论文信息
论文解读
AI 辅助整理- 方法
- 31省2011-2023面板数据,空间杜宾模型与中介模型。
- 发现
- 数字经济显著降低本地及邻省农业碳排放,效应存在区域与时段异质性。
- 机会点
- 可深入探究2019年后本地减排效应减弱的原因及跨区域技术扩散的传导路径。
论文摘要
为探究数字经济促进农业低碳转型的空间效应与作用机制,本文基于2011—2023年中国31个省份的面板数据,测度数字经济发展水平与农业碳排放量,并分析其空间分布特征,运用空间杜宾模型和中介模型考察数字经济对农业碳排放的影响效应及作用机制。结果表明:数字经济与农业碳排放均呈现“高—高、低—低”的空间集聚格局,农业碳排放具有显著的正向空间溢出效应。数字经济显著降低本地及邻近省份的农业碳排放,且存在明显的区域异质性。在粮食主产区,数字经济仅抑制本地农业碳排放;在非粮食主产区,数字经济对本地碳排放的抑制作用较弱,但可通过跨区域技术扩散降低相邻地区碳排放。2011—2018年,农业碳排放呈现空间集聚格局,数字经济对本地和邻近省份均有显著减排效应;而2019—2023年,农业碳排放转向空间竞争格局,数字经济的本地直接减排效应逐渐减弱并变得统计不显著,对邻近省份的抑制作用亦有所减弱。财政支农强化了数字经济的碳减排效应,经济发展水平越高,这一抑制效应越强。农村创业活力与农业全要素生产率在数字经济驱动农业碳减排过程中发挥显著的中介作用。研究结论为强化数字经济的农业碳减排作用提供实证支持,为完善农业低碳发展政策体系提供参考。
全文
Frontiers in Sustainable Food SystemsFrontiers in Sustainable Food Systems 01 frontiersin.org
TYPE Original Research
PUBLISHED 17 November 2026
DOI 10.3389/fsufs.2026.1942827
OPEN ACCESS
EDITED BY
Shivani Garg,
Kurukshetra University, India
REVIEWED BY
Daleng Xin,
Shandong Normal University, China
Urvashi Kashyap,
Kurukshetra University, India
Prashant Ravish,
Kurukshetra University, India
CORRESPONDENCE
Fu Zeng
648714912@qq.com
RECEIVED 20 July 2026
REVISED 28 October 2026
ACCEPTED 02 November 2026
PUBLISHED 17 November 2026
CITATION
Hou X, Xu Z, Zhang Q, Zhang D and
Zeng F (2026) Spatial effects and
mechanisms of the digital economy’s
impact on agricultural carbon emissions.
Front. Sustain. Food Syst. 10:1942827.
doi: 10.3389/fsufs.2026.1942827
COPYRIGHT
© 2026 Hou, Xu, Zhang, Zhang and
Zeng. This is an open-access article
distributed under the terms of the
Creative Commons Attribution License
(CC BY). The use, distribution or
reproduction in other forums is
permitted, provided the original
author(s) and the copyright owner(s) are
credited and that the original publication
in this journal is cited, in accordance
with accepted academic practice. No
use, distribution or reproduction is
permitted which does not comply with
these terms.
Spatial effects and mechanisms
of the digital economy’s impact
on agricultural carbon emissions
Xuefeng Hou 1, Zihan Xu 1, Qiang Zhang 2, Dianfeng Zhang 3 and
Fu Zeng 2
1School of Economics and Management, Yingkou Institute of Technology, Yingkou, China, 2School of
Economics and Management, Anqing Normal University, Anqing, China, 3School of Management,
Qufu Normal University, Rizhao, China
To explore the spatial effects and mechanisms of the digital economy in promot-
ing low-carbon agricultural transformation, this paper measures digital econ -
omy development and agricultural carbon emissions based on panel data for
31 provinces in China from 2011 to 2023, and analyzes their spatial distribution
characteristics. Spatial Durbin and mediating models are employed to investigate
the effects and mechanisms of the digital economy on agricultural carbon emis-
sions. The results show that both the digital economy and agricultural carbon
emissions present a “high-high and low-low” spatial agglomeration pattern, and
agricultural carbon emissions have a significant positive spatial spillover effect.
The digital economy significantly reduces agricultural carbon emissions in local
and neighboring provinces, with obvious regional heterogeneity. In major grain-
producing areas, the digital economy only inhibits local agricultural carbon emis-
sions. In non-major grain-producing areas, the digital economy has a weaker
effect on local emissions but can reduce carbon emissions in adjacent areas
through cross-regional technology diffusion. From 2011 to 2018, agricultural
carbon emissions exhibit a spatial clustering pattern, while the digital economy
significantly reduces emissions in both the local and neighboring provinces. By
contrast, in 2019–2023, agricultural carbon emissions shift toward a spatially
competitive pattern: the direct local reduction effect of the digital economy
gradually diminishes and becomes statistically insignificant, while its inhibitory
effect on neighboring provinces also weakens. Fiscal support for agriculture
strengthens the carbon reduction effect of the digital economy, and a higher
level of economic development enhances this inhibitory effect. Rural entrepre -
neurship vitality and agricultural total factor productivity play significant mediat-
ing roles in the process of digital economy-driven agricultural carbon emission
reduction. The conclusions provide empirical support for strengthening the role
of the digital economy in agricultural carbon emission reduction and guidance
for improving the policy system of low-carbon agricultural development.
KEYWORDS
agricultural carbon emissions, digital economy, mediating model, spatial durbin model,
spatial effect
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 02 frontiersin.org
1 Introduction
Against the background of the coordinated advancement of the
“dual carbon” goal and the rural revitalization strategy, the manage-
ment of agricultural carbon emissions (ACE) has become crucial for
balancing ecological protection and high-quality agricultural develop-
ment (Guo and Guo, 2024). As a foundational sector of the national
economy, agriculture generates substantial carbon emissions through
fertilizer application, agricultural machinery operation, straw burn-
ing, and other production practices, which constitute a non-negligible
component of China’s carbon emission reduction system (Ke et al.,
2023). At present, the low-carbon transformation of agriculture has
emerged as a key field for achieving the “dual carbon” goal. How to
reduce carbon emission intensity while ensuring food security has
become an urgent practical challenge (Y ang et al., 2022; Wang et al.,
2020), requiring effective emission reduction pathways to be explored
from diverse perspectives. Supported by data flows and digital tech-
nologies, the digital economy (DE) is increasingly integrated into all
stages of agricultural production (Agrawal et al., 2025). Ranging from
precise monitoring in smart farmland and digital scheduling of agri-
cultural machinery operations to low-carbon management of agricul-
tural product supply chains, digital technologies provide new solutions
for improving agricultural production efficiency and resource inten-
sification, and have gradually become an important driving force for
promoting green agricultural development (Alam et al., 2023; Yu et
al., 2026).
The DE can overcome spatiotemporal barriers and facilitate cross-
regional technology flows. As it improves agricultural factor allocation
and transforms production practices, its effects on ACE are complex
and spatially interdependent. Recent studies have examined these
effects from several perspectives, including direct impacts, mediation
pathways, spatial spillovers, and contextual constraints. Despite these
contributions, several important gaps remain in existing literature.
Theoretical analyses tend to focus on individual mechanisms rather
than develop an integrated framework that encompasses direct effects,
spatial spillovers, mediation, moderation, and threshold effects.
Empirical studies likewise tend to test these effects separately, with
relatively few examining them systematically within a common frame-
work. Heterogeneity analyses generally follow conventional geograph-
ical groupings of eastern, central, and western China, while
policy-oriented analyses based on grain production functional zones
remain limited. Mediation analyses predominantly examine a single
pathway, and few identify multiple parallel indirect channels through
which the DE mobilizes rural production factors. To address these
gaps, this study uses panel data for 31 Chinese provinces from 2011 to
2023 and develops an integrated theoretical framework comprising
four dimensions: direct effects, spatial spillovers, internal transmission
mechanisms, and external constraints. Building on this framework,
spatial Durbin, mediation, moderation, and threshold models are
employed to investigate the direct and spatial spillover effects of the
DE on ACE. The analysis further tests the mediating roles of rural
entrepreneurship activity and agricultural total factor productivity,
examines the moderating role of fiscal support for agriculture, and
explores the threshold effect of economic development. Regional het-
erogeneity between major and non-major grain-producing regions, as
well as temporal heterogeneity, is also examined. The study aims to
provide theoretical and empirical support for differentiated agricul-
tural low-carbon development policies and coordinated green agricul-
tural development across regions.
2 Literature review
In recent years, research on the DE has expanded substantially,
and ACE have attracted growing scholarly attention. Therefore, this
paper systematically reviews the relevant literature in three areas.
First, existing studies on the DE have widely recognized its posi-
tive effects across key areas of economic development. At the industry
level, the DE is regarded as an important driver of transformation
throughout the agricultural value chain (Leng and Tong, 2022) and of
higher farmers’ incomes (Wang and Zhou, 2025). At the macroeco-
nomic level, it promotes urban–rural integration through both long-
term and spatial spillover effects ( Bian et al., 2025 ), supports
high-quality economic development by stimulating technological
innovation (Ding et al., 2021), and contributes to the development of
new quality productive forces ( Wang, 2024). Positive effects on
employment and common prosperity have also been demonstrated
(Lu and Zhou, 2023 ; Chen and Zhang, 2023 ; Wei et al., 2025 ).
However, these studies primarily emphasize growth and efficiency
gains. The underlying mechanisms involving resource allocation and
changes in production practices remain less fully explored. The litera-
ture also tends to focus on macroeconomic growth and industrial
upgrading, leaving the micro-level mechanisms of factor transmission
insufficiently examined. Although spatial spillovers have been docu-
mented, empirical analyses still emphasize local effects, with less atten-
tion to how cross-regional effects arise.
In terms of spatial patterns, studies employing the entropy weight
method, TOPSIS, and the Dagum Gini coefficient reveal regional dis-
parities in China’s DE development, characterized by higher levels in
eastern China than in western China and a widening absolute east–
west gap (Tong and Tang, 2025). Interregional differences account for
much of the overall disparity, while R&D investment and economic
conditions are major drivers of spatial differentiation (Luo and Zhou,
2022). Si et al. (2026) further found that regional DE development is
uneven and exhibits significant spatial correlation. Moreover, DE
development shows significant convergence nationwide and in eastern
and central China, whereas convergence in western China is not sta-
tistically significant. Although these descriptive analyses help clarify
the spatial pattern of the DE, most studies remain focused on measur-
ing regional disparities and describing their evolution, with limited
attention to the economic consequences of such spatial differentiation
for coordinated regional development. Moreover, the mechanisms
through which spatial disparities in DE development generate cross-
regional socioeconomic spillovers remain insufficiently explored.
