Chickens, eggs and CDOs

After a turkey, successful construction of a 1100 piece K’nex rocket rollercoaster, many turkey sandwiches, several 10 person games of farkle, too many goodbyes and two “could have been much worse” flights, I can get back to the question of whether the causality in the CDO market ran from the “dumb” money to the low spreads or not.

What is missing from the argument that “dumb” money caused low spreads is an analysis of how synthetic assets (e.g. CDS on MBS and CDOs) affected the market.  By definition a CDS has a short and a long side, so from a purely theoretic perspective it should have no effect on the price of the underlying — but this theoretic argument is based on a world without asymmetric information, so it is likely to be misleading (since one thing that’s pretty clear at this point is that the buyers and sellers of CDOs (and CDS via CDOs) definitely did not have similar information — the buyers were “dumb” because they relied on credit ratings).

Most participants in the market argue that the CDO market “needed” synthetic assets in order to meet the demand of the “dumb” money.  But what this implies is that, in the absence of synthetics, spreads would have fallen even further and that, in the absence of synthetics, the demand would have extinguished itself as spreads fell to the level of treasuries and there was literally no point in buying CDOs because the yield advantage was completely gone.  In other words, in the absence of synthetics, market forces would have worked to eliminate the “dumb” money demand for CDOs by eliminating the spread differential between Treasuries and CDOs entirely.

In other words, the role of synthetics was to artificially sustain the yield spread over Treasuries (small as it was) offered by CDOs.  By keeping the yield spread up and delaying the function of market forces, synthetics kept the “dumb” money in the market and helped create a situation where the market collapsed abruptly (rather than slowly having the air pushed out of it by yield spreads that fell steadily to zero, discouraging demand in a natural manner).

In short, I think the “dumb” demand for CDOs was sustained for months and years, because synthetic assets created an artificial supply of CDO assets that kept yields at an artificially high level — in order to attract and sustain that demand.

Subprime vs subprime

The financial industry’s way with words — that is, the industry’s way of making sure that words never mean what you think they mean — has tripped up the academics once again.  The recent debate between Raghuram Rajan and Paul Krugman over the role played by Fannie Mae and Freddie Mac seems to hinge on the fact that they are talking at cross purposes.

The commonly understood meaning of “subprime” when referring to securitized mortgages is to be found in wikipedia:

Varieties of underlying mortgages in the pool:

  • Prime: conforming mortgages: prime borrowers, full documentation (such as verification of income and assets), strong credit scores, etc.
  • Alt-A: an ill-defined category, generally prime borrowers but non-conforming in some way, often lower documentation (or in some other way: vacation home, etc.) (Article on Alt-A)
  • Subprime: weaker credit scores, no verification of income or assets, etc.

There are also jumbo mortgages, when the size is bigger than the “conforming loan amount” as set by Fannie Mae.

Since a “conforming mortgage” is one that meets the criteria set by Fannie Mae and Freddie Mac, the traditional meaning of subprime is a loan that does not meet Fannie Mae and Freddie Mac’s criteria for a reason other than loan amount.  (Alt-A came much later than sub-prime.)  For this reason it is common to see the words “prime” and “conforming” used interchangeably in mortgage articles.

Thus using the traditional approach, any MBS issued by Fannie or Freddie is, by definition, not a sub-prime MBS.  When Fannie and Freddie invested in sub-prime they did so by buying up privately issued sub-prime MBS.  It seems obvious that in this market Fannie and Freddie were two of many, many guilty parties.

What makes the whole discussion complicated is that (I think) the GSE’s loan criteria became less discriminating over time.  Thus, if one wants to fix the definition of sub-prime based on the GSE’s standards in, for example, 1998, one can claim that the GSEs themselves were issuing “non-conforming” MBS.  On the other hand, it is unarguable that even when the GSEs were issuing “non-conforming” MBS, the loan criteria were always noticeably stricter than those in the privately issued subprime MBS market.  The bottom line is that the lending behavior of the GSEs is not well suited to a casual assertion that they started issuing subprime MBS, but instead deserves a carefully researched dissertation that breaks loans into far more granular categories than prime, subprime and Alt-A.

Tanta is of course the go-to source for anyone who wants to understand the nitty-gritty details of the mortgage market.  For an extremely thorough discussion of what precisely it means for a loan to be sub-prime or Alt-A see here, and here for a thorough discussion of what precisely the GSEs were doing when they securitized loans — see in particular the paragraph that starts:  “Propaganda from certain other market participants aside, you cannot just put any old loan in a GSE MBS.”

