Do Current Cmbs Pricing Conventions Make Sense?

®

Do Current CMBS Pricing Conventions Make Sense? Peter Rubinstein TOWER

OF

BABEL

Bond trading used to be much easier. Virtually all fixed income securities were offered on a spread to Treasuries basis, which facilitated comparisons across markets. Now, methods of pricing are all over Rubinstein the map. In commercial mortgagebacked securities (CMBS), investment-grade bonds are priced on a spread to swaps basis (using an interpolated average life and the “J” curve method), but below-investment-grade bonds are priced on a spread to interpolated on-the-run Treasuries, using the “I” curve method. CMBS IOs (interest only) are also priced as a spread to interpolated Treasuries, but illogically, to a much longer maturity than their cash flow average lives, which can grossly understate the spread earned in comparison to other bonds, particularly in a steep yield curve environment. ABS pricing is worse. They use swaps, “E” curves, “N” curves, EDSF and, on occasion, still price to unique, off-the-run Treasuries. Similarly, residential MBS, derivatives, corporate bonds, and other fixed income securities all use their own, distinct pricing conventions. Custom pricing conventions, like any specialized language, have benefits, but there are also costs and inefficiencies. Few players now understand all of the pricing conventions, which makes comparisons across markets a difficult, time consuming and exacting task at best. Converting all prices to a common basis is not a trivial task. In this article, we ask whether the new conventions used to price CMBS are (1) still logical, and (2) worth the added cost and complexity. With that goal in mind, we start by looking at how we got here, what “here” really means, and then discuss where we might go from here, and why.

A BRIEF HISTORY MESS ANYWAY?

OR

HOW DID WE GET INTO THIS

Most of you recall the unprecedented dislocation of fixed income spreads (when measured with respect to Treasury yields) in the Fall of 1998. The volatility was disorienting because the data was sending out strong but conflicting signals. On the one hand, wide spreads suggested high levels of risk.1 On the other, the economic 50 CMBS WORLD™

data was strong, indicating a low level of risk1. The message in the economic data was supported by the fact that triple-A yields were largely stable: the spread widening came solely from the rally in Treasuries. In CMBS, the confusion was amplified. The market was too young to independently arrive at a risk premium. An outside proxy for credit risk was needed. Confusion alone would probably not have changed market pricing conventions, but in addition, the turmoil caused a lot of people to lose a lot of money. Most trading desks had hedged their positions by selling Treasury bonds short. Many issuers hedged their pipeline the same way. Obviously, some investors also used Treasuries to hedge. They all took a bath because the hedges lost money, and the assets either lost money, or barely appreciated. Faced with the poor hedging results and conflicting informational content, the market began to search for an alternative benchmark. The search gained urgency when government surpluses started to shrink the supply of Treasury bonds. A number of alternatives were examined, such as Agency debentures, but the swaps market offered certain key advantages. The market was large, had broad participation, and many players who may never actually have traded in swaps still used swap spreads on a regular basis to establish an alternative benchmark for relative value. Further, movements in swap spreads appeared to be well correlated with other asset spreads. Some participants actually claimed that swap spreads caused other asset spreads to move. In March 1999, the ABS market offered its first deal priced to swaps (AMEX 99- 1). CMBS and other segments of the ABS market soon followed.

WHERE ARE WE NOW? A SWIPE

AT

SWAPS

Was the shift to swaps good? Initially, it appeared so. Numerous regression studies were produced to show the high correlation between swap spreads, CMBS spreads, and many other fixed income spreads. The claim was made that swaps are a good proxy for credit risk and this became a major supporting argument for using swaps as a benchmark. But there were problems that were overlooked too. First, the correlations did not always hold. Some of the historical regression studies looked good because they skipped periods of time where swap spreads did not correlate well. That should have been

Do Current CMBS Pricing Conventions Make Sense? (cont.)

