Summary

• Public pension plans tend to focus disproportionately on investment performance whilepaying only cursory attention to liability performance. We believe public pensionliabilities can, and deserve to, play a more central role in portfolio construction andholistic performance measurement.
  
  
• Assets and liabilities are inextricably linked through the expected return on assets(EROA) discounting mechanism. Realized investment returns drive plan assets, but alsoplan liabilities as a consequence of their impact on asset valuations and thus forward-looking expected returns.
  
  
• This paper introduces Liability Aware Investing (LAI) as a broad portfolio constructionand risk management framework that explicitly incorporates this linkage between planassets and liabilities and funded status volatility as a holistic measurement of totalplan risk that quantifies the expected tracking error between assets and liabilities.
  
  
• LAI is NOT corporate pension liability-driven investing (LDI) applied to public plans.Rather, LDI can be conceptually understood as a narrow application of the broaderLAI framework.
  
  
• Asset class risk characteristics are altered when shifting from a traditional asset-onlyframework to an LAI framework. For example, some like public equity become moreattractive while others like cash become less attractive. LAI leads to moderate assetallocation changes relative to traditional asset-only mean-variance optimizationtechniques. In this sense, it can augment, rather than replace, current industry practices.
  
  
• LAI may justify taking on higher levels of asset volatility than a plan would otherwiseaccept before considering liability impacts, and thus support higher allocations toreturn-seeking assets. For example, the funded status volatility of public equity islower than its asset volatility.
  
  
• LAI can be a helpful additional tool for communicating total plan risk and performanceto investment committees and other plan stakeholders.
  
  
• This piece presents the groundwork and initial conceptual framework for LAI includingbroad principles and risk analytics. In practice, we expect adoption and implementationmethods across the public plan universe will vary based on unique plan characteristics,governance and beliefs as well as over time.
  
  

Introduction

Pension liability and keeping your eye on the ball

On May 26, 1959, Harvey Haddix pitched 12 perfect innings for the Pittsburgh Pirates against the defending National League Champion Milwaukee Braves, retiring the first 36 batters on 115 pitches—82 of them for strikes. While many consider the game to be the best pitching performance in Major League Baseball history, the Pirates actually ended up losing to the Braves 1-0. In a similar respect, public pension funds that focus solely on investment performance are only watching one-half of each inning. To truly gauge a winning performance, one must look up and occasionally check the scoreboard to see how the liabilities are performing. Admittedly, this hasn’t exactly been easy. Asset performance can be measured and refreshed in real time, while liability valuations are often perpetually stale, updated at most on an annual basis, and often conspicuously absent from asset/liability modeling (ALM) studies. However, we believe that liabilities can, and deserve to, play a more central role in the strategic asset allocation and holistic performance measurement process.

Over the years, and as corporate pension plans have broadly reached a consensus on liability-driven investing (LDI), many academics and practitioners have explored how LDI could be transposed into the public pension world, without a wholesale importation of the current framework, which discounts liabilities on corporate bond yields and leads to solutions heavily concentrated in high-quality public long duration fixed income. We believe these proposed methods (for example, cash flow matching) certainly have their merits, but may not be widely applicable or useful to all plans (for example, a plan open to new participants) or rather may serve as a constituent of a broader liability-focused framework.

Liabilities are inextricably linked to the asset portfolio through the discounting mechanism used to present value future benefit payments, the expected return on assets assumption (EROA), which itself is informed by actual market returns. In fact, traditional corporate pension LDI can be understood as a narrowly defined application of this framework where the asset portfolio is limited to a (hypothetical) liability-immunizing portfolio of high-quality fixed income, rather than the (actual) total portfolio itself. We propose that portfolio construction and risk management explicitly incorporate this dynamic by modeling how the EROA assumption, and in turn liability valuation and volatility, will respond to realized market returns. Short-term negative asset returns will generally result in an increase in long-term expected returns, increasing the liability discount rate and reducing the value of the pension liability. The inverse chain of events occurs, as illustrated in Exhibit 1, in reaction to short-term positive asset returns. This positive correlation between asset returns and liability returns can be harnessed to proactively manage the volatility of plan funded status. Again, if you take the preceding description and replace the broader phrase “asset returns” with the more targeted phrase “high-quality fixed income returns,” you have in effect described corporate pension LDI.

