Estimating the average treatment effect of adopting stress tolerant variety on rice yield in China

ZHOU Jie-hong, TANG Li-qun, Xiaohua Yu,

1 Center for Agricultural and Rural Development, Zhejiang University, Hangzhou 310058, P.R.China

2 Department of Agricultural Economics and Rural Development, University of Goettingen, Goettingen 37073, Germany

1. Introduction

Climate extremes, characterized by drought and flood, have significant adverse effects on agricultural production (Liu and Chen 2000; Wanget al. 2007; Longet al. 2011) and have become a major challenge to sustainable development of agriculture (Lin 1997; Stern 2006; Mendelsohn and Dinar 2009; Pan 2011; De Salvoet al. 2013; Chen 2015; Schwan and Yu 2017). The frequency of the extremes is predicted to increase and the challenge to agriculture and international food security has been pronounced (World Bank 2013).The total area suffering from drought globally will increase between 15–44% in the future (IPCC 2014). In China, the annual average crop area suffering from drought has more than doubled since the 1950s, followed by flood (MWR 2014). Juet al. (2007) report that the direct economic losses due to meteorological disaster amount to 100 billion CNY each year, accounting for an estimated 3–6% of GDP,and the areas affected by drought and flood respectively account for 17.6 and 8.1% of the total grain acreage, while the proportions for each province respectively are about 5–19% and 2–10%. Drought and flood are the most severe weather events faced by China’s rice producers (Huanget al. 2015), especially with the frequent occurrence of climate extremes, even if the irrigation condition has been improved under current technical level, the losses of wheat,corn and rice yields are expected to be 3–7%, 1–11% and 5–12%, respectively (Juet al. 2007). Rice is the main staple food in China, which produces nearly 30% of the world’s total rice output (FAO 2014), but it is particularly vulnerable to climate extremes. It was found that a 1°C rise in average temperature during the growth period in South China would decrease paddy rice yields by 2.5–3.5% (Zhouet al.2010). The rising temperatures, especially extremely high temperatures during the ripening period, could accelerate rice growth, shorten the growth period, lead to milky white grains and immature grains, and ultimately result in a decline in rice yields (Zhouet al. 2010; Kawasaki and Uchida 2016). Historical record datasets also show that rice yield loss caused by drought increased at a rate of 4.6% over the period of 1951–2010, while the rice yield loss resulting from flood grew at a rate of 3.8% (NBSC 2012). Hence we especially shed light on rice production in this study.

Overcoming the challenge to food security caused by increasing climate extremes has drawn much attention from researchers. Recent studies have identified a variety of effective adaptation measures being taken by farmers to cope with climate change, such as diversifying crop varieties (Bradshwaet al. 2004; Bryanet al. 2009; Chenet al. 2014; Baiet al. 2015), adjusting the timing of sowing and harvesting (Smit and Skinner 2002; Challinoret al.2007; Tubielloet al. 2007; Deressaet al. 2009), increasing input use and changing plant densities (Cuculeanuet al.1999; Smit and Skinner 2002; Mezaet al. 2008; Seo and Mendelsohn 2008). Particularly, farmers appear to undertake two main types of adaptation measures, including physical measures and non-physical measures, to cope with climate extremes, such as investment in engineering adaptation measures in irrigation infrastructure (Wang 2005;Wu and Liu 2015) and farm management or other measures demanding less investment (Wanget al. 2014; Huanget al.2015). However, most studies focused on the determinants of adaptation decisions, the effectiveness of adaptation practices has not been well evaluated. For example,Deressaet al. (2009) find that household characteristics and access to extensions influence farmers’ adaptation decisions in Ethiopia. Similarly, Chenet al. (2014) indicate that farm characteristics and local government policies influence farmers’ adaptation decisions in China. Though some studies have analyzed the impact of adaptation on crop yield (Yesufet al. 2008; Di Falcoet al. 2011; Pan 2011;Chen 2015; Huanget al. 2015), yet to what degree these adaptation measures can help mitigate the impact of climate extremes remains unclear.

