Saturday, March 20, 2010

Debunking Gary Gorton's "Fire Sale" Thesis

A few weeks ago I looked at Gary Gorton's Somewhat Flawed Take on Shadow Banking. One big problem I had with his paper (which now has been enshrined as part of the public record at the Financial Crisis and Inquiry Commission) is his analysis of a very curious phenomenon during the first half-year of the worst part of the crisis. Namely, at several points, AAA rated corporate bonds paid bigger yields than AA rated. This is a real "upside down topsy turvy" kind of thing. Bonds that achieve a triple AAA rating -- the highest possible -- are supposed to be really safe, almost U.S. Treasury safe. AA, the next ranking down, denotes a bit more risk. And a bit more risk means that investors demand a higher yield.

But over two periods -- one in the fall of 2008 and the other early in 2009 -- investors were getting a dramatically lower yield (well, dramatic in the bond world) for AA than AAA. That's Alice in Wonderland stuff, it seems.

Gorton explains the market weirdness on page 13 of his paper to the commission. (First, here's background you need: just before the excerpt below, he was discussing the sudden withdrawals of funds from the shadow banking system, withdrawals that took the form of repo haircuts on structured debt -- if you're lost by that last sentence, see my earlier post on Gorton's paper, linked above):
Faced with the task of raising money to meet the withdrawals, firms had to sell assets. They were no investors willing to make sufficiently large new investments, on the order of $2 trillion. In order to minimize losses firms chose to sell bonds that they thought would not drop in price a great deal, bonds that were not securitized bonds, and bonds that were highly rated. For example, they sold Aaa‐rated corporate bonds.

These kinds of forced sales are called “fire sales” – sales that must be made to raise money, even if the sale causes to price to fall because so much is offered for sale, and the seller has no choice but to take the low price. The low price reflects to distressed, forced, sale, not the underlying fundamentals. There is evidence of this. Here is one example. Normally, Aaa‐rated corporate bonds would trade at higher prices (lower spreads) than, say, Aa‐rated bonds. In other words, these bonds would fetch the most money when sold. However, when all firms reason this way, it doesn’t turn out so nicely.

The figure below shows the spread between Aa‐rated corporate bonds and Aaa‐rated corporate bonds, both with five year maturities. This spread should always be positive, unless so many Aaa‐rated corporate bonds are sold that the spread must rise to attract buyers. That is exactly what happened!!
When I read this, his "fire sales" thesis didn't smell right. Before we see how it falls apart -- I chose to look at a puzzling period in March 2009 when it appears that the worst outbreak of "fire sales" occurred, according to his own graphic -- let's make a couple of stops.

Why does this matter anyway? This "fire sale" section is buried on page 14 of his presentation.

Because -- and this is very important -- there are two dominant views of this financial crisis. They are so critical to shaping perception, they are practically worldviews. Depending on which you hold, you're likely to propose a different set of policies to revive the financial system. They are:

At the heart of this crisis was a liquidity crunch. The banks peddled this line furiously. Credit markets seized up and poor bankers were innocent bystanders, stuck with valuable assets they could only sell at ... here it comes ... fire-sale prices. The market was valuing their assets, worth 90 cents on the dollar, at 40 or 50 cents! If you believe this: you probably thought that Geithner's PPIP proposal made sense and was going to work, and that the big banks just needed a little breathing room and an infusion of liquidity (which the Fed happily provided in spades) and they'd get back on their feet and start lending normally again.

At the heart of this crisis were plain ol' crappy, overvalued assets. Their values had become inflated by a combination of loose money spurred by too-low Fed interest rates, rating-agency complicity, opaque markets (RMBS, CMBS), and maybe a dash of housing-bubble mania. If you believe this: you probably thought that tougher measures needed to be taken with the big banks, that these crappy assets had to be honestly accounted for and cleared at market prices, and that large swaths of the banking system were insolvent.

Notice most smart observers would assign some truth to both; they are not mutually exclusive concepts. However, people tend to align themselves with either the first or second idea. I am in the second camp, firmly, and Gorton seems to be more in the first.

