Economy
AI Will Reshape Economics Research, But Not Evenly
By Bear Creek Staff, . Bear Creek Capital Network.
Ingar Haaland, writing in a post published by Marginal Revolution, has laid out a detailed forecast for how artificial intelligence will restructure economics and the social sciences, and the picture is sharply uneven depending on where a researcher sits in the profession.
At the top, Haaland expects little disruption and considerable gain. Elite research will remain human-led, with AI deployed for scale: running qualitative interviews across large populations, delivering behavioral interventions in the field, and processing large text datasets. The work gets better; the researcher stays in charge.
Field experiments, in his view, rise in relative value. So does the ability to build relationships with firms and run real-world tests, including tests of AI pipelines themselves. The premium shifts toward people who can get things done outside a university computer.
Administrative data research also becomes more valuable, but Haaland flags a concentration problem. The gains may flow mainly to a small group of established authors who have already negotiated access to government agencies and can run local models on sensitive data at scale. For PhD students, he is blunt: they can probably forget about it.
The most striking upside he identifies is in economic history. Haaland predicts "an enormous race and big rewards for digitizing archives" as a source of both new research strategies and new descriptive facts. That is a field where the bottleneck has always been the cost of turning physical records into usable data, and AI substantially lowers that cost.
Review and verification systems, he adds, will be transformed. AI-assisted checking of papers, code and data at the submission stage, rather than after acceptance, is likely to become standard.
The hardest landing, in his forecast, is for research that can already be largely automated: downloading and analyzing public data, running simple survey experiments, writing theory models. Haaland describes this as work that will be "outsourced to public agencies who can answer their own questions without the need for 'peer-reviewed' research." For researchers whose output sits in that category, the competitive pressure is not gradual.
The framing Marginal Revolution chose to introduce the post is itself a signal: "With math essentially being delegated to OpenAI" is the premise, not the conclusion. That is a significant assumption, and the forecast rests on it holding.
What Haaland's predictions do not address is timing. The access asymmetries he describes, between established authors and PhD students, between institutions with government data agreements and those without, already exist. AI may accelerate them rather than create them, which would make the distributional effects arrive faster than the profession's hiring and funding structures can adjust.
For investors and analysts who rely on academic economics for forecasting frameworks or policy signals, the more immediate question is whether the verification improvements he describes will actually raise the reliability of published findings, or simply raise the speed at which unreliable findings get produced and checked. That remains an open question, and no current reporting settles it.
Bear Creek Capital Network publishes content for informational and educational purposes only. Nothing in this article constitutes financial, investment, tax or legal advice, or a recommendation to buy or sell any security. Consider your own circumstances and consult a qualified professional before making financial decisions.