Cat Food or Caviar: Sustainable Withdrawal Rates in Retirement

INVESTMENT MANAGEMENT RESEARCH Cat Food or Caviar: Sustainable Withdrawal Rates in Retirement May 2017 Katelyn Zhu, MMF Senior Analyst, Portfolio Construction CIBC Asset Management Inc. katelyn.zhu@cibc.ca Vjosana Klosi, MFE, CFA Director, Portfolio Construction CIBC Asset Management Inc. vjosana.klosi@cibc.ca Imagine you depend exclusively on your backyard apple tree to feed yourself in your retirement. While you have a pretty good idea of roughly how many apples the tree will bear each year, there will be some bumper crop years as well as the occasional drought or snap freeze years. On the consumption side, you know exactly how many apples you need to eat to survive without depleting too much of the harvest on fancy pies or caramel apple frappuccinos. The same concept applies to your retirement portfolio. Determining a sustainable consumption rate is possible once you quantify the expected annual yield and the residual stockpiles you will have left after each year s consumption. You can also improve potential harvests by pruning weaker branches. In your portfolio, pruning can be done by selecting and refining appropriate risk-adjusted investments. Establishing realistic withdrawal rates for individual investor risk profiles can help define an investor s asset allocation as he or she moves through various stages of retirement. In addition, a realistic withdrawal rate is essential in determining whether the current asset mix and investment goals are sustainable. An annual withdrawal rate of 4% has historically been widely accepted as the appropriate rate 1. This is based on a typical balanced 60% equity and 40% fixed income portfolio. Given recently reduced expectations on long-term returns, the 4% withdrawal rate for each of the five investor risk profiles established in the Long-Term Strategic Asset Allocation (LTSAA) paper may no longer be optimal. Longevity risk is now also a key factor to consider. This risk is defined as the probability, at any given withdrawal rate, of a portfolio being depleted before the end of an investor s life. Longevity risk measures the ability for retirees to meet their portfolio return requirements without eroding their principal. In this paper, we focus on testing the validity of the widely-accepted 4% annual withdrawal rate for these pre-established investor profiles. What is an Appropriate Withdrawal Rate for Retirees? A happy retirement doesn t require oodles of money, nor should it mean fighting the cat for food. Shelley Fralic Determining a withdrawal rate that can meet an investor s longterm financial goals is a key consideration, particularly for those investors approaching retirement. The first step in retirement planning is a thorough portfolio allocation assessment, which should appropriately reflect the investor s time horizons and risk tolerance. The second step is ensuring that by combining an appropriate asset allocation with realistic withdrawal rates, the investor can achieve the intended financial outcome with a high confidence level. This confidence interval is generally defined as a minimum 90% probability of success. This approach helps retirees determine appropriate investment strategies that can keep up with their spending requirements, while minimizing the risk of exhausting savings. Figure 1: Factors That Impact Investment Goal Success Projected Returns Asset Allocation Withdrawal Rates Investment Goals 90% Success Failure 1

As shown in Figure 1 (p. 1), projected returns also have a profound influence on the potential success of investors goals. Given the current low interest rate environment, a primary concern is that longterm expected returns have been reduced. Retirees, who typically have a considerable portion of their portfolio in fixed income, will be impacted the most by these lowered return expectations. Historical fixed income returns have averaged 5.8% 2, fueled by a prolonged period of declining interest rates. As it is widely expected that rates are beginning to reverse course, the forecast return for the Canadian bond market is now 2. 