On selling “crap”

After reading the Lex column’s defense of the not-so-uncommon practice of peddling “crap”, I want to make two points:

(i)  Because the asymmetric information is built into many business relationships, it’s impossible to outlaw the sale of “crap” and I don’t think anybody will argue that first goal of financial regulation should be to put an end to transactions involving “crap”.  In the initial “caveat emptor” decision the judges explicitly recognize the limits of the law in practical matters of contract.  According to Jon Faust:

What is striking is that caveat emptor arises as a legal principle mainly because of the tangle the courts would get into if they tried to enforce a more ambitious standard of right and wrong.

(ii) There’s a big difference between arguing that peddling “crap” is a deplorable practice that it is impossible to end, and arguing that peddling “crap” has economic benefits.  (I criticize the “two people made a deal, therefore it must be good” approach to markets here.)

Issuing a security that is designed as a short vehicle and then marketing it to long investors without full disclosure about the design of the security — that is, deliberately creating asymmetric information  — quite simply cannot be viewed as a positive contribution to “the liquidity and vitality of our financial system“.  In fact as Steve Waldman argues the Abacus CDO — by giving Paulson the opportunity to trade off the public ABX market at below market prices — undermined public information about the state of the subprime mortgage market.

There’s not much that can be done about eliminating the practice of selling “crap”.  We can, however, force over the counter markets and any other dark markets to shrink dramatically and drive as much trade as possible to an environment where it will be publicly observed.

The problem with structured finance

It’s widely recognized that structured finance was used to arbitrage regulations.  Less well recognized is that fact that structured financial products were also used to arbitrage the ignorance or the ego of investors.

CDOs — at least in the quantity they have been issued in recent years — are inherently suspect as innovations that add economic value.  The reason for these doubts are simple:  If the assets underlying the CDO are priced to reflect fundamental values and all of the CDO tranches are also priced to reflect fundamental values, then what the CDO as a product brings to the market is the opportunity for investors to choose their preferred level of credit risk exposure to the assets in the CDO.  To believe that CDOs are a value adding financial innovation one must also believe that the gains from this division of credit risk into tranches are more than sufficient to cover the multi-million dollar cost of creating the CDO.*

The question is whether investors actually demand that wide variety of exposure.  The CDO needs different investors who want (i) a very safe, diversified portfolio of assets that pay a yield just a little higher than Treasuries and (ii) a relatively high fixed income return on a product that concentrates risk — and is likely to behave more like equity than a bond, and a range of possibilities in between.

Thus, the very concept of a CDO (as a product that distributes risk in an economically efficient manner) requires that there exists a full spectrum of investors who prefer different levels of risk including equity (the first loss position), equity-like risk with fixed income returns (the lower mezzanine tranches, that are unlikely to get any recovery in case the deal goes bad), and low risk, low returns on a diversified portfolio (the senior tranche — or super senior tranche in a hybrid or synthetic CDO).  In practice it seems to be relatively rare that all the different groups of investors exist for given CDO, because the banks frequently end up holding at least one of the tranches.

If it is not the case that the full spectrum of investors exists and that the gains from serving the needs of these investors more than cover the fees of creating the CDO, then the “value” of the CDO is likely to come from selling to one or more groups of investors a product whose risks they do not understand.  While a CDO that arbitrages such misunderstandings is likely to be very profitable for the investment bank that issues it, these profits clearly do not represent economically efficient allocation of risk.  In fact, after the recent crisis there is support for the view that trade in mispriced CDOs actually reduces social welfare.

Some more specific examples have popped up on the web recently:

(i)  Felix Salmon has a post up on:  super senior CDOs.  As I commented there

I think the reason the super senior “had” to exist in the 2006-2007 environment is because that’s where the risk was most underpriced. Many of the CDO “investors” were interested in high yield AAA assets, that is the leveraged senior, but not super senior segment of the CDO. But if market makers had tried to sell protection on this segment of the CDO only, that protection would have tended to be expensive for the same reason that it paid investors well relative to the super senior.

Selling packages that included large super senior tranches allowed the structured financiers to earn their salaries by keeping costs down for protection buyers, while also meeting the needs of CDO “investors” looking for high yield assets. Unfortunately they ended up warehousing large quantities of the residual super senior risk in the banks.

