Tuesday, September 29, 2009

AIGFP's Mistakes

I want to expand on something I said in the comments to my last post. In discussing AIGFP's excuse for not understanding the risk they held, I said that it was broadly a combination of three factors:

  1. AIGFP not understanding or particularly caring about the declining underwriting standards in the subprime market (a mortgage was a mortgage to them);
  2. AIGFP not fully understanding how the (somewhat new) collateral posting process on these trades worked; and
  3. Old-fashioned excessive optimism.
The first two factors require a bit more explanation. Why would a huge financial institution like AIG take on $100+ billion in risk without understanding, or particularly caring about, material terms of the deals? You have to understand the players in the ABS CDS market at the time, and how AIGFP viewed itself in the market. The natural protection sellers for CDS on ABS or CDOs were the monoline insurers (MBIA, Ambac, etc.), since "wrapping" deals is what they do. The market had been trying to develop standardized documentation for ABS CDS for a while, and the monolines had been notoriously difficult to work with. The market never did settle on a single template for ABS CDS, and that was primarly because of the monolines. Instead, the market settled on two forms: the so-called "dealer form" and the "end user form." Ironically, the monolines were adamantly opposed to this because — I kid you not — they said they didn't want to be exposed to market risk.

AIG billed itself as the protection seller that was easy to work with. AIG was willing to use the dealer form, which was more favorable to protection buyers. It was even willing to fully collateralize its ABS CDS trades with shockingly loose Credit Support Annexes (CSAs), on which more below. The monolines were sometimes heavily involved in the negotiations over the underlying deals that they would wrap, so they had at least some say on the quality of the loans that went into the deal (which isn't to say that they were prudent about loan quality, because they obviously weren't). Since AIG wanted to be seen as the hassle-free alternative, it was perfectly willing to simply write protection on deals that the Street brought them. This, from Michael Lewis's article, is actually pretty accurate:
When traders asked Frost [the head of AIGFP's CDS sales] why Wall Street was suddenly so eager to do business with A.I.G., says a trader, “he would explain that they liked us because we could act quickly.” (emphasis mine)
AIG's hands-off attitude meant that it had no line-of-sight to the mortgage market, or to the mortgages underlying their deals. So when the lending standards in the subprime market plummeted, AIG was the last to find out.

In the end, though, it was the terms of the Credit Support Annexes (CSAs) on the CDS trades that really killed them, because they left AIG massively exposed to market risk. CSAs govern the passing of collateral between counterparties to a swap transaction. Standard CSAs provide the parties with the right to demand collateral whenever the party's "Exposure" — defined as the cost of replacing the trade in the market — exceeds a certain threshold. The CSAs on AIG's trades required them to post collateral based on the market value of the underlying cash bonds (i.e., the subprime RMBS), and included moderate thresholds in the 5-10% range, if they included a threshold at all. This meant that as the ABS and CDO markets collapsed and liquidity dried up, AIG had to post more and more collateral — even if, as was sometimes the case, the underlying ABS had yet to suffer any principal or interest shortfalls.

Honestly, it's a mystery why AIG agreed to fully collateralize these trades with CSAs, because it was a phenomenally stupid idea, and it literally brought down the entire company. Again, it left AIG completely exposed to the market risk of the underlying ABS and CDOs, and as everyone knows now, the ABS and CDO markets completey collapsed. What's more, by agreeing to CSAs that provided for collateral posting based on the market price of the underlying securities, rather than the market price of the CDS, they probably deprived themselves of the ability to use the much-more-liquid ABX index as a reference price. Instead, they almost certainly had to use the utterly collapsing prices in the cash ABS and CDO markets. (No word yet on whether they understood the difference between exposure to market risk and the amount of market risk, but all signs point to no.)