3. This could have a significant impact on realistic retirement goals and associated investment decisions. Setting up Retirement Goals Age is an issue of mind over matter. If you don t mind, it doesn t matter. Mark Twain In addition to an appropriate withdrawal rate, the unique circumstances of an investor s retirement goals require an appropriate risk profile and associated portfolio. Using the five profiles from the LTSAA paper shown in Figure 3 below, we have captured the range of possible outcomes based on various withdrawal rates. Given that a number of variables would result in different suggested allocations and withdrawal rates, we have outlined our assumptions in Figure 2 below. Given the five investor profiles and associated asset allocations contained within the LTSAA paper and the above-described retirement and longevity assumptions, we can analyze realistic withdrawal rates. This fundamental analysis will help investors make informed portfolio decisions, with a greater probability of selecting investments that will sustain investors retirement requirements. Withdrawal Rate Methodology There have been numerous studies examining retirement withdrawal rates. They can be divided into two groups: those based purely on long-term historical data for equity and debt returns, and those based on forward-looking return estimates. Historical data, such as that referenced in William Bengen s 1994 study, concluded that a 4% withdrawal rate for income portfolios has a 9 probability of success within 30 years after retirement. However, this needs to be stress tested given current estimated asset class returns. Figure 2: Assumptions Used in Withdrawal Rate Testing Variables Affecting Withdrawals Life Expectancy Emergency Funds Inflation Projections Description The client is projected to have a 30-year life expectancy after retirement. This would not account for a younger retirement age or extended longevity. A longer retirement period would imply a longer investment horizon and require either lower withdrawal rates or riskier profiles with a higher allocation to equities. Expense requirements are held constant and will not account for unexpected or sudden capital needs. Inflation is assumed to be constant at 2% per year, in line with the Bank of Canada s current inflation target. Figure 3: Recommended Asset Allocation for Five Investor Risk Profiles Expected Return 2 Capital Preservation 1 to 3 year time horizon 2 Income 3 to 5 year time horizon Income and 5 to 7 year time horizon 4 6 7 to 10 year time horizon Aggressive 10+ year time horizon Global Equities Money Markets Canadian Bonds Global Bonds High Yield Bonds Canadian Equities Risk 2

Investors must begin by selecting a risk profile that aligns with their circumstances, risk tolerance, time horizon, financial situation, income needs, liquidity and other constraints. Our 2017 LTSAA determined asset allocation recommendations based on the efficient frontier shown in Figure 3 (p.2). These serve as the base portfolios for our analysis. These investor portfolio models combine Canadian and global fixed income, Canadian money market (cash), Canadian and global equities, and U.S. high-yield debt. In addition to traditional asset classes, investors may also consider alternative investments such as infrastructure, real estate and multi-sector fixed income to improve their risk/return profile. IMR s LTSAA paper discusses these in greater detail, and provides a further methodology for asset allocation decisions. After investors establish an appropriate asset allocation profile, the next step is to determine an appropriate withdrawal rate with reasonably low longevity risk. The optimal withdrawal rate is the highest possible rate that still maintains strong probability (defined as 90% or greater) of residual or legacy savings at the end of the investor s projected lifetime. To estimate s for an average 30-year retirement period, we need to apply our estimated portfolio growth rates based on expected returns for each asset class. For any given path of expected future, we subtract a predefined annual withdrawal percentage. If at the end of any given year the estimated portfolio balance is either zero or below the target legacy amount, that is recorded as a failure year. The number of projected failure occurrences during the estimated 30-year retirement period determines the portfolio s success rate for any given withdrawal rate. As we increase the withdrawal rates in increments of 0.1%, we record the probability of success within a spectrum of potential withdrawal rates for each investor profile. To ensure a high degree of confidence in withdrawal rate probabilities, we use Monte Carlo 4 analysis to simulate potential outcomes for the portfolio. All asset prices are simulated simultaneously and their performance is dependent on the returns of all other asset classes, volatility and correlation assumptions. The combined returns of each asset class will determine the remaining asset composition of the portfolio at the end of each year. The performance of the portfolio, combined with the predetermined withdrawal rate, determines the