Harnessing the positive correlation between asset and liability returns

Exhibit 1: Pension schematic: Relationship between asset returns and liability values

Source: J.P. Morgan Asset Management. For illustrative purposes only. 

As it roughly stands now, a plan assuming a 7.0% expected return will construct a portfolio that meets the target, including alpha, while attempting to minimize portfolio volatility or some other measure of asset drawdown risk. But what if we considered how funded status might evolve by measuring this association between assets and liabilities? It doesn’t change the need for a 7.0% return, or the near-term employer contribution rates, but we find it leads to reasonably different portfolio solutions and may facilitate improved long-term economic decision making and pension health. In the remainder of the paper we will refer to this concept of asset/liability linkage as Liability Aware Investing (LAI) and the holistic measurement of total plan risk as funded status volatility, which quantifies the expected tracking error between plan assets and liabilities.

In order to be constructive and additive to current practice, we believe public pension LAI should prudently balance practical implementation with theoretically sound underpinnings. To that end, the utilization of funded status volatility for portfolio construction and risk measurement:

LAI for Public Pension Plans

What makes the LAI framework so different from existing frameworks?

In its application, it’s not radically different, which makes it all the more accessible. The main differentiator is the explicit measurement of funded status volatility, accounting for the liability by modeling its volatility and correlation to other exposures vis-à-vis the asset allocation. The liability is market-based in the sense that asset allocation and market performance drive forward-looking expected returns, which are then used to calculate the expected return assumptions and liability discount rate (Exhibit 2). To be more precise, we are focused on the relationship between an asset class’s realized returns and the subsequent change in expected return assumption. For example, during the one-year period ending September 30, 2020², global equity returned +10.4% and JPMorgan’s 2021 LTCMA dropped the asset class assumption by 140bps, from 6.50% to 5.10% (Exhibit 3). The valuation impact reduced the assumption by around 225bps, but was offset by changes in other return building blocks, mainly an expectation of margins improvement. While asset class return expectations are driven by both cyclical and structural factors, the vast majority of year-over-year changes are a consequence of valuation adjustments. This insight is a foundational principle of the LAI framework.

Asset allocation and asset class returns impact both assets and liabilities

Exhibit 2: Pension schematic: Drivers of pension assets and liabilities

Source: J.P. Morgan Asset Management. For illustrative purposes only. 

Relationship between realized returns and return expectation changes

Exhibit 3: JPMorgan’s long-term capital market assumptions 

Source: J.P. Morgan Asset Management. For illustrative purposes only. 

A simple example: Assumption stability ≠ funding stability

In order to demonstrate the concept of funded status volatility for public pensions, we provide a simplified, albeit unrealistic, example (Exhibit 4). A fully funded plan experiences a 12% drawdown in the portfolio, reflected in the reduced asset value. This market pullback also improves the valuation component of the portfolio’s forward-looking expected return—the assets it holds are now more attractive than just prior to the drawdown.³ Assuming a nominal liability (we will use nominal liabilities, in contrast to real, in the remaining body of the paper but address inflation linkage in the appendix; see call out box) with a PV1 of 12 (present value of 1% change in EROA; see call out box) and a 1.0% increase in the expected return, the liability value declines by an equivalent amount and full funding is maintained. In theory, if this relationship were to hold across the set of all possible asset returns, we could say that this plan is immunized from an LAI perspective.

While capital market assumptions from asset managers and consultants tend to fluctuate at least annually, most public plan systems keep their assumptions relatively constant, a sort of lagged declining step function, in the name of “stability.”

Immunization occurs when asset changes are completely offset by liability changes

Exhibit 4: Stylized example of public plan LAI

Source: J.P. Morgan Asset Management. Calculations assume a nominal Liability PV1 of 12. For illustrative purposes only. 

This example shows that stability in the discount rate might in fact drive instability in the funding level and, in turn, volatility in the actuarially determined employer contribution (ADEC) rate. In some cases, more frequent and market-driven liability discounts facilitate reduced funded status volatility. For the plans that use a five-year or longer smoothing period for their actuarial value of assets (AVA), this more “mark-to-market” approach to assets and liabilities may display higher funded status volatility. However, this pattern would be expected to reverse for lower risk portfolios and for GASB disclosures where the plan assets (fiduciary net position) are based on market value.

