Adaptation to Climate Change and its Impacts on Wheat Yield: Perspective of Farmers in Henan of China

This paper explored farm households’ autonomous climate change adaptation strategies and corresponding impacts on wheat yield. Based on a survey of 314 wheat farmers in rural China, results show that Chinese wheat farmers have a high rate of climate change awareness and adoption of climate change adaptation measures. Farmers’ cultivated area, cognition level and information accessibility on climate change significantly affect their adaptation decisions. However, these farmers are given limited adaptation strategies, mainly including increasing irrigation, and using more chemical fertilizer and pesticides. Through employing a simultaneous equations model with endogenous switching, we find farmers’ adaptation to climate change is maladaptive with negative effects on wheat yield. This study, therefore, suggests policymakers be mindful of farmers’ maladaptive responses to climate change and provide effective adaptation measures, to help farmers cope with the risks of climate change and ensure farmer’s livelihood security and sustainable agriculture development.


Introduction
Previous research has indicated that climate change, especially the increase of extreme weather phenomena, has a profoundly adverse effect on agriculture production [1][2][3][4][5][6][7]. A recent study shows that depending on the severity of climate change conditions, the average yield of barley wheat could fall from 3% to 17% worldwide [8]. How to overcome the risks of climate change on agriculture has gained extensive attention from researchers all over the world [2]. There are two main adaption categories, which are actors-private adaptation and public adaptation [9]. Private adaptations are autonomous adaptation-practices made by households, and public adaptations are planned adaptation-decisions made by governments [10].
Climate change is hitting farmers the hardest of all, especially in the least developed and developing countries [1,5]. Farmers autonomously adjust their production practices in response to climate change impacts depending on their existing knowledge and technology [11,12]. Several studies have researched the determinants of farmers' adaptation decisions and measures to enhance their adaptive capacity [13][14][15][16]. Chen et al. (2014) found that, in China, farm characteristics and local governments' policies are key factors affecting farmers' climate change adaptation decisions.
This study focuses on Henan province due to its significance in China's cereal production. Located in central China, Henan Province is one of the thirteen major grain cultivating areas. Based on relevant statistics, Henan Province's cultivated land area and grain yield are both ranked second in the country [22]. Henan Province is located in the junction area of the subtropical and warm temperate zone, with clear climate transitional characteristics [21].
This study is based on a survey conducted on 360 farm households located within six traditional cropland zones in Henan province in January 2016. Rejecting the invalid ones, 314 questionnaires were collected with a return rate of 87.2%. The sampling procedure consisted of five steps. First, six cities (AnYang, KaiFeng, XuChang, ZhouKou, NanYang, XinYang) were selected based on geographical and climatic distribution characteristics, scattered in the east, south, west, north and central of Henan. Second, each county was randomly chosen from each city, so six counties (NeiHuang, LanKao, ChangGe, XiHua, DengZhou, Xi) were selected. Third, we randomly selected three rural towns from each county. Fourth, one sub-district division (village) was randomly selected from each town (see Figure 1). Finally, 20 farmers were randomly selected from each village from a list of farmers collected from the village committee.
In order to collect information concerning farmers' characteristics and their actual adaptation strategies, a pre-tested structured questionnaire was used. We asked each farmer the following three contingent questions to ensure that their production adjustments were actual responses to climate change not due to other pressures [12]. In order to collect information concerning farmers' characteristics and their actual adaptation strategies, a pre-tested structured questionnaire was used. We asked each farmer the following three contingent questions to ensure that their production adjustments were actual responses to climate change not due to other pressures [12].

