Effects of Rural Population Aging on Agricultural Carbon Emissions in China

Abstract

The “double carbon” goal (China aims to achieve carbon peak by 2030 and carbon neutrality by 2060) puts forward new requirements for the low-carbon development of agriculture. However, with the increasing aging of the rural population and the gradual aging of the agricultural labor force, determining the best means of achieving the target of reducing agricultural carbon emissions is particularly urgent. Based on the IPAT identity relationship (method of decomposing environmental impact (I) into socio-economic variables: population (P), affluence (A), and technology (T)), aging of the rural population, rural residents’ income, and agricultural technology innovation were selected as threshold variables. Using provincial panel data from 2003 to 2020 in China, this study empirically analyzed the impact of rural population aging on agricultural carbon emissions through a threshold–STIRPAT expansion model. The results showed that agricultural carbon emissions showed an inverted U-shaped growth trend from 2003 to 2020 and reached a peak in 2016. Baseline regression found that rural population aging has a significant emission reduction effect on agricultural carbon emissions. In addition, rural residents’ income and agricultural technology innovation have significant positive and negative impacts on agricultural carbon emissions, respectively. Using the three environmental factors as threshold variables, it was found that there is a significant threshold effect. The emission reduction effect of rural population aging weakens with the deepening of the aging degree but is enhanced with the improvement of rural residents’ income and agricultural technology innovation. In view of these findings, policy suggestions are put forward for agricultural low-carbon development that alleviates the effects of rural population aging, increases rural residents’ income, and strengthens agricultural technological innovation.

1. Introduction

Global climate governance is a major environmental issue, with carbon emissions generated by human activities being the main cause of climate change and the greenhouse effect. Although the current debate regarding global governance of climate change is mostly focused on the industrial and transportation sectors, the development of agriculture at the cost of high emissions cannot be ignored. In 2021, the Food and Agriculture Organization of the United Nations (FAO) released a report at the COP26 Climate Summit, which states that in the past 30 years, global greenhouse gas emissions from agriculture and food production increased by 17%. As early as 2019, anthropogenic CO₂ emissions from agri-food systems already accounted for 31% of global emissions. Agriculture has, thus, become an important source of global greenhouse gas emissions. China’s agriculture-related carbon emissions account for about a quarter of global emissions [1]. As the second largest source of agricultural greenhouse gas emissions [2], China has given prominence to addressing climate change in national governance [3]. In September 2020, China put forward the “dual-carbon” goal of achieving carbon peaking by 2030 and carbon neutrality by 2060. In 2021, the Chinese government issued the Opinions on Fully, Accurately and Comprehensively Implementing the New Development Concepts to Achieve Carbon Peak Neutrality and the Action Plan for Achieving Carbon Peak Before 2030; both plans explicitly stated that carbon emission reduction and sequestration in agriculture should be promoted. At the same time, China’s chronically low fertility rate, longer life expectancy, and accelerated urbanization process have made the problem of the aging rural population become increasingly prominent. According to the China Statistical Yearbook, the level of aging in rural China reached 17.7% in 2020, which is significantly higher than that in urban areas. In this context, ways of ensuring food security, clarifying the relationship between food security and emissions reduction, realizing the transformation of agricultural production to green, low-carbon and sustainable development, and effectively responding to global climate change are all particularly important issues.

The existing literature mainly studies China’s agricultural carbon emissions from two perspectives. Firstly, research has measured agricultural carbon emissions and explored the temporal and spatial characteristics and internal influencing factors. Most studies believe that in recent years, China’s total agricultural carbon emissions continued to increase and the growth rate decreased [4], while the efficiency of agricultural carbon emissions increased in most provinces [5]. The level of industrial agglomeration, agricultural development, agricultural financial support, and public investment have negative spatial spillover effects on agricultural carbon emissions [6]. Secondly, research has focused on agricultural carbon sequestration and emission reduction potential. Wu Xianrong et al. (2015) [7] found that the cost of agricultural carbon emission reduction is lower in Tibet and Qinghai, where the level of economic development is relatively low.

When discussing the relationship between population and carbon emissions, most scholars discussed the impact of population size, population structure, and labor quality on carbon emissions [8,9,10,11,12]. When it comes to population aging, scholars have drawn different conclusions due to differences in study area, time, and variable selection. Feng Yitao et al. (2023) [13] found that the aging of China’s population has a significant inhibitory effect on carbon emissions from 2005 to 2019. In contrast, Guo Hongwei et al. (2023) [14] believed that the increase in the aging population promoted carbon emissions from 1995 to 2019 in China. In terms of regions, the aging of the eastern region had a significant negative effect on carbon emissions, while in the central and western regions it had the opposite effect. Zheng Heran et al. (2022) [15] applied data from 32 developed countries during 2005–2015 and found that the elderly played a leading role in driving greenhouse gas emissions. In addition, some scholars have also discussed the non-linear relationship between the two factors. Yang Ting et al. (2020) [16] found that the higher the level of population aging, the greater the offset effect of population aging on carbon emissions. Wang Qiang et al. (2021; 2022) [17,18] reported population aging as the threshold variable to explore the non-linear relationship between industrial structure, urbanization, economic growth, energy consumption, and carbon emissions. When further focusing on the relationship between the rural population and agricultural carbon emissions, Ma Libang et al. (2022) [19] found that the improvement of the rural population quantity and quality could promote the growth of an agricultural net carbon sink, while population aging would lead to the decline of an agricultural net carbon sink.