AIG also apparently didn't fully appreciate how the process for determining the "market value" of the cash ABS/CDOs — which determines how much collateral they have to post — would work if liquidity completely dried up. When liquidity actually did dry up in 2007, AIG discovered that, per the CSAs, the dealers had the final say on what the market value was. So when, say, Deutsche Bank told AIG that it had to post an additional $25mm of collateral because, according to Deutsche Bank's correlation trading desk, the market value of the underlying CDO had falled from $300mm to $275mm, AIG had no recourse. In the end, it had to post the collateral. (I understand that AIG disputed this reading of the CSA language, which isn't uncommon when a firm is facing large collateral calls, but you can only play that game for so long.)

I know the mistakes that AIG made on the collateral posting probably seem like highly technical mistakes, but believe me, they're not — they're serious, fundamental mistakes that no moderately sophisticated investment shop should ever make.

Friday, September 25, 2009

Risk held at AIGFP was not a surprise

Gillian Tett is going way out of her way to avoid admitting that she (or any financial journalist, really) could have and should have known that AIG Financial Products was holding a substantial portion of the risk in the credit derivatives market. In today's FT, she argues that the opacity of the CDS market prevented anyone from knowing about the huge risk concentration at AIGFP in advance. I won't deny that the CDS market is opaque and that more transparency is needed, but the risk concentration at AIGFP was neither a secret nor a surprise—and it certainly shouldn't have been a surprise to a financial journalist who is supposedly an expert on the credit derivatives market. The "hoocoodanode?" card won't play on this one. Tett cites a Fitch survey of the CDS market from 2007 to make her point about AIGFP:

[W]hat is ... striking is that a well-respected Fitch survey before the collapse of the credit bubble suggested that AIG was just the 20th largest credit default swap player in the sector, based on gross notional outstanding volumes. No wonder those billions of lossmaking contracts subsequently came as such a shock.
Tett is being very disingenuous here. First, not until a few paragraphs later does Tett admit that the 2007 Fitch survey was ranking counterparties by gross notional rather than net notional amounts. The Fitch survey wasn't wrong—AIG was only the 20th largest counterparty in the CDS market by gross notional amount. What's wrong is to suggest that a ranking of counterparties by gross notional amount is the same as a ranking of "the largest players in the CDS market." More importantly, though, what Tett conveniently fails to mention is that this very same Fitch survey clearly explained that AIGFP was the 800-pound gorilla in the credit derivatives market, and that by the end of 2006, AIGFP had sold a net total of $384 billion in CDS protection. From the 2007 Fitch survey (emphasis mine):
Global Insurance The global insurance/reinsurance sector, representing principally AIG Financial Products, remained a large seller of protection, registering an aggregate gross sold position of USD503bn. On a net basis, the sector stood at USD395bn sold, up slightly from the USD383bn tallied at year-end 2005. Structured finance synthetic CDOs and corporate synthetic CDOs accounted for 93% of the sold volume. In terms of quality, of the total gross amount sold, 93% was in the super-‘AAA’ segment, versus 91% reported in last year’s survey.
In addition, 68% of the tenor sold ranged from one year to four years (compared with 44% at year-end 2005) though that was driven by AIG’s dominant “footprint” in this market. Excluding AIG Financial Products’ sizeable position, the global insurance industry had a net position of only USD11bn (USD21bn gross sold), down from the USD15bn (USD29bn gross sold) compiled at year-end 2005. Some of the decline can be attributed to a smaller sample than in the previous year.
By product on a sold basis, single-name CDS consisted of 63% of the volume, while corporate synthetic CDOs and structured finance synthetic CDOs made up just 16% of the total. Similarly, the credit profile, excluding AIG, of the industry changed, with ‘BBB’ representing 40%, ‘A’ at 35%, and ‘AA’ at 10%, while speculative grade represented 8%, and ‘AAA’ 6% of the notional amount sold.
The Fitch survey—which is publicly available (with a free Fitch account)—was published in July 2007, and was very well known in the markets. If you're a financial journalist covering the credit derivatives market, there's no excuse for not knowing about AIGFP until September 2008.

