Sunday, 21 August 2011

Information-Insensitive Debt and the Strange Case of Haircuts

Now for part 6 of my rather exhaustive (and exhausting) look at Gary Gorton's theory of information-insensitive debt.



This post is rather granular, to show inside-out how a very dubious assertion at the theory's center leads Gorton to what (I think) is a wrong-headed interpretation of what occurred during the height of the financial crisis.



We start with haircuts -- in the repo market, not at the local barber shop. Anyone who needs a refresher on how repurchase agreements work, go here.



HAIRCUTS FOR DUMMIES



Per Gorton, the haircut is the "percentage by which an asset's market value is reduced for the purpose of calculating the amount of overcollateralization of the repo agreement."



That's very gnarly sounding. Here's an unpack that gets at the gist of the matter.



When you "deposit" say $100 million in a shadow bank through a repurchase agreement, the bank essentially posts collateral (to guarantee your funds in case it defaults). The haircut can be thought of as a way to ensure you get ALL your money back. So you may receive $105 million of securities for that $100 million -- a haircut of 4.8 percent (5/105). If the shadow bank collapses overnight, you're holding $105 million of securities to make you whole on $100 million -- not a bad proposition, it seems.



Haircuts vary with the nature of the collateral debt. During the financial crisis, for AAA corporate debt, they were minor (about 5 percent, according to Gorton). A graph of haircuts on asset-backed securities, on the other hand, resembles a Stairmaster in profile. They climbed from zero percent to about 40 percent when the financial system was in extremis. That deep 40 percent haircut was the shadow banking equivalent of a large amount of money being sucked out of the system -- a bank run, in other words.



WHY A HAIRCUT? HINT: THE ANSWER MAY SURPRISE YOU



Now comes a question that turns out to be more interesting than it first appears. Namely, what are these haircuts based upon? This is where things get curious. You might assume, if you're a common-sensical markets person, that “depositors” demand haircuts because if their counterparty in the repo agreement fails (a la Lehman), they need to be compensated for the fact that the securities may not really be worth what they were told. Or, an alternative explanation could be that they’re afraid the value of the securities might drop while they’re holding them.



Both interpretations, however, would be incorrect, according to Gorton.



Gorton informs us that:

Haircuts are a function of the default probabilities of the two parties to the transaction, as well as of the information-sensitivity of the collateral.
Now, before we analyze that, here's another excerpt you need to read, to get the full picture of where Gorton is coming from (this is also from his paper titled Haircuts, the bold is mine):

Keep in mind that the collateral offered in repo is valued at market prices. If the bonds become riskier, and their prices go down, then they would be valued at these lower prices. Furthermore, if there is more uncertainty about their price in the future, that risk can be addressed with a higher repo rate. Repo rates can and did go up (see Gorton and Metrick (2009)). Why should repo collateral be haircut? And why should these haircuts go up? Our answer, following Dang, Gorton, and Holmström (2010a,b), is that a haircut amounts to a tranching of the collateral to recreate an information-insensitive security so that it is liquid. The risk that is relevant here is different than the risks we usually think about, which are related to the payoff on the security. A haircut addresses the risk that if the holder of the bond in repo, the depositor, has to sell a bond in the market to get the cash bank, he may face a better informed trader resulting in a loss (relative to the true value of the security). This risk is endogenous to the trading process. It is not the risk of loss due to default. Consequently, the price cannot adjust to address this risk.
PAINTING ONESELF IN A CORNER WITH EFFICIENT MARKETS THINKING



The first thing you should have noticed: Gorton very much appears to be some form of EMHer (Efficient Markets Hypothesis, or the belief that market prices are efficient and reflect all existing public information). (As a reformed EMHer, I can spot a member of the species. This is precisely how they talk: "If the bonds become riskier, and their prices go down, then they would be valued at these lower prices." They don't simply make a point; their stating of a proposition has a whiff of the evangelical.)



Being an EMHer, though, paints him in a corner, starting with his explanation of the two factors contributing to haircuts. Because for an efficient markets guy, the "default probabilities of the two parties to the transaction" -- reason #1 for haircuts -- shouldn't matter at all.



After all, if the collateral for your deposit is a security at market price, that's what someone would buy it for at that moment. And you're only holding the security overnight -- or for a few days -- so where's the risk? Of course, the price may change. But Gorton covers that in the longer excerpt above, saying you'll demand a higher repo rate to compensate for that risk. The more detailed excerpt, in fact, appears to conveniently forget about counterparty risk.



So, back to square one: why do you need a haircut?



BETTER INFORMED TRADERS ON THE LOOSE!



This is where the theory starts crumbling around the edges. Remember, the centerpiece is information sensitivity, so that's the Procrustean bed Gorton has to fit his analysis into. Here's how he explains why a shadow banking “depositor” requires a haircut: if forced to sell the debt (my bold again), "he may face a better informed trader resulting in a loss (relative to the true value of the security)."



