The Commonplace
Home Papers Evidence Explore Trends Syntheses Digests About 🎲 Workforce Futures
← Papers
Direction, evidence grade, and study type are AI-generated labels (gpt-5-mini), not human-verified. Syntheses are LLM-written. "Tensions" are machine-detected candidates, not confirmed contradictions. A research-acceleration tool, not peer review. How this is built →

Innovation and productivity — not AI adoption — predict export success among Vietnam's manufacturing SMEs; technology purchases alone do not translate into stronger export performance.

Internal capabilities, digital transformation, and SME export performance: Evidence from Vietnam’s manufacturing industries
Dinh Thi Mung, Tran Quang Minh · June 15, 2026 · Problems and Perspectives in Management
openalex correlational medium evidence 7/10 relevance Full text usable extracted full text DOI Source PDF
Using a 2015–2023 industry-level panel for Vietnam's manufacturing SMEs, the study finds that innovation activity and higher labor productivity are positively associated with export performance, while digital transformation, AI adoption, and FDI show no significant direct effects.

Type of the article: Research ArticleAbstractExport upgrading has become a pressing concern for small and medium-sized enterprises (SMEs) in emerging economies. Digital transformation and artificial intelligence (AI) are often presented as fast-track solutions. Yet evidence on whether these technologies actually improve export performance remains inconclusive. In many industries, technology adoption does not automatically translate into stronger foreign market outcomes, especially where internal capabilities differ substantially.The purpose of this study is to examine how digital transformation, AI adoption, innovation activity, labor productivity, and foreign direct investment (FDI) relate to SME export performance in Vietnam’s manufacturing sector over the period 2015–2023. Using industry-level panel data (63 observations) and pooled OLS and fixed-effects estimations, the analysis evaluates both internal capability factors and external structural influences.The results reveal a clear pattern. Innovation is positively associated with export performance (β = 45.61, p < 0.01), and labor productivity exerts a significant positive effect (β = 24.57, p < 0.05). By contrast, digital transformation, AI adoption, and FDI do not display statistically significant direct effects in the baseline specification. The explanatory power of the model remains substantial (R² between 0.642 and 0.701), suggesting that capability-related factors account for a meaningful share of export variation across industries.Taken together, the findings indicate that technology adoption alone is insufficient. In practice, SMEs cannot simply invest in digital tools and expect immediate export gains. Sustained innovation efforts and productivity-enhancing routines appear to be the more decisive foundations of export competitiveness.

Summary

Main Finding

Industry-level evidence from Vietnam’s manufacturing SMEs (2015–2023) shows that internal capability factors — innovation and labor productivity — are positively and significantly associated with export performance, while digital transformation, AI adoption, and FDI show no statistically significant direct effects in the baseline models. The authors conclude that technology adoption alone is insufficient to boost SME exports without complementary capability building.

Key Points

  • Sample & scope: Industry-level panel of Vietnamese manufacturing industries (2015–2023), 63 observations; focus on SME-intensive sectors.
  • Main explanatory variables: innovation (share of firms with product/process/business-model innovation), labor productivity (output per worker), digital transformation (industry investment in digitalization/automation), AI adoption (share of firms using AI applications), and FDI (industry-level foreign investment).
  • Main estimates (pooled OLS / fixed-effects; models control for industry and year effects):
    • Innovation: β = 45.61, p < 0.01 — positive and highly significant.
    • Labor productivity: β = 24.57, p < 0.05 — positive and significant.
    • Digital transformation: not statistically significant in baseline.
    • AI adoption: not statistically significant in baseline.
    • FDI: not statistically significant in baseline.
  • Model fit: R² between 0.642 and 0.701, indicating substantial explanatory power from capability-related variables.
  • Interpretation: Adoption of digital tools or AI does not automatically translate into higher exports; internal absorptive capacity (innovation routines, productivity-enhancing practices) appears decisive.
  • Limitations acknowledged by authors: industry-level (not firm-level) data, potential endogeneity / reverse causality (learning-by-exporting), restricted to manufacturing, possible measurement noise in proxies (especially for AI and digital transformation), relatively small sample (63 obs).

Data & Methods

  • Data sources: Vietnam Statistical Yearbooks, National Innovation Center reports, World Bank databases; processed dataset available on request.
  • Unit of analysis: manufacturing industries (Vietnam’s official classification).
  • Timeframe: 2015–2023, chosen to cover pre/post digital policy phases.
  • Variables:
    • Dependent: annual export value by industry.
    • Key independent variables: innovation (share of innovating firms), labor productivity (output per worker), digital transformation (industry investment in digitalization/automation), AI adoption (proportion of firms implementing AI), FDI (industry-level realized foreign investment).
  • Empirical approach: panel regressions (pooled OLS, fixed-effects, random-effects). Industry and year fixed effects included to control for time-invariant heterogeneity and common shocks.
  • Diagnostics: descriptive stats, correlation checks, multicollinearity and heteroskedasticity tests; authors caution on remaining endogeneity and interpret results as associations rather than definitive causal estimates.

