When marketing a CDO deal, the first words spoken to the investor are often: “The most important aspect of picking a CDO is selecting a man- ager; so look at the track record of this manager.” But it is very difficult
430 OTHER CDO TOPICS
for investors to assess a manager on track records alone, as they do not necessarily allow easy comparison. The best one can hope to establish is that a manager has been managing that particular asset class for a long period of time, their investment approach can be articulated clearly, and risk management parameters are strictly adhered to.
There is good reason to be very skeptical about track records. They contain three biases—“creation bias,” “survivorship bias,” and “size bias.” We briefly discuss these biases.
“Creation bias” refers to the fact that an investment can take funds that were run in another form—say a bank trust fund, a limited partner- ship, or an insurance company “separate” account—and convert them into a mutual fund. The mutual fund can now claim the track record of the old entity. Obviously, only “good performers” are morphed into mutual fund form, creating an upward bias to returns. Here is another example. Mutual fund management complexes often start incubator funds. Ten new funds may be created with different in-house managers, each with a small amount of seed money. Then, the mutual fund parent waits to see which are successful. Suppose after a few years, only three produced notable total returns. Then only these are marketed, with their attendant track records. There is nothing dishonest about it. It allows a mutual fund to take a chance on a number of young portfolio managers, and discover one who displays real talent. Plus, since the seed money comes from the mutual fund complex’s pockets, investors have not lost money due to poor performance. But it clearly creates an upward bias to reported returns, and makes it all that much more diffi- cult for investors to pick managers.
The bad track record for a fund can be wiped out by merging it into a sibling fund, selling it, or returning money to investors. These tech- niques all wipe out a fund’s bad track record and create “survivorship bias” which can take another form: If a manager with a good tack record leaves one firm, both the fund and the manager can take credit for the favorable track record. (The manager does have to show that they had full control over investment decisions at the former fund.) Nevertheless, it is really difficult to apportion credit or blame for a fund’s performance between a particular manager and the parent fund company. Success in one environment does not necessarily translate into success in another. For example, an outstanding manager who had an army of talented analysts may do far less well in an environment with no support, where he is his own analyst. Moreover, good track records are “adopted” by both the departing manager and the fund left behind, bad track records are used by neither, as the fund gets quietly merged into another one under a different manager.
c22-EvaluateCDO Page 430 Monday, March 6, 2006 11:28 AM
The last bias is “size bias.” As funds do better they tend to accumulate more money to invest. However, they may not be able to replicate that per- formance with more assets, and returns are not dollar weighted, and are hence overstated. For example, assume a fund initially sized at $10 million earned 50% one year. The fund then attracts a considerable amount of new money, say $85 million, which when added to the $5 million “profit”
and the $10 million invested at inception, gives it $100 million at the end of the first year. Assume it then proceeds to lose 20% the following year.
The fund will report a 9.54% annualized return, as a dollar invested at inception is worth $1.20 over 2 years [(1.5 × 0.8 = 1.20), corresponding to an annual return of 9.54%]. However, the average investor lost money.
The fund made $5 million for a small number of investors the first year, then lost $20 million for a much large number of investors the second year.
Again, we see that it is hard to evaluate track records.
While this discussion is very quick, there is a great deal of academic literature on these biases as they pertain to the mutual fund area.1 The same biases apply to fixed income funds.
Common Sense
The key to evaluating manager performance is to use common sense. Do not be duped by performance numbers. Here is what to look for:
■ Make sure the firm has a track record with every asset class it is includ- ing in the CDO, and that the money manager is not stretching into asset classes in which they have not historically been active.
■ Make sure the firm has a disciplined, consistent approach to investing, which is followed in good times and bad.
■ Look at the stability of both the firm and the manager. A management team that has been at a firm for a long period of time, with significant equity, is less likely to leave. (Ideally, CDO investors would like to handcuff managers to the firm for the life of their deal. One obviously cannot do that, but bigger manager stakes mean there is less likelihood of leaving.) Moreover, the longer a group of people has been working together, the less chance of a sudden shift in strategy.
There is an assumption on the part of investors that Wall Street dealers who underwrite CDOs act as gatekeepers, allowing only the top- notch performing managers to pass through their pearly gates. That
1 A fuller discussion of these issues can be found in Burton G. Malkiel, A Random Walk Down Wall Street (New York: Norton, 1999).
432 OTHER CDO TOPICS
blind trust, however, is to some extent misplaced. More money manage- ment firms wish to manage CDOs than there is dealer pipeline capacity.
Thus, a dealer wants to underwrite CDOs (from managers) they believe will sell quickly.
However there are often other considerations, including overall quality of the relationship between the dealer and a money manager, as well as help the asset manager can provide in marketing the deal and taking some of the equity. Consider two money managers; one has a very good track record, the other only an average one. The manager with the average track record will take all the equity in the CDOs, plus some of the subordinate securities. The manager with the better track record wants the dealer to market all the equity. Who will the under- writers pick? It’s a no brainer—the manager with the average track record who is willing to provide more help in underwriting the deal.