path of the. Forecast Return Methodology and Assumptions When estimating withdrawal rates, careful thought needs to be given to return expectations. Given that market history contains extreme situations such as the 2008 financial crisis, we believe that a forward-looking methodology is preferable to historical modeling. This methodology best captures the current and projected global and financial market environments. As part of this analysis, we estimated expected returns for cash, U.S. high-yield, Canadian and global bonds as well as Canadian and global equities. To estimate cash returns, we used recent interest rate history as a starting point. The 1-year Treasury rate has been in a long-term downtrend since the 1980s, and has remained near zero since 2011. The current short-term rate is 0.6. Based on current demographics and sluggish economic growth, we believe rates are unlikely to revert to their longer-term historical average of in the near future. As a result, our expected return for cash is only 1.3%. Forecast fixed income returns are lower than historical averages and anchored around the index yield-to-maturity. Returns have been adjusted to account for high-yield credit risk and 10-year expected rates for Canadian and global debt. Equity returns are calculated using the Dividend Discount Model, which is estimated by the sum of dividend yields, earnings growth and valuation premiums (see Figure 4 below). Further details on return assumptions can be found in IMR s LTSAA paper. Figure 4: Dividend Discount Model Equity Returns Dividend Yield Real Earnings = + + Valuation Premium To best capture the diverging correlation and volatility of asset classes during both stable and crisis periods, we have separated risk projections into two distinct time periods. These periods are characterized as the normal and turbulent regimes. In order to create these long-term volatility forecasts, our model assumes that there is a 22% chance of a turbulent regime 5. During a turbulent regime, volatility estimates for each asset class are higher than their long-term volatility based on historical returns. The expected volatility for investment-grade bonds and cash is assumed to be higher than the average historical volatility. For equity and high-yield bonds, the expected volatility is assumed to be 70% higher. This approach captures the tail events of a normal distribution, to account for the potential impact of an economic crisis. Figure 5: Long-term Asset Class Returns (as of December 31, 2016) Equities Fixed Income Cash High Yield Global Canadian Global Canadian Canadian U.S. Returns 6.4% 5.6% 1.9% 2. 1.3% 6 4.7% Risk (Normal Regime) 9.8% 10.0% 3.8% 7 3. 1.2% 7.6% Risk (Turbulent Regime) 17.1% 17.4% 4.9% 4. 1. 13.3% 3

Recommended Withdrawal Rate There are two ways to be rich: one is to have great wealth, the other is to have few wants. Dr. David Myers The following graphs illustrate the portfolio failure probability at various withdrawal rates for three (highest, middle and lowest) of the five investor profiles through 30 years of withdrawals. With the previously highlighted assumptions, at 4% or lower withdrawal rates, the failure rate is minimized but not eliminated. These graphs show that, based on the above projected portfolio returns for the various asset allocations, 4% withdrawals are no longer optimal to ensure a high probability of success throughout a 30-year retirement period. Figures 6, 7, 8: Probability of Failure for 3 Investor Profiles at 3- Withdrawals Capital Preservation Probability of Failure The capital preservation profile is the most conservative profile, with 80% of the portfolio allocated to fixed income securities. Based on this profile, a 3. withdrawal rate would have a 17% probability of failure within the 30-year retirement span. A 3% withdrawal rate would have a near-100% chance of success. 0% 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Years In Retirement 3% 3. 4% Income and Probability of Failure 50% 40% This medium-risk profile is recommended for investors who are looking for a balance of capital growth and capital preservation, with income potential a secondary consideration. The debt and equity mix for this profile is 5 fixed income and 4 equities. The recommended withdrawal rate for clients with this allocation is between 3. and 3.7. At 4%, the probability of success breaches the acceptable threshold of 90% for a 30-year retirement period. 0% 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Years In Retirement 3. 4% 4. Aggressive Probability of Failure 40% 3 2 0% 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Years In Retirement 3. 4% 4. This profile is recommended for risk-taking clients seeking longterm growth through capital gains. Clients should be willing to accept significant fluctuations in the value of their investments. The portfolio consists of high-yield debt and 8 equities. The higher expected returns reduce the probability of failure. Here we see a 90% success rate given a 30-year retirement period and a withdrawal rate just shy of 4%. 4