Variety choice is a major adaptation strategy. In other words, farmers can adopt new variety with strong resistance to reduce risks from climate extremes (Selvaraj and Ramasamy 2006). Given that the stress tolerant varieties are of shorter duration, and have ability to withstand high heat, drought, flood and other unfavorable weather conditions, crop breeding for stress-tolerance variety has attracted considerable research attention in the recent past(Lybbert and Bell 2010; Baiet al. 2015). In the case of rice,it has been reported to have a yield advantage of 5–28%over the existing varieties (Virk and Witcombe 2007; Prayet al. 2011; IRRI 2013). The adoption of excellent variety with strong tolerance is a main adaptation measure of farmers, which can mitigate the harmful effects of climate change on rice (Wu and Zhou 2004; Wang 2005). Besides,several studies have examined the factors affecting farmers’choice of seed varieties. Menget al. (2005) indicated that yield potential was the top concern when farmers in Guangxi Zhuang Autonomous Region make a decision of seed purchase, which could help maximize the profit. Yuan and Yan (2009) found that farmers’ maize seed choice behavior was heavily motivated by increasing high yield, other factors including labor, marketing, local cultivation knowledge,livelihood strategy and the awareness of risk and so on.Similarly, Cao (2011) indicated that yield potential was a major driver for adoption behavior in China, significantly related to labor force and the age of household head. These studies, however, did not make clear that to what extend farmers’ adoption behavior is affected by climate change,particularly, the increasing trend of climate extremes.

Given the increasing severity of climate extremes and the potential role of stress tolerant variety in mitigating climate risks, it is important to identify the factors influencing farmers’ adoption of stress tolerant variety, and to evaluate whether their adoption can really reduce rice yield loss. The adoption of stress tolerant variety responding to extreme climate events can be considered as an effective adaptation strategy to climate extremes, an issue which is only studied in a limited way in the current literature. Particularly, the adoption behavior could be endogenous, but it has not been well examined in the literature. Except for the study by Huanget al. (2015), in which they investigated how rice farmers adjust their farm management practices in response to extreme weather events and determine whether their adjustments affect the mean, risk, and downside risk of rice yield. Different from Huanget al. (2015) who use an endogenous switching model, we assume the technical innovation for adoption is Hicksian neutral and we also take endogenous adoption into account to estimate the average treatment effect (ATE) of stress-tolerant variety adoption by an instrumental variables (IV) regression.

Empirically, we shed light on the impact of adoption of stress tolerant varieties on rice yield in China by using a three-year panel dataset collected from 1 080 Chinese rice farmers in four major rice producing provinces in China: Zhejiang and Jiangsu in the coastal area of eastern China, Sichuan in Southwest China, and Hunan in Central China. We are particularly interested in identifying factors influencing farmers’ adoption behavior and evaluating whether their adoptions can reduce rice yield loss. The nature of panel data enables us to compare the adoption behaviors in different years in respond to different weather situations, while controlling for unobserved heterogeneities.It is methodologically superior to the cross section analysis,prevalent in the literature (e.g., Wanget al. 2009, 2013).

Rest of the paper is organized as follows. Section 2 illustrates the empirical strategy and introduces the data and sampling method used in this study. Section 3 provides estimation results for the ATE of farmers’ adoption of stress tolerant variety on land productivity, and offers some related discussions on policy implication. The final section concludes the whole research.

2. Model and data

2.1. Model specification

Basic model with exogenous participationThere are two broad streams of literature which model the impact of climate change on agricultural production. One stream is called Ricardian method, which implicitly takes into account all adaptation measures, observable or unobservable, in the impact analysis (e.g., Deschenes and Greenstone 2007;Wanget al. 2009, 2013). One stream is called the production approach, which explicitly incorporates adaptation to production process (e.g., Holstet al. 2013). The latter is more flexible, as it can analyze the direct impact of adaptations.Following Kim and Chavas (2003), Di Falco and Chavas(2009) and Holstet al. (2013), we use the production function approach with consideration of adaptation behavior. If we assume adoption of stress tolerant variety (1 for adoption and 0 otherwise) is Hicksian neutral and exogenous, following the suggestion of Wooldridge (2010, Chapter 21, p. 919), the ATE model can be specified as:

Where,yis the rice yield (kg ha–1);Ais a dummy variable denoting the adoption of stress tolerant variety (1 for adoption, and 0 otherwise), andτis the average treatment effect (ATE) for all sample.Xis a set of explanatory variables,including: 1) farm characteristics including characteristics of household head (gender, years of education and years of plant experience), agricultural labor, and type of rice planted(single-seasoned and double-seasoned); 2) production inputs (labor, land, fertilizer and pesticide, machinery and other inputs) specified in logarithm; 3) year dummies for 2013 and 2014 to control for the effects related to time,such as technological change.Xis the ground mean value ofXfor all sample.βandγare two vectors of parameters to be estimated.µis the error term that captures measure errors, unobserved heterogeneities, and uncertainties, and satisfiesE(µ)=0.

The treatment effect of adoption can be depicted in Fig. 1.Given the same input, ATE is the pure increase in output.

ATE with endogenous participationThe above ATE model has a strong assumption that the adoption behavior is exogenous, which, however, might not be true in this context. The adoption behavior, which is linked to climate extremes, could be endogenous. It violates the “ignorability”requirement of ATE model (Wooldridge 2010, Chapter 21).Following the suggestion of Wooldridge (2010, Chapter 21,p. 943), we could use an instrument variable regression to correct the bias.

First, given instrumental variablesZ, we run a probit model

In this study, we use the variable of “local access to public services on new rice variety” as the instrument. We argue that this variable is significantly correlated with the adoption behavior but not affect the production behavior.Weak instrumental will be tested as well.

Note that all standard errors are heteroscedasticityrobust standard errors, in order to correct trivial effect on the standard errors (Wooldridge 2010, Chapter 21, p. 943).

Finally, the marginal effect for input variables in eq. (1) is:

Eq. (3) shows that the marginal effects for adopters and non-adopters are different, and they areβ+γandβ, respectively. The differenceγis exactly the average treatment effect of adopting stress-tolerant variety.

2.2. Data and sampling methods

We use a stratified sampling method to select rice farms in order to make the samples more representative. Rice in China is mainly planted in the Northeast Plain, the Yangtze River Basin and Southeast coastal area, respectively,accounting for 12, 64, and 22% of the national cultivating area. Heilongjiang in the Northeast region; Hunan, Hubei,and Jiangxi in the Central region; Jiangsu, Zhejiang, Anhui,Guangxi, and Guangdong in East region and Sichuan,Yunnan in Southwest region are the 11 major provinces(autonomous region) of rice production, together accounting for over 80% of the national total production (NBSC 2015).Climate change has impact on these major rice production areas at various degrees. For instance, the potential rice output in Northeast China may increase due to global warming, but the yields in other three major regions might be adversely affected (Tanget al. 2000). Therefore, taking full consideration of regional crop production systems and climate situations, we selected four provinces from the three major regions with high risk of rice yield loss: Zhejiang and Jiangsu in the coastal area of eastern China, Sichuan in Southwest China, and Hunan in Central China. We then conducted a large-scale household survey regarding the impact of adaptation to climate change on rice production during the period from October 2014 to May 2015.

From each of the provinces selected, six counties are randomly chosen following three standards. First,we identified counties that had experienced at least one episode of climate extremes (low, moderate, severe) year in the past three years (2012, 2013 or 2014). According to China’s national standard for natural disasters (CMA 2004),the severity of climate extremes has three categories: low(10–30% of yield loss), moderate (30–50% of yield loss)and severe (greater than 50% of yield loss). Second, we only kept those which had experienced one normal year of weather in the past three years and randomly selected six counties from each province. Then three towns are randomly selected from the chosen counties based on the condition of agricultural production infrastructure of ‘good’, ‘medium’and ‘poor’, respectively. Finally, three villages are randomly selected from these towns and 15 households were randomly selected from each chosen village for face-to-face interviews.Therefore, a total of 1 080 rice farms from 72 villages in 24 counties were interviewed. Excluding the incomplete samples, the final sample used in our analysis includes 1 057 households from 68 villages in 24 counties (Table 1).

The information collected in the survey include: 1)characteristics of households and farms (e.g., gender,education and years of experience of household head,agricultural labor, rice type); 2) detailed rice production cost information (e.g., land, labor, fertilizer and pesticide,machinery service, other inputs); 3) rice yield; 4) farmers’adoption behavior for stress tolerant variety in the past three years (2012–2014); and 5) availability of public services related to the extension and technical guidance for new rice variety which was collected in the village level survey.