Second, research on ACE has primarily focused on identifying its
determinants. Rural human capital exhibits a significant inverted
U-shaped relationship with ACE, with agricultural mechanization and
fiscal support for agriculture serving as mediating and moderating
mechanisms, respectively (Lu et al., 2025). The effect of agricultural
production efficiency on ACE intensity is also nonlinear, with the
direction of the effect varying across efficiency levels (Zhu and Huo,
2022). Green finance has been shown to reduce ACE intensity,
although its effects vary across quantiles (Yu et al., 2025; Zhou et al.,
2025). Coordinated advances in production and energy technologies
are expected to provide an effective pathway for reducing ACE (Song
et al., 2023). In addition, environmental regulation is an important
instrument for achieving China’s carbon peaking and carbon neutral-
ity goals and promoting the green transformation of agriculture, and
has been shown to significantly reduce agricultural carbon intensity
(Hao and Y ang, 2026). Although these studies have enriched our
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 03 frontiersin.org
understanding of the determinants of ACE, two limitations remain.
First, existing studies tend to examine individual determinants sepa-
rately and lack a systematic analytical framework that integrates eco-
nomic, social, technological, and other relevant factors. Second, many
studies implicitly assume that the effects of individual determinants
on ACE are independent and localized, thereby overlooking interac-
tions among determinants and system-wide transmission
mechanisms.
Studies on the spatial characteristics of ACE have documented
substantial regional disparities and spatial dependence across China.
Carbon emission trends differ markedly across grain production func-
tional zones, with interregional differences accounting for most of the
overall disparity (Huan et al., 2025). ACE generally exhibit a spatial
pattern characterized by higher levels in eastern than in western China
and in southern than in northern China, while regional differences in
agricultural development and mechanization are identified as key
drivers of this spatial differentiation (Huang et al., 2025; Zhang et al.,
2023). Su et al. (2023) further demonstrated significant positive inter-
provincial spatial autocorrelation and positive spillover effects in
ACE. Although these spatial analyses provide an important founda-
tion for the present study, most remain focused on describing spatial
patterns and decomposing regional disparities, without explicitly
incorporating spatial effects into the analysis of causal relationships.
Third, existing studies have provided preliminary evidence of the
carbon-reduction effect of the DE on ACE and reveal multiple under-
lying mechanisms. Some studies suggest that the DE reduces the
carbon intensity of agricultural energy use by alleviating agricultural
capital misallocation and promoting green technological innovation
(Li and Gao, 2024; Li J. et al., 2024; Li H. et al., 2024; Tong et al., 2024).
The development of the rural DE can also reduce ACE by improving
agricultural socialized services, with stronger effects in regions char-
acterized by particular levels of economic development and techno-
logical investment (Jin et al., 2024). Y ang et al. (2025) emphasize the
mediating role of labor migration, whereas Li et al. (2026) identify a
potential “efficiency rebound” effect, whereby improvements in
human capital may weaken the emission-reduction effect of the DE,
and further document a significant positive moderating role of the
regional innovation environment. In addition, fiscal support for envi-
ronmental protection and science and technology (Y ang and Wen,
2025), stronger environmental regulation ( Guo et al., 2024 ), and
greater coordination between environmental governance and fiscal
expenditure (Guo et al., 2025) have all been shown to strengthen the
agricultural carbon-reduction effect of the DE. Nevertheless, existing
studies tend to focus on direct effects or individual transmission
mechanisms, while systematic analyses integrating spatial spillover,
mediation, moderation, and threshold effects within a unified frame-
work remain limited. Mediation analyses have largely concentrated on
channels such as capital allocation and green technological innova-
tion, with relatively little attention paid to multiple parallel pathways
through which the DE activates rural production factors.
Heterogeneity analyses also continue to rely heavily on conventional
eastern, central, and western regional groupings, with limited policy-
oriented empirical evidence based on grain production functional
zones. Moreover, the linear moderating role of fiscal support for agri-
culture and the nonlinear threshold effect of economic development
have yet to be systematically distinguished and examined.
Overall, existing research on the DE and ACE provides a substan-
tial foundation, including separate examinations of direct effects,
mediation pathways, spatial spillovers, and conditional effects. Four
limitations remain. First, theoretical analyses typically focus on indi-
vidual or partial mechanisms, separating direct impacts and internal
transmission from spatial spillovers and external constraints. An inte-
grated framework encompassing direct, mediated, spatial spillover,
moderating, and threshold effects is still lacking. Second, empirical
studies generally test selected effects independently rather than sys-
tematically examining spatial, mediation, moderation, and threshold
effects within a common framework. This limits their ability to explain
how strongly, through which pathways, and under what external con-
ditions the DE affects ACE in both local and neighboring regions.
Third, heterogeneity analyses typically use conventional geographical
groupings of eastern, central, and western China. Comparatively little
attention is paid to grain production functional zones as a policy-
relevant basis for differentiation. Most studies also examine full-sam-
ple effects without considering temporal variation associated with
successive digital rural policies. Limited attention to temporal hetero-
geneity further reduces the relevance of the findings to agricultural
policy. Fourth, mediation studies predominantly examine capital allo-
cation, green technological innovation, or another single pathway. Few
identify two distinct parallel indirect channels, leaving the role of rural
factor mobilization insufficiently explored.
In response to these limitations, this study makes four incremen-
tal contributions. (1) At the theoretical level, it develops an integrated
framework comprising direct effects, spatial spillovers, internal trans-
mission mechanisms, and external constraints. The framework
explains the direct, mediated, and spatial spillover effects of the DE on
ACE while incorporating the moderating role of fiscal support for
agriculture and the threshold effect of economic development. (2) At
the empirical level, provincial panel data are used to test the direct,
spatial spillover, mediation, moderation, and threshold effects identi-
fied in the framework. Compared with studies examining only
selected effects, this design aligns theoretical mechanisms with mul-
tidimensional empirical testing and provides a more comprehensive
account of the effects of the DE on ACE. (3) For heterogeneity analy-
sis, the study moves beyond conventional geographical groupings in
two ways. It estimates separate models for grain production functional
zones to examine regional differences from a food security policy per-
spective, and it investigates temporal heterogeneity in relation to the
implementation of digital rural policies. Identifying these temporal
patterns improves the targeting and practical relevance of agricultural
low-carbon policies. (4) For transmission mechanisms, rural entrepre-
neurship activity and agricultural total factor productivity are exam-
ined as parallel mediators, identifying two indirect channels through
which the DE mobilizes rural production factors and reduces ACE.
3 Theoretical analysis
The effects of the DE on ACE constitute a complex system of inter-
related mechanisms operating at multiple levels. This study develops
a unified theoretical framework comprising four dimensions: direct
effects, spatial spillovers, internal transmission mechanisms, and two
types of external constraints.
The direct effect is the core baseline effect, describing how the DE
reduces ACE within a province. Spatial spillover effects capture the
associated impacts on neighboring provinces as the DE extends
beyond geographical boundaries. Rural entrepreneurship activity and
agricultural total factor productivity provide indirect emission
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 04 frontiersin.org
reduction channels through which the DE reorganizes rural produc-
tion factors.
However, the magnitude of these effects is not fixed; rather, it is
shaped by external conditions. First, the moderating effect of fiscal
support for agriculture represents a linear external constraint: changes
in fiscal support can continuously strengthen or weaken the overall
emission reduction effect of the DE. Second, the threshold effect of
economic development represents a nonlinear, stage-dependent con-
straint: the emission reduction potential of the DE can be fully realized
only after economic development exceeds a specific threshold.
3.1 Direct impact mechanism
As a new production factor, data can overcome constraints in tra-
ditional agricultural resource allocation. By collecting and processing
full-process production data such as soil fertility and climatic condi-
tions, data-driven systems can precisely match agricultural input
schedules to crop growth requirements. Combined with big data ana-
lytics and satellite monitoring technologies, it enables dynamic adjust-
ments to planting structures, irrigation, and fertilization rates, thereby
improving resource allocation. By reducing unnecessary agricultural
inputs and the associated carbon emissions, it helps achieve “increas-
ing yield without increasing carbon. ” At the digital technology level,
cloud computing, the Internet of Things (IoT), and intelligent equip-
ment are deeply embedded in all links of agricultural production (Guo
et al., 2026). On the one hand, blockchain technology is used to trace
agricultural material quality, helping farmers select low-carbon fertil-
izers, organic fertilizers, and other high-quality production materials.
On the other hand, high-precision sensors and intelligent connected
systems capture crop growth status in real time and calculate the opti-
mal application rates of chemicals to reduce carbon leakage.
Meanwhile, the adoption of intelligent agricultural machinery and
unmanned aerial vehicles (UAVs) enables fully automated precision
operations, which not only reduces the disturbance of soil tillage on
carbon pools but also lowers energy consumption in machinery oper-
ations, promoting the transformation of agricultural production from
extensive to low-carbon modes. Implementing these technologies also
requires improvements in producers’ capabilities. Furthermore, digital
networks overcome information barriers by integrating agricultural
databases and low-carbon technical tutorials on Internet platforms
with “online + offline” training models. This not only popularizes low-
carbon concepts and practical skills among producers but also
enhances their environmental awareness and digital tool proficiency,
enabling farmers to adopt low-carbon production methods.
Based on the above analysis, Hypothesis H1 is proposed: The DE
can effectively reduce ACE.
3.2 Spatial spillover effect
Neighboring regions possess natural adaptability for the transmis-
sion of carbon mitigation effects due to their close geographical prox-
imity and similar agricultural production models. Digital technologies
effectively break spatial and temporal constraints, significantly lower-
ing the threshold for the dissemination of low-carbon knowledge and
technologies, thereby enabling the rapid penetration of digital emis-
sion-reduction experiences into surrounding areas (Li J. et al., 2024;
Li H. et al., 2024). Through the construction of digital platforms, the
real-time sharing of agricultural policies, low-carbon technologies,
and other information can be realized, assisting neighboring regions
in drawing on the mature experience of core areas to reduce trial-and-
error costs. Meanwhile, digitalization promotes cross-regional scien-
tific research cooperation and the sharing of green technologies,
supporting the rapid adoption of new energy agricultural equipment
and low-carbon planting techniques in surrounding regions. Relying
on big data technology, cross-regional integration of idle agricultural
machinery, high-quality agricultural supplies, and other resources can
be achieved. With efficient cross-regional allocation mechanisms,
neighboring regions are able to obtain various resources required for
low-carbon production at a lower cost ( Liang and Qiao, 2025 ).
Furthermore, guided and supported by the regional coordinated
development strategy, various regions have smoothly promoted cross-
regional digital agriculture cooperation projects and established low-
carbon technology alliances, strengthening inter-regional resource
sharing and further amplifying the cross-regional radiation of carbon
mitigation effects. Ultimately, while reducing local ACE, the DE also
generates a significant inhibitory effect on ACE in surrounding areas
(Wang et al., 2025).
Based on the above analysis, Hypothesis H2 is proposed: The
inhibitory effect of the DE on ACE presents significant spatial spillover
effects.
3.3 Moderating effect of fiscal support for
agriculture
Government intervention theory suggests that the application
of the DE in agriculture requires complementary public funding.
Fiscal support for agriculture can address gaps in rural digital
infrastructure, subsidize purchases of smart agricultural machin -
ery, and support research and development in agricultural digital
technologies, thereby shaping the external institutional conditions
for agricultural digitalization. Specifically, increased fiscal support
can finance new rural digital infrastructure, address weaknesses in
existing infrastructure, and create a foundation for integrating
digital technologies into agricultural production. Subsidies for
smart agricultural machinery can also reduce equipment costs for
farmers and agricultural operators, accelerating the adoption of
unmanned aerial vehicles and intelligent fertilization equipment.
