Just How Safe are ‘Safe Withdrawal Rates’ in Retirement?

This study considers one of the cornerstone questions in the retirement income debate; namely, what’s a safe withdrawal rate for retirement? The much celebrated 4% Rule has become a popular heuristic that has provided a quick shortcut to ‘solving’ this most difficult of retirement planning problems.

The pioneering work in the field was contributed by Bengen (1994).² Using historical simulation, the study shows that the retirement portfolios of people who retired during the period 1926 through 1976 and withdraw 4% of the initial balanced portfolio value every year (adjusted for inflation) could be sustained for at least three decades.³ The second group of studies that provide support to the 4% Rule are known as the Trinity studies.⁴ These studies use a simple, but highly informative, approach to investigate withdrawal rates with respect to different asset allocations, and several time horizons. In their most recent paper, Cooley, Hubbard, and Walz (2011) consider an observation period from January 1926 through December 2009. This study suggests that retirees who plan to make annual inflation adjusted withdrawals should stay within the 4% to 5% range.⁵

As we entered the 21^st century, the sequence of major events that have occurred in the first decade included: 9/11; the dot.com bubble; the sub-prime crisis and the GFC. This unfortunate succession of events across the early 2000s resulted in a level of wealth compression in investment portfolios not seen for many years. This period of heightened volatility underscored the importance of path dependency to the sustainability of retirement income (Basu and Drew, 2009; Basu, Doran and Drew, 2012). It has given rise to a far more critical assessment of the 4% Rule (and its variants).

The work of Spitzer, Strieter, and Singh (2007) and Spitzer (2008) has been important in developing a line of argument that suggests the 4% Rule may be an oversimplification of a complex process that involves the analysis of risk tolerance, asset allocation, withdrawal size and expected returns. Using a bootstrap approach, these studies examine a myriad of withdrawal rates finding that the fixed 4% rule is not always safe and that dynamic approaches to the withdrawal rate may assist the retiree. Harris (2009) finds that sequencing risk is a key determinant of the sustainability (or otherwise) of safe withdrawal rates, with rates varying in the range of two to four percent.⁶

Pfau (2010) conducted the first major study to examine the issue of safe withdrawal rates from a larger selection of countries. This study replicates the methodology of Bengen (2006) by using the Dimson, Marsh, and Staunton (2012) data from 1900 through 2008 for 17 developed countries. The analysis provides some interesting results finding that the 4% withdrawal rate is not safe when using the original Bengen (2006) maximum safe withdrawal rate criterion. The findings show that, even with the assumption of perfect foresight, the maximum safe withdrawal rate exceeds 4% in only 4 of the 17 countries, range between 4% and 2% in a further eight countries, and are less than 2% in five countries.

In order to provide positive insights into what is, at its core, a normative question, our review of the literature suggests that it is prudent for researchers to investigate capital markets that have very, very long historical data series and, if possible, markets with different return distributions. For this reason, we have non-randomly selected five countries to stress test the 4% Rule. We ranked all 19 countries in the Dimson, Marsh and Staunton (2012) database in descending order based on their respective annualised performance (real accumulated returns) of stock returns for the period 1900 through 2011 (a total of 112 years) and those countries representing the key percentile levels (minimum; first quartile; median; third quartile and maximum) are selected to test safe withdrawal rates under different asset allocations and investment horizons (Table 1).⁷

Given the centrality of inflation (and its relationship to stocks, bonds and bills through time), we use real returns throughout the study. Specifically, instead of using nominal rates of return and then adjusting withdrawals each year for inflation, we elect to use real returns thereby avoiding the annual inflation adjustment. Annual withdrawal rates ranging from one percent through 10 percent (in increments of 100 basis points) are considered across investment horizons of 10, 20, 30, and 40 years. Given that Australians are living longer lives (and many Australians retire before 65 years of age), we argue it is important to include the 30 and 40 year horizons to provide positive insights into the robustness of safe withdrawal rates across longer horizons. We consider the Golden or 4% Rule for stock allocations ranging from 0% to 100% (in increments of 25%) for each of our representative countries (rebalanced annually, for brevity, we only report the 50:50 results in this paper) and report maximum safe withdrawal rates (or “SAFEMAX” as in Bengen, 2006).⁸ Finally, we assume that retirees make an initial withdrawal at the commencement of each year. That is, the initial withdrawal amount is equal to the specified withdrawal rate times the starting balance of the portfolio (Pfau, 2010).

We have highlighted (bold and underline) five countries: Australia (“AUS”); New Zealand (“NZL”); Netherlands (“NLD”); Japan (“JPN”); and Italy (“ITA”), in the table as they represent annualised performance levels that most closely correspond to key percentiles in the distribution of the annualised performance of stock markets over the long run.⁹

Figure 1 shows the evolution of $1 invested in each of these countries over the full length of the data set (note the logarithmic scale on the y-axis).