Realize that the Wall Street dealer community does require at least a minimum performance threshold. The manager’s investment philosophy and track record do have to be good enough to market the deal. More- over, since dealers are looking at the overall quality of the relationship between the dealer and asset manager, as well as an asset manager’s willingness to take down some of the equity, it is natural that larger, bet- ter established money management firms are likely to have an edge. This is a good thing for investors, per our common sense tests above.
Managing a CDO Portfolio
So far, we have examined what to focus on when looking at an individ- ual deal—making the case that rather than focusing on the manager’s track record, focus on the performance of outstanding CDO deals, and how the manager has balanced his interests with those of the notehold- ers. We now shift gears, and examine the argument that not only should CDO buyers look at individual deals, but they should look at their CDO holding in a portfolio framework.
The key to managing a CDO portfolio is diversification. One of the few indisputable facts is that the types of securities purchased (the style) is key—far more important than skills of a particular manager. Roger Ibbotson, one of the key researchers in the performance area, writes:
. . . relying on past performance is not as simple as it appears. The investment styles of mutual funds typically explain more than 90 percent of the variation in returns.
Just knowing that a fund is a large or small capitalization fund, a growth or value fund, an international stock fund, or a combination of these categories largely explains its
c22-EvaluateCDO Page 432 Monday, March 6, 2006 11:28 AM
performance. The skill of the manager is demonstrated rel- ative to the fund’s investment style . . .2
While it is indisputable that style matters, there is a question as to whether good or poor performance in one period is indicative of the per- formance going forward. That is, are some managers just far superior to others? While there have been studies of mutual funds that have exam- ined this issue, in short, the debate seems to be whether style (asset class) accounts for 90% or 99% of return variation. There is no disputing the fact that it is the key factor. Bottom line—diversify across asset classes.
Do Not Ignore CDO’s with Low-Diversity Scores
Many investors buying a large number of positions still tend to look at each purchase individually. Yes, it is important to look at each deal.
Some parameters, however, may be unacceptable if a particular deal was the only one purchased, and less important when the security will become part of a portfolio. Diversity is one such parameter.
In fact, it is important to look at holdings on a consolidated basis.
Adding deals with low diversification may, in some circumstances, help a CDO portfolio. For example, a REIT-only deal may have a low-diver- sity score, but if it was part of a larger CDO portfolio, and REIT hold- ings elsewhere are limited, then purchasing it may actually increase diversification. By contrast, if one purchased three high-yield deals within a short period of time, each with very high-diversity scores, the additional diversification provided by buying all three deals may actu- ally be limited, as they may own substantially the same securities.
In point of fact, one can argue that debt holders are actually better off holding a portfolio of deals with low diversity scores rather than a portfolio of deals with higher diversity scores and overlapping holdings.
This is because the rating agencies tend to require more subordination on a deal with a lower diversity score. However, when an investor purchases a large number of CDOs, they are creating their own diversification.
There is a secondary effect. In trying to raise diversity scores, manag- ers often venture into asset classes or industries they are less familiar with. We saw in Chapter 8 that a disproportionate amount of ABS CDO downgrades were from ABS CDOs with high-diversity scores. These deals added off the run asset classes such as mutual fund fee securitizations and franchise loan securitizations to boost diversity scores and yield.
The practical advice for managing a CDO portfolio is:
2 Roger Ibbotson, “Style Conscious,” Bloomberg Personal (March/April 2001).
434 OTHER CDO TOPICS
1. An investor should not shun low diversity score deals since the investor also creates his own diversification.
2. An investor should look at holdings in his CDO portfolio on a consoli- dated basis.
CONCLUSION
With the rapid growth of the CDO market over the past few years, a fair number of portfolio managers have amassed large CDO portfolios.
However, unlike other portfolios, CDO positions tend to be accumu- lated one-by-one rather than within a portfolio framework.
In this chapter, we argue the importance of evaluating each CDO deal and each CDO manager. However, it is less important to focus on a manager’s overall track record and more important to focus on common sense items—length of time the management team has been together, whether they have followed a consistent investment philosophy, impor- tance of interest risk management, and so on. It is also critical to focus on the manager’s reaction to potential conflicts of interest in previous CDO deals they have managed.
Again, one should realize that the key determinant of CDO perfor- mance is likely to be the type of assets held, not the contribution of any individual manager or deal. Thus a portfolio of CDOs should consist of different types of collateral, each purchased when it is cheap. In addi- tion, portfolio diversification should be judged in a total portfolio framework, not on a CDO-by-CDO basis. Thus it is very important to aggregate across CDOs in a portfolio to see the collateral composition and diversification.
c22-EvaluateCDO Page 434 Monday, March 6, 2006 11:28 AM
CHAPTER 23
435
Quantifying Single-Name Risk Across CDOs
ingle-name risk in collateralized debt obligations (CDOs) arises from the presence of the same credit in the portfolios of different CDOs. If the same credit appears in several, or all, of the CDO portfolios an investor owns, there may be an unexpectantly large exposure to that particular credit. In the extreme, if every CDO had the same exact set of underlying credits, there would be no diversification benefit from own- ing different CDOs. In this chapter, we quantify the extent of collateral overlap among a sample of U.S. collateralized loan obligations (CLOs) and structured finance (SF) CDOs (i.e., CDOs backed by asset-backed securities, commercial mortgage-backed securities, and residential mort- gage-backed securities). We then propose a simple and consistent mea- sure of single-name risk applicable across CDO tranches of various seniorities. Finally, we review a more complex, high tech approach to CDO single-name risk.
