Does Cleaner Household Energy Promote Agricultural Green Production? Evidence from China

Cleaner household energy for agricultural green production can significantly alleviate energy poverty and food security, thus contributing to global sustainable development. Using survey micro-data collected from Sichuan Province, the ordered probit model, OLS model, and instrumental variables approach were applied for empirical analysis. The results show that: (1) cleaner household energy significantly enhances farmer’s agricultural green production awareness and improves agricultural green production levels, which is still significant after treating endogenous issues with the conditional mixing process estimation method and 2SLS model; (2) health plays a partially mediating effect of cleaner household energy on agricultural green production awareness and agricultural green production levels; (3) environmental protection awareness and digital literacy have a moderating effect and reinforce the positive impact of cleaner household energy on agricultural green production awareness and agricultural green production levels. This research suggests that governments can enhance the impact of cleaner household energy on agricultural green production through price and subsidy mechanisms.


Introduction
Household energy consumption and agricultural production are the two main sources of carbon emissions [1,2]. The long-term use of non-clean energy by households and traditional agricultural production practices produces large amounts of greenhouse gases, which are a major cause of frequent climate extremes [3]. In order to solve the issues of air pollution, environmental damage, energy poverty, health crises, and food security caused by non-clean energy consumption and traditional agricultural production, many countries are embarking on new "green revolutions" in the energy and agriculture sectors: cleaner household energy (CHE) [4] and agricultural green production (AGP) [5]. CHE can eradicate energy poverty [6], reduce harmful and greenhouse gas (GHG) emissions [7], accelerate human and social capital accumulation [8], and improve the health and well-being of the population [9]. AGP can also increase farmers' income and alleviate poverty [10], reduce CO 2 and environmental pollution [11], and improve the quality of agricultural production and food security indices [12]. Therefore, CHE and AGP are important ways to achieve the UN's sustainable development goals (SDGs) of "no poverty", "zero hunger", "affordable and clean energy", and "climate action" [13,14].
The existing literature on factors influencing farmers' AGP has highlighted that the spread of digital technology [15] and the use of the Internet [16] has significantly increased the farmers' awareness of AGP. Moreover, political identity [17], social networks [18], and social contacts [19] accelerate the accumulation of farmers' social capital, which is positively correlated with farmers' willingness to participate in AGP [18]. On the other hand, the adoption of modern agricultural-technology-related training programs can effectively help been developed in Western Sichuan, making it an important region for the production of pastoral products in China. Sichuan Province has always been a typical "big agriculture" (agriculture, pastoralism, forestry, and fishing) region in China [41]. Therefore, this paper is representative of the study of the impact of CHE on AGP, using Sichuan as an example.
Although there are some studies discussing the effects of CHE on AGP, further refinement is needed. China is one of the world's leading energy consumers and agricultural producers, but there is a relative lack of evidence for China in current studies on CHE and AGP. At the same time, most of the existing research on green agricultural development is based on a sample of macro statistics and there is a comparative lack of analysis of microdata [42]. Therefore, the researchers took the survey micro-data from Sichuan Province as its sample and empirically analyzed whether and how CHE influences farmers' AG which is rich in theoretical value.
The remainder of this paper includes: an overview analysis of energy consumption and green agriculture development in Sichuan (Section 2); data and methods (Section 3); empirical analysis and discussion (Section 4); further research (Section 5); and conclusions and policy recommendations (Section 6).

An Overview of Energy Consumption and Green Agriculture Development in Sichuan Province
According to Figure 1, the changes in non-clean energy production (EP) and consumption (EC) in Sichuan Province from 2010 to 2020 are less volatile, maintaining an overall growth trend, and are consistently greater than the amount of energy produced, indicating that Sichuan Province consumes more energy and that non-clean energy still plays an important role in economic development. production in Central and Eastern Sichuan is dominated by the cultivation of grain crops (i.e., rice, maize, soybeans, and wheat) and pig breeding, making it an important grainproducing and pork-supplying area in China. Livestock farming has long been developed in Western Sichuan, making it an important region for the production of pastoral products in China. Sichuan Province has always been a typical "big agriculture" (agriculture, pastoralism, forestry, and fishing) region in China [41]. Therefore, this paper is representative of the study of the impact of CHE on AGP, using Sichuan as an example.
Although there are some studies discussing the effects of CHE on AGP, further refinement is needed. China is one of the world's leading energy consumers and agricultural producers, but there is a relative lack of evidence for China in current studies on CHE and AGP. At the same time, most of the existing research on green agricultural development is based on a sample of macro statistics and there is a comparative lack of analysis of micro-data [42]. Therefore, the researchers took the survey micro-data from Sichuan Province as its sample and empirically analyzed whether and how CHE influences farmers' AG which is rich in theoretical value.
The remainder of this paper includes: an overview analysis of energy consumption and green agriculture development in Sichuan (Section 2); data and methods (Section 3); empirical analysis and discussion (Section 4); further research (Section 5); and conclusions and policy recommendations (Section 6).

