The Role Of Long-Maturity TIPS In Retirement Portfolios by Steven Sapra & Niels K Pedersen –PIMCO
All things equal, we believe long-duration Treasury Inflation-Protected Securities (TIPS) should play an important role in the asset allocation of workers who are within 10 to 20 years of retirement. Long TIPS effectively help hedge retirees against fluctuations in the real wealth required to sustain a given level of consumption in their retirement years. As such, long-duration TIPS act as the “risk-free asset” in the context of retirement savings, and, for investors on a reasonable savings path, an allocation to this asset class may increase the likelihood of a successful retirement outcome.
While the secular tailwind of declining real and nominal yields, which have contributed to favorable returns of long-duration fixed income over the last 20 to 30 years, has most likely come to an end, the prospective returns of long-maturity TIPS are still sufficient to justify a meaningful allocation, in our view. Alternatives such as cash or shorter-duration bonds are expected to have worse performance over the next 10 years than long TIPS, under our base case view. Risk assets such as equities and higher-yielding fixed income, which have the potential to outperform long-duration TIPS, should also be considered for retirement savings portfolios. However, their inclusion increases the uncertainty of final outcomes, particularly as one approaches retirement.
Although changes in market-wide risk premia will inevitably cause both the relative and absolute valuations of asset classes to change in the future, the role of long-maturity TIPS as the “risk-free” asset will remain undiminished. For example, a rise in the equity earnings yield is likely to translate into higher future risk-adjusted returns. A DC investor may optimally allocate more to equities as a result. We believe long TIPS, however, will continue to be the most appropriate asset for hedging the liability, irrespective of valuations, and, as such, should always serve an important role in retirement investors’ portfolios.
The Relationship Between Real Wealth And Real Consumption In Retirement
To evaluate the current state of a retirement plan, we can invoke a basic principle of financial planning, which is that the present value of retirement consumption should be less than or equal to the present value of assets set aside to finance it. For a worker requiring a real income stream of CR, lasting T years, and assuming a constant real interest rate, r, this condition implies a minimum wealth level at retirement, WR, of
The same principle applies N years before retirement, which implies the following condition:
where SR-N is the worker’s future real savings. If we define as the “real price index” for K years, we can restate these conditions as follows:
We can consider the current retirement savings to be consistent with a worker’s retirement objective if Equation 4 holds. In other words, an investor is on track to his retirement goal if the current value of his wealth plus the present value of future savings is greater than or equal to the present value of his anticipated future consumption. If Equation 4 does not hold, then the deficit can only be closed by either lowering consumption expectations in retirement (CR) or increasing savings (SR-N). This concept is similar to how a corporate pension plan compares the value of its asset portfolio with its expected liabilities to determine the plan’s funded status.
Figure 1 shows the impact of real yields on the retirement price index, along with the liability duration, assuming 20 years of retirement income. Each line represents the number of years to retirement, ranging from zero (at retirement) to 20 years. The value of the liability is directly influenced by the level of real interest rates in the economy, with lower rates translating to greater liability values and vice versa. Additionally, the further one is from retirement, the greater the duration, or interest rate sensitivity, of the liability. For example, at retirement, the duration of an investor’s liability is about 10 years.
Empirical Performance Of Asset Classes And The Real Liability
Figure 2 shows the retirement price indexes relevant for three different ages – 55, 60 and 65 – from March 1999 to June 2016, alongside the evolution of 10- and 20-year real yields.1 The index is computed assuming 20 years of retirement income. Over this period, the wealth required to finance a given retirement income increased substantially, driven by the steady decline in real interest rates. Real rates were 4% in 1999, then hovered around the 2% level from 2003–2007, before recently falling to an unprecedented range of nearly ?1% in 2012. Since 2012, real rates have rebounded somewhat, ranging between 0% and 1%, resulting in a reduction of the cost of retirement from its peak in 2012.
