Provinces with deeper digital–intelligent integration emit less carbon per unit of output, partly because digitization spurs industrial upgrading and green innovation; benefits also leak to nearby provinces but vary by geography, calling for regionally tailored policies.
Research on the Pathways and Spatial Effects of Digital–Intelligent Integration on Carbon Emission Intensity
Using Chinese provincial panel data (2014–2023), greater digital–intelligent integration is associated with lower carbon intensity, with industrial upgrading and green-technology innovation partly mediating the effect and geographically limited positive spillovers to neighboring provinces.
In the context of global efforts to achieve carbon neutrality, understanding how digital–intelligent integration influences carbon emissions is crucial for advancing the ecological transition. Using panel data from 30 provincial-level regions in China (2014–2023), a digital–intelligent integration index was constructed via entropy weighting and a coupling coordination model. Employing fixed-effects, mediation, and spatial Durbin models, the analysis shows that digital–intelligent integration is significantly associated with lower carbon intensity, a result that is robust to endogeneity concerns and alternative specifications. Industrial structure upgrading and green technology innovation were identified as mediating pathways. Furthermore, digital–intelligent integration generates positive spatial spillovers, reducing carbon intensity in neighboring provinces. Notably, these spillovers are geographically constrained and vary significantly across the regions. These findings indicate the need to formulate regionally differentiated strategies to harness the specific mechanisms through which digital–intelligent integration operates in different contexts.
Summary
Main Finding
Digital–intelligent integration significantly reduces carbon intensity across Chinese provinces (2014–2023). This effect is robust to alternative specifications and endogeneity checks, operates partly through industrial-structure upgrading and green-technology innovation, and produces positive but geographically constrained spatial spillovers to neighboring provinces that vary by region.
Key Points
- Sample: panel of 30 provincial-level regions in China, 2014–2023.
- Dependent variable: carbon intensity (carbon emissions per unit of GDP).
- Core explanatory variable: a digital–intelligent integration index constructed with entropy weighting and a coupling coordination model.
- Main econometric approaches: fixed-effects panel models, mediation analysis, and spatial Durbin models.
- Results:
- Higher digital–intelligent integration → lower carbon intensity (statistically significant).
- Mechanisms: industrial structure upgrading and green-technology innovation are identified as significant mediators.
- Spatial effects: integration reduces carbon intensity in neighboring provinces (positive spillovers), but these effects are geographically limited and heterogeneous across regions.
- Robustness: findings hold under alternative specifications and after addressing endogeneity concerns.
Data & Methods
- Data: provincial-level panel data for 30 Chinese regions, covering 2014–2023.
- Index construction: a composite digital–intelligent integration index built using entropy weighting and a coupling coordination framework to capture the interaction between digitalization and intelligence (AI-related capabilities).
- Main models:
- Fixed-effects panel regressions to estimate within-province effects over time and control for time-invariant heterogeneity.
- Mediation analysis to test whether industrial-structure upgrading and green-technology innovation transmit the effect of digital–intelligent integration on carbon intensity.
- Spatial Durbin models to capture spatial dependence and quantify spillover effects to neighboring provinces.
- Identification/robustness: robustness checks and procedures to mitigate endogeneity concerns (alternative specifications, robustness tests, and spatial controls) were performed; results remain consistent.
Implications for AI Economics
- Mechanisms matter: AI and digital-intelligent integration reduce emissions not only via direct efficiency gains but importantly by speeding industrial upgrading and spurring green-technology innovation — channels that economists and policymakers should model and measure explicitly.
- Spatial externalities: AI/digital investments produce cross-jurisdictional benefits (positive spillovers). Economic models and policy designs need to account for geographically bounded spatial externalities when allocating resources and designing incentives.
- Heterogeneity and targeting: heterogeneous regional responses imply that one-size-fits-all policy is suboptimal. Region-specific strategies (e.g., focusing on digital infrastructure in lagging regions, scaling green-tech R&D in innovation hubs) will better leverage AI-driven decarbonization.
- Measurement and evaluation: the study’s composite index and coupling approach provide a template for quantifying digital–intelligent integration in empirical studies; spatial econometric methods are important for capturing diffusion and externalities.
