Agricultural water-saving irrigation (WSI) techniques becomes the main way to build drought resilience in rural areas. The effectiveness of government support in promoting the adoption of water-saving technologies is controversial. There is a research gap concerning farmers with different pro-environmental intentions to the adoption of water-saving behaviors under the influence of government support. This study is based on a survey in Yulin, Shaanxi, China, using 404 sample households. Pro-environmental intentions have been introduced into this paper by using the entropy weight method, and propensity score matching has been used to overcome the confounding between government support and farmers' adoption. Our findings reveal that farmers who received government support exhibit a significantly higher probability of adopting WSI techniques, ranging from 80 to 81.6%, representing a notable increase of 29.0–30.7% compared to those without government support. Furthermore, government support significantly affects farmers with weak pro-environmental intentions by 40.5%, but it does not have a significant impact on farmers with stronger pro-environmental intentions. These insights emphasize the significance of subsidies and advocate for an approach that tailors policies based on varying levels of pro-environmental intentions. This is particularly crucial considering that farmers with weaker intentions may be inclined to adopt the technology but face higher efficiency risks.

  • Government support significantly boosts water-saving irrigation (WSI) adoption among farmers, increasing the probability by 29.0–30.7%.

  • Farmers with weak pro-environmental intentions experience a significant 40.5% increase in adopting WSI techniques due to government support.

  • These findings can inform the design of effective WSI policies.

Water scarcity already significantly impacts agricultural development, with 72% of all water withdrawals allocated to the agriculture sector. Approximately 3.2 billion people reside in agricultural regions experiencing high to very-high water shortages or scarcity, of which 1.2 billion people live in severely water-constrained agricultural areas (FAO, 2020). According to the IPCC Sixth Assessment Report, there is high confidence that human influence has increased the frequency of concurrent heatwaves and drought extremes since the 1950s (Masson-Delmotte, 2021). Farmers are grappling with severe water shortages, and the adoption of water-saving irrigation (WSI) techniques can greatly enhance their resilience.

WSI equipment encompasses both public infrastructure, such as machine wells and ditches, and private facilities, including pipelines and field irrigation equipment. However, the high investment cost of both infrastructure and private facilities has emerged as the primary obstacle to promoting WSI technologies (Bjornlund et al., 2009). Consequently, governments are increasingly providing subsidies to alleviate this financial burden. Commonly used policies to promote WSI are direct subsidies, purchase incentives for water-saving irrigation equipment listed in government catalogs, tiered water pricing systems, and rewards for the adoption of water-saving practices.

However, the relationship between government subsidies for WSI and farmers’ behavior remains uncertain, and it may influence the implementation of the policy in different groups. Upon reviewing related literature, studies have shown that government support, including subsidies and extension services policies, has a significant and positive association with the probability of adoption (Koundouri et al., 2006; Genius et al., 2014; Cremades et al., 2015; Bahinipati & Viswanathan, 2019). However, several studies have documented opposite conclusions. Malik and Namara found that subsidies hardly promoted adoption among very poor farmers (Namara et al., 2007; Malik et al., 2018). Han (2018) pointed out that the percentage of farmers who continued to adopt WSI techniques and received government subsidies was smaller than the percentage of farmers who gave up adopting WSI techniques and received subsidies.

Our hypothesis is that controversial conclusions regarding adoption behaviors among farmers are the result of unexplored variables and improper methodologies. Empirical studies have shown that pro-environmental intentions have a positive effect on farmers’ pro-environmental behaviors (Bayard & Jolly, 2007; Deng et al., 2016). Pro-environmental intentions have been the subject of extensive research in the literature, with the Value-Belief-Norm theory proposing that people’s behaviors are driven by intentions, which are a combination of their values, beliefs, and social norms (Schwartz, 1977). Specifically, Stern’s theory of basic human values identifies three value orientations that can motivate pro-environmental intentions: egoistic, social-altruistic, and biospheric (Stern et al., 1993). While previous studies have explored the effects of each specific pro-environmental value orientation as a mediating variable, limited research has been conducted on how these three value orientations interact to shape individuals’ overall pro-environmental intentions and their subsequent adoption of water-saving technologies (Hedlund et al., 2012; Chua et al., 2016; Coşkun et al., 2017).

To address this gap, our study uses Yulin City in China as a case study, introduces WSI-related pro-environmental intentions as a new variable, and utilizes the entropy weight method (EWM) to quantify its importance. Building on Stern and Schwartz’s theories, we propose that pro-environmental intentions stem from three value orientations: egoistic, social-altruistic, and biospheric. Firstly, the egoistic value orientation is considered the primary motivator for human behavior. Secondly, individuals have a general value orientation towards the welfare of others, motivating them to engage in pro-environmental behavior. Finally, individuals who care about nonhuman species or the biosphere are more likely to exhibit pro-environmental behaviors. Our study integrates these three value orientations into pro-environmental intentions as a new variable to investigate their combined influence on farmers’ adoption of water-saving technology. We employ the EWM to quantify farmers’ pro-environmental intentions. By examining the collective effects of these variables on farmers’ overall pro-environmental intentions, we aim to contribute to a more nuanced understanding of the factors that influence sustainable agricultural practices.

