Recovering Hidden Compliance Costs from Legal Text and Censored Data

Official statistics often miss how much regulation really costs. In Germany’s OnDEA database, any rule with estimated annual compliance costs below €1 million is recorded as zero, effectively erasing a large number of small but meaningful obligations. Previous research on regulatory burdens has focused on aggregate totals and rule counts but has largely ignored how data conventions, such as cost censoring, systematically hide the everyday burdens of compliance. This omission leaves a major blind spot in the study of regulatory governance and the politics of measurement. This paper addresses that gap by reconstructing the missing costs for businesses, public administrations, and citizens using a model explicitly designed to handle censored data. Combining random forests with an inverse hyperbolic sine (IHS) transformation, the approach estimates plausible values for omitted entries while preserving the overall structure of the dataset. The findings show that regulations once recorded as costless in official reporting actually impose substantial hidden burdens across society. Most arise from one-time administrative or technical requirements such as registrations, certifications, and IT system upgrades. While some of these measures implement EU directives, the majority are rooted in national legislation. Together, they reveal how data conventions and measurement practices can obscure the everyday impact of regulation. By integrating modern machine-learning imputation with administrative cost accounting, this study provides both a methodological and empirical contribution. It offers a replicable framework for recovering hidden compliance costs. It also highlights how decisions about data architecture and measurement shape our understanding of the state’s regulatory footprint and, ultimately, define what counts as burden in governance.