The issue came up yesterday when Fed Chairman Ben Bernanke testified before the Senate Banking Committee. Senator Elizabeth Warren cited a Bloomberg report that put the number at $83 billion to the 10 largest U.S. banks. The Bloomberg figure is extrapolated from the finding of an IMF study that the backstop provided to banks lowers their cost of borrowing by approximately 0.8 percentage points.
Matt Levine at Dealbreaker makes the provocative claim that “The Too Big to Fail Subsidy is Negative Sixteen Billion Dollars”. This comes in the second round of Levine’s tit-for-tat with Bloomberg. His original critique started off with a reasonable and incisive drill down into the numbers. Now, after an effective rejoinder by Bloomberg, he abandons the two main points from his original critique and substitutes new ones.
First, he compares the interest expense as a % of total liabilities at the five largest banks against the ratio at five smaller banks. Lo and behold, the interest expense is higher at the five largest banks. The conclusion: “The too-big-to-fail banks are subsidizing us!”
But wait a second. The business models of the two sets of banks are entirely different from one another. A simple comparison like that is specious.
Levine knows this. Right after he drags us through the calculation and draws the conclusion, he then announces that “That’s silly of course…,” as he now takes note of the contrasting elements of those two business models.
This transitions him to his second, entirely different point which is that “the TBTF banks use long-term, expensive funding to appropriately cushion themselves from the risks of their capital market businesses.”
The trouble is that he has no evidence for this whatsoever. All he shows us is that the two sets of banks are different. There is no systematic reckoning done to establish what would be an appropriate funding model for either type of bank. There is no benchmark on which to lay the claim of an “appropriate cushion”. There’s nothing. A critique that began by taking the high ground on using the right figures for comparison and properly taking into account differences across a sample set, now abandons all burden of proof.
Probably the fault lies with me for taking Levine’s post too seriously. It’s full of snark and sarcastic hedges. In fact, right after putting out the claim that the subsidy is negative, the headline coyly toys with us saying “…Or Possibly Some Other Number.”
Definitely some other number. As Levine had noted in his earlier, more serious post, there is research out there. And none of it, to my knowledge, gets a negative number.
 He had two main points. First, the 0.80 percentage point figure is an average across banks with different balance sheets and credit ratings, while the banks covered in the Bloomberg report do not match that average. Levine revises the cost of borrowing advantage down to 0.31 percentage points. Second, the funding advantage documented in the IMF study only applies to long-term debt issued by those banks, whereas Bloomberg applied that number to all of the banks’ liabilities. Bottom line: Levine’s adjustments produce a revised subsidy figure of $3.7 billion.
Bloomberg replied to both criticisms. On the first point, it accepted Levine’s contention that the banks it examined were not average, but dug deeper into the detailed figures to arrive at the conclusion that the right figure was therefore a 0.50 percentage point gain. On the second point, it accepted, too, Levine’s contention that a bank’s different liabilities might enjoy differently sized benefits from the taxpayer backstop, and it cited an FDIC study documenting the funding advantage on certain types of deposits.