Fed Stress Tests: More Mess than Early Warning

Imagine if federal rules required power plants to tell the Environmental Protection Agency how much carbon they wanted to emit, and then the EPA ran a complicated, secret model that calculated just the right amount of carbon for each particular plant.

Plants that asked for the "right" amount, could proceed. But any plant that requested more than the EPA's estimate would be rejected.

This "Price Is Right" approach — bid up to, but not over, an amount that you have no way of knowing — is how the Federal Reserve's stress tests are conducted, and it makes no sense.

The capital stress tests the Fed just administered to the country's 19 largest banks are too complex, costly, time-consuming and redundant.

Here's how the process works. The Fed draws up numerous mind-bendingly awful assumptions about what could happen to the economy, employment, home prices and the like. It gives these assumptions to the largest banks. After building sophisticated, expensive models — that their examiners must sign off on — the banks "stress" their portfolios according to the Fed's assumptions and calculate aggregate loss figures. Bankers tell me this exercise takes "tens of thousands" of man-hours.

A bank then figures out how much capital, over and above the regulatory minimums, it would have after subtracting those losses, and tells the Fed how it plans to use that excess capital.

Then the Fed repeats the exercise using its own models.

It plugs a bank's data in, applies the assumptions to provide the "stress" and then estimates a bank's losses. The Fed then figures out how much excess capital a bank has. If the plan the bank submitted to buy back stock or pay dividends exceeds that amount — even by a little bit — the Fed rejects the bank's plan.

Unsettling headlines are written about how a bank "failed" the Fed's test, investors sell the stock, confidence is shaken. The bank's executives rush around trying to figure out how to contain the damage.

All this happens because a bank overestimated how much capital it would end up with after an economic apocalypse? And in the case of Citigroup — the biggest victim of this "process" — it overestimated by a tiny amount.

According to the Fed, Citi would have ended up with a 4.9% Tier 1 capital ratio after subtracting the capital it wanted to pay shareholders. The threshold for "passing" was 5%.

"There is so big a difference between 4.9% and 5% that franchise-significant strategic mandates are imposed?" asked Karen Shaw Petrou, co-founder and managing partner of Federal Financial Analytics. "The Fed is so convinced that its tests are so perfect that it in fact issues edicts based on 10 basis points?"

Remember, Citi did "pass" the test. The bank even had some amount of excess capital, just not as much excess as the Fed wanted. Still, the Fed not only turned down the bank's request in total, it did not tell Citi how it calculated its number or give the bank a chance to appeal the result.

"If they are going to override your results anyway, what's the point?" asked Clifford Rossi, a banker turned academic at the University of Maryland. "If the Fed's intention was to say, 'Look, we want to instill a sense of confidence among the American public and investment community in these banks in this difficult time,' there are better ways to do that."

Requests to interview a Fed official about how the process might be improved went unanswered. For all the central bank's efforts to become more transparent, it's still hard to get someone on the phone to field questions it would rather not be asked.

It'd be simpler to have the Fed do these tests and leave it to the banks to stress their portfolios however they want. But the Dodd-Frank Act requires two tests: one by supervisors and another by the companies. So there is not much to be done about the redundancy.

The law, however, does not require this "Price Is Right" guessing game. So the Fed could simply take the banks' data, run its supervisory model and tell the banks how much capital they have above the regulatory requirements. Then the banks could decide what to do with that excess capital. It would be simple and, if not totally transparent, at least more so than the current setup — and a whole lot less likely to trigger scary headlines.

The Fed ought to make its model public. If it shows dramatically higher losses than a bank's own model projects, shouldn't the bank know why? I have no idea if the Fed's model is better than the bankers', but I don't understand keeping it secret.

Rossi said the Fed should invite an independent panel of experts to vet its model.

"The model would be revealed for all its pluses and minuses," he said. "The banks would submit their data, the Fed would run its model and then they would have a conversation. It would create more transparency and it would reduce regulatory burden for the banks themselves."

Stress testing is good. It can be even better.