The AI Readiness Paradox: Oil Wealth Offers No Clear Shortcut to Higher Worker Productivity

Research covering 14 economies found that rising readiness scores were associated with lower output per worker during 2021–2024, questioning expectations of an immediate economic payoff from better preparation.

A higher score for artificial intelligence readiness can suggest that a government is building the rules, digital systems, and expertise needed to use the technology, but it cannot show whether workers are producing more. Research covering 14 economies found that rising readiness scores were associated with lower output per worker during 2021–2024, questioning expectations of an immediate economic payoff from better preparation.

The paper, "Government AI Readiness, Resource Dependence, and Labor Productivity: A Comparative Panel Analysis of 14 Post-Soviet and Comparator Economies, 2021–2024," published in the journal Economies, examines whether this relationship changes with resource dependence and government effectiveness. The findings describe changes within countries over a short period, without establishing that improvements in readiness caused productivity to fall.

The distinction between preparing for AI and actually using it is central to understanding the results, because the readiness index measures government capacity and supporting conditions rather than adoption by businesses or workers. A country can improve its strategy or governance score without immediately changing how goods and services are produced across its economy.

Four Years of Evidence Across Uneven Economies

The researchers combined Oxford Insights' Government AI Readiness Index with World Bank economic and governance data, producing 56 country-year observations across Armenia, Azerbaijan, Georgia, Türkiye, Kazakhstan, Uzbekistan, the Kyrgyz Republic, Tajikistan, Estonia, Latvia, Lithuania, Moldova, Romania, and Bulgaria. Eleven were former Soviet republics, with Türkiye, Romania, and Bulgaria providing comparisons across different historical and institutional backgrounds.

Readiness scores covered government strategy and capacity, the technology sector, and data infrastructure, with labour productivity measured as inflation-adjusted GDP per employed person using purchasing power parity. The analysis accounted for persistent differences between countries, changes shared across years, government effectiveness, trade openness, urbanisation, and internet use, using a wild cluster bootstrap to assess uncertainty with so few countries.

The sample was selected for data availability rather than regional representativeness, and its differences were substantial, with productivity ranging from about $15,295 to $98,713 per worker. More productive countries generally had higher readiness scores when compared with one another, making the negative relationship found within countries a different finding from that broader comparison.

Higher Readiness Did Not Bring an Immediate Productivity Lift

At the sample's average oil-rent level, each additional point on the 100-point readiness index was associated with approximately 0.32% lower labour productivity, after the model's adjustments. At zero oil rents, a reference point close to the position of most sampled economies, the association was approximately 0.40% lower productivity per point, or 3.9% lower for a ten-point increase.

The negative association remained when researchers removed each country in turn, excluded both major oil exporters, and restricted the analysis to 2021–2023, indicating that no single country or the final year alone explained it. Adjusting readiness scores for differences between annual editions preserved the finding, although estimates became less precise with a one-year lag, fewer controls, or additional education and inflation controls.

A breakdown of the index showed that the relationship came mainly from its Government component, which captures strategy, governance, public-sector digital capacity, and adaptability, with no clear association for the combined technology-sector and data-infrastructure components. A one-point readiness increase was also associated with roughly 0.26 percentage points lower GDP growth at average oil rents, without establishing relationships with unemployment or high-technology exports; greater trade openness and internet use were linked to higher productivity in the main model.

Oil Revenues Offered No Reliable Productivity Advantage

Oil dependence was concentrated in Azerbaijan and Kazakhstan, where oil rents represented approximately 21% and 14.8% of GDP in 2021, compared with less than 1% in every other sampled economy. These rents describe the economic surplus from oil production rather than government revenue alone, and the researchers used 2021 values because later observations were unavailable in the source database.

The model suggested a more favourable readiness–productivity relationship at higher oil-rent levels, but support for this pattern was weak and disappeared when either Azerbaijan or Kazakhstan was removed. The positive estimates at those countries' rent levels could not be clearly distinguished from zero, leaving no reliable evidence that oil wealth strengthens the productivity benefits associated with readiness.

A model-implied turning point near 13% oil rents also lacked empirical support, because no sampled economy occupied the large gap between roughly 1% and 15%. Government effectiveness did not clearly change the readiness–productivity relationship either, leaving the analysis unable to establish a dependable institutional or resource-based advantage.

Preparation Needs Skills, Implementation, and Better Measurement

One possible explanation is that governments and businesses incur training, organisational, and infrastructure costs before new technologies deliver measurable gains, producing an early dip sometimes described as a productivity J-curve. The researchers did not directly measure those investments or workplace AI use, so the results cannot confirm this explanation or predict whether and when productivity gains will emerge.

Changes in commodity markets, other economic shocks, and revisions to the readiness index offer alternative explanations, and the index's close relationship with national income complicates interpretation further. Its indicators and treatment of missing information changed between editions, meaning that adjustments improve comparability without fully resolving the measurement problem.

The four-year window, small sample, concentration of oil dependence in two countries, and absence of firm-level evidence limit how widely the findings can be applied. Exploratory analyses using national AI-strategy adoption did not establish causation, and additional threshold searches could not identify a credible resource-dependence tipping point.

The paper's policy considerations draw on the wider literature as well as these results, encouraging governments to track workforce skills, organisational changes, data infrastructure, and actual implementation alongside readiness scores. Resource-rich countries would need programme-level evaluation before assuming oil-funded digital projects improve productivity, and longer datasets with consistent measures and workplace evidence would help establish what preparation delivers.