Reflecting our forecast returns in Figure 9 below, closer-tooptimal withdrawal rates are in the range of 2.-3. for the low-risk profiles, increasing to 3.0%-4.2 for the higher-risk profiles. As we increase the allocation to riskier asset classes in the growth-oriented profiles, recommended withdrawal rates increase for a given success rate. This is because the increase in risk is accompanied by higher expected returns. The higher expected returns outweigh the impact of increased volatility and allow investors to withdraw at higher rates while maintaining a high probability of success. We have defined an acceptable minimum success rate as a 90% probability that residual assets remain in the portfolio 30 years following retirement. Greater risk aversion and success probability will necessitate a lower withdrawal rate. According to a survey by the Financial Advisor Journal involving 1100 retirees and pre-retirees between the ages of 50 and 65, the fear of outliving their money is the main anxiety facing retirees 8. Based on this survey, 5 of the respondents said that they want 0% risk of outliving their retirement savings. For this group of investors, a higher confidence level when determining withdrawal rates would be more prudent. Below we show optimal annual withdrawal rates based on four success rates. Figure 9: Recommended Withdrawal Rates for 5 Investor Profiles Probability of Success Capital Preservation Income Income & Aggressive (success probability > 8) (success probability > 90%) (success probability > 9) (success probability = 100% 9 ) 3.2-3. 3.-3.7 3.-3.7 3.7-4% 4%-4.2 3.2-3. 3.-3.6 3.-3.7 3.-3.7 3.7-4% 3%-3.2 3.2-3. 3.2-3. 3.2-3. 3.2-3. 2.-3% 2.7-3.2 2.2-2.7 2.2-2.7 2.2-2.7 10 Figures 10 and 11 depict the effects of 3.7 and 3. withdrawal rates on a hypothetical $500,000 portfolio invested in the Income and profile. The shaded areas represent the dispersion of projected residual s after 30 years, based on one million Monte Carlo simulations. At 3.7 withdrawals, the principal is expected to last at least 30 years with a greater than 90% probability of success. The initial withdrawal amount of $18,750 is increased by 2% each year for inflation, to a final year withdrawal of $33,297. At 3., initial withdrawals are $17,500 but the probability of success increases to approximately 94% and expected residual s increase. The charts below demonstrate that reduced withdrawal rates can have a material impact on expected s. Figure 10, 11: Distribution of Projected Ending Portfolio Values on $500,000 Invested in the Income and Profile Distribution of Ending Portfolio Value Median ending is $300k 3.7 Withdrawal Rate 3. Withdrawal Rate Probability of ending > $890k is 0 100 200 Distribution of Ending Portfolio Value 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 0 Median ending is $390k Probability of ending > $1.08M is 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 Probability of zero residual is Ending Portfolio Value ($Thousand) Probability of zero residual is 6% Ending Portfolio Value ($Thousand) 5

Additional Considerations As with any investment portfolio, appropriate withdrawal rates will also depend on the investor s unique circumstances. Investors who are highly dependent on portfolio returns should lean towards more conservative portfolios and withdrawal rates to avoid any sudden fluctuations in their investments. Additionally, investors with long-term legacy goals over and above their retirement expenses have the flexibility to either invest in higher-yielding, higher-risk portfolios and/or accept a lower probability of meeting legacy goals. Another key consideration is the portfolio s performance during the first few years of the investment, as a considerable initial loss will limit the portfolio s ability to meet its goals. We have estimated the impact of material losses within the first five years on the probability of success using three of the five investor profiles. As shown in Figure 12 below, if portfolios experience a loss of more than in the first five years, the overall success rate is low, even with a conservative withdrawal rate of 3.. As such, investors are encouraged to use withdrawal rates as a guideline, and must be flexible to changing circumstances or market returns during retirement. Figure 12: Probability of Loss Within First 5 Years and Subsequent Success Rates with 3-4% Withdrawals Investor Profile Capital Preservation 3% 3. 