Through a review of the existing literature, it can be found that previous studies focused on the population structure and population size when exploring the relationship between population factors and carbon emissions. In recent years, scholars began to refer to population aging; however, most of them were part of the research of this paper and failed to directly analyze the relationship between aging and carbon emissions. Moreover, most of the papers studied the whole of society or urban residents, but failed to focus on agriculture and rural farmers to explore the relationship between rural population aging and agricultural carbon emissions. In addition, existing studies on agricultural carbon emissions mainly focus on the measurement of emissions or emissions efficiency, exploring its influencing factors, and analyzing its temporal and spatial distribution characteristics; past studies rarely explored the internal impact of certain influencing factors on agricultural carbon emissions. By referring to previous studies and considering that there is no simple linear relationship between rural population aging and agricultural carbon emissions, this paper constructs the threshold–STIRPAT expansion model on the basis of existing studies and selects population, affluence, and technology as three important environmental factors based on IPAT identity. Using the panel data of 31 provinces and cities in China from 2003 to 2020, the specific impact of rural population aging on agricultural carbon emissions was explored when these three factors were altered.

The marginal contribution of this paper may be to explore the non-linear impact of rural population aging on agricultural carbon emissions by introducing two important variables—rural residents’ income and agricultural technology innovation—from the perspective of agriculture and rural areas, combined with the realistic background that rural labor population aging is gradually deepening. The paper then tries to clarify the influence mechanism of income and innovation in theory, suggesting countermeasures to realize a “double reduction and increase in efficiency” in the agricultural field that may also provide a reference for the sustainable development of agriculture in other countries.

2. Theoretical Foundation

In the short term, the increasing aging of the rural population reduces the size of the working-age labor force; thus, it is necessary to adjust agricultural production behavior to obtain the expected economic benefits. For example, mechanical planting has been employed to lessen the pressure of planting, removing the dependence on the labor force. In addition, when comparing the young and old rural labor forces, the older labor force tends to have a lower awareness of environmental protection. To pursue high crop yields, they often choose to increase the input of agricultural materials, such as fertilizers and pesticides, which are the main sources of agricultural carbon emissions. Studies have also found that there is a substitution relationship between aged labor and input factors, such as fertilizer machinery [20]. In the long term, due to the transfer of rural working-age workers to cities for non-agricultural employment, a large number of elderly rural workers not only have to rely on their own old-age support [21], but also have to take into account the family responsibility of taking care of left-behind children [22]. Farmers often adjust their land-use decisions according to their own conditions to alleviate livelihood pressures. Firstly, the planting structure is adjusted. On one hand, planting more time- and labor-saving cash crops reduces the size of food crops, though some cash crops require a greater fertilizer input. On the other hand, some elderly farmers in mountainous areas choose to plant trees, which not only alleviates the planting pressure but also reserves land for the elderly rural labor force. Trees can absorb and fix CO₂ in the atmosphere and have the function of absorbing carbon and releasing oxygen, conducive to realizing agricultural carbon sequestration and emission reduction. Secondly, land transfer takes place. Large-scale grain production and the efficient use of land can improve the adoption rate of low-carbon production technologies [23], which may lead to the reduction in carbon emissions.

When the income of rural residents is low, the prerogative of the rural elderly labor force is to increase their income, pay less attention to environmental issues, and use fertilizers to ensure food production as they lack spare funds to hire machinery, thus affecting agricultural carbon emissions. With the increase in income, the rural aging population has more money to choose mechanical production. Research also shows that with the increase in income, the rural aging population and agricultural mechanization show a U-shaped relationship from negative to positive [24]. Secondly, the increase in income can effectively improve the consistency of environmental concerns and environmentally friendly behavior [25]. Improving farmers’ livelihoods makes the elderly rural labor force focus not only on profit, but also provide green and pollution-free crops for relatives and friends, reducing the input of chemical fertilizers and pesticides in the production process and, thus, reducing agricultural carbon emissions.

Agricultural technological innovation can help ameliorate the dependence on the quantity of traditional agricultural production factors, making the utilization of production input factors more effective; thus, innovation alleviates the impact of the labor shortage caused by rural population aging on agricultural production and reduces the input of fertilizers, pesticides, and other materials. The current level of technology is the main reason for excessive agricultural carbon emissions from agricultural machinery and equipment [26]. With the continuous innovation and promotion of technology, new and clean energy can be promoted to replace traditional fossil fuel energy, improving the energy consumption structure and utilization efficiency, reducing the cost of low-carbon production behaviors, such as deep tillage and loose mechanical and straw returning to the field, and improve low-carbon efficiency [27]. In addition, methane and nitrous oxide released from the soil during crop production are important components of agricultural carbon emissions, while the variety of seeds, as key input factors, also determines the carbon emissions released from the soil to some extent. Therefore, agricultural carbon emissions will also be alleviated with the improvement and promotion of low-carbon emissions varieties.

In conclusion, this paper proposes the hypothesis that the influence of rural population aging on China’s agricultural carbon emissions is non-linear and affected by the threshold adjustment effects of rural population aging, rural residents’ income, and agricultural technology innovation. The details are shown in Figure 1.

3. Materials and Methods

3.1. Agricultural Carbon Emission Measurement Methods

This paper only focused on one narrow agricultural concept—the plantation industry—and drew on the studies of Tian Yun and Yin Minhao (2022) [6], Li Bo and Zhang Junbiao (2011) [28], and Min Jisheng and Hu Hao (2012) [29] to measure agricultural carbon emissions from two aspects.