Dedicated funding can encourage research institutions and enter -
prises to develop agricultural digital technologies and overcome
technical bottlenecks. It can also ensure the stable operation of
agricultural meteorological monitoring systems and the timely
release of information, providing data for digital decision-making.
Together, these measures improve infrastructure, equipment, tech -
nology, and information provision, facilitate the application of the
DE in agriculture and rural areas, enhance the digitalization and
automation of agricultural production, and indirectly strengthen
its effect on ACE reduction.
Based on the above analysis, Hypothesis H3 is proposed: The level
of fiscal support for agriculture exerts a moderating effect in the pro-
cess whereby the DE reduces ACE.
3.4 Threshold effect of economic
development level
Development stage theory suggests that the effectiveness of
digital technologies depends on a region’s material resources,
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 05 frontiersin.org
farmers’ purchasing power, and supporting industries. Economic
development determines local fiscal capacity, rural household
incomes, and the availability of complementary inputs for agricul -
tural digitalization. At the low stage of economic development,
regional factor supporting capacity and technological adaptation
conditions are insufficient. Local governments lack the fiscal
capacity to build rural digital infrastructure, such as fiber-optic
networks and Internet of Things base stations, while farmers’ lim -
ited incomes constrain purchases of smart machinery. Agriculture
also tends to be characterized by small-scale, extensive operations,
making digital emission reduction technologies costly and diffi -
cult to adopt. These constraints limit the carbon mitigation effect
of the DE. As economic development advances, stronger local
fiscal capacity supports infrastructure improvements and equip -
ment subsidies, while farmers’ purchasing power increases.
Agricultural production also shifts toward larger-scale operations,
and new agricultural business entities are better positioned to
adopt precision planting and smart management technologies.
Meanwhile, greater emphasis on high-quality, sustainable devel -
opment, demand for green agricultural products, and environ -
mental policies stimulate the adoption of digital low-carbon
technologies. Collectively, these conditions strengthen the effect
of the DE on ACE reduction, creating a clear contrast with less-
developed regions.
Based on the above analysis, Hypothesis H4 is proposed: The
effect of the DE on ACE reduction exhibits a threshold associated with
economic development.
3.5 Indirect influence mechanism
Factor reconfiguration theory suggests that, by introducing data
as a new production factor, the DE can overcome traditional barri -
ers to rural factor mobility and reshape the allocation of land, labor,
and capital. On the one hand, digital platforms reduce information
search costs and barriers to entrepreneurship in rural markets. They
stimulate entrepreneurial activity and encourage new business
models, such as green crop and livestock production and agricul -
tural waste recycling, thereby promoting a shift toward low-emis -
sion production. As rural entrepreneurship activity increases,
low-carbon production practices become more widespread, indi -
rectly reducing total ACE ( Xu et al., 2024 ). On the other hand,
endogenous growth theory identifies technological progress and
improved factor allocation as key drivers of production efficiency.
Digital technologies improve agricultural factor allocation through
intelligent monitoring, cross-regional machinery scheduling, and
precise matching of agricultural inputs to production needs. As
agricultural total factor productivity rises, production gradually
moves away from an extensive, high-input, high-emission model
toward an intensive model characterized by greater efficiency and
lower resource consumption. This structural shift indirectly con -
tributes to carbon emission reductions.
Based on the above analysis, Hypothesis H5 is proposed: The
DE can effectively reduce ACE by enhancing rural entrepreneurial
vitality and improving agricultural total factor productiv -
ity (ATFP).
The direct effect, spatial spillovers, moderating effect, threshold
effect, and two indirect transmission pathways together form an inte-
grated framework explaining how the DE affects ACE. The mecha-
nisms are illustrated in Figure 1.
4 Research design
4.1 Variable selection
4.1.1 Explained variable
ACE: Following Li et al. (2011), this paper mainly focuses on the
direct and indirect carbon emissions generated by six carbon sources
in crop planting, including chemical fertilizers, pesticides, agricultural
films, diesel oil, irrigation, and tillage. The carbon emissions of crop
production are calculated based on the corresponding carbon emis-
sion coefficients of each carbon source. The specific formula is shown
in Equation 1.
φ= ∑ = ∑×i iiACE ACE T (1)
In Equation 1, ACE denotes the total carbon emissions of crop
production, iT represents the input quantity of the i-th carbon source,
and φi refers to the carbon emission coefficient of the i-th carbon
source. The main carbon emission sources and corresponding
carbon emission coefficients for crop production are presented in
Table 1.
4.1.2 Core explanatory variable
Following Shao et al. (2024) , Hou et al. (2023) , and Wang et al.
(2026), this study constructs an evaluation indicator system, con -
sidering data availability and timeliness, from three dimensions:
digital infrastructure, digital industrialization, and industrial digi -
talization. The entropy weight method is adopted to measure DE
development.
The calculation proceeds as follows: First, the original data are
standardized. Because the indicators differ in units and ranges, min–
max normalization is applied to m years, k provinces, and n indicators
to eliminate the influence of these differences on the composite assess-
ment. Positive and negative indicators are standardized using
Equations 2, 3, respectively.
( )
−= = =− min 1,2, , 1,2, ,max minijt jijt ij
j
XXB i nt mXX
W hen X isa positive indicator (2)
( )
−= = =− max 1,2, , 1,2, ,max minj ijtijt ij
j
XXB i nt mXX
W hen X isanegative indicator (3)
where ijtB denotes the standardized data; maxjX denotes the
maximum value in year t across k provinces for indicator j ; and
minjX denotes the minimum value in year t across k provinces for
indicator j. Second, the standardized indicator values are normalized
and their entropy values are calculated. The entropy of each indicator
( = 1, 2,jn ) is given by Equation 4:
( ) ( )= =
= =
=−= × ∑∑ ∑∑11
11
1 ln , whereln mk ijtj ijt ijt ijt mkti ijt
ti
BE pp pmk
B
(4)
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 06 frontiersin.org
Third, the entropy values are used to calculate the weights of the
indicators for the DE index, as shown in Equation 5:
( )
−=
−∑
1
1
1
j
j n
j
j
Ew
E
(5)
Finally, the DE index for each province is calculated, as shown in
Equation 6:
=∑
1
n
it j ijt
j
DE w p
(6)
Under the entropy weighting method, a higher DE index indicates
a higher level of DE development.
The specific indicator system is presented in Table 2.
4.1.3 Control variables
To ensure the accuracy of regression results and account for other
potential factors affecting ACE, this paper selects rural human capital
level, agricultural industrial structure, environmental regulation,
agricultural technological level, and agricultural mechanization as
control variables for empirical analysis. Specifically, the rural human
capital level is measured by the average years of schooling in rural
areas (Golovina et al., 2022). The agricultural industrial structure is
represented by the proportion of the added value of the primary
TABLE 1 Main carbon emission sources and carbon emission coefficients for crop production.
Carbon source Carbon emission coefficient Reference
Diesel oil 0.59 kg/kg IPCC (2013)
Chemical fertilizer 0.89 kg/kg Oak Ridge National Laboratory
Pesticide 4.93 kg/kg Oak Ridge National Laboratory
Agricultural film 5.18 kg/kg
Institute of Agricultural Resources and Ecological Environment,
Nanjing Agricultural University
Irrigation 266.48 kg/hm2 Duan et al. (2011)
Tillage 312.60 kg/km2 Li et al. (2011)
FIGURE 1
Influence mechanism of the DE on ACE.
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 07 frontiersin.org
TABLE 2 Indicator system of the DE.
First-level indicator Second-level indicator Third-level indicator Definition Direction
DE Digital Infrastructure Internet broadband access rate Number of Internet broadband access ports/Regional permanent
population
+
Internet broadband penetration rate Number of Internet broadband subscribers/Regional permanent
population
+
Mobile telecommunication facility scale Capacity of mobile telephone exchanges +
Long-distance optical cable line length Long-distance optical cable line length +
Number of web pages Raw statistical data +
Number of domain names Raw statistical data +
Digital Industrialization Per capita total telecommunication business volume Total telecommunication business volume/Regional permanent population +
Mobile phone penetration rate Raw statistical data +
Number of corporate entities in information transmission,
software and information technology services
Raw statistical data +
Employment proportion of information and software
industry
Employees in information transmission, software and information
technology services/Total urban employment
+
Number of granted domestic patent applications Raw statistical data +
Number of accepted domestic patent applications Raw statistical data +
Industrial Digitalization Peking University Digital Inclusive Finance Index Raw statistical data +
Proportion of enterprises with e-commerce transactions Raw statistical data +
E-commerce sales volume Raw statistical data +
Number of websites per hundred enterprises Raw statistical data +
Added value of secondary and tertiary industries Added value of secondary industry + Added value of tertiary industry +
Investment in scientific and technological innovation R&D expenditure of industrial enterprises above designated size +
Express delivery volume Raw statistical data +
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 08 frontiersin.org
industry in GDP . Environmental regulation is measured by the ratio
of fiscal environmental protection expenditure to total fiscal expen-
diture. The agricultural technological level is characterized by the
number of agricultural science and technology patents. Agricultural
mechanization is measured by the total power of agricultural
machinery.
4.1.4 Moderating variable
Referring to the research of Tian and Liao (2024) , this paper
adopts the ratio of government fiscal expenditure on agriculture, for-
estry and water affairs to the total output value of agriculture, forestry,
animal husbandry and fishery to measure the level of fiscal support
for agriculture.
4.1.5 Threshold variable
Following Xu and Y ang (2025), this paper selects per capita GDP
to measure the level of economic development.
4.2 Data sources
The above indicators are mainly collected from the China Energy
Statistical Y earbook, China Statistical Y earbook, China Rural Statistical
Y earbook, China Science and Technology Statistical Y earbook, pro-
vincial statistical yearbooks, and provincial annual statistical data
released by the National Bureau of Statistics. The average growth rate
method is adopted to interpolate missing provincial data. The defini-
tions and descriptive statistics of all variables are summarized in
Table 3.
4.3 Spatial correlation analysis of DE and
ACE
4.3.1 Global Moran’s I index of DE and ACE
Global Moran’s I index is calculated separately for the DE and
ACE to examine their spatial autocorrelation. In many spatial econo-
metric studies of agriculture and the DE, geographical proximity
between provinces is treated as a basic condition for technology spill-
overs and factor mobility. The Euclidean distance weight matrix is
constructed from geographical distances between provincial cen -
troids. Because it does not rely on economic or agricultural indicators,
the matrix is exogenously specified, thereby minimizing potential
endogenous relationships between the spatial weights and the depen-
dent or core explanatory variables. This study accordingly adopts the
Euclidean distance weight matrix. As shown in Table 4, the global
Moran’s I statistics for both the DE and ACE are significant at the 1%
level, with values ranging from 0.3 to 0.5. These results indicate strong
spatial autocorrelation in both variables and support the use of a spa-
tial econometric model to examine the impact of the DE on ACE.
