Please Note: Blog posts are not selected, edited or screened by Seeking Alpha editors.

Dynamic Withdrawal Policies for Retirement Portfolios


A common approach to retirement planning is to determine an amount that can be initially withdrawn from a retirement portfolio, and increased each year by the rate of inflation. For example, Monte Carlo simulation might show that someone withdrawing $35,000 a year (in today’s dollars) from a $1,000,000 portfolio with some given asset class allocation would result in a terminal portfolio value after forty years of at least $250,000 in 85% of the simulations. 

The problem with modeling a retirement plan based off of a fixed initial withdrawal rate that adjusts for inflation each year is that it does not account for the wide range of potential outcomes.  Some of these will be negative outcomes – such as the 15% of outcomes where the afore mentioned portfolio dropped below $250,000 – but others will be positive, in that the portfolio will grow at a higher rate, making the initial 3.5% withdrawal rate effectively fall over time. 

Now most people would not be content to keep withdrawing the same (inflation adjusted) value from a portfolio when it has significantly increased in value.  In this article, I am going to use Monte Carlo simulation to look at the effects of implementing a dynamic withdrawal policy, which in the previous example could be stated as:

Withdraw the greater of either $35,000 (adjusted annually for inflation) or 3.5% of the portfolio’s value at the start of a given year.

In this example, I am going to use a sample portfolio with the asset class allocation shown below, where it is assumed that the portfolio will be rebalanced annually to maintain the allocation.  In the screenshot below, LCDS refers to large cap domestic stocks, LCFS to large cap foreign stocks, REIT to real estate investment trusts, and STDB to short-term domestic bonds.  Note that the rate of return statistics for each position are real (inflation-adjusted) values.  The portfolio, which will now be imported into a Monte Carlo simulator, has an expected real arithmetic rate of return of 5.2%, with a standard deviation of 10.6%.

I ran two simulations on this portfolio, one where an inflation-adjusted amount of $35,000 is withdrawn each year, and another where the greater of either $35,000 in today’s dollars or 3.5% of the portfolio’s value at the start of a given year is withdrawn.  The results are shown in the following table:

 

 

 

%  Depleted

Terminal value at 85% confidence level

Static Withdrawal

8.79%

$339,646

Dynamic Withdrawal

8.97%

$300,143

 

 

As you can see, there is not a huge effect on the downside scenarios.  Now let’s look at the upside.  The following chart plots the static withdrawal scenario in blue, and the dynamic withdrawal scenario in red. The portfolio values at an 85% confidence level (i.e., the portfolio values were at least this high in 85% of the simulations) are plotted against the primary y-axis, and the average withdrawal amount is plotted on the secondary y-axis.  You can see that by year 10, on average, the dynamic withdrawal policy results in a $44,167 withdrawal amount vs. $35,000 for the static policy.

 

It is pretty clear that the dynamic withdrawal policy allows a retiree to participate in positive outcomes while at the same time not adding significant risk.  We have been using a dynamic withdrawal rate for our own retirement portfolio since 2005, and to date, are satisfied with the results.

The portfolio in this article, which gives a reasonable balance between return and rate of return volatility, could be implemented using the following ETFs available from iShares: ACWX for the LCFS asset class, IVV for the LCDS asset class, ICF for the REIT asset class, and IJR for the SCDS asset class.  The STDB asset class can be tracked using Vanguard's VBISX

If you would like to try out the software I used in this article, you can visit my website at

 web.me.com/briangaudet/ThePatientInvestor/Home.html

 

Aside from the portfolio modeling and Monte Carlo software, there is also a mean variance optimizer and company analysis tool.

I have no position in any of the funds discussed in this article.