0 cumulative citations
View corpus contextAI-linked activity is associated with higher export markups for Chinese manufacturers, helping some firms escape low-value traps. 'Intelligent empowerment' delivers bigger markup gains than pure automation, with the biggest payoff for state-owned, high-productivity, and quality-focused exporters to developed countries.
Artificial Intelligence and Enterprise Export Price Markup in China’s Economy: Based on Large Language Models
openalex
correlational
medium evidence
8/10 relevance
Summary only
summary available; pdf_status=not_found
DOI
Source
PDF
Using China customs and listed-firm data (2007–2016) and LLM-extracted annual-report indicators of AI activity, the paper finds firms with stronger AI signals have higher export price markups—driven by efficiency gains and innovation but partially offset by tougher competition and greater transparency— with larger benefits for SOEs, high-productivity firms, and those exporting to developed markets.
Citation observations
Cumulative provider counts captured on specific dates; providers are never combined.
1 cumulative citations
View corpus contextChina’s exports face long-standing challenges of low quality, low price, and low profit margins. Amid intensifying global economic competition and deeper value chains, breaking the ‘low-value-added lock-in’ to boost export enterprises’ profit margins is urgent for addressing foreign trade bottlenecks. This study systematically examines artificial intelligence (AI)’s effect and mechanism on export enterprises’ price markup. It integrates 2007–2016 data from China’s Customs Database and Shanghai and Shenzhen A-share manufacturing listed companies, combined with AI indicators built by extracting annual report text via large language models. Results show AI significantly raises export enterprises’ price markup, with the ‘intelligent empowerment’ model having a stronger promotional effect than ‘machine replacing human’. Mechanism analysis reveals AI’s dual impacts: positive effects from efficiency improvement and technological innovation, and negative effects from intensified market competition and higher information transparency. Heterogeneity analysis finds AI benefits state-owned enterprises, high-productivity enterprises, quality-competitive enterprises, and those exporting to developed countries more. Additionally, it reduces markup dispersion and improves resource allocation efficiency. Overall, AI drives export enterprises to break low-value-added dilemmas, optimise resource allocation, and further boost China’s export economy.
Summary
Main Finding
AI adoption significantly increases export firms’ price markups in China (2007–2016). The “intelligent empowerment” mode of AI (augmenting human decision-making and processes) raises markups more strongly than the “machine replacing human” mode. AI’s net effect reflects two opposing channels — positive gains from efficiency and innovation versus negative effects from intensified competition and greater information transparency — with the positive channels dominating in the sample. AI also narrows markup dispersion and improves resource-allocation efficiency, helping firms break a long-standing low-value-added trap in China’s export sector.
Key Points
- Data span: 2007–2016, combining China Customs trade records and Shanghai & Shenzhen A‑share manufacturing listed companies.
- AI measurement: firm-level AI indicators constructed by extracting and coding annual report text using large language models.
- Main effect: AI adoption is associated with higher export price markups.
- Modes of AI:
- Intelligent empowerment (AI as augmentation) has a larger positive effect on markups than substitution (machines replacing human labor).
- Mechanisms (dual effects):
- Positive channels: productivity/efficiency improvements and technological innovation raise markups.
- Negative channels: intensified market competition and increased information transparency reduce markups.
- Heterogeneous effects: AI’s markup-raising effect is stronger for state-owned enterprises, high-productivity firms, firms competing on quality, and firms exporting to developed-country markets.
- Aggregate effects: AI adoption reduces dispersion in markups across firms and improves resource-allocation efficiency in the export sector.
Data & Methods
- Data sources: matched firm-level panel from China Customs (trade records) and A‑share manufacturing listed firms (financials and annual reports), 2007–2016.
- AI indicators: created by automated extraction and coding of annual-report language using large language models to identify AI-related activities and modalities (e.g., intelligent empowerment vs machine replacement).
- Empirical strategy: firm-level panel analysis examining the relationship between AI indicators and export markups, with mechanism tests and heterogeneity analyses. Robustness checks and tests for dispersion/resource allocation effects were performed (details in the paper).
- Identification: the study uses within-firm variation over time in documented AI activity to link AI exposure to markup outcomes and explores plausibly exogenous cross-firm and cross-market heterogeneity to support inference.
Implications for AI Economics
- For theory:
- AI affects firm pricing power through competing channels — productivity/innovation gains that raise markups and competitive/transparency effects that compress them. Models of AI’s macro and micro impacts should incorporate both channels.
- Distinguish AI modes (augmentation vs substitution) in theoretical and empirical work: augmentation can be more pro-profit than substitution.