Ah, so the real problem is a "better informed trader." But what does that phrase mean? And what does it imply? I'm not sure whether Gorton tries to use it in a special way, but let's assume he doesn't. In that case, a "better informed trader" would presumably be a trader who knows the worth of the debt better than you do. And, it appears, you're worried that his information will be negative and push his offer price lower.



So a better informed trader knows something about the value of the security that you don't, so you're afraid that the debt may be mispriced, and that's why you demand a haircut?



No, Gorton would probably demur, it's not quite that. He tells us "This risk is endogenous to the trading process. It is not the risk of loss due to default." See, the market price plus the repo rate has already captured the risk of loss due to default, according to Gorton. But then, what is the nature of the knowledge possessed by a better-informed trader? When the donut cart makes its rounds at corporate headquarters? Because, seriously, when I hold a debt security, I'm concerned mainly with one thing: getting paid what I'm due, when I'm due it (and the probability of that occurring).



(Another thing: what is the "true value" of the security? What relationship does it have to the "market price"? Which should I care about? If the true value is $100 million but the market price is $200 million, why should I mind paying $200 million as long as other traders in the market are willing to pay that, especially if I possess the security for only a day or two?)



The other problem with these "better-informed traders" is that it stretches credulity that they suddenly appear on the horizon, a sagacious glint in their eyes, waiting to take advantage of you. Presumably they were also there the day before. So why weren't they pushing down the "market price" before? And if these better-informed traders are feared, why not find some of the apparently dumb traders of the day before who helped set the "market price" -- and simply sell to them instead, if they're so enamored of the security?



AND SO A THEORY VEERS INTO ABSURDITY



Gorton wants to convince us that the securities were fairly priced (they capture all the risk of loss due to default, remember) and that depositors extracted giant haircuts of 40 percent for fear that, if they got stuck with the debt, the only traders they encounter may possess an "information advantage." He never clarifies why, if this information advantage necessitates such a large haircut, the better informed traders aren't already profiting from their (considerable) advantage by trading in the market.



So Gorton basically says that market prices on asset-backed securities (a key kind of collateral in the shadow banking system) during the financial crisis were accurate. He makes this claim even though there were some 100,000 of them -- sui generis problems abound* -- and trading in particular ones probably got pretty thin and the value of asset-backed securities is often derived from a model (hence "mark to model") and investors were just starting to realize these things had been misrated and were probably lousier than they thought and ... you get the idea.



*(Brief aside: The uniqueness of these assets, and the difficulty accounting for them, was why Paulson scuttled his original plan for TARP, as Hernando de Soto recounted in Businessweek: “When then-Treasury Secretary Henry Paulson initiated his Troubled Asset Relief Program (TARP) in September 2008, I assumed the objective was to restore trust in the market by identifying and weeding out the "troubled assets" held by the world's financial institutions. Three weeks later, when I asked American friends why Paulson had switched strategies and was injecting hundreds of billions of dollars into struggling financial institutions, I was told that there were so many idiosyncratic types of paper scattered around the world that no one had any clear idea of how many there were, where they were, how to value them, or who was holding the risk.”)



ANOTHER WAY OF LOOKING AT THE MELTDOWN IN THE SHADOW BANKING SYSTEM



Is there an alternative explanation of what happened?



Yes, and it might go like this: Before the financial crisis, the haircut was zero on asset-backed securities because it was practically unthinkable that the large investment bank opposite you on the repo transaction would fail. And so, because the counterparty is deemed safe, the asset-backed securities being offered for repo aren't examined too carefully. The "magic pig" phenomenon starts to set in. "Sure, they're worth what we claim," Mr. Investment Banker says. "Would you like to see our math-heavy, extremely complex model or just take our word that you'll get repaid?" And you say: "I'll take your word, no problem."



But then the investment banks start looking shakier, and the asset-backed securities begin looking dodgier as well. You realize too that the banks are frightfully interconnected, boosting risk further. So you begin looking askance at asset-backed securities that you suspect aren't at "market price" at all. Further, you expect you'll have trouble reselling them. This would naturally lead you to demand deep haircuts.



Now you may resist going as deep as 40 percent -- that's pretty severe -- until you get really, really scared. What would scare you the most? If you think that the collateral may be mispriced, the scariest thing would be seeing one of those investment banks go under. Say Lehman Brothers. Until then, if you think the chance of the investment bank failing is remote, you may not extract much of a haircut for the mispriced securities. Who cares what they’re really worth? But once it becomes clear you may get stuck with this collateral -- the game changes totally.



You need a deep discount, and 40 percent would be reasonable. Gorton would have you think that such a discount implies crazy sale prices. This alternative explanation doesn't need to invoke a fire sale to make sense. It would, however, suggest there was a bit of the "magic pig" in those asset-backed securities.



Next: Many words later, a short conclusion: why should anyone care so much about this arcane subject?