As for methodology, every micro-econometric approach has to overcome heterogeneity and the following selection bias. The fact is that only one of the potential outcomes is observed for each farmer, and another outcome, the counterfactual outcome, cannot be observed. To address this issue, Rubin (1974) proposed the ‘counterfactual framework’, also known as the ‘Rubin causal model’ and Rosenbaum & Rubin (1985) proposed propensity score matching (PSM). PSM serves as a practical way to solve the selection bias by matching participants and nonparticipants who have similar treatment characteristics. This allows us to attribute the difference in outcomes to the treatment, enabling a precise examination of the impact of government support on the adoption of WSI techniques. Consequently, we can compare the behavior of farmers who receive government support with those who do not, facilitating the assessment of counterfactual results. Besides, the PSM model is less restrictive than regression models, as it does not require any specific distribution or parameter constraints, as well as it does not mandate a detailed explanation of the exogenous variables for identifying causal effects (Chen et al., 2022).

This article investigates the impact of government support on the adoption of agricultural WSI techniques by farmers, with a focus on the role of farmers’ pro-environmental intentions. Firstly, we quantify farmer’s pro-environmental intentions by using the EWM. Secondly, we employ the PSM method to analyze the impact and degree of influence of government support on farmers’ adoption of WSI techniques. Finally, we provide separate PSM estimations for farmers with different pro-environmental intentions to examine the differential treatment effect. Our findings contribute to a deeper understanding of farmers’ decision-making processes and their values, which can inform the design of effective policies to promote WSI techniques and alleviate water scarcity.

This article is structured as follows: Section 2 outlines the EWM and PSM procedures and provides details on the study area and the data. Section 3 presents the empirical results, and Section 4 summarizes the findings and their policy implications. In Section 5, we discuss the unique cultural context in China, examining the influence of social constraints and industrial development on the research

In this section, we first introduce the EWM, which will be used to determine the weight of each item in the questionnaire regarding value orientations through objective weighting. Then, we explain the PSM and its model.

Estimate farmers’ pro-environmental intentions by using the entropy weight method

To assess farmers’ pro-environmental intentions, we conducted a survey using questionnaires. One section of the survey was dedicated to examining farmers’ three value orientations under pro-environmental intentions: social-altruistic, biospheric, and egoism, or self-interest. Table 1 presents the items used in the survey. We quantified farmers’ pro-environmental intentions by applying the EWM. This method allows us to assign weights to the different items in the survey, based on the relative importance of each item in determining farmers’ pro-environmental intentions. The EWM has been extensively used in previous studies to estimate latent variables, including pro-environmental intentions.

Table 1

Survey items used to measure farmers’ pro-environmental intentions.

OrientationNameMeasurement
  1 = Strongly agree; 2 = Agree; 3 = Neither agree nor disagree; 4 = Disagree; 5 = Strongly disagree 
Egoistic E1 If farmers save irrigation water, they can reduce irrigation costs. 
 E2 Adopting WSI techniques is good for our family 
 E3 If our family adopts WSI techniques, we can increase our farming income. 
Social-altruistic S1 If friends or relatives persuade me to adopt WSI techniques, I will follow their advice. 
 S2 If farmers conserve irrigation water, they can ensure water security for future generations. 
 S3 If farmers conserve irrigation water, they can ensure water security for farmers or other water users downstream. 
Biospheric B1 If farmers keep pumping water for irrigation, the groundwater will be drained. 
 B2 By saving irrigation water just by my family, we can also alleviate water shortage and maintain soil and water. 
 B3 If farmers conserve irrigation water, it will help protect the environment. 
OrientationNameMeasurement
  1 = Strongly agree; 2 = Agree; 3 = Neither agree nor disagree; 4 = Disagree; 5 = Strongly disagree 
Egoistic E1 If farmers save irrigation water, they can reduce irrigation costs. 
 E2 Adopting WSI techniques is good for our family 
 E3 If our family adopts WSI techniques, we can increase our farming income. 
Social-altruistic S1 If friends or relatives persuade me to adopt WSI techniques, I will follow their advice. 
 S2 If farmers conserve irrigation water, they can ensure water security for future generations. 
 S3 If farmers conserve irrigation water, they can ensure water security for farmers or other water users downstream. 
Biospheric B1 If farmers keep pumping water for irrigation, the groundwater will be drained. 
 B2 By saving irrigation water just by my family, we can also alleviate water shortage and maintain soil and water. 
 B3 If farmers conserve irrigation water, it will help protect the environment. 

The EWM method used in this study is based on the concept of entropy, which was originally proposed by Shannon (1948). Shannon entropy is a measure of the uncertainty or randomness of information. In the EWM method, the entropy weight is determined by considering both the randomness and the dispersion of an indicator. Indicators with greater dispersion are given greater weight in the comprehensive evaluation. This allows for maximum compression of the evaluation system based on the evaluation results without sacrificing accuracy (Zhang et al., 2014). The following steps were followed in this study:

Step 1. Calculate contribution degree of each toward indicator :
formula
(1)
Step 2. Calculate the entropy of indicator .
formula
(2)
where the constant is generally taken as , and .
Step 3. Calculate the divergence coefficient for indicator .
formula
(3)
where represents the contribution divergence of each alternative toward indicator
Step 4. Determine the weight coefficient .
formula
(4)
Step 5. Evaluate the composite score of each sample.
formula
(5)

Model setting and propensity score matching

Model setting

This article selects variables from four sectors based on existing research, including farmers’ demographic and socio-economic characteristics, technical knowledge, willingness to adopt techniques, and pro-environmental intentions. Farmers’ demographic and socio-economic characteristics include five variables: age, education, agricultural experience, agricultural income ratio, and irrigation conditions. Knowledge and willingness include four variables: motivation, cognition level, perceived difficulty of operation, and willingness to install. The specific indicators for pro-environmental intentions will be processed later. Table 2 provides a specific definition and descriptive statistics of each variable.

Table 2

Definition of variables.

VariableDescriptionType
Whether to adopt (Y) 1 if adopted WSI techniques Dummy 
Government support 1 if the local government promoted or subsidized WSI techniques Dummy 
Age Age of the farmers in number of years Continuous 
Education Farmers’ education level: 1 = primary education; 2 = lower secondary education; 3 = upper secondary education; 4 = postsecondary nontertiary education Categorical 
Agricultural experience Agricultural experience of farmers in number of years Continuous 
Agricultural income The distribution of farm income in gross household income: 1= [0, 10%]; 2 [10, 20%]; 3 [20, 30%]; 4 [30, 50%]; 5 [50, 100%] Categorical 
Irrigation condition Is it convenient to irrigate farmland? 1 = no such irrigation; 2 = very inconvenient; 3 = inconvenient; 4 = average; 5 = convenient; 6 = very convenient Categorical 
Motivation Do you take the initiative to understand WSI technology? 1 = never; 2 = occasionally; 3 = sometimes; 4 = often; 5 = always Categorical 
Cognition level I know this technology very well 1 = no; 2 = a little bit; 3 = I know something; 4 = I know a lot Categorical 
Perceived difficulty For our family, it is not difficult to adopt irrigation water-saving technology: 1 = Strongly disagree; 2 = Disagree; 3 = Neither agree nor disagree; 4 = Agree; 5 = Strongly agree Categorical 
Willingness of installation Are you willing to install and use WSI facilities in the next 3 years? 1 = Strongly disagree; 2 = Disagree; 3 = Neither agree nor disagree; 4 = Agree; 5 = Strongly agree Categorical 
VariableDescriptionType
Whether to adopt (Y) 1 if adopted WSI techniques Dummy 
Government support 1 if the local government promoted or subsidized WSI techniques Dummy 
Age Age of the farmers in number of years Continuous 
Education Farmers’ education level: 1 = primary education; 2 = lower secondary education; 3 = upper secondary education; 4 = postsecondary nontertiary education Categorical 
Agricultural experience Agricultural experience of farmers in number of years Continuous 
Agricultural income The distribution of farm income in gross household income: 1= [0, 10%]; 2 [10, 20%]; 3 [20, 30%]; 4 [30, 50%]; 5 [50, 100%] Categorical 
Irrigation condition Is it convenient to irrigate farmland? 1 = no such irrigation; 2 = very inconvenient; 3 = inconvenient; 4 = average; 5 = convenient; 6 = very convenient Categorical 
Motivation Do you take the initiative to understand WSI technology? 1 = never; 2 = occasionally; 3 = sometimes; 4 = often; 5 = always Categorical 
Cognition level I know this technology very well 1 = no; 2 = a little bit; 3 = I know something; 4 = I know a lot Categorical 
Perceived difficulty For our family, it is not difficult to adopt irrigation water-saving technology: 1 = Strongly disagree; 2 = Disagree; 3 = Neither agree nor disagree; 4 = Agree; 5 = Strongly agree Categorical 
Willingness of installation Are you willing to install and use WSI facilities in the next 3 years? 1 = Strongly disagree; 2 = Disagree; 3 = Neither agree nor disagree; 4 = Agree; 5 = Strongly agree Categorical 

The following is the model used to illustrate the relationship between the influencing factors and the adoption behavior:
formula
(6)
formula
(7)
where denotes whether the farmer uses WSI techniques, is an underlying and unobserved stimulus index for the th observation, is the functional relationship between and , which determines the probability of adoption, is a dummy variable for government support, if the farmer received government support and otherwise, and denotes the pro-environmental intentions of farmer . The vector represents the farmer’s characteristics, and is a normal random disturbance term.

It is worth noting that the relationship between government support and the adoption of the techniques might be interdependent, potentially resulting in nonrandom treatment of samples. Hence, the traditional regression analysis method, ordinary least squares, would cause biased estimations. Compared with the traditional linear regression method, PSM can effectively overcome the ‘selection bias’ caused by biased estimation and sample ‘selfselection’. PSM has an advantage in solving the endogeneity problem. Unlike traditional regression methods, PSM does not require pre-assumptions about function form, parameter constraints, or error term distribution, nor does it require a strict generation of explanatory variables.

PSM

The idea of PSM is to match the samples of the treatment group and the comparison group according to the propensity score so that the main characteristics of the control group and the test group are as similar as possible. Propensity score is a single index variable that summarizes the observed characteristics into the matching algorithms. The comparison group can be used to estimate the expected outcome values without treatment for those who actually participated in treatment. The most important issue in the study is to specify the treatment effect for individual , defining as follows:
formula
(8)
where and denote the adoption possibility for farmers with and without government support, respectively. Since the combination of counterfactual situation and selection bias, we normally focus on the average treatment effects on the treated (ATT), which can be expressed as follows:
formula
(9)
where is the adopter when they received government support, and is obtained after matching, which denotes the farmers in the experimental group who have not received government support. To quantify the effect, this study implements the propensity score, a single index variable that summarizes the observed characteristics into the matching algorithms. The propensity score () is estimated using the following logit model:
formula
(10)
where represents the vector of observable variables.

