I return to a point contained in my original letter that both Colin Snowdon and Nick Baxter ignored, presumably because it disarms their arguments. At GMAC-RFC, we piloted automated underwriting in 2003 and fully rolled it out in 2004.
We benchmarked our delinquency performance over nearly five years and many billions of pounds against Council of Mortgage Lenders and Fitch Ratings indices. And we asked our portfolio buyers to compare the performance of our loans with those they acquired or originated separately. In all instances, the automated approach produced higher quality delinquency performance at the portfolio level, in some cases by a margin of 50%.
If Baxter has come across a few instances where the automated approach has performed badly, then these might either be exceptions that do not invalidate the general trend or be the subject of poor automated underwriting systems.
Neither invalidates the fact that based on the above large sample, automated underwritten loans performed better. I feel that you need to have run a high volume mortgage lender that has undertaken both manual and automated underwriting to compare the two.
There are many reasons why the traditional paper-based manual underwriting approach produces poorer de- linquency performance.
For a start, there are several websites available where applicants and their brokers can produce high quality references, payslips and IDs that are completely undetectable and utterly false. Presumably, if Baxter came across a file with that documentation he might take comfort, as do the rating agencies and the regulator, amazingly.
But this is 2008. The frauds I have experienced in recent times have involved paper or human beings. In addition, every day of every week, underwriters are subject to broker leverage, influence and error, which produces inconsistent decision-making. By contrast, a good automated underwriting approach will analyse consumers’ attitude to credit and ability to manage debt, which are far more important than arbitrary paper-based income assessments in any event torn to shreds by subsequent interest rate rises.
If there is insufficient existing credit to analyse then applications are rejected. My concern is that those without direct experience of both underwriting approaches might, with the best of intentions, use the current liquidity crisis to take us back to a cautious manual approach. They may think it’s a comfort zone but I know from experience it’s a danger zone.