Title: Forecasting mortgage defaults: evidence from UK portfolio-level data

Authors: Alexios Makropoulos; Anastasios Savvopoulos; Peter L. Jones

Addresses: Robert Gordon University, Aberdeen AB10 7QE, UK ' Leeds Building Society, Leeds LS1 5AS, UK ' Robert Gordon University, Aberdeen AB10 7QE, UK

Abstract: This paper considers an equilibrium/error correction modelling (ECM) approach to identify determinants of mortgage portfolio default rates at firm level. Using mortgage portfolio data from a UK lender we estimated the lender's portfolio concentration weights in different UK regions to estimate portfolio-adjusted unemployment rate and portfolio-adjusted house price index. The modelling results suggest that the portfolio-adjusted unemployment rate and house price index along with Council of Mortgage Lenders' (CML) default rates, interest rates and household savings are useful determinants of firm-level default rates within an ECM framework. These findings may be of practical use for forecasting or back-populating default rates in financial institutions and especially for the smaller lenders where a long time-series with default rates may not be available. However, modelling and forecasting at portfolio level should be made carefully, as we argue that data quality and changes in the lending policy should be carefully monitored.

Keywords: credit risk; default forecasting; error-correction modelling; ECM; mortgage defaults; UK; United Kingdom; unemployment rate; house prices; default rates; interest rates; household savings; data quality; lending policy.

DOI: 10.1504/IJCEE.2015.068677

International Journal of Computational Economics and Econometrics, 2015 Vol.5 No.2, pp.199 - 210

Received: 30 Dec 2013
Accepted: 17 Sep 2014
Published online: 08 Apr 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article