The story so far is that a few days ago Bloomberg View claimed that the ten biggest U.S. banks got an annual subsidy of $83 billion from being too big to fail. That claim seemed silly to me, and I said so, and this weekend Bloomberg responded to that post saying, and I quote, "we weren't kidding." Apparently the people who keep the blogging rulebook believe that I now have to write a post in response to their response to my response to their original claim, and so this is that post. Actually this is that footnote, whatever.1
Up here let's be super super naïve and just ask: how much do too big to fail banks pay to fund their balance sheets, and how much would they pay if they were smaller and failier and less government-supported? One dumb way to go about answering that is to actually just look at the cost of funding of some banks. We can start with the big five that Bloomberg uses - JPMorgan, BofA, Citi, Wells Fargo, and Goldman - and compare them to some smaller banks. Since Bloomberg seems to believe that Fitch believes that the TBTF banks would be rated around BBB- were it not for their TBTF-ness, we can compare them to some banks rated BBB- by Fitch. I chose five BBB- rated bank holding companies pseudorandomly from Fitch's web page: Associated Banc Corp, TCF Financial Corp., First Horizon National Corporation, First Niagara Financial Group, and Zions Bancorporation.2 Then I just looked at how much those banks paid for their funding (interest expense, preferred dividends), compared to how much the big five banks pay.
Here are some average numbers:
Data and calculations are here. Now this oversimplifies a lot of things, of course, but as some wise people once said, "Our experience with such calculations has taught us that the simple approach typically gives you pretty much the same answer as the complicated approach."3
The differences are striking, aren't they? The small banks average less than 2% of the assets of the big ones, and at BBB- are rated 4 to 6 notches lower. But their blended average cost of funding is 20 basis points lower - not 80 basis points higher, as Bloomberg claims - than that of the big five too-big-to-fail banks.4 For the five banks combined, that 20bps of assets works out to over $16 billion per year. The too-big-to-fail banks are subsidizing us!
That's silly of course: the reason JPMorgan and BofA and Citi pay more for funding is some combination of
- Some of their funding sources don't trust Fitch's ratings implicitly, and think that funding Citi is riskier than funding a small boring BBB- rated regional bank, and
- They rely on different sources of funding: instead of just "issue checking accounts and make loans," they fund using repo and wholesale deposits and CP and relatively more long-term debt than the regional banks do, and those sources of funding have a higher cost than, y'know, checking accounts.5
The first source of the TBTF "penalty" that we've identified suggests that Bloomberg's approach - which, at base, relies on Fitch's analysis of (1) how safe banks are and (2) how much of that safety derives from government support - is wrong. Fitch may see more of a too-big-to-fail subsidy than the market does. Sometimes ratings agencies make mistakes.
But the second is more important. If you take Bloomberg's approach at face value, then JPMorgan pays 80bps less for long-term debt than Associated Bank Corp does (2.2% vs. 3%, to make up numbers), and 80bps less for six-month CDs (0.5% vs. 1.3%, again, to make up numbers), and 80bps less for checking accounts (0.0% vs. 0.8%), and so on down the line. But that would only mean that JPMorgan pays 80bps less for its funding than Associated Bank Corp does if JPMorgan used the same mix of funding.
But it doesn't. JPMorgan has way more long-term debt - something like 10% of its assets. (Goldman has 17%.) It relies less on checking account deposits and more on capital markets. And it ends up paying about 8 basis points more, not 80 basis points less, than Associated.
What does that mean? JPMorgan will tell you - at their investor day today, for instance - that it means they're conservatively funded with, you guessed it, a fortress balance sheet:
This may not be the whole story - arguably consumer checking accounts, for all that they are demand liabilities, are a more conservative source of funding than repo lending. But it's not wrong, either: the big capital markets banks, whether due to personal upstandingness, regulatory pressure, or market discipline, really do seem to fund longer-term, more conservatively, and more expensively than their non-too-big-to-fail brethren. Perhaps if they lost access to implicit government support, they'd find it more expensive to issue long-term debt. Perhaps then they'd stop, and rely more on deposits.
And then perhaps they'd shut down their capital markets businesses and become Associated Banc Corp.? And maybe that'd be a good thing? If you want to break up the banks, by all means, keep wanting to break up the banks. (Good luck!) But don't do it on the basis that the big banks recklessly take TBTF-subsidized deposits and gamble them in capital markets. The opposite - that the TBTF banks use long-term expensive funding to appropriately cushion themselves from the risks of their capital markets businesses - seems closer to the truth.
Remember That $83 Billion Bank Subsidy? We Weren't Kidding [Bloomberg View]
earlier: Why Should Taxpayers Give Big Banks A Subsidy of $83 Billion Per Year, Or Any Other Made-Up Number For That Matter?
