Arnold Kling  

Quote of the Day, II and III

Quote of the Day... Priceless Anecdote...

The highly quotable Nassim Taleb has a new essay.

In some situations, you can be extremely wrong and be fine, in others you can be slightly wrong and explode. If you are leveraged, errors blow you up; if you are not, you can enjoy life.

Taleb has a message for Robin Hanson:

"prediction markets" are for fools. They might work for a binary election, but not in the Fourth Quadrant.

The fourth quadrant refers to a matrix where payoff functions are either simple or complex and probability distributions are either well-behaved or highly skewed.

In the first quadrant, the payoff function is simple and the probability distribution is well behaved. When Tony LaRussa orders a squeeze bunt, he is making a decision in the first quadrant.

In the second quadrant, the payoff function is simple, but the probability distribution is highly skewed. If you buy a backup generator, you know what the benefits will be if the power goes off. But it is difficult to estimate how much use you will get out of the backup generator, even based on past data on the number of power outages in your area. The longest power outage you have experienced in your lifetime is not the upper-most estimate of the distribution.

In the third quadrant, the payoff function is complex, but the probability distribution is well behaved. To me, bets on football games seem to belong in this quadrant. I don't understand the payoff structure that the guys on the radio are talking about ("I'll give the three." "I'll take the under.") I don't know what any of it means, and I don't really care, but I'm assuming it's more complex than just betting on which team is going to win. However, my guess is that the underlying probabilities are well behaved. It is in this quadrant where Taleb grants that "statistical methods work surprisingly well."

Finally, we have the fourth quadrant, where the payoff function is complex and the probability distribution is not well behaved. Taleb's point is that too many finance people are eager to play in the fourth quadrant, believing that they are in the third quadrant.

Taleb wants to argue that the current mortgage meltdown is a case of financial firms playing in the fourth quadrant without realizing it. House prices behaved in an unprecedented manner, and people were caught unprepared.

I will grant that people were caught unprepared, but I do not think that the recent house price declines are outside what seemed plausible based on historical experience. I think that what happened in the mortgage industry is that we moved from the first quadrant to the third quadrant, and then folks decided not to use the statistical models that had worked well for years in mortgage default analysis.

When you make loans with a 20 percent down payment and all the risk of default is held by one institution, the payoff function is relatively straightforward. When you make loans with little or no down payment, and you create "structure securities" in which the default risk is split unevenly, the payoff function gets really complex. In an adverse environment, nobody knows what the securities are really worth. That is why all those mortgage securities today are trading at such a huge discount relative to what is likely to be their fair value.

I think that the Foster-Van Order model worked fine for assessing mortgage risk. It's just that Wall Street used something different, and Freddie Mac's CEO decided that the Foster-Van Order model was getting in the way of Freddie's mission to support affordable housing.

Comments and Sharing

COMMENTS (10 to date)
Adam writes:

Am I wrong in my perception that Taleb is really just an arrogant jerk who doesn't add much to any discussion?

For instance, the reasonable response to his comment about prediction markets would be to look for a track record in those circumstances he considers "fourth quadrant". If my limited exposure to the man has given me any accurate understanding of his position, this is how he would probably reply:

Using evidence is for naive empiricists. Also, even if evidence did ever mean anything, you have misapplied when a situation is in the fourth quadrant, which is only ______ (whenever the mood to categorize it thusly hits Taleb, certainly not in a case where it would make him potentially vulnerable to criticism).

I mean I started Black Swan and couldn't get more than a third of the way through it because it was full of that sort of self-indulgent tripe; but that does mean I didn't take the time to find out if there was anything of value in it beyond that. Is there any way to put his hypothesis about prediction markets to a test, or is his criticism as up in the clouds and untouchable as he always seems to make them?

Adam writes:

I also have to wonder if Google's internal prediction markets are "foolish" by Taleb's criteria.

sourcreamus writes:
That is why all those mortgage securities today are trading at such a huge discount relative to what is likely to be their fair value.
If this is true, there is an opportunity to make a good bit of money. Who is positioned to make this money?
Arnold Kling writes:

Institutions that have capital to deploy and the ability to assess the risk of the mortgages that are backing securities should be able to pick up some bargains. My guess is that in a few years we will find that some institutions made out really well by buying in this market. But if you don't know what you're buying, you can get killed. It isn't a free lunch.

Barkley Rosser writes:


You are right. He also likes to threaten to sue people who point out how bogus some of the more specific advice that he hands out is, alhthough he is very quotable.

caveat bettor writes:

I think we should have a more masonomic respect for prediction markets, i.e. prediction markets fail; let's use prediction markets.

Prediction markets are inferior to which benchmark? Anecdotal hindsight bias???

The Sharpe & Information Ratios are useless without a good benchmark. There is a reason why Empirica Capital folded.

btw, I wouldn't conflate "quotable" with "readable".

Max writes:

He seems to dislike parametric assumptions of any sort, but also stress tests. His fourth quadrant, is described verbally, which makes it very squishy (what is complex or fat tailed, is then subjective). His rules come down to 'expect anything', which is true at one level, but also entirely useless. You need a specific alternative, not just a criticism, and all Taleb is saying is that any specific model is imperfect.

Robin Hanson writes:

When someone "quotable" makes a claim, but offers no theoretical or empirical support for that claim, is it really worth bothering to reply?

Ajay writes:

Max, what part of this statement from the linked essay do you find non-specific? "It appears that financial institutions earn money on transactions (say fees on your mother-in-law's checking account) and lose everything taking risks they don't understand. I want this to stop, and stop now." He also says that models in the third quadrant can work well, which makes your statement that "all Taleb is saying is that any specific model is imperfect" false. Do you have something invested in the particular techniques and institutions that he's calling bullshit on? Because that's the only reason I can think of for your false statements critiquing him.

Marc in Asia writes:

Arnold, I think your last sentence needs editing.

Freddie Mac's CEO decided that the Foster-Van Order model was getting in the way of Freddie's mission to enrich Freddie management.

You can't really believe that Frannie have been acting in the interests of housing affordability.

On Taleb, I'd say that while he makes some very good points and important angles to view the story from, he's not nearly as original as he so profoundly claims. In the case of this four quadrant framework, I think he is basically right. Banks, Wall street, and the regulators were mostly working with tools designed for quadrant three -- where payoff are complex (think structured derivatives) but underlying risks are well-behaved. They underestimated the possibility that they were playing in quadrant four; that the underlying risks might contain some surprises.

Arnold is also right that these shouldn't have been surprises. We haven't seen true quadrant four from the underlying markets (i.e. house prices, ability to make payments), just something a bit on the edge of quadrant three. This makes not having been prepared all the more inexcusable.

Comments for this entry have been closed
Return to top