4.3.2 Local Moran’s I index of DE and ACE
To further analyze the spatial agglomeration characteristics of the
DE and ACE, Figure 2 plots the local Moran’s I scatter plots of the DE
and ACE for 2011 and 2023. It can be seen from Figure 2 that the
Moran’s I values of the DE and ACE are concentrated in the first and
third quadrants, indicating that the DE and ACE have maintained a
long-term “high-high” and “low-low” distribution trend.
The provinces consistently in the first quadrant in terms of the
DE are Shanghai, Beijing, Zhejiang, Jiangsu, Guangdong, Fujian
and other provinces and cities. These provinces are all located in
the eastern region, with a solid economic foundation and obvious
TABLE 3 Variable definition and descriptive statistics.
Category Variable Code Unit Observations Mean Std. dev. Min Max
Explained
Variable
ACE ace
10 thousand
tons
403 270.8991 192.7048 6.6130 867.5698
Key independent
variables
DE de - 403 0.1345 0.1156 0.0072 0.7466
Controlled
variables
Rural human
capital level
edu Ye a r 403 7.7092 0.8276 3.8038 10.0653
Agricultural
industrial structure
pi % 403 9.6995 5.2292 0.2035 25.8381
Environmental
regulation
er % 403 2.8577 0.9514 1.0301 6.8141
Agricultural
technological level
tec Piece 403 3030.189 3148.383 7 16,651
Agricultural
mechanization
level
fmp
10 thousand
kilowatts
403 3376.849 2937.055 93.97 13353.02
Moderating
Variable
Fiscal support for
agriculture
fd - 403 0.3002 0.3853 0.0705 2.0911
Threshold
Variable
Economic
development level
pgdp Yuan 403 61996.39 34820.22 10146.45
201929.8
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 09 frontiersin.org
advantages in the development of the DE, thus forming a “high-
high” agglomeration characteristic. The provinces that have
always been in the first quadrant in terms of ACE are Yunnan,
Guangxi, Xinjiang, Guangdong, Sichuan, Heilongjiang, Anhui,
Hubei, Jiangsu, Henan, Shandong and other provinces. Most of
these provinces are major agricultural provinces, with large agri -
cultural production and generally high levels of ACE, forming a
“high-high” agglomeration pattern.
4.4 Model specification
To eliminate heteroscedasticity, several variables are processed by
logarithmic transformation (Rosopa et al., 2013 ). Through spatial
econometric model testing methods such as Lagrange Multiplier (LM)
test, Likelihood Ratio (LR) test, the test results presented in Tables 5–7
are obtained. Finally, the two-way fixed effects spatial Durbin model
(SDM) is comprehensively selected for empirical analysis (Equation 7).
αρ β β φ
φ µηε
=
= + ++ + +
+++
∑
∑
6
11
2
6
2
ln lnit it it s sit it
s
s sit i t it
s
ace W ace de Z Wde
WZ
(7)
ln itace represents the ACE of province i in year t ; ρ is the spatial
autoregressive coefficient, which denotes the spatial spillover effect of
ACE in one province on ACE in its geographically adjacent provinces.
itde denotes the DE of province i in year t . itZ is the vector of control
variables, including rural human capital level (lnedu), agricultural
industrial structure (pi), environmental regulation (er), agricultural
technological level (lntec), and agricultural mechanization level
(lnfmp). W is the Euclidean distance weight matrix selected in this
paper. µi and ηt represent the spatial fixed effect and time fixed effect,
respectively, and εit is the random disturbance term.
5 Spatial effect analysis of the impact
of the DE on ACE
5.1 Benchmark regression analysis
To further verify the robustness of the results, Table 8 reports the
regression results of the two-way fixed-effects Spatial Autoregressive
model (SAR), Spatial Error Model (SEM), and spatial Durbin model
(SDM) models using the Euclidean distance weight matrix. The results
of the three models are generally consistent. Focusing on the SDM, the
spatial autoregressive coefficient ρ is positive and significant at the
10% significance level, indicating that ACE generate a positive spatial
spillover effect on neighboring provinces. On the one hand, neighbor-
ing provinces often share similar natural conditions and market
demands. High-carbon agricultural production models adopted in
one province (e.g., high fertilizer input, intensive farming) are easily
imitated by neighboring provinces to reduce trial-and-error costs and
increase output, directly driving a linked rise in carbon emissions. On
the other hand, factors required for agricultural production, such as
fertilizers and agricultural machinery, flow frequently across neigh-
boring regions. Increases in high-carbon inputs in one province are
transmitted to neighboring provinces through factor supply chains,
indirectly raising their carbon emission intensity. The core explana-
tory variable, the DE, has a significant inhibitory effect on ACE in
both the local province and its neighboring provinces, confirming
research hypotheses H1 and H2.
5.2 Spatial effect decomposition
According to the analysis by Dolores (2011), the direct and indi-
rect effects of the DE on ACE can only be obtained by decomposing
the SDM using the partial derivative method. The results are presented
in Table 9.
From the direct effects, the coefficient of the core explanatory
variable, the DE, is −0.319 and significant at the 1% level. This indi -
cates that for each one-unit increase in the DE, ACE in the local
province decrease by 31.9%, indicating that the DE exerts a signifi -
cant inhibitory effect on local ACE. Among the control variables,
the coefficients of rural human capital level, agricultural industrial
structure, and agricultural technological level are significantly neg -
ative. This suggests that improvements in rural education and agri -
cultural technology effectively reduce ACE. Additionally, an
increase in the output value of the primary industry promotes agri -
cultural carbon reduction, possibly because rising output is often
accompanied by the modernization of agricultural production
methods, optimized resource utilization efficiency, and technologi -
cal upgrading, ultimately lowering ACE. The coefficient of environ -
mental regulation is negative but not significant, indicating that
environmental regulation contributes to agricultural carbon reduc -
tion to a certain extent, though the effect is not statistically signifi -
cant. Further efforts are needed to enhance the effectiveness of
environmental regulation in reducing ACE in the future. The coef -
ficient of agricultural mechanization level is significantly positive,
meaning that higher agricultural mechanization is associated with
greater ACE. One possible explanation is that mechanization pro -
motes larger-scale, more intensive agricultural production. By low -
ering labor costs, large machinery can encourage farmers to expand
cultivated areas and increase fertilizer, pesticide, and other inputs.
TABLE 4 Global Moran’s I index of DE and ACE.
Year DE ACE
Moran’s I p-
value
Moran’s I P-value
2023 0.375 0.000 0.406 0.000
2022 0.379 0.000 0.406 0.000
2021 0.382 0.000 0.401 0.000
2020 0.383 0.000 0.388 0.000
2019 0.369 0.000 0.378 0.000
2018 0.368 0.000 0.375 0.000
2017 0.401 0.000 0.371 0.000
2016 0.413 0.000 0.367 0.000
2015 0.408 0.000 0.357 0.000
2014 0.410 0.000 0.355 0.000
2013 0.417 0.000 0.343 0.000
2012 0.453 0.000 0.337 0.000
2011 0.446 0.000 0.329 0.000
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 10 frontiersin.org
Excessive input expansion raises emissions from the production
and application of agricultural inputs, potentially outweighing the
emission reductions achieved through machinery operations and
producing a positive net effect on ACE. This finding also suggests
that the carbon mitigation potential of agricultural mechanization
in China has not yet been fully realized. Mechanization does not
automatically imply low-carbon production; complementary
changes toward greener production practices are essential.
For the indirect effects, the coefficient of the DE is −1.859 and
significant at the 1% level. This indicates that for each one-unit
increase in the DE, ACE in neighboring provinces decrease by
185.9%, showing that the DE has a significant inhibitory effect on
ACE in neighboring provinces. The indirect effect is substantially
larger than the direct effect. This may reflect the broad applicability
of digital technologies and their capacity to generate cross-regional
knowledge spillovers. DE development in one province can affect
several neighboring provinces through demonstration effects, pro -
duction network linkages, and competitive pressures. Close
regional economic ties and frequent factor flows during the sample
period can further reinforce these mechanisms. Because the indi -
rect effect summarizes spillovers across other provinces, their com -
bined average magnitude can exceed the local direct effect. Among
the control variables, agricultural industrial structure has a signifi -
cantly negative impact on ACE in neighboring provinces. An
increase in the primary industry output value of one province often
reflects improvements in agricultural technology and the
FIGURE 2
Local Moran’s I scatterplots for the DE in 2011 and 2023, and for ACE in 2011 and 2023.
TABLE 5 LM test results.
LM test Value Degrees of
Freedom
p value
Moran’s I 9.337 1 0.000
LM-error 80.995 1 0.000
Robust LM-
error
67.503 1 0.000
LM-lag 14.081 1 0.000
Robust LM-lag 0.590 1 0.443
TABLE 6 LR test results.
LR test LR statistic
results
P value
H0: SDM reduces to
SAR
89.24 0.0000
H0: SDM reduces to
SEM
107.24 0.0000
TABLE 7 LR test results.
LR test LR statistic
results
P value
H0: The spatial fixed
effect is superior to the
two-way fixed effect
18.72 0.0439
H0: The time fixed
effect is superior to the
two-way fixed effect
1187.92 0.0000
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 11 frontiersin.org
formation of large-scale production models. These mature experi -
ences in agricultural modernization spread to neighboring prov -
inces, driving them to optimize their agricultural production
methods and reduce carbon emissions. Environmental regulation,
agricultural technological level, and agricultural mechanization all
have positive effects on ACE in neighboring provinces. Strict envi -
ronmental regulations in one province may drive high-energy-
consuming, high-emission agricultural activities to relocate to
neighboring provinces, raising their carbon emission levels.
Meanwhile, improvements in one province’s agricultural technol -
ogy may attract talent and resources from neighboring provinces,
weakening their technological capacity and indirectly increasing
their ACE. Furthermore, the expansion of agricultural mechaniza -
tion in one province may generate demonstration effects in neigh -
boring provinces. When neighboring provinces adopt mechanized
equipment, they may increase the use of fertilizers and pesticides,
leading to higher carbon emissions.
5.3 Robustness tests
This paper conducts robustness checks using the two-way fixed-
effects model, the carbon emission weight matrix, and the spherical
distance matrix, respectively. The results are presented in Table 10.
Given the strong spatial agglomeration characteristics of ACE, this
paper constructs a carbon emission distance weight matrix, the spe-
cific method of which is shown in Equation 8.
=
−
1
a
ij
W
ace ace
(8)
where andijace ace denote the average ACE of provinces i and j,
respectively, over the 2011–2023 sample period. aW denotes the ACE
spatial weight matrix.
The results obtained using the two-way fixed-effects model, the
agricultural carbon emission spatial weight matrix, and the spherical
distance matrix are largely consistent with the spatial effect decompo-
sition results. Specifically, both the direct and indirect effects of the DE
are significantly negative, further verifying the robustness of the
regression results.
