top of page

Playing Limbo with Your Withdrawal Rate?

For over a decade, we've seen many in the personal finance space advocate for very low withdrawal rates for those dependent on their portfolio for income. Larry Swedroe, for instance, has put forward that '3% is the new 4%', suggesting a 33% reduction in the amount that retirees' can spend with relative safety from a portfolio comprised mainly of stocks and bonds. This argument appeals to investors who are already rather risk averse, but it seems to have also gained some traction among those who don't fully understand the reasoning underlying such low withdrawal rates nor, more importantly, the ramifications of employing them.

In this post, I attempt to explain why many have suggested that retirees should withdraw less from their portfolio going forward than they've been able to in the past, the implications of using such low withdrawal rates, and what investors might consider to remedy the situation.

Historic Safe Withdrawal Rates

Bill Bengen's seminal research on the topic of safe withdrawal rates (SWRs) in 1994 has been followed by a slew of work examining SWRs in myriad contexts. But there has been much misunderstanding of precisely what Bengen and others meant by the term 'SWR'. This method is also referred to as a constant-dollar withdrawal in that the first year's withdrawal is a percentage of the portfolio's balance but that subsequent withdrawals are for the same dollar amount, adjusted for inflation. For instance, this means that an investor employing a constant-dollar withdrawal rate of 5% from a portfolio of $1 million would withdraw $50,000 in the first year and continue to withdraw $50,000 in later years, adjusting the amount for inflation.

Bengen and others have largely found that for U.S. investors, the SWR over a 30 year period was about 4%, depending on the asset allocation (i.e., the stock/bond split) used. Note that 'success', as Bengen and most others have defined it, with this method is not depleting the portfolio before the 30 year period (or whatever historic period is being studied) is finished. Retirees in 1966, the worst starting year for U.S. retirees employing this method, would have completely depleted their portfolio by the 30th year. But it should be pointed out that in most 30 year periods in U.S. history, the constant-dollar withdrawal rate that turned out to be 'safe' was well above 4%, often 6% and sometimes exceeding 9%.

However, the outcome has often been less favorable for investors in other nations. Investors whose portfolios only consisted of assets in their home country often had experienced significantly lower SWRs. Note that many of the very low SWRs of the past occurred in nations that were heavily involved in a bad way in World War 1, World War 2, or both, and likely shouldn't be relied upon. If investors had spread their holdings globally, the 30 year SWR for investors in most developed nations was roughly 3.5%.

By the nature of the methods used to derive it, the SWR over a given time frame (e.g., the last 100 years) represents the lowest constant-dollar withdrawal rate that succeeded in every 30 year period within that time frame. In other words, it is representative of the worst historic performance of the assets being analyzed. For this reason, many believe it to be very unlikely that their withdrawal period will be worse than the worst historic period on record, though it is certainly possible that things could turn out otherwise.

Another vital point to keep in mind is that 'failure' in the context of a SWR does not mean 'completely depleting one's portfolio' in the real world, because virtually no sane people will voluntarily continue to withdraw the same amount from their portfolio if the latter has been performing poorly. Rather, essentially everyone reduces their withdrawals if their portfolio is suffering.

Bond Yields

One of the most prominent justifications put forward by many for lower SWRs going forward is low bond yields. Bonds' returns can be predicted with a reasonable degree of precision (i.e., within an annual return of about +/- 1% for 10 year Treasuries) and are precisely known in advance for Treasuries or TIPS held to maturity, making it seemingly easy to use them to estimate forward SWRs.

The problem with this logic is that inflation-adjusted (i.e., real) bond returns have not historically been a good predictor of the SWR for any given period of time. Consider that while Treasuries returned about -1.6% real from 1941-1981, a 4% constant-dollar rate succeeded in every period within that time frame. Rather, the factor with a far larger impact on SWRs has been the performance of the stocks within investors' portfolios.

