Loan to Value Caps and Government-Backed Mortgage Insurance: Loan-Level Evidence from Dutch Residential Mortgages

Leo De Haan,Mauro Mastrogiacomo

doi:10.1007/s10645-020-09367-w

Abstract

Using loan level data on mortgage loans originated by Dutch banks during 1996 to 2015, we analyse the determinants of the incidence of non-performance. We find that both the originating loan-to-value ratio (OLTV) and the debt-service-to-income ratio are significantly positively associated with the probability of non-performance. The results suggest that mortgages with government-loan-guarantees perform better. Moreover, several mortgage loan and borrower characteristics, such as the (interest-only) loan type and the underwater status of the borrower, increase credit risk. Our model predictions suggest a novel policy implication: in order to avoid acceleration of non-performance probabilities, the OLTV-limit should be set to about 70–80% for uninsured mortgages, and to about 90% for those with mortgage insurance.

Highlights

A growing amount of literature is currently focusing on the relationship between indebtedness and credit risk
Using loan level data on residential mortgage loans of all Dutch banks originating from 1996 to 2015, we analyse the determinants of the incidence of non-performance
Using loan level data on residential mortgage loans originated by all Dutch banks during 1996 to 2015, we analyse the determinants of the incidence of non-performance, i.e., arrears or defaults

Summary

Introduction

A growing amount of literature is currently focusing on the relationship between indebtedness and credit risk. Most studies focus on cross country comparisons and rely on differences across countries, to estimate the relationship between LTV and DSTI-caps and risk. These studies are either conducted using macro data (Stanga et al 2017) or micro data (Japelli et al 2008). On the contrary, can pin-down a treatment and control group, but in the case of Japelli et al (2008), for instance, the survey nature of the data does not allow a specific study of tail-risk, which is a problem as defaults in several countries occur in < 3% of the relevant population

Methods

Results

Conclusion