is nearly ¥600 per mu, and then the cost of improved seeds with improved fertilizers and pesticides is about ¥1,100 per mu[4]. The cost here includes all material costs. It is easy to understand for farmers. The income per mu is around ¥1,000 with traditional seeds and almost ¥2,000 with improved seeds. Insurance premium and indemnity of current local agricultural insurance are ¥15 and ¥200 per mu, respectively. The experimental parameters were designed based on these values and ratios.

The living cost is ¥5,000 for every family, and was covered by the earning of last year. When the farmers make the technology decision at the beginning of next year, they do not need to consider it and have enough money to buy improved seeds (11,200>11,000). The subjects were told this assumption before the experiment began.

The weather is simplified as two kinds: bad weather and normal weather. The probability of bad weather is 30 percent (based on the historical weather data of study area). To prevent farmers from being influenced by experimental setting; we label traditional and improved seeds as No. 1 and No. 2, respectively. Table II shows the experiment parameters.

3.3.4 Experiment process

The experiment includes four parts: a game to help farmers understand index insurance, field experiment of technology choice, a game to test farmers’ risk preference and a short questionnaire survey.

3.3.4.1 The game of introducing weather index insurance

Most farmers are not familiar with weather index insurance in study area; poor understanding of insurance will influence the experiment efficiency. The challenge we faced in experimental design was interpreting index insurance clearly to the participants. We played a game to help them understand index insurance terms and basis risk before the formal experiment.

First, weather index insurance was introduced to the participants. The weather index insurance is a new product. Its indemnity is calculated according to the contract weather index (such as rainfall or temperature) which is highly correlated with yield. When the crop growth period rainfall/temperature of the nearest weather station is less or higher than the contract index, all insured farmers will receive the same indemnity no matter how much loss they suffered.

Second, the participants simulated production, and then the representative drew lots to determine the crop growth period temperature/rainfall. Afterwards, compare it with the contract index; if it is more or less than the contract index, all the participants got the same indemnity. Third, the participants drew lots to determine personal loss, and then compared it with his indemnity. The difference was the basis risk which is an important factor affecting the performance of weather index insurance (Xiao and Yao, 2018).

Finally, we assessed the participants’ understanding through a quiz. If the participants could not answer the questions correctly, we helped them understand it until they gave correct answers before the formal experiment began.

According to local climate, weather index insurance is designed as a rice temperature and rainfall composite index insurance. The insurance clauses took the form of the contract of Guoyuan Insurance Company which is the first company to provide index insurance in China.

3.3.4.2 Technology selection experiment

We introduced experiment task to the subjects and handed out flyers with experimental parameters to them for reference. The subjects selected seeds to simulate production with initial endowment, and then draw lots to determine the weather. The experimenters calculated net income for them at the end of each round and carried it over as start-up capital for the next round. This process lasted four rounds for control group. The first two rounds of treatment group were the same with control group. In the last two rounds, the subjects were provided with weather index insurance and then they made seeds decisions.

3.3.4.3 The game to test farmers’ risk preference

Farmers’ risk preference influences their production decisions (Dercon and Christiaensen, 2011; Ward and Singh, 2015). Based on the research of Holt et al. (2002), Brick and Visser (2015), we played a game to test farmers’ risk preference.

The subjects participated in a lottery game with two options. Option A is to obtain a certain amount of money which increased from ¥3 to ¥25. Option B is a gamble which has seven black balls (present ¥0) and three white balls (present ¥50). The probability of drawing black and white balls is 70 and 30 percent, respectively.

The subject was asked to choose from two options to decide which one he is more preferred: to get 3¥ directly, or to participate in the game of drawing. If the subject chose option B, then the next time the amount of money in option A would increase two more ¥ and the subject continued to choose. The game ended when the subject chose option A. The more times Option B is chosen, the higher farmer’s risk tolerance is, because there is more uncertainty in Option B than in Option A. The times of choosing option B is farmer’s risk preference score which varies from 0 to 12. The higher the score is, and the farmer is more willing to take risks (Table III).

3.3.4.4 A short questionnaire survey

Finally, a questionnaire survey was conducted to collect the information about households’ characteristic. Each experiment lasted for 60 min. After the experiment, each participant got ¥80 as compensation which equals two and half hours wages of local labor.

4. Empirical analysis of weather index insurance’s influence on technology adoption

4.1 Selection of DID model

Panel-Difference in Difference (Panel-DID) was used to analyze the impact of index insurance. DID assesses the influence of policy or external events by comparing the difference between treatment group and control group. Weather index insurance is an external variable. It will influence the technology choices of the treatment group before and after insurance was provided; moreover, it makes farmers’ technical adoption of treatment group differ from control group. DID was used to investigate the impact of risk shock and insurance on farmers’ venture capital investment by Hill and Viceisza (2012); following their study, logistic regression (LOGIT), fixed effect model (FE) and random effect model (RE) are used for DID regression to ensure the robustness of the model. The panel-DID model can be established as:

(1) T e c h i t = β 0 + β 1 T + β 2 I i t + β 3 T × I i t + β 4 X i t + ε i t .

Techit represents the agricultural technology choice of the ith farmer in the tth period, T is a dummy variable of group (control group is 0; treatment group is 1). I denotes a dummy variable of period (Before providing with weather index insurance, I= 0; otherwise I=1). T×I indicates the net effect of weather index insurance on the agricultural technology adoption. Xit are control variables, including individual characteristics of households, family characteristics, risk preferences, etc.

4.2 Dependent variables

We measured farmers’ technology adoption through the selection of traditional seeds and improved seeds: when the farmer chooses the traditional seeds, it is 0, otherwise it is 1. Farmers’ agricultural technology adoption in four rounds is shown in Table IV.