Thursday, September 24, 2009

Memories (Casey Mulligan edition)

Remember this NYT op-ed by the University of Chicago's Casey Mulligan last October? Well, as bad as it was then, it's even worse now. (Keep in mind that this is a full professor of economics in the #1 ranked economics department in the country):

The non-financial sectors of our economy will not suffer much from even a prolonged banking crisis, because the general economic importance of banks has been highly exaggerated. ... Although banks perform an essential economic function — bringing together investors and savers — they are not the only institutions that can do this. Pension funds, university endowments, venture capitalists and corporations all bring money to new investment projects without banks playing any essential role. [ed: Wow. Just wow. I didn't think it was possible to be a professional economist and also not understand a single thing about how a financial market works. Apparently I was wrong.] What’s more, it’s not as if banking services are about to vanish. When a bank or a group of banks go under, the economywide demand for their services creates a strong profit motive for new banks to enter the marketplace and for existing banks to expand their operations. (Bank of America and J. P. Morgan Chase are already doing this.) ... And if it takes a while for banks and lenders to get up and running again, what’s the big deal? Saving and investment are themselves not essential to the economy in the short term. Businesses could postpone their investments for a few quarters with a fairly small effect on Americans’ living standards. How harmful would it be to wait nine more months for a new car or an addition to your house?
Ah, memories.

Saturday, September 12, 2009

Commentator or Fiction Writer?

Simon Johnson has another barely-coherent article in The New Republic about financial crises past, present, and future. Johnson has never been interested in making serious, fact-based arguments, but this article has to be his worst. I think this part amused me the most:

During the heady days of summer 1927, the Fed had done something else that would contribute to the Great Depression: It lowered interest rates.
The "heady days of summer 1927"? The U.S. was in a recession in the summer of 1927. I guess if the facts don't fit your conspiracy theory, you just change the facts. He also manages to describe the LTCM crisis without ever even mentioning that a consortium of fourteen Wall Street banks bailed out LTCM to the tune of $3.625 billion. (Johnson conveniently omits any reference to this massive debt-for-equity swap, even though he's previously argued that, with debt-for-equity swaps, "the problem of moral hazard . . . would at once be forgotten.") Instead, he focuses on the moral hazard created by the Fed's decision to cut rates 50bps in the weeks following the LTCM crisis. I kid you not. These two examples are just the tip of the iceberg. TNR's editors should be thoroughly embarrassed.

As demonstrated by Donald Barlett and James Steele of Vanity Fair. They devote 6,500 words to the (shocking!) revelation that no one knows exactly what each bank did with its TARP money, because Treasury hasn't been "tracing" TARP funds. Apparently no one pulled Barlett and Steele aside and explained to them that money is fungible—once TARP money is credited to a bank's Federal Reserve account, it becomes indistinguishable from the rest of the money the bank has in that account. When you take $20 out of the ATM, do you know exactly which of your deposits the $20 came from? No, because money is fungible. Treasury could have asked the banks what they did that, without TARP, they would not have, but since there's no way to verify a bank's claim, all the banks would've claimed that they spent their TARP money on apple pie and American flags. So after 6,500 words, the basic conclusion is: Treasury injected a lot of money into a lot of banks, and a handful of those banks have business practices that, while not against the law, you may find unsavory.