An Overview of Energy Consumption and Green Agriculture Development in Sichuan Province
According to Figure 1, the changes in non-clean energy production (EP) and consumption (EC) in Sichuan Province from 2010 to 2020 are less volatile, maintaining an overall growth trend, and are consistently greater than the amount of energy produced, indicating that Sichuan Province consumes more energy and that non-clean energy still plays an important role in economic development. According to Figure 2, the proportion of Sichuan's agricultural GDP in the national agricultural GDP from 2010 to 2020 has generally maintained a growing trend and has increased significantly, illustrating the important role of Sichuan Province in China's agricultural development. According to Figure 2, the proportion of Sichuan's agricultural GDP in the national agricultural GDP from 2010 to 2020 has generally maintained a growing trend and has increased significantly, illustrating the important role of Sichuan Province in China's agricultural development.  Figure 3 portrays the changes in grain production in Sichuan Province from 2010 to 2020. It can be seen that grain production in Sichuan Province is growing faster. The results in Figures 2 and 3 show that the level of agricultural development in Sichuan Province has been increasing in recent years and significantly improving China's agricultural development. Agricultural green total factor productivity (AGTFP) is a commonly used indicator at the macro level to measure the level of development of green agriculture, which covers the use of fertilizers, pesticides, and machinery in agricultural production, as well as carbon emissions [43]. This study used a static SBM model to calculate the AGTFP for each province in China from 2010 to 2019, and then calculated the ten-year average value of AGTFP for each province. Figure 4 portrays the ten-year AGTFP for 30 Chinese provinces (excluding Tibet, Taiwan, Hong Kong, and Macau), and it can be seen that the best development of green agriculture is in Guangxi province while the worst is in Gansu province, with Sichuan Province ranking 13th, indicating that green agriculture in Sichuan Province is at an intermediate level in China [40]. 5   It can be seen that grain production in Sichuan Province is growing faster. The results in Figures 2 and 3 show that the level of agricultural development in Sichuan Province has been increasing in recent years and significantly improving China's agricultural development.   Figure 3 portrays the changes in grain production in Sichuan Province from 2010 to 2020. It can be seen that grain production in Sichuan Province is growing faster. The results in Figures 2 and 3 show that the level of agricultural development in Sichuan Province has been increasing in recent years and significantly improving China's agricultural development. Agricultural green total factor productivity (AGTFP) is a commonly used indicator at the macro level to measure the level of development of green agriculture, which covers the use of fertilizers, pesticides, and machinery in agricultural production, as well as carbon emissions [43]. This study used a static SBM model to calculate the AGTFP for each province in China from 2010 to 2019, and then calculated the ten-year average value of AGTFP for each province. Figure 4 portrays the ten-year AGTFP for 30 Chinese provinces (excluding Tibet, Taiwan, Hong Kong, and Macau), and it can be seen that the best development of green agriculture is in Guangxi province while the worst is in Gansu province, with Sichuan Province ranking 13th, indicating that green agriculture in Sichuan Province is at an intermediate level in China [40]. 5  Agricultural green total factor productivity (AGTFP) is a commonly used indicator at the macro level to measure the level of development of green agriculture, which covers the use of fertilizers, pesticides, and machinery in agricultural production, as well as carbon emissions [43]. This study used a static SBM model to calculate the AGTFP for each province in China from 2010 to 2019, and then calculated the ten-year average value of AGTFP for each province. Figure 4 portrays the ten-year AGTFP for 30 Chinese provinces (excluding Tibet, Taiwan, Hong Kong, and Macau), and it can be seen that the best development of green agriculture is in Guangxi province while the worst is in Gansu province, with Sichuan Province ranking 13th, indicating that green agriculture in Sichuan Province is at an intermediate level in China [40]. According to Figure 5, the AGTFP in Sichuan Province maintained a year-on-year upward trend from 2010 to 2019, indicating that the level of green agriculture development in Sichuan Province has improved faster and is at a higher level after 2017, which implies that agricultural production in Sichuan Province is gradually changing from traditional to green agriculture. In summary, the macro data analysis shows that non-clean energy consumption and the level of green agricultural development in Sichuan Province are on the same growth trend. Does this mean that non-clean energy consumption can promote green agricultural development? Or, does cleaner energy consumption suppress AGP? This study addresses these questions using micro-data.  According to Figure 5, the AGTFP in Sichuan Province maintained a year-on-year upward trend from 2010 to 2019, indicating that the level of green agriculture development in Sichuan Province has improved faster and is at a higher level after 2017, which implies that agricultural production in Sichuan Province is gradually changing from traditional to green agriculture.  According to Figure 5, the AGTFP in Sichuan Province maintained a year-on-year upward trend from 2010 to 2019, indicating that the level of green agriculture development in Sichuan Province has improved faster and is at a higher level after 2017, which implies that agricultural production in Sichuan Province is gradually changing from traditional to green agriculture. In summary, the macro data analysis shows that non-clean energy consumption and the level of green agricultural development in Sichuan Province are on the same growth trend. Does this mean that non-clean energy consumption can promote green agricultural development? Or, does cleaner energy consumption suppress AGP? This study addresses these questions using micro-data.  In summary, the macro data analysis shows that non-clean energy consumption and the level of green agricultural development in Sichuan Province are on the same growth trend. Does this mean that non-clean energy consumption can promote green agricultural development? Or, does cleaner energy consumption suppress AGP? This study addresses these questions using micro-data.

