Bank of America Spent Some Future Money Last Year
Bank of America announced earnings this morning (release, presentation, supplement), and those earnings were disappointing even after stripping out a bunch of weird charges that totaled $1.2 billion. But let's talk about the weird charges, isn't that more fun? Here they are:
Results for the most recent quarter include three adjustments that, in aggregate, reduced revenue in the fourth quarter of 2014 by $1.2 billion (pretax) and lowered earnings per share by $0.07. These adjustments were a $578 million negative market-related net interest income (NII) adjustment, driven by the acceleration of bond premium amortization on the company's debt securities portfolio due to lower long-term interest rates; a one-time transitional charge of $497 million related to the adoption of funding valuation adjustments on uncollateralized derivatives in the company's Global Markets business; and $129 million in net DVA losses related to a tightening of the company's credit spreads.
Debit valuation adjustment (DVA) is a famously weird charge, since Bank of America loses money (for accounting purposes) when its credit gets better. This is a nice countercyclical income cushion, though one of questionable economic practicality, and one that is so confusing that even Bank of America can't quite keep it straight.
But the other ones are a bit less well-known. One is funding valuation adjustment (FVA), which has become popular among all the best banks in recent years. We talked about it at absurd length when JPMorgan adopted it last year. But here is Bank of America's succinct explanation:
In the fourth quarter of 2014, the Corporation adopted a funding valuation adjustment on uncollateralized derivatives in the Corporation's Global Markets business. This methodology seeks to account for the value of funding costs today rather than accruing the cost over the life of the derivatives. The adoption resulted in a one-time transitional charge of $497 million recorded in the fourth quarter of 2014.
The idea is: Your client has a derivative liability to you that it does not collateralize. (Say it's an interest-rate swap where the client agreed to pay you a high fixed rate, and you agreed to pay it a floating rate, and rates floated down and the contract is now worth a lot of money to you.) You hedge this derivative with some roughly opposite trade, because you are not a maniac. And your hedge liability is collateralized, so you are posting cash collateral on one side and not receiving cash on the other side. Despite what people think about global megabanks having access to free money, coming up with this cash actually costs you something. So every day you have to pay a little bit of money to fund your hedge.
How do you account for that? Well, one way to do it is, every day you pay a bit of money, so every day you have an expense. And at the end of the quarter you add up all the money you paid for derivatives funding that quarter, and reduce your income by that amount. Your cost of funding for the quarter is the amount that you actually paid for funding in the quarter.
The other way to do it is to project how much those little daily expenses will come to, over the life of the derivative, and then reduce the "fair value" of the derivative by the projected present value of all those little expenses. And each quarter, you update that projection, and if it's gone up then that reduces your income, and if it's gone down then that increases your income. For accounting purposes, your cost of funding for the quarter is, roughly, the change in the expected amount you'll pay for funding in the future.
Both approaches are currently allowable, but the second approach -- "FVA" -- is becoming more popular. And in the quarter where you start doing that -- start accounting for all your future funding costs, and stop just paying the expense as you go -- then you have a big one-time charge against income, as Bank of America did this quarter.
The other weird charge is the "$578 million negative market-related net interest income (NII) adjustment, driven by the acceleration of bond premium amortization on the company's debt securities portfolio due to lower long-term interest rates." This is not especially weird -- Bank of America does it every quarter, sometimes positive, sometimes negative -- but it's a bit counterintuitive, and unusually large this quarter. The idea is, you own a prepayable bond -- a mortgage-backed security, say -- that matures in 10 years, with a face amount of $100 and a coupon of $5 per year. Because rates are low, you paid $110 for this thing, $10 more than you'll get back at maturity. Accounting requires that you amortize the $10 over the next 10 years, reducing your interest income. (So, loosely, each year you get $5 in cash interest but reduce your interest income by $1 to account for the premium.)
But then rates go even lower, and people get more excited about paying back this prepayable bond, and now you expect it to be paid back in five years instead of 10. So now accounting requires you to amortize the $10 premium over five years instead of 10, roughly doubling the charge that you take against your interest income. When long-term interest rates drop a lot, as they did toward the end of last year, this charge can be big. It's not a cash expense -- I mean, you already paid the $10 -- but it does reflect a certain economic reality.
It's not particularly fun to write about bank earnings, but the philosophy of bank earnings is an endlessly rich field. Last time Bank of America announced earnings, we discussed how abstract and uncertain those earnings were, and we turned out to be right: Bank of America's third-quarter earnings were later reduced by changing expectations about future fines for past conduct. Why are bank earnings so abstract? Why are they so much more abstract than, I don't know, retailer or oil company earnings?