Table IV shows that the improved seeds adoption rates are different within the group and between groups. Overall, the improved seeds adoption ratios are all more than 70 percent in four rounds of two groups, and higher than those of traditional seeds. Through focus group interviews, we learned that seeds companies often organized new seeds promotional activities. The technicians in technology extension stations provided technical training program sometimes. Farmers in study area are familiar with improved seeds and willing to accept them.

The growth rates of the improved seeds adoption ratio between the two groups are different. In the last two rounds of the treatment group, the ratio of the farmers’ choosing improved seeds substantially increases, compared with those of the first two rounds. After index insurance was provided, the ratio of the farmers’ choosing improved seeds increases from 70.35 and 73.26 percent in the first and second rounds to 81.98 and 85.47 percent in the third and fourth rounds of the treatment group. In the control group, the experimental context remains unchanged, and the improved seeds adoption is relatively stable. The ratio of farmers’ choosing improved seeds increases from 76.74 percent in the first round to 80.81 percent in the fourth round. The change is much smaller than that of the treatment group.

There are differences in the rate of improved seeds adoption between two groups. In the first two rounds, the ratios of improved seeds adoption in the control group are higher than those of the treatment group by 6.39 and 5.81 percent, respectively. After the farmers were provided with insurance, the rates of adopting improved seeds in treatment group are higher than those of the control group by 2.91 and 4.66 percent in the last two rounds. The findings preliminarily indicate that weather index insurance influences farmers’ technology selection, but further empirical tests are needed.

4.3 Control variables

Based on the research of Ward and Singh (2015), and Meng, individuals and households characteristics, natural disaster experience and risk preference are included as control variables. The household endowments of rural households have been set as experimental parameters; they are not taken as control variables. Descriptive statistics of the main variables are shown in Table V.

Table V shows that most of the subjects are male (about 84.4 percent), and the mean of education is 7.305 years with only 36.7 percent completed junior high or higher education (⩾9 years). The average experience engaging in planting is about 27 years and the average rate of agricultural labor is 85 percent. In other words, most of the samples’ income comes from agriculture. The average household farm size is about 80 mu. About 60 percent of the farmers have suffered natural disasters in the past five years, and the loss is nearly 40 percent of normal annual income.

The average risk appetite of the farmers is about 6 (on a scale of 0–12). There is no significant difference in the individual characteristics of the two groups.

According to the meteorological data, the probability of bad weather in the experiment is 30 percent. The weather of the former year will affect the technological choices of the next year; the lag phase of weather is used as a control variable.

The previous studies of Dercon and Christiaensen (2011), Ward and Singh (2015) and (Ahsan et al., 1982) found that risk attitude affects farmers’ production behavior. To investigate the impact of different risk attitudes on farmers’ technology adoption, all subjects were classified into three categories based on the study of Binswanger (1980) and the scores of the test. Farmers with risk scores between 0~4, 5~8 and 9~12 were classified as risk-averse, risk-neutral and risk-preferring farmers, respectively. Risk-averse farmers account for the highest proportion of 42.73 percent. Risk-neutral households account for 18.9 percent, the lowest proportion (see Table VI).

4.4 Model regression results and interpretation

This research focuses on the regression coefficient of T×I, which expresses the net effect of weather index insurance on farmers’ agricultural technology adoption. Since agricultural development, natural conditions, and production methods are different in Heilongjiang and Jiangsu Province, DID model is adopted to analyze both pooled and by province samples.

It can be seen from Table VII that the coefficients of T×I of the pooled sample and the Heilongjiang sample in three models are all significant, but that of the Jiangsu sample is not significant. It means that weather index insurance plays an important role in promoting the adoption of improved seeds after controlling the time-varying effect and the differential effect in Heilongjiang Province, and there are regional differences in the impact of weather index insurance.

In the experiment, farmers’ adoption rate of improved seeds increases after the insurance is provided, which verifies the research hypothesis. The result is consistent with previous studies by Hill and Viceisza (2012) and Norton et al. (2014), which indicates when farmers are provided with insurance they are more willing to allocate resource endowment in risky investment. Improved seeds would cost more investment, in the absence of risk-sharing strategies, farmers will suffer a higher loss if bad weather occurs, which constrains the farmers’ advanced technology adoption; weather index insurance, as a risk management tool designed specifically for natural disasters, can effectively help farmers diversify natural risks and reduce post-disaster losses. Table VII shows after they are provided with index insurance which effectively spread the natural risks, new technologies adoption are much higher than before.

In terms of statistical criteria, three models are all significant; which means that the research results are robust. In view of the variable significance and informativeness of the model results, the logit model can provide a full picture of farmers’ advanced technology adoption; we focus on the discussion of it to investigate the major drivers in addition to weather index insurance which influence agricultural technology adoption. In the logit model, the variable of risk preference has significant positive effects on technology adoption both in pooled and by province samples. It indicates that risk-preferring farmers have preference for advanced technology; they are more concerned about benefits other than risk, and will prefer to new technologies with high risk and high return rather than traditional technologies with low risk and low return. For each additional unit of farmer’s risk tolerance, the possibility of adopting new technology is increased by 1.0, 0.9 and 1.5 percent in Jiangsu, Heilongjiang and pooled samples, respectively, which is consistent with the theoretical hypothesis.

The results also show that advanced technology adoptions are driven by changes in weather, household farm size and the age of farmers. In Table VII we see that the coefficient of weather is significant at 5 percent level in both the pooled and Heilongjiang samples, indicating that if the weather of the previous year is normal, more new technologies would be adopted by the farmers, which is consistent with our expectations. The variable of household farm size has a significant positive impact, which denotes that large-scale farmers can achieve more significant scale effects by adopting technology, and thus they are more inclined to adopt new technologies. Age negatively affects the technology adoption significantly, indicating that the older farmers are more reluctant to accept new technology.