As a result of this secular decline in real yields, the wealth required to maintain the retirement incomes of a 55-, 60- and 65-year-old worker grew by a staggering 90%, 60% and 40%, respectively, between 1999 and 2016. One million dollars of retirement savings could have supported about $70,000 of real retirement spending in 1999. By 2012, that number had fallen below $50,000, and more recently the number is just over $50,000. Hence, focusing purely on one’s wealth would have given the false impression that the investor was equally as welloff in 1999 as in 2012. Looking at the problem through the retirement income lens, however, shows that real wealth has been a poor proxy for prospective retirement consumption.
To further illustrate this point, consider a worker who was 55 in September 2005, worked for the next 10 years and retired in September 2015. Furthermore, assume that he was fully funded relative to his retirement objective in 2005 and hence on track for a successful retirement. From 2005 to 2015, this worker’s liability grew at a real annualized growth rate of 3.0% per year, which we can compare with the real returns he would have earned by investing in four different asset classes over the same period: cash (three-month Libor), equities (S&P 500 Index) core bonds (Barclays U.S. Aggregate Index) and long TIPS (Barclays 10+ Year TIPS Index). Table 1 shows the real performance of each asset class, along with the liability.
Over this period, equities were the only asset class that outperformed the liability. With the benefit of hindsight, the worker could have more than kept up with the retirement liability by investing 100% of his retirement savings in the S&P 500. However, equities were also by far the most volatile asset class, both in terms of total volatility and tracking error to the liability portfolio, at 19.8% and 24.4%, respectively. At the beginning of 2009, a worker holding 100% U.S. equities would have appeared to be heading for a very dire outcome. In fact, not until age 63 – in 2013, following a strong rebound in U.S. equity market performance – would the worker have been in a surplus situation. If the worker had retired in September 2011 rather than September 2015, he would only have been able to afford 60% of his desired real retirement income. Long-duration TIPS, on the other hand, also would have fully funded the worker’s liability but done so with much less volatility. Long TIPS almost perfectly matched the liability return and, at 2.2%, had the lowest tracking error versus the liability. A core bond portfolio nearly kept up with the growth in the retirement liability, at 2.8% and 3.0%, respectively, but with significantly more tracking error than long-duration TIPS. This high level of tracking error is largely attributable to the substantial duration mismatch between core bonds and the liability. The suboptimal nature of cash investments over this period is evident. The very low real return of ?0.20% per year did not come close to keeping up with the liability performance, resulting in a cumulative shortfall of nearly 30% over the 10-year period.
Asset-Liability Performance In A Forward-Looking Context
An important question for many investors is whether the current low levels of real yields make long-duration fixed income and real duration unattractive. It seems to be a foregone conclusion that yields will mean-revert fast enough that duration risk must be a losing proposition over the next 10 years. PIMCO’s forward-looking Capital Market Assumptions (CMAs) can help address this question. Figure 3 shows our expected returns for various asset classes in nominal, real and liability-relative space for the next 10 years. We also include our return assumptions for duration-matched TIPS, which explicitly match the overall duration of the liability each year.
Under our base case, long-duration TIPS continue to look attractive relative to the returns of cash, Treasuries and TIPS. The underlying assumption that drives this result is that real rates will rise slowly to long-term levels that are somewhat below their historical averages. Our forecast is that the 10-year TIPS yield will be 1.1% in 10 years, versus the long-term average of 1.8%. This results in an expected real return to long TIPS and the liability of just under 2%. Cash is again expected to be a notable underperformer. Indeed, under our base case, cash is expected to underperform the liability by about 2% per year, making cash a very unattractive investment currently, unless one expects a major, sudden sell-off in interest rates. Equities have the highest expected return of all the asset classes, both on an absolute basis and relative to the liability return. The expected relative return over the next 10 years is modest, however, at just 0.4% per year. Equities may once again prove to provide the highest return and be justified ex post, but the expected return benefit appears to be small given current valuations. We do not suggest that equities or equity risk be excluded from the asset allocation but note that our forwardlooking return expectations are low relative to the historical norm, especially for U.S. equities.
Rather than focusing on just the average performance, Figure 4 shows the range of outcomes that may be realized given our simulations of forward returns.