On one hand, we considered the carbon emissions caused by farmland input. This study investigated the carbon emissions caused by the direct consumption of agricultural diesel or the indirect consumption of agricultural energy, specifically agricultural irrigation, as well as the carbon emissions caused by using agricultural materials, including chemical fertilizers, pesticides, and agricultural film. On the other hand, we considered soil carbon emissions released during three types of food production. Firstly, we considered how agricultural plowing destroys the soil organic carbon pool and causes a large amount of organic carbon to be lost to the air, increasing carbon emissions. Secondly, methane emissions from paddy fields caused during the planting of rice were considered. Due to the different temperature conditions in different regions, the methane emission rate in the growth cycle of rice varies in different regions, which can be divided into three categories: early, middle and late. Thirdly, nitrous oxide released from soil during crop planting was considered. In this paper, only five major crops, including rice, wheat, soybean, corn, and vegetables, were studied, among which wheat was divided into spring wheat and winter wheat.

The specific calculation equation is as follows:

$$C=\sum C_i=\sum T_i\times {\delta }_i$$

(1)

In Equation (1), $C$ represents the agricultural carbon emissions, $T$ represents the actual use of carbon sources, $\delta$ represents the carbon emission coefficient, and subscript $i$ represents the various types of carbon sources. In addition, for convenience, C, N₂O, and CH₄ are all converted into CO₂ in this paper. According to the fifth assessment report of the Intergovernmental Panel on Climate Change (IPCC), the conversion ratios for C, N₂O and CH₄ to CO₂ are 3.6667, 298, and 25, respectively.

3.2. Construction of the Threshold–STIRPAT Expansion Model

Based on the STIRPAT model of random form proposed by Dietz et al. (1994) [30], the IPAT equation was used to transform the model into a linear expansion model in logarithmic form to reduce the heteroscedasticity problem, as follows:

$$I=\alpha P^b{A}^c{T}^{d}e$$

(2)

$$\mathrm{l}\mathrm{n}I=\mathrm{l}\mathrm{n}\alpha +b\mathrm{l}\mathrm{n}P+c\mathrm{l}\mathrm{n}A+d\mathrm{l}\mathrm{n}T+\mathrm{l}\mathrm{n}e$$

(3)

In Equations (2) and (3), I, P, A and T represent environmental pressure, population structure, affluence, and technology level, respectively, where $\alpha$ is a constant and $e$ is the error term.

Combined with the research purpose, the STIRPAT model was extended in two aspects. Firstly, the variables in the STIRPAT model were redefined. Environmental pressure ($I$), population structure ($P$), affluence ($A$), and technology level ($T$) were re-defined as agricultural carbon emission (carbon), rural aging (aging), rural income (income), and agricultural technology innovation (tech). Secondly, considering the important impacts of urbanization (urban), agricultural trade (trade), industrial structure (structure), energy input (energy), and medical level (doctor) on agricultural carbon emissions, the STIRPAT extension model was established by introducing these variables into the model as control variables:

$$\begin{array}{c}\mathrm{l}\mathrm{n}carbon=\mathrm{l}\mathrm{n}\alpha +b\mathrm{l}\mathrm{n}aging+c\mathrm{l}\mathrm{n}income+d\mathrm{l}\mathrm{n}tech+e\mathrm{l}\mathrm{n}urban\\ +f\mathrm{l}\mathrm{n}trade+g\mathrm{l}\mathrm{n}structure+h\mathrm{l}\mathrm{n}energy+i\mathrm{l}\mathrm{n}doctor+\mathrm{l}\mathrm{n}e\end{array}$$

(4)

Considering that the influence degree of rural population aging on agricultural carbon emissions may change with aging, rural residents’ income, and the improvement of the agricultural technology innovation level, the threshold model and STIRPAT expansion model were combined according to Liu Chao et al. (2022) [31], while the threshold–STIRPAT expansion model was constructed as:

$$\mathrm{l}\mathrm{n}carbon=\beta +b\mathrm{l}\mathrm{n}aging·I(q\le \lambda )+c\mathrm{l}\mathrm{n}aging·I(q>\lambda )+dX+\epsilon$$

(5)

In Equation (5), $I\left(·\right)$ is the indicative function, and the sample is divided according to the threshold value. $q$ is the threshold variable, including the aging of the rural population, rural residents’ income, and agricultural technology innovation. $\lambda$ is the threshold value. If there are double or multiple thresholds, more can be introduced. $X$ is the set of other variables affecting agricultural carbon emissions, $\beta$ is a constant, and $\epsilon$ is the error term.

3.3. Index Selection and Data Source

Since the concept of a “low-carbon economy” was first put forward in 2003, China has regarded energy conservation and emission reduction as the focus of environmental protection work and constantly strengthened the establishment of a sound system of laws and policies related to low carbon [26]. This paper used panel data from 31 provinces from 2003 to 2020 to empirically explore the impact of rural population aging on agricultural carbon emissions.

Explained variable: Agricultural carbon emission (carbon), in which the corresponding carbon emission coefficients of agricultural energy utilization, agricultural material input, land consolidation, and irrigation were obtained from Li Bo and Zhang Junbiao et al. (2011) [28]. The corresponding carbon emission coefficients of rice and other crops were derived from the research of Min Jisheng and Hu Hao (2012) [29]. Data on the actual use of agricultural diesel oil, chemical fertilizers, pesticides, agricultural film, irrigated area, and crop sown area of each region in that year were sourced from the National Bureau of Statistics.

Core explanatory variable: The core explanatory variable was the aging of the rural population(aging), measured by the proportion of the rural resident population aged 65 years olf or older. The data came from the China Population and Employment Statistical Yearbook published by the National Bureau of Statistics.