Implications for AI Economics

  • Complementarities matter: The findings reinforce that AI/digital adoption needs complementary internal capabilities (innovation routines, skilled labor, managerial systems) to affect export outcomes. AI should be studied as part of a bundle of complementary investments, not in isolation.
  • Heterogeneous and conditional effects: Null average effects for AI and digitalization point to likely heterogeneity — positive effects may exist for industries or firms with sufficient absorptive capacity. Future AI-economics work should model and test interactions (AI × innovation, AI × skills, AI × firm size).
  • Measurement and data recommendations:
    • Move to firm-level panel data to capture within-industry heterogeneity and allow causal designs.
    • Use richer AI adoption measures (intensity, types of AI, use-cases) and time lags to capture delayed productivity/export effects.
    • Track intermediate outcomes (product quality upgrades, entry to new markets, productivity growth, mark-ups) to illuminate mechanisms.
  • Identification strategies: To establish causality, employ quasi-experimental methods (instrumental variables, difference-in-differences exploiting exogenous policy rollout or infrastructure shocks, event studies), or randomized interventions for capability-building programs coupled with AI deployment.
  • Policy implications relevant to AI deployment:
    • Prioritize complementary investments (training, innovation grants, managerial support) alongside subsidies for digital/AI tools.
    • Targeted programs for SMEs to build absorptive capacity (innovation systems, process improvement) may be more effective for export upgrading than blanket digitalization incentives.
    • Encourage linkages that facilitate effective technology transfer (structured FDI spillovers, supplier development programs) rather than assuming passive benefits from foreign presence.
  • Research agenda: quantify heterogeneity of AI effects across firm types and industries; study dynamic paths (how innovation → productivity → exports unfold over time after AI adoption); evaluate cost-effectiveness of combined AI + capability-building interventions.

Summary takeaway: For AI economics, this study is a reminder that observed null average effects of AI/digital adoption on exports can reflect missing complementarities and heterogeneity; rigorous microdata and causal designs that account for absorptive capacity are needed to assess when and how AI raises international competitiveness.

Assessment

Paper Typecorrelational Evidence Strengthmedium — Findings are based on panel regressions with fixed effects, which provide credible associations and control for time-invariant industry heterogeneity, but the study lacks a clear causal identification strategy (e.g., instruments or plausibly exogenous shocks), uses a small aggregated sample (63 observations), and may be vulnerable to time-varying confounders, measurement error in AI adoption, and reverse causality. Methods Rigormedium — Appropriate baseline econometric techniques (pooled OLS and fixed effects) are used and model fit is reported, but the analysis does not address endogeneity beyond fixed effects, provides no robustness checks or alternative identification (based on the abstract), and works with a small number of industry-year observations which limits statistical power and inference. SampleIndustry-level panel data for Vietnam's manufacturing sector covering 2015–2023 (63 industry-year observations), using aggregated measures of SME export performance, digital transformation, AI adoption, innovation activity, labor productivity, and FDI; data sources and variable construction are not detailed in the abstract. Themesinnovation adoption productivity IdentificationIndustry-level panel (Vietnam manufacturing, 2015–2023) analyzed with pooled OLS and industry fixed-effects estimators to control for time-invariant heterogeneity; no instrumental variables, natural experiment, or other exogenous variation exploited for causal identification. GeneralizabilitySingle-country (Vietnam) — may not transfer to other institutional or market contexts, Manufacturing sector only — excludes services and other sectors where AI effects differ, SME-focused / industry-level aggregation — ecological aggregation may mask firm-level heterogeneity, Short/limited panel (63 observations) — limited temporal and cross-sectional variation, Observational design and possible measurement error in AI adoption limit external validity

Claims (6)

ClaimDirectionOutcomeConfidence & EvidenceDetails
Innovation is positively associated with export performance (β = 45.61, p < 0.01). Firm Revenue positive export performance
Reading fidelity high
Study strength medium
n=63
β = 45.61, p < 0.01
0.3
Labor productivity exerts a significant positive effect on export performance (β = 24.57, p < 0.05). Firm Revenue positive export performance
Reading fidelity high
Study strength medium
n=63
β = 24.57, p < 0.05
0.3
Digital transformation, AI adoption, and foreign direct investment (FDI) do not display statistically significant direct effects on export performance in the baseline specification. Firm Revenue null_result export performance
Reading fidelity high
Study strength medium
n=63
0.3
The explanatory power of the model is substantial (R² between 0.642 and 0.701), suggesting capability-related factors account for a meaningful share of export variation across industries. Firm Revenue positive model explanatory power for export performance variation
Reading fidelity high
Study strength medium
n=63
R² between 0.642 and 0.701
0.3
Using industry-level panel data (63 observations) and pooled OLS and fixed-effects estimations, the analysis evaluates internal capability factors and external structural influences for Vietnam's manufacturing SMEs over 2015–2023. Other positive study methodology / analytic approach
Reading fidelity high
Study strength high
n=63
0.5
Technology adoption alone is insufficient for improving SME export performance; sustained innovation efforts and productivity-enhancing routines are the more decisive foundations of export competitiveness. Firm Revenue positive export performance (policy/practice implication)
Reading fidelity high
Study strength speculative
n=63
0.05

Notes