My simple model of How To Be A Bank goes something like (1) amass assets that are numerous and volatile enough to make your management rich and happy and (2) give as much money back to shareholders as you can, consistent with (1). If that were your model and you were building your capital plan what feelings would you feel about this:
The Federal Reserve on Friday kicked off the next round of its annual “stress test” for big banks, releasing instructions on how the process will work.
Included is a new opportunity for banks to alter their proposals to pay dividends or buyback shares before the Fed decides to approve or reject their overall capital plans. … Under the new instructions, banks will have “one opportunity to make a downward adjustment to their planned capital distributions from their initial submissions” before the final decision to accept or reject a bank’s capital plan is made, the Fed said.
I propose a strategy that goes like:
- take your best guess at how much capital you’ll be allowed to distribute, call it $X;
- submit a plan to distribute $2X;
This appears to be a reaction to the sad fate of Citigroup in the last stress test. In the world of the early-2012 stress tests – and here I think that I am 100% accurately representing what the stress tests said though if you are not a connoisseur of stress tests you might find the rest of this sentence a little strange – Citi was a robust and hearty bank that could have easily withstood a severe economic downturn even after paying out $9 billion in capital distributions to shareholders, but it could not do so after paying out $10 billion in capital distributions. You can quibble with the numbers but the concept is perfectly sensible: some amount of money is probably enough to keep Citi safe and some other, smaller amount of money is not.
Citi’s problem was that the 2012 stress tests were a one-shot deal, and that it told the Fed it was going to pay out $10 billion. Since Citi minus $10 billion was not, per the stress test methodology, quite safe, the Fed had no choice but to say no, which in turn left the media no real choice, when you account for the temptation to kick Citi around a bit, but to say things like “Citi failed its stress test, hahahaha/panic!” Months later, Vikram Pandit was out and one widely discussed cause for his firing was that he failed the stress test, which really means not “under his watch Citi was adjudged unable to withstand financial stress” but rather “he asked regulators to give back more money to shareholders than they’d let him.” But it sounds like the former and is thus embarrassing. If, on the other hand, the Fed had given him a second chance, after telling him roughly how far off he was, he could have just dialed back to a $9 billion buyback and gotten a passing grade.1
So one thing to think is that if the 2013 rules were in place for the 2012 stress tests, Vikram Pandit might still be employed at Citi.
Are these changes a good thing? Oh I don’t know. On the one hand, duh, obviously: the Fed and Citi have a joint interest in getting Citi’s capital levels right, though Citi would probably rather err on the side of too low and the Fed on the side of too high. If Citi could afford to give back $9 billion to shareholders, and the Fed would be happy with it doing that, but it was prevented from doing so solely because the Fed had a number in mind and Citi failed to guess that number, that seems stupid and inefficient. Why don’t you just tell me the name of the movie you’ve selected, Fed?
On the other hand if there’s a good place for stupidity and inefficiency this may be it. One great example of a bank and a quasi-regulator working together seamlessly and efficiently is ABN Amro’s collaboration with S&P to get its CPDO deals rated AAA. As Felix Salmon puts it:
a whole slew of S&P alumni ended up at the Dutch bank, putting together deals precisely designed to be the absolutely worst possible product in the world which could still, somehow, get a triple-A rating. In order to do that, you need people who know how the ratings sausage is made — and the easiest way to do that is to simply hire them.
But the second-best way, one which ABN also took advantage of, is to negotiate with them back and forth until you get to their limit: ABN didn’t just lob in a security structure and get a yes or no from S&P; it lobbed in a structure and negotiated it painstakingly until it got the result it wanted.2 If your regulator or quasi-regulator is a black box, it’s quite risky to ship off to them the thing that you think is the absolute worst possible thing they’d approve. But if it’s an iterative process then you’ll almost by definition be able to end up at the worst thing they can live with.
The change is only a partial concession to banks. Bank executives wanted a fuller opportunity to consult with Fed officials before the results are made public. … Fed officials also declined to change their stance not to share the models they use to produce the results.
“Fuller opportunity to consult” = “just put me in a room with them for an hour and I’ll persuade them of anything”; “share the models” = “just give me the model for a week and I’ll figure out how to game it.” When Citi failed this year’s test they made similar noises about how they “plan to engage further with the Federal Reserve to understand their new stress loss models” and “strongly encourage the public release of these models and the associated benchmarks and assumptions.” And I pointed out that you can see what they had in mind: the stress test required 5.0% stressed tier 1 capital, and Citi ended up with 4.9%. If they’d understood the model 0.1% better, they’d have passed!
It’s perfectly rational for the Fed to just not want Citi to know the model 0.1% better. Stress tests, to the extent they have any purpose, are meant to show a bank’s ability to survive some uncertain but alarming future state of the world. Giving them certainty about the model and its results seems like cheating. Optimizing around the disaster model’s weaknesses is not quite the same as preparing for disaster.
Anyway, the Fed’s approach strikes me as a pretty good compromise: a one-shot revision of your capital plan is not susceptible to too much model gaming or persuasion, but avoids over-harsh binary outcomes like Citi’s. But compared to last year, it does represent a movement toward the banks’ desired open-source, fully-negotiated outcome. It’s not all the way there, but that’s okay: it’s a good start to the negotiation.
1. Though with an asterisk – “Fed officials also said that a bank’s original request, and the fact that it would have failed the test, would be released to the public along with the revised capital distribution plan.” This strikes me as a lot less damaging than failing and not being able to do buybacks: the message of a fail is “you were too aggressive”; the message of a pass-with-asterisk is “you were the right amount of aggressive.” Everybody loves the right amount of aggressive.
Also I think this says “we’ll tell you how far off you are before you have to resubmit”:
In a change from prior years, following the Federal Reserve’s assessment of the initial capital plans, CCAR firms will have one opportunity to make a downward adjustment to their planned capital distributions from their initial submissions before a final Federal Reserve decision is made.
2. Incidentally, from reading that CPDO opinion it seems like the negotiations revolved far more around assumptions than structure: not “how high can the coupon be?” but “what volatility should we assume?” That seems like a bad place to be. The stress-test negotiation is at least limited to one axis, and that axis is an economic decision rather than an assumption. The Fed gets to hold banks to its own assumptions instead of being worn down by their lobbying.