Unusually flat

Most market commentators talk about the U.S. Treasury yield-curve steepness in the context of the short end: T-bill yield versus the 2-year Treasury yield. But given that U.S. pension plans have approximately 80% of their liability risk associated with yields beyond 10 years, a more important metric is the 10-year Treasury yield versus the 30-year Treasury yield (below).

Presently, this measure of yield-curve steepness is near zero, which is unusual. We can deduce that it:

  • Generally ranges between zero and 125 bps
  • Rarely falls below zero
  • Tends to exhibit mean reversion characteristics over long cycles that are roughly tied to the U.S. economy

Why is this of interest to U.S. companies that use liability-driven investing (LDI) strategies?

For companies with pension plans, the concept driving LDI investing strategies is to focus on matching assets to their pension plan’s liabilities in order to help secure the benefits and control the impact of the plan on the company’s financial health over time.

For many pension plans, the LDI portfolio is built by creating a custom LDI benchmark against the plan’s liabilities using a “slice” approach. As an example, if a plan has the same value of assets as liabilities and 50% of its portfolio is in LDI strategies, the plan might choose to target “hedging” a 50% slice of its liability. This approach provides a simple structure for measuring LDI performance: The “return” of the LDI portfolio can be compared with the “return” of the liability, thereby indicating the LDI strategy’s effectiveness.

But this approach has a hidden consequence. It might be clear that the LDI portfolio is not hedging the full liability (in the example above, 50% was hedged), and so the pension plan is positioned for an increase in Treasury yields. What might be less appreciated is that the LDI slice approach also positions the pension plan in favor of a Treasury yield curve steepener—in other words, it’s positioned to benefit if long yields increase more than shorter yields, as shown in the table below.

ScenarioImpact vs. 50% slice of liabilitiesImpact vs. all liabilities
Parallel shift
1% rise in yields across the curve
0.00%5.90%
Steeper curve
Yields shorter than 10 years fall 50 bps & yields over 10 years rise 50 bps
0.55%2.00%
Sources: Bloomberg, Allspring, and the ICE BofA Mature U.S. Pension Plan AAA-A Corporate Curve Discounted Index as of 28-Feb-23; best-fit LDI benchmark calculated by Allspring using combination of Bloomberg indexes for intermediate and long Treasuries, Separate Trading of Registered Interest and Principle Securities (STRIPS), and credit.

We observe:

  • Parallel shift: The results are as expected—the hedge is seen to work well against the slice of liabilities it was scaled to hedge, and the impact versus all liabilities is as expected given the 50% level.
  • Steeper curve:
    • The result for the 50% slice of liabilities reflects a fairly close match. We do, though, see a small positive effect, which is typical given that liabilities have a different key rate profile than a straightforward LDI benchmark (due to the presence of coupons and the composition of the corporate bond market, among other factors).
    • The result versus all liabilities clearly shows the pension plan’s exposure to yield-curve shape due to the partial hedge and term structure of the liabilities.

Our view on this is that best practice is to separate the decision about overall duration from change in shape of the yield curve and have the two managed explicitly and directly within the LDI portfolio. This isn’t completely straightforward, and there’s some tension between the two. But—especially if interest-rate derivatives like Treasury futures are used—exposure to the level of the Treasury curve and to its shape can be controlled separately. One casualty of this would be the simplicity of measuring the LDI portfolio against a slice of the liability because more complicated performance reporting would be needed. However, many LDI investors have developed total plan risk and attribution frameworks, and many consultants have systems in place to support this approach.

Where does this leave pension plans and their LDI strategies?

As discussed, most U.S. corporate pension plans have a degree of yield-curve steepener built into their LDI strategies. As we’ve shown, adopting a long-end Treasury yield-curve steepener may make sense in the current market environment.  But as we’ve also noted, this risk position isn’t always going to be desired or supported by market conditions.

Therefore, we suggest pension plans consider this issue so that:

  • It’s recognized that yield-curve steepness is a design feature that can be incorporated into their LDI strategy going forward.
  • Tools and measurement frameworks are put in place to enable the plan to be managed in the future.
  • When the time is deemed right, the LDI benchmark is changed to target both outright duration and curve steepness to enable more deliberate interest rate management.

Summary

  1. Corporate pension plans tend to be positioned in favor of steepening of the Treasury yield curve. Was this a deliberate decision? Is it managed over time?
  2. Current market conditions show the long Treasury yield curve is unusually flat and hence might be expected to steepen. Therefore, pension plan positioning currently appears sensible.
  3. We believe best practice is to separate the decision about overall duration from change in shape of the yield curve so that both aspects are explicitly considered.
  4. The tools for managing these aspects within an LDI strategy aren’t complex, but in our view they do need to be built into a custom mandate with an LDI manager responsible for monitoring, managing, and reporting.

CFA® and Chartered Financial Analyst® are trademarks owned by CFA Institute.

A STRIP bond is a debt obligation whose principal and each interest payment are removed (stripped) and become separate securities that are sold to investors.

Leave a Reply

Your email address will not be published. Required fields are marked *

You might also like: