Academic journal article Journal of Economics and Economic Education Research

A Reexamination of Safe Retirement Withdrawal Rates

Academic journal article Journal of Economics and Economic Education Research

A Reexamination of Safe Retirement Withdrawal Rates

Article excerpt


One of a retiree's primary financial concerns is that they will outlive their retirement savings. Bengen's 1994 article "Determining Withdrawal Rates Using Historical Data" provided an answer to that question - four percent of the total portfolio at retirement with future annual withdrawals adjusted for inflation. He determined that level of withdrawal was sustainable for thirty years based on return data from rolling thirty year periods from 1926 to the time of the article. Other authors (Guyton and Klinger, 2006, Robinson, 2007, and Tezel, 2004) have addressed this issue subsequent to Bengen and have introduced adjusting withdrawal rates based on current conditions and providing information on the effects of higher withdrawal rates. All have used historical data to provide a basis for their conclusions.

More recent contributions have taken different approaches. Pfau (2011) determines maximum withdrawal rates at the time of retirement based on market conditions at the time of retirement. He develops a model that determines maximum withdrawal rates based on the ten year cyclically adjusted earnings yield, the nominal bond yields for ten year government bonds, and the ten year dividend yield. His model finds maximum withdrawal rates as high as 8.8 percent in 1982 and under two percent in 2003. Blanchard and Blanchard (2008) use average returns and their standard deviation and predictions of future returns from a variety of different investments. They calculate the probability of exhausting ones resources based on investments with a given mean return and standard deviation. Finke et al (2013) conclude in a time when real returns for bonds are at or below zero, the four percent withdrawal rate may no longer be valid.

What these studies have in common is they make assumptions about rates of return, inflation, and life expectancy and then evaluate whether the four percent withdrawal rate and a variety of alternatives are valid. This paper uses a different methodology to evaluate variations of the four percent methodology and the endowment and minimum distribution methods. First, Bengen (1994) and a number of other authors used a series of actual returns in their studies, usually a thirty year duration. Rather than using a thirty year time frame to determine the validity of a given withdrawal rate, the necessary duration of retirement savings is determined based on the age of the female spouse based on census data and the actual life expectancies of each spouse. The thirty year time frame used in other studies is not relevant if the actual years in retirement are twenty or thirty-five. Second, history may not be a good predictor of future returns, but volatility in the markets is a certainty. To determine rates of return, historical monthly information is used, but the rate of return is selected randomly from returns from a twenty-six year period, September 1987 to December 2013. Therefore, actual rates of return are used in this study, not in the order that they occurred, and are different for each observation.


To evaluate the propriety of the various methodologies for determining withdrawal rates, a model was developed that utilizes a Monte Carlo simulation with ten thousand observations.

The first step was to determine the life expectancy for each observation. In the first iteration of the model, it was assumed that the male would retire at the age of sixty-five. The age of his spouse was then determined based on a uniform distribution applied to US Census data on differences in age between a husband and wife (2000). The life expectancy of both members of the couple was then determined based on a uniform distribution applied to mortality tables provided by the Social Security Administration (2012). The number of years the retirement income would be required was then based on the maximum of the life expectancy of the couple.

The second step was to determine the rate of return for each month during the couple's maximum life expectancy. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.