Threshold variables: According to the IPAT identity relationship, starting from the three important aspects that affect the environment (population, affluence and technology, rural population aging, rural residents’ income, and agricultural technology innovation) were selected as threshold variables. The income of rural residents (income) was measured by the per capita disposable income of rural residents. Considering the lack of relevant data on agricultural mechanization innovation level, some innovative crop varieties could alleviate soil carbon emissions released in the production process and reduce the dependence on input factors, such as fertilizers and pesticides. Therefore, this paper only used the innovation ability of new plant varieties to quantify agricultural technology innovation. (tech) This measure was the weighted average of the number of applications and authorizations for new plant variety rights. Relevant data came from the China Statistical Yearbook and Seed Industry Big Data Platform (http://202.127.42.145/bigdataNew/. accessed on 10 January 2023).

Control variables: (1) Urbanization (urban) was measured by the proportion of urban residents in the total population. The process of urbanization is also a process of industrial and population agglomeration, which is often accompanied by the expansion of demand and economic development, increasing energy consumption and, thus, an increase in carbon emissions. In addition, with the gradual increase in the urbanization level, the rural working-age labor force moves to cities on a large scale, which increases the aging level of the rural elderly population; different age groups have different consumption tendencies, environmental attitudes, and production decisions, thus affecting agricultural carbon emissions. (2) The trade in agricultural products was divided into the import trade (import) and export trade of agricultural products (export). This was measured by dependency, that is, the proportion of import or export trade of agricultural products in the added value of the primary industry. The trade of agricultural products causes the transfer of carbon emissions between exporting and importing regions, thus affecting carbon emissions. (3) Industrial structure(structure) was calculated by the proportion of added value of the secondary and tertiary industries to GDP. For a long time, China’s economic growth mainly relied on secondary industries, among which the annual average carbon emissions of the manufacturing industry accounted for 26% [32], making it a carbon-emission-intensive industry. However, tertiary industries have a relatively low energy intensity, which can significantly reduce the carbon emission intensity [33]. Therefore, upgrading the industrial structure is closely related to carbon emissions. (4) Energy input (energy) was measured by the rural electricity consumption per hour. At present, the proportion of rural hydropower is relatively small, with the majority of energy being generated by thermal power, which requires large amounts of carbon combustion, thus increasing agricultural carbon emissions. (5) Medical level (doctor) was measured by the number of rural doctors and healthcare workers per 10,000 rural population in each region. The improvement in the medical level has alleviated the contradiction between supply and demand for rural medical factors, creating a livable environment, and maintaining the health of the elderly population in rural areas. However, the carbon emissions of the medical industry are relatively high. According to the Roadmap of China’s Medical Carbon Neutrality in 2022—Medical Institutions published by the Yiou Think Tank, the carbon footprint of China’s medical system alone accounts for 17% of the country’s carbon emissions. The relevant data were retrieved from the China Statistical Yearbook, China Monthly Statistical Report on Import and Export of Agricultural Products (December), and China Rural Statistical Yearbook.

Since Equation (3) is in logarithmic form, all variables are processed logarithmically in this paper. The descriptive statistics of the processed variables are shown in

Table 1. Descriptive statistics of variables.

Variable	Observations	Mean	Standard Error	Minimum	Maximum
lncarbon	558	13.058	1.154	9.376	14.684
lnaging	558	2.316	0.327	1.531	3.261
lnincome	558	8.884	0.676	7.391	10.461
lntech	558	3.310	1.371	0.000	6.443
lnurban	558	3.898	0.346	2.743	4.501
lnimport	558	1.804	1.709	0.000	7.425
lnexport	558	1.637	1.250	−1.689	5.040
lnstructure	558	4.483	0.069	4.186	4.602
lnenergy	558	13.673	1.571	8.366	16.817
lndoctor	558	2.596	0.485	0.808	3.998

Note: Carbon was calculated by authors and relevant data came from Li Bo and Zhang Junbiao et al. (2011) [28], Min Jisheng and Hu Hao (2012) [29], National Bureau of Statistics; Aging data were obtained from China Population and Employment Statistical Yearbook published by the National Bureau of Statistics; Data of income, urban and structure were obtained from China Statistical Yearbook; Tech data came from Seed Industry Big Data Platform (http://202.127.42.145/bigdataNew/ accessed on 10 January 2023); Import and export data are from China Monthly Statistical Report on Import and Export of Agricultural Products (December); Energy and doctor data are from China Rural Statistical Yearbook.

.

4. Results and Discussion

4.1. Analysis of the Agricultural Carbon Emission Calculation Results

Agricultural carbon emissions calculated above are shown in Figure 2. The results showed that, China’s agricultural carbon emissions were on the rise from 2003 to 2015, reaching their highest level in 2015 at 25.24238 million tons; this figure represents a total increase of 39.244% and an average annual growth rate of 3.055%. From 2016 to 2020, China’s agricultural carbon emissions have declined, which is consistent with the results of Ding Baogen et al. (2022) [34]. The biggest drop was observed in 2019 (4.067%). Agricultural carbon emission intensity (measured by the ratio of agricultural carbon emissions to agricultural output value per million Yuan) has been on a downward trend, decreasing from 12.191kg/million Yuan of agricultural output value in 2003 to 3.049kg/million Yuan of agricultural output value in 2020. China has made some advances in agricultural carbon emission reduction in recent years.

In terms of sources of agricultural carbon emissions, agricultural materials, such as fertilizers, pesticides and agricultural films, are the main sources of agricultural carbon emissions in China, accounting for 81.957% of the total. It is worth noting that although the real reduction in chemical fertilizers and pesticides began in 2016, the proportion of chemical fertilizers and pesticides has been decreasing year-by-year since 2009, from 69.036% in 2009 to 66.703% in 2020. These figures indicate that with the continuous development of the social sciences and technology in China in recent years, the utilization rate of agricultural materials has gradually increased and people’s requirements for food consumption have gradually changed from “enough to eat” to “eat well and eat healthily”, increasing the demand for greener pollution-free agricultural products.

