Tuesday, April 03, 2007

They Have Their Models!

I was watching a very interesting interview with Mr James Grant, editor of the Grant's Interest Rate Observer and a truly brilliant forecaster, and was stunned by some of what I heard. For all we have heard on the subprime mortgage market risks, most articles have focussed on the revenues of individual subprime mortgage providers. We have never really heard just how significant the subprime market is, and many articles have alluded to the subprime being a small fraction of the overall mortgage market.

Sub-prime and Alt-A [better than subprime, but below prime] represent 40% of the $8 trillion mortgage market. Hundreds of billions of dollars of CDOs were sold... You take 70% of a pile of BBB- (marginal)... 70% of this junk, is AAA. [How do they do that?] They have their models!

What is a CDO? CDO is a collaterized debt obligation - basically a package of mortgage loans that has been studied by credit rating agencies and assigned a certain risk. So basically agencies have used mathematical models to study recent correlations between the securities in that portfolio to classify the collection of junk as equivalent grade to a US Treasury. The problem, as Mr Grant points out, is that these models are based on what can be considered an anomalous recent period of lax lending, and the model predictions will turn out to be wrong. Indeed, a caption during the story pointed out that about 10% of subprime mortgages are delinquent by 90 days as of December 2006.

Mr Barry Ritzhold takes the focus to the effects on the economy, and the effect of the reset of interest rates on $2 trillion dollars worth of mortgage debt, that will result in monthly payments rising 10-50%. He cites a study by a title insurer, First American Corp., that projects that one in eight ARMs (adjustable rate mortgage) will end up in default. That's a stunning number, and in an economy that's already a little overstretched, may well cause a recession, in his opinion.

The risks associated with computer modeling though stretch further than the mortgage market. A bevy of hedge funds and "quant" investors have increasingly been relying on mathematics for investing. As one of my professors who teaches a water quality modeling course keeps reiterating, creating a model without the right data to support it is an exercise in academic gymnastics ... and foolhardiness.

No comments: