Pension plan sponsors have broadly embraced the notion that they should take less risk relative to liabilities as they seek to reduce the financial implications of funded status volatility. In practice, many plans are accomplishing this through a gradual shift away from return-seeking strategies that rely heavily on exposure to equity risk, and toward strategies that use long duration fixed income as a direct hedge to the liability. Many plans are following a “glide path” strategy that links improvements in the funded status of the plan to specific shifts in the asset allocation, with an ultimate target of 80%, 90% or even 100% fixed income.
If not now, then someday soon, many plan sponsors may need to address an important question: how to design a fixed income investment program that can successfully manage to keep a plan fully funded as it gradually winds down. They may find it is not quite as simple as putting all of the assets into duration-matched bonds.
Identifying the Characteristics of a Successful “End-State” LDI Strategy
In a traditional pension asset allocation, the dominant source of risk between assets and liabilities is the large equity allocation, which tends to be both highly volatile and largely uncorrelated with the liability. As a plan’s asset allocation becomes increasingly tilted toward bonds, however, it becomes ever more important to limit the mismatch between the risk characteristics of the fixed income portfolio and the risk characteristics of the liability. For a plan that is fully invested in bonds, this mismatch between the fixed income portfolio and the liability typically represents nearly all of the funded status risk, with the exception of some actuarial risk related to mortality and participant behavior that cannot be hedged in the market.
Importantly, we believe there are two dimensions to effectively matching the bond portfolio to the liability:
- First, the bond portfolio should have the same key risk factors as the liability so that a change in the mark-to-market value of the liability should be fully offset by a change in the mark-to-market value of the fixed income assets. This will ensure that the plan’s funded status does not change as market conditions fluctuate.
- Second, the assets should provide a return that is at least as high – if not slightly higher – than the discount rate on the liability, to ensure that the plan assets keep pace with the value of the liability even if market conditions are stable.
Matching Risk Factors
In our view, the first of these goals is relatively straightforward for most typical pension liabilities. The fixed income manager needs to construct a portfolio that matches the liability across four key risk factors:
- Duration – sensitivity to changes in interest rates
- Credit Spread Duration – sensitivity to changes in corporate credit spreads
- Curve Exposure – sensitivity to changes in the shape of the yield curve
- Convexity – sensitivity of duration to large changes in yields
Using a mix of liquid fixed income market sectors including (but not limited to) corporate bonds, Treasuries and interest rate derivatives, we believe it is possible to identify a portfolio benchmark that is a close fit for these risk factors. In practice it may not be possible to precisely match all four risk factors simultaneously, so it is worth prioritizing them to some extent. The first two factors – duration and credit spread – tend to be the most significant in terms of their ability to drive changes in the value of the liability and thus we believe should be the priority in any liability-driven investing (LDI) strategy.
Figure 1 below shows a simple example of a customized blend of market indexes that in aggregate may offer a close fit to the risk characteristics of a typical pension liability.
The relatively simple blend of three benchmarks – one intermediate, one long and one very long – can provide a set of factor exposures that closely matches the pension liability and helps ensure that mark-to-market changes in value are closely aligned.
Matching Returns
The second goal – matching or exceeding the “yield” implied by the liability discount rate – can actually be surprisingly difficult depending on the characteristics of the liability and the shape of the yield curve. Broadly speaking, a fixed income benchmark that is closely matched to the risk factors of the liability cash flows (as in the example above) can often fail to provide the yield necessary to keep pace with the liability over time.
Returning to the same hypothetical example, compare the yield on the liability (i.e., the discount rate) with the yield the assets (i.e., the current yield to worst for the blended benchmark).
Plan sponsors should be understandably concerned that the overall yield on the assets is lower than the rate of return on the liability. In a steady-state world, the plan would underperform the liability and funded status would decline, resulting in the failure of the primary goal of the “end-state” strategy – keeping the plan funded and ultimately paying off the benefit obligations.
Why Doesn’t the Yield Measure Up?
As mentioned above, the characteristics of the liability and the shape of the yield curve tend to be the drivers of the yield mismatch. Generally, the longer the liability and the steeper the yield curve, the greater the degree to which the yield on the liability can be expected to exceed that of an appropriate risk matching benchmark. A more detailed explanation requires a look under the hood of the liability discounting mechanism itself.
Liability discounting begins with the benefit cash flows, estimated by an actuary out to 75 years or more. Calculating a present value requires a discount curve comprising high quality corporate bond yields at each maturity point. If the portfolio were able to invest the present value of each cash flow in bonds yielding exactly the same as that equivalent discount rate, the liabilities would be perfectly matched.
The problem is that the investable universe of corporate bonds does not match the typical distribution of liability cash flows, and effectively disappears after 30 years. The discounting methodology seeks to “solve” this problem by simply assuming an extension of the yield curve out to the end of the liability cash flow stream. In practice, the yield curves commonly used for liability discounting either extend the 30-year yield indefinitely (the “flat-line” approach) or extrapolate the trend of the yield curve beyond the 30-year point (allowing for a continued increase in rates) – see Figure 3.
The end result is that a meaningful portion of the liability may end up being discounted at yields that are at least as high, and often higher, than the highest investable yield available in the corporate bond market. This is true even if one were to assume the pension portfolio had the option of holding zero coupon corporate bonds out to the 30-year maturity point. In practice, an investor may have to choose from among a set of options – including more diversified long duration corporate benchmarks and/or zero coupon Treasury bonds – that provide investable yields well below that of the discount rates applied to long liability cash flows.
Observing the Yield Differential in Practice
It is beyond the scope of this paper to delve into the details of the process by which the liability discount curves are constructed. To the extent that these yield curves tend to be biased toward higher yields than are actually available in the market, there is a practical challenge to be addressed. First, it is necessary to assess the size of this yield gap.
The right comparison to make in our view is between the yield available in the corporate bond market across the various maturity ranges and the yield on the Pension Protection Act (PPA) curve for an equivalent duration. The PPA curve is composed of spot rates, which are the functional equivalent to a zero coupon rate at each point on the curve. For this reason, we believe it is inappropriate to compare the 30-year PPA yield with the yield on a 30-year maturity corporate bond – because the bond’s duration may be only 15 years or so. It is more appropriate to compare the yield on the 30-year bond to the 15-year PPA rate.
This comparison serves to highlight the problem: While the yields on the PPA curve tend to be relatively well aligned with the market yields for the range of durations at which market yields exist, there is almost no meaningful supply of investable corporate debt with a maturity beyond 30 years and a duration beyond 15 years. (And in practical terms the problem is worse – the duration of the overall long corporate index is below 13 years.)
Figure 4 below sets out the market yields for investment grade corporate bonds in five-year maturity buckets (showing both the full investment grade universe and the PPA-equivalent AAA-A segment). It then compares those investable yields to the PPA yields at the same durations, and in comparison to longer duration PPA yields that are effectively non-investable.
In our experience, approximately 15%–25% of the liability present value can result from longer-dated cash flows that are discounted at yields that are effectively non-investable. In the example above, the incremental yield applied to this segment of the liability can exceed market yields by as much as 100 basis points.
How Should a Plan Respond?
While the end-state strategy poses complex challenges, a glide path approach to pension investing can still offer significant improvement in risk management. Given the relatively limited size of the yield shortfall, we believe there are a number of possible strategic responses:
A. Target a level of overfunding in the plan that allows a fully risk-factor-hedged portfolio to offset the yield gap.
B. Allow active management of the fixed income portfolio to target excess return sufficient to make up the gap.
C. Sacrifice some elements of risk matching (such as convexity and curve risk matching) to allow the portfolio to concentrate in higher yielding assets.
D. Allow the credit portion of the portfolio to move down in credit quality (and up in yield) to a degree sufficient to make up the gap.
E. Maintain some degree of exposure to higher risk/return assets, such as equity, so that over time performance can potentially stay ahead of liabilities.
F. Allow for the use of leverage.
G. Implement some combination of the above.
Option A simply accepts the state of affairs and relies on the over-capitalization of the plan to cover the difference. This is probably the lowest risk strategy but also potentially the costliest, since it would require a plan to reach levels of funding over and above that required by current law. The plan can “spend down” this overfunding gradually over time as the liability comes due. For plans that are already meaningfully overfunded, this may be the best strategy.