After estimating the propensity score, we match the farmer households that have received government support with those that have not received government support by using matching algorithms. Various algorithms can be used to match samples, such as k-nearest neighbor matching, caliper matching, k-nearest neighbor matching within the caliper, and nuclear matching. Caliendo and Kopeinig suggest that different methods have different trade-offs between bias and efficiency. Therefore, the results of different matching methods may vary, but if the results obtained using multiple matching methods are similar or consistent, it indicates that the matching results are robust, and the sample validity is good (Caliendo & Kopeinig, 2008). To ensure the balance of the distributions of relevant variables between the control and treatment groups, it is necessary to test the balance before and after the matching process. A low pseudo- after matching indicates that the regression explains the situation well, and there are no systematic differences in the distribution of covariates. In addition, sensitivity analysis is commonly used to check for hidden bias (Becker & Caliendo, 2007).

Study area and data source

This study takes Yulin City as a case study. The location of Yulin City is shown in Figure 1. Yulin city includes 12 smaller administrative areas known as county-level areas within the city. These county-level areas are like sub-districts or districts within the city itself. Yulin is grappling with severe water scarcity. A large part of its arable land lacks irrigation infrastructure, and the region has poor climate resistance. According to the United Nations, water stress occurs when annual water supplies drop below 1,700 m per person. When annual water supplies fall below 1,000 m per person, it is considered water scarcity, and when it drops below 500 m, it is termed ‘absolute scarcity’. In 2020, Yulin City had per capita water supplies of only 790 m, indicating that the population is facing water scarcity. Despite the gradual decrease in the proportion of agricultural output value in Yulin, water use in the agricultural sector, mainly for planting, still accounts for the majority, and agricultural water use efficiency remains low (Rui, 2016). This underscores the significant contradiction between the demand for agricultural irrigation water supply and industrial development in Yulin and other resource-based cities in developing countries. Therefore, there is a pressing need to strictly control agricultural water use while promoting water-saving irrigation techniques. Strictly controlling agricultural water use and accelerating the adoption of WSI techniques have become imperative in these regions.

The data used in this study were obtained from a survey of 406 farmers who were randomly selected from the study area. The sample was selected based on population size, with 10–50 households randomly drawn in each county-level area to ensure fair representation. To ensure the authenticity and validity of the questionnaire, it was optimized through a presurvey, and investigators were trained. The survey involved face-to-face interviews with farmers, and the questionnaire covered several modules, including farmers’ demographic and socioeconomics characteristics, government support details, knowledge and attitudes toward water-saving irrigation techniques, willingness to install, and their pro-environmental intentions. After removing invalid samples, 404 valid questionnaires were used for analysis. The descriptive statistics of these four aspects are presented in Tables 3 and 4.

Table 3

Descriptive statistics of sample households.

VariableSample meanStandard deviationMinMax
Dependent variable     
Whether to adopt(Y) 0.119 0.324 – – 
Independent variable     
Government support 0.094 0.292 – – 
Age 57.114 11.036 25 86 
Education 1.389 0.634 
Agricultural experience 35.329 12.39 70 
Agricultural income 2.594 1.766 
Irrigation condition 4.527 1.671 
Motivation 1.485 0.901 
Cognition level 1.463 0.733 
Perceived difficulty 2.876 1.248 
Willingness of installation 3.433 1.238 
VariableSample meanStandard deviationMinMax
Dependent variable     
Whether to adopt(Y) 0.119 0.324 – – 
Independent variable     
Government support 0.094 0.292 – – 
Age 57.114 11.036 25 86 
Education 1.389 0.634 
Agricultural experience 35.329 12.39 70 
Agricultural income 2.594 1.766 
Irrigation condition 4.527 1.671 
Motivation 1.485 0.901 
Cognition level 1.463 0.733 
Perceived difficulty 2.876 1.248 
Willingness of installation 3.433 1.238 

Note:.

Table 4

Descriptive statistics and weights of items.

Item nameMaxMinMeanStandard deviationWeightWeight for each orientation
E1 3.579 0.875 0.0976 0.2506 
E2 3.973 0.838 0.0736  
E3 3.772 0.817 0.0794  
S1 3.651 0.818 0.0828 0.2961 
S2 3.522 0.914 0.1101  
S3 3.522 0.887 0.1032  
B1 3.663 1.069 0.1547 0.4533 
B2 1.985 0.729 0.2225  
B3 3.866 0.829 0.0761  
Item nameMaxMinMeanStandard deviationWeightWeight for each orientation
E1 3.579 0.875 0.0976 0.2506 
E2 3.973 0.838 0.0736  
E3 3.772 0.817 0.0794  
S1 3.651 0.818 0.0828 0.2961 
S2 3.522 0.914 0.1101  
S3 3.522 0.887 0.1032  
B1 3.663 1.069 0.1547 0.4533 
B2 1.985 0.729 0.2225  
B3 3.866 0.829 0.0761  

Note: E1–E3 represent three items related to egoism, S1–S3 represent three items related to social altruism, and B1–B3 represent three items related to pro-biosphere. See Table 1 for detailed descriptions.

Estimation of farmers’ pro-environmental intentions

The results show the weights of each item and the descriptive statistics. The biospheric value orientation had the highest weight of 0.4533, followed by social altruistic at 0.2961 and egoistic had the lowest weight at 0.2506. This suggests that individuals in the study placed a greater emphasis on their environmental values and the well-being of the biosphere as opposed to more self-centered values. In addition, the mean scores for most items fell between 3.5 and 3.9 on a five-point scale, indicating a moderate level of pro-environmental intention. B2 had a relatively low mean value, suggesting that farmers believe they have good initiative in every behavior. This positive outlook bodes well for future actions. The standard deviations ranged from 0.729 to 1.069, suggesting a moderate to high degree of variability in responses. Overall, these results suggest that individuals in the study had moderate pro-environmental intentions, with a greater emphasis on environmental values and the well-being of the biosphere. The distribution of scores on farmers' value orientations can be seen in Figure 2.

Estimation of the impact of government support on WSI techniques adoption by using the propensity score matching method

The empirical PSM analysis of the impact of government support on WSI techniques adoption was preceded by a specification of the propensity scores for the treatment variables. A logit model was employed to predict the propensity scores, and the results are reported in Table 5. This article, in line with other literature discussing PSM, treated propensity score as a device to balance the observed distribution of covariates (Becerril & Abdulai, 2010; Lee, 2013). Consequently, we refrain from delving into a discussion of the score itself. The result shows that the logit model has predicted 92.08% of observations correctly. As shown in Table 5, five variables, educational level, motivation, cognition level, perceived difficulty, and willingness to install, have a significant relationship with whether farmers receive extension services or subsidies from the government.

Table 5

Logit estimation results of propensity score.

CoefficientStandard error-value
Age −0.034 0.048 −0.71 
Education 0.593* 0.329 1.8 
Agricultural experience 0.01 0.0428 0.23 
Agricultural income −0.008 0.134 −0.06 
Irrigation conditions 0.047 0.173 0.27 
Motivation 0.485** 0.249 1.94 
Cognition level 1.221*** 0.345 3.53 
Perceived difficulty 0.350* 0.206 1.7 
Willingness of installation 0.781*** 0.299 2.93 
Pro-environmental intentions scores −0.203 0.62 −0.33 
Constant −8.846*** 2.665 −3.32 
LR  111.81 
Prob >  
Pseudo  0.4552 
Correctly classified 92.08% 
CoefficientStandard error-value
Age −0.034 0.048 −0.71 
Education 0.593* 0.329 1.8 
Agricultural experience 0.01 0.0428 0.23 
Agricultural income −0.008 0.134 −0.06 
Irrigation conditions 0.047 0.173 0.27 
Motivation 0.485** 0.249 1.94 
Cognition level 1.221*** 0.345 3.53 
Perceived difficulty 0.350* 0.206 1.7 
Willingness of installation 0.781*** 0.299 2.93 
Pro-environmental intentions scores −0.203 0.62 −0.33 
Constant −8.846*** 2.665 −3.32 
LR  111.81 
Prob >  
Pseudo  0.4552 
Correctly classified 92.08% 

To use the propensity score as the matching tool, one must ensure that the common support condition is imposed, which ensures that there is overlap in the range of propensity scores across treatment and comparison groups after matching. The before and after matching probability density function graphs of propensity scores for the comparison group and the treatment group are presented in Figure 3. We intuitively concluded, based on Figure 3, that the two groups exhibit significant overlap after matching, thereby avoiding poor matches. A balancing test will be processed later to ensure the satisfaction of the assumption of common support.
Figure 1

The location of Yulin city.

Figure 1

The location of Yulin city.

Close modal
Figure 2

Distribution of farmers’ pro-environmental intentions scores.

Figure 2

Distribution of farmers’ pro-environmental intentions scores.

Close modal
Figure 3

Probability density function graph of the propensity scores.

Figure 3

Probability density function graph of the propensity scores.

Close modal

A different number of samples would be lost due to different matching algorithms. Table 6 shows the maximum sample loss under different matching algorithms. Given that the treatment group and the comparison group still retain 313 matching samples in total, we can say that the level of data loss is within a reasonable range.

Table 6

Matching results.

Unmatched samplesMatched samplesTotal
Comparison group 88 278 366 
Treatment group 35 38 
Total 91 313 404 
Unmatched samplesMatched samplesTotal
Comparison group 88 278 366 
Treatment group 35 38 
Total 91 313 404 

Note: Comparison group represents the farmers who were not supported by the government. Treatment group represents the farmers who were supported by the government.

Given the lack of consensus on the optimal matching algorithm, we employed multiple algorithms to gauge their impact. The rationale behind this approach is that if the results obtained from various matching algorithms exhibit similarity or consistency, it suggests the matching results are robust. Therefore, we adopted three commonly used algorithms: nearest-neighbor matching, kernel matching, and radius matching. The parameters for these algorithms were selected based on the relevant literature (Caliendo & Kopeinig, 2008; Austin, 2014).