(ii) In a series of posts on Goldman’s Abacus CDO, Steve Waldman discusses another case of welfare reducing structured finance.  He argues that the Abacus CDO that is the subject of the SEC lawsuit allowed Paulson (the hedge fund manager) to take out a short position on subprime that was significantly underpriced relative to a comparable position on the public ABX markets.  By hiding the fact that the CDO was selected as a short vehicle for Paulson, Goldman induced ACA Capital, IKB and ABN Amro to take on the offsetting long position for compensation well below that available on public markets — this leads one to conclude that these counterparties did not understand that they were just serving as a cheap means for Paulson to take on a position comparable to shorting the ABX.

Effectively Waldman argues that Goldman used this Abacus CDO as a vehicle to create information asymmetry in the market.

(iii) Critics of structured finance have long argued that many of these “innovations” are profitable precisely, because they are carefully designed to exploit misinformation and hubris in financial markets.  Satyajit Das takes this view in his book, Traders, Guns and Money and gives a multitude of examples.  Two classic cases:  in 1993 Proctor and Gamble thought it was lowering its cost of funds when it entered into a transaction that involved the sale of interest rate puts;  in the early 1990s Orange County tried to raise its investment returns by entering into highly leveraged interest rate swap transactions.  Both of these cases ended up in court when interest rates rose and vast sums were lost.

It’s hard to believe that this needs to be stated, but here goes:  When a financial product is used to arbitrage misinformation or hubris, it does not contribute positively to economic welfare.  When such transactions result in bankruptcies there is little question that they are harmful to economic welfare.

These transactions do not add valuable pricing information to the economy for the simple reason that the basis of the transaction is a failure to understand the product.  If prices created by such transactions are viewed by other participants in the economy, they may lead to further misallocation of resources.

When structured finance innovations are designed to arbitrage misinformation, the markets in these products are inefficient and are likely to generate significant economic costs via unnecessary bankruptcies.  For this reason it is important to shine a light on these markets, not only through clearing and exchange trading of standardized products, but also by requiring delayed public reporting of transactions and transaction prices for those products permitted to trade over the counter.

*Note that tranched MBS are not necessarily subject to the same criticism because they distribute prepayment risk across investors — that is, the tranches are differentiated by their expected maturity — and thus they cater to the fact that different investors have different investment horizons.  In my view it’s much more intuitive that investors can be distinguished by their maturity preferences, then by their desire to take on different levels of credit risk.

Goldman’s role in the ABS CDO market

If Lehman’s ABS CDO data (as presented by Barnett Hart) is accurate, then Goldman’s role in the ABS CDO market is both interesting and unique.

First recall that ABS CDOs (along with CDO squareds) were the worst performing CDOs by far and that this potential for truly disastrous performance is obvious to anyone who understands their structure.  (ABS CDOs will one day probably be used to define “cliff risk”.) (Note Nomura document h/t Alea.)

Second, according to the Lehman data the ABS CDO league table for 1999 – 2007 (from Appendix A2 Panel D I calculate Total Balance/Sum of Total Balance) reads as follows:
Merrill Lynch     16.5%
Goldman Sachs    13%
Citigroup       12.3%
Wachovia     6%
Credit Suisse    5.7%
UBS     5.4%
Bear Stearns    4.3%
Deutsche Bank  3.8%
RBS    3.7%
Lehman   3.5%

Observe that the top three ABS CDO originators were responsible for 40% of the market and that Goldman was the number two originator.  Table 4 of the Barnett Hart paper presents S&P ABS CDO information and indicates that a disproportionate share of Goldman’s ABS CDO origination took place in 2005 and 2006.

Third, it is clear that Merrill and Citi believed that the ABS default cliff was far enough away that senior losses were unlikely to occur.  The evidence of this is the fact that Merrill and Citi carried large quantities of super senior ABS CDO risk on their balance sheets — and both had to be rescued in no small part because of their ABS CDO losses.

Goldman is a very different story.  When Goldman originated ABS CDOs it was apparently careful to lay off the senior risk onto other parties (the growing difficulty of this undertaking probably explains the relative decline of Goldman’s ABS CDO origination in 2007).

How big is the difference between CDOs and CDS?

Steve Waldman has a great post  up deconstructing the Abacus CDO that was the source of the SECs charges against Goldman.  What has become clear is that Abacus was not a synthetic CDO as those of us outside the industry understood them.  It was a bespoke CDO — a CDO tranche designed to meet the needs of a particular client with the investment bank taking on the responsibility of hedging or laying off the risk relating to the rest of the CDO.