Interestingly, while cash, bonds, long bonds and TIPS have different stand-alone volatilities, they are all characterized by fairly similar return volatility when measured versus the retirement liability. In all cases, the 10th and 90th percentiles represent deviations of under +/?15% versus the liability. The distribution of long TIPS and duration-matched long TIPS shows the least amount of return variation, with relative returns between ?3.4% and + 2.9% and ?2.0% and 1.8%, respectively. Figure 4 also again highlights the high degree of volatility associated with the relative return of equities. For instance, in 10% of our simulations, equities underperformed the retirement liability by 29% or more in a given year. In the other extreme, in 10% of right tail outcomes, equities beat the liability return by 31%. Defined contribution (DC) investors may still be tempted to hold a large allocation to equities due to their higher expected returns. But in reality, the ultimate outcome is highly uncertain and, as such, a high allocation to equities may result in a “feast or famine” retirement.
Robustness Analysis
To examine the robustness of the conclusions in the previous section, we include a set of alternative scenarios that represent perturbations to our baseline assumptions for real interest rates and inflation. The scenarios are defined as follows:
(i) Real rates and nominal rate scenarios – Our baseline assumption is for a terminal real and nominal 10-year yield of 1.1% and 3.2%, respectively. For robustness, we shift the real and nominal yield curves down from the base case by 50 and 100 basis points (bps) and up by 50, 100 and 150 bps. Expected inflation is held constant. This results in expected 10-year real yields ranging between 0.1% and 2.6% and 10-year nominal yields between 2.2% and 4.7%.
(ii) Inflation scenarios – Our baseline assumption for inflation is 2.1%. We shift our 10-year inflation rate between 1.0% and 3.5%. This is accomplished by shifting the nominal rate by the same quantities as in the interest rate scenarios but holding the real rate fixed at the base case value of 1.1%.
Table 2 compares the current, terminal and average values for the scenarios. The top section shows our assumptions for the interest rate scenarios, while the second panel shows the inflation assumptions. The base case discussed in the prior section is highlighted in red. By construction, inflation is held fixed in the first set of scenarios at the baseline view of 2.1%, while real rates are fixed in the second.
As before, we estimate forward-looking returns for the following asset classes – cash, equities, core bonds, long bonds, TIPS, long TIPS and duration-matched TIPS – under each scenario. Figure 5 shows the results for the interest rate and inflation scenarios.
As expected, long TIPS and duration-matched TIPS show consistent performance across all scenarios, effectively matching the liability performance.2 Of course, while long TIPS are the most consistent performer, this does not mean that they are the best-performing asset class across all scenarios. In fact, equities are expected to outperform long TIPS in all but the ?50 bps and ?100 bps rate scenarios. Both bonds and TIPS outperform long TIPS in the +100 bps and +150 bps scenarios, as both benefit disproportionally from the larger duration mismatch to the liability when rates rise. Long nominal bonds underperform across all scenarios, a result of the low breakeven inflation rates embedded in nominal bonds relative to our baseline inflation forecast of 2.1%. Cash underperforms in all scenarios except the +150 bps scenario, in which it just matches the liability portfolio. Hence, even in our most severe rate scenario, cash at best keeps pace with the retirement liability.
Panel B of Figure 5 shows the average return results for the inflation scenarios. Both TIPS and long TIPS maintain a constant relative return throughout the scenarios, a result of inflation impacting both TIPS and the liability in an identical fashion. However, TIPS systematically underperform as a result of their lower carry compared with long TIPS. Cash underperforms in all scenarios, but it is particularly poor in the high inflation regimes. Equities fare better on a relative basis when inflation is low and underperform when inflation is high. Nominal bonds and nominal long bonds underperform the liability by a significant margin in the high inflationary scenarios due to their fixed coupon nature. Positive inflationary shocks are unambiguously bad for nominal bondholders, as they reduce value through the impact of higher yields while at the same time increasing the value of the real liability.
