Using household-level data to guide borrower-based macro-prudential policy

In 2019, Luxembourg introduced borrower-based instruments in the macro-prudential toolkit to constrain credit to households who exceed a certain limit on their loan-to-value ratio, on their (mortgage) debt-to-income ratio or on their debt service-to-income ratio. This paper analyses the impact of setting these limits at different levels, using household-level data from Luxembourg. We calculate these debt burden ratios for individual households who recently purchased their main residence using data from the Household Finance and Consumption Survey conducted in 2010, 2014 and 2018. On January 1, 2021 authorities imposed a legally binding limit on the loan-to-value (LTV) ratio for new mortgages. This may be 80%, 90% or 100% depending on the category of borrower. Had the least restrictive LTV limit envisaged by the law (100%) been applied in 2018, credit would have been rationed to 24% of households with recent mortgages on their main residence. This limit would have required a 7% reduction in the overall debt of this group of households. Had the most restrictive LTV limit envisaged by the law (75%) been applied in 2018, credit would have been rationed to 64% of households with recent mortgages on their main residence, requiring an 18% reduction in overall debt in this group. More generally, we evaluate how well borrower-based instruments can target those households that are financially vulnerable (according to conventional measures from the literature). By simulating an adverse scenario, we find that combining several ratios one could better target households that were not financially vulnerable in the benign conditions of 2018 but would become vulnerable after a shock to income. However, any borrower-based instrument inevitably generates some classification errors (either granting credit to households that are financially vulnerable or constraining credit to households that are not financially vulnerable). Using different assumptions on policymaker preferences, we apply the signals approach to derive limits that are “optimal” in the sense of minimising classification errors.