CLO portfolios, even from CLOs issued in different years, tend to have a lot of underlying borrower names in common, and this is espe- cially the case for CLOs managed by the same manager. In this chapter, we name the most common underlying borrowers among a sample of CLOs.
For SF CDOs, the single name of interest is the originator of the ABS, CMBS, and RMBS assets in the portfolio. Many defaults in struc- tured finance have been originator-driven, related to too-lenient under- writing standards or even to fraud by the originator. However, the high credit quality of the collateral within SF CDOs ameliorates some of the single-name originator risk. So we look closely at SF CDO risk to origi- nators of downgraded collateral.
S
436 OTHER CDO TOPICS
Our measure of single-name risk compares the concentration of a sin- gle credit across CDO portfolios to the credit protection afforded the par- ticular tranches an investor owns. On this basis, our analysis shows little reason to be concerned about single-name risk for the CLOs and SF CDOs we studied, except perhaps at the level of subordinate debt and equity.
COLLATERAL OVERLAP IN U.S. CLOs
We examined the U.S. CLO portfolio of a major CDO investor. This investor owned the first six CLOs we list (in an opaque manner1) in Exhibit 23.1. Note that this investor bought two CLOs from Manager A, a 1999 vintage and a 2001 vintage. To test some conclusions about how managers buy and allocate loans across the CLOs they manage, we added a 2003 CLO from Manager A to our study. So all together we looked at seven CLOs, six purchased by a particular investor and one additional CLO. Among these seven CLOs, three have the same man- ager. These CLOs have collateral portfolios ranging between $350 mil- lion and $1 billion. All of these CLOs are healthy, none of their debt tranches have been downgraded, and all their tranches are passing their par and interest coverage tests by good margins.
To quantify single-name risk, we first looked at the percent of a CLO portfolio, by par amount, made up of obligations of the same credit. For a particular credit, collateral overlap is defined as the aver- age percent of that single name in two CLOs. Thus, since 1.0% of 1999 A’s portfolio is comprised of Invensys and 2.3% of 2000 B’s portfolio is comprised of that same credit, the collateral overlap from Invensys
1 Why the secrecy? To protect the investor’s privacy, because we are relying on un- derlying collateral data we cannot verify, because we wish to make general points about the CLO market rather than specific points about individual managers, and because we do not want to increase our hate mail from disgruntled CDO managers.
EXHIBIT 23.1 CLO Vintages and Managers in Study 1999 Manager A
2000 Manager B 2001 Manager A 2001 Manager C 2002 Manager D 2002 Manager E 2003 Manager A
c23-QuantSingleName Page 436 Monday, March 6, 2006 11:28 AM
between the two CLOs is calculated as 1.7%. The total collateral over- lap between 1999 A and 2000 B is the sum of these individual name col- lateral overlaps.
Note that our measure of single-name risk focuses on exposure to a particular name across CDOs. It does not, however, address single- name risk within a particular CDO. For example, if a CDO portfolio consisted of only one credit, the CDO’s debt and equity holders would bear huge single-name risk. But if that credit was not also in another CDO, our measure of single-name risk across CDOs would show 0%
collateral overlap between the two CDOs. In practice, single-name risk within a particular CDO is handled by that CDO’s concentration limits.
In this chapter we are is concerned about single-name risk arising from the presence of a particular name across multiple CDOs.
To determine whether two CLOs contain the same credit, we first looked to the “Issuer” data field in INTEX’s asset detail. Unfortunately, INTEX does not have a single unique issuer identifier across all CDOs.
Nor does it group affiliated legal entities, such as holding companies and their subsidiaries. Therefore, we grouped single-name risks “by eye.” Also, the portfolio for one of the CLOs was not available on INTEX, so we obtained it from Moody’s EMS database.
Exhibit 23.2 shows that, by far, the greatest collateral overlap occurs among the three CLOs managed by the same manager. Pairwise overlap between Manager A’s three CLOs is 71%, 68%, and 58%, respectively. The next highest collateral overlap between any two CLOs is 45%, from two CLOs issued in 2001. Apparently, having the same manager produces higher collateral overlap between CLOs than does being in the same vintage year.
The far right column in Exhibit 23.2 shows the average collateral overlap between a particular CLO and the other six. This varies from 30% to 46%. But across all possible pairs of CLOs, the collateral over- lap varies from 25% to 71% and averages 38%. Ignoring the collateral overlap between the three CLOs managed by the same manager, collat- eral overlap varies from 25% to 45% and averages 34%.