Bank for International Settlements study finds household debt surveys misstate loan participation and amounts while delinquency is reported accurately

The Bank for International Settlements published a working paper assessing how accurately households report debts in surveys by matching respondents in Chile’s Household Finance Survey with their administrative banking loan records. It finds substantial gaps between survey and registry data on both whether borrowers hold particular loans and the amounts reported, especially for consumer credit, while delinquency status and many maturity measures are reported with much higher accuracy. Using 8,047 matched respondents from the 2011, 2014 and 2017 survey waves linked to monthly bank loan records from 2003–2018, the paper shows that 9.3% of respondents report a mortgage in the survey that is not present in the registry, while 14.1% have consumer debt in the registry that they do not report in the survey. For borrowers with positive loans in both sources, the interquartile error range for original loan amounts spans -31% to 18% for mortgages and -0.7% to 59.3% for consumer installment loans; adjusting for rounding to multiples of 10 materially reduces these gaps and sets the median error to zero in several categories. Discrepancies are larger when the interviewee is not the highest-income household member, and the mismeasurement shifts risk diagnostics: the survey implies higher mortgage and lower consumer debt than the registry, and a 40% debt-service-to-income threshold classifies 20.1% of borrowers as risky in the registry but not in the survey, compared with 4.6% in the opposite direction.