5.4 Endogeneity test
To address potential endogeneity arising from reverse causality,
the one-period lag of the DE is used as an instrument (Liu and Sheng,
2026). The model is estimated using two-stage least squares, and the
results are reported in Table 11. The one-period lag of the digital
economy index results in the exclusion of 31 observations from 2011,
reducing the sample size for the instrumental variable regression to
372. The first-stage F-statistic is 2692.14, well above 10, indicating a
strong association between the instrument and the DE. In the second
stage, the coefficient of the DE is negative and significant at the 5%
level, indicating that the DE significantly reduces ACE. This finding is
consistent with the baseline results and supports their robustness after
accounting for potential endogeneity.
TABLE 8 Model estimation results.
Variables SAR SEM SDM
de
−0.200**
(0.095)
−0.187*
(0.102)
−0.282***
(0.096)
lnedu
−0.758***
(0.142)
−0.744***
(0.147)
−0.578***
(0.134)
pi
−0.367
(0.435)
−0.141
(0.440)
−1.395***
(0.438)
er
−1.644**
(0.656)
−1.760**
(0.683)
−0.925
(0.624)
lntec
−0.001
(0.012)
0.002
(0.013)
−0.029**
(0.012)
lnfmp
0.226***
(0.030)
0.257***
(0.034)
0.181***
(0.031)
W × de - -
−1.583***
(0.274)
W × lnedu - -
0.002
(0.399)
W × pi - -
−3.760***
(1.284)
W × er - -
3.134***
(1.207)
W × lntec - -
0.059*
(0.033)
W × lnfmp - -
0.185***
(0.067)
ρ 0.373*** - 0.151*
R-squared 0.4450 0.3365 0.6696
Log-L 499.7369 492.0057 536.7577
Obs 403 403 403
The values in parentheses are standard errors. , , and *** denote significance at the 10%,
5%, and 1% levels, respectively.
TABLE 9 Decomposition of effects.
Variables Direct effect Indirect effect
de
−0.319
(0.099)
−1.859***
(0.368)
lnedu
−0.587***
(0.129)
−0.113
(0.431)
pi
−1.441***
(0.439)
−4.421***
(1.557)
er
−0.866
(0.601)
3.518**
(1.411)
lntec
−0.028**
(0.011)
0.064*
(0.039)
lnfmp
0.188***
(0.030)
0.246***
(0.069)
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 12 frontiersin.org
5.5 Heterogeneity analysis by
grain-producing regions
Since ACE levels are affected by grain production functional
zones, this paper divides China’s 31 provincial-level administrative
regions into major grain-producing regions and non-major grain-
producing regions for heterogeneity analysis. The results are presented
in Table 12.
In major grain-producing regions, the spatial autoregressive coef-
ficient ρ is significantly negative. This indicates that an increase in
ACE in one province exerts an inhibitory effect on ACE in neighbor-
ing provinces. A possible explanation is that due to limited resources
in major grain-producing regions, the expansion of production in
high-emission provinces intensifies resource competition and captures
market share, leading to a decline in agricultural output and, conse-
quently, reduced carbon emissions in neighboring provinces. The
direct effect coefficient of the DE is significantly negative, while the
indirect effect coefficient is positive but not significant. This shows that
the DE inhibits local ACE, but its promoting effect on neighboring
provinces is not statistically significant. Major grain-producing
regions bear the primary responsibility for grain production, with
agriculture accounting for a large share of their economies. In these
regions, the DE can exert its full effect by optimizing agricultural pro-
duction processes through digital technologies and reducing resource
waste, thereby significantly suppressing local ACE. Agricultural digi-
talization in major producing regions is primarily focused on ensuring
local food security, with limited cross-regional resource flows. As a
result, digital technologies struggle to drive coordinated regional
development, leading to an insignificant impact of the DE on ACE in
neighboring provinces.
In non-major grain-producing regions, the spatial autoregressive
coefficient ρ is positive but not significant, indicating that the positive
impact of ACE in one province on those in neighboring provinces is
TABLE 10 Robustness tests.
Variables Two-way fixed
effects model
SDM (Carbon Emission Distance
Matrix)
SDM (Spherical Distance Matrix)
Direct effect Indirect effect Direct effect Indirect effect
de
−0.259**
(0.104)
−0.346***
(0.098)
−0.261*
(0.141)
−0.368***
(0.103)
−3.431***
(0.809)
lnedu
−0.766***
(0.157)
−0.492***
(0.131)
−0.397
(0.294)
−0.674***
(0.132)
−0.095
(0.799)
pi
−0.215
(0.480)
−0.431
(0.393)
−6.273***
(1.242)
−1.729***
(0.473)
−9.212***
(3.410)
er
−1.487**
(0.724)
−0.699
(0.594)
4.231***
(0.954)
−0.664
(0.613)
5.099*
(2.785)
lntec
0.009
(0.013)
0.001
(0.011)
0.060**
(0.029)
−0.029**
(0.012)
0.040
(0.073)
lnfmp
0.290***
(0.031)
0.261***
(0.028)
0.129**
(0.055)
0.212***
(0.030)
0.570***
(0.151)
R-squared 0.6592 0.1671 0.5833
Log-L - 533.8428 531.2621
Obs 403 403 403
TABLE 11 Instrumental variable regression results.
Variables First stage: de Second stage:
lnace
L.de 1.0240***
(0.0127)
de −0.6595**
(0.2686)
Control variables Ye s Ye s
F-statistic 2692.14 -
R2 0.9779 0.8955
N 372 372
TABLE 12 Heterogeneity analysis by grain-producing regions.
Variables Grain-producing
areas
Non-grain-
producing areas
Direct
effect
Indirect
effect
Direct
effect
Indirect
effect
de
−1.008***
(0.184)
0.327
(0.252)
−0.249*
(0.131)
−1.024***
(0.341)
Control
Variables
Control Control
ρ −0.543*** 0.031
R-squared 0.5379 0.6400
Log-L 310.3362 287.3751
Obs 169 234
Major grain-producing regions (13 provinces): Heilongjiang, Jilin, Liaoning, Inner Mongolia,
Hebei, Henan, Shandong, Jiangsu, Anhui, Jiangxi, Hubei, Hunan, Sichuan. Non-major grain-
producing regions (18 provinces): Beijing, Tianjin, Shanghai, Zhejiang, Fujian, Guangdong,
Hainan, Chongqing, Guizhou, Yunnan, Xizang, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang,
Guangxi, Shanxi.
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 13 frontiersin.org
statistically insignificant. This may be because agriculture accounts for
a relatively low share of the economy in non-major grain-producing
regions, where industries are diversified. Unlike major producing
regions, these areas lack the strong spatial linkages driven by grain
production, weakening the spatial correlation of ACE across prov-
inces. The direct effect coefficient of the DE is significantly negative,
but its absolute value is small, suggesting that the DE inhibits local
ACE. In non-major grain-producing regions, the role of the DE in
empowering agricultural production is limited due to the lower share
of agriculture, resulting in a moderate carbon reduction effect on local
agricultural emissions. The indirect effect coefficient of the DE is sig-
nificantly negative. This indicates that non-producing regions have
closer economic ties, allowing digital technologies to spread easily
across regions, driving improvements in agricultural production effi-
ciency and effectively suppressing ACE in neighboring provinces.
5.6 Temporal heterogeneity analysis
The Outline of the Digital Rural Development Strategy, issued in
2019, marked a shift from localized digital agriculture pilots to coor-
dinated nationwide implementation. Before 2019, digital transforma-
tion was piloted in selected regions, with limited policy coverage.
These changes in the policy environment may have reshaped the
mechanisms through which the DE supports low-carbon agriculture.
We therefore use the transition between 2018 and 2019 to examine
temporal heterogeneity in the effects of the DE on ACE.
The period-specific results in Table 13 reveal pronounced differ-
ences in spatial dependence in ACE, the local emission reduction
effect of the DE, and the strength of spatial spillovers. For 2011–2018,
the spatial autoregressive coefficient ρ is 0.406 and significant at the
1% level, indicating positive spatial clustering in ACE and a tendency
toward similar emission levels across neighboring regions. The direct
effect of the DE is −0.824 and significant at the 1% level, indicating a
substantial reduction in local ACE. Early agricultural digitalization
initiatives, together with policies promoting greener and lower-input
production, may have enabled pioneering regions to optimize input
structures and production practices through precision farming and
intelligent management. The indirect effect is −3.239 and also signifi-
cant at the 1% level, indicating that DE development reduces ACE in
neighboring provinces. Early green digital production practices may
have generated strong demonstration effects. Their relatively low
adoption barriers and high replicability could have supported low-
carbon transformation in neighboring provinces, producing substan-
tial cross-regional emission reduction spillovers.
During 2019–2023, nationwide implementation of digital rural
policies accelerated agricultural digitalization across provinces, and
the estimated effects of the DE on ACE changed. First, the spatial
autoregressive coefficient ρ is −0.279 and significant at the 10% level,
with ρ changing from significantly positive to significantly negative.
This suggests a shift from spatial clustering toward spatial competi-
tion. Widespread digitalization may have intensified interregional
flows of digital technologies, green agricultural inputs, and labor.
Digitally advanced regions could increasingly attract resources for
green agriculture, drawing them away from neighboring regions and
generating an offsetting spatial pattern of ACE. Second, the direct
effect of the DE is −0.003 and is not statistically significant, suggesting
that local emission reduction benefits have weakened. As agricultural
digitalization becomes more widespread, the marginal benefits of digi-
tal technologies for production optimization and carbon mitigation
may decline, leaving less scope for additional production-side reduc-
tions. As applications mature and mitigation channels become better
established, local emission reduction capacity may approach a steady
state, rendering the direct effect statistically insignificant. Third, the
indirect effect is −1.679 and significant at the 1% level, but its absolute
magnitude is smaller than in the earlier period. The DE therefore con-
tinues to reduce neighboring provinces’ ACE, although spillovers have
weakened. Compared with the differentiated and potentially transfor-
mative technology spillovers of the earlier stage, nationwide digitaliza-
tion may have narrowed regional gaps, reduced diffusion barriers, and
made regional digital capabilities more similar. The marginal benefits
of cross-regional diffusion consequently diminish: the spillover effect
remains statistically significant, but its magnitude declines.
It should be particularly noted that the relatively low goodness-of-
fit (R2) values in the period-specific heterogeneity regression results
are attributable to the following reasons: Dividing the panel into two
subperiods reduces the number of observations in each regression. In
addition, the 2019–2023 period includes several external shocks, such
as the COVID-19 pandemic, changes in agricultural policy, and
extreme weather events. These factors may have contributed to short-
term fluctuations in agricultural carbon emissions beyond those cap-
tured by the explanatory and control variables. These considerations
offer possible explanations for the lower R2, which falls to 0.0289 in
2019–2023.
5.7 Moderating effect
The impact of the DE on ACE may be moderated by fiscal support
for agriculture. Therefore, this paper incorporates the level of fiscal
support for agriculture to explore its moderating effect on the relation-
ship between the DE and ACE. The corresponding formula is shown
in Equation 9, and the regression results are presented in Table 14.
( )
( )
αρ β β
βφ φ
φ µηε
=
= + + + ×+
-
- ×+
+++
∑
∑
12
7
12
3
7
3
ln lnit it it it it
s sit it it it
s
s sit i t it
s
ace W ace de de fd
Z Wde W de fd
WZ
- ×+
(9)
TABLE 13 Temporal heterogeneity regression results.
