AI expansion in five Asia‑Pacific economies is associated with occupational upgrading—boosting high‑skill jobs while eroding middle‑skill employment—but the magnitude and direction vary significantly across countries.

Augment or Replace? Uneven Labour Market Consequences of AI Expansion Across Asia-Pacific Economies

Seyram Amedede, Nguyen Thuy Ha, YAO Tang, Xiaofen Tan · September 16, 2026 · International Journal of Innovative Science and Research Technology (IJISRT)

openalex correlational low evidence 7/10 relevance Full text usable extracted full text DOI Source PDF

Structured author observations

Linked only from stored provider relations; the raw author line above is never matched by name.

OpenAlex

Latest observation: September 18, 2026

Seyram Amedede provider ID
Nguyen Thuy Ha provider ID
YAO Tang provider ID
Xiaofen Tan provider ID

Semantic Scholar

Latest observation: September 18, 2026

Seyram Amedede unresolved corpus identity
Ha Thuy Nguyen unresolved corpus identity
Yao Tang unresolved corpus identity
Xiao-Fen Tan unresolved corpus identity

Using a PCA-based AI intensity index and country fixed-effects for five Asia-Pacific economies (2010–2025), the paper finds AI intensity is positively associated with high-skill employment shares and negatively associated with middle-skill employment shares, with substantial cross-country heterogeneity.

Citation observations

Cumulative provider counts captured on specific dates; providers are never combined.

0 cumulative citations

OpenAlex · Observed September 18, 2026

View corpus context

0 cumulative citations

Semantic Scholar · Observed September 18, 2026

View corpus context

The rapid diffusion of artificial intelligence (AI) has intensified debates regarding its implications for employment and the future of work. While some scholars argue that AI complements human labour by enhancing productivity and creating new employment opportunities, others contend that it displaces workers through automation and task substitution. This study examines the labour-market consequences of AI across selected Asia-Pacific economies, focusing on whether AI primarily augments or replaces labour. Drawing on the theoretical foundations of skill-biased technological change, routinebiased technological change, and task-based approaches to technological transformation, the study investigates the relationship between AI intensity and occupational employment structures using panel data from China, Japan, South Korea, Hong Kong, and Taiwan over the period 2010-2025. A composite AI Intensity Index is constructed using principal component analysis, while fixed-effects panel regression models are employed to assess the effects of AI development on high-skill, middle-skill, and low-skill employment shares. The findings indicate that AI development is positively associated with high-skill employment but negatively associated with middle-skill employment, suggesting that AI simultaneously generates labour augmentation and labour displacement effects. These relationships remain robust across alternative specifications, including lagged AI measures and leave-one-country-out estimations. The results further reveal significant cross-country heterogeneity, indicating that the labour-market consequences of AI are shaped by differences in institutional arrangements, innovation capacity, and economic structures, while robustness analyses confirm that the findings are not driven by any single economy. The study contributes to the emerging literature on AI and employment by demonstrating that the effects of AI are neither uniformly beneficial nor uniformly disruptive but are instead context-dependent and manifested through processes of occupational restructuring. The findings underline the importance of adopting nuanced and comparative perspectives when evaluating the implications of AI for labour markets and workforce transformation.

Summary

Main Finding

AI expansion across five advanced Asia–Pacific economies (China, Japan, South Korea, Hong Kong, Taiwan; 2010–2025) is associated with simultaneous labour-market augmentation and displacement: higher AI intensity correlates with larger shares of high‑skill employment and smaller shares of middle‑skill (routine‑intensive) employment. Effects vary substantially across countries, indicating context dependence.

Key Points

Research question: Does AI development augment or replace labour across Asia–Pacific economies, and how does it affect occupational structure?
Theoretical framing: draws on Skill‑Biased Technological Change (SBTC), Routine‑Biased Technological Change (RBTC), and task‑based approaches; expects both complementarities with high‑skill labour and substitution of routine middle‑skill jobs.
Main empirical findings:
- Positive relationship between AI intensity and share of high‑skill employment (supports H1).
- Negative relationship between AI intensity and share of middle‑skill (routine‑intensive) employment (supports H2).
- Significant cross‑country heterogeneity in effects (supports H3); results robust to lagged AI measures and leave‑one‑country‑out checks.
Contributions:
- Extends empirical AI–employment work beyond North America/Europe to Asia‑Pacific.
- Constructs a multidimensional AI Intensity Index (patents, publications, VC, AI hiring).
- Provides comparative evidence that AI simultaneously augments and displaces labour through occupational restructuring.

