PD / LGD correlation (empirical)

[[email protected]]

Hi Everybody

I have some serious difficulties with a credit risk task where I try to estimate the correlation (first linear) between PD of a residential real estate mortgage portfolio and LGD. For me it is plausible, that as e.g. the economy worsens PD of the portfolio increases and LGD increases as well. I want to empirically check this intuition and find some evidende for this assumption. The null hypothesis would therefore be H0: roh(PD, LGD) = 0.

If a house price index is taken as estimation (proxy) of the changes in LGD over the years, to measure correlation only PD needs to be estimated from external data.

However, I do not find a reliable estimation method to come up with PD data for the years. I read trough the forum already and found some very interesting inputs. I would be very grateful for any help in order to estimate PD for a portfolio of real estate loans with external data.

Thank you very much!

Gabe

 

[Sidharth]

Hi Dailay

I believe you are thinking on the right lines, as economy conditions go bad people income and asset prices will go down.

LGD like you said can be derived from house price index, similarly reason for default will be fall in Income due to which Loan service ratio of borrower will go down - Amount of Loan/Income

Hence PD will be indirectly proportional to Income, Do you any such index which reflects Earnings of average household ??

Note- Income is only on the major contributor , there will be other as well.

Sidharth

 

[AG]

Gabe, for retail portfolios, logistic modelling/scorecard building is a good technique to estimate the PD....

However, if you are working on developed market mortgage portfolios, I'd suggest you use the FICO score as a proxy for PD.... The alignment is: at a score of 660 the odds is 15:1 and it doubles every 15 points... If you want to find the PD of an obligor with a score of 690 say, the odds is 60:1, i.e. the PD=1/(60+1)....

Hope it's useful

AG