Table 7 presents the estimated results of PSM, indicating the differences in the adoption of water-saving irrigation (WSI) techniques between the two groups of farmers. The average treatment effect on the treated (ATT) shows that in the absence of government support, the probability of WSI technique adoption is estimated to be 50.1–52.6%. However, with the government support, the probability of adopting WSI techniques increases significantly to 81.6–82.8%, which represents an increase of 29–32.7%. These results suggest that government support has a positive impact on promoting the adoption of WSI techniques among farmers.

Table 7

Estimation of ATT.

Matching algorithmTreatment groupComparison groupDifference value
Nearest four neighbors matching 0.816 0.526 0.29 2.52 
Radius matching (caliper 0.1) 0.815 0.509 0.307 3.98 
Kernel matching (bandwidth 0.06) 0.8 0.501 0.299 3.66 
Matching algorithmTreatment groupComparison groupDifference value
Nearest four neighbors matching 0.816 0.526 0.29 2.52 
Radius matching (caliper 0.1) 0.815 0.509 0.307 3.98 
Kernel matching (bandwidth 0.06) 0.8 0.501 0.299 3.66 

Note: ATT is average treatment effects for the treatment group.

Comparison group represents the farmers who were not supported by the government. Treatment group represents the farmers who were supported by the government.

, , and represent significance at the 10, 5, and 1% levels, respectively.

A sensitivity analysis was conducted to test the robustness of the PSM results, as unobserved variables that affect both the treatment variable and the outcome variable simultaneously may cause ‘hidden bias’, resulting in unrobust matching estimators (Becker & Caliendo, 2007). To address this problem, the bounding approach developed by Rosenbaum (2002) was used. The effect is significant, and the hidden bias is excluded under . The significance of the effect becomes even more significant when increases. According to the sensitivity analysis results, the treatment effect is significant (at 5% level) in the nearest four neighbors matching, radius matching, and kernel matching within in the ranges of 1–3, 1–17, and 1–17, respectively. These ranges are all wide enough to ensure the robustness of the PSM results in this study, as well as indicating a significant treatment effect.

Moreover, it is important to evaluate how well the randomization in samples is, to stochastically balance all covariates. The results of balancing tests are presented in Table 8. The substantial reductions in the mean standardized bias (column 6) indicate that the groups are strongly balanced. The third column presents the pseudo from the propensity score estimation, and the numbers after matching are relatively low. The fourth column shows the likelihood-ratio test estimation, and the p values of the test show that the overall significance of regression covariates could always be rejected after matching but never be rejected before matching. Column 5 presents Rubin’s , and Rubin (2001) recommends that be between 0.5 and 2 for the samples to be considered sufficiently balanced.

Table 8

Quality indicators of matching algorithms.

Matching algorithmPseudo Rubin’s Mean standardized bias
Nearest neighbors matching Before matching 0.448 0.6 80.4 
 After matching 0.027 0.984 1.11 8.9 
Radius matching Before matching 0.448 0.6 80.4 
 After matching 0.031 0.98 1.72 10.6 
Kernel matching Before matching 0.448 0.6 80.4 
 After matching 0.03 0.983 0.81 10.4 
Matching algorithmPseudo Rubin’s Mean standardized bias
Nearest neighbors matching Before matching 0.448 0.6 80.4 
 After matching 0.027 0.984 1.11 8.9 
Radius matching Before matching 0.448 0.6 80.4 
 After matching 0.031 0.98 1.72 10.6 
Kernel matching Before matching 0.448 0.6 80.4 
 After matching 0.03 0.983 0.81 10.4 

To gain additional insights into the varying impacts of government support on farmers’ adoption behavior with different pro-environmental intentions, we also employed the radius matching algorithm on two distinct groups of farmers based on their intentions. Group 1 consists of farms with lower scores, indicating a stronger inclination towards adopting WSI techniques, while Group 2 comprises farmers with higher scores, indicating a relatively weaker intention towards WSI techniques. The results of this analysis are presented in Table 9.

Table 9

ATT in farmer with different pro-environmental intentions.

Farmer categoryTreatment groupComparison groupDifference valueCritical value of Rubin’s
0.75 0.487 0.263 1.85* 1.00–5.30 1.9 
0.864 0.459 0.405 2.83*** 1.00–1.70 1.37 
Farmer categoryTreatment groupComparison groupDifference valueCritical value of Rubin’s
0.75 0.487 0.263 1.85* 1.00–5.30 1.9 
0.864 0.459 0.405 2.83*** 1.00–1.70 1.37 

Note: ATT is average treatment effects for the treatment group Category 1 includes farms with stronger inclination toward adopting WSI techniques. Category 2 includes farmers with weaker intention towards WSI techniques. , , and represent significance at the 10, 5, and 1% levels, respectively.

As presented in Table 9, the impact of government support on farmers with stronger pro-environmental intentions adopting WSI techniques is not statistically significant. However, when government support is provided, farmers with lower pro-environmental intentions are more likely to adopt WSI techniques. Specifically, the adoption probability of WSI techniques is increased by 40.5% for farmers with lower pro-environmental intentions. These results indicate that government support plays a positive role in promoting the adoption of WSI techniques, especially for farmers with lower levels of environmental awareness. This may be because government support provides the necessary resources and incentives to reduce the adoption costs of WSI techniques, prompting even those farmers who may be initially less inclined to embrace this technology to consider its adoption. On the contrary, farmers who already possess strong environmental awareness may be more influenced by other factors, and government support may not have a significant impact on their decision-making. However, it should be acknowledged that farmers with weak pro-environmental intentions face risks in effectively and sustainably utilizing the WSI technique to achieve its maximum environmental benefits.

