Pension funding has improved materially over the past decade, and for many plan sponsors that means that the “end-game” strategy that once seemed a distant objective is now within reach. But caution may be needed, as some key conditions which gave rise to pension de-risking strategies no longer apply. Before continuing to push capital into low returning and increasingly inefficient LDI programs, a careful review of alternative options is called for. Many sponsors may wish to re-think the wisdom of hibernation and termination when the benefits of a well-funded and stabilized pension are added to the mix.
Most pension plans across the U.S. are still following the same incremental de-risking plan they adopted in the years immediately after the global financial crisis. Specifically, replacing return-seeking assets such as equity with long-duration fixed income that seeks to match the liability. In re-evaluating these glide path strategies in the current environment, it appears that most plans have de-risked far enough to avoid serious risks and to continue would simply increase costs and inefficiency with no discernable benefit, says Jared Gross, Managing Director and Head of Institutional Portfolio Strategy for JP Morgan Asset Management. II recently asked Gross to describe the course correction he’s urging pensions to make before they face the consequences of an outdated investment approach.
You have had a ringside seat at the evolution of pension strategy over the past 20 years. How far has it come since then?
Jared Gross: Prior to early 2000s, plan sponsors were focused on generating high returns with a traditionally diversified asset allocation – mainly stocks and core bonds. In many respects, this approach was appropriate given the rules in place at the time. Accounting and regulatory standards were mild, with only a limited pass-through from funding changes to required contributions or the financial statements. Pension investors generally didn’t consider liabilities when allocating assets. After all, a long period of strong returns had left most plans in a surplus position.
This era ended with the tech bust in the early 2000s, and the decline in interest rates that rapidly followed. The fallout led to a massive downturn in the funding of defined benefit plans; most went from being overfunded before 2000 to significantly underfunded just a year or two later. That was the first wake-up call. No one really had to confront the asset-liability mismatch in pensions before. Not too many years later, the global financial crisis delivered another blow, and sponsors began responding by starting to de-risk their asset allocations, closing and freezing defined benefit plans, and moving employees to defined contribution plans that offered the promise of more predictability.
How have you seen that de-risking process evolve?
I think we have seen two broad phases of pension de-risking with liability driven investments, and I believe we are about to enter a third.
The first phase, which I refer to as LDI 1.0, involved moving from shorter to longer duration bonds during the period immediately following the global financial crisis. With LDI portfolios offering high levels of yield, this was an unusual opportunity to reduce risk while also increasing return. Within a few years, most plan sponsors had moved their fixed income to a long-duration benchmark, often starting with the Long Government Credit Index.
Since then, we have seen a long process of patient de-risking that I refer to as LDI 2.0 or the “Glidepath Era,” in which plans have just been waiting for incremental improvements in funding before shifting capital from return-seeking strategies to LDI strategies. They’ve developed increasingly customized hedging strategies, many of which involve custom fixed income benchmarks, derivative overlays, and completion management. While the complexity of these hedging strategies has increased, their diversification has not. Most plans have reached a point where the exposure to investment grade corporate credit is the largest risk in the asset allocation.
Today, however, we need to reconsider the glide path approach for two reasons. First, higher funding levels and lower volatility asset allocations have reduced pension risk to a very low level already. And second, the rules that govern mandatory plan contributions have been dialed back to the point that contribution risk is minimal, if not nonexistent. It is very hard to see the value in further concentration in long-duration bonds.
If you don’t see plans continuing to de-risk, then what actions do you see them taking?
Given this backdrop, plan sponsors need a course correction. They’ve already done an enormous amount of de-risking – consider that volatility is dramatically reduced even in plans that are only 50% hedged. Pushing further down the glide path rapidly becomes inefficient. That “last mile” is very expensive; giving up significant returns to reduce volatility only slightly makes it an extremely costly form of hedging. If the risk of large mandatory contributions was a serious concern, this behavior might be justified, but it is not.