Okay then: why doesn't Gorton's explanation smell right in the first place?

This is sort of interesting. Let's go meta-finance for a moment.

There are many classes of assets. And there is money, the ultimate liquid asset, extremely flexible and wonderfully fungible. One way of looking at assets is to see how "money like" they are (I'm not precisely talking about liquidity here, so I'm going to stick with "money like"). By "money like" I mean how much "objective value" (as expressed in a unit of money, say dollars) the asset contains. So if we all were guaranteed $2 for any loaf of bread (that met certain criteria of course, pertaining to such attributes as dimensions), and the bread could be redeemed at "bread banks" (a groaner, I know), loaves of bread would be very "money like."

Okay, let's take Beanie Babies. Gong! Not very "money like" at all. At the height of the fad, you might receive $50 for a rare stuffed penguin that might not fetch 15 cents a year later. Let's slide quickly down the asset scale. House -- more "money like." Share of stock: even more "money like." A plain-vanilla corporate bond: very "money like."

For what is an ordinary "bullet" corporate bond? A series of interest payments, usually semi-annual, and then a lump sum spit out at the end of its lifespan of maybe 5, 10, 20, or even 30 years. And these payments are all made in ... money. Bonds don't pay you in Toblerones, iPads, or dental floss. It's all money, money, money.

Now what's a simple way to think about a bond's coupon? Say IBM's 5-year plain-vanilla bond pays 4 percent interest and is rated AA. Part of that coupon -- let's say 2.5 percent -- reflects interest rate and inflation risk over the five-year life of the bond. The remainder -- and this is where it gets interesting -- is for credit risk. Since IBM is perceived as really safe, it pays out 1.5 percent interest for credit risk. If it were Johnson & Johnson, let's say, and rated AAA, and perceived as really, really safe, it might pay only 1.3 percent.

Now which of these bonds would you, Joe Investor, buy? You might think the IBM at 4 percent instead of the Johnson & Johnson at 3.8 percent. The IBM pays more, right? But that company also stands a higher chance of defaulting on its debt. There's no free lunch. You take more risk; you get more money.

But what if the Johnson & Johnson, with its AAA rating, paid 4 percent, and IBM, a whole grade lower, paid only 3.8 percent. That's a no brainer. Bonds are very "money like," remember? You don't "like" a bond more because it's prettier, has a dormer window on the third floor, or would look good on a pendant. You like it for its money-ness. So you'd snap up Johnson & Johnson at 4 percent and marvel at your wonderful luck.

Which brings us to our main subject ...

Why did yields on AAA corporates balloon out over AAs if that doesn't make any sense? Was it indeed a "fire sale" as Gorton claims, a mass rush to the exits, with perfectly good AAAs being chucked overboard at whatever price the market would offer?

I did some research. I did so knowing that Gorton's thesis would stand under three conditions: (1) the AAA corporates were "true" AAAs and not lower grades masquerading as AAAs (2) the universe of AAAs was reasonably diverse in nature (as in, not skewed to one particular industry) (3) there was a reasonably large number of AAA corporates trading (as in, you might have a diverse selection of bonds, but if there are only six of them, that's too few and the volatility of small numbers could explain the strange inversion between AAA and AA yields.)

I could have looked at the period in late 2008, but instead I chose an even more dramatic one, where an even larger inversion occurred (even more "fire sales"!). If you look at Gorton's graphic on page 14, you'll see the icicle-shaped spike in early March of 2009. Briefly, AAA corporate bonds paid more than 2 percentage points more than AAs! That's a huge spread. Gorton would have us think that investors were just dumping truckloads of perfectly good AAAs and the market, facing a glut of the bonds, irrationally pushed their prices way down (and their yields up).

So I looked behind the curtain, using a handy Bloomberg machine. And what I found was nothing like what he suggested was going on.

First, using Merrill Lynch indexes, I saw evidence of the inversion easily enough. The components of Merrill Lynch's AAA index may not exactly match what Gorton looked at, but the curves seem pretty close, so I think we're on fairly solid ground (as you'll soon see, there aren't a lot of AAAs in the first place, so I'm confident there's large overlap in both our data sets; I may graph my numbers later and drop in that visual here). Also I focused on AAAs because that's where the real story is: while AA yields did creep higher during that period, the AAAs zoomed past them and were mainly responsible for creating the gap.