Another way to understand LAI mechanics is as a countercyclical funding dampening effect. The expected return is reduced, a costly action in isolation, after favorable returns in the asset portfolio. In contrast, after large asset drawdowns, a corresponding increase to expected returns provides cost relief to plan sponsors who otherwise might need to dramatically boost contribution rates. As we will discuss later, a plan need not necessarily adopt a dynamic expected return assumption for this framework to be worthwhile as there are economic benefits from holding a higher expected return portfolio.

A simple example: Stocks, bonds & cash

In order to explore these characteristics, we start with a small subset of core asset classes—global equity, U.S. aggregate bonds and cash— and draw out insights about their role in an LAI framework. Exhibit 3 outlined the historical realized returns and subsequent changes to the expected return assumptions. Exhibit 5 summarizes both the asset and resulting LAI metrics for these asset classes as well as some basic portfolios.

Public equity funded status volatility is lower than asset volatility

Exhibit 5: Asset class characteristics (Nominal liability PV1 of 12)

Source: J.P. Morgan Asset Management. For illustrative purposes only. 

In the discussion below we introduce several novel LAI metrics that drive funded status volatility. A brief description is provided although we strongly recommend that readers explore the Appendix section titled “Is there specific analytical support for the LAI funded status volatility drivers” for a more in-depth explanation.

From these metrics we can make some inferences regarding the application of an LAI framework in comparison to a traditional asset-only framework:

a) Cash isn’t risk-free: The funded status volatility of cash is more than 10x the asset volatility. Cash has a low A/ER Sensitivity and positive A/ER Correlation, both of which increase risk within the LAI framework. Poor realized cash returns are unlikely to be counterbalanced by increasing return expectations. They will generally beget further poor returns, and vice versa.

b) Equities aren’t as risky: The funded status volatility of public equity is lower than the asset volatility. A/ER Sensitivity (21.7) and Correlation (-47) indicate that expected return changes will tend to counterbalance realized returns and do so to an extent that is meaningful.

c) Traditional de-risking has limited efficacy: In a simple, two-asset portfolio, substituting core bonds for public equity reduces funded status volatility but at a much diminished rate relative to asset volatility. For example, going from 70/30 to 30/70 stock/bond reduces asset volatility by 600bps but reduces funded status volatility only by 400bps. The substitutions are improving A/ER Correlation but significantly reducing the A/ER Sensitivity measures.

Expanding the opportunity set

Moving past our simple example of LAI, we now open the door to a broader set of asset classes, gauging their risk through both the traditional asset-only lens and the proposed LAI funded status volatility lens. Exhibit 6 plots the governing properties of funded status volatility: A/ER Sensitivity, Correlation and asset volatility (bubble size). Based on these characteristics, we would expect assets plotting in the upper left, where return assumption changes are more responsive to realized market returns (high A/ER Sensitivity) and their relationship is more negatively correlated (low A/ER Correlation), to be more attractive in LAI space than asset-only space. Similarly, assets plotting in the lower right quadrant are expected to be less attractive. Positive A/ER Correlation leads to high funded status volatility, particularly at high levels of asset volatility. Thus, we would expect assets in the top right to be relatively less attractive as well.

LAI characteristics explain changes in asset class risk profile

Exhibit 6: LAI asset class characteristics: A/ER sensitivity, correlation and asset volatility (bubble size)

Source: J.P. Morgan Asset Management. For illustrative purposes only.  

Exhibit 7 plots the asset volatility versus funded status volatility of each asset class and confirms these findings. Most asset classes plot fairly close to the parity line, again showing that higher levels of asset volatility are associated with higher levels of funded status volatility. Interestingly, we can see that long duration bonds exhibit low funded status volatility, reminiscent of the corporate pension LDI framework. This is partly because expected returns are closely related to the starting yield level. Thus yield changes will drive both actual returns and expected return changes in a predictable and proportional manner. The opposite is true for asset classes like cash and hedge funds, which in some sense can be thought of as cash-plus strategies and falter for the same reasons.

Exhibit 7: Changing the lens: Asset volatility vs. funded status volatility

Source: J.P. Morgan Asset Management. For illustrative purposes only.  

