From liability to volatility driven investing

Transcription

1 Asset Management White Paper November 2015 From liability to volatility driven investing A practical way to improve LDI outcomes for pension plans. by Neil Olympio, CFA, FIA, CMT As more and more corporate pension plans have adopted a Liability Driven Investing (LDI) framework, asset glide paths (AGP) have also become more common to provide investment discipline designed to close the gap on funded status. In this paper we follow a sequential journey that may parallel the actual steps a pension plan s staff would be taking to gain investment recommendations approval within its organization. We also explore our research, which suggests that AGP should evolve to directly target funding ratio volatility for the benefit of the pension plan. Exhibit 1 summarizes the results of this paper by plotting the four approaches examined on a risk-and-return basis. We begin with a 60/40 approach as 60% equity and 40% bonds is a common historical strategic allocation that serves as a good reference point for our discussion. This approach will be compared to AGP or an asset glide path; this is the primary step pension plans have taken and reflects an improvement in funding ratio return with additional funding ratio risk. We then take the 60/40 approach and apply a static target funding ratio volatility ( TFRV ) overlay, which improves return characteristics as well as brings the funding ratio risk back down. Finally, we discuss the use of a dynamic TFRV overlay in combination with an AGP, which both raises the return expectations and reduces the risk to a decidedly superior position on the chart. By this point, we expect that you may be interested in just what alchemy is being practiced, and we believe that as you absorb the results of our research as presented in this paper, you will find that this research is both compelling and practical. Exhibit 1: Funding ratio risk/return profile 3.0% Funding ratio return Dynamic TFRV Static TFRV AGP 0 60/40 6% Funding ratio risk 1

2 A somewhat surprising result that AGP may not reduce risk as compared to 60/40 Asset glide paths (AGP) come in different forms; most will have a funding ratio trigger, others have rate-based triggers, and some will have both. The principle behind the AGP is that the allocation will dynamically shift from riskier assets to assets with more liability hedging characteristics, helping reduce the funding ratio volatility as the position improves. There is an expectation that riskier assets like equities will outperform liabilities, driving the funding ratio up and thereby closing the gap between assets and liabilities. One of the risks is that riskier assets might underperform at a time when the allocation to these assets is high. In setting the path to our new approach to pension de-risking, we decided to anchor our analysis in a common base case scenario the archetypical 60/40 portfolio of stocks and bonds. We realize that strategic asset allocations today are far more granular and incorporate a wider range of assets, including real estate, private equity and hedge funds. But for the scope of this study, we thought a 60/40 portfolio would be most effective in setting a standard. Our hypothetical pension plan starts with a 70% funding ratio, and a 60/40 allocation of stocks (S&P 500 Index) and bonds (Barclays US Long Duration Credit A+ Index) (the static 60/40 model). We assume that this pension plan s liabilities will behave in line with our fixed income index, and we also assume that it tactically rebalances to maintain its 60/40 split if the allocation Exhibit 2: 60/40 and traditional glide path (AGP) Key funding ratio results June 2001 September /40 AGP Funding ratio at % Simulated Return % (annualized) Max Drawdown % Average Volatility % Min Volatility % Max Volatility % Transaction costs (p.a) % Information Ratio Asset glide path Funding ratio Equity allocation % Credit allocation % Less than 50% % 60% % 70% % 80% % 90% % 105% Greater than 105% Funding ratio 80% /40 AGP Funding ratio volatility 50% Jun-01 Jun-03 Jun-05 Jun-07 Jun-09 Jun-11 Jun-13 Sep-15 Jun-01 Jun-03 Jun-05 Jun-07 Jun-09 Jun-11 Jun-13 Sep-15 1 Returns are net of transaction costs. 2 Information ratio is defined as the annualized funding ratio return divided by the average annualized volatility. Volatility measures are annualized calculations, using 252 day window. The overall equity exposure was constrained so as not to exceed 100% or fall below 0% for this case study. Please see disclosures for important additional information. 2

