Wednesday, March 5, 2008

Transparency and New Conforming Loan Limits

HUD had 30 days from the date the law was passed, February 13th, to designate the median home prices and otherwise resolve all the ambiguity. However, the rumor mill indicates they may not get their act together until this summer. Additionally, there is currently discussion about not actually changing the conforming loan limit (per se), but rather "adding" a new loan classification - maybe they will call it conforming-jumbo.

All in all, that might be along the right lines, just not enough. The huge ripple effect through the financial community was partially driven by a lack of transparency - the claims that no one really knew what kinds of loans were in a portfolio. The argument is that if more were truly known about the loans, they could have been better rated for risk and priced accordingly. But, what can create the required transparency?

I have two recommendations. Context & Labelling.

Context (weighed heavily in my PhD research).  Consider that a number without units (context) is worthless.  What is my speed?  5.  5 what?  Feet / second?  Miles / hour?  Although equity in a home is considered when the loan is first given, it seems to be an unknown when a loan is sold (re-sold) into the secondary market.  After all, if people are getting 100% financing via a 1st-mortgage for 80% and a 2nd-mortgage for 20% (to avoid paying mortgage insurance) then there is no equity in the property.  That is a very different 1st-mortgage than the one where someone has put down 20%.

Labelling. Label the loans by the key attributes. (This is really just a derivative of context when you look closely). Start with Loan Amount by Credit Score, in units of, say, $50,000 and 50 pts respectively. (E.g., Loan $250-300K & 500-550 FICO vs $550-600K & 750-800 FICO.) Such classification would provide the much needed transparency. Then, it is fairly easy for the market to "buy" a bucket of loans and to know exactly what they are getting. And the invisible hand of the market can efficiently price such loans. Better yet, loans that the market doesn't like will dissappear entirely for lenders who resell their loans or else will be priced higher (e.g., hard-money loans).

We must realize there already is (an N-dimensional 'grid') classification for loans based on #years, interest rate, payment schedule, interest only, etc. So all I'm really offering is that we add more attributes (context). In fact, as a programmer, I believe everything but the equity in the home is already available.

Of course, the more attributes to be measured and the increase in the number of divisions in each attribute result in a rapidly growing quantity of unique labels. E.g., 12 (credit score buckets) x 20 (price divisions) x 10 (equity divisions - say, by 10% percent) x 10 (loan term options) x NNNN = 24,000 x NNNN. Seems like a lot, but we seem to be able to have 'zillions' of different mutual funds. And besides, it is just a database and the numbers identified are quite small for our computers.

Now for the fun... (Another part of my research and work history is about information in real time.) What we will then need is to dynamically re-classify loans based on changes in equity. E.g., a $100,000 home equity loan removes that much equity out of the house, reducing the equity, say from a 25% down-payment to 5% remaining equity. This just made that first loan quite a bit more risky. This would take effort to implement and track, but it would stabilize the overall system by creating (and managing) the much needed transparency.


No comments: