I’m always thankful for the feedback I receive on the site and tools, as it’s a critical part of making sure they are as accurate and helpful as possible. One such bit of feedback today identified some special situations with the Withdrawal Rates calculator where the mathematical accuracy perhaps could lead investors to inappropriate conclusions. That was a great catch and a nice bit of insight, and I’ve updated the calculator to address this issue.
To explain the issue, let me first show what a “good” Withdrawal Rates chart looks like:
Note that the Safe WR numbers always are a little smaller the longer the retirement period you look at, and that they approach a steady-state number similar to the Perpetual WR. This is expected and a good sign that the data sufficiently describes the long term trend with no real outliers. Now contrast that to this portfolio:
That’s quite a bit different. What withdrawal rate should an investor choose? Can a portfolio that has a lower SWR at 20 years really support a higher SWR over 30 years? The short answer is no.
Importantly, however, nothing is fundamentally wrong with the math. Looking at every full 30-year retirement period, the safe withdrawal rate was actually higher than for 20-year periods. The discrepancy is a result of the quantity and timing of the varying-length retirement periods and a bit of impractical chart logic. Displaying the WRs in buckets like this does not account for the very simple fact that if there’s a more recent timeframe where a retiree ran out of money in 20 years, there’s no possible way the portfolio will eventually recover at 30 years once it has enough data to qualify for that bucket on the chart.
To account for this in a way that makes the results as helpful as possible, I’ve introduced the following logic to the calculator:
- Each subsequent retirement period (from smallest to largest) checks to see if the calculated SWR is higher or lower than the previous shorter retirement period. If it is higher, the result is discarded in favor of the smaller number. If it is lower, the new lower number is displayed. Note that this follows the same underlying SWR logic that records the single worst retirement period.
2. The PWR is also capped at the the SWR value. This was not an issue until I introduced #1 above, but is necessary to prevent misleading results.
In short, rather than simply displaying the results for the limited bucket of retirement periods of a given duration, the chart calculates the lowest safe withdrawal rate over ever-growing timeframes.
The goal is to maintain the value of the longer withdrawal rate periods for showing how withdrawal rates may be lower the longer you look, while avoiding any situation where the results may mislead anyone to believe the withdrawal rate is higher than it really is based on timing of the sample. The resulting chart for the example above looks like this:
As you can see, the longer withdrawal rates are all capped at the lowest previous safe withdrawal rate. The results for the long-term withdrawal rates are essentially set aside for over-stating the potential result, but keeping those time periods on the chart allows one to track when longer terms do result in lower rates. Because of how the SWR tracks the single worst retirement period, even a single new low matters.
Regular users of the tool should note that the change does not affect the numbers of any of the sample portfolios on the site at all. And rest assured that the basic math is unchanged. All it does is properly interpret the results to avoid confusion with certain asset allocations, especially involving highly volatile assets like Emerging Markets.
Thanks again to all who provided feedback on this issue. If the explanation is confusing or you find any more things that you feel need to be addressed please don’t hesitate to contact me.