Modelling Adaptation to Climate Change and Wheat Yield
Following Di Falco et al. (2011), a climate change adaptation decision and its effects on wheat yield can be simulated using a two-stage framework [12,19,23,24].
In the first stage, we employed a selection model for climate change adaptation decisions. It assumes that the risk-averse farmer household will implement climate change adaptation strategies if it generates net benefits, and the net benefits can be represented by a latent variable A*.
Farm i will choose to adopt climate change adaptation strategies (Ai = 1) if Ai * > 0, and 0 otherwise. The vector Z represents variables that influence farmers' adaptation decisions. According to empirical literature on the determinants of farmers' climate change adaptation decisions [12,15,18], this study chose farm household characteristics and climate information provided by extension agents as the dependent variables. Household characteristics include gender, age, education, labor share, the area under cultivation and climate cognition. The information from government mainly included weather warnings about frost and drought.
In the second stage, the effect of adaptation on wheat yield was modeled via the production technology. The simplest approach would have been to apply the ordinary least squares (OLS)

Modelling Adaptation to Climate Change and Wheat Yield
Following Di Falco et al. (2011), a climate change adaptation decision and its effects on wheat yield can be simulated using a two-stage framework [12,19,23,24].
In the first stage, we employed a selection model for climate change adaptation decisions. It assumes that the risk-averse farmer household will implement climate change adaptation strategies if it generates net benefits, and the net benefits can be represented by a latent variable A*.
Farm i will choose to adopt climate change adaptation strategies (A i = 1) if A i * > 0, and 0 otherwise.
The vector Z represents variables that influence farmers' adaptation decisions. According to empirical literature on the determinants of farmers' climate change adaptation decisions [12,15,18], this study chose farm household characteristics and climate information provided by extension agents as the dependent variables. Household characteristics include gender, age, education, labor share, the area under cultivation and climate cognition. The information from government mainly included weather warnings about frost and drought.
In the second stage, the effect of adaptation on wheat yield was modeled via the production technology. The simplest approach would have been to apply the ordinary least squares (OLS) method, taking adaptation as a dummy variable in the food productivity equation. However, assessing the impact of adaptation on wheat yield through the OLS approach may have created many potential problems. For example, adaptation may be potentially endogenous, which if true will lead to biased estimates [18]. In addition, problems such as sample selection bias and inconsistent estimates might rise and confound the results [12].
According to Di Falco et al. (2011), a simultaneous equation model of climate change adaptation and its impact on wheat yield with endogenous transformation was estimated with full information maximum likelihood. In our study, the variables regarding climate cognition and climate information were used as selection instruments. Table A1 shows that climate cognition and climate information significantly affected farmers' adaptation decision, but they have no significant impact on the wheat yield of non-adaptors. Therefore, they can be considered as valid selection instruments.
In this paper, the endogenous switching regression model was selected to estimate the impact of climate change adaptation on wheat yield. Adaptors and non-adaptors have different yield functions.
where y 1i and y 0i are the wheat quantity produced per hectare specified in log for adopters and non-adopters respectively. x i is a vector of inputs specified in log form (e.g., seeds, fertilizers, manure, labor), β is the parameter vector to be estimated, and ε is the error term. and Equations (7) and (8) represent the actual expectations observed in the sample. Equations (9) and (10) represent the counterfactual expected outcomes. In addition, the average treatment effect on the treated (TT) can be calculated using the difference between Equations (7) and (9). Similarly, the difference between Equations (10) and (8) can be calculated as the average treatment effect on the untreated (TU) for the household that actually did not adapt. For the group of "adopters", the effect of base heterogeneity is the difference between Equations (7) and (10). Similarly, for the "non-adopters" group, the effect of base heterogeneity is measured using the difference between (9) and (8).
More details about the endogenous switching regression model can be found in Di Falco et al. (2011). Table 1 shows definitions and descriptive statistics of the surveyed farmer households. It shows that, on average, 91.7% of the farmers perceived climate changes and 82.8% of households adopted adaptation strategy in response to the changes for wheat cultivation; and 43% of households received external climate change information. Farmers adopted several measures to adapt their wheat crop to climate change. Overall, the major strategies were to increase the frequency and amount of irrigation, increase the use of chemical fertilizer and pesticides, and change crop variety. Moreover, about 63.3% of the farmers adopted more than one adaptation strategy and 2.3% of the households adopted more than three adaptation strategies. Overall, farmers were informed of the information about temperature rise and rainfall decrease at the research sites. More details about the farmer's perception of climate change, the impacts of climate change on wheat production and farmer's adaptation practices are presented in Figure A1. In addition, we collected detailed production data in different production stages. Labor input was classified by household labor and employment. The household average wheat planting area was 0.771 hectare, and the average wheat yield was 6827 kg/hectare, which was above the national average at 5471 kg/h [20]. Major inputs from farmers were chemical fertilizer, household labor, and machinery, and they had low rental and employment expenses. The average age of household heads was 55 years old, and sixty percent of them received more than nine years' education.

