Correlation, Causation, and Change

Reality reared its ugly head overnight last night as the Europeans realized that they have some trouble on their hands in the form of U.S. denominated mortgage backed securities.

The equity markets took a pummelling as the beneficiaries of global liquidity (i.e. broker-dealers) faced tougher prospects for an easy out: U.S. Stocks Tumble on Credit Concerns; Banks, Brokers Retreat

Interestingly, at the same time that more bad news emerges around CDO's and other asset-backed pools of securities, many quantitative hedge funds are apparently taking large hits as their statistical models face a changing marketplace: Highbridge, Goldman `Quant' Hedge Funds Lose Money

To me this is unsurprising and driven by the same trend causing the implosion in the CDO markets. Most of these "quantitative" hedge funds are driven by modeling techniques that utilize a high degree of statistical analysis. Statistics is great for telling a story about what has happened and also powerful for picking up potentially unnoticed correlations (and therefore - in theory - causation). However, this power is harnessed and utilized in the context of looking at historical data.

The challenge facing the credit markets and statisticians generally is that things seem to have changed. For whatever reason (I have speculated about many possible theories below), the underlying credit dynamics of the housing market and the surrounding financing markets have had a dramatic impact on the mispricing of risk generally, and now reality is coming home to roost. It is not surprising that as these underlying changes play out, models based on data collected in a different historical context sometimes fail.

Maybe part of the reason why these models don't work as planned is because the system is built on the backs of sometimes not entirely rational creatures - american homeowners.

This article offers an interesting take on why home owners are lured into loans they may not be able to afford: The Psychology of Subprime Mortgages

As the author states: "I think a big part of the reason sub-prime loans remain so seductive, even when the financial terms are so atrocious, is that they take advantage of a dangerous flaw built into our brain. This flaw is rooted in our emotional brain, which tends to overvalue immediate gains (like a new house) at the expense of future costs (high interest rates). Our feelings are thrilled by the prospect of a new home, but can't really grapple with the long-term fiscal consequences of the decision. Our impulsivity encounters little resistance, and so we sign on the bottom line. We want the house. We'll figure out how to pay for it later."

Such a statement sure sounds like it makes sense and is predictable to a human being like you or me. But unfortunately, statistics speaks a different language than common sense...and as it turns out that linguistic gap may have dramatic consequences.