Table 2 provides a description statistic for variables included in the empirical models. Of the 1 057 farms, most of household heads are male-dominated and relatively lowlevel educated (middle school or below), but have rich rice plant experiences (an average of 20 years). Most of farms are dominated by single-seasoned rice. Each farm has two agricultural labor forces on average.

The average rice yield is 6 026.01 kg ha–1, lower than the 2014 national average rice yield, which is 7 274 kg ha–1,indicating the substantial yield loss caused by the occurrence of climate extremes in the past three years, or our sample may not represent the whole nation. The average cost of labor, land, chemical fertilizers and pesticides, mechanical service and other inputs is approximately 1 122 CNY d–1,3 671 CNY ha–1, 3 606 CNY ha–1, 2 513 CNY ha–1and 422 CNY ha–1, respectively. Particularly, labor cost, land cost, chemical fertilizers and pesticides cost are the three highest, consistent with the clear upward trend of the three costs in recent years. However, only one-fourth of rice farms in our study can access to the public service related to the extension and technical guidance for new rice variety at village level, suggesting that the current public services are generally low and there is still much room to improve.

Table 1 Distribution of surveyed rice farms

3. Results and discussion

As shown in Table 3, the second column in Table 3 reports the estimation results for the adoption eq. (2), which is a probit model helping explain why some farmers adopt stress tolerant variety and others not. The third and fourth columns present, respectively, the estimated coefficients of rice yield functions (1) by taking or not taking into account the endogeneity of adoption behavior (with and without the instrument variable). As expected, most of the coefficients are consistent with our expectations and the current literature (e.g., Huanget al. 2008, 2015; Holstet al. 2013;Baiet al. 2015).

3.1. The factors affecting farmers’ adoption behavior

In the results of adoption eq. (2), we are particularly interested in the effects of different severities of climate extremes on farmers’ adoption decision. Though some previous studies (e.g., Di Falcoet al. 2011) did not find strong relationship between climate change variables and farmers’ adaptation decisions, we have different evidence.The coefficients for low, moderate and severe climate extremes are 0.382, 0.971, and 1.716, and all statistically significant. It is consistent to our common sense that rice farmers are more likely to adopt stress tolerant variety when they have experience of suffering from more serious climate extremes. This finding is also consistent with the study by Huanget al. (2015) in which more farmers were found to adjust their farm management practices in severe drought and flood years than in normal years.

Regarding various inputs, it is interesting to find that there are only two inputs variables which are significant:land, fertilizer and pesticide, suggesting that the more inputs on land and fertilizer and pesticide in rice production, the lower likely to adopt the new variety. It is possible that due to the constant budget constraint and uncertainties, the adoption of new variety, which is usually more expensive and more risk, would certainly reduce the possibility of farmers’ adoption on stress tolerant variety. Particularly,farmers’ adoption is found significantly negative to fertilizer and pesticide cost and its value is up to –2.529, which is different from the finding that fertilizer input has positive effect on the likelihood of adaptation (Huanget al. 2015).However, both the variables of labor cost and machinery cost are statistically insignificant, contrary to the study by Huanget al. (2015).

Household characteristics could affect the adoption behavior. According to the estimation results, agricultural labor forces have significant and positive effects on the probability of adopting the new variety. It is understandable that more labor force, whose cost is rising rapidly in recent years, indicates stronger ability against climate extremes,so that farmers have higher motivation to adopt stress tolerant variety. The negative sign of household head plant experience in agriculture indicates that farmers who have more years of experience in rice production are less likely to adopt the new variety. It is possible that the new variety demands new knowledge for planting, while the experience accumulated from the old variety may not work. Surprisingly, other variables such as the gender of household head and the education of household head are not statistically significant, contrary to the findings that male head of households is negatively correlated with farmers’adoption decisions (Baiet al. 2015; Huanget al. 2015),highly educated people tend to take adaptation measures to improve rice yield (Huanget al. 2015).

For farm characteristics, the estimated coefficient for double-seasoned rice is positive and statistically significant,suggesting that farmers with doubled-seasoned rice are more likely to adopt stress tolerant variety, this may be mainly determined by rice’s labor-intensive characteristics.Farmers who adopt new variety could have more time to work in city or spend more efforts on rice cultivation and farm management.

The coefficients for the year dummies of 2013 and 2014 are –0.117 and 0.057, respectively, though they are not statistically significant, while the value of which indicate an increasing likelihood of adoption behavior, confirming that year dummy variables have positive effects on adaptation in past three years (Huanget al. 2015).

Finally, we take the estimated coefficient for the instrument variable (IV)- local access to public services on new rice variety. As an instrumental variable, it should be correlated with selection behavior, but not the error terms in the output function. We argue that the variable of “local access to public services on new rice variety” is significantlycorrelated with the adoption behavior but not affect the production behavior. The estimated value is 2.932 and statistically significant at 0.1% level and the result implies that local access to public services on new rice variety could help increase the likelihood of farmer adoption. That is, the IV is not a weak instrument.

Table 2 Descriptive statistics of the sample (N=1 057)

Table 3 Estimations of farmers’ adoption on stress tolerant variety and its impact on rice yield

3.2. The effects of farmers’ adoption on rice yield

To evaluate the effects of adoption on rice yield, we first used at-test to examine whether there is a significant difference between the average rice yield for adopter and non-adopter farmers. The average output value of rice cultivation for adopter farmers is relatively higher than that of the farmers without adoption at a 0.1% confidence level(Appendix A). We then used an ATE model to evaluate the effects of adoption behavior on rice yield. We also applied the IV regression model to effectively identify the effects of adoption on rice yield by taking into account the endogeneity of adoption behavior. In yield equations, most estimated coefficients are statistically significant with expected signs.The estimated coefficients of yield functions without and with the instrument variables are reported in the third and fourth columns in Table 3, respectively. As expected, the signs in both OLS and the IV regressions are very similar,though there are some magnitude differences. However,we use the Hausman test to test the endogeneity problem.The value ofχ2test is 133.54 and statistically significant at the 0.1% level, indicating that we can reject the null hypothesis of no systematic difference between OLS and IV regressions. We hence confirmed the endogeneity of adoption behavior. Because the IV regression is consistent,while the OLS is not, so the following discussion is based on the IV regression results.

First, the most important parameter of our interest is the estimated coefficient for adoption choice, which is 0.155 at the significance of 0.1% level. It empirically suggests that the yield for the adoption of stress tolerant variety could be 15.5% higher than that for the non-adopters, when facing the same severity of climate extremes and other conditions being equal.

The coefficients for low, moderate and severe climate extremes are –0.686, –1.594 and –1.897. They are all statistically significant. It makes sense that more severe climate extremes bring more yield loss for non-adopters.According to eq. (3), the marginal effects for low, moderate and severe climate extremes are –0.043, –0.045 and –0.878.It implies that the impact of climate extremes on adopters is significantly smaller, and specifically, the effects of low and moderate climate extremes now basically are very small.

Regarding the coefficients for different inputs, we find that no statistically significant terms either for non-adopters or for adopters (interaction terms). It is possible that rice yield in China has reached a very high level. Further increase in inputs will not help rice output substantially, and future increase of rice output in China will mainly depend on the improvement of productivity and efficiency since agricultural land is diminishing due to urbanization. In addition, there are no significant interaction terms between different inputs and treatment variables, which also confirm that Hicksian technology assumption for the adoption of stress-tolerant varieties holds.

The coefficient for agricultural labor is –0.025 and also statistically significant, which represents the marginal effect for non-adopters, while the marginal effect for adopters is 0.020. It implies that rice farms with conventional variety have abundant labors, while the farms adopting stress-tolerant varieties demand more agricultural labors in the family. The former might be more traditional and conservative smallholders, while the latter are more likely to have better knowledge about and are willing to adopt new technologies.

Finally, the coefficients for the two year dummies are 0.197 and 0.195, respectively for 2013 and 2014, and also statistically significant. As the interaction terms are –0.158 and –0.169, the marginal effect for the adopters are 0.039 and 0.026. It implies that the yield for 2013 and 2014 is significantly higher than that for 2012. In addition, the yield increase is smaller for adopters. It also makes sense that a major risk for rice production is climate extremes but adoption of stress tolerant variety could offset the windfall effect of extreme weather conditions to some degree.

4. Conclusion

Using a panel data survey from 1 080 rice farmers conducted in four provinces in China, this article investigates the contribution of adopting stress tolerant variety in response to extreme climate to the rice yield loss reduction (or yield increase). Different from the current literature, we assume Hicksian neutral technology of adopting stress-tolerant variety and employ an IV regression to estimate the average treatment effect (ATE) on rice yield for adopting stress-tolerant variety by taking into consideration of the endogeneity of adoption behavior.

The results of adoption behaviors reveal that farmers’adoption decision of stress tolerant rice variety mainly depends on the severity of climate extremes, local access to public service on new variety, the number of labor forces, and the cost of inputs. The former three factors could incentivize farmers to adopt the new variety, while the cost of inputs for rice production discourages farmers to adopt.

More importantly, we find that farmers who adopted the stress tolerant variety increased rice yield by 15.5% in the IV regression. It suggests that adopting stress tolerant variety could generally increase rice production and contribute to the reduction in rice yield loss.

The estimation results also indicate that the possible benefit of adopting stress tolerant variety for non-adopters is much smaller than those adopters. Adoption of new variety demands more agricultural labor forces, more new knowledge, more intensive management and higher seed costs, thus the benefit might not overcome the learning costs and adoption costs (Yu and Zhao 2009). Therefore, further expansion of the stress tolerant rice variety calls for more government action on extension services.

In addition, we find that the output elasticities for all physical inputs are very small in terms of point estimation,and none is statistically significant. It implies that further increases of these inputs would have very small effect on expansion of rice output, as the rice yield in China has reached very high level in the world. In addition, no significant interaction terms between adoption choice and physical inputs implies that our assumption of Hicksian neutral technology is valid.

However, famers’ adoption of stress tolerant rice variety to a certain degree may be affected by some other factors,such as different seed varieties, market price, and access to new varieties. However, we cannot empirically test this point because the data used in this study lack the information about the specific stress tolerant rice variety information.These will be definitely studied in the future.

Acknowledgements

The research is supported by the National Natural Science Foundation of China (NSFC, 71773109 and 71273234)and the Key Project of the Ministry of Education of China(16JJD63007).

Appendixassociated with this paper can be available on http://www.ChinaAgriSci.com/V2/En/appendix.htm

Bai J F, Xu Z, Qiu H G, Liu H. 2015. Optimising seed portfolios to copeex antewith risks from bad weather: Evidence from a recent maize farmer survey in China.Australian Journal of Agricultural and Resource Economics, 59, 242–257.

Bradshaw B, Dolan H, Smit B. 2004. Farm-level adaptation to climatic variability and change: Crop diversification in the Canadian prairies.Climatic Change, 67, 119–141.

Bryan E, Deressa T T, Gbetibouo G A, Ringler C. 2009.Adaptation to climate change in Ethiopia and South Africa:Options and constraints.Environmental Science and Policy,12, 413–426.

Cao R. 2011. Empirical study on the factors influencing rape framers’ adoption of new variety.Economic Perspective, 8,146–147. (in Chinese)

Challinor A J, Wheeler T R, Craufurd P Q, Ferro C A, Stephenson D B. 2007. Adaptation of crops to climate change through genotypic responses to mean and extreme temperatures.Agriculture,Ecosystems & Environment, 119, 190–204.

Chen H, Wang J X, Huang J K. 2014. Policy support, social capital and farmers’ adaptation to drought in crop production of China.Global Environmental Change, 24, 193–202.

Chen S. 2015. The impact of climate change on the wheat productivity: Evidence from Huang-Huai-Hai Plain.China Rural Economy, 7, 4–16. (in Chinese)

CMA (China Meteorological Administration). 2004.Trial Procedures for the Early-warning Signal Issuance of Unexpected Meteorological Disasters. China Meteorological Press, Beijing. (in Chinese)

Cuculeanu V, Marica A, Simota C. 1999. Climate change impact on agricultural crops and adaptation options in Romania.Climate Research, 12, 153–160.

Deressa T T, Hassan R M, Ringler C, Alemu T, Yesuf M. 2009.Determinants of farmers’ choice of adaptation methods to climate change in the Nile Basin of Ethiopia.Global Environmental Change, 19, 248–255.

Deschenes O, Greenstone M. 2007. The economic impacts of climate change: Evidence from agricultural output and random fluctuations in weather.The American Economic Review, 97, 354–385.

Di Falco S, Chavas J P. 2009. On crop biodiversity, risk exposure and food security in the highlands of Ethiopia.American Journal of Agricultural Economics, 91, 599–611.

