Arnold Kling  

Today's Reading List

PRINT
Who Are These People?... Selection Bias in Blogging...

All from Mark Thoma.

The shortest read is from Robert Waldmann.


A sudden decline in the liquidity of assets can create problems as firms can't unwind leveraged positions without extreme market disruption. If the assets had always been illiquid, those leveraged positions would never exist. I think that would be a good thing.

The term Austro-Keynesian comes to mind. Read the whole post.

The most frustrating read is from Bethany McLean. I think she is right in her description of the feud between Fannie and policymakers. I think she is wrong in making it sound like the decision to put Fannie and Freddie into conservatorship was arbitrary. They had lost the confidence of investors, and that is the one risk that they could not possibly overcome. I knew they were dead back in July.

Finally, the read that has everything is from Joe Nocera. It has a major suits-vs.-geeks theme. He features Nassim Taleb. He talks about LTCM. One quote:


There was everyone, really, who, over time, forgot that the VaR number was only meant to describe what happened 99 percent of the time. That $50 million wasn't just the most you could lose 99 percent of the time. It was the least you could lose 1 percent of the time.

This is a really good point. Value at Risk in some sense measures the boundary of the 99 percent confidence interval for your portfolio. Other things equal, it's better to have a low VaR than a high one. But it is conceptually wrong to think of it as anything like "the most you can lose." When I was at Freddie Mac, I preferred stress-test methodologies to VAR. A stress test shows how much you lose in a specific scenario. See The Risk Disclosure Problem.

Another good quote:


Guldimann, the great VaR proselytizer, sounded almost mournful when he talked about what he saw as another of VaR's shortcomings. To him, the big problem was that it turned out that VaR could be gamed. That is what happened when banks began reporting their VaRs. To motivate managers, the banks began to compensate them not just for making big profits but also for making profits with low risks. That sounds good in principle, but managers began to manipulate the VaR by loading up on what Guldimann calls "asymmetric risk positions." These are products or contracts that, in general, generate small gains and very rarely have losses. But when they do have losses, they are huge. These positions made a manager's VaR look good because VaR ignored the slim likelihood of giant losses, which could only come about in the event of a true catastrophe. A good example was a credit-default swap

Not surprisingly, I really like the Nocera piece.

UPDATE: TIll Guldimann elaborates on his views of the financial crisis.


In the past banks were the principal intermediaries of money and the national regulators supervised them. Their sources of funds (liabilities) were local customer deposits, their uses (assets) loans. Regulators controlled the financial system by a) allowing only banks to accept deposits, b) specifying the minimum required capital banks must hold and c) influencing interest rates by injecting or removing funds from the banking system. This simple system has changed dramatically

The changes include disintermediation, derivatives, market value accounting, and asymmetric compensation.


Comments and Sharing






COMMENTS (5 to date)
Greg Ransom writes:

"The term Austro-Keynesian comes to mind."

You assume that the "Keynesian" adds something true that isn't already in Hayek.

Can you tell us what?

Some of us think Knight nailed it: what is true in Keynes isn't new, what is new in Keynes isn't true.

Freshman accounts of "classical economics" are comical for their bizarrely false accounts of macro as it existed at the time Keynes.

So, straight up. Tell me what is missing in Keynes that isn't in Hayek?

Expectations? It's in there. Liquidity preference? It's in there. Genuine uncertainty? It's in there.

So, what is it?

John Thacker writes:

I wish that the Nocera piece had mentioned the martingale betting strategy, perhaps when discussing LTCM. It's a good way that laymen can understand the risks of low-probability events, and what Taleb means when he talks about how if you have a "black swan" (massively negative event that occurs with some probability bounded below by a constant) you will eventually crash.

I.e., for other readers, imagine that you play roulette, or blackjack, or any other game. You start out betting $1 to win $1. Whenever you lose, you double your bet, so that you're betting $2, and if you win you win $2; minus the $1 you lost before, you net $1. If you lose a second time, you double again. Once you finally win, you win $1, and then you go back to betting $1.

It seems like a surefire strategy for winning over the long run, since you'll eventually get a win, after any streak of losses. The problem is liquidity-- eventually you'll have such a streak of losses in a row that you'll bust your bankroll, and not be able to cover the next bet.

Note that a strategy like this could show no risk according to VaR, so long as you made enough bets in the time period that VaR was measuring, say a day. In the case of a fair bet, 50% chance of winning, the odds of seven losses in a row are 1/128, less than 1%. But you can make the perceived VaR zero on any bet, so long as you can make enough quickly. (You're just "stuffing more risk into the tails.")

VaR is obviously flawed in the sense that it makes the martingale betting strategy look good, which any measure that ignores liquidity must do. Considering that the martingale betting strategy has been invented many times in history, that's a dangerous thing.

Anonymous writes:

There's a second way of gaming VaR: overfitting.

Not every correlation that occurs in the sample data is significant. If you buy a random portfolio, the correlations from your historical sample are the best estimate of the future correlations. But if you search your sample for assets that show the best correlation, that correlation is almost bound to be an overestimate. You can then build a portfolio that your model says is hedged, but that's an artifact you looked for and found in your sample set.

Jim Glass writes:

...banks began to compensate them not just for making big profits but also for making profits with low risks. That sounds good in principle, but managers began to manipulate the VaR by loading up on what Guldimann calls "asymmetric risk positions."

These are products or contracts that, in general, generate small gains and very rarely have losses. But when they do have losses, they are huge. These positions made a manager's VaR look good because VaR ignored the slim likelihood of giant losses...

How to pay for your kid's college tuition by betting on college football or basketball games:

Every week, find the longest shot -- say, a 30-1 -- then bet on the underdog to lose.

You'll make only about 3 cents per dollar bet, but you can mortgage your house and bet the whole proceeds. If your house is worth $400,000, you'll make about $12,000 per week. Nice! Who wouldn't enjoy a 97% chance of making $12,000 each week? After a dozen weeks you've made over $140 grand -- with a near 70% probability of success. Not bad odds! If you've got the gonads.

You just have to remember to quit then. If winning week-after-week seduces you into thinking you've found the golden road to riches, so you decide to carry your strategy over into next season, and before then you buy yourself a summer home and a yacht and a Beamer on credit figuring you'll pay for them with your next few bets, and then that's the year when on opening day Appalachian State beats Michigan...

mk writes:

The Nocera piece is really interesting and illuminating.

The scariest but most intriguing idea is that VaR's problems are a feature, not a bug. Is anyone in finance really incentivized to predict crashes? Or is everyone in finance incentivized to find a good "cover story" to drive up asset prices so everyone can go home happy with a new Lexus?

Finance is supposed to be a zero-sum game pitting different investment managers against each other. Theoretically this is what stabilizes asset values to their "true value." But you might also predict some amount of coordinated behavior to drive up asset prices, so that all investment managers benefit. Wouldn't it be reasonable to expect some such coordinated behavior? If enough such behavior occurs, a bubble starts, and defection (shorting the bubble) ceases to be profitable in the short-term, because prices are rising so fast.

VaR is an interesting story, but the real question is what the abuses of VaR say more generally about our financial system, and the incentives of the players therein.

Comments for this entry have been closed
Return to top