Threshold Reporting and Censoring in Regulatory Cost Indicators: Evidence from the German OnDEA Database

Aggregate compliance-cost figures from regulatory impact assessment (RIA) systems are routinely cited as measures of regulatory burden, but they are produced under reporting rules that condition the depth of cost estimation on a relevance threshold. Using Germany’s OnDEA database (2006–2025; 762 unique federal regulations, 29,440 Vorgaben), this paper documents two consequences of that institutional design. First, threshold-based reporting censors the lower tail of the cost distribution: an interval-censored regression with ministry and policy area fixed e!ects implies that published aggregates understate a reconstructed counterpart by approximately 11.6% for business regulations and 21.2% for public administration — a consolidated €14.2 billion (13.1%) understatement on a published total of €108.4 billion. Second, the same model is used to test whether the censoring mass conceals “artificially downgraded” regulations whose true annual cost exceeds the €1 million threshold but whose reported cost is zero or near-zero. Under the properly specified interval-censored MLE no regulation meets the strict criterion (posterior → 0.90 and reported cost €0), and only seven meet a moderate one (posterior → 0.50 and reported cost below €1 million). Within-bin shares of zero cost regulations fall monotonically with posterior probability, the calibration pattern that would be implausible under widespread strategic downgrading. The seven moderate candidates are descriptively linked to ministry and party of the responsible minister but the set is too small to support distributional inference. The paper makes two contributions to the regulatory-governance literature: it quantifies how a transparent reporting rule biases a widely-cited indicator, and it shows that informal estimates of a large “strategic downgrading” mass are an artefact of model choice rather than a feature of the data.