This brings us to an inflection point in pension strategy. What I think of as LDI 3.0, or as we describe it to our clients, “Pension Stabilization”, is a more balanced approach than traditional LDI. It aims for a long-term blend of modest excess return and low volatility – but not the lowest possible volatility. It’s definitely not a journey back to 1999 when plans were seeking high returns and ran 12-15% annualized funding volatility. A stabilization strategy begins when a plan approaches full funding and adopts a sensibly de-risked and diversified asset allocation, allowing for the gradual improvement in funding across time while preserving numerous positive financial benefits for the sponsor. I expect that many sponsors will see the appeal of taking this “off-ramp” from an outdated glide path.
Why is it so difficult to completely hedge the risks in a traditional pension plan?
Simply put, it’s impossible to perfectly hedge pension liabilities with financial assets. For starters, you can’t hedge the actuarial risks around participant behavior and people living longer. The sponsor must bear these costs through increased returns or contributions.
Additionally, the fixed income portfolios that are used to hedge liabilities will experience downgrades and defaults. The liabilities won’t change, but there will be performance drag across time. A more diversified stabilization strategy can still out-earn the liabilities over the long run and that drag from credit losses will just be noise. But when assets are concentrated in fixed income, the drag becomes more difficult to outrun. This is the critical flaw in most hibernation strategies. An LDI portfolio comprised of long-duration corporate bonds and Treasuries will slowly burn through its excess capital.
How might pensions approach additional de-risking now?
We really need to get away from the notion that, as the hedge portfolio grows in size it should become more and more concentrated in corporate credit. The exact opposite is true. Hedge portfolios need to become more diversified. Treasuries certainly have a role to play, though their lack of yield is a concern. Long-duration, securitized fixed income and high yield bonds are also useful options, but the list doesn’t end there. It’s instructive to observe the investing behavior of life insurance companies, who are managing the risks of annuity portfolios that are quite similar to pension liabilities; these firms are large holders of securitized assets, crossover portfolios that incorporate high yield bonds, structured credit and mezzanine debt.
The other path to de-risking begins in the return-seeking portfolio. Investors today have access to a far more diverse set of low-to-moderate volatility alternative asset classes that can deliver stable excess performance versus liabilities without the high volatility and low correlation of equities. Many of these asset classes are also capable of delivering a high level of distributable income, which is a valuable resource for a benefit-paying institution.
Defined benefit plans are continuing to decrease as defined contribution plans become more common. Do you see a future for DB plans?
I hope so. The historical volatility of defined benefit pensions and the uncertainty surrounding contributions convinced many sponsors that they were too risky to maintain. This is simply untrue. A well-funded and prudently managed DB plan has powerful benefits that a DC plan simply cannot replicate. When we consider the recovery in DB plan funding and the evolution of pension asset allocation over the past 20 years, there should be a conversation in corporate America about preserving what remains of the DB system. While I am not convinced we will see closed plans reopen or new plans created, it’s worth considering. In a period of rising wages and highly competitive labor markets, the original purpose of defined benefit plans comes back into focus: proving a tax-advantaged, investment-driven vehicle for delivering competitive benefits to employees.
What is the next evolution in pension plan strategy?
I would love to see pension stabilization evolve to become the standard model. I’m encouraged because I do see many well-funded plans adopting elements of this approach. Consultants are getting behind this trend as well, although some continue to view the pieces independently rather than as a holistic framework that can challenge the glide path/hibernation/termination model.
I think we may be close to the high-water mark for traditional LDI strategies. With current interest rates and credit spreads, managers know that a dollar put into a traditional LDI strategy today is almost never going to outperform liabilities going forward. Conditions may change, but from where we stand today, it’s hard to envision those LDI frameworks remaining static. There are just too many better ways develop a risk-efficient pension strategy outside of them.
What do you see as the advantage for plan sponsors in the way JP Morgan collaborates with them on pension stabilization?
JP Morgan has a 150-year history of serving as a fiduciary to our clients; I can’t overstate the importance of that legacy to the present-day JP Morgan Asset Management. We also have the broadest cross-section of investment strategies of any manager in the business, each of which is built around the premise that highly skilled, independent investment teams can deliver specialized strategies while drawing on the collective resources of the broader organization. Unlike some firms that implictly guide clients to the particular strategy in which they specialize, we want clients to make the best possible strategic decision because we are confident that we will find a way to partner with them in implementing it.