Discussion
Given the favorable results of long TIPS, an obvious question is, Why would a worker saving for retirement hold anything other than long TIPS? According to Vanguard’s How America Saves 2015 survey, the typical 55-year-old has $64,000 in his DC retirement account.3 Such levels of funding imply materially lower income in retirement than most workers expect. As such, the majority of DC investors need to generate returns over and above that of the real liability in order to build the necessary wealth for retirement success. In our base case scenario, the average return to the real liability is 4% per year. In addition to increasing their contribution rate, investors with insufficient savings may need some exposure to return-seeking assets such as equities, credit fixed income, emerging market bonds and even alternatives, most of which have expected returns higher than 4%. However, allocations to such assets, while potentially increasing the expected retirement income, come at the expense of greater outcome uncertainty.
To illustrate this point, Figure 6 shows the risk/return trade-off between high returning equities and duration-matched TIPS for the base case scenario. We sequentially allocate away from equities to long duration-matched TIPS and show the cumulative returns relative to the liability over the next 10 years, along with the 10th and 90th percentiles. As expected, while the excess return falls when allocating away from equities, the variation around the return drops substantially. For example, at 100% equities, the cumulative relative return ranges between ?88% and +137% versus values of ?4% and +5.5% for duration-matched TIPS.
Another question may be the relevance of nominal duration bonds in a DC glide path. Nominal long bonds are expected to underperform long real bonds in every real rate scenario and all but one inflationary scenario (long nominal bonds outperform in the lowest-inflation scenario). Due to their much smaller outstanding market value, TIPS tend to be less liquid than nominal bonds.4 In times of crisis, there has often been a large discrepancy between the performance of real and nominal government bonds, despite the fact that both are U.S. government guaranteed and driven by similar fundamental risk factors. In the majority of market environments, real and nominal yields have been closely correlated. However, in crisis periods this relationship has sometimes broken down. For example, in the 2008 financial crisis the 10-year nominal yield decreased from 3.94% to 2.21%, while the 10-year real yield actually increased from 1.63% to 2.09%, causing large differences in returns.5 Over this same period, the S&P 500 declined by 28%. Therefore, nominal bonds can play an important risk-hedging role, particularly in times of severe stress, while TIPS may not.
Under our baseline New Neutral thesis, the returns to major asset classes are modest. In fact, of the seven asset classes in the prior section, only equities are expected to outperform the liability under the base case. Furthermore, significant outperformance of certain asset classes is only observed in scenarios that deviate strongly from the base case view. However, our results are only shown for core asset classes. Venturing into asset classes outside of core bonds and U.S. equities should offer potentially higher long-run returns and thus greater opportunity for outperformance. Another consideration is the role of alpha, or performance above the manager’s benchmark. To the extent investors allocate to managers who offer the opportunity to add value through skillful asset and sector selection, or via access to less liquid segments of the market, the likelihood of outperforming one’s liability may be higher. This is particularly true when alpha is achieved later in the savings cycle, because there is typically “more money on the table” midcareer and into retirement.
Conclusions
The appropriate benchmark for retirement savings is the discounted value of a stream of required real income in retirement. The amount of real consumption that such a structure can provide is a function of three basic parameters: total savings, time horizon and interest rates. If an investor is unable to generate sufficient wealth by the time he retires, he simply will not be able to enjoy the lifestyle he desires without taking extra risk in retirement. The components of time and interest mean that the retirement savings benchmark is sensitive to the level of interest rates in the economy, with higher rates providing a greater level of consumption and vice versa. We have shown that duration-matched long TIPS are the most appropriate asset class for hedging the investor’s liability, both in a historical and a forward-looking context. While long TIPS generally do not provide an opportunity for adding value in excess of the liability, they may largely insulate the investor from a wide range of interest rate and inflationary scenarios. Higher-returning assets, such as equities, may be necessary for investors who need to generate returns over and above the liability during the savings cycle. But long-duration TIPS are the only asset that tracks this liability with a high degree of accuracy. As such, our research suggests that a properly constructed retirement glide path will allocate significantly to return-seeking assets but will gradually devote a portion of the asset allocation to longduration real bonds as retirement approaches. Such a balance provides the return potential necessary to help ensure investors’ retirements are properly funded, but does so with an appropriate amount of caution by ensuring that a portion of their asset allocations are invested in the “risk-free” retirement benchmark.
See the full PDF below.