Descriptive Statistics
In this study, farmers that adopted at least one adaptation action were termed "adopters" and those not adopting any strategy as "non-adopters". Table 2 shows differences in household characteristics between adopters and non-adopters. The average wheat yield for non-adopters was significantly higher than that of adopters. It is also evident that some inputs, such as the cost of employment expense and irrigation, was significantly higher for non-adopters than that of adopters. However, adopters had a higher perception of climate change and corresponding impact on wheat production, and access to climate change information.

Estimates of Climate Change Adaptation and Wheat Yield Equations
The endogenous switching regression model estimates adaptation decisions and yield outcome equations jointly [12]. Table 3 shows the results of endogenous switching regression.  Note: *, ** and *** represent the statistically significant at 10%, 5% and 1%, respectively; t-value in parenthesis.
The second column of Table 3 presents the estimated results of the adaptation selection equation representing the determinants of adopting climate change adaptation. The coefficient of area is positive and statistically significant, suggesting that farmers with larger cultivation area were more likely to employ climate change adaptation strategies. The effects of climate cognition and climate information were both positive and statistically significant, indicating that farmers who are aware of climate change and could obtain information about climate change were more likely to adapt to climate change.
The estimates presented in the third and fourth column of Table 3 account for the endogenous switching in the wheat yield function. The estimated coefficients of the correlation coefficients ρ 0 or ρ 1 are both not significantly different from zero, indicating there may be no sample selectivity bias in the sample [18]. However, the differences in the coefficients of the wheat yield equation between the adopters and non-adopters suggest heterogeneity in the sample [6,18]. The results in Table 3 indicate that area is an important factor in explaining lower wheat yield in both adopter and non-adopter groups. However, gender, education, farm manure, household labor, irrigation, and rent appear to have differentiated impacts on the wheat yield of adopters and non-adopters. The results in the third column indicate that education and irrigation are significant and positive factors in wheat yields among adopters. However, household labor input seems to have a negative and significant effect on the wheat yield of non-adopters. Table 4 presents the expected farmers' wheat yield under actual and counterfactual conditions and the estimated results of the average treatment effects and base heterogeneity effects. Cells (a) and (b) represent the expected wheat yield observed in the sample. Cell (c) represents the expected wheat yield of the adopters if they had decided not to adapt, and cell (d) represents the expected wheat yield of the non-adopters if they decided to adapt. Adopters would have produced about 1911 kg/ha (29%) more if they had not adapted. Similarly, non-adopters would have produced about 1039 kg/ha (14%) less if they had adapted. Note: Standard errors in parenthesis and t-value in square brackets. *** represent the statistically significant at 1%.
In addition, the last row of Table 4 shows that adopters would have produced significantly more than the non-adopters in the counterfactual case. The significant heterogeneity effects imply that, regardless of the issue of climate change, the adopters are "better producers" than the non-adopters caused by some important sources of heterogeneity. The finding is consistent with Di Falco et al.

