(Co-authored with Lekha Chakraborty)

Artificial intelligence (AI) is rapidly emerging as a general-purpose technology with the potential to drive significant productivity growth and reshape labour markets worldwide. Yet, as historical technological shifts have shown, its benefits are often unevenly distributed—both across countries and within them. In G20 economies, nations that are structurally better prepared for AI adoption are already experiencing stronger labour productivity gains. However, in the early phases of diffusion, these same countries tend to exhibit wider gender wage gaps, highlighting the risk that AI could exacerbate existing inequalities without deliberate policy intervention.

Traditional indicators of AI progress, such as patent registrations or venture capital inflows, provide valuable but incomplete insights. They focus primarily on innovation and commercialisation but overlook the macro-structural complementarities that determine how effectively countries can absorb and scale AI technologies. To address this gap, our study constructs a comprehensive AI Composite Index that captures the foundational capabilities enabling productive AI integration (Dubey and Chakraborty,2026).

The index aggregates five equally weighted indicators, each normalised annually on a 0-1 min-max scale to ensure cross-country comparability and focus on relative positions over time. These components are: AI-related patents per million population, sourced from the World Intellectual Property Organisation, reflecting a country’s position at the innovation frontier; venture capital investment in AI as a share of GDP, from OECD data, capturing private-sector confidence and commercialization efforts; gross capital formation as a percentage of GDP, from the World Bank, proxying the physical infrastructure—such as data centres and computing hardware—complementary to AI deployment; public expenditure on education as a share of GDP, also from the World Bank, measuring human capital essential for leveraging AI in knowledge-intensive sectors; and internet users as a percentage of the population, again World Bank data, indicating the maturity of digital infrastructure for data transmission and broad-based participation in the digital economy.

This multidimensional approach reveals substantial and persistent heterogeneity across the G20. Frontier economies like the Republic of Korea consistently lead, benefiting from sustained investments across all dimensions—particularly research intensity, education, and connectivity. The United States maintains a robust second position, while China has demonstrated impressive upward trajectory since the mid-2010s, rapidly closing gaps in several areas though not yet matching leaders in all complementarities. Advanced economies including the United Kingdom, Japan, and Germany form a middle cluster with similar scores. Emerging markets generally lag, with countries like India and Indonesia constrained by lower education spending, capital deepening, and digital penetration.

Over the 2012–2023 period, the G20 average index rose gradually from around 0.36 to 0.43, reflecting diffuse improvements in digital infrastructure and human capital accumulation. However, dispersion remained wide, and convergence limited. Leading nations continued to extend their advantages, while middle- and lower-ranked economies progressed at comparable paces. These patterns align with stylized facts from individual components: AI venture capital remains overwhelmingly concentrated in the United States and China, with rapid but late surges in some emerging markets; patent activity is dominated by technological pioneers; and complementary investments show uneven diffusion.

The empirical analysis links this preparedness measure to macroeconomic and distributional outcomes using panel regressions with country and year fixed effects. A comprehensive set of controls—trade openness, unemployment rates, private credit to GDP, services sector employment share, fiscal balances, broadband subscriptions, GDP per capita, and female labour force participation—addresses potential omitted variable bias and endogeneity concerns. The index’s focus on pre-existing structural factors further mitigates reverse causality risks.

Results show a robust positive association between AI preparedness and labour productivity, measured as log output per worker. The coefficient remains statistically significant across specifications, even in saturated models, ranging from approximately 0.39 to 0.73. Services sector employment emerges as a key mediating channel, consistent with evidence that AI adoption has been most extensive in knowledge-intensive services. These findings indicate that productivity gains from AI are conditional on strong complementarities: economies with deeper human capital, digital networks, and investment climates are best positioned to translate technological potential into aggregate growth. This complements emerging evidence on AI’s role in human capital formation across G20 economies (Dubey and Chakraborty, forthcoming).

More cautionary are the distributional implications, particularly for gender equity. Higher AI preparedness is associated with wider gender wage gaps during initial diffusion stages, with a coefficient significant at the 10 percent level even after extensive controls. Female labour force participation enters positively and significantly, suggesting that greater workforce inclusion does not automatically yield pay parity and may reflect occupational segregation or unequal access to AI-complementary high-skill roles.

This pattern echoes historical experience with technological change, where new general-purpose technologies initially reward skills unevenly distributed across genders (Goldin, 2014). It also underscores that women’s empowerment—intrinsically linked to broader development—requires targeted policies alongside technological progress (Duflo, 2012). Without intervention, AI risks amplifying rather than ameliorating systemic wage disparities.

The policy implications are clear and urgent. For emerging G20 economies, narrowing preparedness gaps demands prioritised public investment in education, broadband infrastructure, and capital formation. Fiscal space for such spending can be created through gender budgeting and efficient resource mobilisation. Frontier economies, meanwhile, must complement their structural advantages with inclusive measures: reskilling programmes prioritising women, expanded STEM education for girls, affordable care economy infrastructure to support female labour force participation, and regulations addressing occupational segregation in AI-related fields.

Continuous monitoring of gender-disaggregated labour market indicators is essential as AI diffuses further. Early evidence of widening gaps should trigger corrective actions—such as affirmative hiring quotas or wage transparency mandates—before inequalities become entrenched.

AI represents a historic opportunity to elevate living standards and advance human development across the globe. Yet the G20 experience demonstrates that this potential is neither automatic nor equitable by default. Countries that proactively build structural foundations while embedding inclusivity in their AI strategies are most likely to achieve shared prosperity. The alternative risks a future of diverging outcomes, where productivity surges for some while others are left behind.

As policymakers navigate this transformative era, the empirical evidence calls for balanced approaches: harnessing AI’s growth dividend while safeguarding against its distributional risks. The choices made today will shape whether AI becomes a force for convergence or divergence in human development.


References

Dubey, Rohan and Lekha Chakraborty (2026). “Measuring Artifical Intelligence”, NIPFP Blog, National Institute of Public Finance and Policy (NIPFP), New Delhi.

Dubey, Rohan and Lekha Chakraborty (forthcoming). “AI Preparedness in G20 Economies and Labour Productivity: An Empirical Analysis.” mimeo, National Institute of Public Finance and Policy (NIPFP), New Delhi.

Duflo, Esther (2012). “Women Empowerment and Economic Development.” Journal of Economic Literature, 50(4): 1051–1079.

Goldin, Claudia (2014). “A Grand Gender Convergence: Its Last Chapter.” American Economic Review, 104(4): 1091–1119.