Absence of Record, Record of Absence: Positive-Unlabeled Estimation of Integrity-Law Enforcement Against Elected Officials

Abstract

Romania's integrity agency publishes findings that an elected local official is incompatible or in a conflict of interest, and the statutory consequence is termination of the mandate. Whether that consequence occurs is hard to observe, because removal produces a published administrative act, while non-removal produces no document at all. Absence of record is therefore not record of absence, and the missingness is not ignorable. This article treats a registry of 748 officials named in agency communications between 2014 and 2026 as positive-unlabeled data. The label frequency is estimated from the overlap of two search channels, court records and authority records, at 0.41. Correcting for this, integrity removal follows about 7 per cent of reports, but removal reaching the office-holder while the targeted mandate was still running follows under 2 per cent. The median procedure takes 946 days; thirteen of seventeen dated removals took effect after the mandate had ended anyway. For the in-mandate figure to reach 5 per cent, the search would have had to locate only 15 per cent of removals. The consequence is less refusal than being outlasted.