Variables 2011–2018 2019–2023
Direct
effect
Indirect
effect
Direct
effect
Indirect
effect
de
−0.824***
(0.158)
−3.239***
(0.804)
−0.003
(0.216)
−1.679***
(0.518)
Control
variables
Control Control
ρ 0.406*** −0.279*
R-squared 0.1463 0.0289
Log-L 426.2155 299.3425
Obs 248 155
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 14 frontiersin.org
where ×it itde fd denotes the interaction term between the DE and
the level of fiscal support for agriculture, which has been centered. The
meanings of all other variables are consistent with those in Equation 7.
As can be seen from Table 14, the coefficients of both the direct
and indirect effects of the interaction term between the DE and fiscal
support for agriculture share the same sign as the coefficients of the
DE itself, both being negative. This indicates that fiscal support for
agriculture acts as an “amplifier” in the process through which the DE
inhibits ACE. An increase in its intensity not only strengthens the
direct emission reduction effect of the DE on local ACE but also
enhances the indirect inhibitory effect on neighboring provinces
through spatial spillover effects. Thus, Hypothesis H3 is supported.
The increase in fiscal support for agriculture provides favorable finan-
cial backing for the development of the DE, promotes inter-regional
technology exchange and resource flow, creates a favorable environ-
ment for coordinated development, and effectively amplifies the agri-
cultural carbon mitigation benefits of the DE.
5.8 Threshold effect
To verify whether the impact of the DE on ACE exhibits nonlinear
characteristics, this paper employs a threshold effect model for empiri-
cal testing (Fang et al., 2022). Economic development levels influence
various aspects such as technological progress, industrial structure,
and environmental governance, leading to “phased differences” in the
agricultural carbon reduction effect of the DE. Therefore, this paper
selects the level of economic development as the threshold variable to
explore how the effect of the DE on ACE changes across different
stages of economic development. The single-threshold effect model is
constructed as follows (Equation 10):
( ) ( )αβ θ β θ
βε
= + ⋅ ≤+ ⋅ >+
+∑
12
7
3
ln ln lnit it it
s sit it
s
ace de I pgdp de I pgdp
Z
(10)
Where 1(•) denotes the indicator function, which takes the value
of 0 if the expression in parentheses is false and 1 otherwise. Based on
whether the threshold variable ln pgdp is greater than θ, the research
sample is divided into two intervals, with regression coefficients β1
and β2, respectively. The meanings of all other variables are consistent
with those in Equation 7.
The first step is to conduct the threshold effect test, and the
results are shown in Table 15. It can be found that the economic
development level passes the single-threshold test at the 10% signifi-
cance level but fails the double-threshold test, indicating that there is
a single threshold effect in the impact of the DE on ACE, with a
threshold value of 10.8357. The second step is to perform the thresh-
old regression estimation, and the results are presented in Table 16.
When the level of economic development is below 10.8357, the coef-
ficient of the DE on ACE is −0.125. When the level of economic
development exceeds 10.8357, the coefficient becomes −0.143. It is
evident that an increase in the level of economic development
strengthens the inhibitory effect of the DE on ACE, thus supporting
Hypothesis H4. A possible reason is that higher economic develop-
ment provides material support for agricultural digitalization and
low-carbon development, promoting agricultural modernization. At
the same time, economic development improves farmers’ technical
capabilities, thereby facilitating coordination across the entire agri-
cultural chain and enhancing the agricultural carbon reduction effect
of the DE.
In practical terms, the threshold of 10.8357 constitutes a criti -
cal economic development threshold for the effectiveness of the
DE in reducing ACE. Below this threshold, limited local fiscal
capacity restricts investment in rural digital infrastructure, the
promotion of agricultural low-carbon technologies, and the devel -
opment of farmers’ digital skills. These constraints prevent the full
realization of the carbon mitigation potential of digital technolo -
gies. Above the threshold, a stronger economic base provides
material support for agricultural digitalization and the low-carbon
transition. It enables investment in rural networks and smart
machinery, as well as digital skills training for farmers, thereby
supporting agricultural modernization. Higher economic develop -
ment also improves farmers’ capacity to adopt and apply green
production technologies, promotes coordination across the agri -
cultural value chain, and strengthens the effect of the DE on ACE
reduction.
From a policy perspective, this threshold indicates that the agri-
cultural carbon mitigation benefits of the DE are not realized auto-
matically or unconditionally. Provinces below the threshold should
complement digital infrastructure investment with efforts to
strengthen the local economy and increase fiscal support for agricul-
ture and rural areas. Provinces above the threshold should use their
economic advantages to integrate digital technologies more effectively
into agricultural production and realize their carbon mitigation
potential. Agricultural digitalization and low-carbon transition poli-
cies should therefore be differentiated across regions and adapted to
local conditions.
6 Mechanism analysis of the impact of
the DE on ACE
6.1 Mediating variables
6.1.1 Rural entrepreneurship activity
Based on the research by Huang and Yin (2023), rural entrepre-
neurship activity is measured by the ratio of the sum of rural
TABLE 14 Regression results of moderation effects.
Variables Direct effect Indirect effect
de
−0.283***
(0.085)
−1.322***
(0.260)
fd
0.020
(0.039)
0.306***
(0.102)
de × fd
−0.923***
(0.124)
−1.622***
(0.296)
Control Variables Control
ρ −0.162*
R-squared 0.7500
Log-L 594.0005
Obs 403
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 15 frontiersin.org
individual entrepreneurs and employees in rural private enterprises to
the rural population.
6.1.2 Agricultural Total factor productivity (ATFP)
Based on the study by Li (2025), ATFP is calculated using the
super-efficiency Slack-Based Measure (SBM) model.
6.1.3 Data sources
Data for all the above indicators are obtained from the China
Statistical Y earbook, China Rural Statistical Y earbook, provincial
statistical yearbooks, and the provincial annual statistical data
released by the National Bureau of Statistics of China. Missing
data for certain provinces are imputed using the average growth
rate method. Variable definitions and descriptive statistics are
summarized in Table 17.
6.2 Model construction
Following the recommendation on mediation effect testing by
Jiang (2022) , this paper adopts the two-step approach to examine
the mediating effects in the relationship between the DE and
ACE. When testing the transmission mechanism of “DE → medi-
ating variable → ACE, ” the two-step method does not require an
explicit analysis of the direct effect between the mediating variable
and ACE; instead, their relationship is explored through theoreti -
cal reasoning. Specifically, this study constructs two-way fixed-
effects models to analyze the impact of the DE on the two
mediating factors—rural entrepreneurship activity and agricul -
tural total factor productivity, respectively. The corresponding
specifications are shown in Equations 11, 12 .
αβ β γζ ε
=+ + +++ ∑
6
1
2
it it s sit i t it
s
ea de Z
(11)
αβ β γζ ε
=+ + +++ ∑
6
1
2
it it s sit i t it
s
atfp de Z
(12)
The variable itea represents the rural entrepreneurship activity of
province i in year t ; itatfp represents the agricultural total factor pro-
ductivity of province i in year t ; itde denotes the level of DE develop-
ment of province i in year t ; itZ is a vector of control variables, whose
definitions are consistent with those in Equation 2; γi and ςt refer to
province fixed effects and time fixed effects, respectively; and εit is the
random error term.
6.3 Mediation effect test
6.3.1 Mediation effect test based on rural
entrepreneurship activity
The coefficient of the DE is positive and statistically significant at
the 1% level ( Table 18), indicating that the DE significantly boosts
rural entrepreneurship activity.
Increased rural entrepreneurship activity promotes the develop-
ment of non-agricultural industries such as rural e-commerce and
rural tourism, thereby raising the proportion of low-carbon industries
in the rural economy. Newly established agricultural enterprises tend
to adopt low-carbon operational models, and green organizational
structures are more likely to receive capital support facilitated by green
finance. This trend provides momentum for reducing ACE intensity
and supports sustainable agricultural development (Xu and Zhang,
2025). With the popularization of the internet and advances in fintech,
farmers now have more abundant and diverse channels to access
entrepreneurial opportunities. As farmers’ willingness to start busi-
nesses grows, the adoption of modern production technologies also
increases, effectively unlocking the innovative value of digital tech-
nologies (Dai and Guo, 2023) and driving the green transformation of
agricultural production methods (Wang and Y e, 2024).
6.3.2 Mediation effect test based on agricultural
total factor productivity
As shown in Table 18, the coefficient of the DE is positive and
significant at the 1% level, indicating that the DE significantly
improves ATFP .
The improvement in ATFP promotes the rational allocation of agri-
cultural resource factors and enables specialized resource utilization.
Meanwhile, optimized resource allocation facilitates the scaling-up of
agricultural production, which helps achieve economies of scale in
agrochemical use and thereby reduces overall agricultural machinery
power and fuel consumption. In addition, higher ATFP optimizes pro-
duction processes and improves efficiency. The adoption of modern
agricultural machinery and automation technologies raises the
TABLE 15 Threshold effect test results.
Threshold
variable
Threshold
number
F statistic P value Critical values Threshold
value
95%
confidence
interval1% 5% 10%
Level of economic
development
Single-threshold 42.24 0.0600 59.5271 43.3635 37.8160 10.8357 (10.8196,10.8358)
Double-threshold 21.87 0.1300 105.1277 60.9334 25.9827 - -
TABLE 16 Threshold regression estimation results.
Threshold variable lnace
deI
( ≤lnpgdp 10.8357)
−0.125***
(0.037)
deI
( ≥lnpgdp 10.8357)
−0.143***
(0.037)
Control variables Control
R2 0.5649
N 403
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 16 frontiersin.org
mechanization and automation levels of agricultural production, not
only reducing manual labor time and labor input but also lowering
carbon emissions generated during production (Li and Zhang, 2025).
Through technological innovations such as precise positioning and
intelligent analysis, smart agricultural machinery enables real-time
adjustments of operating parameters and accurate matching with envi-
ronmental conditions, effectively improving operational efficiency and
energy utilization. The upgrading of agricultural machinery technology
is not only an important manifestation of agricultural technological
progress but also a key pathway to achieving the synergistic effect of
pollution reduction and carbon mitigation (Gao and Lyu, 2023).
Overall, both rural entrepreneurship activity and agricultural total
factor productivity play mediating roles in the process through which
the DE curbs ACE, supporting Hypothesis H5.
7 Conclusions and recommendations
7.1 Research conclusions
First, this paper conducts an in-depth analysis of the impact
mechanism of the DE on ACE. Second, based on the panel data of 31
provinces in China from 2011 to 2023, this paper measures the levels
of the DE and ACE respectively, and explores the spatial distribution
characteristics of the DE and ACE. Third, the spatial Durbin model is
used to empirically analyze the spatial effects and regional heterogene-
ity of the DE on ACE. In addition, this paper explores the moderating
effect of the level of fiscal support for agriculture in the impact of the
DE on ACE, and further examines the threshold effect of the level of
economic development in this impact process. Finally, the mediating
model is adopted to empirically test the transmission mechanism of
the DE affecting ACE through rural entrepreneurship activity and
agricultural total factor productivity. The research results show that.
