Models cannot replace brains

I liked this article on how the scientific models for cod population were bogus. A friend of mine who studied Earth and Planetary Sciences at a good university left the field because she found the models they used to be totally bogus as well (yes, these are the models used in Global Warming). I developed a computer model looking at proteomics -- it was bogus also. Obviously the risk models used by banks were similarly bogus, although the "Government will bail us out" parameter seems to be working as advertised.

This is not to knock models -- it valuable to make assumptions that are implicit, vague, and inconsistent; explicit, precise, and consistent. But models don't replace brains.

I think a good example is the traditional subprime mortgage: there are people out there with bad credit, but good willingness (and ability) to repay. One test of this willingness and ability is to require a significant downpayment -- if you have the discipline to save for an extended period of time, you will repay your loan. This makes the second part of the test, a low "teaser" rate, make sense: if you can make payments for two years, then your credit score improves and you can escape the reset by refinancing into a regular mortgage. If you cannot make payments, then you are higher risk, and have to pay a higher rate to compensate for that.

This is also why just looking at interest rates does not tell you much about credit expanding or contracting, you also need to look at how the downpayment requirements are changing.