Wednesday, February 13, 2013

- Misinforming The Uninformed

USA today is a color printed daily that has a reputation for skipping across the surface of a story like a stone on a still pond. It will give you the highest level of the story so you can then tell people you 'know what's going on', but it would never give you enough detail to understand a story's context.

Compared to the NYTimes, that might actually be better, since no context also means no distorted context. But few would disagree, even on the left, that USAToday is the perfect news medium for today's 'low information' voter.

Which is why this story bothers me so much. Take this quote:

Physicists and economists led by Didier Sornette, chair of Entrepreneurial Risks at ETH Zurich, have developed models that can identify market bubbles and predict when they will burst. Sornette's team has successfully predicted several recent crashes, including the worldwide market crash in October 1997, the dot-com bust in April 2000 and the crash in September 2008.

This is completely false. The truth of the matter is that they have a process which "they claim" will predict market bubbles and has done so only on an ex-post "back-testing " basis. This is what we in the industry refer to as "fitting the curve". Their model will be next to useless in predicting the next 'bubble burst'.

Imagine that you want to write a computer program to fly a plane over the rocky mountains while using a minimum of fuel. It takes more fuel to increase your altitude so only doing so when you have to would make your computer model seem more efficient. If you REALLY custom cut your software to rise only when necessary, the results will look spectacular on paper. Make allowances for every tiny hollow, dry riverbed or broken tree, and your code will look amazingly accurate. But when you then tried to use your software to fly over the Alps and the dry riverbeds and broken trees weren't' where you expect them to be, you'd crash.

That's the principle behind over fitting. And in my opinion it's the most common problem in quantitative finance. It's at the point these days that we with experience can tell how badly something is overfit with just a few summary statistics. There is only so much information in the data. And inexperienced quants have a habit of mistaking data fitting for 'new discovery'.

Then there is this mammoth piece of wrongheadedness:

Economists already use some tools to try to identify bubbles, and the Federal Reserve uses its control over interest rates to try to avoid crashes. But these decisions would be even more effective if policymakers used the full range of predictive tools available, including those produced by Sornette's team.

Granted, doing so would be expensive. But if we consider what we spend on intelligence and defense, the resources needed to improve economic security are minuscule.

We have made this kind of investment in science before, with both the Manhattan Project in the 1940s and the space race in the 1960s. We need to make a similar commitment to financial research.

A big push for government funding is the LAST thing we actually need. There are already thousands of scientists and analysts out there right now trying to figure out how to 'predict' the next market bubble, but they work in the private sector. Some of them work for me. And only those who work in Physics and not markets have the intellectual vanity to make such a ridiculous claim. The data we analyze in the markets is qualitatively different from the data in physics. The rules that Physicists believe are universal (and are in physics) are not in finance. In other words, past results (more often than not) are no indication of future returns.

If the scientists above really could predict the next bubble, then they wouldn't need to go begging for funding from government. People would line up to buy their service in any form they were willing to sell it - and the regulators wouldn't need to use it because the markets would correct themselves. But the fact is they cannot predict it. That's why they go begging to government for funding - because there is no one else who will offer it to them unless they produce results.

People in government suffer from the same sort of intellectual vanity that physicists do. So the scientists who cannot produce results go to the people who do not require results, in order to waste a little public money on a fruitless effort, designed from the outset to fail. and when you want to sell this dubious mish mash of misunderstanding, you run to a publication like USA today and the low information voter. This is jus the kind of pointlessness that makes Snooki-Ameri cans feel like they've accomplished something and made the world better.

I got a resume the other day from an applicant that had spent his career in government and NGO's. He claimed he successfully predicted the last 7 market crisis. Billionaires (Paul Tudor Jones for example) have built entire careers upon having successfully predicted only one of those crashes, so it was a very bold claim to read from someone who had only worked in government. If it were true - even in part - then there was really no reason that I shouldn't have been working for him.

When I pressed him about it it turned out that he didn't actually predict a date, he didn't predict an amount, and he didn't predict the breadth of any of the things he said he did. In other words all he had done, was write a paper describing a vague market excess which will be unlikely to persist. That's the kind of thing that is common knowledge in the finance industry prior to any genuine market bubble. We all know it's there, what we don't know is when it will break.

That's all these physicists can tell us, but to them this is a revelation. It isn't. It's just the thing we all already knew.

We don't need an economic Manhattan project. A bunch of government funded academic intellectuals is the last thing we need mucking around where they aren't qualified. What we really need is for the government to get out of the way and let the rest of us get back to fixing the market ourselves. Physicists are generally smart people. But if the financial markets were as easy to "manage" as they believe, then we would have all those problems solved already. The author of this idiotic piece needs to learn humility. And he won't learn it by over fitting data in a government funded cocoon.

2 comments:

ikaika said...

Also - they refer to the 1998 failure as some eort of success.
They didn't predict anything, but everyone in the industry predicted that if decimalization and odd-lot representations (coupled with exchanges competing within exchanges) would ultimately lead to the POS electronic market we currently enjoy.

I attended two meetings last week. One on HFT where the proponants said tthere was not enough but they claimed to have added liquidfity (I had to pick myself up off the floor I was laughing so hard at that one).
And the opponants of HFT offered the solution that we just give up.
The following meeting was about Latin America.
On the panel were HFT people complaining vigorously that they don't have access to Brazil.
Other panel members lamented that the Latin exchanges were not as functional as ours...
Again - I laughed out loud and was almost escorted out.
Brazil has an orderly functional exchange compared to our Reg NMS disaster.
Rant over

ps: I hung out with Dennis Gartman the other night - nice guy.

Anonymous said...

Heck, *I* predicted the dot.com bust and the housing crash, and I have no background in finance, etc. I just saw the wackiness going on and though, "Hmmm... Tulips..." When people through too much money at something, and JUST KEEP DOING IT, eventually people are going to realize it is NOT worth the inflated prices. Common sense should kick in at some point.

(Good lord! During the dot.com craze, "investors" were throwing money at the flimsiest of cot.com business plans.)