3 drifts outside a tolerance band. The start date for our analysis is June 2001 for the practical reason that the daily index data for the chosen fixed income index became available at that time. We then compare the outcome of the static 60/40 portfolio with a similar portfolio that also starts with a 70% funding ratio, but that incorporates a de-risking AGP (the AGP model). Based on our proprietary capital market assumptions, the glide path incrementally shifts the pension plan s assets out of equities and into bonds as the pension plan s funding ratio improves, according to the funding ratio trigger points listed in Exhibit 2 (previous page). The results of this glide path were surprisingly lackluster from a volatility perspective. By following the path, the AGP model ends up in better shape compared to the static 60/40 portfolio s 64% funding ratio in September That said, a 74% funding ratio using the AGP model is only a four percentage point increase from the level the pension plan started with over 14 years ago in June What is most interesting, from start to finish, the over 14-year ride was anything but smooth. Because the glide path increased the plan s allocation to equities in 2008 (i.e., re-risking), volatility was actually worse as compared with the static 60/40 portfolio, with a maximum drawdown of -28.3%, and a funding ratio volatility hurtling toward 40% in This overweight allocation to equities (versus 60/40) is also what helped the funding ratio bounce back from While sticking with the glide path meant that the AGP model plan was better off in 2015 as compared with the static 60/40 approach, most pension plans would appreciate a better risk/ return outcome than this. New paradigm addition of target funding ratio volatility (TFRV) overlay provides significant benefit In our previous example of the AGP model, the glide path systematically shifted assets into bonds (de-risking) or into stocks (re-risking) according to predetermined funding ratio trigger points. We saw how this approach dramatically increased the plan s funding ratio volatility. Since most pension plans would prefer a smoother ride, we next tested our idea of maintaining a specific volatility target using a simple futures overlay. In this approach, we began with a 60/40 portfolio and applied a static TFRV overlay (the static TFRV model). For a like-to-like comparison, we chose a funding ratio volatility target of 13.5%, which is the average volatility of the 60/40 portfolio from June 2001 to September To maintain this particular volatility target across time, we implemented a fairly straightforward overlay comprised of S&P 500 equity futures and US Treasury futures. Generally, these instruments are both highly liquid and inexpensive to trade. If the funding ratio volatility increased above our 13.5% target, then our futures overlay simply increased exposure to the less volatile asset, and vice-versa. To help reduce potential transaction costs from frequent trades, we created a tolerance band around both the physical and notional exposures. 3

4 Exhibit 3: Static Target Funding Ratio Volatility (TFRV) Key funding ratio results June 2001 September /40 AGP Static TFRV Funding ratio at % Simulated Rtn % (annualized) Max Drawdown % Average Volatility % Min Volatility % Max Volatility % Transaction costs (p.a) % 0.8 Information Ratio Static TFRV approach 60/40 Physical Allocation Static TFRV Overlay Futures: Equity and Treasury Funding ratio 100% /40 AGP Static TFRV Funding ratio volatility 50% Jun-01 Jun-03 Jun-05 Jun-07 Jun-09 Jun-11 Jun-13 Sep-15 Jun-01 Jun-03 Jun-05 Jun-07 Jun-09 Jun-11 Jun-13 Sep-15 1 Returns are net of transaction costs. 2 Information ratio is defined as the annualized funding ratio return divided by the average annualized volatility. Volatility measures are annualized calculations, using 252 day window. The overall equity exposure was constrained so as not to exceed 100% or fall below 0% for this case study. Please see disclosures for important additional information. Exhibit 3 shows how our basic futures overlay in the static TFRV model produced noticeable improvements, both in terms of a better return outcome (an 81% funding ratio in 2015) and superior volatility outcomes (drawdown and volatility range) compared with both the static 60/40 portfolio, and the AGP model portfolio. Rather than letting a pension plan s asset allocation decisions drive its funding ratio volatility, our risk paradigm flips that equation around. By embracing a risk-aware approach, a pension plan first identifies its preferred volatility target, and then allows this volatility target to dictate the pension plan s net asset exposures, using a common futures overlay. Combining asset glide path with target volatility glide path provides a dramatic improvement The next stage of our research comprises two steps. First, we take the AGP model portfolio, which decreases equity exposures as funding ratios improve, and combine it with our volatility overlay. Second, instead of maintaining a static volatility target of 13.5%, we create a dynamic volatility glide path, where the targeted volatility systematically decreases as the funding ratio improves (the dynamic TFRV model). For example, if the pension plan achieves a 70% funding ratio, then we shift the equity allocation to 60%, and our futures overlay targets a funding ratio volatility of 8%. 4

5 Exhibit 4: Asset glide path and dynamic TFRV Key funding ratio results June 2001 September 2015 AGP Dynamic TFRV Funding ratio at % Simulated Rtn % (annualized) Max Drawdown % Average Volatility % Min Volatility % Max Volatility % Transaction costs (p.a) % 0.9 Information Ratio Glide path Funding ratio Equity allocation % Funding ratio % (for AGP) Target volatility (for Dynamic TFRV) Less than 50% % 60% % 70% % 80% % 90% % 105% 15 2 Greater than 105% 0 0 Funding ratio 100% AGP Dynamic TFRV Funding ratio volatility 50% Jun-01 Jun-03 Jun-05 Jun-07 Jun-09 Jun-11 Jun-13 Sep-15 Jun-01 Jun-03 Jun-05 Jun-07 Jun-09 Jun-11 Jun-13 Sep-15 1 Returns are net of transaction costs. 2 Information ratio is defined as the annualized funding ratio return divided by the average annualized volatility. Volatility measures are annualized calculations, using 252 day window. The overall equity exposure was constrained so as not to exceed 100% or fall below 0% for this case study. Please see disclosures for important additional information. Exhibit 4 illustrates how adding a dynamic volatility overlay on top of the asset glide path improves the pension plan s outcome across two dimensions. First, the average funding ratio volatility across the over 14-year period drops from 17.5% to 8.0%, with the maximum drawdown improving from -28.3% to -17.9%. Second, funding ratio return increases from 0.3% to 2.3% per annum over the period, leaving the dynamic TFRV model with a funding ratio of 98% in September By decreasing risks and increasing returns, the investment strategy s information ratio increases from 0.02 to 0.29 (in funding ratio space). Summary comparing asset allocation and target volatility paradigms In this paper, we have focused on how one should alter the exposure to risk assets and less risky assets with help from an overlay, based on realized volatility. The focus has not been on optimizing the allocations within each bucket; thus, we have kept our research very direct using US equities and US long credit bonds. 5

6 Exhibit 5: Summary of Results Key funding ratio results June 2001 September 2015 Asset allocation driven Volatility-driven 60/40 AGP Static Dynamic TFRV TFRV Funding ratio at % Simulated Rtn % (annualized) Max Drawdown % Average Volatility % Min Volatility % Max Volatility % Transaction costs (p.a) % Information Ratio Funding ratio risk/return profile Based on simulated results Funding ratio return 3.0% Dynamic TFRV Static TFRV AGP 60/40 6% Funding ratio risk 60/40 Constant Equity Allocation of 60% AGP Asset allocation follows a traditional glide path Static TFRV Targeting an average funding ratio volatility of 13.5% Dynamic TFRV Targeting a dynamic funding ratio volatility level 1 Returns are net of transaction costs. 2 Information ratio is defined as the annualized funding ratio return divided by the average annualized volatility. Volatility measures are annualized calculations, using 252 day window. The overall equity exposure was constrained so as not to exceed 100% or fall below 0% for this case study. Please see disclosures for important additional information. Exhibit 5 summarizes the results across our four approaches to help facilitate easier comparisons. At a broad level, our study examined two paradigms asset allocation versus target volatility. Starting with the traditional asset allocation paradigm, we first compared the outcomes of a static 60/40 portfolio with a similar portfolio that follows an active asset glide path. Next, we switched paradigms and examined how maintaining a static volatility target across time, using a common futures overlay, improved the outcome of the 60/40 portfolio, while also outperforming the asset glide path. Lastly, we created a dynamic combination of the active asset allocation approach with a target volatility overlay, which produced the best risk-adjusted returns of all four test cases. The target volatility overlay helped reduce the allocation to equities in 2008, which explains why its maximum drawdown is not as extreme as compared with the two asset allocation approaches. We believe the improved information ratios of our TFRV overlays present a strong case for a risk approach that focuses on controlling volatility. 6