Data & Methods

Sample: Panel of five economies — China, Japan, South Korea, Hong Kong, Taiwan — annual country‑level data, 2010–2025.
Dependent variables: employment shares by skill category
- High‑skill employment share (HSE)
- Middle‑skill employment share (MSE)
- Low‑skill employment share (LSE)
Main independent variable: AI Intensity Index
- Components: AI patents, AI research publications, AI venture capital investment, AI‑related hiring activity.
- Composite index constructed via Principal Component Analysis (PCA).
Controls: GDP, national R&D expenditure, industrial‑structure indicators (sectoral composition), and other macro controls to isolate AI effects.
Estimation strategy:
- Fixed‑effects panel regressions (country and/or year FE) to exploit within‑country variation over time.
- Robustness checks: lagged AI measures (to address timing), leave‑one‑country‑out estimations (to ensure results not driven by any single country).
- Heterogeneity analysis across countries to assess context dependence.
Conceptual approach: task‑based lens linking AI intensity to occupational upgrading (complementarity) and routine‑task displacement (substitution).

Implications for AI Economics

Measurement: Multidimensional measures of AI activity (innovation, publications, investment, hiring) provide richer signals than single proxies; future empirical work should adopt composite indices or multiple indicators to capture AI intensity more fully.
Distributional effects: AI does not produce uniform labour gains or losses — it reallocates employment toward high‑skill occupations while compressing middle‑skill routine jobs. Economic models of AI should incorporate simultaneous complementarity and substitution mechanisms (task heterogeneity).
Policy relevance: Findings emphasize the need for targeted labour‑market policies (reskilling/upskilling, occupational mobility support, education aligned with non‑routine cognitive skills) and institutional arrangements that shape how countries capture AI complementarities.
Comparative research: Cross‑country heterogeneity underscores that institutional context, innovation capacity, and industrial composition critically mediate AI’s labour effects; future studies should prioritize comparative and country‑specific mechanisms (e.g., regulation, labour market flexibility, firm organization).
Research agenda: Strong case for micro‑level analyses (firm and task data), causal identification of adoption versus local AI activity spillovers, and exploration of mitigation policies (training programs, social insurance) to manage occupational transitions.

Assessment

Paper Typecorrelational Evidence Strengthlow — Observational correlational design with a very small number of cross-sectional units (five countries) limits statistical power and external validity; while fixed effects and robustness checks reduce some confounding, no credible exogenous variation or instrumental strategy is used to establish causality and measurement/error and omitted-variable bias remain plausible. Methods Rigormedium — The authors use standard and appropriate tools (PCA to construct a multidimensional AI intensity index; fixed-effects panel regressions; plausible control variables; robustness checks with lags and leave-one-out), but the small number of countries, potential endogeneity of AI measures, aggregate occupational outcomes, and absence of stronger identification strategies (instruments, natural experiments, difference-in-differences exploiting policy or adoption shocks) weaken rigor. SampleCountry-year panel of five advanced Asia-Pacific economies (China, Japan, Republic of Korea, Hong Kong, Taiwan) from 2010 to 2025; dependent variables are employment shares by skill category (high-, middle-, low-skill) drawn from harmonised occupational statistics; main independent variables are AI-related patents, research publications, venture-capital investment, and AI-related hiring combined into a PCA-based AI Intensity Index; controls include GDP, national R&D expenditure, and sectoral/industrial structure indicators. Themeslabor_markets innovation IdentificationPanel fixed-effects regressions on a country-year panel (China, Japan, South Korea, Hong Kong, Taiwan; 2010–2025) using a PCA-constructed composite AI Intensity Index (patents, publications, VC, AI hiring) as the main explanatory variable, with controls (GDP, R&D, industrial structure); robustness checks reported include lagged AI measures and leave-one-country-out estimations. No exogenous variation, instruments, or quasi-experimental design are reported. GeneralizabilityOnly five technologically advanced Asia-Pacific economies—findings may not generalise to other countries (low-income or less AI-intensive economies)., Country-level aggregates (occupation shares) mask within-occupation/task heterogeneity and firm- or industry-level variation., Small cross-sectional sample (N=5) limits robustness to outliers and reduces external validity., Potential measurement error in composite AI intensity (choice and weighting of indicators) and in mapping occupations to skill categories., Observational design limits causal interpretation—results may reflect correlated economic trends or reverse causality.

Claims (5)

Claim	Direction	Outcome	Confidence & Evidence	Details
AI development is positively associated with the share of high-skill employment in the selected Asia-Pacific economies. Employment	positive	Share of high-skill employment	Reading fidelity high Study strength medium	n=80 0.3
AI development is negatively associated with the share of middle-skill employment. Employment	negative	Share of middle-skill employment	Reading fidelity high Study strength medium	n=80 0.3
The study finds simultaneous labour augmentation and labour displacement effects associated with AI development. Job Displacement	mixed	Occupational employment structure by skill group	Reading fidelity high Study strength medium	n=80 0.3
The estimated relationships remain robust when lagged AI measures are used and when the analysis excludes one country at a time. Employment	positive	Association between AI intensity and occupational employment shares	Reading fidelity high Study strength medium	n=80 0.3
The relationship between AI development and labour-market outcomes varies across the five Asia-Pacific economies studied. Employment	mixed	Cross-country variation in occupational employment outcomes associated with AI development	Reading fidelity high Study strength medium	n=80 0.3