Friday, March 20, 2009

Collective Punishment for AIG

I have heard the argument that those at AIG should not get bonuses because they destroyed the firm, or because they destroyed the firm and in doing so helped precipitate the current economic calamity to boot. This sort of argument doesn’t make sense to me. The vast majority of those in the bonus pool at AIG had nothing to do with precipitating the firm’s failure. They were marketing insurance products, managing call centers to handle customer inquiries, and other exciting stuff like that. They just happened to live in the same corporate city-state as the evil-doers. Pulling their bonuses based on such an argument is collective punishment.

A more reasonable argument is that without the government assistance, AIG would have gone bankrupt. And if it had gone bankrupt, those who are pulling in bonuses not only would have had no bonus, they likely would have had no job. So then, the argument goes, why should the government’s bailout money – which of course is tax payer money – go to give out-sized bonuses?

That makes sense. But then we come to some follow-up questions.

One is why Paulson didn’t include compensation controls as one of the terms for keeping AIG afloat. You could ask the same thing of Geithner, who frankly is taking on far more grief than he deserves, but the time to have done this was back when the government bought the majority stake in the company.

A second is why the argument stops with the boundaries of AIG. We should ask who beyond AIG would have gone bankrupt if the government did not keep AIG from default, and make the same demands on bonuses that are being paid there.

Think of it this way: If time had not been so tight, the creditors would also have been in the bailout meetings. These creditors would have include those on the hook in the event of default due to their CDS exposure. The meeting would have started off with Paulson saying, "We can pull AIG from the brink. It will take a lot of taxpayer money to do so. We want concessions all around, both from AIG and from its creditors, and especially from those creditors that will go under with it.

That is the correct route to collective punishment. A route that starts with questions like this:

True or False: If AIG had gone into default, Goldman Sachs would also have failed.

Tuesday, March 10, 2009

The Fat-Tailed Straw Man

My Time article about the quant meltdown of August, 2007 started with “Looks like Wall Street’s mad scientists have blown up the lab again.” Articles on Wall Street’s mad scientist blowing up the lab seem to come out every month in one major publication or another. The New York Times has a story along these lines today and had a similar story in January.

There is a constant theme in these articles, invariably including a quote from Nassim Taleb, that quants generally, and quantitative risk managers specifically, missed the boat by thinking, despite all evidence to the contrary, that security returns can be modeled by a Normal distribution.
This is a straw man argument. It is an attack on something that no one believes.
Is there anyone well trained in quantitative methods working on Wall Street who does not know that security returns have fat tails? It is discussed in most every investment text book. Fat tails are apparent – even if we ignore periods of crisis – in daily return series. And historically, every year there is some market or other that has suffered a ten standard deviation move of the "where did that come from" variety. I am firmly in the camp of those who understand there are unanticipatable risks; as far back as an article I co-authored in 1985, I have argued for the need to recognize that we face uncertainty from the unforeseeable. To get an idea of how far back the appreciation of this sort of risk goes in economic thought, consider the fact that it is sometimes referred to as Knightian uncertainty.
Is there any risk manager who does not understand that VaR will not capture the risk of market crises and regime changes? The conventional VaR methods are based on historical data, and so will only be an accurate view of risk if tomorrow is drawn from the same population as the sample it uses. VaR is not perfect, it cannot do everything. But if we understand its flaws – and every professional risk manager does – then it is a useful guide for day-to-day market risk. If you want to add fat tails, fine. But as I will explain below, that is not the solution.
So, then, why is there so much currency given to a criticism of something that no one believes in the first place?
It is because quant methods sometimes fail. We can quibble with whether ‘sometimes’ should be replaced with ‘often’ or ‘frequently’ or ‘every now and again’, but we all know they are not perfect. We are not, after all, talking about physics, about timeless and universal laws of the universe when we deal with securities. Weird stuff happens. And the place where the imperfection is most telling is in risk management.
When the risk manager misses the equivalent of a force five hurricane, we ask what is wrong with his methods. By definition, what he missed was a ten or twenty standard deviation event, so we tell him he ignored fat tails. There you have it, you failed because you did not incorporate fat tails. This is tautological. If I miss a large risk – which will occur on occasion even if I am fully competent; that is why they are called risks – I will have failed to account for a fat tailed event. I can tell you that ahead of time. I can tell you now – as can everyone in risk management – that I will miss something. If after the fact you want to castigate me for not incorporating sufficiently fat tailed events, let the flogging begin.
I remember a cartoon that showed a man sitting behind a desk with a name plate that read ‘risk manager’. The man sitting in front of the desk said, “Be careful? That’s all you can tell me, is to be careful?” Observing that extreme events can occur in the markets is about as useful as saying “be careful”. We all know they will occur. And once they have occurred, we will all kick ourselves and our risk managers and our models, and ask “how could we have missed that?”
The flaw comes in the way we answer that question, a question that can be stated more analytically as “what are the dynamics of the market that we failed to incorporate.” If we answer by throwing our hands into the air and saying, “well, who knows, I guess that was one of them there ten standard deviation events”, or “what do you expect; that’s fat tails for you”, we will be in the same place when the next crisis arrives. If instead we build our models with fatter and fatter tailed distributions, so that after the event we can say, “see, what did I tell you, there was one of those fat tailed events that I postulated in my model”, or “see, I told you to be careful”, does that count for progress?
So, to recap, we all know that there are fat tails; it doesn’t do any good to state the mantra over and over again that securities do not follow a Normal distribution. Really, we all get it. We should be constructive in trying to move risk management beyond the point of simply noting that there are fat tails, beyond admonitions like “hey, you know, shit happens, so be careful.” And that means understanding the dynamics that create the fat tails, in particular, that lead to market crisis and unexpected linkages between markets.
What are these dynamics?
One of them, which I have written about repeatedly, is the liquidity crisis cycle. An exogenous shock occurs in a highly leveraged market, and the resulting forced selling leads to a cascading cycle downward in prices. This then propagates to other markets as those who need to liquidate find the market that is under pressure no longer can support their liquidity needs. Thus there is contagion based not on economic linkages, but based on who is under pressure and what else they are holding. This cycle evolves unrelated to historical relationships, out of the reach of VaR-types of models, but that does not mean it is beyond analysis.
Granted it is not easy to trace the risk of these potential liquidity crisis cycles. To do so with accuracy, we need to know the leverage and positions of the market participants. In my previous post, "Mapping the Market Genome", I argued that this should be the role of a market regulator. But even absent that level of detail, perhaps we can get some information indirectly from looking at market flows.
No doubt there are other dynamics that lead to the fat tailed events currently frustrating our efforts to manage risk in the face of market crises. We need to move beyond the fat-tail critiques and the ‘be careful’ mantra to discover and analyze them.