4.2. Analysis of the Empirical Regression Results

4.2.1. Model Foundation Test

Before threshold regression, the variance inflation factor (VIF) was first measured. The results showed that the maximum VIF value was 5.863, less than 10, and the mean value was 3.326, less than 5, indicating that there was no multicollinearity problem for each variable and the variable selection was reasonable to a certain extent. Secondly, the LLC and IPS methods were used to test the stability of the panel data. The results are shown in

Table 2. Unit root test results of panel.

Variable	LLC Test				IPS Test
	Original Sequence		First Difference Sequence		Original Sequence		First Difference Sequence
	t-Statistic	p Value	t-Statistic	p Value	t-Statistic	p Value	t-Statistic	p Value
lncarbon	3.874	1.000	−6.779	0.000	4.763	1.000	−6.213	0.000
lnaging	0.998	0.841	−6.362	0.000	0.979	0.164	−9.469	0.000
lnincome	−7.229	0.000	−9.642	0.000	−3.980	0.000	−6.918	0.000
lntech	−2.952	0.002	−10.925	0.000	−2.413	0.008	−11.913	0.000
lnurban	−11.802	0.000	−83.651	0.000	−11.497	0.000	−31.445	0.000
lnimport	−4.025	0.000	−7.693	0.000	−1.259	0.104	−7.674	0.000
lnexport	2.891	0.998	−8.885	0.000	1.059	0.855	−8.088	0.000
lnstructure	2.028	0.979	−5.862	0.000	2.420	0.992	−6.483	0.000
lnenergy	−3.347	0.000	−4.862	0.000	0.246	0.597	−5.187	0.000
lndoctor	−10.689	0.000	−13.229	0.000	−6.351	0.000	−10.790	0.000

. The first-order difference sequences of all the variables passed the unit root test and the data was stable.

4.2.2. Baseline Regression Analysis

Before the threshold effect test, individual effects, random effects, and Hausman tests were carried out on the panel data. The results showed that the mixed regression model and random effects model were rejected and the fixed effects model was determined as the optimal model. The regression results are shown in

Table 3. Results of baseline regression.

Variable	Coefficient	Robust Standard Error	t-Statistic	p Value
lnaging	−0.098 **	0.047	−2.100	0.036
lnincome	0.132 ***	0.031	4.190	0.000
lntech	−0.045 ***	0.010	−4.530	0.000
lnurban	0.002	0.049	0.050	0.962
lnimport	−0.072 ***	0.022	−3.230	0.001
lnexport	0.016	0.016	1.010	0.311
lnstructure	−0.545 *	0.288	−1.890	0.059
lnenergy	0.101 ***	0.027	3.730	0.000
lndoctor	0.318 ***	0.029	10.820	0.000
Constant term	12.595 ***	1.181	10.670	0.000

Note: ***, **, and * represent a significance level of 1, 5, and 10%, respectively.

.

Rural population aging and the agricultural technology innovation coefficient were both significantly negative, effectively reducing agricultural carbon emissions, while the rural residents’ income coefficient was significantly positive, that is, the increase in income increased agricultural carbon emissions. Among the control variables, the urbanization coefficient was positive but not significant. The possible reason is that the urbanization level is gradually rising and the rural working-age labor force is transferring to cities on a large scale. On one hand, energy consumption is reduced by reducing rural demand; on the other hand, the input factors of the agricultural production labor force are reduced and the input of alternative factors, such as fertilizers and pesticides, are increased, thus increasing agricultural carbon emissions. This results in no significant impact on urbanization. The coefficients of energy input and medical level were significantly positive, which increased agricultural carbon emissions. For agricultural trade, the coefficient of agricultural import trade was significantly negative. The possible reason is that, currently, China mainly imports land-intensive products, such as oil and cotton, which can effectively relieve environmental pressures, thus reducing agricultural carbon emissions [35]. The coefficient of agricultural export trade was not significant, indicating that there was no significant correlation between agricultural export trade and agricultural carbon emissions, which is consistent with the research conclusions of Liu Qian and Wang Yao (2012) [36]. The industrial structure coefficient was significantly negative, that is, the positive impact of transforming and upgrading tertiary industries on carbon emissions is mitigated by the negative impact of the secondary industries.

4.2.3. Threshold Effect Regression Analysis

Based on the above results, this paper further examined the threshold effect of rural population aging, rural residents’ income, and agricultural technology innovation on the emission reduction effect of rural population aging. The threshold effect was tested 300 times by bootstrapping, the results of which are shown in

Table 4. Threshold effect self-sampling of 300 test results and confidence interval.

Argument	Threshold Variable	Hypothesis Testing	RSS	MSE	F Value	p Value	Threshold Value	95% Confidence Interval
lnaging	lnaging	Single threshold	7.130	0.013	41.110	0.043	1.968	[1.963, 1.970]
		Double threshold	6.825	0.013	24.190	0.147
		Triple threshold	6.633	0.012	15.650	0.593
	lnincome	Single threshold	6.735	0.013	96.080	0.000	8.154	[8.141, 8.161]
		Double threshold	5.885	0.011	78.010	0.000	10.037	[9.988, 10.046]
		Triple threshold	5.210	0.010	69.950	0.507
	lntech	Single threshold	7.366	0.014	44.780	0.067	4.213	[4.174, 4.222]
		Double threshold	6.831	0.013	42.300	0.033	5.783	[5.669, 5.787]
		Triple threshold	6.701	0.012	10.490	0.850

. The rural population aging threshold variable model was only significant in the single threshold test estimation, that is, there was a non-linear relationship and a single threshold of 1.968, with a corresponding value of 7.156% from the logarithmic form. The three threshold variable models of rural resident income and agricultural technology innovation were significant in the single and double threshold test estimations, that is, there was a non-linear relationship and double threshold, with double threshold values of 8.154, 10.037 and 4.213, 5.783, respectively. The corresponding values were 3480, 22,860 and 66.559, 323.732, respectively. The hypotheses proposed in this paper are, thus, verified.