a tip-off about the consistency of using swaps as a benchmark. This consistency problem has now showed up several times in real life. For example, Treasuries rallied after the World Trade Center attack, just like in 1998, due to a flight to quality, but this time around, swap spreads tightened, just the opposite of what would be expected if swaps were a good proxy for credit risk. In CMBS, spreads to swaps decoupled after the World Trade Center attack and widened significantly, yet they remained stable to Treasuries. Similarly, the 10-year part of the swap curve is periodically distorted by unusually high demand from MBS servicers, originators and hedge funds, all of whom can use swaps to offset the loss of duration suffered from rapid prepayments. And, as we know from past experience, corporate issuance (or the lack thereof), the slope of the Treasury curve, the supply of Treasury bonds, and other factors, many of which are unrelated to credit, influence the level and shape of the swaps curve. Note also the interpretation of the high correlations in these regression studies has not always been right. Regressors in a regression are called “explanatory” variables, but they don’t actually explain anything, a caveat highlighted in almost every econometrics textbook.2 Even when there is a causal link, the direction of causality is not specified by the regression. In other words, if you switched the “dependent” and “independent” variables in a typical swap spread versus CMBS spread regression, you would get the same high R2, “proving” the absurd statement that the movement in CMBS spreads “causes” swap spreads to move! Theory

What is missing from the swaps studies is a theory that establishes a direct cause and effect linkage between swap spreads and CMBS (or other asset) spreads. The theory doesn’t exist, however, because the relationship between swaps and the rest of the fixed income market is not causal. What actually drives pricing in the markets, beyond the pure and simple supply and demand for money which is captured in overall interest rates, are multiple sources of risk that come in two flavors: systematic risk and unsystematic risk. These risks often act independently of each other and generally occur simultaneously, which has induced some of the odd results we have seen over the past few months, for example, the credit induced rally in Treasuries coupled with the tightening in swaps. To understand these impacts, you need to look at the effect of each risk on a stand-alone basis, and then realize that at any point in time, either one can dominate in determining the ultimate prices in the market. Systematic risks operate like the tide causing all ships to rise and fall together. They stem from macro level events, like the current war and recession we are now in. Since systematic risks cause all bonds to respond together, under pure systematic risk,

swaps (which are credit sensitive) respond along with other bonds, and so perform well as both a benchmark and as a hedge. There is a twist, however. Under systematic risk, Treasury bonds usually break away from the rest of the market because they are the only (credit) risk-free alternative in the world. As the only safe haven, an exaggerated amount of demand for safety gets channeled into the Treasury market, causing an exaggerated widening in spreads to Treasuries. This is exactly what happened in 1998, and shows why swaps, under pure systematic risk, are a more stable benchmark and better hedge than Treasuries. Note also that in these circumstances, it appears as if swaps

“

Custom pricing conventions, like any specialized language, have benefits, but there are also costs and inefficiencies.

”

cause other assets to widen, but that is only because the swaps market is the most liquid and highly watched market next to Treasuries. We simply see the widening there first. Swaps spreads are a good signal, but not a cause.

On the other hand, markets also experience unsystematic risks, which are often the result of technical factors specific to one or just a few markets. The swaps market is subject to these specific, unsystematic risks just as much as any other asset, and sometimes the impact of unsystematic risks dominate in pricing, which explains why swaps sometimes fail as a benchmark and a hedge, and why swap spreads sometimes appear to behave erratically. It explains, for example, why, in the post World Trade Center attack, swap spreads tightened (instead of widening to reflect increased systematic credit risk as in 1998) as Treasuries rallied. The combination of the steepening curve and demand by mortgage market participants for receive-fixed swaps dominated the credit induced widening. So empirically, what we have seen over the past few months is that both treasury bonds and swaps sometimes fail to work as expected, both as benchmarks and as hedges. How to Proceed

The fact that swaps are merely a signal, and the fact that risks specific to the swap market can make them a poor benchmark and a poor hedge is a complete reversal of the thinking in 1998 when we were convinced that swaps were the answer. The essence of the problem is that there is no one number that will always function as a stable benchmark, which means we must compromise. The best we can do is identify the performance criteria that define a good benchmark, and then see how the various alternatives stack up. I suggest starting with the following criteria, but do not claim this to be a complete list: SUMMER 2002 51

®

Do Current CMBS Pricing Conventions Make Sense? (cont.)