7 Appendix #1: Performance given recent market volatility The last few weeks in August-September saw increased volatility, mainly driven by concerns about slower growth and potential policy missteps. It is an environment in which we would expect the Volatility Driven Investing (VDI) approach to deliver good results. Let us reaffirm that good results are not defined by superior returns since it is a risk management approach first and foremost. We define good results as the ability to achieve the agreed upon volatility target. With a carefully chosen target, it is expected that the approach would help minimize drawdown in extreme market conditions. Below is an updated study through September 30, 2015 showing how the VDI approach behaved since the previous version of this paper (data was as of February 27, 2015). We estimate that many pension plans had a funding ratio between 85% and 90% (on a corporate bond basis) as of February 27, Below we present the simulated results using both starting points. Up until August 2015, the charts show that the Dynamic TFRV approach evolved in line with the AGP, under both starting points. They decoupled in August, which has been the worst month year-to-date for U.S. equity market performance (as measured by the S&P 500 total return index). Using a starting funding ratio of 85%, both strategies end up around the same funding level as of September close; Starting point: 85% Funding ratio 89% Dynamic TFRV AGP Starting point: 90% Funding ratio Dynamic TFRV AGP 96% Feb-15 Mar-15 Apr-15 May-15 Jun-15 Jul-15 Aug-15 Sep-15 so from a return perspective, in 2015, the difference is minimal. From a plan sponsor s experience however, it is interesting to see that the funding ratio did not drop as much during the most volatile period of the year. The plan s experience is similar for plans that had a funding ratio of 90% at the end of February, albeit with a higher return. This is because the volatility target was lower and the TFRV approach protected the plan even more so; as evidenced by a lower drawdown compared to the 85% starting funding ratio case. In both cases the VDI approach fulfilled its primary objective of funding ratio volatility mitigation at the time when it was needed the most. The behavior exhibited so far in the study (starting in 2001) is that of tracking the AGP during less volatile periods and reducing the volatility during the most volatile periods, in line with expectations. This is encouraging not only for investors who expect volatility to remain elevated for some time, but also for investors who have no particular view, yet are keen to dampen the impact of negative surprises on the funding level Feb-15 Mar-15 Apr-15 May-15 Jun-15 Jul-15 Aug-15 Sep-15 7

8 Appendix #2: Time period sensitivity We are fully aware that a model that looks compelling in one time period can fall flat on its face in another. This is why testing models in different time periods is crucial. To test the timing sensitivity of our models, we altered the start date in two-year increments, starting in June 2001, then June 2003, then June 2005, until we had seven test cases, including the initial case and the six illustrated in the table below. Next, for each time period, we compared the performance of the traditional asset glide path portfolio sans overlay, with one that incorporated our dynamic volatility overlay. First, looking at returns only, the dynamic TFRV model portfolio performed at least as well as the AGP model portfolio in each time period test case. Second, when we only look at risk, our dynamic TFRV model is the hands down winner across all seven time periods. Maximum drawdown, average funding ratio volatility, and maximum funding ratio volatility are lower in each time period compared with the AGP model approach. The most interesting statistic of all, however, is the information ratio (IR), which combines risk and returns in a single measurement. Our dynamic TFRV model produces better IRs in all seven time periods. We do expect to see periods of underperformance relative to the asset glide path, in particular in a strong, upward-trending volatile equity market. This is because the dynamic TFRV portfolio could reduce equity exposure enough so that the AGP model portfolio would produce a better return. Even then, the results would have to be compared on a risk-adjusted basis. Timing matters Key funding ratio results to June 2003 June 2005 June 2007 June 2009 June 2011 June 2013 AGP Results Simulated Return % (p.a.) Max Drawdown % Average Volatility % Max Volatility % Information Ratio Dynamic TFRV Results Simulated Return % (p.a.) Max Drawdown % Average Volatility % Max Volatility % Information Ratio Returns are net of transaction costs. 2 Information ratio is defined as the annualized funding ratio return divided by the average annualized volatility. Volatility measures are annualized calculations, using 252 day window. The overall equity exposure was constrained so as not to exceed 100% or fall below 0% for this case study. Please see disclosures for important additional information. 8