To further clarify the influence size, rural population aging, rural residents’ income, and agricultural technology innovation were taken as threshold variables to carry out threshold model estimation; the regression results are shown in

Table 5. Threshold regression estimation results.

	Model 1	Model 2	Model 3
Threshold variable $q$	lnaging	lnincome	lntech
lnaging ($q$≤ η)	−0.236 ***
	(0.050)
lnaging ($q$ > η)	−0.150 ***
	(0.046)
lnaging ($q$ < $\lambda$1)		−0.078 **	−0.034
		(0.037)	(0.045)
lnaging ($\lambda$1 < $q$≤$\lambda$2)		0.018	−0.077 *
		(0.036)	(0.044)
lnaging ($q$ > $\lambda$2)		−0.081 **	−0.155 ***
		(0.037)	(0.044)
lnincome	0.124 ***		0.149 ***
	(0.030)		(0.028)
lntech	−0.042 ***	−0.011
	(0.010)	(0.008)
Control variable	Control	Control	Control
Observations	558	558	558
R-squared	0.563	0.636	0.582

Note: ***, ** and * represent a significance level of 1, 5, and 10%, respectively. Standard error in parentheses.

.

There was a non-linear negative correlation between rural population aging and China’s agricultural carbon emissions. However, as the threshold of 7.156% was crossed, the emission reduction effect of rural population aging weakened (Model 1). The possible reasons are that, on the one hand, with the gradual deepening of aging, an increasing proportion of the rural aging population quit agricultural production and carried out land transfer, promoting the large-scale operation of land and improved mechanization level, thus increasing agricultural carbon emissions. On the other hand, an increasing proportion of the rural aging population choose cash crop varieties that save time and labor and are easy to grow. Compared with food crops, fertilizer input is increased and the N₂O content released by soil also increases.

When the income level of rural residents was lower than threshold 3480 CHN, rural population aging increased by 1% and agricultural carbon emissions decreased by 7.8%. When the income level was between 3480 and 22,860 CHN, the coefficient of rural population aging was positive but not significant. Upon crossing the threshold, the inhibition effect of rural population aging on agricultural carbon emissions was significantly enhanced (Model 2). This may be because when the income is too low, the rural elderly labor force are more inclined to plant by themselves, lack employment funds, and use less machinery, as shown by the reduced emissions. With the increase in income, the aging population in rural areas first choose to improve their lives rather than adjust the agricultural production mode. With a continuous increase in income and a solution to livelihood problems, they then choose to change the planting structure and promote environmentally friendly behaviors, reducing agricultural carbon emissions.

When agricultural technology innovation was lower than the threshold value of 66.559, the influence of rural population aging on agricultural carbon emissions was negative but not significant. However, when the threshold value was crossed, the increase in rural population aging by 1% reduced agricultural carbon emissions by 7.7% until it exceeded the threshold value of 323.732. If rural population aging increased by 1%, the step effect of rural population aging on emission reduction rose to 15.5% (Model 3). This indicates that rural population aging does not inhibit agricultural carbon emissions due to the improvement of agricultural technological innovation; rather, only when the threshold is crossed does the emission reduction effect of rural population aging appear.

4.3. Robustness Test

To verify the robustness of the previous estimation results, the following two robustness tests were carried out: Firstly, the dependent variable was replaced and the dependency ratio of the elderly population in rural areas was taken as a measure of rural population aging. The 300 self-sampled tests showed that there were double thresholds. The regression results for Models 1–3 are shown in

Table 6. Robustness test regression results.

	Change-Dependent Variable				Adjustment Empirical Method
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
Threshold variable $q$	lnaging	lnincome	lntech			lnincome		lntech
lnaging ($q$ < $\lambda$1)	−0.354 ***	−0.098 ***	−0.075 **
	(0.046)	(0.032)	(0.038)
lnaging ($\lambda$1 < $q$ ≤ $\lambda$2)	−0.280 ***	−0.016	−0.111 ***
	(0.042)	(0.031)	(0.037)
lnaging ($q$ > $\lambda$2)	−0.232 ***	−0.102 ***	−0.175 ***
	(0.039)	(0.032)	(0.037)
lnaging				1.695 ***	1.852 ***	−1.462 ***	1.675 ***	0.677 **
				(0.255)	(0.254)	(0.418)	(0.260)	(0.323)
lnaging 2				−0.367 ***	−0.382 ***	0.346 ***	−0.368 ***	−0.146 **
				(0.051)	(0.052)	(0.093)	(0.052)	(0.068)
lnincome	0.133 ***		0.161 ***	0.115 ***		0.078 ***	0.059 **	0.133 ***
	(0.029)		(0.028)	(0.030)		(0.027)	(0.028)	(0.030)
lntech	−0.037 ***	−0.008		−0.045 ***	−0.030 ***	−0.038 ***		−0.050 ***
	(0.009)	(0.008)		(0.009)	(0.009)	(0.009)		(0.009)
lnaging*$q$						−0.773 ***		−0.217 *
						(0.246)		(0.120)
lnaging2*$q$						0.050		0.027
						(0.057)		(0.026)
Control variable	Control	Control	Control	Control	Control	Control	Control	Control
Observations	558	558	558	558	558	558	558	558
R-squared	0.588	0.637	0.583	0.572	0.560	0.659	0.554	0.593

Note: ***, **, and * represent a significance level of 1, 5, and 10%, respectively. Standard error in parentheses. Models 1–3 are regression results with lnaging, lnincome, and lntech as threshold variables after replacing core explanatory variables. Models 4–5 and 7 are baseline regression of square term (lnaging2) of rural population aging variable. Model 6 is regression result of increasing lnincome’s interaction with lnaging and lnaging2. Model 8 is regression result of increasing lntech’s interaction with lnaging and lnaging2.