With respect to pension stabilization, the merits of the approach speak for themselves. As clients look to implement this type of program, they will need to reconsider their exposures to traditional hedging and return-generating assets as well as new forms of diversifiers. For hedge portfolios, we’ve been at the forefront of long-duration securitized investing and have a long history of success in high yield and mezzanine debt. Within the realm of alternative diversifiers, we tend to highlight the powerful income generation and diversification benefits of core real assets – with a very strong platform across real estate, infrastructure, and transportation.
JP Morgan also has a long history of working with institutional clients in discretionary management and delegated solutions. Pension plans may not have the internal resources and operational flexibility to execute a sophisticated asset allocation that takes advantage of the full investment opportunity set. A number of sponsors are currently using our delegated platform and its full spectrum of assets to outperform liabilities in a risk-aware manner, and we think this business model will continue to grow.
The current path to hibernation and termination is a dead end. Specialized LDI managers aren’t going to give plan sponsors the unbiased advice that they need. Diverse, forward-thinking managers and consultants can help redirect clients to change course with full-spectrum strategies that maintain positive returns while still managing volatility. We want plans to get fully funded and stay fully funded. Pension stabilization can achieve both and help ensure their pensions are once again able to be a source of value.
In the interview above, JP Morgan’s Jared Gross questions whether pension funds still need to aggressively de-risk given that funded status has generally improved since LDI strategies first became a go-to for such plans. And, indeed, despite the Covid pandemic, funded status is looking better these days.
According to “The State Funding Pension Funding Gap: Plans Have Stabilized in Wake of Pandemic,” a whitepaper by The Pew Charitable Trusts, the gap between the cost of pension benefits that U.S. states have promised their workers and what they have set aside to pay for them dropped in 2021 to its lowest level in more than a decade.
Pew estimates that state retirement systems are now over 80% funded for the first time since 2008. Such progress would be significant in any year, but the improvement in fiscal 2021 occurred during a recession in which many analysts predicted that revenue losses related to the COVID-19 pandemic would increase retirement fund shortfalls. Instead, Pew found an increase in assets of over $500 billion in state retirement plans, fueled by market investment returns of more than 25 percent in fiscal 2021 (the highest annual returns for public funds in over 30 years) and substantial increases in contributions from taxpayers and public employees to pension funds.
Upward trend for U.S. states
These contribution increases, which came after years of states shortchanging their systems, are part of an upward trend over the past 10 years. Pew research shows that contributions have increased an average of 8% each year over the past decade, boosting assets and paying down debt. In the four states with the most financially troubled pension systems – Illinois, Kentucky, Pennsylvania, and New Jersey – contributions increased by an average of 16% a year over the same period.
Nearly every state has also enacted benefit reforms to lower costs, including cutting benefits for newly hired public workers. Officials in many states have also become more disciplined about managing pension finances, using tools such as stress testing to determine how twists and turns in the economy might affect pension funds. As a result, Pew found that for the first time this century, states are expected to have collectively met the minimum pension contribution standard. This means that even before the market rally during the pandemic, payments into state pension funds were sufficient to cover current benefits and reduce pension debt, a milestone called positive amortization. The improvement in pension funding levels also has led to the highest aggregate funding ratio since the 2007-09 recession. Based on data from the fiscal year that ended June 30 in most states, Pew estimates that the funded ratio – the dollars held in pension funds compared to dollars promised in retiree benefits – has risen above 80%, a level higher than any point since the 2007-09 recession. As a result, Pew projects that state retirement systems will have less than $1 trillion in accumulated pension debt for the first time since 2014.
