Demystifying decumulation: A practical guide

Among investors pursuing long-term objectives like retirement, there is an understandable tendency to focus on the accumulation of assets, and the process, while by no means easy, is fairly straightforward, with a limited number of variables to consider. But as populations around the world age and increasingly need to tap into accumulated wealth, there will be a growing focus on and demand for sound decumulation practices — and here the investment challenge is decidedly more complex.

The defining characteristic of a decumulation strategy is the need to support regular withdrawals from a portfolio regardless of the performance of the underlying investments. Unless the strategy generates a sufficient return, these withdrawals will eventually erode the capital below a desired amount or even erode it completely (longevity risk). Even if the return is equal to the withdrawal rate, erosion could still occur, as the return replacement process is affected by path dependency (also known as sequencing risk), which is in turn determined by the interaction between the withdrawals and specific characteristics of the strategy’s realized return. These characteristics, which we’ll refer to collectively as the portfolio distribution, include returns, volatility, skewness, peakedness, and the left tail of the returns.

In this paper, we examine the scope of the decumulation investing challenge and the impact of each portfolio distribution characteristic. We then suggest practical methods for improving the likelihood of success in decumulation strategies, focusing on portfolio construction, process design, and implementation. These insights may be of use to investors constructing their own decumulation strategies or those evaluating third-party strategies.

The impact of withdrawals and the path-dependency effect

Figure 1 illustrates the role that withdrawals play in the decumulation investing challenge. The orange line shows a US$1 million hypothetical portfolio that generates a -10% return and then experiences a withdrawal of 4% of the initial portfolio value (US$40,000). The portfolio would then need to generate a return of more than 16% to get back to the US$1 million starting point, compared to an 11% return for a portfolio with no withdrawals (dark-blue line). For comparison, we also included an accumulation portfolio (light-blue line), in which capital is contributed periodically and portfolio drawdowns are more limited.

Figure 1

Next, we look at how the distribution of returns creates the path-dependency effect (Figure 2). We start with two hypothetical US$1 million portfolios and then assume withdrawals of 4% of the initial portfolio value (US$40,000) annually, spread over equal amounts on a monthly basis. We assume that Portfolio A returns exactly 5% (annualized) every month while Portfolio B has a distribution of returns centered around 5% (annualized). While the expected returns are the same, Portfolio A has a 100% probability of capital preservation over 30 years and Portfolio B has only a 71% probability of the same outcome. This result, highlighted in the top chart of Figure 2, is entirely due to path dependency, which is a function of the portfolio distribution.

The lower chart of Figure 2 examines the impact of the portfolio distribution in more detail, showing that even if the withdrawal amount and return target are equal (highlighted in orange), there is only about a 50% chance of preserving capital through time. Why? The actual return is in a range due to the distribution, so 50% of the time, withdrawals will not be matched by an equivalent (or higher) return. To improve these odds, investors need to improve the distribution, which we focus on next.

Figure 2

The critical role of the portfolio distribution

The key variables in a decumulation strategy, as we’ve established, are the level of withdrawals and the realized returns of the strategy. Lower withdrawal rates will, all else equal, lessen the erosion and reduce the longevity risk. But this option may not be feasible, depending on the ongoing consumption needs of the decumulation investor.

With this in mind, we opted to focus on how the realized returns, a function of the portfolio distribution, can be modified to improve the outcome of decumulation strategies. Specifically, we looked at how five characteristics of the portfolio distribution can impact the success of a decumulation strategy: returns, volatility, skewness, peakedness, and the left tail of the returns.

We used Monte Carlo simulations to analyze outcome probabilities for several hypothetical portfolio scenarios, assuming varied distribution aspects and, unless otherwise noted, using a 5% expected return (annualized), 7% volatility (based on the assumed risk level of a simple 40% equity/60% bond mix, which we would consider an appropriate reference mix for the 5% return target, and our capital market assumptions), and a normal distribution (0 skew and 0 excess kurtosis) as the starting point. Each scenario was based on a 30-year period with monthly observations, 10,000 simulations, and an initial portfolio value of US$1 million. We assumed annual withdrawals of US$40,000 (4% of the initial portfolio), taken in equal amounts on a monthly basis. (Please see the “Important Disclosures – Capital Market Assumptions” at the end of the paper for additional information.) Figure 3 summarizes our analysis and findings (more detailed results available upon request.)

Figure 3

Pursuing a better decumulation strategy

Next, we’ll focus on what these research findings mean for portfolio construction, process design, and implementation of a decumulation strategy.

Portfolio construction

The starting point for all portfolio construction is the expected return and risk, which for decumulation strategies is complicated by the rate of withdrawal. A key finding from the above analysis is that decumulation strategies should target an expected return above that of the withdrawal rate to help mitigate the erosion of capital over time. Broadly speaking, the higher the return above the withdrawal rate, the greater the likelihood of success. However, higher returns are also often associated with higher risk, which can worsen portfolio outcomes, as we’ve discussed. Therefore, volatility should be as low as possible for a given level of return, implying maximization of the Sharpe ratio.

The analysis also suggested that decumulation portfolio design should aim for negative skew and higher peakedness. We think this focus on negative skew, which stems from the importance of where the mass of the distribution is concentrated (as opposed to the length of the tails), differs from standard practice, which tends to focus on the tails. In practice, we acknowledge the difficulty in building portfolios that target specific return distributions, but we believe it should be a consideration during portfolio construction.

Thus far, we’ve focused our analysis on multi-asset portfolios, which we prefer given the opportunity to blend assets to pursue a favorable distribution shape and potentially reduce risk. Investors who prefer to use a single asset class need to bear in mind its distribution characteristics. Figure 4 shows the skewness and peakedness¹ for various segments of the equity, bond, and commodities² markets. Overall, the dispersion in distribution characteristics of the different asset classes is high. Investment-grade credit, US high-yield bonds, and EM sovereign debt, in particular, have higher degrees of negative skew and peakedness, both of which we consider desirable for decumulation strategies.

Figure 4

Governance and risk-mitigation processes

Decumulation strategies have a long time horizon, and it is therefore important to structure the governance process in a way that will minimize the temptation or even the ability to make changes in reaction to short-term events. For example, responding to a portfolio decline by reducing the portfolio’s risk level (a common behavior) could worsen the path-dependency risk by lowering the expected return. One way of avoiding this is to design a governance process that calls for infrequent action on asset allocation, performance, and manager decisions, and keeps the focus on long-term, forward-looking data. We also think that any ongoing monitoring should include a review of any changes in the portfolio distribution and the consistency of the role that different building blocks are playing in the portfolio.

Turning to risk mitigation, our analysis highlighted the risks that volatility and large drawdowns pose for decumulation strategies. For the former, volatility-control techniques may play a role. For the latter, investors may want to consider active asset allocation, constant proportion portfolio insurance (CPPI), and put options. Active asset allocation may help avoid drawdowns by moving capital away from certain areas before negative returns occur. CPPI strategies shift capital between a “risky” portfolio and a “low-risk” portfolio at a rate intended to protect against drawdowns of a specific size. Lastly, option strategies are designed to pay out in the event of a significant market drawdown, potentially limiting the impact on portfolio returns.

Note that any benefit gained by reducing left-tail outcomes through these methods will need to be weighed against the cost associated with the protection, which shifts the mean of the distribution to the left.

Implementation

Once the portfolio has been constructed and the process design is in place, the portfolio can be implemented in a way that contributes to a beneficial portfolio distribution. For example, strategies that specifically focus on generating income may produce a distribution that supports decumulation strategies. (Note that this is not about using “natural” income to pay out the withdrawals, but rather about how targeting income may provide beneficial distributional characteristics). To demonstrate, we compared a traditional multi-asset mix (60% equity/ 40% fixed income) with a 60/40 mix that targets higher income (using high-dividend stocks, global bonds, high-yield bonds, and covered calls).³ In this hypothetical example, adding an income goal to the portfolio did not materially change the mean, standard deviation, or skewness, but it did have a significant impact on kurtosis (3.5 vs 2.3), which would lead to more peaked distributions and potentially be beneficial for decumulation strategies.

While income distributions are typically more peaked, they can also have larger tails. Therefore, when targeting income, we think it is important to have a mechanism in place to avoid drawdowns such as those discussed earlier. In fact, for income investors, it can sometimes be beneficial to seek total return funds with income-like distributions, which can avoid reliance on the more constrained “income” opportunity set.

Decumulation investors may also want to consider strategies that seek to manage drawdowns and avoid negative returns, both critical for decumulation strategies. In Figure 5, we look at how a variety of equity implementations could affect the portfolio distribution. Specifically, the charts show distribution characteristics for seven equity factor indices — all flavors of the MSCI USA Index, which we use here to represent different types of equity managers. The results vary greatly, suggesting that decumulation investors should be deliberate about manager selection in each asset class and think about the right “levers” to pull. For example, investors may want to tilt toward lower-volatility and income-orientated strategies when implementing their equity allocation.

Figure 5

Putting theory into practice

The insights we’ve shared on portfolio construction, process design, and implementation may be of use to investors establishing their own decumulation strategy or to those evaluating third-party decumulation strategies. To help, we have summarised what we believe are the five critical steps for constructing a decumulation strategy and five questions to ask a third-party manager in Figure 6. We would welcome the opportunity to discuss these ideas and our research in greater detail.

Figure 6

¹Peakedness is measured here as the height of the return distribution, as a percentage of the highest return distribution (Japanese govt. bonds). | ²Despite having less appealing distribution characteristics, commodities, given their inflation sensitivity, often still play an important role in portfolios for investors with real spending objectives. | ³Traditional mix based on: 60% MSCI ACWI, 40% Bloomberg Global Aggregate. Higher-income mix based on: 60% MSCI ACWI High Dividend Yield, 20% Bloomberg Global Agg, 10% Bloomberg Barclays US Corp High Yield, 10% CBOE S&P 500 Buywrite (minus S&P 500 return). Monthly returns in USD from January 2001 to December 2021. Analysis is for illustrative purposes and does not represent the results of an actual strategy. Hypothetical portfolios are developed with the benefit of hindsight (e.g., knowledge of market conditions) and thus have many limitations. PAST RESULTS ARE NOT NECESSARILY INDICATIVE OF FUTURE RESULTS AND AN INVESTMENT CAN LOSE VALUE. Investments cannot be made directly into an index. Performance of actual strategies managed in these styles may differ substantially from the performance presented. This does not include the effect of fees. If fees and expenses were reflected, the performance would be lower.