0 cumulative citations
View corpus contextChinese manufacturing firms that meaningfully embed AI raise carbon productivity, chiefly by generating green innovations and improving energy efficiency; gains concentrate in high‑emission firms, low‑competition industries and regions with weaker environmental regulation.
Artificial Intelligence Embedding and Corporate Carbon Productivity: Mechanisms of Technological Progress and Efficiency Optimization
openalex
correlational
medium evidence
7/10 relevance
Summary only
summary available; pdf_status=paywall
DOI
Source
PDF
Higher AI embedding in Chinese listed manufacturing firms is associated with increased corporate carbon productivity, primarily via green technological innovation and improved energy utilization efficiency, with larger effects in high‑emission firms, less competitive industries, and regions with looser environmental regulation.
Citation observations
Cumulative provider counts captured on specific dates; providers are never combined.
ABSTRACT The driving mechanism of artificial intelligence (AI) embedding on manufacturing firms' carbon productivity has emerged as a critical pathway for promoting industrial green transformation and high‐quality development. Drawing upon absorptive capacity theory and the resource‐based view (RBV), this study constructs a dual‐mediator model driven by technology and efficiency. Grounded in the technology‐organization‐environment (TOE) framework, we integrate multidimensional indicators and employ the entropy method to establish an AI embedding evaluation index system that effectively distinguishes substantive AI embedding from symbolic AI embedding. Using a sample of A‐share listed manufacturing firms on the Shanghai and Shenzhen stock exchanges from 2008 to 2023, we systematically investigate the effect of AI embedding on corporate carbon productivity and its underlying transmission mechanisms. The results indicate that AI embedding drives the improvement of corporate carbon productivity primarily through two pathways: the enabling path of green technological innovation and the integrating path of energy utilization efficiency. The former creates incremental low‐carbon resources through technological breakthroughs, whereas the latter optimizes the allocation of existing resources through factor reorganization. Heterogeneity analysis reveals that the positive association between AI embedding and carbon productivity is more pronounced in firms with high carbon emissions, in industries with low competition, and in regions with less stringent environmental regulations. This study extends the explanatory boundary of the RBV regarding the construction of corporate green capabilities in the digital context and aims to provide both theoretical support and practical implications for achieving the synergistic goals of carbon reduction and efficiency improvement in the manufacturing sector.
Summary
Main Finding
AI embedding in manufacturing firms raises corporate carbon productivity primarily through two mediated channels: (1) an enabling path via green technological innovation (creating incremental low‑carbon resources), and (2) an integrating path via improved energy utilization efficiency (optimizing allocation of existing resources). The effect is stronger for firms with high carbon emissions, in low‑competition industries, and in regions with less stringent environmental regulation.
Key Points
- Conceptual framing: integrates absorptive capacity theory and the resource‑based view (RBV), within a technology‑organization‑environment (TOE) framework.
- Dual mediators: green technological innovation (technology channel) and energy utilization efficiency (efficiency channel).
- Measurement: constructs a multidimensional AI embedding index that differentiates substantive vs symbolic AI embedding using the entropy method.
- Empirical scope: sample of A‑share listed manufacturing firms on Shanghai and Shenzhen exchanges, 2008–2023.
- Heterogeneity: larger carbon‑productivity gains where firms emit more carbon, industry competition is low, or regional environmental regulation is less strict.
- Contribution: extends RBV to explain construction of corporate green capabilities in a digital/AI context and identifies concrete transmission mechanisms linking AI adoption to environmental performance.
Data & Methods
- Sample: publicly listed manufacturing firms (A‑share) on Shanghai and Shenzhen stock exchanges, 2008–2023.
- AI embedding index: multi‑indicator composite built under TOE dimensions; entropy method used to weight indicators and to separate substantive embedding (meaningful integration) from symbolic adoption (superficial use).
- Empirical model: a dual‑mediator framework testing direct effect of AI embedding on corporate carbon productivity and indirect effects via (a) green technological innovation and (b) energy utilization efficiency.
- Heterogeneity tests: subsample analyses by firm carbon intensity, industry competition level, and regional environmental regulation stringency.
Implications for AI Economics
- Measurement matters: distinguishing substantive from symbolic AI adoption is crucial for estimating AI’s real environmental and productivity impacts; composite, well‑weighted indices (e.g., entropy method) improve inference.
- Mechanisms: AI contributes both to innovation (new, low‑carbon capabilities) and to reallocation/efficiency gains — models of AI’s economic impact should capture both technology‑creation and factor‑integration channels.
- Targeting policy and investment: prioritizing substantive AI embedding in high‑emission firms/regions and in less competitive industries can yield larger carbon‑productivity returns; policymakers should couple AI adoption incentives with support for green R&D and energy‑efficiency measures.
- Regulation design: looser environmental regulation can coincide with stronger AI‑driven gains in carbon productivity, suggesting complementary roles for regulation and digital investment; careful regulatory design can balance incentives for efficiency improvements vs. technology upgrades.
- Theoretical extension: provides empirical support for extending RBV to digital capabilities — AI can be framed as a firm resource that, via absorptive capacity, builds green capabilities and improves environmental performance.
Assessment
Paper Typecorrelational
Evidence Strengthmedium — The paper uses a long firm‑level panel (2008–2023), a carefully constructed AI embedding index (entropy weighting and substantive vs symbolic separation), mediation analysis, and heterogeneity/robustness tests which lend credibility to the correlations and mechanism story; however, causal identification is limited because adoption is endogenous (selection, reverse causality, omitted confounders), measurement error in the composite AI index is possible, and no clearly exogenous variation (IV/DID/RCT) is reported to rule out alternative explanations.
Methods Rigormedium — Strengths: systematic index construction (entropy method), explicit mediation framework, panel scope and heterogeneity analyses, and linkage to theoretical frameworks (RBV, TOE, absorptive capacity). Weaknesses: apparent lack of a convincing exogenous identification strategy to address endogeneity, potential measurement and omitted-variable bias, and limited discussion (in the supplied text) of dynamic endogeneity or robustness to reverse causality.
SampleFirm‑year panel of publicly listed A‑share manufacturing firms on the Shanghai and Shenzhen stock exchanges, 2008–2023; uses firm financials, emission/carbon data, patent/innovation indicators, energy consumption/utilization measures, and region/industry identifiers to construct AI embedding index and mediators.
Themesproductivity innovation
IdentificationObservational panel analysis using a constructed multi-indicator AI embedding index and mediation regressions to estimate direct and indirect effects on corporate carbon productivity (mediators: green technological innovation and energy utilization efficiency); includes heterogeneity and robustness checks across firm, industry, and region controls and likely firm/year fixed effects — no clear exogenous source of variation (no instrumental variable or natural experiment) is described.
GeneralizabilitySample limited to publicly listed Chinese manufacturing firms (A‑shares) — excludes SMEs, unlisted firms, services, and non‑Chinese contexts., Chinese regulatory, market, and industry structures (including environmental regulation enforcement) may limit transferability to other countries., Index construction and the substantive vs symbolic AI distinction depend on chosen indicators and weighting (entropy method) — measurement choices may not generalize., Findings apply to a 2008–2023 window; rapid changes in AI capabilities and diffusion after the sample may affect external validity., Heterogeneity results suggest context dependence (firm emissions, competition, regulation), so average effects mask considerable variation across settings.
Claims (7)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| AI embedding in manufacturing firms increases corporate carbon productivity. Firm Productivity | positive | Corporate carbon productivity |
Reading fidelity
high
Study strength
medium
|
not reported
|
| Green technological innovation mediates the positive relationship between AI embedding and corporate carbon productivity. Innovation Output | positive | Corporate carbon productivity through green technological innovation |
Reading fidelity
high
Study strength
medium
|
not reported
|
| Improved energy utilization efficiency mediates the positive relationship between AI embedding and corporate carbon productivity. Firm Productivity | positive | Corporate carbon productivity through energy utilization efficiency |
Reading fidelity
high
Study strength
medium
|
not reported
|
| The carbon-productivity effect of AI embedding is stronger among firms with high carbon emissions. Firm Productivity | positive | AI-related gains in corporate carbon productivity by firm carbon-emission level |
Reading fidelity
high
Study strength
medium
|
not reported
|
| The carbon-productivity effect of AI embedding is stronger in industries with low competition. Firm Productivity | positive | AI-related gains in corporate carbon productivity by industry competition level |
Reading fidelity
high
Study strength
medium
|
not reported
|
| The carbon-productivity effect of AI embedding is stronger in regions with less stringent environmental regulation. Firm Productivity | positive | AI-related gains in corporate carbon productivity by regional environmental-regulation stringency |
Reading fidelity
high
Study strength
medium
|
not reported
|
| The study measures AI embedding using a multidimensional composite index that distinguishes substantive AI embedding from symbolic adoption. Adoption Rate | other | AI adoption or embedding rate |
Reading fidelity
high
Study strength
low
|
not reported
|