Martin Hodula, Milan Szabo, Lukáš Pfeifer, Martin Melecký
This paper studies the effects of regulatory recommendations concerning maximum (i) loan-to-value (LTV), (ii) debt-to-income (DTI) and (iii) debt service-to-income ratios (DSTI) on new loans secured by residential property. It uses loan-level regulatory survey data on about 82,000 newly granted residential mortgage loans in the Czech Republic from 2016 to 2019 to estimate the average effects of the Czech National Bank’s regulatory recommendations and their heterogeneous effects depending on borrower, loan, bank and regional characteristics. The studied response variables include the mortgage loan size and lending rate and the value of the property with which loans are secured. The machine learning method of causal forests is employed to estimate the effects of interest and to identify any heterogeneity and its likely drivers. We highlight two important facts: (i) value-based (LTV) and income-based (DTI and DSTI) limits have different impacts on the mortgage market and (ii) borrower, loan, bank and regional characteristics play an important role in the transmission of the recommended limits.
JEL codes: E44, G21, G28, G51, R31
Keywords: Borrower-based measures, causal forests, Czech Republic, macroprudential recommendations, residential mortgage loans
Issued: March 2022
Download: CNB WP No. 3/2022 (pdf, 1.4 MB)