Variables of gender and agricultural labor rate have significant positive effects in the pooled and the Jiangsu samples; gender has a significant positive effect on technology adoption at 1 and 5 percent levels, respectively. This is consistent with the previous findings by Li et al. (2010) and Zhu et al. (2015). The rate of agricultural labor is also the main driver of technology adoption, the more agricultural labor, the more dependent the household income is on agriculture, and improved technologies are more crucial to income increase.

Variables of education and suffering natural disasters have significant positive effects only in the Jiangsu sample, the possible reason is that farmers in Jiangsu Province have longer education experience than those of Heilongjiang Province (see Table VIII), and they can better understand the advanced technology, which results in a significant positive effect of education; Jiangsu sample is less likely to suffer natural disasters than that of Heilongjiang, and their perception of extreme weather event may not be as strong as that of Heilongjiang farmers, then they are more liable to adopt new technology. These variables should be considered when implementing weather index insurance to promote farmers’ agricultural technology adoption.

Table VII shows that the same variable has different influences in pooled and by province samples. In order to find the possible reasons, a sample t-test is conducted to compare the individual or households characteristics differences between samples of the two provinces. Only the significant variables are listed (see Table VIII).

It can be seen from Table VIII, the rate of agricultural labor and the average household farm size in Heilongjiang Province reaches 94.40 percent and 101.53 mu, which are 1.45 and 2.3 times of those in Jiangsu Province, respectively. That is, farmers in Heilongjiang rely more on agriculture and there would be more human capital and resources invested in agriculture which is the main source of the households income. Relying on agricultural income makes farmers pay more attention to advanced agricultural technology. Natural disasters increase the risks of agricultural investment and farmers have limited resources to mitigate risks when they adopt new risky agriculture technology. Large-scale farmers are faced with more risk than their smaller counterparts, arising from the considerable up-front investment. The compensation mechanism of weather index insurance alleviates the risks, and then induces farmers to take risky, yet profitable technology. Once weather index insurance is provided to farmers, it will stimulate them to invest in technology. Then the insurance has a significant impact on Heilongjiang sample which contains more large-scale farmers.

In Heilongjiang sample, 79.90 percent of them have suffered natural disasters in the past five years, which is 2.07 times of farmers in Jiangsu Province. That is to say, for farmers in Heilongjiang Province, adopting new agricultural technologies is more likely to suffer losses due to the higher incidence of natural disasters. Weather index Insurance reduces farmers’ exposure to risk, then their perception of the degree of risk would change, which can result in behavioral adaptation such as adoption of risky agricultural technology (Hill et al., 2013). The farmers’ average education experience in Heilongjiang Province is 6.9 years, while that in Jiangsu Province is 8.954 years. Different educational experiences lead to diverse acceptance of new technologies and insurance; farmers have different attitudes toward technology choices even if they are covered by insurance simultaneously, which contributes to regional differences in the impact of weather index insurance on technology adoption.

In addition to the above factors, the difference in the impact of weather index insurance on technology adoption between the two provinces may also be caused by some unobserved characteristics such as the development of insurance and credit market. In fact, the insurance and credit markets in Jiangsu Province are more developed than those of Heilongjiang Province[5]. Farmers in Jiangsu Province can more easily gain access to the financial market. Weather index insurance has less marginal utility for them; As a result, the stimulation of weather index insurance in Jiangsu Province is not as significant as that in Heilongjiang Province.

4.5 Placebo testing

The DID model avoids endogenous problems effectively, but the control group and the treatment group must meet the common trend hypothesis. That is, there may be fixed differences between the two groups but their time trends should be consistent. If there is a significant difference in time trends which are caused by non-policy factors between two groups, the results will be bias. Placebo testing was used to test the applicability of the DID model. The pre-insurance samples of the control group and the treatment group are assigned to “pseudo-control group” and “pseudo- treatment group” randomly, and then the DID model was used to estimate two groups. If the coefficient of the interaction item is significant, it means the time trends between groups are different, it is not suitable for the DID model. Otherwise, the estimation results of the DID model are reliable.

Table IX shows the interaction items of the “pseudo-processing group.” The period dummy variables are not significant in these three models, indicating that the control group and the treatment group have similar time trends and the DID model is suitable.

5. Robust test

t-test is performed to assess the robustness of the empirical results.

Table X shows that p (|T|>|t|)<0.01, which indicates there is a significant difference in technical adoption between two groups. Pr(T<t) =0.0000<0.01, indicating that the average willingness of farmers in treatment group to adopt new technologies is greater than that of the farmers in control group. It indicates the results are robust and further certifies the hypothesis that weather index insurance has a catalytic effect on the adoption of agricultural technology for rural households.

6. Conclusion and discussion

A field experiment was conducted to test the impact of weather index insurance on farmers’ improved technology adoption. DID model was used to evaluate the effect. The placebo testing and the t-test further confirmed that results are robust to various estimation methods. The findings are as follows.

First, weather index insurance has a positive impact on the adoption of improved seeds – the key inputs influencing the yields. The results show that the rate of improved technology adoption increased after the weather index insurance was provided. The number of households with weather index insurance apply improved seeds is greater than that of farmers without insurance. Second, weather index insurance has a significant positive effect on farmers’ technology adoption in Heilongjiang Province, but no significant effect on farmers in Jiangsu Province after control variables are added. Third, household farm size, gender, age, rate of agricultural labor, and weather and risk reference are important factors affecting farmers’ technology adoption. Similar results were found by Giné and Yang (2009), Carter et al. (2011), Miranda and Gonzalez-Vega (2011) and Hill and Viceisza (2012) in Africa, southeast Asia, etc.

