My friend Who Does This Stuff For A Living points out that my reasonable-sounding question actually has a perfectly straightforward answer: traders have an incentive to “go rogue” only if they’re losing money, and that incentive disappears once they make it back. It’s only the losing positions that keep getting doubled-up.
UBSâ€™s $2.3 billion unauthorized trading loss is back in the news. Yesterday, CEO Oswald GrÃ¼bel accepted responsibility for the loss and resigned.
When the loss was reported last week, Mark asked, â€œWhy donâ€™t rogue traders ever make money?â€
â€œThis should be statistically impossible,â€ Mark suggested. â€œSince risks are symmetric and transactions costs tiny, if there were really trading strategies that could reliably lose billions of dollars, you could make huge sums just by making the opposite bets. This suggests to me that big banks only classify trading activity as â€˜unauthorizedâ€™ when it loses huge amounts of cash.â€
I disagree and think that a simple model of how banks monitor and compensate proprietary traders explains why there arenâ€™t $2.3 billion unauthorized trading gains.
1. Proprietary traders are given a certain amount of capital to trade at the beginning of the year. Thus, if they have gains during the year, they have more capital to trade; if they have losses, their available capital declines.
2. Traders are paid a year-end bonus based on their annual trading profits.
3. Banks have risk control systems that set risk-taking guidelines and attempt to monitor trading. If a trader is observed making larger or riskier bets than is permitted, he may have his capital reduced, he may forfeit any bonus he would otherwise be due, and he may be fired or suspended. The risk control rules also include per-trade and per-period stop-loss limits. If a trader exceeds a stop-loss limit, he is shut down, either temporarily or permanently.
4. The longer unauthorized trading continues, the more likely it is to be detected.
There are three traders at a bank: A, B, and C. Six months into the year, A has trading profits, B is flat, and C has losses. None has engaged in rogue trading thus far. How do their incentives to engage in rogue trading differ?
Trader A: Because A has profits, he has accrued a bonus and is effectively trading his own money alongside the bank’s. Good alignment of interests as far as these things go.
Trader B: Weaker alignment because B has no accrued bonus. If B loses money over the next 6 months, none of the losses comes out of his pocket. However, B earns a bonus on the first dollar of profits.
Trader C: Poor alignment. Because C is under water, he can only earn a bonus if he makes a big profit over the next 6 months. He has an incentive to swing for the fences. And because he has losses, he has less capital to trade, making it very difficult for him to earn a big profit without taking greater trading risks than the bank’s risk control rules permit.
You see why massive rogue trading gains are unlikely? The losing trader, C, has the strongest incentive to cheat, and also the strongest incentive to make big bets. If he starts cheating, then the longer his losing streak, the more he will bet, and the greater the chance he gets caught. Moreover, if C engages in rogue trading and somehow gets back into bonus territory, he has an incentive to stop screwing around and return to making small, authorized trades.
It still may seem puzzling that a trader can lose billions. Losses on that scale require a losing streak that would seem to violate the laws of chance. And financial markets seem efficient enough that there shouldnâ€™t be trading strategies that reliably loseâ€”otherwise, as Mark pointed out, one could make a bundle by adopting the opposite trading strategy. However, UBS-like trading losses tend to be neither random nor systematic in a way that one could reliably profit from taking the other side of the loss-generating trades. The losses are ex-ante random but ex-post systematic.
Imagine a trader who in February 2011 believed that the yield curve would steepen, and he expressed his view by going short 10-yr treasuries and long 2-yr treasuries. He started to lose money, began hiding his bad trades, and doubled down after losses like a Martingale bettor. At each point in time, the outcome of the next bet appeared randomâ€”when the yield on 10-yr treasuries reached 2%, was it obvious the yield would go to 1.7%?â€”so there was no ex-ante free lunch from taking the other side. But with the benefit of hindsight it’s clear that the 7-month flattening of the yield curve was not a random walk.
Now imagine a trader who decided to bet the other way. Ex post, he turns out to have a systematically winning strategy. If he doubled up his winning bets the way the “rogue” doubles up his losing bets, he’d make ton of money for the bank.
But he’d also get himself fired. There’s no way he could conceal from his bosses that he achieved his huge win by grossly violating his trading limits. The laws of chance make huge rogue gains possible. But they’re not incentive-compatible, so they don’t happen.
Gotta love blogging. All you have to do is express a random ignorant opinion, and an expert appears and straightens you out. It’s like the reason a hiker should always carry a deck of playing cards; if he gets lost, he just sits down and deals himself a hand of Solitaire, and someone is certain to come along to tell him to play the black jack on the red queen.