- Policy implications for AI governance: subsidizing digital–intelligent adoption, aligning AI deployment with green-innovation incentives, and coordinating interregional policies can amplify decarbonization benefits. Cost–benefit assessments should internalize spillovers and heterogeneity.
- Research opportunities: further work could sharpen causal identification (e.g., quasi-experiments or IVs), examine firm- or sector-level mechanisms, explore dynamic adoption and diffusion patterns of AI-driven green technologies, and quantify welfare and distributional consequences of AI-enabled decarbonization.
Assessment
Paper Typequasi_experimental
Evidence Strengthmedium — Uses panel fixed effects, mediation analysis, and spatial econometrics which provide credible correlational evidence and address some confounding and spatial dependence, but lacks a clearly exogenous source of variation (natural experiment or convincingly exogenous instrument) so causal interpretation remains tentative and vulnerable to remaining omitted variables, measurement error in the constructed index, and reverse causality.
Methods Rigormedium — The study applies a solid set of empirical tools (entropy-weighted index construction, coupling coordination model, fixed effects, mediation tests, spatial Durbin) and reports robustness checks; however, the constructed digital–intelligent integration index may contain measurement choices that drive results, details on endogeneity correction are not fully specified here, and causal leverage would be stronger with an explicit exogenous shock or valid instrument.
SampleAnnual panel of 30 Chinese provincial-level regions from 2014–2023 (≈300 province-year observations); constructed a province-level digital–intelligent integration index using entropy weighting and a coupling coordination model; outcome is provincial carbon intensity; mediators include measures of industrial-structure upgrading and green-technology innovation; spatial relationships modeled using interprovincial proximity/weight matrix.
Themesinnovation adoption
IdentificationPanel two-way fixed-effects (province and year) to control for time-invariant and common time shocks, extensive covariates, robustness checks addressing endogeneity (authors report alternative specifications and checks), mediation analysis to identify pathways, and a spatial Durbin model to capture spatial spillovers and spatial dependence.
GeneralizabilityChina-only provincial data — findings may not generalize to other countries with different institutional, industrial, or energy contexts, Provincial-aggregate analysis masks within-province, sectoral, and firm-level heterogeneity, Constructed digital–intelligent integration index depends on chosen indicators and weighting method, limiting comparability to other measures of digitalization/AI adoption, Relatively short panel (2014–2023) may miss longer-term dynamics and structural breaks (e.g., major policy changes or technological shocks), Spatial spillover patterns depend on China's geographic and policy landscape and may differ in other settings
Claims (8)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| The study uses panel data from 30 provincial-level regions in China covering 2014–2023 to analyze the relationship between digital–intelligent integration and carbon intensity. Other | null_result | high | dataset coverage (30 provinces, 2014–2023) |
n=30
0.48
|
| A digital–intelligent integration index was constructed using entropy weighting and a coupling coordination model. Other | null_result | high | construction of digital–intelligent integration index |
n=30
0.48
|
| Digital–intelligent integration is significantly associated with lower carbon intensity. Organizational Efficiency | negative | high | carbon intensity |
n=30
0.48
|
| The negative association between digital–intelligent integration and carbon intensity is robust to endogeneity concerns and alternative model specifications. Organizational Efficiency | negative | high | carbon intensity (robustness of estimated association) |
n=30
0.48
|
| Industrial structure upgrading and green technology innovation were identified as mediating pathways through which digital–intelligent integration reduces carbon intensity. Organizational Efficiency | negative | high | carbon intensity (mediated by industrial structure upgrading and green technology innovation) |
n=30
0.48
|
| Digital–intelligent integration generates positive spatial spillovers, reducing carbon intensity in neighboring provinces. Organizational Efficiency | negative | high | carbon intensity in neighboring provinces (spatial spillover effect) |
n=30
0.48
|
| The spatial spillover effects are geographically constrained and vary significantly across regions. Organizational Efficiency | mixed | high | heterogeneity of spatial spillover effects on carbon intensity across regions |
n=30
0.48
|
| Policy implication: regionally differentiated strategies are needed to harness the mechanisms through which digital–intelligent integration reduces carbon intensity in different contexts. Governance And Regulation | positive | high | policy formulation (regional differentiation) |
n=30
0.08
|