This series of blogs is derived from my discussions on the LinkedIn Quant Finance: What is the best approach to handling CMBS &/or RMBS Credit Risk analysis? discussion forum.
In a previous comment I had found that FICO socres were not useful, Russell, however, found that FICO scores were extremely good predictors of defaults. Here I clarifying what & how I did my analysis:
If I’m correct all the data I used was non-agency, 2005, 2006 and some 2007. More than 100 ‘tapes’ from service providers and banks themselves. There were in total about 330,000 residential mortgages assets. And I think they were all subprime – thats why the incomplete data problem.
The probability distribution of the universe of underwritten FICO scores was identical to the probability distribution of the foreclosed (no deliquents) underwritten FICO scores. I did not look at vintage because firstly FICO scores should have been enough information, and secondly there is another problem with vintage.
I think the difference in our analyses is that I used FICOs at underwriting and Russell used FICOs at default. We will know in a future comment, and I’m not sure what Russell meant by “projected defaults”.
I would definitely like to hear from other analysts on how they handled RMBS defaults versus FICO scores so that all of have a very clear idea on how to use FICO scores in the most beneficial manner for the industry.
Discalimer: This blog is purely for informational/educational purposes and is not intended to either negate or advocate any product, service, political position or persons.
QuantumRisk Blog Posts by Benjamin T Solomon is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License.
Based on a work at quantumrisk.wordpress.com.