These results have some straightforward implications from a policy perspective: first, sound risk management methods will help farmers reduce the consequences of weather risks and smooth income fluctuations so as to promote adoption of income-raising technologies. Promoting weather index insurance can improve the risk tolerance of farmers and then increase their risky investment with high return. This will improve agricultural productivity, and enhance rural households’ well-being in China.

Second, when promoting weather index insurance, regional differences should be considered. Empirical analysis shows that there are regional differences in the role of weather index insurance. It means that weather index insurance should not be unified like traditional agricultural insurance. Adequate regional investigation should be done before initiating an index insurance product pilot project. Providing different risk-sharing measures or suitable insurance contracts for different region may improve the operational efficiency of the insurance product. Weather index insurance should be preferentially promoted in major grain growing areas with a high natural risk where farmers are more willing to accept index insurance, which in turn will affect farmers’ technology adoption. It will improve productivity in the agricultural sector and ensure food security in China.

Third, because the household farm size has a significant positive effect on technology adoption, from the perspective of insurance marketing strategy, the insurance companies should target large-scale farmers first which are more dependent on farming income than others. Index insurance can help them spread risk when they adopt advanced technologies. In addition, large-scale farmers are the role models in rural communities and have a peer effect on small farmers. Small-scale farmers will wait and see the response of large-scale farmers and make decisions, and then the recognition of insurance qproducts by large farmers plays a vital role in the implementation of index insurance in rural China.

This research is subject to some limitations. First, the experimental parameters are designed according to the actual situation to simulate reality. But the willingness in the experiment does not mean it will be put into action in reality. Second, due to the diversity of China’s climate, geography and economic environment, rural households are heterogeneous in rural China. Whether the conclusion can be generalized beyond the study area is naturally questionable. Third, limited by the experimental design, unobserved characteristic and external environment variables which may cause the difference between the two provinces were not included in the model. This may cause the results unable to give a complete explanation of the index insurance impact difference. A study including more diverse samples and considering more external environmental factors is needed to gain a full understanding of index insurance’s effects, and then allows more precise policy recommendations.

The geographical distribution of the samples

Province County Number of sample towns Number of sample villages Number of sample households
Heilongjiang Mulan 4 8 75
Hulan 3 6 71
Tonghe 2 4 63
Jiangsu Guanyun 4 8 87
Jurong 3 6 48

The experiment parameters setting unit: RMB

Cost Earnings normal weather Earnings bad weather Living cost Insurance premium Insurance indemnity
Traditional seeds (No. 1) 6,000 13,000 0 5,000 200 3,000
Improved seeds (No. 2) 11,000 21,000 0 5,000 200 3,000

Source: Authors’ calculations, based on data collected by the field experiment

Risk preference test unit: RMB

Order Option A Option B (draw a ball)
 1 ¥3 Black ball: ¥0; white ball: ¥50
 2 ¥5 Black ball: ¥0; white ball: ¥50
 3 ¥7 Black ball: ¥0; white ball: ¥50
 4 ¥9 Black ball: ¥0; white ball: ¥50
 5 ¥11 Black ball: ¥0; white ball: ¥50
 6 ¥13 Black ball: ¥0; white ball: ¥50
 7 ¥15 Black ball: ¥0; white ball: ¥50
 8 ¥17 Black ball: ¥0; white ball: ¥50
 9 ¥19 Black ball: ¥0; white ball: ¥50
10 ¥21 Black ball: ¥0; white ball: ¥50
11 ¥23 Black ball: ¥0; white ball: ¥50
12 ¥25 Black ball: ¥0; white ball: ¥50

Farmers’ technical choices and differences in two groups

Traditional seeds Improved seeds Difference Traditional seeds Improved seeds Difference
Group Number of samples Ratio (%) Number of samples Ratio (%) Number of samples Ratio (%) Number of samples Ratio (%) Number of samples Ratio (%) Number of samples Ratio (%)
The first round The second round
Control 40 23.26 132 76.74 92 53.48 36 20.93 136 79.07 100 58.14
Treatment 51 29.65 121 70.35 70 40.70 46 26.74 126 73.26 80 46.52
Difference 11 6.39 −11 −6.39 −22 −12.78 10 5.81 −10 −5.81 −20 −11.62
The third round The fourth round
Control 36 20.93 136 79.07 100 58.14 33 19.19 139 80.81 106 61.62
Treatment 31 18.02 141 81.98 110 63.96 25 14.53 147 85.47 122 70.94
Difference −5 −2.91 5 2.91 10 5.82 −8 −4.66 8 4.66 16 9.32

Source: Authors’ calculations, based on data collected by the field experiment

Descriptive statistics of the main variables

All samples Control group Treatment group
Variables Definition Mean (SD) Mean (SD) Min. Max. Mean (SD) Min. Max.
Weather 1st year (disaster =0, normal=1) 0.733 (0.443) 0.75 (0.434) 0 1 0.715 (0.453) 0 1
2nd year 0.631 (0.483) 0.622 (0.486) 0 1 0.639 (0.481) 0 1
3rd year 0.765 (0.425) 0.825 (0.381) 0 1 0.703 (0.458) 0 1
4th year 0.639 (0.481) 0.587 (0.494) 0 1 0.692 (0.463) 0 1
Individual characteristics Province (H=0, J=1) 0.385 (0.487) 0.335 (0.473) 0 1 0.435 (0.497) 0 1
Age (years) 51.302 (10.017) 51.823 (10.139) 26 77 50.782 (9.895) 27 74
Gender (male=1, female=0) 0.844 (0.371) 0.829 (0.377) 0 1 0.859 (0.366) 0 1
Education (years) 7.305 (2.856) 7.059 (2.851) 0 16 7.553 (2.847) 0 16
Household production characteristics Years of planting (years) 27.938 (12.449) 28.341 (12.499) 2 55 27.535 (12.423) 2 57
Rate of agricultural labor (%) 0.854 (0.239) 0.863 (0.231) 0 1 0.845 (0.246) 0.2 1
Household farm size (mu) 79.897 (120.95) 93.860 (129.38) 1 1,200 65.933 (110.51) 2 1,100
Natural disasters If suffer natural disasters in the past five years (yes=1, no=0) 0.644 (0.479) 0.624 (0.486) 0 1 0.665 (0.473) 0 1
Risk preference Risk preference (0–12) 6.079 (4.681) 6.658 (4.541) 0 12 5.500 (4.762) 0 12