First, both the DE and ACE exhibit strong spatial autocorrelation,
with high–high and low–low clustering. Provinces consistently located
in the first quadrant for the DE are all in eastern China, in line with
earlier findings of stronger DE development in the east than in the
west (Liu et al., 2025). Provinces consistently located in the first quad-
rant for ACE are predominantly major agricultural provinces, reflect-
ing a positive association between the scale of agricultural production
and total carbon emissions.
Second, ACE have a positive spatial spillover effect ( Su et al.,
2023). The DE significantly reduces ACE in both local and neighbor-
ing provinces, consistent with evidence that digital technologies gen-
erate cross-regional knowledge spillovers (Shao et al., 2025). In major
grain-producing areas, ACE have a negative spatial spillover effect,
and the DE can only significantly inhibit ACE in the province itself; in
non-major grain-producing areas, the spatial spillover effect of ACE
is not significant, and the emission reduction effect of the DE on ACE
in the province itself is weaker than that in major grain-producing
areas. However, due to close regional economic links, the cross-
regional dissemination of its digital technology can inhibit ACE in
neighboring provinces. These differences suggest that the emission
reduction effects of digital technologies depend on both local agricul-
tural scale and absorptive capacity, as well as economic connectivity
across regions. Meanwhile, as digital rural development has pro -
gressed from initial exploration to nationwide implementation, the
spatial pattern of ACE has shifted from clustering toward competition.
The emission reduction pathways of the DE have also changed across
stages. In the earlier period, strong local emission reductions were
accompanied by substantial spillovers to neighboring provinces. In the
later period, the local effect became insignificant, leaving weaker
cross-regional spillovers as the principal emission reduction channel.
These findings highlight the dynamic, stage-dependent contribution
of the DE to low-carbon agricultural development.
Third, the level of fiscal support for agriculture plays the role of an
“amplifier” in the process of the DE inhibiting ACE. The increase in
its intensity can not only strengthen the direct emission reduction
effect of the DE on ACE in the province itself, but also enhance the
indirect inhibitory effect on ACE in neighboring provinces through
spatial spillover effects. The level of economic development has a
single threshold effect in the process of the DE inhibiting ACE. With
the improvement of the level of economic development, the inhibitory
effect of the DE on ACE is continuously enhanced. This threshold
effect is consistent with the theoretical expectation that the carbon
mitigation benefits of digital infrastructure depend on the stage of
economic development and absorptive capacity (Xia et al., 2026; Yuan
et al., 2026). More developed regions are better able to absorb and
apply digital technologies, thereby amplifying the carbon mitigation
benefits of the DE.
Fourth, the DE inhibits ACE by improving rural entrepreneurship
activity and agricultural total factor productivity. Both rural entrepre-
neurship activity and agricultural total factor productivity play a
mediating role in the impact of the DE on ACE. The mediating role of
TABLE 17 Variable definitions and descriptive statistics.
Variable
category
Variable name Variable
symbol
Unit Observations Mean Std. dev Min Max
Mediating
variable
Rural
entrepreneurship
activity
ea % 403 28.1393 45.1115 1.7971 308.6651
ATFP atfp - 403 0.6151 0.2261 0.2247 1.0647
TABLE 18 Regression results of mediation effect test.
Variables ea atfp
de
0.735***
(0.181)
0.286***
(0.095)
Control Variables Control Control
Time fixed effects Ye s Ye s
Individual fixed effects Ye s Ye s
R-squared 0.4251 0.8214
Obs 403 403
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 17 frontiersin.org
agricultural total factor productivity is consistent with earlier findings
that the DE reduces ACE intensity by improving total factor produc-
tivity (Feng et al., 2025). However, agricultural technological progress
can have opposing effects on emissions: advances in mechanical tech-
nologies may increase emissions, whereas resource-saving and bio-
chemical technologies can reduce them (Zhu et al., 2024). The type of
agricultural technological progress promoted by the DE therefore
influences its net emission reduction effect. This complex mechanism
warrants further investigation. Policies encouraging return migrants
to start rural businesses have also been shown to significantly reduce
ACE intensity (Luo et al., 2026 ). Our finding that the DE reduces
emissions by stimulating rural entrepreneurship activity complements
this policy evidence, suggesting that digitally enabled entrepreneur-
ship is an important micro-level driver of agriculture’s low-carbon
transition.
7.2 Policy recommendations
7.2.1 Strengthen digital infrastructure
construction and regional coordination to unlock
the spatial carbon emission reduction effect
In view of the spatial agglomeration characteristics of the DE and
ACE, carbon emission reduction should be promoted with the devel-
opment of the DE across the whole region as the core. On the one
hand, increase investment in digital infrastructure in key agricultural
production areas, make overall arrangements for infrastructures such
as 5G agricultural private networks, agricultural big data platforms
and Internet of Things monitoring systems, and simultaneously popu-
larize digital technologies including intelligent agricultural machinery
and precision water and fertilizer management. Through the digital
transformation of the whole process of farming, management and
harvesting, it improves the efficiency of agricultural production and
the accuracy of resource utilization within the province, and strength-
ens the direct inhibitory effect of the DE on local ACE from the
source. On the other hand, break the barriers to the flow of digital
resources among regions and establish a cross-provincial coordinated
development mechanism for digital agriculture. For example, promote
the joint construction of a digital technology sharing library between
digitally developed eastern regions and agricultural areas in central
and western China to share digital achievements such as smart plant-
ing plans and carbon footprint monitoring tools. Meanwhile, build a
cross-provincial agricultural digital cooperation platform to facilitate
the cross-regional flow of digital technology, talents and data ele -
ments, enable the carbon reduction effect of the DE to radiate to
neighboring provinces, form a spatial carbon reduction pattern of
“local emission reduction plus cross-regional coordination, ” and fully
release the overall inhibitory effect of the DE on regional ACE.
7.2.2 Implement regionally differentiated policies
to accurately unlock the carbon reduction
potential of DE
Targeted policies should be formulated according to the heteroge-
neous characteristics of major grain-producing areas and non-major
grain-producing areas. For major grain-producing areas, increase
investment in agricultural production digitalization, focus on links such
as precise fertilizer application and intelligent scheduling of agricultural
machinery, optimize resource allocation through digital technology, and
strengthen local emission reduction effects. At the same time, improve
the inter-regional allocation mechanism of agricultural production
materials to reduce fluctuations in emission reductions caused by
resource competition. For non-major grain-producing areas, relying on
their advantage of close economic ties, promote the integration of digital
technology with characteristic agriculture, enhance farmers’ digital
application capabilities, and establish a cross-regional digital technology
promotion alliance to amplify the radiation effect of technology and
assist emission reduction in neighboring regions.
7.2.3 Optimize fiscal support for agriculture and
economic environment to consolidate carbon
reduction guarantees
Make use of the amplifying effect of fiscal support for agriculture,
and increase financial support for the digital transformation of agri-
culture. For instance, set up special subsidies for digital agriculture to
focus on supporting projects such as the research and development of
smart agricultural machinery and digital technology training.
Meanwhile, improve the cross-regional overall planning mechanism
of fiscal funds to strengthen the emission reduction spillover support
for neighboring provinces. In view of the threshold effect of economic
development level, for economically underdeveloped areas, priority
should be given to improving the foundation for DE development, and
attracting digital enterprises to settle down through preferential taxa-
tion and talent introduction policies. For economically developed
areas, promote the in-depth integration of digital technology and agri-
cultural green technology to further enhance the carbon reduction
efficiency of the DE.
7.2.4 Foster rural entrepreneurship and improve
production efficiency to smooth the carbon
reduction transmission path
Build digital empowerment platforms focusing on the two medi-
ating variables of rural entrepreneurship activity and agricultural total
factor productivity. On the one hand, carry out rural digital entrepre-
neurship training, set up entrepreneurship support funds, and encour-
age farmers to develop green agricultural projects by utilizing digital
technology, such as agricultural product e-commerce and smart plant-
ing cooperatives, so as to drive the upgrading of agricultural produc-
tion modes through entrepreneurship. On the other hand, promote
the improvement of agricultural total factor productivity with the help
of digital technology, such as optimizing planting structure through
big data analysis and realizing precise control of production processes
via the Internet of Things, reducing resource waste and carbon emis-
sions, and forming a virtuous cycle of “digital technology --efficiency
improvement-carbon emission reduction. ”
8 Research limitations and future
research directions
Using provincial panel data from China for 2011–2023, this study
employs spatial Durbin, mediation, and threshold models to examine
the spatial effects, mechanisms, and heterogeneous impacts of the DE
on ACE. Several limitations remain. Nevertheless, several limitations
should be acknowledged. First, the DE index primarily covers digital
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 18 frontiersin.org
infrastructure, digital industrialization, and industrial digitalization,
without fully capturing applications of emerging technologies such as
big data, artificial intelligence, and blockchain. It may therefore incom-
pletely represent the contribution of the DE to ACE reduction. In addi-
tion, a small number of missing provincial observations are imputed
using the average growth rate method. This method assumes smooth
changes over time and cannot capture short-term fluctuations caused
by policy changes or external shocks. Imputed values may differ from
actual observations, introducing uncertainty into the empirical results.
Second, the analysis relies on provincial macro-level panel data and
lacks micro-level data on counties, villages, and agricultural business
entities. It therefore cannot fully reveal the behavioral responses of farm-
ers and emerging agricultural businesses, and its micro-level founda-
tions require further investigation. Third, the sample and analytical
setting are specific to China’s agricultural production, DE policies, and
institutional environment. Other countries and regions differ substan-
tially in production systems, farm size, digital infrastructure, agricul-
tural low-carbon governance, and conditions for technology adoption.
The estimated emission reduction effects, spatial spillovers, and thresh-
old characteristics therefore cannot be directly generalized to other
countries or economies.
Future research could improve the measurement of the DE by
incorporating additional emerging technology dimensions.
Combining household survey data, remote sensing observations, and
platform data would enable analysis across multiple spatial scales and
help clarify the micro-level mechanisms of digitally enabled carbon
mitigation. Further examination of mediation pathways and dynamic
effects, together with cross-country and cross-regional comparisons,
could improve external validity and provide more targeted guidance
for the digital transformation of low-carbon agriculture.
Data availability statement
The original contributions presented in the study are included in
the article/supplementary material, further inquiries can be directed
to the corresponding author.
Author contributions
XH: Conceptualization, Data curation, Resources, Software,
Supervision, Writing – original draft. ZX: Investigation, Validation,
Writing – review & editing. QZ: Writing – original draft, Methodology,
Validation. DZ: Investigation, Project administration, Visualization,
Writing – review & editing. FZ: Formal analysis, Methodology, Project
administration, Resources, Validation, Writing – original draft.
Funding
The author(s) declared that financial support was received for this
work and/or its publication. This research was funded by the Anhui
Provincial Philosophy and Social Science Planning Y outh Project
“Research on Mechanisms and Pathways of Digital Economy
Empowering the Development of New Quality Productive Forces in
Anhui Agriculture” (No. AHSKYQ2025D101).