- For measurement:
- Text-derived AI indicators via LLMs provide a scalable way to measure firm-level AI activity across large administrative and financial datasets.
- For policy and industry:
- Policies that encourage intelligent-empowerment AI (augmentation, process redesign, managerial analytics) may yield stronger export-price and value-added gains than policies simply promoting automation.
- Complementary investments (skills, R&D, quality upgrading) amplify AI’s positive effects, especially for firms active in developed-country markets.
- Attention is needed to distributional and competitive effects: smaller or low-productivity firms may be disadvantaged as AI concentrates gains in higher-productivity and state-owned firms.
- For future research:
- Examine longer-term dynamics, causal identification strategies (e.g., exogenous shocks to AI adoption), sectoral variation beyond listed manufacturing, and labor-market/distributional consequences of the AI-driven re-allocation.
Assessment
Paper Typecorrelational
Evidence Strengthmedium — Large administrative and financial panel data and careful heterogeneity/mechanism analysis give credible associative evidence that AI-related activity correlates with higher export markups, but the study lacks a clear exogenous identification strategy (random assignment, natural experiment, or convincing instrument), so reverse causality, omitted variables (e.g., management quality, unobserved product upgrading), and measurement error in LLM-derived AI indicators remain plausible threats to causal interpretation.
Methods Rigormedium — Strong points: integration of large customs and listed-firm datasets, multi-year panel, disaggregation of AI modes ('intelligent empowerment' vs 'machine replacing human'), mechanism exploration and heterogeneity analysis. Concerns: AI exposure measured from annual-report text may capture strategic signaling rather than realized adoption; limited to listed manufacturing firms (selection); potential endogeneity not fully addressed in the description; results end in 2016, predating major post-2016 AI advances.
SampleFirm-level panel combining China Customs trade data (2007–2016) with financial and disclosure data for Shanghai and Shenzhen A-share manufacturing listed companies; AI indicators constructed by extracting and classifying annual report text via large language models; analysis confined to exporting, listed manufacturing firms over 2007–2016.
Themesproductivity innovation adoption org_design
IdentificationAssociational analysis linking firm-level export and financial outcomes (China Customs + Shanghai/Shenzhen A-share manufacturing listed firms, 2007–2016) to AI indicators constructed from annual report text using large language models; likely estimated with panel regressions and controls (time and firm fixed effects, observables) and heterogeneity and mechanism tests, but no clearly stated exogenous source of variation or instrument.
GeneralizabilityRestricted to listed manufacturing firms — excludes private, small, and unlisted exporters., China-specific institutional and export-market context may not transfer to other countries., Study period ends in 2016, before rapid diffusion of recent generative-AI tools, so findings may understate current AI impacts., AI measure based on annual-report language may reflect disclosure behaviour or strategic signalling rather than realized technology adoption., Focus on goods exports; findings may not apply to services or digital exports.
Claims (13)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| AI significantly raises export enterprises' price markup. Firm Revenue | positive | price markup |
Reading fidelity
high
Study strength
medium
|
not reported
|
| The 'intelligent empowerment' model of AI has a stronger promotional effect on price markup than the 'machine replacing human' model. Firm Revenue | positive | price markup |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI increases price markup through efficiency improvement (positive mechanism). Firm Productivity | positive | price markup (via efficiency improvement) |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI increases price markup through technological innovation (positive mechanism). Research Productivity | positive | price markup (via technological innovation) |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI can decrease price markup by intensifying market competition (negative mechanism). Market Structure | negative | price markup (via market competition) |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI can decrease price markup by increasing information transparency (negative mechanism). Market Structure | negative | price markup (via information transparency) |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI adoption yields larger markup increases for state-owned enterprises. Firm Revenue | positive | price markup |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI adoption yields larger markup increases for high-productivity enterprises. Firm Productivity | positive | price markup |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI adoption yields larger markup increases for quality-competitive enterprises. Firm Revenue | positive | price markup |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI adoption yields larger markup increases for enterprises exporting to developed countries. Firm Revenue | positive | price markup |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI reduces markup dispersion across export enterprises. Organizational Efficiency | negative | markup dispersion |
Reading fidelity
high
Study strength
medium
|
not reported
|
| AI improves resource allocation efficiency among export enterprises. Organizational Efficiency | positive | resource allocation efficiency |
Reading fidelity
high
Study strength
medium
|
not reported
|
| Overall, AI helps export enterprises break the low-value-added lock-in, optimises resource allocation, and further boosts China's export economy. Fiscal And Macroeconomic | positive | export economy performance / low-value-added lock-in |
Reading fidelity
high
Study strength
speculative
|
not reported
|