Because most of the tranches of a bespoke CDO are not sold, but are held on the books of the originating bank, bespoke CDOs are very different from synthetic CDOs — in particular their price structure is not market-tested.  (See Steve Waldman for further details.)  This raises three questions,

(i)  To what degree have the investment banks been issuing bespoke CDOs, but presenting them to the public (and possibly to counterparties) as synthetic CDOs?

(ii)  Were the losses that were attributed to super senior CDOs at Merrill Lynch and Citigroup really just unhedged (or perhaps poorly hedged) CDS exposure — just like AIG?

(iii)  After 2008 is there any reason to believe that the investment banks as a group will ever be able to manage their CDS exposure wisely?

Abacus and the credit crisis

Just a quick note on one of the reasons the Goldman-Abacus case is so interesting.  The Abacus program definitely plays at least a small role in the credit crisis.

(i) Goldman itself notes that IKB was a notable Abacus counterparty  (See Part I here).

(ii) One of the first SIVs to fail was IKB’s Rhineland funding.  According to Acharya Schnabl and Suarez, Rhineland funding specialized in financing CDOs.

(iii)  In August 2007, when Rhineland couldn’t role over it’s commercial paper, IKB was the guarantor of the commercial paper.  This caused IKB to be one of the first banks to fail due to the subprime crisis.  In short, IKB was one of the banks that set off the asset backed commercial paper collapse of 2007 — a market that ended up requiring extraordinary accommodation from the Fed.

So there’s no question that IKB was right there at the heart of the credit crisis when it started.

Cognitive Dissonance in Structured Finance

The discussion emanating from the Goldman fraud allegation shines a light on the contradictory arguments that are used to defend the market in synthetic “assets”.

When one asks whether synthetic CDO tranches were a good idea, the standard response is:  Synthetic CDOs were necessary to meet the demands of investors:  The demand for AAA assets was “unlimited” and there’s no way the supply of cash assets could have filled that need, so synthetics were created to meet investors’ needs.

On the other hand when one observes that using a synthetic CDO to market a short vehicle as if it were appropriate for a long investor is dishonest, the response is:  But everybody knows that synthetic assets are backed by shorts, so it’s a case of buyer beware.

Clearly both stories of how synthetic CDOs work cannot be true at the same time.  Either synthetic CDOs are a benign development that allow financiers to better meet the needs of investors, or they are a particularly dangerous product where the investor always needs to be scrupulously second guessing the intentions of everyone else involved in the transaction.  It is precisely because outsiders are concerned about the latter — that is, how dangerous synthetic products can be for investors — that they question their right to exist and are told “Oh no, synthetics just meet a genuine investor need”.  Then when the SEC documents how noxious synthetics can be, we are told that the structurers of the product should be indemnified by the buyer’s duty to understand the product.  Well, if the latter is the case, then what possible justification is there for such natural vehicles for legalized fraud to exist?

The structured finance folk need to get their stories straight.

On synthetics in Maiden Lane III

Calculating the exposure of a CDO to synthetic assets is complicated for two reasons:  (i) first because, not only can the CDO itself use swaps to generate synthetic exposure, but also the CDO and RMBS tranches in which the CDO invests may include synthetics; and (ii) secondly, because the tranche structure of CDOs complicates things.

Because it is easier to create synthetic exposure to an asset than to originate an actual loan (remember creating a synthetic asset involves selling protection on an asset, not buying it — thus you just need to find counterparties willing to pay small premia for protection), I will generally assume that the synthetic exposure of a CDO or RMBS is close to the limits permitted in the deal documents.  This is an assumption and therefore subject to correction if the actual data is ever made public.

The collateral underlying the Broderick I CDO is 20% CDO, 80% RMBS.  20% of this collateral may be in the form of synthetic assets.  Since the industry (and undoubtedly Merill Lynch in particular as a major CDO issuer) had a great need to place junior CDO tranches most likely it was the RMBS that was referenced synthetically, not the CDOs.  So as a working assumption lets consider that the Broderick I CDO is 20% CDO, 20% synthetic referencing RMBS and 60% RMBS.

The thing to remember is that the 20% CDO collateral is likely to also be 20% synthetic.  I’m not going to make any assumptions about the synthetic exposure in the RMBS, because I haven’t found reliable information on the issue, but there is no question that some synthetic RMBS were issued.  Thus Broderick I could easily be backed by 24% synthetic assets — and possibly more.

But it’s important to understand that 24% would be a low estimate of Broderick’s exposure to synthetic assets.  This is because the structure of a CDO is designed to concentrate risk by increasing the exposure of the junior investors to losses..