Notes: H represents Heilongjiang Province; J represents Jiangsu Province; Household farm size is farmers’ planting scale in their real life

Source: Authors’ calculations, based on data collected by the field economics experiment

Risk preference distribution

Risk-averse Risk-neutral Risk-preferring
Risk score Number (person) Ratio (%) Risk score Number (person) Ratio (%) Risk score Number (person) Ratio (%)
0 50 14.53
1 56 16.28 5 17 4.94 9 26 7.56
2 14 4.07 6 11 3.20 10 12 3.49
3 9 2.62 7 22 6.40 11 1 0.29
4 18 5.23 8 15 4.36 12 93 27.03
Total 147 42.73 65 18.90 132 38.37

Source: Authors’ calculations, based on data collected by the field economics experiment

Panel-DID model regression results of by province and pooled samples

(1) LOGIT (2) FE (3) RE
JS HLJ TOTAL JS HLJ Total JS HLJ TOTAL
Variables Coefficient Margin Coefficient Margin Coefficient Margin Coefficient Coefficient Coefficient Coefficient Coefficient Coefficient
T×I 0.394 (0.283) 0.055 (0.048) 0.827* (0.240) 0.130* (0.038) 0.476** (0.179) 0.071** (0.026) 0.091 (0.037) 0.118*** (0.029) 0.108*** (0.031) 0.062 (0.034) 0.114*** (0.027) 0.086*** (0.027)
Weather (Lag one phase) 0.320 (0.281) 0.046 (0.040) 0.522** (0.229) 0.082** (0.036) 0.415** (0.182) 0.059** (0.025) 0.045 (0.033) 0.034 (0.026) 0.034* (0.022) 0.048 (0.032) 0.044* (0.025) 0.039*** (0.025)
Province 0.301 (0.197) 0.046 (0.034) 0.024 (0.056)
Age −0.016 (0.022) −0.002 (0.003) −0.042** (0.021) −0.007** (0.003) −0.026** (0.015) −0.005** (0.003) −0.00058 (0.005) −0.006 (0.005) −0.005 (0.003)
Gender 1.378*** (0.479) 0.194*** (0.064) 0.261 (0.234) 0.039 (0.035) 0.501** (0.213) 0.069** (0.033) 0.226* (0.125) 0.050 (0.059) 0.059 (0.062)
Education 0.122*** (0.045) 0.017*** (0.006) −0.0292 (0.0388) −0.004 (0.006) 0.037 (0.029) 0.007 (0.009) 0.020** (0.009) −0.005 (0.010) 0.009 (0.011)
Years of planting −0.005 (0.014) −0.00065 (0.002) 0.139 (0.174) 0.002 (0.003) 0.007 (0.012) 0.003 (0.005) −0.001 (0.003) 0.003 (0.004) 0.004 (0.001)
Agricultural labor 0.830** (0.421) 0.117** (0.057) 0.907 (0.645) 0.014 (0.098) 0.576* (0.463) 0.086* (0.051) 0.108 (0.091) 0.024 (0.146) 0.082 (0.074)
Household farm size 0.005 (0.002) 0.00065 (0.0003) 0.002* (0.001) 0.003* (0.002) 0.004*** (0.001) 0.0005*** (0.0003) 0.00022 (0.00026) 0.00020 (0.00021) 0.003 (0.005)
If Suffered Natural disasters 0.723** (0.304) 0.102** (0.042) 0.293 (0.283) −0.044 (0.043) 0.153 (0.186) 0.031 (0.036) 0.092 (0.058) −0.049 (0.062) 0.023 (0.045)
Risk Preference 0.074** (0.029) 0.010** (0.004) 0.055** (0.025) 0.009** (0.004) 0.073*** (0.018) 0.015*** (0.009) 0.011* (0.006) 0.010* (0.006) 0.013** (0.006)
Constant −0.939 (0.372) 0.064*** (0.813) 0.723*** (0.029) 0.739*** (0.018) 0.726*** (0.021) 0.344*** 0.954*** 0.665*** (0.173)
F/Wald χ2 45.32 37.35 71.69 2.81 5.76 11.93 32.12 32.65 50.67
Prob>F/χ2 0.000 0.000 0.000 0.039 0.001 0.000 0.002 0.000 0.000

Notes: The values in brackets are standard errors. *,**,***Significant at the 10, 5 and 1 percent levels, respectively

Source: Authors’ calculations, based on data collected by the field economics experiment

Variables descriptive statistics of by province sample and t-test results

Heilongjiang province Jiangsu province t-test
Variables Mean Min. Max. Mean Min. Max.
Education (years) 6.919 (2.724) 0 16 8.954 (3.192) 0 16 −5.98***
The rate of agricultural labor (%) 0.944 (0.170) 0 1 0.655 (0.315) 0 1 18.88***
Household farm size (mu) 101.531 (119.573) 10 1,200 43.981 (113.291) 1 710 9.58***
Disasters in the past five years (yes=1; no=0) 0.799 (0.401) 0 1 0.385 (0.487) 0 1 14.90***

Notes: The values in brackets are standard deviations. We selected some scale farmers from each village due to the rapid development of large-scale farmers in Jiangsu Province, and thus the average household farm size is much larger than that of only including small farmers. *p<0.1; **p<0.05; ***p<0.01

