• 07 Mar 2013 at 6:20 PM
  • Banks

Fed Kicks Off Awkward Week For Banks

One of the nice things about last year’s Fed bank stress tests was that they were released, and everyone was like “OMG Citi failed!!,” and then we all calmed down and realized that all that meant was that Citi’s capital return plans had failed, so it couldn’t launch a big share buyback, but it wasn’t going to be smashed into dust as a warning to its compatriots. That turned out to be cold comfort for Vikram Pandit but was soothing for the rest of us. This year, in part to avoid the Vikram thing, the rules have changed: today the straight-up stress test results were released, while the Fed will approve or reject capital return proposals next week, and there’ll be a lot of weird disclosure gamesmanship in the interim. Early signs point to Citi being out of the doghouse, and Goldman possibly being in it.

Also Ally Bank failed, sorry! Legit failed, not failed pro forma for capital return. So, smashed into dust.

Here is a chart you may or may not find amusing:

This chart is intended to answer the question: how many a 1-in-100 terrible days would these banks need to have in order to get the Fed’s estimated trading losses?1 The answer is somewhere between one and two hundred top-1%-terrible days between now and Q4 2014, which by my watch is seven quarters or ~440 trading days from now.

Are you reassured? I mean, don’t be, nobody believes VaR, what are you nuts? The basic interpretation of this should be something like: the world looks scarier in the Fed’s “adverse scenario” than it does in the banks’ VaR estimates, not because the Fed’s adverse scenario is banks having regular horrible days over and over again for six months, but because the adverse scenario involves tail risks and correlations that would make these VaR numbers useless. That’s as it should be.

There’s a potential advanced interpretation, which is something like: why does Goldman’s value-at-risk look twice as risky as Citi’s? Various possibilities present themselves, ranging from “different banks have different portfolios and the Fed stresses different types of risks differently to reflect their actual macroeconomic scenario,” through “Goldman is better at optimizing VaR models to minimize risk-weighted assets than it is at optimizing for the Fed’s adverse scenario,” up to “the Fed is cutting Citi a break after last year’s unpleasantness.”2

Anyway the key stress test results are here:

With 5% tier 1 common and 3% tier 1 leverage ratios being particularly worthy goals for those seeking capital return. Note in particular the relatively lame performance of GS and MS, who both scrape close to the 5% minimum for tier 1 common. What those banks have in common, I suppose, is that they’re not really banks; they’re investment banks awkwardly shoehorned into banking regulation. I occasionally go around saying things like “traditional banking activities like lending are the riskiest activities!” but the Fed obviously disagrees:

Of the $462 billion of adverse-scenario losses among these eighteen banks, the Fed estimates that $97 billion will come from trading portfolios. Only five of those banks even have material trading portfolios.3 I suppose those five banks are a bit sadder about that fact today than they were yesterday.

Dodd-Frank Act Stress Test 2013: Supervisory Stress Test Methodology and Results [Fed]
Fed Stress Tests Show 17 of 18 Banks Weathering Severe Recession [Bloomberg]
Fed’s Stress Tests Point to Banks’ Improving Health [DealBook]
Stress test results: Banks could lose nearly half a trillion dollars [Fortune]
Goldman Sachs and Morgan Stanley Near Bottom of Stress Tests [NetNet / John Carney]
Goldman exposed to $20bn loss in a crisis

1. So this uses reported average 2012 VaR numbers from the banks’ 10-Ks, grossing up 95% VaR numbers to 99% stupidly (NORMSINV(.99)/NORMSINV(.95)) where only 95% numbers are reported. The dark blue bars represent the Fed’s “Trading and counterparty losses” divided by 99% VaR; the light blue bars represent “Trading and counterparty losses” plus “Realized losses/gains on securities (AFS/HTM)” divided by VaR. Some calculations and references are here. Because of differences in how these things are reported, who puts what credit exposures where, how hedging VaRs etc. are calculated, these things are to some approximation meaningless, but whatever.

2. These answers are all undermined by the laziness of my math, which is really atrociously lazy. If you wanted to, you could try to actually, like, break out the Fed’s stress scenarios and their impact on credit, rates, equities, etc.; compare those to the components of VaR, do a better job of identifying trading vs. loan/etc. VaR in the banks’ reports, etc. etc. Or you could not. I didn’t.

3. Oh, six, Wells Fargo you can come too, why not.

3 comments (hidden to protect delicate sensibilities)
Show all comments ↓

Comments (3)

  1. Posted by quant me maybe... | March 7, 2013 at 8:36 PM

    After reading this Matt. I have to conclude that you are the Hunter S. Thompson of the finance world.

    I imagine that this is your typical day:
    <img src="http://i1231.photobucket.com/albums/ee514/audiem2491/hunter-day-in-the-life_zps831d37ba.jpg&quot; /img>

  2. Posted by TFTY | March 7, 2013 at 9:24 PM

    Replace cocaine with making graphs and you nailed it

  3. Posted by Level 3 CFA | March 8, 2013 at 1:44 PM

    close but not quite Matt. 99% VaR measures the estimated max trading loss 99 days out of 100. It doesn't tell you how much will be lost on that trading day where you exceed VaR. So you can't add up the VaR and conclude they would need 600 consecutive days of exceeding VaR to reach the Feds stress test levels. The 1987 stock market crash was a -27 standard deviation event (99% VaR is a -2.33 standard deviation event)