The use of active management (Option B) may make sense, not only because of the potential for higher returns but also because the alternative – passive investing – does not offer the flexibility to actively manage portfolio risk such as potentially mitigating losses associated with downgrades out of the benchmark (aka “fallen angels”) and “jump-to-default” (i.e., higher-rated securities defaulting before they are downgraded) from within the portfolio. One key potential benefit of active management is reducing the portfolio’s exposure to downgrades and defaults through the credit selection process. We believe the possible increase in tracking error relative to a passive index strategy should be modest compared with the potential for improved performance.
Shifting away from the optimal risk hedging portfolio to allow for more yield (Option C) is also feasible – for example, by investing exclusively in a long credit index and eschewing the shorter duration and STRIPS portfolios that allow for a more precisely tailored liability hedge. Fortunately, the degree of risk mismatch necessary is probably modest from a plan sponsor’s point of view, although the true cost of the mismatch would not be known until the portfolio experiences significant market volatility. If the volatility was driven by profound changes in the shape of the yield curve or the term structure of credit spreads, the resulting change in funded status could be meaningful.
Reducing credit quality in order to increase the yield on the portfolio (Option D) can make sense within reasonable constraints. Allowing the benchmark to include the full spectrum of investment grade credit (including BBB-rated bonds) may be appropriate in some cases, given the modest historical increase in default/downgrade risk and the increase in yield that they tend to provide. The broader benchmark also provides a more complete set of sector exposures for an active manager to use in the portfolio construction process. Further, we have found strong evidence to suggest that the broader investment grade credit universe does as good a job (and possibly a better job) of tracking liability performance than a more constrained benchmark limited to only A-AAA rated securities.
Further reduction in credit quality to include high yield bonds should be taken with caution, given the meaningful historical increase in default risk. That said, a modest shift into high yield could go a long way toward fixing the yield gap and may be a superior alternative to retaining an equity allocation (Option E) – given the ability of a high yield allocation to contribute to the hedging of key risk factors such as duration and credit sensitivity, along with the prospect of mitigating some of the default risk through active management.
Conversely, the rationale for holding equities, which tend to deliver return primarily through price appreciation (not dividends) and typically have high levels of volatility, for the purpose of offsetting a yield differential, is limited, in our view – particularly if period-over-period downside risk is meaningful to the sponsor.
Finally, a plan may be able to employ leverage (Option F) in the form of a derivative overlay to deliver a hedge that incorporates both long duration and a high credit content. In practice this would likely entail the use of a cash bond portfolio that is heavily weighted to investment grade credit and an overlay that employs interest rate swaps or Treasury futures to achieve a desired duration and curve exposure. Also possible would be the inclusion of additional credit exposure via the use of credit default swaps.
Conclusion
Looking ahead to the end of the glide path, we believe that a 100% funded plan being able to invest in a risk-matched bond portfolio and simply “walk away” is somewhat unrealistic. Perhaps more likely is that plans will stop somewhere short of 100% LDI fixed income and rely on a combination of manager alpha and a modest degree of beta diversification in their effort to keep the plan assets ahead of the liabilities over time, while accepting that – in the real world at least – the perfect hedge will always be just out of reach.
Appendix: Risk Factors
With respect to the credit risk embedded in the liability discount rate, it is important to adjust for the high quality of the bonds used in the discount curve. PIMCO uses a more nuanced measure of credit sensitivity known as “beta-adjusted credit spread duration,” which captures the reality that higher quality bonds tend to exhibit less spread volatility (i.e., lower beta) than the overall credit market, and lower quality bonds tend to exhibit greater spread volatility (i.e., higher beta) than the market as a whole. Because pension liabilities are discounted using only very high quality bonds, estimates of the credit sensitivity of pension liabilities must be adjusted downwards.
There are a number of ways to assess the risk associated with changes to the shape of the yield curve, with the most comprehensive being a granular breakdown of risk into “buckets” across the yield curve, known as key rate duration analysis. For the purpose of simplicity, we are using an alternative approach that shows the sensitivity of a liability or portfolio to a steepening of the yield curve.