At the end of my initial post on financial innovation, I listed a handful of financial innovations that I considered beneficial. (I didn't mean to list both zero-coupon bonds and Treasury STRIPS, since STRIPS are obviously zeros—the most important zeros, in fact.) I was mainly focusing on financial products that no one has yet mentioned as being beneficial in the financial innovation debate, which is why I omitted obvious examples like ETFs. Anyway, here, at last, is my argument for why medium-term notes and zero-coupon bonds are clearly beneficial financial innovations. (Next up: project finance and puttable bonds.) Medium-Term Notes A medium-term note (MTN) is a debt security offered to investors continuously over a period of time as part of an MTN program. Essentially, a company establishes an MTN program by filing a so-called "shelf registration" with the SEC describing, broadly, the types and total amount of securities they reasonably expect to sell under the program in the next two years. (Rule 415 covers shelf registrations.) After the SEC declares the shelf registration effective, the company can sell a wide variety of securities under their shelf-registered MTN program for the next two years on a moment's notice, and without having to file separate registration statements for each debt issue. Unlike conventional corporate bonds, which are underwritten, MTNs are generally distributed on an agency basis, with 3 or 4 investment banks acting as agents for each MTN program. An issuer will post offering rates with its agents for a range of different maturities: 9 months to 1 year, more than 1 year to 18 months, more than 18 months to 2 years, and so on up to any number of years. Issuers can change their offering rates in response to, say, a lack of investor interest, or a change in interest rates. An investor interested in purchasing notes at the offered rate will contact the issuer's agent, who then arranges the transaction. Investors are generally allowed to choose the exact maturity date within a given maturity range, which allows them to more precisely match the maturities of their assets and liabilities. (This is obviously an (extremely) abridged description of MTNs. For people who want a more complete description, I've extracted the chapter on MTNs in Marcia Stigum's Money Markets—mainly because I've just discovered how to extract sections from PDF files, and now I'm an extracting machine! For an example of an active MTN program, see here, or see Freddie Mac's MTN program here.) The main reason I immediately thought of MTNs as a beneficial financial innovation—though there are better reasons, described below—is a process in the MTN market called "reverse inquiry." Say a mutual fund needs $6.5 million of A3-rated paper with a maturity of 4 years and 6 months, but no such paper is currently available in the MTN market. The mutual fund could contact one of the major MTN agents, who would then go to the A3-rated issuers it represents with the mutual fund's proposed terms. If an issuer is satisfied with the proposed terms, a deal can be struck and the issuer can sell the custom-tailored security out of its MTN program that day. (The issuer is essentially taking the securities it registered with the shelf registration statement "off the shelf.") Reverse inquiries are actually quite common, and they account for a large share of total MTN transactions. The most exotic MTNs usually result from reverse inquiries—a fund manager may be exposed to some weird risk which necessitates paper that's linked to a random equity index or exchange rate or something. But this isn't a case of investment banks dreaming up exotic (and risky) securities and selling them to unsuspecting investors. Exotic MTNs that result from reverse inquiries are, by definition, driven by the needs of investors. More broadly, MTNs bridged the gap between commercial paper and traditional corporate bonds. For regulatory reasons, commercial paper can't have a maturity at issue of more than 9 months. That meant that prior to MTNs, any company that wanted to issue debt with a maturity of over 9 months had to register each debt issue separately, which was—and still is—both expensive and time-consuming. One of the reasons that traditional corporate bonds trade in such size ($100 million is an unofficial minimum) is that the legal, accounting, and printing costs of registration are so high. This is on top of the generous underwriting fee that seasoned issuers (the biggest beneficiaries of MTNs) paid to their "usual" investment bank. These costs made it prohibitively expensive to sell smaller and shorter-dated securities (i.e., maturities of 9 months to 2 years). This minimum-offering-size issue, as well as other conventions in the corporate bond market, such as the clustering of maturities around 2, 3, 5, 7, and 10 years, really limited issuers' options, and ultimately raised their borrowing costs. MTNs changed all that, making it cost-effective for companies to sell smaller debt issues at non-standard maturities, and often with terms that better suit both the issuer's and the investor's specific needs. MTNs also increased competition among underwriters by allowing issuers to engage multiple investment banks as agents for their MTN programs. The increased competition significantly reduced investment banks' fees, lowering issuers' borrowing costs even further. (The notion that increased competition can lead to lower underwriting fees seems almost quaint nowadays, but it was not always so.) That, to me, is a beneficial financial innovation. Zeros (and STRIPS) Zero-coupon bonds (known as "zeros") are bonds that pay no interest, but instead promise a single payment at maturity. Since investors receive no coupon payments over the life of the bond, zeros are sold at a discount to their face value. Any conventional couponed bond can be considered a series of zero-coupon bonds—each coupon is simply a promise to make a single payment on a specified date. Consequently, a conventional 5-year note can be "stripped" into its 10 coupon payments and a principal repayment, creating 11 separate zero-coupon bonds. Merrill created zeros in 1982, when it stripped a series of Treasury long bonds. Other investment banks quickly started selling privately stripped Treasury zeros as well. The Treasury Department finally started allowing direct stripping of Treasuries in 1985, when it instituted its STRIPS (Separate Trading of Registered Interest and Principal Securities) program. Treasury zeros, which are commonly referred to as STRIPS, are created via the Fed's book-entry system—when a Treasury is stripped, each coupon payment and the principal payment becomes a separate zero-coupon bond with its own CUSIP. Stripped Treasuries can also be reconstituted via the book-entry system. So how have zeros been beneficial? They eliminate reinvestment risk, which has made them perfect for pension funds and insurance companies, which need long duration and certainty in their returns. STRIPS, in particular, offer maximum certainty: the U.S. Treasury is the safest borrower in the world, and, as zero-coupon bonds, STRIPS also eliminate reinvestment risk. The certainty and long duration of STRIPS have been absolutely crucial to the ability of pension funds and insurance companies to reliably and predictably match the cash flows of their assets to the benefits payments they make. Zeros are more sensitive to changes in interest rates than conventional couponed bonds, which can lead to short-term volatility in prices, but that generally hasn't been an issue for long-term, buy-and-hold investors like pension funds. Also, zeros prices are something of a theoretical building-block in modern financial models. The large and liquid market in STRIPS, for instance, has provided accurate prices across a range of maturities, which is actually pretty important, especially for banks and other investment funds who rely on things like the zero-coupon Treasury yield curve in their pricing models. (In fact, the question of why it took the financial industry so long to create a product as fundamental and seemingly obvious as zero-coupon bonds is still something of a mystery in finance circles.) This isn't the most important benefit of zeros, to be sure. But it's not insignificant either, and I think we can all agree at this point that better financial models are a good thing.

