Martin S. Fridson

Forecasting Default Rates on High-Yield Bonds

J ~ N G. J ~ N S S O N is an assistant vice president at Merrill Lynch & Co. in New York. here is a voluminous research on the prediction of bond defaults, in the sense of classifylng individual issuers by their propensity to default.’ Much less T work has been devoted to forecasting aggregate default rates. This represents a significant information gap for high-yield bond portfolio managers, given the demonstrated relationship between aggregate default rates and market performance.2 Several sources assert causal relationships between fluctuations in default rates and economic or financial factors, even if few analysts attempt to incorporate the factors into formal models. Typically, such comments focus on business conditions and the quality of outstanding bonds. Spahr and Sunderman [1994] claim that “default losses tend to vary with economic condltions and cycles .... During economic downturns both the frequency and severity of default losses are high.” Altman and Nammacher [1987], on the other hand, find it difficult “to establish a clear connection between the aggregate business failure rates in the United States and corporate default rates,” even though the two series are “seemingly related.” Howe [3988] suggests that, along with the condition of the economy, the general level of corporate debt may influence the default rate. Similarly, Lucas and Lonslu [1991] argue that in appraising the potential impact of a recession, the quality profile of the highyield universe should be considered. They link the higher default rate of 1990-1991, compared to the rate experienced during the more severe recession of 1982, to less concentration in the most default-resistant segment (issues rated Ba) and poorer measures of interest coverage and financial leverage. Lucas and Lonski also

Spread Versus Treasuries and the Riskiness of High-Yield Bonds

J ~ N G. J ~ N S S O N is an analyst in high yield credit research at M e d Lynch & Co. in New York. ridson and Bersh [1994] show that the high-yield spread versus Treasuries, measured against its historical average, is not a useful market-timing tool. Backtesting over the period 1985-1992, the authors determined that purchases of the Merrill Lynch High Yield Master Index at wider-than-average spreads resulted in quarterly returns slightly lower than (but statistically indistinguishable from) the quarterly returns on purchases made when the spread was wider than average. TheJFI editor commented at the time: “While we understand that it is conceptually possible, the almost total absence of usefulness of the yield spread is striking, and perhaps disturbing for traders and researchers” (Breeden [1994, p. 11). This comment is immensely gratikng, as we are not keen on producing research that confirms the conventional wisdom. Upon hrther reflection, the truly dsturbing result would be to discover that excess returns consistently result from a trading rule as simple as, “Buy when the spread versus Treasuries is wider than its historical average.” The spread, after all, represents information that is readly available to institutional investors. It can be found on the Bloomberg terminals, for example. Easy arbitrage profits fiom such data would seriously challenge existing notions of informational efficiency in the capital markets. A hidden assumption fatally undercuts the widely held belief that the hgh-yield sector must be cheap if its risk premium (the yield spread over default-free U.S. Treasury bonds) exceeds its past mean level. Those who so argue are in effect positing that the correct spread, at all times, is the hstorical average. If the spread over Treasuries is truly a risk pre-

Contingent Claims Analysis

The authors examine whether contingent claims analysis (CCA) completely explains the price relationships within a company’s capital structure. According to CCA, a company’s bond and stock values must invariably move in the same direction. Sometimes CCA suggests the presence of an anomaly or pricing discrepancy. CCA proponents assert that such an anomaly provides an opportunity for gains through riskless arbitrage. The authors contend that CCA provides only a partial representation of reality because the model cannot capture some factors that justify a price above or below the level that the model predicts.

Semiannual Seasonality in High-Yield Bond Returns

Exploitable seasonal patterns in high–yield bond returns are not limited to the widely studied “January effect.” On average, the high–yield sector outperforms ten–year Treasuries by a wider margin between December 1 and May 31 than between June 1 and November 30. Within the high–yield sector, single–Bs outperform double–Bs on average between December 1 and May 31, and underperform them between June 1 and November 30. The author finds that the semiannual seasonality effect does not appear to reflect seasonality in capital flows to the high–yield sector, but arises rather from a seasonal pattern in Treasury bond returns.