The following aspects of this study require further discussion. First, the level of industrial development and the maturity of supporting facilities are factors that influence the farmers’ pro-environmental behaviors (Guo Lijing, 2014). When farmers are located within the same city, their choices in adopting WSI techniques are limited. However, when considering a broader study area with a more extensive range of options available to farmers, factors such as the maturity of WSI technology – especially the maturity of equipment in the government subsidy list – along with the level of convenience and improvements in operation and maintenance collectively influence farmers’ behavior. Particularly among those who already possess a certain level of awareness, they are less likely to make hasty decisions and instead prioritize long-term benefits.

Second, apart from economic considerations, the impact of social constraints on the behavior of Chinese farmers should not be underestimated. Peer effects contribute to an individual’s decision-making regarding agricultural technique adoption (Xu et al., 2022). It may also contribute to the possibility that farmers cannot accurately express their pro-environmental attitudes during surveys due to the influence of other information, such as the attitudes of their neighbors and village committees. This, in turn, can affect the assessment and categorization results. However, whether farmers will continue to be swayed by collectivism to the extent of adopting WSI technologies remains uncertain. Exploring how to incorporate peer effects into the study could be included in future research.

Our study underscores the significance of recognizing these dynamics when formulating policies to promote sustainable water management. However, translating these insights into effective policy measures demands thoughtful consideration of the feasibility and accessibility of data collection methods. Future research and policy initiatives may explore innovative approaches to bridge the gap between individual intentions and policymaker knowledge. Besides, it is crucial to acknowledge that these findings can offer insights for regions beyond China, but applying the results to other cultural contexts requires culture-sensitive strategies.

In conclusion, this study has delved into the impact of government support on the adoption of WSI techniques among Chinese farmers, examining the moderating effect of farmers’ pro-environmental intentions. The results emphasize the significantly positive influence of government support on WSI technique adoption, with a more pronounced impact observed among farmers with weaker pro-environmental intentions. This highlights the need for targeted policies to promote WSI technique adoption, considering the heterogeneity of farmers’ pro-environmental intentions.

Based on these findings, several recommendations are proposed for policymakers and practitioners. First, it is crucial to not only maintain but also consider the increasing government support for WSI technique adoption. This can be achieved through financial incentives, technical guidance, and various forms of support. Second, government should identify the groups of farmers with different pro-environmental intentions. Multiple levels of government should not only prioritize subsidies but also recognize the risk faced by farmers with weaker pro-environmental intentions. The insufficient awareness of the pro-environmental intentions may result in a lack of motivation for learning and maintenance of WSI techniques, then poses a potential risk of the low efficiency of water saving. Efforts to develop farmers’ pro-environmental intentions should be coupled with additional extensive services. These services may encompass environmental protection seminars, technical training, and maintenance training, aiming to ensure a more thorough and effective utilization of WSI techniques and reduce the risk of ineffective subsidies and equipment utilization.

In summary, our study underscores the pivotal role of government support in promoting WSI technique adoption among Chinese farmers. Policymakers and practitioners should focus on providing targeted assistance to farmers with weaker pro-environmental intentions to maximize impact.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.