Data
For its sample, this study uses data from the "Household Energy Consumption and Green Agriculture Development Survey" conducted in 21 cities (states) in Sichuan Province from January to April 2022. The survey was commissioned from the municipal (state) agricultural and rural bureaus and the township government departments and collected data were based on the basic personal and household information, agricultural (pastoral) production, health, and household energy use of farmers in the agricultural and pastoral areas of Sichuan Province. The survey actually distributed 600 questionnaires; 523 questionnaires were returned and 498 questionnaires were valid, accounting for 95.22% of the valid questionnaires. The reliability test value of the survey data of Cronbach-α = 0.801, and the validity test value of KMO = 0.776 demonstrated that the survey data is reliable. First, according to the questions "Which of the following crops do you grow?" and "Which of the following livestock (poultry) do you raise?" if respondents selected both "not farming" and "not breeding", respectively, they were not involved in agricultural production and were excluded from the data. Then, in the remaining sample, we refer to agriculture and pastoralism collectively as agricultural production. Later, we removed extreme values from the data, filled in missing values, and normalized the data. Finally, a total of 454 effective sample data points were obtained for empirical analysis. Figure 6 displays the study area and sample distribution of this research.

Data
For its sample, this study uses data from the "Household Energy Consumption and Green Agriculture Development Survey" conducted in 21 cities (states) in Sichuan Province from January to April 2022. The survey was commissioned from the municipal (state) agricultural and rural bureaus and the township government departments and collected data were based on the basic personal and household information, agricultural (pastoral) production, health, and household energy use of farmers in the agricultural and pastoral areas of Sichuan Province. The survey actually distributed 600 questionnaires; 523 questionnaires were returned and 498 questionnaires were valid, accounting for 95.22% of the valid questionnaires. The reliability test value of the survey data of Cronbach-α = 0.801, and the validity test value of KMO = 0.776 demonstrated that the survey data is reliable. First, according to the questions "Which of the following crops do you grow?" and "Which of the following livestock (poultry) do you raise?" if respondents selected both "not farming" and "not breeding", respectively, they were not involved in agricultural production and were excluded from the data. Then, in the remaining sample, we refer to agriculture and pastoralism collectively as agricultural production. Later, we removed extreme values from the data, filled in missing values, and normalized the data. Finally, a total of 454 effective sample data points were obtained for empirical analysis. Figure 6 displays the study area and sample distribution of this research.

Explained Variables: Agricultural Green Production (AGP)
In the current study, AGP is usually measured by indicators such as fertilizer [11], pesticide [44], green manure use [42], and straw treatment [45]. The common variables of AGP include "green production awareness" [18], "green production efficiency" [24], "green production willingness" [46], and "green production behavior" [47]. In this work, the researchers measured AGP in terms of "agricultural green production awareness (AGPA)" and "agricultural green production levels (AGPLs)". AGPA refers to farmers' knowledge of green production; the AGPL refers to farmers' green production index. The specific definitions of explained variables are reported in Table 1.

Explanatory Variables: Cleaner Household Energy (CHE)
Referring to Zang et al. [48], the researchers measures CHE in two ways in this work: First, compare the frequency of use of clean energy (FCE) with that of non-clean energy (FNCE), set "CHE_1 = 2" if FCE > FNCE, set "CHE_1 = 1" if FCE = FNCE, and set "CHE_1 = 0" if FCE < FNCE. Second, compare the proportions of clean energy (PCE) and non-clean energy (PNCE) in the household energy portfolio to consider a shift towards cleaner household energy if PCE > PNCE. The specific definitions of CHE are reported in Table 1.

Control Variables
In this paper, personal characteristic variables such as "age" and "gender", and household characteristic variables such as "number of agricultural labors (AL)" were seelected as control variables. The specific definitions of all control variables are reported in Table 1.
The mean, standard deviation, and maximum and minimum values of all variables selected in this research are reported in Table 2. A total of 262 (57.71%) respondents have a high degree of agricultural green production awareness and 18.94% of farmers still lack agricultural green production awareness. The mean of the agricultural green production level is 0.52, the minimum value is −1.12, and the maximum value is 0.70, indicating that the overall agricultural green production level is low and the gap is obvious. More than half (50.88%) of households have a higher frequency of clean energy consumption than non-clean energy and 82.82% of households use a higher proportion of clean energy than non-clean energy, indicating that household energy consumption is shifting towards green. Most of the respondents are married (84.14%), middle-aged (average age was 42.33 years), male (72.25%), and have a lower education level. Only 97 (21.37%) have a high school or university education but 32.82% of family members have a college education. The mean duration that respondents have participated in agricultural production is about 22 years. Of those respondents, 82.16% have accepted agricultural production training and 296 (65.20%) have the experience of going outside to work. The mean land area of the interviewed households is 5.75 acres and 309 households (68.16%) have two laborers engaged in agricultural production. There is a large gap in the income of family agricultural production subsidies and most households (63.66%) do not purchase agricultural insurance. In the past 3 years, 15.42% of the households have suffered from agricultural disease and the agricultural production of 31.50% of the households has been negatively affected by COVID-19. The non-agricultural income of the interviewed households is relatively high and the gap is small (standard deviation = 0.60). More than half of the respondents (56.83%) were relatively healthy but 43.17% of the respondents have a low level of health, and 166 households (36.56%) have patients with respiratory illness. A total of 67.40% of the respondents have environmental protection awareness but the majority (64.10%) of the respondents have poor digital literacy.  Agricultural production loan (APL) Has your household applied for and received a loan from a bank for agricultural production in the last 3 years? 1 = Yes; 0 = No.
Non-agricultural production income (NAPI) What was your household's total income last year in approximately RMB? (Total household income); What was your household's agricultural production income last year in approximately RMB? (Agricultural production income). Ln (Total household income-Agricultural production income)

Methods
The explained variable AGPA is an ordered multi-categorical variable and it is a commonly used econometric model, such as the ordered probit model (O-probit) [49]. The explained variable AGPL is the continuous variable and the classical model is ordinary least squares (OLS). So, we construct O-probit models for CHE_1 and AGPA (Equations (1) and (2)) and OLS models for CHE_1 and AGPLs (Equation (3)).
We take AGPA as an example, referring to Wen et al. [50], and then construct the mediating effect and moderating effect model. Equations (4) and (5) represent the mediating role and moderating effect of some variables on the impact of CHE_1 on AGPA and AGPLs.
AGPA is agricultural green production awareness, AGPL is the agricultural green production level, CHE_1 is cleaner household production, MV is the mediating variables, RV is moderating variable, CV is the control variable, and ω, β, λ are regression coefficients.