Regression results of placebo test

Variables (1) OLS (2) FE (3) RE
T×I 0.041 (0.061) 0.041 (0.035) 0.041 (0.035)
Control variables Control Control Control
F/χ2 4.73 1.35 16.87
Prob>F/χ2 0.0000 0.2600 0.1547
T′×I 0.041 (0.061) 0.041 (0.044) 0.041 (0.044)
Control variables Control Control Control
F/χ2 6.71 14.95 53.49
Prob>F/χ2 0.0000 0.0000 0.0000

Notes: The values in brackets are standard errors. *,**,***Significant at 10, 5 and 1 percent levels, respectively. T×I denotes the interaction items of group dummy variable and period dummy variable of “pseudo-control group” and “pseudo-treatment group”; T’×I indicates the interaction items of the group and the period dummy variable of the sample before providing the insurance

Source: Authors’ calculations, based on data collected by the field experiment

t-test of weather index insurance’s impact on farmers’ technology adoption

Variable Observations Mean SE SD
Control group (A) 688 0.644 0.018 0.479
Treatment group (B) 688 0.777 0.016 0.416
Diff.=Mean (A)−Mean (B) −0.134 0.024
H0: diff.=0 t= −5.526
Ha: diff.<0 Ha: diff.=0 Ha: diff.>0
Pr(T<t)=0.0000 Pr(|T|>|t|)=0.0000 Pr(T>t)=1.0000

Source: Authors’ calculations, based on data collected by the field experiment

Notes

1.

Per capita arable land is calculated based on the data from China Statistical Yearbook 2018, it equals to the total cultivated land area of the province divided by the total population of the province.

2.

Except for following the sample selection rules of this study (see Section 3.2), we selected some scale farmers (more than 50 mu) from each village in Jiangsu Province for two reasons: First, the average planting scale of another sample area (Heilongjiang Province) is higher than the national average from the official statistics. Second, the Jiangsu Provincial Agriculture Commission has invested a lot of funds and issued many policies to encourage land transfer and farmers to expand farm scale. We selected some scale farmers from each village in Jiangsu Province so that the sample distribution can reflect the reality.

3.

This 30 percent is calculated based on experimental data. In the focus group interviews pre and post the experiment, we learned that approximately half of the farmers like to keep using the same variety they are familiar with when choosing seeds in reality. Then we believe that the effect of status quo bias was not significant in our experiment. In other words, farmers’ technology choices also have certain status quo bias in their real life.

4.

Cost of high-yield hybrid rice seeds is nearly ¥300 per mu, cost of improved compound fertilizer is ¥500 per mu, and cost of improved herbicide and pesticides is ¥150 and ¥150 per mu, respectively.

5.

In 2017, agricultural loan of Jiangsu was ¥2,827.12bn (Statistics Bureau of Jiangsu Province, “Continuous Optimization of the Total Financial Fiscal Growth Structure”, September 2018), and Heilongjiang, ¥851.83bn (Banking Regular Press Conference held by Li Lan, Deputy Director of the Banking Regulatory Bureau of Heilongjiang Province on March 15th, 2018). The insurance premium income and the total indemnity were ¥269.02bn and ¥91.51bn (Statistics Bureau of Jiangsu Province, “Continuous Optimization of the Total Financial Fiscal Growth Structure”, September 2018), respectively in Jiangsu Province; whereas those in Heilongjiang were ¥93.14bn and ¥240.5bn (Heilongjiang Province Finance Office official website, “2017 Heilongjiang Province insurance premiums more than ¥931bn to pay more than ¥24bn”, 21 March 2018) the same period.

References

Ahsan, S.M., Ali, A.G. and Kurian, N.J. (1982), “Toward a theory of agricultural insurance”, American Journal of Agricultural Economics, Vol. 64, pp. 520-529.

Asfaw, S., Bekele, S., Franklin, S. and Leslie, L. (2012), “Impact of modern agricultural technologies on smallholder welfare: evidence from Tanzania and Ethiopia”, Food Policy, Vol. 37, pp. 283-295.

Barnett, B.J. and Mahul, O. (2007), “Weather index insurance for agriculture and rural areas in lower-income countries”, American Journal of Agricultural Economics, Vol. 89, pp. 1241-1247.

Barnett, B.J., Barrett, C., B., Skees, C. and Jerry, R. (2008), “Poverty traps and index-based risk transfer products”, World Development, Vol. 36 No. 10, pp. 1766-1785.

Binswanger, H. (1980), “Attitudes toward risk: experimental measurement in rural India”, American Journal of Agricultural Economics, Vol. 62, pp. 395-407.

Bourdillon, M., Hebinck, P., Hoddinott, J., Kinsey, B., Marondo, J., Mudege, N. and Owens, T. (2002), “Assessing the impact of HYV Maize in resettlement areas of Zimbabwe”, Discussion Paper No. 161, International Food Policy Research Institute, Washington, DC.

Brick, K. and Visser, M. (2015), “Risk preferences, technology adoption and insurance uptake: a framed experiment”, Journal of Economic Behavior & Organization, Vol. 118, pp. 383-396.

Carter, M.R., Lan, C. and Alexander, S. (2011), “The impact of interlinked insurance and credit contracts on financial market deepening and small farm productivity”, working paper, Applied and Agricultural Economics Association Annual Meeting, Pittsburgh, PA.

Carter, M.R., Lybbert and Travis, J. (2012), “Consumption versus asset smoothing: testing the implications of poverty trap theory in Burkina Faso”, Journal of Development Economics, Vol. 99 No. 2, pp. 255-264.

Carter, M., Guirkinger, C. and Serfilippi, E. (2014), “Discontinuity of preferences and insurance demand: results from a framed field experiment in Burkina Faso”, working paper, University of California, Davis, CA.