. The direction and relative size of the core explanatory variables did not change, which is consistent with the previous analysis, and the three hypotheses proposed above were verified again.

Secondly, the empirical method was adjusted to test the research conclusion again. In view of the non-linear relationship between rural population aging and agricultural carbon emissions, the square term of the rural population aging variable (lnaging2) was introduced in the regression. Subsequently, in order to verify the conclusion that the relationship between rural population aging and agricultural carbon emissions was affected by the moderating effect of rural residents’ income and agricultural technology innovation, the interaction terms of rural population aging and its square term were added to the regression, respectively. At the same time, in order to reduce the influence of multicollinearity, this paper carried out centralized processing on the interaction variables. The regression results for Models 4–8 are shown in Table 6. Rural population aging and its coefficient of square term were both significant at the 1% level and the symbols were opposite, confirming a non-linear relationship between rural population aging and agricultural carbon emissions. According to the results of Model 4, there was an “inverted U-shaped” relationship between the aging level and agricultural carbon emissions, which first increased and then decreased when the other control variables were unchanged.

In Models 6 and 8, the adjustment effect of the primary term was significantly negative, that is, the inflection points and symmetry axis of the inverted U-shaped curve of the main effect were negatively affected by the income of rural residents and the level of agricultural technology innovation. In other words, when rural residents’ income and agricultural technology innovation are at a higher level, rural population aging plays a role in emission reduction. This result is basically consistent with the threshold regression results above, which further confirm the robustness of the study’s conclusion.

4.4. Heterogeneity Analysis

On one hand, due to the large differences in agricultural carbon emissions among provinces and cities in mainland China, in order to more accurately examine the relationship between them, this paper adopted a K-means clustering analysis to divide the 31 provinces in China into two categories according to their regional agricultural carbon emissions in 2020 (

Table 7. Classification of agricultural carbon emission levels in different regions.

Category	Area
Regions with high levels of agricultural carbon emissions	Hebei, Inner Mongolia, Jilin, Heilongjiang, Jiangsu, Anhui, Shandong, Henan, Hubei, Hunan, Guangdong, Guangxi, Sichuan, Yunnan, Shaanxi, Xinjiang
Regions with low agricultural carbon emissions	Beijing, Tianjin, Shanxi, Liaoning, Shanghai, Zhejiang, Fujian, Jiangxi, Hainan, Chongqing, Guizhou, Xizang, Gansu, Qinghai, Ningxia

). Moreover, through traversing the contour coefficient, it was found that the contour coefficient of the two categories was 0.732, proving a good clustering effect. The estimated results for Models 1 and 2 are shown in

Table 8. Regression results of heterogeneity test.

	Classification of Agricultural Carbon Emission Levels		National Bureau of Statistics Division
	High Level	Low Level	Eastern	Central	Western
	Model 1	Model 2	Model 3	Model 4	Model 5
lnaging	−0.218 ***	−0.014	−0.101 *	−0.082	−0.276 ***
	(0.062)	(0.060)	(0.059)	(0.060)	(0.080)
lnincome	0.137 ***	0.070 *	0.011	0.230 ***	0.411 ***
	(0.046)	(0.039)	(0.044)	(0.055)	(0.049)
lntech	−0.029 **	−0.059 ***	−0.079 ***	−0.054 ***	0.001
	(0.014)	(0.012)	(0.016)	(0.015)	(0.012)
Control variable	Control	Control	Control	Control	Control
Observations	288	270	198	144	216
R-squared	0.693	0.577	0.541	0.805	0.789

Note: ***, **, and * represent a significance level of 1, 5, and 10%, respectively. Standard error in parentheses. According to classification criteria of the National Bureau of Statistics, western region consists of 12 provinces: Gansu, Ningxia, Qinghai, Xinjiang, Shaanxi, Sichuan, Chongqing, Guizhou, Yunnan, Inner Mongolia, Guangxi, and Tibet. Central region consists of eight provinces: Anhui, Jiangxi, Henan, Hubei, Hunan, Jilin, Shanxi, and Heilongjiang. Eastern region comprises 11 provinces: Jiangsu, Zhejiang, Shandong, Beijing, Shanghai, Tianjin, Guangdong, Liaoning, Hainan, Fujian, and Hebei.

. In Model 1, the coefficient of rural population aging was significantly negative at the 1% level, indicating that rural population aging in areas with high agricultural carbon emissions can effectively reduce agricultural carbon emissions, while in Model 2 the coefficient of rural population aging was not significant. The reason may be that most regions with high agricultural carbon emissions are large agricultural provinces with rich land resources. With the increase in rural population aging, it is easier to organize large-scale land management, improving the utilization efficiency of input factors and, thus, reducing agricultural carbon emissions. However, in areas with low agricultural carbon emissions, rural households have relatively high non-agricultural employment income as agriculture is not their main source of family income. As a result, the elderly population does not rely on land to survive and, thus, has negligible impact on agricultural carbon emissions.