9 Appendix #3: Re-risking versus no re-risking There have been several debates on the topic of re-risking. When the funding ratio deteriorates, should a pension plan s exposure to hedging assets be reduced as per the glide path, or should it be kept unchanged? In discussions of this paper, people have realized that our approach both de-risks and re-risks as appropriate considering the glide path. Some market participants view the glide path as a one-way path, when others may believe that re-risking makes sense. We are very much in favor of re-risking for reasons that go beyond the scope of this paper; however, for completeness, we decided to investigate whether the volatility-based approaches would still show favorable results if we assumed no re-risking. Previously in the AGP model, if the funding ratio were to drop from 70% to 60% for example, we would have increased the equity allocation by 10%. Here, we assumed that re-risking was not allowed, therefore, if the funding ratio dropped from 70% to 60% the equity allocation would be kept unchanged at 70%. In this case the physical allocation to equities can only be reduced, never increased. The re-risking versus no re-risking debate only affects the AGP and dynamic TFRV models. The summary of results in the table below shows that re-risking would have produced better results. In spite of the information ratios dropping for both models, the dynamic TFRV still exhibits superior results against the AGP model. By not re-risking, the average volatility for the AGP model reduces as expected. The issue however, is that when the equity market bounced back off the lows (March 2009), the no re-risking AGP model did not participate as much as the initial re-risking AGP model. For example, over the period of 2009 to 2012 the average allocation to equities for the AGP model is approximately 10% higher than that of the no re-risking AGP model. It is therefore not surprising that the AGP model ended up outperforming the no re-risking version. It may seem counterintuitive to see that the drawdown is slightly higher in the no re-risking AGP case. A lower equity allocation for the no re-risking version might intuitively suggest the no re-risking AGP model should have a lower drawdown; however, one also needs to allow for the behavior of the credit component of the portfolio. The higher allocation to credit was enough to cause the drawdown to be higher for the no re-risking AGP model, albeit by a negligible amount. Having a no re-risking constraint on the physical allocation for the dynamic TFRV model implies that the overlay will have to work a bit more. Because all of the same principles mentioned throughout this paper still apply, we are not surprised to see that the no re-risking dynamic TFRV model outperformed the no re-risking AGP model. Re-risking matters Key funding ratio results June 2001 September 2015 AGP AGP no re-risking Dynamic TFRV Dynamic TFRV no re-risking Funding ratio at EoP % Simulated return (annualized) % Max drawdown % Average volatility % Min volatility % Max volatility % Transaction costs (p.a.) % Information ratio

10 Appendix #4: Implementation considerations One of the interesting features of the volatility driven approach is that it can either be used at the total plan level or by using a sleeve of the total assets. In the latter case, assuming no net leverage, it would make sense to scale the liabilities that are being considered. For example, if the strategy is implemented on 30% of the asset pool, it would be challenging to target funding ratio volatility on 100% of liabilities. It is important to note that the study was conducted assuming a more cautious approach where plans would not be comfortable having more than 100% asset exposure (no net leverage). Using the approach on a levered basis would give us more flexibility, especially with a reduced asset pool under our control. One of the practical challenges we ll have to do more research on is the calculation of the funding ratio volatility for plans that have illiquid assets in their portfolio. Other points of consideration would include the management of the collateral and margin requirements for the derivatives position. It is also important to consider the behavior of the futures in times of market stress. On a final note, and as mentioned in Appendix 2, it is expected that at times the strategy will underperform the AGP approach (on a returns basis not necessarily volatility basis); therefore one could also consider blending the AGP approach with the VDI approach. 10

