Bad statistics meets worse journalism

From the OC Register, we have Jonathan Lansner talking about Novel mortgages blamed.

Cagan tracked everything including purchase prices, kinds of loans used and fresh valuations of homes. Then he estimated how many owners will get into deep trouble as their monthly payments increase when they have little or no equity left in their home.

Yes, Cagan sees 18,601 O.C. mortgages going bad through 2011.

18,601 exactly? Not 18,602, or 18,603? Heaven forbid, it should be 18,599. Oh, the horror!

And a forecast out to 2011, eh? That's pretty gutsy. Cagan is a wondrous beacon of light boldly looking out into the mists of time.

Since I do stuff like this for a living, I should explain. Economic analysis (when done well) generally reveals a range of possibilities (a distribution, formally.) While, I can talk about the most likely outcome (mean or median, whatever is most appropriate), the chances of hitting that mean exactly are about as likely as a politician telling the truth. Which is to say, not a fucking chance in hell!

However, newspaper reporting can't deal with such subtleties as a range of outcomes. They want to know crude things like, "will it go up, or will it go down?", and I would probably answer, "That depends on whether you're taking Viagra or not!" (but that's why they don't interview me.)

Secondly, most models have an error associated with them. The error compounds geometrically as you iterate the model (which is to say the model becomes pretty darn useless after a few iterations.) Our amateur statistician has fallen into the above trap by projecting out the model for 5 years. I've even seen professors of statistics from famous schools (which shall remain unnamed) fall into the above trap so it seems to be an endemic arrogance of the profession.

Lastly, all models (particularly economic) are conceived with certain assumptions in mind (either overtly or implicitly.) However, events that are outside these assumptions can radically alter the distribution, and the outcome (the "fat tail" phenomenon.) How many airline profit-projection models took into account the possibility of two planes crashing into two tall towers?

You can never account for all possibilities, but good traders make sure they are protected against these, as yet unknown, "fat tails".

I would like to add that in the light of the above paragraph, the following quote of Donald Rumsfeld (which he took a lot of flak for) makes perfect sense:

"Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns -- the ones we don't know we don't know."

The "fat tails" are the unknown unknowns, and thinking about them is absolutely crucial to developing trading models (which is what I do for a living.)

I'm sure the journalist doesn't understand "standard errors", "geometric compounding of errors", or "fat tails". He's just a parrot repeating stuff without any clear understanding of the concepts.

As for the unknown unknowns, fuggedaboutit!