Basic Regression
In Table 3, the results of models (1) and (2) show that CHE significantly positively improved farmers' AGPA, and the use of clean energy at home fosters farmers' awareness and knowledge of AGP through the development of a green production mindset. The results for the control variables in model (1) show that REL significantly positive affected AGPA, implying that the higher the level of education, the greater the cognizance of AGP [18]. APT significantly positive influenced AGPA at the 1% level and the participation in agricultural knowledge and skills training can broaden farmers' access to information related to AGP [51]. EWO significantly positive enhanced AGPA, and working outside the home broadens farmers' minds to accept new things, making them capable of being more proactive in understanding AGP. AI is positively associated with AGPA at the 5% level, and the purchase of agricultural insurance means that farmers are already endowed with modern agricultural production characteristics, and this group of farmers is more receptive to AGP [27]. AD significantly reduces the AGPA of farmers, who may choose to increase pesticides or fertilizers use rather than diminish agricultural surface pollution when encountering pests and diseases. NAPI was significantly and negatively associated with AGPA at the 1% level; households with a higher NAPI have a lower willingness to participate in agricultural production [20] and do not spend time understanding AGP. Note: standard errors in parentheses, *** p < 0.01, ** p < 0.05, * p < 0. The results of model (3) in Table 3 show that CHE significantly increases farmers' AGPL and clean energy household use saves household members' time to participate in AGP, thus increasing AGPLs. The results of model (3), controlling for variables, further demonstrate that the lower odds of participating in AGP with farmers get higher, consequently decreasing the AGPL [22]. In contrast, APT upgrades farmers' skills in agricultural production, increases their motivation to participate in AGP, and improves AGPLs. Likewise, EWO and AGPLs are significantly positively connected, and working outside the home provides access to more income and market information, increasing the subjective ability and willingness of farmers to participate in AGP [52]. AS significantly enhances AGPLs, and AS not only increases the capital element of AGP but also increases agricultural hedging of agricultural risk and increases AGP willingness [26]. In addition, AD and COVID-19 increase the difficulty and uncertainty of household production decisions, hence AD and COVID-19 significantly reduce AGPLs [53,54]. Finally, NAPI can increase AGPLs by alleviating the financial constraints on household participation in AGP.

Endogenous Treatment
The regressions of CHE_1 on AGPA and AGPLs may have endogenous issues due to omitted variables, reciprocal causality, etc., which may affect the results of the analysis. The instrumental variables method is the classic method for dealing with the endogenous problem in cross-sectional data. Therefore, the question "Has anyone in your household been diagnosed with a respiratory illness? (respiratory illness = RI) 1 = Yes, 0 = No" was selected as the instrumental variable (IV), and both Iv-O-probit and 2SLS models were employed to deal with potential endogenous problems. RI was chosen as IV because the household member with an RI would use less non-clean energy, thus promoting CHE but would not have an impact on the household's agricultural practices, which means that RI is directly related to CHE but not to AGP, satisfying the exogeneity requirement of IV.

Endogenous Issue Treatment of CHE and AGPA Regression
Currently, there are technical difficulties in directly using Iv-O-probit models to deal with the endogenous issue [55], and the conditional mixed process estimation (CMP) method is often applied to overcome technical constraints [56]. The results of the endogenous treatment of CHE_1 and AGPA are reported in Table 4. The value of the auxiliary estimation parameter atanhrho_12 is significantly different from zero (p-value = 0.00), indicating that the basic regression model has endogenous problems and that CHE_1 is an endogenous variable. The results of the first stage of Table 4 show that RI is significantly interrelated with CHE_1 and not with AGPA, satisfying the conditions of the IV method, and the F-value is 55.32, which is greater than the empirical value of 10, indicating that there is no weak IV problem. The second stage results show that CHE_1 still significantly enhances AGPA after dealing with the endogenous problem, with reduced coefficient values and average marginal effects at each cut-off point, indicating that the basic regression overestimates the effect of CHE_1 on AGPA.

