The Government Accountability Office has issued a long-awaited study on the hidden subsidies that large U.S. financial institutions enjoy. Many will read it as evidence that the Dodd-Frank Act has ended the threat that big banks pose to taxpayers and the economy.
That conclusion would be unduly optimistic.
At issue is “too big to fail,” the idea that some banks are so important to the financial system that the government can’t let them collapse. Because creditors assume they’ll get rescued at taxpayer expense, such banks can borrow for less than they otherwise would. This implicit subsidy, a sort of free insurance, makes disasters more likely by encouraging reckless behavior.
One way to assess the too-big-to-fail problem is to estimate how much of a break big banks are getting on their borrowing costs. Last year, using a study that relied on credit-rating data from 2009, Bloomberg View put the value of the subsidy at $83 billion in a typical year for the 10 largest U.S. banks. In March, using various methodologies, the International Monetary Fund put it at somewhere between $16 billion and $70 billion annually in 2011 and 2012 for eight systemically important U.S. banks.
Enter the GAO, which two senators — Ohio Democrat Sherrod Brown and Louisiana Republican David Vitter — asked in January 2013 to come up with its own estimate. Its economists built 42 statistical models in an effort to isolate the effect of size on a bank’s bond yields. Their finding: The largest banks’ funding advantage may have disappeared or even turned negative in 2012 and 2013.
The GAO gets a different answer largely because it uses a different approach. Estimating the subsidy isn’t easy: It requires figuring out what the banks’ borrowing costs would be in the absence of government support — something that can’t be observed. Researchers employ various tricks to tease it out, all with their own imperfections.
Suppose, though, that the GAO’s methods and numbers are right. Does it mean that Dodd-Frank reforms introduced over the past few years have ended too-big-to-fail? No, for two reasons.
First, the reforms envisioned by Dodd-Frank are far from complete. Regulators have increased capital levels, but not enough to prevent distress. They have established a new resolution mechanism that gives the Federal Deposit Insurance Corp. more power to impose losses on bondholders when banks do fail, but investors aren’t convinced it will work in a crisis. Even the FDIC’s vice chairman has expressed doubt that the mechanism could handle multiple large bank failures simultaneously.
Second, in a period of low perceived credit risk and unusual market calm, you’d expect the subsidy to be low. At such times, as the GAO recognizes, investors aren’t concerned about banks going bust, so they might not give a funding advantage to the ones they think are too big to fail.
To test this thinking, the GAO asked what the big banks’ funding advantage would have been in 2013 if credit risk had been as high as in 2008. Its models found an advantage ranging from about one to six percentage points. Applied to just the uninsured liabilities of the five largest U.S. banks, that would be worth $67 billion to $400 billion a year.
In other words, the GAO study suggests that the subsidy will be there for the big banks if and when the credit cycle turns.
The tendency of banks to look fine in good years but then need bailing out in bad ones has long concerned regulators. The solution is to tell banks to build their defenses during upswings so they won’t need help when the cycle flips. To some extent, the Federal Reserve is trying to do this, issuing warnings to banks that have been making a lot of risky corporate loans. A better approach, which Fed Chairwoman Janet Yellen mentioned in a recent speech, would be to employ countercyclical capital buffers — that is, require banks to finance themselves with more money from equity investors, who share in profits but also absorb losses when times are tough.
All told, the GAO has made an excellent contribution to our understanding of the too-big-to-fail problem. But its report needs to be read cautiously: One thing it doesn’t show is that the problem has been solved.