Figure 1:

Turning specifically to the 30 year planning horizon, we report SAFEMAX results (that is, the maximum withdrawal rate that ensured portfolio survivability) for a 50:50 growth/defensive asset allocation (Figure 2). Even with the stellar performance of Australian equities historically, we find success for the 4% Rule in the shortest of timeframes, with horizons greater than a decade exposing the hypothetical investor to some chance of ruin (SAFEMAX for Australia is estimated at 2.96%, some 100 basis points below the golden rule).

Figure 2:

For completeness, we also report SAFEMAX results for New Zealand (3.64%), the Netherlands (3.19%), Japan (0.24) and Italy (0.89). A wider range of asset allocations, investment horizons and payout rates are provided in the full report (Drew and Walk, 2014).

Our results confirm that whatever you think you need as superannuation nest egg, it is almost certainly going to be less than you actually need. The conversation is a difficult one in that, for many investors, their focus is on the asset side (particularly, the return portion) of the equation, not the liability side. We posit that the first challenge in tipping the scales in the retiree’s favour is to get the framing right, moving from a ‘pot of gold’ (asset) mindset to an ‘income replacement’ focus (liability).

It’s time for a difficult conversation. Let’s assume (somewhat heroically) that a couple has a retirement nest egg of $1m today. How can we begin to assist retirees with framing reasonable expectations given different starting balances?

The Association of Superannuation Funds of Australia (“ASFA”) has developed the ASFA Retirement Standard benchmarks that estimate the annual budget needed by Australians to fund either a comfortable or modest standard of living in retirement. It is updated quarterly to reflect inflation, and provides detailed budgets of what singles and couples would need to spend to support their chosen lifestyle. We argue that these benchmarks are a critical component to improving the framing of retirement income decisions. The ASFA Retirement Standard (June Quarter 2013) shows that, in general, a couple looking to achieve a ‘comfortable’ retirement needs a budget of $56,406 a year, while those seeking a ‘modest’ retirement lifestyle need a budget of $32,656 a year (ASFA, 2013).¹⁰

For the purposes of providing a practical perspective to the safe withdrawal debate, we can consider (on a $1m starting balance), a real income requirement of 3.27% (that is, 3.27% of $1m = $32,700 p.a. for 30 years) for a modest income level; and a 5.64% for a comfortable income (5.64% of $1m = $56,400 p.a. for 30 years).

We can consider the ASFA benchmarks as forming a retirement income channel through which retirees are attempting safe passage (in this case, safe passage is avoiding portfolio ruin). Even if we exclude the countries with the lowest safe withdrawal rate results (Japan and Italy), the results on a starting balance of $1m for a couple suggest that the ASFA modest range is vastly more sustainable than the comfortable equivalent. Even at this withdrawal rate, history suggests that a couple would still face somewhere between a 10 to 30 percent chance of portfolio ruin for a 30 year horizon. As a form of ‘ready reckoner’, we include in the table below different starting points, and their safe withdrawal equivalent percentage.

In short, holding a 50:50 portfolio over 30 years, the highest SAFEMAX100 rate we report in this study is from New Zealand at 3.64%. This suggests that even using the BEST result from our sample, a couple with a starting balance of $1.5m would, using history as a guide, still face some probability of portfolio run. We again acknowledge the limitations of the 4% Rule, particularly its deterministic nature. In the real world, retirees face an array of expenses, the frequency of which range from well-known (such as utility bills, insurance costs, general living expenses) to some which are stochastic or random in nature (for instance, major unanticipated health events).

However, the 4% Rule can be used as a very helpful heuristic for retirees (a mental shortcut to assist in our understanding the challenge of income planning). Like many shortcuts they can provide imperfect answers to help us better understand complex problems. As noted by Scott, Sharp, and Watson (2009), the 4% Rule imposes an opportunity cost on retirees and is therefore is inefficient. We would certainly echo their view. The 4% rule helps us initially engage cognitively in the retirement income problem which, as we have seen from this study, is simultaneously complex and dynamic in nature.

We have limited skills in forecasting whether or not the retiree gets the ‘bad’ draw out of the cosmic investing world. We know that if the sequence of returns is against us (particularly when the largest amount of our nest egg is at risk) and the timing is wrong, the reality is that some investor is going to get the worst outcome.

However, there are many levers that can be coordinated to tip the scale in the favour of the retiree, including more dynamic approaches to the:

Withdrawal rate (mortality updating, regular mid-point reviews and updating of the cash flow profile of retirees);

We acknowledge that this is not an exhaustive list. However, these are some of the key levers that our research findings suggest can fall within the gambit of ‘known knowns’.

The days of searching for the retirement income silver bullet are over. In this study, the Golden or 4% Rule works for favourable sequences of returns (let’s be honest, everything works in such markets), ignores asset values of the day and is decoupled from the dynamic nature of the asset-liability mismatch faced by many Australians. However, the 4% Rule does present us with an opportunity to form a baseline which can dramatically improve our framing of expectations of what’s possible in retirement.

For the future, we need to move from a silver bullet approach (such as the 4% Rule) and instead employ a veritable arsenal of weapons (based on dynamism: withdrawal rates; asset allocation; planning horizon; fees and after-tax management; scenario testing; risk management; investment governance) to assist retires in managing and mitigating the asset-liability mismatch in retirement.