The significant improvement in plans’ fiscal position is due in large part to dramatic increases in employer contributions to state pension funds in the past decade, which boosted assets by more than $200 billion. Since 2010, annual contributions to state pensions have increased by 8% annually, twice the rate of revenue growth. And for the 10 lowest-funded states, the yearly growth in employer contributions averaged 15% over this period. As a result, after decades of underfunding and market losses from risky investment strategies, for the first time this century states are expected to have collectively achieved positive amortization in 2020 – meaning that payments into state pension funds were sufficient to pay for current benefits as well as reduce pension debt.
Corporates well-funded, too
According to Milliman’s 2021 Corporate Pension Funding Study, despite a decline of 67 basis points in discount rates, the private single-employer defined benefit plans of the Milliman 100 companies continued to make funded ration improvements in 2020 due to their greater than expected investment gains of 13.4%. The Milliman 100 companies sponsor the 100 largest defined benefit plans among U.S. public companies.
The year-end 2020 funded ration for the Milliman 100 companies settled at 88.4%, up from 87.5% in 2019. According to Milliman, such improvement is remarkable given that it estimated the funded ration had fallen to approximately 81% at the end of July 2020.
Overall, according to the report, 19 plans among the Milliman 100 companies had a funded ratio of at least 100%, compared to 14 plans from its 2020 pension survey.
LDI strategies have historically been used not so much for optimizing asset growth – as Jared Gross of JP Morgan points out in his interview above – but rather to manage the disparity between assets and liabilities.
However, as noted in recent Society of Actuaries paper, “Deep Learning for Liability-Driven Investment,” the “complexity of the liability portfolio often makes asset-allocation optimization a difficult task. Both liability cash flows and liability value are sensitive to economic conditions. For example, pension benefits may be linked to the inflation rate. Lapse rates may be affected by household financial status and therefore the economic environment. Liability discount rates are affected by interest rates and credit market conditions.”
Addressing those challenges, according to the paper, can be aided by machine learning, aka artificial intelligence. The report uses deep learning models and reinforcement learning (RL) to construct a framework for learning the optimal dynamic strategic asset allocation plan for LDI.
Deep learning models are trained to approximate the long-term impact of asset allocation on surplus position. Reinforcement learning is used to learn the best strategy based on a specified reward function. RL is expected to learn the patterns and figure out good strategies to maximize the reward. The dynamic investment strategies will then choose the action that maximizes the reward at each decision point in each scenario. However, these strategies may not be the optimal ones but suboptimal ones, because RL does not walk through the entire space of possible asset allocation paths.
To evaluate the effectiveness of RL compared to traditional strategic asset allocation methods, the report uses a sample DB plan modeled with economic scenario generation, dynamic liability projection, asset allocation, and projection and surplus projection. The comparison between optimal static investment strategies and RL-based dynamic strategies is performed assuming two asset classes: an AA-rated corporate bond portfolio and large-cap public equity. Efficient frontiers are built assuming fixed time horizons and static investment strategies. Both fully connected neural networks and long short-term memory models are used to approximate the reward function in the RL process.
Below are the key findings in “Deep Learning for Liability-Driven Investment.”
- With quarterly scenarios and rebalance, investment strategies suggested by RL are driven by average returns rather than periodic fluctuations. This has been tested by using supervised learning with deep-learning models to estimate asset returns, which is not effective. RL may be used for tactical asset allocation with higher data frequency, but not for strategic asset allocation.
- The drivers of the dynamic strategies by RL are liability development and funding status. Compared to static strategies, these dynamic strategies are reasonable but not necessarily the best choices and may achieve a better risk/return tradeoff in some cases.
- RL can reflect risk appetite by adjusting the reward function. By increasing the weight on negative reward (penalty) in total reward calculation, RL moves from aggressive to conservative strategies. However, the relationship between level of risk aversion used in the utility function and the weight on negative reward needs to be explored further.
- RL can incorporate liability complexity, which is challenging for dynamic programming. RL can handle multiple asset classes with reduced training time, which is challenging for a full-blown asset allocation grid search approach.
- The flexibility of AI models allows the inclusion of detailed liability information, refined modeling of the relationship between asset and liability, and other information that is important for decision-making. The new approach opens the door to deriving a dynamic strategy for complicated LDI problems.