The Economics of Compliance: Measuring Market and Innovation Effects of the EU Artificial Intelligence Act's High-Risk Classification System

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This dissertation measures the early economic footprint of the European Union's Artificial Intelligence Act (Regulation [EU] 2024/1689), the first comprehensive, risk-tiered statute to impose binding obligations on artificial intelligence systems. The problem it addresses is that scholarly and policy debate over the Act has been overwhelmingly legal and normative, leaving the economic magnitude of its high-risk compliance obligations—documentation, risk management, conformity assessment, and human oversight for Annex III use cases—essentially unquantified even as enforcement began. The purpose of the study is therefore to produce the first firm-level econometric evidence on whether these obligations raise compliance costs, concentrate markets, provoke lobbying and arbitrage, and reshape innovation. Using only open-access sources, the study assembled a panel of 800 publicly traded, artificial-intelligence-active firms observed over fiscal years 2018 through 2026, yielding 5,816 firm-years, and combined two-way fixed-effects difference-in-differences and triple-difference models with an event study of equity-market reactions to eight legislative and enforcement milestones. The findings are deliberately mixed and are reported with their fragility intact. The first hypothesis, that compliance costs rise for directly regulated firms, is partially supported but fragile: operating-expense intensity rose among European Union–exposed Annex III firms (δ = 4.40, p = .044), but the estimate does not survive the Holm correction (p = .156) or permutation inference (p = .264) and rests on only 27 treated firms. The second hypothesis, that concentration rises, is not supported; the Herfindahl–Hirschman Index coefficient carries the opposite sign to the prediction (β = −839.1). Lobbying intensified markedly (filings rose 114% from 2019 to 2025), while revenue arbitrage and innovation effects were not detected. The event study was the most precise evidence, with cumulative abnormal returns of −2.61% around the final parliamentary vote and −2.81% around entry into force. The study implies that the Act's early incidence is narrow, diffuse, and best monitored as the post-enforcement window lengthens.