New research on the macroeconomic implications of advances in generative artificial intelligence (AI) concludes that the impact on GDP is likely to be modest. What’s more, according to the study by Daron Acemoglu of MIT prepared for the journal Economic Policy, there is no evidence that AI will reduce labour income inequality. Indeed, some groups, notably low-education white, native-born women, are predicted to experience small declines in real wages.

The study begins by noting the rapid spread of ChatGPT, with an estimated 100 million monthly users only two months after launch, and forecasts from some quarters of huge economic gains from new advances in AI even within the next 10 years.

More modestly, Goldman Sachs predicts a 7% increase in global GDP, equivalent to $7 trillion over ten years, and 1.5 percentage point per annum faster economic growth in the United States. McKinsey Global Institute forecasts that the total impact of AI and other automation technologies could produce a 1.5 to 3.4 percentage point rise in average annual growth over the same horizon.

Acemoglu’s research evaluates these claims, starting from a task-based model of AI’s effects, working through automation and task complementarities. It establishes that as long as AI’s microeconomic effects are driven by cost savings and productivity improvements at the task level, its macroeconomic consequences for GDP and aggregate productivity gains can be estimated by what fraction of tasks are affected and average task-level cost savings.

Using existing estimates on exposure to AI and productivity improvements at the task level, these macroeconomic effects appear non-trivial but modest – no more than a 0.71% increase in over ten years.

But even these estimates could be exaggerated, because early evidence is from ‘easy-to-learn’ tasks, whereas some of the future effects will come from ‘hard-to-learn’ tasks, where there are many context-dependent factors affecting decision-making and no objective outcome measures from which to learn successful performance. As an example, how to boil an egg or write computer subroutines are easy for outside learning; problems like diagnosing the cause of a persistent cough are much harder.

Taking this key distinction into account, predicted gains in total factor productivity over the next ten years are even more modest – less than 0.55%.

The research then explores the wage and inequality effects of generative AI, concluding that even if it increases productivity in various tasks, these are unlikely to translate into higher wages or lower inequality. Rather, more favourable wage and inequality effects will depend on the creation of new tasks for workers in general and, more specifically, for middle and low-pay workers.

Empirically, the study finds that AI advances are unlikely to increase inequality as much as previous automation technologies because their impact is more equally distributed across demographic groups. But there is also no evidence that AI will reduce labour income inequality – and low-education white, native-born women, are predicted to experience small declines in real wages. AI is also predicted to widen the gap between capital and labour income.

Finally, the research notes, it is important to consider the negative social value of some new AI tasks – such as misinformation, deep fakes and manipulative advertising – and their potential macroeconomic effects.

---

The Simple Macroeconomics of AI

Authors:

Daron Acemoglu (MIT)