Endogenous Issue Treatment of CHE_1 Impact on the AGPL
The Hausman test for regression of CHE_1 on AGPLs was 0.035, which is less than 0.05, indicating that there is an endogenous problem in the basic regression, and CHE_1 is an endogenous variable. The results of the first stage of Table 5 illustrate that RI is significantly linked with CHE_1 but not with AGPLs, satisfying the IV method; the F-value is 103.2, which is greater than the empirical value of 10, indicating that there is no weak IV problem. The results of the second stage show that CHE_1 is still significantly positively associated with AGPLs after dealing with the endogenous issue, but the coefficients of AGPLs are reduced, implying that the effect of CHE_1 on AGPLs is overestimated in the basic regression. Note: standard errors in parentheses, *** p < 0.01, ** p < 0.05. AGPA = agricultural green production awareness; CHE_1 = frequency of clean energy use = cleaner household energy; RI = IV = respiratory illnesses; CV = control variables. Note: standard errors in parentheses, ** p < 0.05. AGPL = agricultural green production level; CHE_1 = frequency of clean energy use = cleaner household energy; RI = IV = respiratory illnesses; CV = control variables.

Robustness Test
In this work, the researchers redefined CHE, i.e., replacing CHE_1 with CHE_2, and then conducted a robustness test regression of CHE_2 on AGPA and AGPLs. The results in Table 6 show that after replacing the core explanatory variables, CHE still significantly enhances AGPA, and significantly improves AGPLs. The results of the robustness tests for the control variables are essentially the same as in the basic regression and are not reported. 454 Note: standard errors in parentheses, *** p < 0.01, ** p < 0.05. AGPA = agricultural green production awareness; AGPL = agricultural green production level; CHE_2 = proportion of clean energy use = cleaner household energy; CV = control variables.

Mediating Effect Analysis
CHE is very valuable to improving the health of household members [57], then reducing households' expenditure on health care and increasing capital and labor factor inputs to the AGP [34], thereby having a positive effect on the AGP. Thus, health may play a mediating role in the effects of CHE_1 on AGPA and AGPLs. The researchers selected the data measurement "Health" with the question "Do you think you are currently healthy? a = 1 = very unhealthy, b = 2 = unhealthy, c = 3 = average, d = 4 = healthy, e = 5 = healthy", and then takes "Health" as a mediating variable to conduct a mediating effect test. The results of model (3) in Table 7 show that CHE_1 is significantly and positively associated with "Health", indicating that CHE improves the health of families. In models (4) and (5), both CHE_1 and "Health" were significantly and positively correlated with AGPA and AGPLs, which means that "Health" plays a partial mediating role in the effect of CHE_1 on AGPA and AGPLs. This result passed both the Sobel test and the bootstrap test. Note: standard errors in parentheses, *** p < 0.01, ** p < 0.05. AGPA = agricultural green production awareness; AGPL = agricultural green production level; CHE_1 = frequency of clean energy use = cleaner household energy; MV_1 = mediating variable = health; CV = control variables.

Moderating Effect Analysis
Environmental protection awareness (EPA) is positively correlated with clean energy use and green production [22,31], implying that EPA may influence the effect of CHE on AGP. We chose the question "How does your household dispose of the waste generated in your life? a = dumping, b = burning, c = burying, d = disposal by a professional organization" to measure EPA; if "a", "b", and "c" are selected, the respondent is considered to have no EPA and is set to "EPA = 0", and if "d" is selected, the respondent is considered to have EPA and is set to "EPA = 1". EPA was used as a moderating variable to test whether EPA play a moderating role in the effect of CHE_1 on AGPA and AGPLs. Table 8 reports the results of the moderating effects analysis for EPA.
The results in Table 8 show that EPA was significantly and positively associated with AGPA and AGPLs, demonstrating that EPA significantly enhances farmers' AGPA, and significantly increases AGPLs. The interaction term between EPA and CHE_1 was significantly positively interconnected with AGPA and AGPLs at the 1% level, indicating that EPA plays a moderating role in the effect of CHE_1 on AGPA and AGPLs, and reinforces the positive effect of CHE_1 on AGPA and AGPLs.
Farmers with high digital literacy (DL) have access to more information on clean energy use and green production from the Internet, and the use of the Internet can promote CHE [58] and AGP [16], meaning that DL may play a moderating role in the impact of CHE on AGP. We selected the question "Can you participate in online shopping through your smartphone or computer? a = yes, b = no"; if "a" is selected, the respondent is considered to have DL and is set to "DL = 1", and if "b" is selected, the respondent is set to "DL = 0". DL was used as a moderating variable to test whether DL plays a moderating role in the effect of CHE_1 on AGPA and AGPLs. Table 9 reports the results of the moderating effects analysis for DL. Note: standard errors in parentheses, *** p < 0.01, ** p < 0.05. AGPA = agricultural green production awareness; AGPL = agricultural green production level; CHE_1 = frequency of clean energy use = cleaner household energy; EPA = environmental protection awareness; CV = control variables. Note: standard errors in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1. AGPA = agricultural green production awareness; AGPL = agricultural green production level; CHE_1 = frequency of clean energy use = cleaner household energy; DL = digital literacy; CV = control variables.
The results in Table 9 demonstrate that DL was significantly and positively associated with AGPA and AGPLs, showing that DL significantly enhances farmers' AGPA, and significantly increases AGPLs. The interaction term between DL and CHE_1 was significantly positively interconnected with AGPA and AGPLs at the 1% level, suggesting that DL plays a moderating role in the effect of CHE_1 on AGPA and AGPLs, and reinforces the positive effect of CHE_1 on AGPA and AGPLs.

