Before the Great Recession and Great Bear Market of 2007 to 2009, it was considered to be eminently reasonable and prudent for people to take a withdrawal rate of 4% from their portfolios after they retired. But after the market collapse, some have suggested that those withdrawals should be closer to 3% if the money is going to last throughout a person’s lifetime. But I suggest that the sustainable withdrawal rate is closer to 5%, and also that those investors in the withdrawal phase are likely underallocating to stocks because of their misguided reliance on risk-tolerance questionnaires.

One of the more common ways of finding a suitable withdrawal rate starts with a forecast of the future rates of return and the level of inflation. But such an approach means the analyst will be making predictions about the future. I make no attempt here to forecast the future returns of stocks, bonds or inflation.
Instead, my analysis relies on historical data, specifically the monthly total returns provided by Global Financial Data (a San Juan Capistrano, Calif., firm) from September 30, 1875, through June 1, 2013. A look at robust data from the last 138 years allows us ways to think about the future (it contained, for instance, three different depressions).

By using past returns, I am implicitly assuming that future returns will be similar to past ones, though the way we get there may be different. (The stocks in my analysis are represented by the S&P 500, while for bonds I use 10-year U.S. Treasurys and for inflation the All Urban Consumer Price Index.)
Let’s assume we have a couple who have just retired with $1 million of investable liquid assets (Figure 1). They want to invest it in some combination of stocks and bonds.

In this example, the withdrawal period of 353 months (29.42 years) could be thought of as the couple’s prime retirement years. At the end of this period, an end-of-life annuity contract would kick in, funding any potential remaining years.

Simulations And Results
Using the 138-year data set, I looked at exactly 1,300 separate and distinct 353-month-long retirement windows. Each window framed a uniquely different financial journey, and many of these passed through one of the depressions. Figure 2 shows the results the couple would have seen if they had invested everything in stocks.

Note that if the couple annually withdrew $41,940 (an initial 4.194% of the starting principal), they had a 5.00% probability of running out of money before the end of their almost 30-year retirement window. Even more instructive is what happened if they included bonds in their portfolio, which significantly increased the likelihood of exhausting their money (unless they drew down at the lowest percentages). These results are shown in Figure 3.

If the couple shifted from 100% stocks to a 50/50 stock/bond mix, they increased their probability of running out of money from 5.00% to 10.31% at the initial 4.194% starting principal withdrawal rate (the $41,940 figure). Figure 4 shows the lowest risk portfolios for different starting annual dollar withdrawal rates. Recall that the term “lowest risk” means the portfolio has the least likelihood of running out of money before the end of the retirement window.

For initial withdrawal rates at or below 3.75% ($37,500 of the $1 million initial investment), the lowest risk portfolio was 70% stocks and 30% bonds. Such a portfolio generated a 0.00% probability of failure. In other words, it never failed during the 1,300 unique trails that incorporated the three different depressions. However, for withdrawal rates at or above 4.516%, the lowest risk portfolio was 100% stocks. Initially, one might find such a result counterintuitive, i.e., that stocks deliver a more conservative and safer journey than bonds. But recall that in extended withdrawal scenarios (in this example, almost 30 years long), growth is of absolute importance and the stability that results from the inclusion of bonds is of minimal relative importance. For extended periods, growth trumps stability. Thus, the all-equity solution dominates mixed stock/bond portfolios.

Conclusions
These data might also suggest that the couple should adopt a more conservative portfolio and a lower withdrawal rate (3.75% instead of 4.75%). For example, perhaps they feel that the odds they will run out of money must be kept below 5.00%. In which case, they would adopt an 80%/20% stock/bond mix and an annual dollar withdrawal rate (increased each month by inflation) of $41,920 on their initial $1 million investment. But that conclusion is excessively conservative. Because once they have embarked on their journey of withdrawals during retirement, if they fall too far behind, encountering a rare market crisis (of which there is less than a 5% chance), then they can easily adjust their annual consumption level down.

Consider the following example. Assume that the couple prefers to establish a high $47,500 annual withdrawal rate (with monthly inflation increases) on their initial $1 million investment. History since 1875 says they face a 15.77% probability of running out of money before the end with a 100% equity portfolio.
They would have run out, for example, during one recent retirement window from December 31, 1972, to May 31, 2002. This was a particularly traumatic period. It began with a challenging 21-month bear market (from December 31, 1972, through September 30, 1974) and then ended with a second horrific 21-month bear market (from August 31, 2000, until May 31, 2002). Retirees were hit twice, once at the beginning of the period and once again at the end. A couple taking $47,500 annual withdrawals would have run out of money during their 237th month (only 67% of the way through their retirement). They would have ended the 353-month period with an investment balance of negative $641,012.

But this problem has a relatively easy fix—one that imposes little pain. The couple would have to create a glide path for their investment—assuming what the steady rate of return would be. If they simply compared the value of their ongoing investment balance to that glide path, then they could make adjustments. Say, if their balance fell below 85% of the glide path amount, they could reduce their monthly withdrawal by 10.3%. If they withdrew $47,500 each year or $3,958.33 each month, for instance, they would bring it down to a sustainable monthly $3,550.62. The couple would only need to do this once during their 353-month withdrawal period, and they would reduce the probability of running out of money to 0.00% (assuming they have an all-stock portfolio).

Similar results can be shown for almost all of the other problematic periods, i.e., remarkably small incremental adjustments are all that is needed to the monthly withdrawal rates to keep the couple well within a sustainable trajectory. The approach would have worked throughout all of the 1,300 unique trials observed in the return data since 1875.

To those who suggest sustainable withdrawal rates might be closer to 3%, I suggest that the data support withdrawal rates closer to 5%. And for those who advise their retired clients to allocate most of their assets to bonds, I suggest that they re-evaluate their simulations. The data suggests they should do otherwise. That means traditional risk tolerance questionnaires have the potential to do serious harm to investors during retirement years that require growth. 

Rob Brown, PhD, CFA, is the Chief Investment Strategist for United Capital Financial Advisers, LLC. The views expressed are his, and may not reflect the position of United Capital.