Last quarter I said that "a bank is a collection of contracts that provide for future cash flows," that a bank's earnings statement reflects the change in value of those contracts and that the earnings statement is therefore "just a set of social conventions about predicting the future." But that is not obviously or always true. If you are a very simple bank, and you take deposits and make loans and your borrowers pay you interest and you pay your depositors interest, then your earnings are mostly just about how much money you take in from the borrowers minus how much you pay out to the depositors. Money comes in, money goes out, and what's left is earnings. Just like a regular company.
And of course Bank of America has plenty of that too. Most of its $51 billion of "interest income" is probably, you know, cash it gets for interest, and many of its expenses -- the $11 billion of interest expense, the $34 billion for compensation -- are also probably pretty cash-y, for the most part. But there is a modern trend toward abstraction in finance, and in particular the kind of abstraction where you take things that were once measured by adding up how much cash comes in the door and subtracting how much goes out, and turn them into things that are measured based on changes in their expected value under some valuation model. Your income is less about the cash that your contracts paid you last quarter, and more about the change in the expected value of the cash that your contracts will pay you in the future. You telescope the full lifetime value of the contract into the present moment, as best you can.
So Bank of America lost $578 million this quarter on changes in its expectations of future bond prepayments. And it lost $497 million, not even on changes in its expectations of future derivatives funding costs, but on changing to a future-expectations model of those costs. For a long time, Bank of America -- and most banks -- accounted for derivatives funding costs on a sort of pay-as-you-go basis. But now using the "funding valuation adjustment" -- adjusting the value of the derivative for expected future funding costs -- is much trendier.
There are lots of good philosophical reasons for this! Financial contracts really are, fundamentally, about predicting and modifying the future: They instantiate, bet on, hedge against, whatever, your expectations about the future. If you're just looking at the cash flows from a bank's contracts each quarter, you are missing essential information about the value of the contracts, and thus of the bank. Really, you're missing the purpose of those contracts, and thus of the bank. But if you value the contracts based on their future value -- and if you reflect changes in that future value in the bank's current income -- then you move necessarily into abstraction. A bank's income becomes more the output of a series of models, based on a series of assumptions, and less just a straightforward accounting of money coming in minus money coming out. Its income becomes -- hopefully -- more meaningful, but also more uncertain. And harder to understand.
It's fun to abstract this in a world of zero-ish short-term interest rates, but FVA is really about the cost of funding above the risk-free interest rate. (Or, equivalently-ish, above the rate that you receive on your collateral, e.g. Fed Funds.) So FVA is based on the bank's funding spread, the amount by which its cost of funds exceeds the risk-free rate otherwise used to discount its derivatives.
Plus an adjustment to reflect the difference between your prior expectations of this quarter's costs and the actual costs. And time-value effects.
Oh my lord is this loose. I mean, the whole thing is probably a collection of mortgage-backed securities, the principal amortizes, the amortization of premium is really at a constant interest rate rather than a straight-line dollar amount, and both the pre- and post-rate-drop "maturity" here are just fictions and the real thing is based on repayment speeds, etc. Don't, like, do any accounting based on this. But conceptually it works a little like this.
If you want to do more accounting, consider reading Statement of Financial Accounting Standards No. 91. Paragraph 19, and Cases 3 and 4 in Appendix B, might be of particular interest.
Obviously a ton of comp isn't exactly, with deferrals and vesting and so forth.
This is related to Nick Dunbar's narrative of the run-up to the financial crisis as being about the move from an essentially portfolio approach to credit risk -- make loans, mostly get paid back, write off the ones that go bad -- to an essentially market-based approach -- make loans, buy credit default swaps, mark everything to market.
Of course, it mostly accounted for its derivatives on a fair-value basis. It's not like a swap is reflected in income just based on this quarter's cash flow. It's just that previously FVA was not part of the fair-value calculation, for various reasons.
One of those reasons was an identification of "the risk-free rate" for discounting derivatives with "the funding rate for banks." That identification has been blown apart a bit by the Libor scandal, which made it clear that bank funding rates and the risk-free rate and Libor are all very different things. We talked about this a bit more last time we discussed FVA.
The identification has also been blown up by collateralization, which argues against discounting at Libor, which is after all notionally a bank unsecured lending rate, and for discounting at Fed Funds, which is more like an interest-on-collateral rate. For more on discounting debates in derivatives pricing, try this great Risk article, if you can get past the paywall.
This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.
To contact the author on this story:
Matt Levine at email@example.com
To contact the editor on this story:
Zara Kessler at firstname.lastname@example.org