Di Falco S, Veronesi M, Yesuf M. 2011. Does adaptation provide food security? A micro perspective from Ethiopia.American Journal of Agricultural Economics, 93, 829–846.

FAO. 2014. FAOSTAT. Statistics Database of Food and Agriculture Organization of the United Nations, Rome.

Freedman D A. 2006. On the so-called “Huber Sandwich Estimator” and “Robust Standard Errors”.The American Statistician, 60, 299–302.

Holst R, Yu X, Grün C. 2013. Climate change, risk and grain yields in China.Journal of Integrative Agriculture, 12,1279–1291.

Huang J, Hu R, Cao J, Rozelle S. 2008. Training programs and in-the-field guidance to reduce China’s overuse of fertilizer without hurting profitability.Journal of Soil and Water Conservation, 63, 165–167.

Huang J, Wang Y, Wang J. 2015. Farmers’ adaptation to extreme weather events through farm management and its impacts on the mean and risk of rice yield in China.American Journal of Agricultural Economics, 97, 602–617.

IPCC (Intergovernmental Panel on Climate Change).2014.Climate Change 2014:Impacts,Adaptation,and Vulnerability.Contribution of Working Group II to the Fifth Assessment Report of the Inter-Governmental Panel on Climate Change. Cambridge University Press, UK.

IRRI (International Rice Research Institute). 2013.Cluster Demonstrations of Stress Tolerant Rice Varieties in Stress Prone Parts of India.Annual Report Submitted to National Food Security Mission, Ministry of Agriculture, Government of India. International Rice Research Institute, New Delhi Office, India.

Ju H, Xu Y, Xiong W. 2007. The impacts of climate change on agriculture in China.Environment Protection, 6A, 71–73.(in Chinese)

Kawasaki K, Uchida S. 2016. Quality matters more than quantity: Asymmetric temperature effects on crop yield and quality grade.American Journal of Agricultural Economics,98, 1195–1209.

Kim K, Chavas J P. 2003. Technological change and risk management: An application to the economics of corn production.Agricultural Economics, 29, 125–142.

Lin E. 1997.Simulation Analysis of The Impacts of Global Climate Change on Agricultural Production in China. China’s Agricultural Science and Technology Press, Beijing. (in Chinese)

Liu M, Chen B. 2000. The analysis on correlativity grain yield fluctuation and its relation with agricultural natural disasters in China in recent years.Journal of Catastrophology, 15,78–85. (in Chinese)

Long F, Yang Z, Peng L. 2011. The empirical study on the impacts of natural disaster on grain production: The case of rice production in China.China Rural Economy, 5, 33–44.(in Chinese)

Lybbert T J, Bell A. 2010. Stochastic benefit streams, learning,and technology diffusion: Why drought tolerance is not the new Bt.AgBioForum, 13, 13–24.

Mendelsohn R, Dinar A. 2009.Climate Change and Agriculture:An Economic Analysis of Global Impacts,Adaptation,and Distributional Effects. Edward Elgar Publishing, Cheltenha.

Meng X, Rao J, Ye J. 2005. Research on the determinants that affect farmers’ selection behavior on new crop variety.Journal of Agrotechnical, 1, 20–26. (in Chinese)

Meza F J, Silva D, Vigil H. 2008. Climate change impacts on irrigated maize in Mediterranean climates: Evaluation of double cropping as an emerging adaptation alternative.Agricultural Systems, 98, 21–30.

MWR (Ministry of Water Resources, People’s Republic of China). 2014.Bulletin of Flood and Drought Disaster in China 2014. China Water Power Press, Beijing. (in Chinese)

NBSC (National Bureau of Statistics in China). 2012.China Statistical Yearbook 2012. China Statistical Press, Beijing.(in Chinese)

NBSC (National Bureau of Statistics in China). 2015.China Statistical Yearbook 2015. China Statistical Press, Beijing.(in Chinese)

Pan G, Gao M, Hu G. 2011. Impacts of climate change on agricultural production of China.Journal of Agro-Environment Science, 30, 1698–1706. (in Chinese)

Pray C, Nagarajan L, Li L, Huang J, Hu R, Selvaraj K N,Napasintuwong O, Babu R C. 2011. Potential impact of biotechnology on adaption of agriculture to climate change: The case of drought tolerant rice breeding in Asia.Sustainability, 3, 1723–1741.

