Focus
Household finance surveys are used across many central banks to measure families' net worth, indebtedness and default risks. These surveys are especially important in countries where non-bank lenders have a substantial presence and where official sources may not measure all the loan obligations of individual families. Furthermore, growth in household debt is strongly associated with a higher probability of banking crises, showing the need for reliable data on this topic.
Contribution
I study differences between self-reported debts via household surveys and official records using a matched data set from survey respondents and administrative banking loans in Chile. This is the first study comparing loans reported in surveys and those in official registries using a national representative sample and with information on all bank loans. Furthermore, I focus on debt differences between surveys and official records across individual borrowers, showing differences according to families' backgrounds.
Findings
I find significant differences between survey answers and official records in the number of loans and debt amounts reported. Discrepancies for debt amounts range from -31% to 18% for mortgages and from -1% to 59% for consumer instalment loans. Borrowers accurately report their loan maturities and delinquency status. Larger discrepancies between survey answers and official records are associated with respondents that have a lower income and are younger relative to other family members. Finally, I show that rounding answers can explain a significant part of these discrepancies.
Abstract
This article advances upon previous studies by using a unique match of a representative sample of individual borrowers from the Chilean Household Finance Survey and their banking loan records. I show that surveys differ from the credit registry, not just in the number of loans reported, but also in their amount, with a substantial degree of heterogeneity. Delinquency status is accurately reported by survey respondents. Furthermore, a substantial fraction of the discrepancies can be explained by rounding error in survey answers. Finally, I find that discrepancies are larger when respondents are not the highest-income member of the family.
JEL classification: C81, D10, D12, E21, G21
Keywords: household finance surveys, mortgages, consumer credit, default, measurement error