To explain, consider a simple tranched securitization of five $1 million mortgages with one junior $1 million investor and one senior $4 million investor.  It should be obvious that the junior investor — because he absorbs losses first — has 100% exposure to each of the five mortgages.  If one of those mortgages is synthetic, then the junior investor has 100% exposure to the synthetic mortgage.  In short, in a CDO you must always remember that only the first priority investor is guaranteed to benefit from diversification of assets.

For this reason when calculating the exposure of subordinate CDO tranches to synthetic assets, what is important is whether the detachment point of the tranche (that is the point at which it stops absorbing losses because it is worth nothing) is lower than the fraction of synthetic assets in the CDO.  If the CDO has 20% synthetic assets and the tranche in question detaches at 10%, then the tranche can be wiped out twice over by losses on synthetic assets alone.  Thus it doesn’t really make sense to claim that the tranche has less than 100% exposure to synthetic assets.

Since the subordinate tranches in Broderick (as a group) detach at 16%, every one of them probably has 100% exposure to synthetic assets.  If the CDOs included in Broderick are similarly structured (and if I am right that these CDOs made maximal use of synthetic assets), then it is fair to say that Maiden Lane’s exposure to synthetic assets via Broderick I is $400 million or 40% of the CDO.

Why does this matter?  Because as I asked in my first post on Maiden Lane III as taxpayers we need to consider these issues:

Are we okay with the fact that financiers who saw how dysfunctional our debt markets were didn’t go off and raise hell at the Fed and SEC, but instead expressed their views via the market?  Are we okay with the fact that when one of their counterparties to the expression of these views couldn’t honor it’s obligations, the taxpayer stepped to validate the existence of this market?  Is there a difference between bailing out banks that were financing real economic activity and grossly underestimated the risks of that activity and bailing out traders who used derivative markets to express their views on the dysfunctionality of the debt markets?

I think we need public disclosure on each of the Maiden Lane vehicle’s exposure to synthetic assets.  So we can have a robust public discussion about the role of government in underwriting synthetic assets.

On Maiden Lane III – 2

This post will continue my effort to understand Goldman Sach’s huge first-priority exposure to a few of the CDOs in Maiden Lane III.

Yves Smith has some nice clues to what was going on, pointing out that the November Blackrock memo at the time of Maiden Lane III’s formation states:  “Access to assets:  Goldman has said that it does not hold the cash CDOs, but has back-to-back swaps on most of the positions”.  (I’ll address the remarkable fact that a 20% synthetic CDO could be considered a “cash CDO” in another post.)  This indicates that Goldman probably sold the first priority exposure in Broderick I on to customers, offering a Goldman guarantee on the returns in the form of a swap.  Goldman then transferred this risk to AIG using another swap.  In other words, this was Goldman’s clients’ CDO exposure that was protected first by a Goldman and then by an AIG swap.  That this is a likely explanation is confirmed by the fact that $7.4 billion of the CDOs in Maiden Lane settled almost a month after the first CDOs were transferred because they were “contingent upon the ability of the related counterparty to obtain the related multi-sector CDOs”.  In short, Goldman probably had to buy Broderick I from its clients before it could make use of the Maiden Lane facility.

However, if we are interpreting the available facts correctly and, if Goldman took almost all of the first priority Broderick I exposure in order to sell it on to clients, then we still need an explanation for why Merrill Lynch rather than Goldman was the firm originating the CDOs.  The answer is probably that Merrill had the collateral and Goldman had the clients.  Although Merrill wasn’t a big player in the mortgage market (and purchased First Franklin in order to change that situation), Merrill was one of the lead issuers of CDOs (2004 thru 2006).  It is likely that Merrill had established an RMBS pipeline while GS had clients to whom senior CDO tranches could be sold.

It occurs to me that because the senior CDOs that Goldman was selling to clients were like covered bonds (that is investors were protected by the guarantee of the bank in case the mortgages themselves went into default) and the legal structure for the covered bond market does not exist in the US, there may have legal reasons for the issuer of the CDO and its guarantor to be distinct parties.

This innocuous explanation of Goldman’s large first priority exposure to Maiden Lane’s CDOs does not, however, obviate the main concern of my previous post:  There is still plenty of reason to be concerned that Maiden Lane III has far too much synthetic exposure for taxpayers’ comfort — in part because the CDOs in question were issued right at the time that synthetic RMBS started to become more common.  More in the next post.