Readiness, Not Restriction: Firm Size, Digital Infrastructure, And the Conditional Effect of Regulatory Intensity on Enterprise AI Adoption in the European Union
TL;DR
Results indicate that enterprise-level characteristics remain important even when national regulatory conditions are broadly comparable, and suggest that policies aimed at strengthening enterprise capacity and digital infrastructure are more likely to narrow this gap.
Abstract
This paper examines why European Union member states with broadly similar regulatory environments show large differences in enterprise adoption of artificial intelligence (AI). Using a ten-country sample drawn from Eurostat's Community Survey on ICT Usage in Enterprises (2023–2025) and the OECD Digital Services Trade Restrictiveness Index (Digital STRI), we test whether firm size, digital infrastructure readiness, and regulatory intensity explain differences in AI adoption across countries and firm-size classes. An initial specification, built by pooling nine conceptually distinct Eurostat AI-use indicators, several measuring non-adoption rather than adoption, and two incompatible denominators into a single dependent variable, produced results that could not be interpreted as measuring AI adoption. After restricting the sample to a single, comparably defined indicator, the usable sample fell from 731 to 65 observations. Because the corrected sample contains only ten country clusters, we cross-check conventional cluster-robust inference against a t-distribution correction and a wild cluster bootstrap using the full population of sign-flip permutations. Under this stricter inference, small and medium enterprises show a robust adoption gap of 10 to 14 percentage points relative to large enterprises, while the interaction between regulatory intensity and firm size remains significant: the small-firm adoption gap narrows as regulatory intensity rises. Basic cloud computing readiness has a large positive point estimate but does not survive the wild cluster bootstrap. We report this discrepancy explicitly and document the data and specification corrections underlying the final analysis. These results further indicate that enterprise-level characteristics remain important even when national regulatory conditions are broadly comparable. The findings suggest that policies aimed at strengthening enterprise capacity and digital infrastructure, rather than adjusting the general level of digital regulation, are more likely to narrow this gap.