Artificial Intelligence at the crossroads: Productivity J-curve and carbon trade-offs in the global economy

Soumyajit Bhunia, Sneha Dey, Priya Mondal · 22 Jun 2026

Where this sits in the debate

Will AI actually boost productivity and growth? · its finding: it depends (why studies differ)
Will AI widen inequality? · its finding supports the claim (why studies differ)

Direction of the finding as read from the abstract by an AI model (AI-written).

Key takeaway AI-written

Using an eight-year OECD country panel and a seemingly-unrelated-regressions approach, finds support for the AI productivity J-curve (slow initial gains from AI patents and robot installations followed by catch-up as organizations adapt), while also showing democratic political systems tend to delay environmental regulation in favor of near-term growth. It links two normally separate strands of the AI debate — productivity diffusion lags and AI's environmental/carbon trade-offs — in one empirical panel.

The paper in three steps AI-written

Written by an AI agent from the paper's abstract, which is the primary source and is quoted below.

The question

Does AI adoption follow a productivity J-curve, with slow gains at first and faster gains once organisations adapt?

The approach

A panel of selected OECD countries from 2013 to 2020, linking AI patents and robot installations to total factor productivity, and political factors to carbon emissions, estimated with seemingly unrelated regressions.

The findings

The results are consistent with a productivity J-curve: early AI adoption goes with slower productivity growth, and gains catch up as organisational changes follow. Democratic systems are associated with delays in strict environmental regulation.

How to read it: observational. Analysis of real-world firm, worker, or market data — realistic settings; causal claims depend on the identification strategy.

Limitations and counterpoints AI-written

Eight years and a selection of OECD countries; AI is measured through patents and robot installations, and the abstract reports associations rather than causal estimates.

Abstract

Artificial intelligence functions as an engine for increasing the productivity and economic growth of a nation. At the same time, it creates economic inequalities. The rapid development of the AI sector has profoundly transformed global systems, fundamentally altered operational paradigms, and paved novel pathways for economic growth. This study is trying to examine the intricate relationship between AI and economic growth, particularly on both the productivity gains and economic inequalities associated with AI adoption. We frame this scenario through the lens of productivity J-curve analysis. Using a panel of selected OECD countries spanning a period of eight years (2013-2020), the present study seeks to shed some light to examine the association among technological, political, and environmental factors, particularly AI patents and robot installation and their effects on productivity, measured by Total Factor Productivity (TFP). Furthermore, this study also focuses on the complex relationship between political factors, specifically democracy and political stability, and their impact on environmental degradation, measured by carbon emissions. To assess this relationship, we apply a novel econometric technique, the Seemingly Unrelated Regression Equation (SURE) estimation approach. Findings from our empirical study support the productivity J-curve hypothesis, where AI adoption at the early phase exhibits slower productivity growth, and the organizational changes catch up over time. It highlights the necessity for sustained organizational adaptation, realizing the full productivity gain of emerging technologies. Democratic systems frequently face delays in enacting and enforcing stringent environmental regulations due to political negotiations. Governments may prioritize economic development, often postponing environmental regulations to boost economic growth.

About this publication

Venue not recorded. Publication venue not recorded by our data source — check the paper's landing page for where it appeared.

Published: 2026 (4 months ago)
Citations: 0 (OpenAlex)