Default and Fraud in Guaranteed Loans

Based on econometric analysis of more than 1.1 million limited-company loans issued under BBLS and CBILS, this paper shows that the central issue was not simply borrower fraud, but how scheme design, portfolio risk, and lender classification behaviour combined to shape what was ultimately recorded as fraud. The results are striking. The decision to prioritise speed, self-certification, and minimal ex ante verification under BBLS generated substantially greater default and fraud exposure than under CBILS, while evidence from logit models, survival analysis, accelerated failure time models, and external validation indicates that challenger and fintech lenders were materially less likely than major banks to classify fraud conditional on default. This is not merely a statistical nuance, but a finding of clear economic significance: a meaningful share of fraudulent exposure appears never to have been formally recorded as fraud, with under-classified exposure estimated at a minimum of £90 million among the limited company portfolio. More broadly, the paper demonstrates that in emergency lending programmes, measured fraud is not simply observed misconduct, but an outcome shaped by institutional incentives, screening intensity, and differences in detection capacity across lenders.