Some math [Google Docs]
1.THING ONE: Bloomberg based its calculations on a paper which in turn relied on Fitch's overall credit ratings and "support ratings," which reflect Fitch's estimate of how much TBTF support each bank gets. I expressed skepticism that Fitch's views of credit generally and government support in particular are either (1) right or (2) accepted by the market broadly. Bloomberg did not address that skepticism. WINNER: ME.
THING TWO: The paper Bloomberg cited attempted to calculate the uplift that the big banks get from expected government support on five-year unsecured bonds. Bloomberg applied this to all liabilities, including things like FDIC-insured deposits, overnight secured funding, etc. I suggested that an 80bps spread (the number they came up with) on five-year unsecured bonds does not necessarily translate into an 80bps spread on overnight secured paper, or FDIC insured deposits. Bloomberg's response is to cite another paper finding a funding advantage to too-big-to-fail banks on uninsured deposits too, from 2005 to 2010.
I have not read the paper closely but I note again that its conclusion sounds odd in a world where 1.2% is a very rich interest rate to get on your deposits. But as Bloomberg say, "The purpose of our analysis, and the study, is to estimate the long-term value of the too-big-to-fail discount," so the fact that it has nothing to do with conditions existing in the present, recent past, or near future does not deter them. I also note that it's quite strange to believe at the same time that (1) non-TBTF banks pay an 80bps premium on five-year unsecured debt and (2) they pay a 120bps premium on demand deposits. That is quite an inversion of the imaginary-TBTF-premium curve.
In any case that applies only to certain types of uninsured deposits, not the rest of the banks' balance sheets. So I feel okay here? Still, I confess I overreached in doing my contrarian math of saying that the TBTF premium applies only to long-term unsecured bonds - surely there's some cost savings for things like deposits and repos - though I doubt it's 80bps. As they say, "it's true that including deposits might skew the results. Problem is, so would excluding them." That's not wrong. I'd call this one a draw.
THING THREE: They computed their 80bps TBTF subsidy based on the average subsidy to any TBTF bank worldwide, as computed by the paper they relied on. I pointed out that the U.S. TBTF banks they actually looked at got a much more modest subsidy, according to that same paper, just because they were highly rated to begin with and, as the paper said, “government support is more ‘valuable’ at lower rating levels.” The right number, I thought, was more like 30bps.
They have two responses. The first is that the big five U.S. banks got six notches of uplift, from BBB- to AA-, which implies 50bps of uplift, closer to my 30bps than their 80. This is based on 2009 data; only one of those banks, Wells Fargo, is actually still rated AA-, with the rest in the A range with as far as I can tell four notches of uplift (from BBB- to A'ish). I think I win this one, unless you really care about 2009 much more than 2013.
The other response is that those numbers come from a very long-term premium for debt ratings:
Because we're focusing on the U.S. and because the experience of the 1920s isn’t necessarily a good indicator of what will happen in the coming years, we might want to use a more relevant measure. Consider the difference between two Bank of America Merrill Lynch indexes that track the yields on actual AA and BBB bank debt in the U.S. Over the 10 years through early 2008, the average gap was 1.13 percentage points. From this perspective, our blind use of 0.8 looks conservative.
This seems fair to me: even as I did it, it struck me as a little silly to claim that the difference between an A and BBB- bank was 30bps on senior unsecured debt, even though that's what Bloomberg's source paper said. If you use actual current spreads, it gives them some breathing room to be wrong on some stuff, since those spreads are wider than historical levels.
2.Methodology: go to Fitch web page, search for U.S. banks rated BBB+/BBB/BBB-, choose first five BBB- names that sounded clearly BHC-y (as opposed to bank-subsidiary-y). I'm willing to believe that this is an unrepresentative sample but so it goes.
3.Bloomberg View, I don't want to tell you how to live your life, but: never say that! It's self-evidently wrong, for one thing ("when I design an airplane, I find ..."), and it has a whiff of math-is-hard, for another, but most importantly: my experience with such (?) calculations (?) has taught me that there are lots of simple approaches and they all contradict each other. My approach in the text is, I submit, simpler than the approach of relying on an econometric paper using Fitch ratings, and it gets the opposite result, so I win, or something.
4.That is, their cost of non-common-equity funding. I am not a huuuuuge believer in the idea that common equity has a high and quantifiable cost but whatever. The non-TBTF banks run an average of 131bps less common equity, so if you're a deep believer in the cost of equity then you might worry about that. What's the cost of equity for banks? Is it 10%, as some people seem to believe? Then the TBTF penalty is only like 6 bps:
5.Also the arithmetic fact that the small banks have more equity and I'm implicitly assigning a zero cost to common equity in the main text. If you charge for equity it's more like a push than a penalty.