Discussion
The current research assumes the belief that capital factors (financial and social capital) [18,19,25] and technological endowment [15,20] are the main factors affecting agricultural green production. At the same time, some of the literature discusses the impact of household energy consumption on agricultural green production. It is found that household energy cleaning can enhance farmers' environmental awareness [31], control agricultural source pollution [32], and improve climate conditions [29], thereby promoting agricultural green production. This work used micro-data from China as a sample to explore the relationship between cleaner household energy and agricultural green production. A core conclusion of this study is that cleaner household energy significantly enhanced agricultural green production awareness and significantly improved agricultural green production levels. This means that cleaner household energy has a promoting effect on agricultural green production, and the conclusions of this work are similar to those of Ma and Yue [32], Yasar et al. [34], Wu et al. [36], etc. The contributions of this paper include: (1) improving research on this topic while also providing new references for future research; (2) not only discussing whether cleaner household energy affects agricultural green production but it also analyzing how cleaner household energy influences agricultural green production through the mediating effect model and moderating effect model. However, there are limitations to this research. The sample data is drawn from one province in China, and the conclusions drawn may only be applicable to China or developing countries. Therefore, a further direction would be to compare and analyze data from developing and developed countries to draw conclusions with differentiation.

Conclusions
Cleaner household energy and agricultural green production are an imperative part of China's "carbon peaking and carbon neutrality goals", as an important way for developing countries to alleviate poverty, address food security, improve livelihoods, and act as an essential means of combating global warming and environmental degradation. The core finding of this study is that cleaner household energy can significantly contribute to agricultural green production, which, with reference to the "green revolution" in the energy and agriculture sectors, can also change energy consumption and agricultural production decisions in the household sector.
This research used 454 points of micro-survey data from Sichuan Province as the sample to analyze the effects and mechanisms of cleaner household energy on agricultural green production awareness and agricultural green production levels by using O-probit and OLS models. The empirical analysis found the following: First, cleaner household energy significantly enhanced agricultural green production awareness and significantly improved agricultural green production levels, which is still valid after addressing potential endogenous issues with the instrument variables method and conducting robustness tests. Second, health partially mediates the impact of cleaner household energy on agricultural green production awareness and agricultural green production levels. Third, environmental protection awareness and digital literacy have a moderating role in the impact of cleaner household energy on agricultural green production and reinforce the positive impact of cleaner household energy on agricultural green production awareness and agricultural green production levels.

Policy Recommendations
Based on the conclusions of this paper, some policy recommendations are proposed as follows: First, accelerate the development and conversion of clean energy, establish a sustainable and inclusive clean energy market mechanism, improve the construction of rural energy infrastructure, increase households' willingness to use clean energy through a subsidy mechanism, and promote the transformation of household energy into clean energy.
Second, improve the policy design for green agricultural development, provide more policy support to farmers involved in green agricultural production, increase farmers' willingness to produce green through a price mechanism, and improve farmers' green production capacity through regular agricultural training.