De Salvo M, Raffael R, Moser R. 2013. The impact of climate change on permanent crops in an Alpine region: A Ricardian analysis.Agricultural Systems, 118, 23–32.

Schwan S, Yu X. 2017. Social protection as a strategy to address climate-induced migration.International Journal of Climate Change Strategies and Management.https://doi.org/10.1108/IJCCSM-01-2017-0019

Selvaraj K N, Ramasamy C. 2006. Drought, agricultural risk and rural income: Case of a water limiting rice production environment, Tamil Nadu.Economic and PoliticalWeekly,41, 2739–2746.

Seo S N, Mendelsohn R. 2008. Measuring impacts and adaptations to climate change: A structural Ricardian model of African livestock management.Agricultural Economics,38, 151–165.

Smit B, Skinner M W. 2002. Adaptation options in agriculture to climate change: A typology.Mitigation and Adaptation Strategies for Global Change, 7, 85–114.

Stern N H. 2006.The Economics of Climate Change:The Stern Review. Cambridge University Press, London.

Tang G, Li X, Guenther F, Sylvia P. 2000. Climate change and its impacts on China’s agriculture.Acta Geographica Sinica,55, 129–138. (in Chinese)

Tubiello F N, Rosenzweig C, Goldberg R A, Jagtap S, Jones J W. 2007. Effects of climate change on US crop production:Simulation results using two different GCM scenarios. Part I: wheat, potato, maize, and citrus.Climate Research, 20,259–270.

Virk S D, Witcombe J R. 2007. Trade-offs between on-farm varietal diversity and highly client-oriented breeding: A case study of upland rice in India.Genetic Resources and Crop Evolution, 54, 823–835.

Wang C, Lou X, Wang J. 2007. Influence of agricultural meteorological disasters on output of crop in China.Journal of Natural Disasters, 16, 37–43. (in Chinese)

Wang J X, Huang J K, Yan T T. 2013. Impacts of climate change on water and agricultural production in ten large river basins in China.Journal of Integrative Agriculture, 12, 1267–1268.

Wang J X, Mendelsohn R, Dinar A, Huang J K, Rozelle S,Zhang L. 2009. The impact of climate change on China’s agriculture.Agricultural Economics, 40, 323–337.

Wang S. 2005. The impacts of future climatic change on agricultural production and corresponding countermeasures.Journal of Liaoning Academy of Education Administration,22, 126–128. (in Chinese)

Wang Y J, Huang J K, Wang J X. 2014. Household and community assets and farmers’ adaptation to extreme weather event: The case of drought in China.Journal of Integrative Agriculture, 13, 687–697.

Wooldridge J M. 2010.Econometric Analysis of Cross Section and Panel Data. The MIT Press, USA.

World Bank. 2013.Turn Down the Heat:Climate Extremes,Regional Impacts,and the Case for Resilience.A Report for the World Bank by the Potsdam Institute for Climate Impact Research and Climate Analytics. The World Bank,Washington D.C., USA.

Wu C Y, Liu F F. 2015. The research of rice farmers’ behavior of adopting engineering adaptation measure to cope with floods under the background of climate change.Journal of Agrotechnical Economics, 3, 15–24.

Wu Z, Zhou Z. 2004. Effects of climatic changes on China’s agriculture and corresponding countermeasures.Journal of South China University of Tropical Agriculture, 10, 7–11.(in Chinese)

Yesuf M, Di Falco S, Deressa T, Ringler C, Kohlin G. 2008.The Impact of Climate Change and Adaptation on Food Production in Low-Income Countries:Evidence from the Nile Basin,Ethiopia. International Food Policy Research Institute, Washington D.C., USA.

Yu X, Zhao G. 2009. Chinese agricultural development in 30 years: A literature review.Frontiers of Economics in China,4, 633–648. (in Chinese)

Yuan J, Yan Q. 2009. Research on factors of influencing farming household’s acceptance for hybrid maize.Journal of Anhui Agricultural Science, 14, 6651–6652. (in Chinese)

Zhou S D, Zhou W K, Zhu H G, Wang C X, Wang Y. 2010.The impact of climate change on agriculture and the countermeasures.Journal of Nanjing Agriculture University(Social Science Edition), 10, 34–39. (in Chinese)