Stock Valuations

Since stocks have had a far larger effect on portfolio returns than have bonds, many have attempted to find a means of forecasting the returns of stock markets. While several variables have been put forward as having been statistically good predictors of stock market returns, those that have gained most attention are valuations. Valuations refer to a means of determining the current value of a financial assets, including individual stocks or entire stock markets. Many valuation metrics exist, but the one that has gained most attention with regard to forecasting the future returns of stock markets is the cyclically-adjusted price to earnings ratio (CAPE) developed by Robert Shiller. An in-depth analysis of this metric is beyond the scope of this post, but one of the problems with this metric is that, since Shiller put it forward in the late 1980s, it has consistently and significantly underestimated the returns of U.S. stocks, which have grown to represent roughly 60% of the all the world's stocks in total value.

Despite the inadequacy of valuation metrics in generating reasonably accurate estimates of forward stock returns, there is another, bigger problem with this entire line of thinking. Due to the impact of sequence of returns risk, even if we knew exactly what the average returns of a portfolio would be for the next 30 years, we would still be unable to generate a precise prediction of what the SWR would be over those 30 years! The reason for this is because the order in which returns occur is vital to how much investors can safely withdraw from a portfolio, and, to date, no one seems to have found a good means of forecasting what the impact of sequence of returns risk will be, and I suspect that nobody ever will due to the apparent randomness in short- and mid-term stock market returns.

Nubble Lighthouse, York, ME

Problems with Using a Low SWR

Beyond the presence of major issues with the methods used to forecast a lower SWR than what has been seen in history, there are several real problems with an investor using a very low (i.e., significantly under 4% for an assumed 30 year retirement) withdrawal rate.

First, it takes most investors significantly longer to build their portfolio to the point that they can implement a 3% withdrawal rate. The time it takes to go from a 4% withdrawal rate to 3% (i.e., the portfolio going from 25 years' of one's spending to 33.3 years' spending) has historically been between 5 and 10 years. Those years can never be recovered, and for investors who are approaching traditional retirement age (~age 65), they represent a substantial portion of their remaining longevity. For instance, if a given 65 year old has a remaining life expectancy of 25 years, 5 years represents 20% of this person's remaining longevity, and 10 years represents a whopping 40%. Unless such an individual truly enjoys this time spent continuing to work to accumulate more retirement assets, the reduction in the risk of needing to reduce one's withdrawals at some point in the future is far less than the risk of working substantially longer than needed. However, those preparing for a retirement lasting longer than 30 years can and likely should consider pursuing an initial withdrawal rate below 4%, probably closer to 3%.

Second, for married couples of traditional retirement age, the likelihood of both surviving 30 years is quite small. Using data from the Social Security Administration, the probability of both opposite-sex spouses aged 65 surviving another 30 years is only 1%. The expenses for a surviving spouse should fall significantly after the passing of the other spouse (e.g., expenses for food, Medicare premiums, medical care, potentially housing), thereby reducing the amount needed to be withdrawn from the portfolio.

Third, even low withdrawal rates can result in a portfolio declining in value and not recovering for a long time. Retirees in the year 2000 who used a 3% constant-dollar withdrawal rate with a portfolio consisting of 30% U.S. stocks, 30% ex-U.S. stocks, and 40% U.S. bonds saw their portfolio decline in inflation-adjusted value by 33% by the year 2003, and though their portfolio went on to recover to its starting value very briefly in late 2007, it then declined again to a low of 61% of their starting, inflation-adjusted balance in early 2009 and didn't fully recover until late 2020. At the end of 2022, the portfolio would be down about 22% from its starting, inflation-adjusted balance.

While higher withdrawal rates would obviously have led to deeper drawdowns and longer recoveries from portfolio declines, this clearly indicates that even a constant-dollar withdrawal rate as low as 3% can result in a portfolio losing significant value in the mid-term and not recovering for a decade or longer.

Final Thoughts

Investors of traditional retirement age should think long and hard about their starting withdrawal rate in retirement. As one's withdrawal rate declines, the risk of needing to make reductions in one's withdrawals in the future also declines, but the guaranteed time needed to achieve a lower withdrawal rate increases substantially, and this time likely represents a big chunk of investors' remaining lifespan.

bottom of page