4% Aggressive Capital Income and Aggressive Capital Preservation Preservation Income and Income and Aggressive Probability of a > loss within the first 5 years Probability of success following a > loss within the first 5 years 0% 0% 4% 0% 0% 4% 0% 0% 4% 100% 83% 78% 3% 37% 54% 0% 8% 27% Conclusion Declining fixed income return expectations based on ultra-low interest rates have impacted what was once thought to be a gold-standard of 4% annual withdrawal rates during retirement. Our research shows that a 4% withdrawal rate is unsustainable at 100% portfolio success rate. For near-100% probability of the sustaining 30 years of inflation-adjusted withdrawals, the annual drawdowns would need to be reduced to between 2.2 and 3.2. This is given the current market conditions and our forecast asset class returns. The 4% rule would only remain sustainable for higher-risk profiles and a higher failure rate of. As with apple harvests, actual and forecast portfolio returns will vary from person to person and year to year. However, sustainability of withdrawal rates can be improved through factors within an investor s control. By including investments with lower correlation such as dividend-paying equities or alternative fixed income asset classes, investors can prune their investment tree to bolster retirement portfolios. Using realistic and sustainable withdrawal targets is, however, key to ensuring that investors can retain sufficiently-sized slices of apple pie with enough servings to last through retirement. 6

1 Bengen, William P. (October 1994). Determining Withdrawal Rates Using Historical Data 2 Historical average annualized return is 5.8% for the period from 1997 to 2016 for the FTSE TMX Canada Universe Bond Index 3 IMR forecasted returns based on the yield of FTSE TMX Canada Universe Bond Index (2.14% as of December 31st 2016) plus a term premium. Refer to the 2017 Long- Term Strategic Asset Allocation paper for details. 4 Monte Carlo analysis is a portfolio simulation technique used to project s and assess the potential effects of financial market volatility. In this example we have used one million simulations. 5 The crises index built by the CAM Asset Allocation and Currency team is used to estimate the probability of being in a turbulent regime. The index uses historical returns for equities, interest rates, commodities, currencies, etc. Based on the average returns, volatilities and correlations, the index identifies periods when price behaviour is outside of normal parameters. 6 The Canadian cash rate is forecasted using the Vasicek interest rate model, assuming a slow upward-moving interest rate curve and yields of 0.64% as of December 31, 2016. 7 The volatility of the global bond index is adjusted downward from its historical average to account for higher historical Sharpe ratio of the asset class. 8 Dynamic Withdrawal Rule- Simplified, Financial Advisor, October 1, 2015. 9 Success probabilities are rounded up to the nearest whole number 10 Due to the high volatility and downside risk of the Income, and Aggressive profile recommended withdrawal rates are lower versus lower-risk profiles at near-100% success probability Renaissance Investments is offered by CIBC Asset Management Inc. The views expressed in this document are the personal views of the author and should not be taken as the views of CIBC Asset Management Inc. This document is provided for general informational purposes only and does not constitute financial, investment, tax, legal or accounting advice nor does it constitute an offer or solicitation to buy or sell any securities referred to. Individual circumstances and current events are critical to sound investment planning; anyone wishing to act on this document should consult with his or her advisor. The information contained in this document has been obtained from sources believed to be reliable and is believed to be accurate at the time of publishing, but we do not represent that it is accurate or complete and it should not be relied upon as such. All opinions and estimates expressed in this document are as of the date of publication unless otherwise indicated, and are subject to change. Renaissance Investments is a registered trademark of CIBC Asset Management Inc. The material and/or its contents may not be reproduced without the express written consent of CIBC Asset Management Inc. 845 renaissanceinvestments.ca 7