I have nothing particularly useful to tell you about Citi's earnings - they were good, yay, well done Vik, one day maybe you'll be able to pay a dividend - so let me ask you some useless things. My favorite useless thing is DVA, which is the thing where if you are a bank you "lose" "money" when your credit improves and you "make" "money" when your credit gets worse, which is in some ways the opposite of right though also not, like, totally away from reality. Citi suffered thereby for its virtue:
Citigroup reported improved first-quarter earnings on Monday, with steady growth in the bank’s globe-spanning consumer businesses and a rebound in investment banking from a poor previous quarter.
Net income was $3.4bn in the first quarter compared with $3.2bn a year earlier as revenue grew just 1 per cent to $20.2bn. Those measures exclude the impact of so-called “debt valuation adjustments” – an accounting rule that makes companies take gains or losses from swings in the price of their own debt. On a reported basis, including DVA, Citi’s net earnings were down at $2.9bn.
So three useless thoughts/questions for you on that:
(1) WTF guys:
Citigroup Inc. (C), the third-biggest U.S. bank, said profit fell 2.3 percent, missing analysts’ estimates on a bigger-than-projected accounting cost.
Why is it so hard to predict DVA? That's $0.95 of actual GAAP earnings vs. IBES of $1.00. You'd think that the main inputs (credit quality changes, balance sheet) are sort of knowable. So I went and tried to figure it out, and this is what I came up with:*
So, yeah: Citi really did have a larger-than-expected DVA loss here, by call it 15 cents; it also had a smaller-than-run-rate DVA impact last quarter. Now I'm willing to agree with everyone else that ha ha accounting anomaly not a real result of operations etc. But this is not just about Citi's spreads getting tighter and that affecting earnings in some mechanical way: Citi's big DVA losses this quarter represent not only accounting noise but also a pretty big increase in the sensitivity of its balance sheet to its own credit - higher than recent quarters, and much higher than last quarter. Does that mean that it's moving to a longer-duration balance sheet? That would be nice - but then what's up with last quarter?
(2) Why is Citi Holdings the wrong way?
Citi Holdings revenues of $874 million in the first quarter 2012 were 47% below the prior year period. Excluding CVA/DVA, Citi Holdings revenues were $786 million, 53% lower than the first quarter 2011.
So Citi Holdings made $88 million of DVA, whereas Citi as a whole lost $1,288mm on DVA, with ($1,376mm) from the investment bank and +$88mm from Citi Holdings as the only components of that. Citi CDS spreads pretty monotonically declined over the quarter so I don't think it's a timing thing. That leaves you with: Citi's "bad bank" somehow made a profit from Citi's credit getting better, which is the opposite of how it usually goes. Did Citi Holdings credit somehow get worse during their wind-down? Or does Citi Holdings have some Citi exposure that is reflected here? Anyway, I don't know, but it's weird, talk amongst yourselves.
(3) One thing that I idly thought to myself was "boy, it's a shame Citi lost so much on DVA, otherwise maybe it would get to that dividend." (As it is it looks like they'll be waiting until 2013.) This of course is too glib; DVA losses are not the thing that's going to push Citi's capital too low to return anything to shareholders - e.g. Basel III doesn't let you count DVA gains in income to determine capital levels (see paragraph 75). Which is pretty obvious: being a worse credit shouldn't be a reason to return more capital.
Or should it? Last week I finally read Engineering the Financial Crisis,** which is good fun and persuaded me of the benefits of countercyclical capital requirements.*** Including DVA in earnings for capital purposes is I submit to you a pleasingly automatic, countercyclical thing with immediate effect: if Citi's CDS tightened by 85bps last quarter, maybe that means that it's only one quarter away from much wider spreads / just a bit too early to pay dividends? And if your CDS widens out because everyone's panicking about you, maybe that's exactly the time that regulators should cut you a break and let you tell everyone how well capitalized you are?
* This is pretty dumb but the notion is, DVA should be something like (a) relevant debt times (b) change in credit spreads times (c) some sort of duration factor. Relevant debt is unclear; I used trading account liabilities plus long-term debt because Citi reports that a lot of its DVA comes from fair-value-option debt rather than derivatives liabilities. So it's some subset of that so pretend it's a constant subset. Change in credit spreads I used Citigroup 5-year CDS, which is debatable. Some sort of duration factor is what I called "normalized DVA gain," just (i) DVA gain (loss) divided by (ii) CDS change divided by (iii) total relevant debt times (iv) 10,000. If you assume that something like .22ish is a normal factor here, then the zero-ish last quarter and .33 this quarter both look odd.
** This is sort of an amazing book. I don't personally think it's like the A+ #1 explanation of the financial crisis or anything, but it's compelling and at the very least every regulator should read it as an ode to epistemic modesty. Also it's full of these wonderfully dry and catty attacks on prominent economists, and who can resist that?
*** It did not intend to persuade me of that, though, which may say something about the book.