And what a gap it became: AAA pulled ahead of AA by 11 basis points on March 2. That spread widened out to 191 on March 4, hit a high on March 5 of 195, then subsided to 135 on March 10 and fell to 74 on March 13. (Note: there are 100 basis points in 1 percent, so 195 basis points is 1.95% -- which may not seem like much, but in bondland, between investment-rating grades, it's huge.)

Then the meat of the investigation began. Who was in the AAA corporate index at that time? A diverse, large number of companies?

Nope. Not by a long shot. Not many companies win a AAA rating, understandably, only the best of the best. I went through a half-dozen screens of the bonds in the Merrill Lynch index as of March 6, 2009 (the screens show a company name, and the specific bond that belongs to the index). And guess what? To call it an index at that time was pretty much a misnomer. It was General Electric.

General Electric bonds (or those of its financing arms) accounted for 71.5 percent of the index weighting. Berkshire came in second at 8.9 percent. That leaves less than one-fifth for everyone else (and there weren't a lot of others, though it hardly mattered because they made up so little of the overall weighting anyway).

(Note: a careful reader of Gorton will see that he looks at only 5-year bonds. Even so, it turns out that when I filtered for that criterion, it mattered little: GE's weighting drops a little, but only to 68 percent, and Berkshire's rises a little, and everyone else is about 20 percent again.)

So the story of AAAs in March 2009 is not really a market story about AAAs ... it's a story about GE. Still, to be fair to Gorton -- it could be that investors were irrationally chucking their solid GE bonds, desperating trying to raise cash, even though GE was a top-tier company. So what was the GE story right about March 5, 2009? A great AAA corporation with the wind at its back, sailing toward a sparkling future? Let's concede that the early part of that year was turbulent, and markets were down, so one might expect a ding or two in the giant's armor.

It turns out GE's armor was more than lightly dinged:
1. Its stock had plunged 61 percent from Jan. 1 to the end of the day March 5, almost three times the drop of the S&P 500.
2. Its credit default swaps were trading sky-high, meaning investors saw GE as a poor credit risk. The swaps were 1,037 basis points on March 5, compared with JPMorgan's 219 -- and JPMorgan was rated only AA! So let's mull that: JPMorgan at the time was considered almost five times safer than GE, even though GE was a higher-rated company.

And, if you're still not convinced, a March 5 Bloomberg story quotes Marilyn Cohen of Envision Capital Management on GE: "It's a leper right now."

A AAA rated leper? Sounds like an oxymoron. What were investors so bent out of shape about? Bloomberg's article says:
Investors are punishing the shares on a presumption, which the company disputes, that GE Capital will need more outside funding to cover potential writedowns and losses in real estate, consumer credit cards and leasing.
There was a fear that GE's finance arm would need to post $12 billion in collateral if its long-term ratings were cut to single A. Whence this downgrade fear? Well, Moody's said on Jan. 27, just a month earlier, that it was looking at lowering GE's rating. And in fact, if you look at the same AAA Merrill Lynch index on April 15, 2009 -- about a month later -- neither GE nor Berkshire are there; they've dropped off.

So did AAA corporates irrationally leap above AAs during this tumultuous period in March, when the greatest spread inversion of the financial crisis occurred? Absolutely not.

There were no "fire sales," as Gorton would have us believe. This wasn't a story about investors in hard times selling grandma's fine china at big discounts so they could afford a little soup for dinner. This is a story of ratings lagging behind reality, legitimate investor concerns, and GE's peculiar circumstances. So next time you hear the "fire sale" argument advanced, approach with full skepticism.

6 comments:

  1. And, just to put some icing on the cake, Berkshire Hathaway has sold an enormous number of puts (worth well into the 11 figures) on a few world indices, and so it wasn't looking particularly AAA either.