Cole, S.A., Giné, X. and Vickery, J.I. (2013), “How does risk management influence production decisions? Evidence from a field experiment”, World Bank Policy Research Working Paper No. 6546, World Bank, Washington, DC, 1 July.

De Nicola, F. (2015), “Handling the weather: insurance, savings, and credit in West Africa”, World Bank Policy Research Working Paper No. 7187, San Diego, CA.

De Nicola, F. and Hill, R.V. (2013), “Interplay among credit, weather insurance and savings for farmers in Ethiopia”, presentation at the American Economic Association Meetings.

Dercon, S. (1996), “Wealth, risk and activity choices: cattle in western Tanzania”, Journal of Development Economics, Vol. 55 No. 1, pp. 1-42.

Dercon, S. and Christiaensen, L. (2011), “Consumption risk, technology adoption and poverty traps: evidence from Ethiopia”, Journal of Development Economics, Vol. 96 No. 2, pp. 159-173.

Elabed, G., Bellemare, M.F., Carter, M.R. and Guirkinger, C. (2013), “Managing basis risk with multi-scale index insurance”, Agricultural Economics, Vol. 44 Nos 4-5, pp. 419-431.

Fadhliani, Z., Luckstead, J. and Wailes, E.J. (2019), “The impacts of multiperil crop insurance on Indonesian rice farmers and production”, Agricultural Economics, Vol. 50 No. 1, pp. 15-26.

Farrin, K. and Miranda, M.J.A. (2015), “Heterogeneous agent model of credit-linked index insurance and farm technology adoption”, Journal of Development Economics, Vol. 116, pp. 199-211.

Farrin, K. and Murray, A.G. (2014), “The effect of index insurance on returns to farm inputs: exploring alternatives to Zambia’s Fertilizer subsidy program”, paper presented to Agricultural & Applied Economics Association, Minneapolis, 27–29 July.

Freudenreich, H. and Mußhoff, O. (2018), “Insurance for technology adoption: an experimental evaluation of schemes and subsidies with maize farmers in Mexico”, Journal of Agricultural Economics, Vol. 69 No. 1, pp. 96-120.

Ghosh, P., Mookherjee, D. and Ray, D. (2000), “Credit rationing in developing countries: an overview of the theory”, in Mookherjee, D. and Ray, D. (Eds), A Reader in Development Economics, pp. 383-401.

Giné, X. and Yang, D. (2009), “Insurance, credit, and technology adoption: field experimental evidence from Malawi”, Journal of Development Economics, Vol. 89, pp. 1-11.

Harrison, G.W. and List, J.A. (2004), “Field experiments”, Journal of Economic Literature, Vol. 42 No. 4, pp. 1009-1055.

Hazell, P. (1992), “The appropriate role of agricultural insurance in developing countries”, Journal of International Development, Vol. 4 No. 6, pp. 567-581.

Hill, R.V. (2008), “Understanding the investment and abandonment behavior of poor households: an empirical investigation”, IFPRI Discussion Papers, No. 00783, IFPRI.

Hill, R.V. and Viceisza, A. (2012), “A field experiment on the impact of weather shocks and insurance on risky investment”, Experiment Economics, Vol. 15 No. 2, pp. 341-371.

Hill, R.V., Hoddinott, J. and Kumar, N. (2013), “Adoption of weather-index insurance: learning from willingness to pay among a panel of households in rural Ethiopia”, Agricultural Economics, Vol. 44 Nos 4-5, pp. 385-398.

Hill, R.V., Neha, K., Nicholas, M., Simrin, M., de Nicola, F., David, J.S. and Patrick, S.W. (2019), “Ex ante and ex post effects of hybrid index insurance in Bangladesh”, Journal of Development Economics, Vol. 136, pp. 1-17.

Holt, C.A. and Susan, K.L. (2002), “Risk aversion and incentive effects”, The American Economic Review, Vol. 92 No. 5, pp. 1644-1655.

Janzen, S.A. and Carter, M.R. (2013), “After the drought: the impact of microinsurance on consumption smoothing and asset protection”, Working Paper No. w19702, National Bureau of Economic Research, Cambridge, MA.

Jensen, N., Barrett, C.B. and Mude, A. (2014), “Basis risk and the welfare gains from index insurance: evidence from northern Kenya”, MPRA Paper No. 59153, University of California, Davis, CA.

Karlan, D., Osei, R., Osei-Akoto, I. and Udry, C. (2014), “Agricultural decisions after relaxing credit and risk constraints”, The Quarterly Journal of Economics, Vol. 129 No. 2, pp. 597-652.

Kassie, M., Shiferaw, B. and Muricho, G. (2011), “Agricultural technology, crop income, and poverty alleviation in Uganda”, World Development, Vol. 39 No. 10, pp. 1784-1795.

Li, B., Zhang, J.B. and Zhang, Y.J. (2010), “An empirical study on the poor farmers’ demand and influencing factors of agricultural science and technology”, Forum on Science and Technology in China (in Chinese), Vol. 5, pp. 127-132.

Lybbert, T.J. and McPeak, J. (2012), “Risk and intertemporal substitution: livestock portfolios and off-take among Kenyan pastoralists”, Journal of Development Economics, Vol. 97 No. 2, pp. 415-426.

Lybbert, T.J., Galarza, F.B., McPeak, J., Barrett, C.B., Boucher, S.R., Carter, M.R., Chantarat, S., Fadlaoui, A. and Mude, A. (2010), “Dynamic field experiments in development economics: risk valuation in Morocco, Kenya, and Peru”, Agricultural and Resource Economics Review, Vol. 39 No. 2, pp. 176-192.

McIntosh, C., Sarris, A. and Papadopoulos, F. (2013), “Productivity, credit, risk, and the demand for weather index insurance in smallholder agriculture in Ethiopia”, Agricultural Economics, Vol. 44 Nos 4-5, pp. 399-417.