My post on the quality of discussions of financial innovation generated a much larger response than I was anticipating (some pro, some less-than-pro). Now people apparently expect me to follow through on my promise to write a follow-up post on financial innovation. I guess I've boxed myself in, so here goes. First, let me just say that I don't consider myself a "defender of financial innovation." In fact, at the retail level, I'm actually firmly anti-financial innovation, and I'm all for the Consumer Financial Protection Agency. My contention is simply that discussions of institutional-level financial innovation over the past year have taken place at an extremely superficial level, and have been premised on a number of basic misconceptions. The point of my last past was to point out two of those misconceptions, not to make a broad defense of financial innovation. (The point was definitely not to "get CDOs off the hook," as I specifically criticized CDO attachment points.) In his 2002 post-mortem on the internet bubble, Michael Lewis has a great passage on the psychology of booms vs. busts, which I think is particularly appropriate right now:

The markets, having tasted skepticism, are beginning to overdose. The bust likes to think of itself as a radical departure from the boom, but it has in common with it one big thing: a mob mentality. When the markets were rising and everyone was getting rich, it was rare to hear a word against the system—or the people making lots of money from it. Now that the markets are falling and everyone is feeling poor, or, at any rate, less rich, it is rare to hear a word on behalf of either the system or the rich. The same herd instinct that fueled the boom fuels the bust. And the bust has created market distortions as bizarre—and maybe more harmful—as anything associated with the boom.
I think the general anti-finance backlash, fueled by this herd instinct, has driven much of the debate over financial innovation. It's definitely made commentators more willing to make broad, sweeping (and wrong) claims about financial innovation. But anyway, on to the substance. Poorly defined terms are one reason that discussions of financial innovation have suffered so much. What constitutes a "financial innovation"? What are we considering acceptable goals/uses for new financial products? What about a product that's created to take advantage of a change in the tax code or regulatory capital rules? (Are financial institutions supposed to not respond to incentives in the tax code or the regulatory capital rules?) Are financial innovations that are theoretically sound but have been misused in practice still considered beneficial, or are we only judging financial innovations based on the way they've been used in practice? I'm willing to define "financial innovation" as a new financial product that enjoys widespread use, and that was not created solely to realize tax or accounting benefits. That ignores the small-scale, one-off style of financial innovation, where an individual transaction is structured in a unique way in order to suit the particular particular parties in the transaction. These transactions often create new kinds of financial products, and the only reason they're considered "customizations" rather than "financial innovations" is that they're never adopted by the wider market. It's also important to determine when it's fair to say that a financial innovation "caused" a certain problem. Did securitization "cause" the housing bubble? Clearly not, since securitization worked fine for 20+ years before the housing bubble. A more interesting question is: If a regulation allows a financial innovation to be used in a way that turns out to be harmful, is that the fault of the financial innovation or the regulation? For example, in a Planet Money podcast on financial innovation, Mike Konczal of Rortybomb, after saying that credit default swaps were also a great innovation, says:
CDSs were also a terrible innovation, because they allowed a lot of investment firms to skirt regulatory regimes. Basel II had certain requirements about how much debt you could take on, and the CDS was almost engineered to be able to sneak around that regulation.
This is actually not accurate, and in a way that's important to how you view CDS as a financial innovation. CDS do not allow banks to "sneak around" or "skirt" Basel II. Basel II explicitly recognizes CDS as an eligible "credit risk mitigant" that banks can use to reduce the amount of capital they're required to hold—and thus increase their leverage (see paragraphs 191–200 of Basel II). The majority (~$350bn) of AIG's CDS book consisted of these so-called "regulatory capital relief" trades, most of which involved European banks buying CDS protection from AIG. But is this the fault of CDS, or is it the fault of Basel II? I'd argue that it was mostly the fault of Basel II, because CDS were developed primarily for nonregulatory reasons—it took the Basel Accords (both I and II) to make CDS the leverage-encouraging instruments they became. Basel II would have encouraged excessive leverage whether CDS had been invented or not, because banks could also have used total return swaps—which predate CDS—to reduce their regulatory capital under Basel II. So while it's fair to say that using CDS for regulatory capital relief encouraged excessive leverage, it's not fair to lay the blame for this at the CDS market's door. I'm curious to see if Mike Konczal has another argument for why CDS were a bad innovation, or if this was the only one. Finally, I want to push back against an argument that I've seen several times, and that Felix Salmon happened to make in the Planet Money podcast. After listing several of financial innovation's alleged sins over the past 20 years, Salmon attempts to deny any benefits by arguing:
[E]fficiency didn't really improve, as is evidenced by the enormous profits that the financial sector made. If they were getting more efficient, you would expect them to be making less money; in fact, they made more.
This is wrong on multiple levels. On a theoretical level, Salmon is confusing the industrial efficiency of the financial services sector with the informational efficiency of asset markets. Industrial efficiency is concerned with whether the services provided by financial institutions are priced at their marginal cost. Informational efficiency is concerned with whether asset prices fully reflect available information (which is what people commonly mean when they discuss financial market efficiency). While industrial and informational efficiency are obviously related, they are by no means the same thing, and the "enormous profits" in the financial sector say very little about the informational efficiency of asset markets. Salmon's argument also presumes, wrongly, that the financial sector has earned consistently high profits for the same activities. The hugely profitable activities in finance are different practically every year. Traders and salesmen in that year's new, "hot" market pull down seven-figure bonuses, but then every other bank on the Street piles into that market, pushing down margins to more normal levels, and two years later traders in that market are applying for jobs on the muni bond desk. In the early '80s it was mortgage trading desks, then it was LBOs and junk bonds, then it was interest rate swaps, then it was anything calling itself a "hedge fund," then it was internet IPOs, etc., etc. The fact that new, different groups within the financial sector made outsized profits each year doesn't show, or really even suggest, that informational efficiency in the asset markets has failed to improve. Okay, in the next post (which I've already written) I'll discuss some of the specific financial innovations I highlighted last time. This post is getting way too long though, so I'm splitting it into two posts.