Valuing Like-Rated Senior and Subordinated Debt

New York. little-remarked-upon feature of Moody’s and Standard & Poor’s bond rating systems is the fact that a given rating can apply to issues with quite A different estimated probabdities of default. For instance, S&P uses its B+ rating not only for senior issues of companies with default probabilities in the high single-B range, but also for subordmated issues of companies with default probabllities in the mid-double-B range. Recopzing that difficulties may arise when ratings do double duty, Moody’s compiles its default rate statistics on a senior-equivalent basis, rather than grouping issues according to their actual ratings. With few other exceptions, however, research on cre&t experience and on the relationship between ratings and yields lumps senior B1 issues with subordinated B1 issues, senior BBissues with subordinated BBissues and so forth. As we shall demonstrate, failure to distinguish between like-rated bonds of different priorities in the capital structure can foster misperceptions about the relationship between agency ratings and bond pricing. In particular, our informal survey of practitioners discloses a widespread belief that a senior bond ought to yield less than a like-rated subordinated bond. In fact, we show the opposite to be true, both in theory and in practice. We also find that the relative premium on senior debt should and does increase as ratings decrease. Finally, we offer a possible explanation for the paradox of how the market seems to arrive by an incorrect line of reasoning at the correct relative valuation of like-rated senior and subordinated bonds.

Downgrade/Upgrade Ratio Leads Default Rate

Fluctuations in the default rate on high-yield corporate bonds are among the factors that explain variance in the sectors returns. Accordingly, several econometric models have been developed to forecast the default rate. One plausible explanatory variable not normally included is the trend in credit rating changes within the speculative-grade category. The ratio of Moody’s Investors Services downgrades to upgrades explains approximately one-half to two-thirds of the change in the default rate two to three quarters later.

High-yield indexes and benchmark portfolios

can’t please everyone,” as the saying goes, but high-yield bond indexes seem to please no one. Whether viewed from the vantage point of an investment advisory firm’s senior management, its clients, or its portfolio managers, or else by a bond salesperson, regulator, journalist, academician, or consultant, the faults of indexes always appear fatal, yet at the same time easily remediable. What explains the paradox of a costly problem that is universally recognized, yet perpetually resistant to seemingly simple and obvious solutions? Perhaps the difficulty is that no high-yield index can satisfy every category of user. The flaw perceived by a portfolio manager, whose performance is measured by the index, may be correctable with one stroke, but only at the cost of magnifying the shortcoming seen by a pension fund sponsor, who is the one doing the measuring. Tinkering with the index, meanwhile, accomplishes little for the chief investment officer, except to intensify the suspicion that somebody is trying to pull a fast one.

Spread Versus Treasuries as a Market-Timing Tool for High-Yield Investors

To calculate the spread versus Treasuries, we begin with the yield to maturity on a defined class of risky (non-default-free) bonds. We then subtract the yield to maturity on riskless bonds, matched for maturity or duration in order to filter out the term structure of interest rates. The difference represents the yield premium that investors are receiving to opt for holding the risky securities. Over time, this spread varies across a considerable range. Instances of unusually wide spreads are succeeded by periods of average or exceptionally narrow spreads. Simulation readily demonstrates that a risky bond purchased at an abnormally wide spread, which then narrows vis-&vis its corresponding Treasury issue, d generally produce a higher total return than the Treasury. Conversely, underperformance generally results h m buying during periods of exceptionally tight spreads that are followed by spread widening. Given these facts, a comparison of the prevailing spread versus Treasuries to the historical average ought to be a good indicator of a risky debt category’s relative attractiveness.

Return Dynamics of Distressed Bonds

This study adds to the small, but growing, body of research on active management of distressed debt. The analysis stratifies the distressed debt universe in various ways to test whether certain subgroups exhibit comparatively low risk or comparatively high returns, thereby representing superior investment opportunities. A small group of distressed bonds rated double-B do provide statistically lower default rates and higher bear market returns than single-B or triple-C issues. The authors also find no comparable advantages for bonds issued by public companies over bonds issued by private companies, which challenges the prevailing investor preference for public companies. And contrary to popular belief, agency ratings do appear to provide information content in the speculative grade market.

Performance of Distressed Bonds

Distressed debt is a large and growing segment of the fixed-income market. In this article, the authors expand the empirical basis for active management with evidence on risk and reward. They calculate a one-year default rate of 22% for the category, whether defined by spread or by price. That rate is more than 20 times as high as non-distressed speculative-grade debt. They also address the longstanding question of whether non-distressed investors should automatically take losses on bonds that fall below a stated price threshold. The evidence on price change argues against this strategy, based on a threshold of 70, but the cost of continuing to hold is a higher default rate than many investors can tolerate. Finally, a strong correlation indicates that distressed investors earn the highest returns on the lowest-priced bonds, but there is a severe penalty for selecting the wrong issues.