Bringing expected returns into the equation, Exhibit 8 calculates efficiency ratios and return per unit of risk for each asset in both the asset-only and LAI space, and many of the same takeaways emerge. Cash (not pictured), core bonds and hedge funds are significantly less efficient in the LAI space, while long treasury and core real estate show marked improvements under the transformation in perspective.

Core bonds and hedge funds are less efficient in LAI space, while long treasury and core real estate show marked improvements

Exhibit 8: asset class efficiency in asset-only versus LAI framework

Source: J.P. Morgan Asset Management. For illustrative purposes only.

Crucially, we have only considered stand-alone asset characteristics, and none of this is to say that those asset classes that look less promising in the LAI space or have unfavorable A/ER Correlation and Sensitivities should be ignored. Returns are still paramount, and at this stage we are simply reformulating how to think about risk. Long duration bonds may be attractive from a risk perspective, but their paltry return expectations will quickly limit their size in a portfolio that aims to earn 7.0% or higher in the current environment. In the following section, we look at the implications of the LAI framework for portfolio construction where both risk and return are considered simultaneously and contemplate practical takeaways.

From theory to practice: LAI implications for portfolio construction

Rather than survey asset class characteristics in a silo, we now consider how they might interact in a total portfolio context (Exhibit 9). We begin with an archetypal public plan portfolio (return expectations are “beta-only” and so fall short of a typical public plan expected return assumption), roughly approximating the average public plan allocation⁴, and formulate asset class substitutions that align with our exploration of LAI characteristics:

• Portfolio B—Replace hedge funds with real estate: Hedge funds and other “cash plus” strategies are less efficient in LAI than the asset-only space. Despite increasing asset volatility, this reallocation dampens funded status volatility by boosting the A/ER Sensitivity.

• Port C—Extend fixed income duration: Redirecting half the U.S. aggregate allocation to long duration bonds improves the portfolio’s A/ER Correlation and further reduces funded status volatility.

• Portfolio D—Rotate out of fixed income into public equity: An additional 2% shift into public equity from U.S. aggregate increases A/ER Sensitivity and reduces A/ER Correlation, boosting returns in exchange for a moderate increase in funded status volatility.

Increasing liability awareness

Exhibit 9: Portfolio reallocations to boost return and reduce funded status volatility at the same level of asset volatility

Source: J.P. Morgan Asset Management. Archetypal portfolio based on data from Public Plans Database. For illustrative purposes only. 

In aggregate, these allocation shifts reposition the portfolio to reduce funded status volatility at the same level of asset volatility while simultaneously increasing return expectations. Different sets of projected assumptions and application to a broader investment opportunity set will inevitably lead to different solutions, but in practice the LAI framework can facilitate a careful balance between funded status volatility, asset volatility and expected return.

In the appendix to this paper, we explore how other plan characteristics may influence portfolio construction through an LAI lens. These topics include low risk/return portfolios, varying the liability PV1 and investing against real, rather than nominal, liabilities.

Putting it all together: LAI takeaways

What does this all mean and how can plan sponsors benefit from an LAI framework? We believe that an LAI framework should augment current practice rather than entirely replace it. While the funded status volatility measure does redefine the “low risk” portfolio, the allocation implications are only moderate in the range of current return targets. This parallels to the corporate pension world, where even dyed in the wool LDI believers need to run what amounts to total return driven portfolios if their return needs are high enough. However, by integrating and socializing the concept of funded status volatility, they have built a framework for measuring risk, quantifying the impact of allocation and market outlook changes as well as a roadmap for how to reimagine the portfolio as the needs of the plan and asset class expected returns evolve. While the preceding analysis and formulation of A/ER Sensitivity and Correlation characteristics have been derived from historical data, capital market forecasters can make projections of these metrics directly, informed by their secular and structural market outlook. The LAI framework also gives credence to rebalancing while confronting market volatility. Not only are you “buying low” and “selling high,” but also curbing liability values on the opposite side of the plan balance sheet. The LAI framework gives us not a radical overhaul of current practice, but an additional tool to measure risk. If plan sponsors want to maximize their chances of beating their liabilities, they must keep track of the score.

Appendix

Exploration of special considerations for the interested reader

Despite its length, we have really only scratched the surface of LAI and its implications. In the following paragraphs, we quickly address some questions and concerns that we were asking of ourselves.