Using Bond Fund Duration Is an Art, Not a Science

Interest-rate risk is top of mind these days given that bond yields have risen across the board thus far in 2018. The 10-year Treasury yield hovered at 2.94% on Sept. 7, up from 2.4% at the end of last year, and the Bloomberg Barclays U.S. Aggregate Bond Index has fallen 1.4% for the year thus far. Looking at duration is a key first step in understanding that dynamic.

One problem with duration is that it doesn't--and can't--always provide the answers investors expect from it. The 2008 financial crisis and the resulting market shocks provided a very stark illustration of the issue. Most diversified bond funds struggled or lost money during the crisis, even though they carried plenty of duration at a time when investors the world over were scrambling into U.S. Treasuries, pushing their yields down more sharply than they had moved in years; their prices skyrocketed.

The long-term government and long-term bond Morningstar Categories made for a painful contrast. Usually they behave similarly, as both focus on bonds with long maturities and high-quality debt. Yet the divergence in their 2008 performance was breathtaking: The average long-term government fund gained nearly 28%, while the average portfolio in the long-term bond category--funds that hold mostly nongovernment but investment-grade bonds--fell by more than 3.7%. Why?

Bumping Up Against the Limitations of Duration Under normal circumstances, most high-quality bond portfolios' effective durations would have provided useful guidance about how those funds would respond to Treasury market shifts during the crisis. Funds in both the long-term government and long-term bond categories occupied a wide range of duration territory, but not enough to explain such a broad dispersion of returns. And if that weren't odd enough, a nearly inverse phenomenon occurred as the market snapped back from the crisis in 2009.

Clearly, this behavior wasn't captured by the go-to risk metric of duration. The primary reason is that duration works best for bonds that are similar, and it will normally provide an accurate picture when comparing prices on the same types of bonds. In other words, if you're using changes in Treasury bond yields as your reference, they're going to have the most predictive value for other Treasury bonds.

And under normal circumstances, a typical duration calculation will work pretty well for investment-grade corporate bonds with yields that aren't too far away from their comparable Treasury counterparts--say, within 2 percentage points. Much farther away than that, however, and the measure begins to lose its value as a gauge of how much movement your bonds will see relative to Treasuries. In fact, in order to be explicit about that, many portfolio managers now specify that they're talking about interest-rate, Treasury, or even "hard" duration, and that those metrics won't be as predictive when it comes to the risk that other bonds won't act like Treasuries. Fortunately, there is a tool to help managers understand the risks associated with non-Treasury bonds.

Spread Duration – Adding to the Tool Kit Instead of estimating how much a bond will gain or lose based on changes in Treasury yields, spread duration describes how much a bond's price is expected to move if there's a change in the gap between its yield and that of a comparable Treasury. If you were looking at a corporate bond with five years of spread duration, for example, that might not seem especially long given how we tend to view standard effective (or interest-rate) duration figures. If the company hit a particularly rough patch during otherwise normal times and its yield spread over Treasuries widened by 300 basis points (or 3%), the bond would be expected to lose 15%. Spreads on the overall Bloomberg Barclays Capital Investment Grade Corporate Bond Index widened to more than 640 basis points in December 2008, though, from around only 100 in February 2007. Assume that kind of spread widening--that is, in the neighborhood of 550 basis points--and the bond would have lost nearly twice that.

Unfortunately, most fund companies don't report spread duration broadly or include it on their websites for individual investors. If you're anxious to know how much spread duration risk is lurking in your fund, though, it's worth calling to ask.

Use With Care Regardless of what you're measuring, though, it's important to be mindful that duration, like almost any other measure, is not a panacea.

1. Duration is an estimate. It's designed to help you gauge how much a bond's price is likely to rise or fall given a sudden change in market yields. The last part is important because the number gets less and less meaningful the longer the time between yield measurements. Even under relatively "normal" conditions, though, a bond's price may not move exactly as its duration predicts.
2. Portfolio durations are just average. The duration of a fund portfolio consists of the weighted average of the durations of its underlying bonds. Those bonds usually have very different weights and maturities, so the gain or loss predicted by duration would be most accurate if market yields changed by exactly the same amount for every bond along the maturity spectrum. Such a situation is rare, though. In fact, it's possible for some rates to rise while others fall (a yield-curve twist), a topic my former colleague Cara Esser discussed in an earlier piece. In that scenario, two different funds with the same average duration could perform very differently.
3. Even securities with the same numerical durations can perform differently from one another. We've already seen that in the corporate bond example above. But there are factors that can affect other bonds, too, especially those with embedded options, such as government-agency mortgages. The more complex a bond's structure--in particular, the presence of very flexible terms such as those giving a homeowner the option to refinance at any time--the more estimation and modeling is involved in arriving at a duration metric. In such cases, it's quite possible that two asset managers would assign different duration estimates for the same mortgage bond. Fortunately, we don't usually see massive dispersion across portfolios as a result of that phenomenon, but it's important to understand that there are usually modeling assumptions--and not just strict formulas--behind duration calculations.
4. Read the fine print. To understand your bond fund better, dig into its most recent portfolio report and try to figure out what it holds (or ask your financial advisor to do it for you). If you find things that you don't recognize, such as interest-only mortgage CMOs or inverse-IOs, look for a clear explanation in your fund's reports about how and why the fund is using them, or pick up the phone and call your fund company. We see them pop up in small amounts from time to time. But given how volatile those complex securities can be if short-term rates and/or prepayments shift notably, you'll want to make sure you're comfortable with your fund's reasoning and risk controls, especially if they appear in concentrations of more than a few percentage points. Any number of other sectors can produce risks not covered by standard duration figures. If your fund has significant allocations to high-yield or emerging-markets debt, for example, it likely carries meaningful spread duration relative to the movement of those markets.

Conclusion In short, duration--or interest-rate duration--can be a useful indicator for estimating the interest-rate risk of bonds that closely track the Treasury market. Any time you deviate from owning Treasuries, you usually get paid more yield and (perhaps) get a better longer-term total return. But you assume more and different risks, and tracking them all becomes more complicated. It therefore makes sense to use duration with caution and to make the effort to know what's in a fund's portfolio, rather than relying on duration or any other single number or grade when evaluating its suitability for your portfolio.