On the other hand, differences in location and the degree of economic development also lead to different impacts of rural population aging on agricultural carbon emissions. This paper further divided the eastern, central and western regions of China into three groups to explore differences in the emission reduction effects of rural population aging among the regions. The estimated results for Models 3–5 are shown in Table 8. In the eastern and western regions, the coefficients of rural population aging were significantly negative, that is, with the deepening of rural population aging, the agricultural carbon emissions can be effectively reduced; rural population aging in the western region had a stronger emission reduction effect. The reason for this may be due to the fact that the western region is mainly composed of basins, mountains, and plateaus where agricultural production is complex and changeable. With the gradual aging of the rural population, the rural population become more inclined to choose tree planting with a lower dependence on the labor force, thus showing a stronger role in reducing emissions. In the central region, the coefficient of rural population aging was not significant. The reason may be that the influence of rural population aging on agricultural carbon emission may be closely related to income, which can also be verified by the above threshold regression results. When the income is at the medium level, there was no significant influence between the two factors.

4.5. Discussion

The above studies have confirmed that rural population aging, income, and innovation have significant threshold-regulating effects on reducing agricultural carbon emissions by population aging. It is worth mentioning that, first of all, the single threshold value of rural population aging was 7.156%; this figure s similar to the 7% standard determined by the United Nations in 1956, proving that an aging society will have a serious impact on the social economy, while also ensuring the accuracy of the model settings and empirical results to a certain extent. Secondly, when the income level of rural residents was between 3480 and 22,860, the impact of rural population aging on agricultural carbon emissions was not significant. On one hand, it may be that compared with the previous low incomes, the increased incomes are barely enough to cover healthcare, savings, retirement, etc., which is inadequate to sway the rural aging population to change their mode of production. On the other hand, at this stage China may face the risk of the “middle-income trap”. This is at the top of the environmental Kuznets curve, where the degree of environmental pollution caused by income reaches a peak, making the promoting effect of income on carbon emissions difficult to offset with the emission reduction effect of income on rural population aging. Moreover, in this stage capital is relatively abundant, the mechanization degree deepens, and environmental protection awareness is not significantly enhanced. The low adoption of low-carbon production behaviors failed to make a significant impact on the two factors. In addition, when agricultural technology innovation was low and did not cross the threshold value, the emission reduction effect of rural population aging was not apparent; however, with the promotion of agricultural technology innovation, rural population aging transforms agricultural carbon emissions from “quantitative change” to “qualitative change”. Only when the threshold value is crossed can the qualitative improvement of the emission reduction effect of rural population aging be realized.

Aging, income, and innovation are not completely independent. The improvement of the innovation level requires higher economic support. With the growth of the economy, the increase in income may also lead to an improvement in fertility intention, thus alleviating population aging. Therefore, this is a complicated process. Deciding how to coordinate the relationship among the three factors to reduce agricultural carbon emissions and achieve sustainable agricultural development should be the focus of further research.

5. Conclusions and Suggestions

Firstly, by measuring the agricultural carbon emissions of 31 provinces in China from 2003 to 2020, it is found that agricultural carbon emissions showed an inverted U-shaped growth trend that peaked in 2016, with agricultural materials, such as chemical fertilizers and pesticides, being the main source of China’s agricultural carbon emissions. Therefore, it is necessary to continue to promote the research and development of degradability and recycling efficiency of agricultural films. We must continue to promote the use of technologies such as testing soil for formulated fertilization and water-saving irrigation to reduce the amount of fertilizers and pesticides used and increase their efficiency, thus reducing energy consumption.

Secondly, the study found that rural population aging is conducive to reducing agricultural carbon emissions but, with the gradual deepening of the aging level, this effect will gradually weaken. Therefore, the comprehensive relaxation of birth restrictions will improve the fertility rate, delay the aging process, and avoid a series of adverse social results caused by the high aging level. In addition, we should actively respond to the pressure and challenges brought by rural population aging by improving infrastructure construction related to the elderly in rural areas; improving the happiness of rural elderly residents and social participation is the best strategy to reduce agricultural carbon emissions.

Thirdly, the study found that when income reaches a certain level, the emission reduction effect of rural population aging can be effectively improved, though the effect is not significant when the rural population is at the middle-income level. Therefore, improving rural residents’ income is still the focus of China’s rural development. Policies to increase income and benefit the people should continue to be effectively implemented to avoid falling into the “middle-income trap”.

Fourthly, the study found that when agricultural technology innovation was lacking, the emission reduction effect of rural population aging was not significant; however, upon crossing the threshold value, the emission reduction effect was significant and stronger. Therefore, we should continue to increase investment in research and development and accelerate the implementation and transformation of agricultural technological innovation advancements. We should strengthen the support for the research and development of low-carbon crop varieties and breakthrough low-carbon farming technologies, such as high-nitrogen crop varieties, and improve factor input technology and utilization efficiency.

In addition, it is worth mentioning that the environmentally conscious public attribute of agricultural carbon emissions makes it difficult to obtain full investment returns from low-carbon production, thus reducing investment enthusiasm. This means that governments need to compensate producers for the environmental costs and set up a carbon emission compensation mechanism. On one hand, at the inter-provincial level major grain-selling provinces should compensate the environmental costs borne by major grain-producing provinces, forming a cross-regional joint agricultural carbon emission reduction compensation mechanism. On the other hand, at the individual level, grain producers should be compensated for carbon emissions, such as production cost compensation, to improve farmers’ enthusiasm in adopting low-carbon production behaviors.