Mendola, M. (2007), “Agricultural technology adoption and poverty reduction: a propensity score matching analysis for rural Bangladesh”, Food Policy, Vol. 32 No. 3, pp. 372-339.

Miranda, M.J. and Farrin, K. (2012), “Index insurance for developing countries”, Applied Economic Perspectives and Policy, Vol. 34 No. 3, pp. 391-427.

Miranda, M.J. and Glauber, J.W. (1997), “Systemic risk, reinsurance, and the failure of crop insurance markets”, American Journal of Agricultural Economics, Vol. 79, pp. 206-215.

Miranda, M.J. and Gonzalez-Vega, C. (2011), “Systemic risk, index-insurance and optimal management of agricultural loan portfolios in developing countries”, American Journal of Agricultural Economics, Vol. 93 No. 2, pp. 399-406.

Miranda, M.J. and Vedenov, D.V. (2001), “Innovations in agricultural and natural disaster insurance”, American Journal of Agricultural Economics, Vol. 83 No. 3, pp. 650-655.

Mukasa, A.N. (2018), “Technology adoption and risk exposure among smallholder farmers: panel data evidence from Tanzania and Uganda”, World Development, Vol. 105, pp. 299-309.

Norton, M., Osgood, D., Madajewicz, M., Holthaus, E., Peterson, N., Diro, R., Mullally, C., Teh, T.-L. and Gebremichael, M. (2014), “Evidence of demand for index insurance: experimental games and commercial transactions in Ethiopia”, Journal of Development Studies, Vol. 50 No. 5, pp. 630-648.

Patt, A., Suarez, P. and Hess, U. (2010), “How do small-holder farmers understand insurance and how much do they want it? Evidence from Africa”, Global Environmental Change, Vol. 20 No. 1, pp. 153-161.

Rosenzweig, M. and Binswanger, H. (1993), “Wealth, weather risk, and composition and profitability of agricultural investments”, Economic Journal, Vol. 103 No. 416, pp. 56-78.

Skees, J., Hazell, P. and Miranda, M.J. (1999), “New approaches to crop yield insurance in development countries”, EPDT Discussion Papers No. 55.

Stein, D. (2011), “rainfall index insurance in India”, PhD thesis, the London School of Economics and Political Science.

Tuo, G.Z. (2016), “How to promote agricultural insurance in China?: Experiences, problems and solutions during 2007-2014”, China Agricultural Economic Review, Vol. 8 No. 2, pp. 194-205.

Ward, P.S. and Singh, V. (2015), “Using field experiments to elicit risk and ambiguity preferences: behavioral factors and the adoption of new agricultural technologies in rural India”, The Journal of Development Studies, Vol. 51 No. 6, pp. 707-724.

Xiao, Y.G. and Yao, J. (2018), “Double trigger agricultural insurance products with weather index and yield index”, China Agricultural Economic Review, doi: https://doi.org/10.1108/CAER-01-2018-0021.

Ye, T., Wang, M., Hu, W.Y., Liu, Y.B. and Shi, P.J. (2017), “High liabilities or heavy subsidies: farmers’ preferences for crop insurance attributes in Hunan, China”, China Agricultural Economic Review, Vol. 9 No. 4, pp. 588-606.

Zhao, Y.F., Chai, Z.H., Delgado, M.S. and Preckel, P.V. (2016), “An empirical analysis of the effect of crop insurance on farmers’ income: results from inner Mongolia in China”, China Agricultural Economic Review, Vol. 8 No. 2, pp. 299-313.

Zhu, M., Qi, Z.H., Wu, Y.L., Li, X.R. and Tang, S.Y. (2015), “An empirical analysis on the new agricultural operators’ influencing factors of agricultural technology – a case study of 395 large rice farmers in Southern Jiangsu Province, China Rural Survey (in Chinese), Vol. 1, pp. 30-38.

Zimmerman, F.J. and Carter, M.R. (2003), “Asset smoothing, consumption smoothing and the reproduction of inequality under risk and subsistence constraints”, Journal of Development Economics, Vol. 71 No. 2, pp. 233-260.

Further reading

Chu, C.H., Feng, S.Y. and Zhang, W.W. (2012), “An empirical analysis of farmers’ adoption of environmentally friendly agricultural technology – taking organic fertilizer and soil testing fertilization as example”, Chinese Rural Economy, Vol. 3, pp. 68-77.

Clarke, D., Dercon and Stefan (2009), “Insurance, credit and safety nets for the poor in a world of risk”, DESA Working Paper 81, Department of Economics, University of Oxford, Oxford.

Li, H.J. (2012), “An empirical analysis of factors influencing farmers’ willingness to adopt circular agricultural technology”, China Rural Survey (in Chinese), Vol. 2, pp. 28-36.

Zhu, M., Qi, Z.H., Luo, L.N., Tang, S.Y., Wu, Y.L. and Li, X.R. (2016), “Analysis on rice farmers’ influencing factors of agricultural technology based in probit-ISM model – a case study of 320 rice farmers in Hubei province”, Journal of Applied Statistics and Management (in Chinese), Vol. 35 No. 1, pp. 11-23.

Acknowledgements

A Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD). Humanities and Social Sciences Foundation of the Ministry of Education of China: Supply-demand Decisions and Incentive Mechanism of Pest Control Services: Based on the Perspective of Cleaner Production (18YJA790025). Nanjing Agricultural University Central Universities Fundamental Research Funds for Humanities and Social Sciences (SKCX2015011, SKJD2014001). National Natural Science Foundation Project “The Weather Index Insurance Demand and Its Influence on Farmers’ Behavior Research” (No. 71573129).

Corresponding author

Jihong Ge can be contacted at: gjh@njau.edu.cn