Patrik Nevsten och Daniel Lönnborg, presenterar sitt examensarbete

A Credit Risk Model for Real Estate Lending Portfolios
- Implemented on the Swedish Residential Market 

Abstract

The purpose of this Master's thesis is to construct a model that,
based and calibrated on official data, is able to calculate and
compare the credit risk exposed in mortgage portfolios with Swedish
real estates as collateral. Having such a model one would be able to
estimate the exposed credit risk in a portfolio with different lenders
and collaterals of different real estate types and regions. Since the
model is calibrated on official data, it would be possible to use on
all major Swedish credit institutes from a comparing perspective.

To succeed with this challenge we decided to take on an actuarial
approach at a portfolio level. This was done since we will be looking
at portfolio losses rather than individual ones, and the main factor
that constrains us is the level of descriptive data possible to
acquire. The choice of framework naturally fell on the actuarial
approach used in CreditRisk+TM. We though found that the econometric
technique, presented in CreditPortfolioViewTM, offers some very
interesting advantages, especially concerning the possibility to
acquire the loss distribution through the use of driving macroeconomic
variables that automatically takes into consideration the correlation
between default probability and the value of the collateral. Finally
the implemented credit risk model was tested on an approximated
Spintab-portfolio.