®

1. Tamper Proof. While this criterion may be obvious, unfortunately it may not be possible to find a benchmark that is truly safe from manipulation. Even the Treasury market suffered manipulation a number of years ago. The more tamper proof, however, the better. 2. Liquidity and High Trading Volume. High trading volume insures that the benchmark can always be observed and that it reflects all information in the market, while liquidity (i.e., a low bid-ask spread) insures that the value is accurately measured. Liquidity also is important because there will be a natural tendency to use the benchmark as a hedging vehicle, although we submit that hedging strategies need to be independent of the choice of benchmark because different assets contain different mixes of risk. 3. Broad Participation. A relatively small number of participants can make a liquid market and sustain a high trading volume, but for a benchmark, it is even better if there are many participants trading in the market. It makes the market more stable and allows more information to be reflected in the marketplace. 4. Transparency. It should be easy to understand how a benchmark’s value is calculated. For example, just about everyone in the fixed income markets knows how to calculate the present value of a Treasury bond at any point in time, but fewer people know how to value a seasoned interest rate swap several years after it was created.3 5. Consistency. The market factors that move the benchmark’s price and the rules that define the benchmark and its marketplace should be stable. For example, Treasury bonds recently failed in this area when the Treasury changed the rules by announcing the suspension of the 30-year bond. It put undue pressure on the 10-year bond and will effectively eliminate the longest benchmark. Of course, swaps have also failed in this area; for example, when mortgage originators recently swamped the swap market with demand for received-fixed contracts to try and offset the duration they lost in the mortgage market.

CMBS market needs to be larger, and the product needs to be more homogenous, before participants can consider buying and selling on just a dollar price basis. Given the impracticality of pricing on a dollar basis, we submit that the second best solution is to abandon swaps and price CMBS on a spread to Treasuries basis. Corporate bonds, agency debentures and other major asset classes are still quoted this way, and CMBS structures and convexity are far more like corporate bonds and agency bonds than most ABS and residential MBS. Further, Treasuries have the best liquidity, best transparency, broadest participation, probably the best protection against tampering, and probably the highest trading volumes. The downside is that Treasury bonds are not always consistent. The U.S. Treasury occasionally changes the rules and there is not much that can be done about that. Note, however, that the distortion caused by periodic flights to quality is well understood and should not cause a problem in the markets. In fact, you can argue that the fact that triple-A yields in 1998 remained relatively constant in the face of a flight to quality is confirmation that the markets can correctly equate the risk and reward embedded in each asset separately from movement in Treasuries. The biggest problem with using Treasuries as a basis, many will argue, is that they can, and sometimes have, failed when used for hedging purposes. So have swaps, and the cost of going in and out of swaps is much more expensive. Realistically, hedging strategies need to be independent of the pricing benchmark because empirically, no benchmark works well all of the time as a hedge. Hedging is still an art as much as a science. ❑

Peter Rubinstein, Ph.D., is Managing Director at Bear, Stearns & Co. Inc. 1

For example, real (i.e., inflation adjusted) oil prices were approaching all time lows, and interest rates had come way down, both of which dropped the cost structure of our economy, made investment more attractive, and put more money in consumers’ pockets. In addition, the government was running a huge surplus, unemployment was approaching lows not seen since 1970, and weak foreign economies made imports cheap.

2

Peter Kennedy, in “A Guide to Econometrics,” 3rd Edition, MIT Press, p. 68, perhaps says it best: “It is usually assumed that movements in the dependent variable are caused by movements in the independent variable(s), but the existence of a relationship between theses variables proves neither the existence of causality nor its direction” (emphasis added).

3

A good discussion of how swap prices are tied to the Eurodollar futures market can be found in Chapter 15 of “Derivatives” by Fred D. Arditti, Harvard Business School Press, 1996.

SO, WHAT SHOULD WE DO? My ideal: new issue 10-year triple-A CMBS should trade in the secondary on an absolute dollar price basis, or perhaps on a dollar price “behind” some other, more liquid, asset. People are perfectly capable of figuring out yields and spreads on their own. Absolute yield levels would also work, although yield levels introduce the complication of making prepayment and default assumptions. Traders tell me that these suggestions are a pipe dream, but remember, GNMA passthroughs did not always trade on an absolute price basis either. Realistically, however, the 52 CMBS WORLD™