Austin
P. C.
(
2014
).
A comparison of 12 algorithms for matching on the propensity score
.
Statistics in Medicine
33
,
1057
.
doi:10.1002/SIM.6004
.
Bahinipati
C. S.
&
Viswanathan
P. K.
(
2019
).
Incentivizing resource efficient technologies in India: evidence from diffusion of micro-irrigation in the dark zone regions of Gujarat
.
Land Use Policy
86
,
253
260
.
doi:10.1016/j.landusepol.2019.04.017
.
Becerril
J.
&
Abdulai
A.
(
2010
).
The impact of improved maize varieties on poverty in Mexico: A propensity score-matching approach
.
World Development
38
,
1024
1035
.
doi:10.1016/j.worlddev.2009.11.017
.
Becker
S. O.
&
Caliendo
M.
(
2007
).
Sensitivity analysis for average treatment effects
.
Stata Journal
7
,
71
83
.
doi:10.1177/1536867x0700700104
.
Bjornlund
H.
,
Nicol
L.
&
Klein
K. K.
(
2009
).
The adoption of improved irrigation technology and management practices—A study of two irrigation districts in Alberta, Canada
.
Agricultural Water Management
96
,
121
131
.
doi:10.1016/j.agwat.2008.07.009
.
Caliendo
M.
&
Kopeinig
S.
(
2008
).
Some practical guidance for the implementation of propensity score matching
.
Journal of Economic Surveys
22
,
31
72
.
doi:10.1111/j.1467-6419.2007.00527.x
.
Chen
J. W.
,
Maldonado
D. R.
,
Kowalski
B. L.
,
Miecznikowski
K. B.
,
Kyin
C.
,
Gornbein
J. A.
&
Domb
B. G.
(
2022
).
Best practice guidelines for propensity score methods in medical research: Consideration on theory, implementation, and reporting. A review
.
Arthroscopy: The Journal of Arthroscopic & Related Surgery
38
,
632
642
.
doi:https://doi.org/10.1016/j.arthro.2021.06.037
.
Chua
K. B.
,
Quoquab
F.
,
Mohammad
J.
&
Basiruddin
R.
(
2016
).
The mediating role of new ecological paradigm between value orientations and pro-environmental personal norm in the agricultural context
.
Asia Pacific Journal of Marketing and Logistics
28
,
323
349
.
doi:10.1108/APJML-09-2015-0138/FULL/XML
.
Coşkun
A.
,
Vocino
A.
&
Polonsky
M.
(
2017
).
Mediating effect of environmental orientation on pro-environmental purchase intentions in a low-involvement product situation
.
Australasian Marketing Journal (AMJ)
25
,
115
125
.
doi:10.1016/J.AUSMJ.2017.04.008
.
Cremades
R.
,
Wang
J.
&
Morris
J.
(
2015
).
Policies, economic incentives and the adoption of modern irrigation technology in China
.
Earth System Dynamics
6
,
399
410
.
doi:10.5194/esd-6-399-2015
.
Deng
J.
,
Sun
P.
,
Zhao
F.
,
Han
X.
,
Yang
G.
&
Feng
Y.
(
2016
).
Analysis of the ecological conservation behavior of farmers in payment for ecosystem service programs in eco-environmentally fragile areas using social psychology models
.
Science of the Total Environment
550
,
382
390
.
doi:10.1016/J.SCITOTENV.2016.01.152
.
FAO
(
2020
).
The State of Food and Agriculture 2020. Overcoming Water Challenges in Agriculture. FAO, Rome
.
Genius
M.
,
Koundouri
P.
,
Nauges
C.
&
Tzouvelekas
V.
(
2014
).
Information transmission in irrigation technology adoption and diffusion: Social learning, extension services, and spatial effects
.
American Journal of Agricultural Economics
96
,
328
344
.
doi:10.1093/ajae/aat054
.
Guo Lijing
Z. J.
(
2014
).
Influence mechanisms and policy interventions on farmers’ pro-environmental behavior: Take straw disposal behavior as an example
.
Agricultural Economic Issues
12
,
78
112
.
doi:10.13246/j.cnki.iae.2014.12.010
.
Han
C. X. H.
(
2018
).
Influence of government subsidies and adoption effect on continuous adoption behavior of water-saving irrigation technology by farmers
.
Resources Science
40
,
1418
1428
.
Koundouri
P.
,
Nauges
C.
&
Tzouvelekas
V.
(
2006
).
Technology adoption under production uncertainty: Theory and application to irrigation technology
.
American Journal of Agricultural Economics
88
,
657
670
.
doi:10.1111/j.1467-8276.2006.00886.x
.
Lee
W. S.
(
2013
).
Propensity score matching and variations on the balancing test
.
Empirical Economics
44
,
47
80
.
doi:10.1007/s00181-011-0481-0
.
Malik
R. P. S.
,
Giordano
M.
&
Rathore
M. S.
(
2018
).
The negative impact of subsidies on the adoption of drip irrigation in India: Evidence from Madhya Pradesh
.
International Journal of Water Resources Development
34
,
66
77
.
Masson-Delmotte
V.
(
2021
).
Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Technical Report. Cambridge University Press, Cambridge, UK
.
Rosenbaum
P. R.
(
2002
).
Overt Bias in Observational Studies
.
Springer
,
New York, NY
, pp.
71
104
.
doi:10.1007/978-1-4757-3692-2_3
.
Rosenbaum
P. R.
&
Rubin
D. B.
(
1985
).
Constructing a control group using multivariate matched sampling methods that incorporate the propensity score
.
The American Statistician
39
,
33
.
doi:10.2307/2683903
.
Rubin
D. B.
(
2001
).
Using propensity scores to help design observational studies: Application to the tobacco litigation
.
Health Services and Outcomes Research Methodology
2
,
169
188
.
doi:10.1023/A:1020363010465
.
Rui
Z.
(
2016
).
The impact of water scarcity on regional industrial structure optimization: The case of Yulin City
.
China Population - Resources and Environment
26
,
333
335
.
Schwartz
S. H.
(
1977
).
Normative Influences on Altruism
.
Advances in Experimental Social Psychology
10
,
221
279
.
doi:10.1016/S0065-2601(08)60358-5
.
Shannon
C. E.
(
1948
).
A mathematical theory of communication
.
Bell System Technical Journal
27
,
379
423
.
doi:10.1002/j.1538-7305.1948.tb01338.x
.
Stern
P. C.
,
Dietz
T.
&
Kalof
L.
(
1993
).
Value orientations, gender, and environmental concern
.
Environment and Behavior
25
,
322
348
.
doi:10.1177/0013916593255002
.
Xu
Z.
,
Zhang
K.
&
Zhou
L.
(
2022
).
Mutual proximity and heterogeneity in peer effects of farmers’ technology adoption: Evidence from China’s soil testing and formulated fertilization program
.
China Agricultural Economic Review
14
, 395̵–415. doi.org/10.1108/CAER-10-2020-0250.
Zhang
X.
,
Wang
C.
,
Li
E.
&
Xu
C.
(
2014
).
Assessment model of ecoenvironmental vulnerability based on improved entropy weight method
.
Scientific World Journal
.
doi:10.1155/2014/797814
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).