Conclusions and Discussion
This study found that over 90% of the wheat farmers are aware of climate change and over 80% of the households have autonomously adopted adaptation strategies. Farmers' cultivated area, climate change cognition and information on climate change significantly determined their adaptation decisions. However, farmers have limited adaptation strategies, mainly including increasing irrigation, and using more chemical fertilizer and pesticides. Based on this study, farmers' climate change adaptation strategies significantly decreased wheat yield, indicating that farmers' climate change adaptation actions may be maladaptive. Some studies also found maladaptive outcomes of some agricultural adaptation actions [25][26][27]. In this study, why did the measures fail to reduce climate risk and have adverse consequences?
The following are the possible reasons for the failure of main adaptation actions.
First, according to Liu et al. (2010), in the grain-filling stage of winter wheat, the frequency and quantity of irrigation should be reduced appropriately [28]. Therefore, the farmers' adaptation action to increase irrigation frequency and amount in response to reduced rainfall may lead to a negative effect on wheat yield, if they increase irrigation in an improper period. Second, in Ethiopia and Nepal, the input of fertilizer is a significant factor in higher food yields [12,18]. However, China has input utilizes about 52 times pesticides per hectare more than Ethiopian and Nepal; and China spent about 7 and 35 times fertilizer per hectare more than Ethiopian and Nepal respectively [29] China ranks first in the intensity of fertilizer used in agricultural production in the world, but fertilizer utilization is about 45%, far lower than the 60% utilization rate in developed countries [30]. Some empirical studies have shown that small farmers are risk-averse and would like to use a higher amount of fertilizer in order to avoid the negative impact of potential climate risks on agricultural production [31,32]. However, most farmer households have limited technical knowledge and a lack of agricultural labor force, depending on their traditional experience and habits, the phenomenon of farmer's overuse of fertilizer is widespread and serious in China [32,33]. Overuse of chemical fertilizers may decline the fertility of arable land, cause water pollution [34], and erode sustainable development of agriculture [25]. Therefore, the adaptation action of increasing fertilizer in response to the climate change risk would raise food production when the use of fertilizer is insufficient, but instead farmers increasing use of fertilizers and pesticides likely have a negative effect on wheat yield and the environment, if the fertilizer and pesticide inputs are excessive and inefficient. Third, some farmers change crop varieties to plant drought tolerant and disease-resistant wheat varieties in response to reduced rainfall and increasing pest and disease damage. However, the drought tolerant and disease-resistant wheat varieties may not be high-yield varieties and the adaptability of a new variety in complex environmental factors may lead to crop failure.
There is no doubt that adaptation is an important part of reducing the negative effects of climate change, sustaining farmers' livelihoods and ensuring agricultural sustainable development [35]. In particular, agriculture in China faces serious resource and environmental constraints, like water shortage and environmental degradation. However, small farmer's autonomous adaptation may be maladaptive and unsustainable. Therefore, the local government in Henan should help small farmers implement appropriate and effective adaptation strategies because government is in charge of agricultural infrastructure construction (irrigation and water conservancy systems, the agricultural product quality monitoring system and the agricultural information system) and agricultural science and technology progress (agricultural science and technology research, and agricultural science and technology popularization) [36,37]. Otherwise, small farmers may waste their efforts and resources without benefit, and may even experience loss if they just rely on themselves. Based on our findings, on one hand, it is urgent to implement scientific irrigation and fertilization to increase the efficient use of fertilizers and water for the interests of farmers and the sustainable development of agriculture. On the other hand, it is advisable to pay more attention to seed variety research and development to provide farmers with wheat varieties that have high yields, drought tolerance, and disease resistance, and guide them to choose suitable varieties according to local conditions. This study investigates the phenomenon of wheat farmers' maladaptation to climate change in Henan of China. Reasons behind the phenomenon, and whether it represents a general condition across different regions or crop varieties, will be explored in further research.    . T-value in parenthesis. *, ** and *** represent the statistically significant at 10%, 5% and 1%, respectively.