11 Assumptions Transaction costs It was assumed that each rebalancing transaction resulted in a cost of 0.1% of the market exposure transacted. Transaction costs were calculated for each transaction, and the returns shown are net of these costs. The transaction costs shown in the presentation represent the averages over the time period (per annum). The transaction costs represent the incremental costs on top of the standard approaches. This is the reason why no costs are shown for the 60/40 and the asset glide path cases. It is assumed that since the plan would incur these transactions anyway, it made sense to only calculate the incremental additional costs. Modeling At times, the target volatility cannot be achieved because of market conditions. For example, the equity and fixed income volatility might be so high that it is not possible to achieve the target. When this happens, we calculate the asset weights such that we get the minimum possible volatility given the market conditions at the time. We assumed that any liability cash flows were met by corresponding asset cash flows over the time period considered. Liabilities were assumed to behave in line with the Barclays A+ US Long Credit Index. Results are sensitive to the design of the glide path; therefore, we aimed to select as realistic a glide path as possible. Tolerance bands of 4% around the physical and notional exposure were used to reduce transaction costs. The choice of the 4% was to prevent the exposure from drifting all the way to an equity allocation that would correspond to a different funding ratio. Disclosures The simulated performance information presented for the static 60/40, AGP, static TFRV and dynamic TFRV model strategies represent simulated performance results for the time period indicated. Simulated performance results were created using the indices and percentage allocations as described. Simulated performance results reflect hypothetical retroactive application of models and do not represent the results of actual trading using client assets. Results are provided for informational purposes only. Simulated results are subject to inherent risks and limitations. Investors should not take the example herein as an indication, assurance, estimate or forecast of future results and actual results may differ materially from the simulated results shown. The models may not reflect the impact that material economic and market factors might have had on UBS Asset Management s decision-making if UBS Asset Management were actually managing clients assets during the period portrayed. The simulated results portrayed are not compared to a benchmark, although clients may choose to compare performance to a benchmark. The simulated performance results reflect the reinvestment of dividends or other earnings. The model results do not reflect other costs or fees such as advisory fees, custody fees, brokerage or other commissions, or other expenses that a client would have paid, unless noted below. A client s return will be reduced by advisory fees and other expenses incurred by the client. The following apply for all the models presented below: The information ratio is defined as the annualized funding ratio return divided by the average annualized volatility. Volatility measures are annualized calculations, using 252 day window. Funding ratio (FR) is defined as the ratio of the market value of the assets to the market value of the liabilities. Market value of the liabilities is calculated using the returns on the Barclays US Long Credit A+ index. It was assumed that any liability cash flow was met by a corresponding asset inflow so the modeling assumed no liability cash flow. Funding ratio return is defined as excess return of assets over liabilities Funding ratio risk is defined as the standard deviation of funding ratio return The views expressed are as of October 2015 and are a general guide to the views of UBS Asset Management. This document does not replace portfolio and fund-specific materials. Commentary is at a macro or strategy level and is not with reference to any registered or other mutual fund. This document is intended for limited distribution to the clients and associates of UBS Asset Management. Use or distribution by any other person is prohibited. Copying any part of this publication without the written permission of UBS Asset Management is prohibited. Care has been taken to ensure the accuracy of its content but no responsibility is accepted for any errors or omissions herein. Please note that past performance is not a guide to the future. Potential for profit is accompanied by the possibility of loss. The value of investments and the income from them may go down as well as up and investors may not get back the original amount invested. This document is a marketing communication. Any market or investment views expressed are not intended to be investment research. The document has not been prepared in line with the requirements of any jurisdiction designed to promote the independence of investment research and is not subject to any prohibition on dealing ahead of the dissemination of investment research. The information contained in this document does not constitute a distribution, nor should it be considered a recommendation to purchase or sell any particular security or fund. The information and opinions contained in this document have been compiled or arrived at based upon information obtained from sources believed to be reliable and in good faith. All such information and opinions are subject to change without notice. A number of the comments in this document are based on current expectations and are considered forward-looking statements. Actual future results, however, may prove to be different from expectations. The opinions expressed are a reflection of UBS Asset Management s best judgment at the time this document is compiled and any obligation to update or alter forward-looking statements as a result of new information, future events, or otherwise is disclaimed. Furthermore, these views are not intended to predict or guarantee the future performance of any individual security, asset class, markets generally, nor are they intended to predict the future performance of any UBS Asset Management account, portfolio or fund. 11

12 UBS Asset Management (Americas) Inc. UBS Tower One North Wacker Drive Chicago, Illinois ubs.com/am-us UBS All rights reserved. C /15 UBS Asset Management (Americas) Inc. is a subsidiary of UBS Group AG.