Conflict of interest
The author(s) declared that this work was conducted in the
absence of any commercial or financial relationships that could be
construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that Generative AI was not used in the
creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this
article has been generated by Frontiers with the support of artificial
intelligence and reasonable efforts have been made to ensure accuracy,
including review by the authors wherever possible. If you identify any
issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated organizations,
or those of the publisher, the editors and the reviewers. Any product
that may be evaluated in this article, or claim that may be made by its
manufacturer, is not guaranteed or endorsed by the publisher.
References
Agrawal, M., Singh, S., and Jain, J. K. (2025). “Digitization in agriculture: leveraging the digi-
tal economy for enhanced supply chain efficiency and sustainability, ” in Dipak Saha, Mrinal
Kanti Das, Sudhir Rana, Kavita Sharma, and Swati Mukherjee Eds., Agile Sustainable
Marketing: Navigating Towards a Greener Future, (Sharjah: Bentham Science Publishers),
121–134. doi: 10.2174/97898153240681250101
Alam, I., Lu, S., Sarfraz, M., and Mohsin, M. (2023). The interplay of green technology
and energy consumption: a study of China’s carbon neutrality and sustainable digital
economy. Energies 16:6184. doi: 10.3390/en16176184
Bian, J., Tang, D., and Fang, Y . (2025). The dual impact and spatial spillover effects of the
digital economy on urban–rural integration. Sustainability 17:545. doi: 10.3390/
su17020545
Chen, L., and Zhang, Y . (2023). Does the development of the digital economy promote
common prosperity?—analysis based on 284 cities in China. Sustainability 15:4688. doi:
10.3390/su15054688
Dai, Z., and Guo, R. (2023). Digital economy, rural entrepreneurship and common pros-
perity. J. Yunnan Agric. Univ. 17, 71–79. doi: 10.12371/j.ynau(s).202208023
Ding, C., Liu, C., Zheng, C., and Li, F . (2021). Digital economy, technological innovation
and high-quality economic development: based on spatial effect and mediation effect.
Sustainability 14:216. doi: 10.3390/su14010216
Dolores, U. M. (2011). Introduction to Spatial Econometrics [Book review]. J. R. Stat. Soc.
A Stat. Soc. 174: 513–514. doi: 10.1111/j.1467-985X.2010.00681_13.x
Hou et al. 10.3389/fsufs.2026.1942827
Frontiers in Sustainable Food Systems 19 frontiersin.org
Duan, H., Zhang, Y ., Zhao, J., and Bian, X. (2011). Carbon footprint analysis of farmland
ecosysterm in China. J. Soil Water Conserv. 25, 203–208. doi: 10.13870/j.cnki.
stbcxb.2011.05.020
Fang, Y ., Cao, H., and Sun, J. (2022). Impact of artificial intelligence on regional green
development under China’s environmental decentralization system—based on spatial
durbin model and threshold effect. Int. J. Environ. Res. Public Health 19:14776. doi:
10.3390/ijerph192214776
Feng, S., Long, Y ., Y asmeen, R., and Y ang, H. (2025). The spatial spillover impact of digital
finance on agricultural carbon emission intensity: evidence from China. Sustainability
17:8583. doi: 10.3390/su17198583
Gao, D., and Lyu, X. (2023). Agricultural total factor productivity, digital economy and
agricultural high-quality development. PLoS One 18:e0292001. doi: 10.1371/journal.
pone.0292001
Golovina, S. G., Poltarykhin, A. L., Zhuravlev, P . V ., and Mikolaychik, I. N. (2022). Income
of the rural population is a condition for the formation of human capital in rural areas.
Sib. J. Life Sci. Agric., 14, 83–102. doi: 10.12731/2658-6649-2022-14-1-83-102
Guo, L., Zhu, H., and Jiang, S. (2026). How does digital technology innovation affect
agricultural carbon emissions?—empirical evidence from China. Environ. Dev. Sustain.,
1–32. doi: 10.1007/s10668-026-07547-3
Guo, S., and Guo, H. (2024). Development of ecological low-carbon agriculture with
Chinese characteristics in the new era: features, practical issues, and pathways.
Sustainability. 16:7844. doi: 10.3390/agronomy14071460
Guo, Z., Ho, C. S., Ling, G. H. T., and Rashid, M. F . (2025). Spatial spillover heterogeneity
and moderated effects of the digital economy on agricultural carbon emissions: evidence
from 30 Chinese provinces. Environ. Dev. Sustain. 28, 27007–27037. doi: 10.1007/
s10668-025-06013-w
Guo, Z., Ho, C. S., Rashid, M. F ., Chau, L. W ., Toh, S. Y ., and Liu, M. (2024). Impacts of
the digital economy on agricultural carbon emissions: empirical analyses from 31 prov-
inces in China. Chem. Eng. Trans. 113, 547–552. doi: 10.3303/CET24113092
Hao, T., and Y ang, H. (2026). Study on the effects, mechanisms, and pathways of environ-
mental regulation on agricultural carbon intensity reduction. Res. Agric. Mod. 47,
995–1011. doi: 10.13872/j.1000-0275.2025.0619
Hou, X., Zhang, D., Fu, L., Zhang, D., and Zeng, F . (2023). Spatio-temporal evolution and
influencing factors of coupling coordination degree between urban–rural integration and
digital economy. Sustainability 15:9718. doi: 10.3390/su15129718
Huang, D., and Yin, K. (2023). Can the development of digital inclusive finance boost
rural entrepreneurial activity? Finance and Economy. , 43–53. doi: 10.19622/j.cnki.
cn36-1005/f.2023.08.004
Huang, J., Lu, H., and Du, M. (2025). Regional differences in agricultural carbon emis-
sions in China: measurement, decomposition, and influencing factors. Land 14:682. doi:
10.3390/land14040682
Huan, H., Wang, L., and Zhang, Y . (2025). Regional differences, convergence character-
istics, and carbon peaking prediction of agricultural carbon emissions in China. Environ.
Pollut. 366:125477. doi: 10.1016/j.envpol.2024.125477
IPCC (2013). Climate Change 2013: The Physical Science Basis. Contribution of Working
Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate
Change [Stocker, T.F ., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A.
Nauels, Y . Xia, V . Bex and P .M. Midgley (eds.)]. Cambridge: Cambridge University Press,
pp. 1535. doi: 10.1017/CBO9781107415324
Jiang, T. (2022). Mediating effects and moderating effects in causal inference. China Ind.
Econ., 100–120. doi: 10.19581/j.cnki.ciejournal.2022.05.005
Jin, M., Feng, Y ., Wang, S., Chen, N., and Cao, F . (2024). Can the development of the rural
digital economy reduce agricultural carbon emissions? A spatiotemporal empirical study
based on China’s provinces. Sci. Total Environ . 939: 173437. doi: 10.1016/j.
scitotenv.2024.173437
Ke, S., Wu, Y ., Cui, H., and Lu, X. (2023). The mechanisms and spatial-temporal effects
of farmland spatial transition on agricultural carbon emission: based on 2018 counties in
China. Environ. Sci. Pollut. Res. Int. 30, 107716–107732. doi: 10.1007/s11356-023-29860-4
Leng, X., and Tong, G. (2022). The digital economy empowers the sustainable develop-
ment of China’s agriculture-related industries. Sustainability 14:10967. doi: 10.3390/
su141710967
Liang, J., and Qiao, C. (2025). Digital economy and high-quality agricultural develop-
ment: mechanisms of technological innovation and spatial spillover effects. Sustainability
17:3639. doi: 10.3390/su17083639
Li, B., and Gao, Y . (2024). Impact and transmission mechanism of digital economy on
agricultural energy carbon emission reduction. Int. Rev. Econ. Finance 95:103457. doi:
10.1016/j.iref.2024.103457
Li, B., Zhang, J., and Li, H. (2011). Research on spatial-temporal characteristics and affect-
ing factors decomposition of agricultural carbon emission in China. China Popul. Resour.
Environ. 21, 80–86. Available online at: https://kns.cnki.net/kcms2/article/abstract?-
v=7DtDJWciuTI1l0CCB8qXnJhIbx0kYHKNR3Q71SpoKABMGRTCwe13yr2LCDe5D
qEDZkTQY_oqgSyaocEJITLnAQoA-AH9ofjRfXUvdfoq4vlGomZl1yx85xF4WXvYF00J-
VOBfFD2NQvHQQ0UineHysnRRHqUM7YpzgwvH28gVk1Acdt2gk9zPzw==&unipla
tform=NZKPT&language=CHS (Accessed November 10, 2026).
Li, H., Li, K., Maneejuk, P ., and Liu, J. (2026). How the digital economy reduces agricul-
tural carbon emissions: mechanisms, threshold effects, and policy implications.
Agriculture 16:478. doi: 10.3390/agriculture16040478
Li, H., Xie, S., and Su, M. (2024). Does digital technology innovation promote low-carbon
development in agriculture?: a spatial econometric analysis based on 31 provinces in
China. Environ. Sci. Pollut. Res. 31, 4478–4499. doi: 10.1007/s11356-023-31369-9
Li, J., Sheng, X., Zhang, S., and Wang, Y . X. (2024). Research on the impact of the digital
economy and technological innovation on agricultural carbon emissions. Land 13:821.
doi: 10.3390/land13060821
Li, Q., and Zhang, T. (2025). Towards agricultural green development: how does digital
economy enhance agricultural green total factor productivity? Front. Sustain. Food Syst.
9:1700716. doi: 10.3389/fsufs.2025.1700716
Liu, X., and Sheng, P . (2026). Fintech, new quality productivity and high-quality eco-
nomic development. Stat. Decis. 42, 5–10. doi: 10.13546/j.cnki.tjyjc.2026.14.001
Liu, Y ., Pu, X., and Zhang, G. (2025). Spatial spillover effect and heterogeneity of digital
economy on agricultural carbon emissions. Sci. Rep. 15:37693. doi: 10.1038/
s41598-025-21487-4
Li, X. (2025). Study on the impact of the digital economy on green total factor productiv-
ity in agriculture. J. Econ. Manag. Sci. 8:54. doi: 10.30560/jems.v8n5p54
Luo, H., Xiang, Z., and Xie, T. (2026). Impact of rural return entrepreneurship pilot poli-
cies on agricultural carbon emissions in China's Y angtze River Economic Belt. Socio Econ.
Plan. Sci. 103:102373. doi: 10.1016/j.seps.2025.102373
Luo, R., and Zhou, N. (2022). Dynamic evolution, spatial differences, and driving factors
of China’s provincial digital economy. Sustainability 14:9376. doi: 10.3390/su14159376
Lu, Y ., and Zhou, L. L. (2023). Can digital economy improve employment structure?—
mediating effect based on a spatial Durbin model. PLoS One 18:e0287849. doi: 10.1371/
journal.pone.0287849
Lu, Z., Pi, Y ., Su, H., Li, Z., Ma, X., Su, X., et al. (2025). Exploring the impact of rural
human capital (RHC) on agricultural carbon emissions: evidence from China. Energy
论文原文与出版信息
Frontiers in Sustainable Food Systems