    I think the best thing to check his thesis is to look at the other four companies that are AAA (ADP XOM JNJ and MSFT) and see what happened to their bonds. I'd be very surprised if there was inversion.

    ReplyDelete
  2. Saw Felix Salmon's comments on Twitter, but disagree with him (which is a rarity).Your piece is interesting & informative on a number of fronts but does not systematically dismantle Gortons arguments.With respect, isn't there a risk when comparing 'held' 'mark-to-myth'AAs that hold relatively low notional value (knowing how fastidious Auditors have been on ensuring marks are accurate)with actually traded AAAs(oops, now another set of auditors are involved we'd better look more closely at the valuation models we've signed off on)? Variations in expectations, are likely to be uncovered in daylight.
    Comparing theory with practice will show distortions;yields being one.
    Don't disagree that assets were overvalued. But valuation is either intrinsic or 'relative' in my mind i.e. based on what others are doing & saying influencing owners representativeness bias.
    So maybe it was a mix of the two arguments. Smarter understanding of risk visibility & a corporates robustness to risk by investors is a starting point. Challenging a Boards well-meaning platitudes to their Management of Corporate Risk is well overdue. And Skepticism is great advice !

    ReplyDelete
  3. You have done a great job of demonstrating specifically that the AA/AAA bond spread inversion in 2009 was not due to liquidity demand.

    But your ambitions seem much greater, or at least if they are not, you have not expressed yourself well, because you seem to be claiming that there were no fire sales of any asset during the crisis (and perhaps even that there never are - its hard to read you.)

    In my opinion, that would be overreach, and if I am wrong you have a long, long way to go to prove it: there must have been dozens, if not hundreds of examples of "reverse flight to quality" in the equity market. I had interpreted these as people closing out their shorts and getting small; are you really capable of disproving them case by case? And what's your story about why the prospect of a collapse of the US banking system would cause USD to rise and gold to fall - rumors of cheap transmutation?

    ReplyDelete
  4. "And what's your story about why the prospect of a collapse of the US banking system would cause USD to rise and gold to fall - rumors of cheap transmutation?"

    A banking crisis is deflationary (there's less money/credit available than people expected), so why should it be good for gold?

    ReplyDelete
  5. "a" -- good point about Berkshire. Originally I meant to discuss what was happening with Berkshire in March of last year. But the post was already getting far too long.

    David ... interesting comment, but I think you may be confusing corporate bonds with the more complex (and opaque) securitizations out there. I don't think the AA corporates were "mark to myth" at all (look at some of the names: Wal-Mart, Pfizer, Shell, Met Life) and what's more the AA index appeared more liquid than the AAA, if you consider the number of discrete bonds underlying (the AA index had 347 at the time vs. one-fourth that amount, or 88, for the AAA). If you need further convincing: remember, the marked jump that created the "inversion" happened on the AAA side. Once you dig some, you find out that, among AAAs, what occurred in March 2009 was really a GE story. However, we both agree on healthy skepticism. :)

    ReplyDelete
  6. Anonymous 10:39 a.m.:

    I probably didn't express myself well, true. When I said "there were no fire sales," I meant in that data set Gorton presented. Of course I'm sort of shooting fish in a barrel there -- I think he chose a bad example if he wanted to prove the existence of "fire sales."

    The larger, and more intriguing, question, is "Were there fire sales elsewhere during the financial crisis, and if so, to what degree"? I do think that a "fire sale" factor helped depress prices somewhat for certain assets. But the more "money like" an asset, the less it seems susceptible to fire-sale pricing. What's the fire-sale price for a $5 bill?

    CMBS and RMBS are interesting cases. They are bonds and "money like" in a way but extremely complex. So when investors demand a premium for that opacity and for the involvement of the securities in a sick housing/commercial market are they being unreasonable?

    I'll pass on gold and the dollar -- not my areas of expertise. I have noticed that the dollar, as a reserve currency, doesn't always behave the way one might expect. And as for gold -- we've been off the gold standard for half a century and there are plenty of days I couldn't begin to tell you why gold moved one way and not the other. The gold market's kind of a mystery to me.

    ReplyDelete