Efficient Frontier
Efficient Frontiers Logo
William J. Bernstein

Credit Risk: How Much? When?


I’m often asked about the place of junk bonds in a diversified portfolio. The standard Chicago-Santa Monica answer to this question is, "Never, ever: Bonds should always be short and of very high quality. Yes, there is a premium from junk bonds, but over the long haul it is not large enough to justify the extra risk. Further, junk has a high correlation with small stocks, so it offers little diversification value." But high-yield is a remarkable asset class and worth examining, particularly regarding its default rates and expected returns.

Consider a portfolio of B-rated bonds yielding 14%; typically, these are bonds which currently are paying their coupons, but have a high likelihood of defaulting or have done so in the recent past. A Treasury bond of similar duration yields 5.5%. Thus, in this example, the Junk-Treasury Spread (JTS) is 8.5%.

Now, let’s take a look at the spread’s history over the past 13 years. (I’m indebted to Jay Diamond at Grant’s Interest Rate Observer for this data.)

The spread depicted in the above graph corresponds roughly to B-rated debt. Note the very wide range of spreads from just under 3% to almost 10%. What does a JTS of 3% mean? Very bad news for the junk buyer, because he or she will have been better off in Treasuries if the loss rate exceeds 3%. And even if the loss rate is only half of that, a 1.5% return premium does not seem adequate to compensate for this risk. There is a wealth of data out there on the bankruptcy/default rate of these beasts, allowing us to evaluate whether the prevailing risk premium amounts to adequate compensation.

It’s important to briefly discuss the meanings of default rate and loss rate. The former is obviously the proportion of companies defaulting per year. But not all companies that default go bankrupt. The recovery rate is the proportion of defaulting companies that do not eventually go bankrupt. So a portfolio’s reduction in return is calculated as the default rate times one minus the recovery rate: if the default rate is 4% and the recovery rate is 40%, then the portfolio’s total return has been reduced by 2.4%. The loss rate, how much of the portfolio actually disappears, is simply the default rate minus the absolute percent of companies which recover.

According to Moody’s, the annual long-term default rate of bonds rated BBB/Baa (the lowest "investment grade") is about 0.3%; for BB/Ba, about 1.5%; and for B, about 7%. But in any given year, the default rate varies widely. Further, because of the changes in the high-yield market that occurred 15 years ago, the pre-1985 experience may not be of great relevance to high-yield investing today. Before the advent of Michael Milken, the overwhelming majority of speculative issues were "fallen angels"—former investment grade debt which had fallen on hard times. But after 1985, most high-yield securities were speculative right from their initial offering. Once relegated to bank loans, poorly rated companies were for the first time able to issue debt themselves. This was not a change for the better. Similar to speculative stock IPOs, these new high-yield bond issues tended to have less secure "coverage" (an ancient accounting term defined as the ratio of earnings before-taxes-and-interest to total interest charges) than the fallen angels of yore, and their default rates were correspondingly higher.

In the halcyon years of the late ’80s, deals of increasing dubiousness got done, with predictable results for bondholders. Default rates did decrease in the 1990s, but they probably will soon be rising again. In any case, it seems likely that the period from 1988 to 1997 represents a full "credit cycle," and it is useful to examine the cumulative default rates of BB and B-rated bonds over this period.

The BB-rated bonds seem to default at about 2% per year, on average, and the B-rated bonds at about 4% per year. Of course, rates can temporarily be much higher: even 8% to 10% per year at times for B-rated debt. Remember, default does not mean total loss though; about 40% of defaulted debt is eventually recovered. So reckon about a 1.2% annual long-term loss rate for BB-rated, and about 2.5% annual for B-rated.

Finally, there is the issue of default-rate volatility. A debt portfolio with an average yield of 10%, a default rate of 5%, and a recovery rate of 50% is mathematically identical to a risk-free asset with a yield of 7.5%. The problem is, defaults are unpredictable and therefore demand a risk premium. Default-rate volatility can be measured conventionally with standard deviation, with one caveat: default rates are serially correlated (that is, they tend to trend), so the SD understates this risk a bit. The below chart (courtesy of Alexandra Berthault of Moody’s) beautifully demonstrates both the volatility and trending phenomena.

We now have the tools to rationally investigate high-yield investing. The risk premium of junk is simply the JTS (Junk-Treasury Spread) minus the loss rate. To begin, take a look at the graph depicting the cumulative default rate of B-rated bonds. Is it rational to invest in a portfolio of B-rated high-yield securities that offers a 3% yield premium over treasuries? (Northeast Investors Trust and Fidelity High Income, for example, are overwhelmingly composed of B-rated bonds.) Only the dullest of blades would do so. With an annual default rate of roughly 7% and a recovery rate of 40%, we’re losing about 4.2% of return (3% minus [.6 x 7%] ) per year. So we’re doomed here to a return lower than Treasuries, at much higher risk. Historically, the JTS is about 4.5%, and even this does not provide an adequate return premium (4.5% minus [7% x .6] is 0.3%). This is the reason why most pundits recommend that long-term investors stay away from junk.

However, in late November, the Junk-Treasury Spread was at 9%; at that level, things look different. Running the same calculation, we subtract from 9% a loss rate of 7% times .6 and get a risk premium of 4.8%. In fact, this sort of analysis understates the case since it does not take into account potential market-value change. A JTS of 3% leaves little room for further narrowing of the spread—it’s much more likely to rise, which will result in price fall and further erosion of return. Contrariwise, a JTS of 9% is quite likely to narrow, resulting in price gain.

Junk bonds are different from stocks in one substantial aspect: a stock or stock market cross-section can remain cheaply valued or richly valued for decades. There is no fundamental reason, for example, why REITs yielding 8% or industrials yielding 1.3% should not continue to do so for decades. But a distressed bond selling at 60% of par with a maturity of eight years which does not default must of necessity yield its coupon plus a capital gain of 67% after eight years. In other words, in the long term the only way of being burned with a properly diversified high-yield portfolio is if the future average default rate is much higher than has been the case in the past. And remember, the recent past includes the 1989-91 debacle, the worst in high-yield history.

Below I’ve plotted the JTS versus the forward five-year difference in returns between junk and Treasuries.

There’s a pretty clear-cut relationship here: As expected, the higher the spread, the greater the advantage of bearing credit risk.

For a believer in efficient markets, these conclusions are profoundly disturbing, but not unprofitable. Although most of the time, it does not pay to take credit risk, there are periods when expected returns are too high to ignore. Yes, the devout efficient marketeer will point out that there’s a reason why one does not often find $10 bills lying on the sidewalk, and that if this junk-bond opportunity were really a free lunch, it would have been arbitraged out long ago. However, there are limits to arbitrage. I ran smack into this limit at a conference of institutional fixed-income managers recently. It was easy to pick out the "spread investors"¾ they were the ones with the deer-in-the-headlights stare and the portfolios suffering from the bond equivalent of irresistible-force-meets-immovable-object. I’m talking about huge mutual-fund-redemption demands running smack into illiquid, impossible-to-price securities. If you're a small investor with modest portfolio exposure to junk, say 1% to 2%, you can afford to wait a few years for prices to recover. These folks looked like they didn’t even have a few weeks.

Belief in the efficient market theory does not relieve one of the duty to estimate asset-class returns. Because of the term structure of high-yield bonds, returns will tend to mean-revert more quickly, and more surely, than equity. Yes, there is risk. But when their long-term expected returns start approaching 5% over Treasuries (as they did not so long ago), it looks like a risk worth taking with a small corner of one’s portfolio. One caveat: Because most of the return, similar to REITs, accrues as ordinary income, junk bonds are appropriate only for tax-sheltered accounts.

Are we market timing? I suppose. It’s the lesser of two evils—I’d rather violate the efficient market hypothesis than ignore appealing expected returns with a relatively short time horizon.

To Efficient Frontier Homepage E-mail to William Bernstein

Copyright © 2001, William J. Bernstein. All rights reserved.

The right to download, store and/or output any material on this Web site is granted for viewing use only. Material may not be reproduced in any form without the express written permission of William J. Bernstein. Reproduction or editing by any means, mechanical or electronic, in whole or in part, without the express written permission of William J. Bernstein is strictly prohibited. Please read the disclaimer.