Plenty has been written about withdrawal strategies. Yet, one dimension has been, in my opinion, under-explored: market timing. And believe me, I’ve read more or less everything ERN has written on the topic. Yes, I can hear it from over here: you can’t time the market! Which is true, but not really. Hear me out before getting the pitchforks.
The main cause of failure of the 4% rule for almost any asset allocation is Sequence of Risk: right after you decide to retire, the market plunges by 20, 30, or 60% over the course of months or a few years. My question is: why would anyone in this situation sell shares? The answer is usually: well, I have to sell something, and I have to keep my allocation constant. Hmmm… but why do you have to keep your asset allocation constant? Why can’t you only sell shares when things go well and only sell bonds or gold or use a cash cushion when things go bad, while you wait for your shares to recover?
I have developed a simple set of rules (which constitute the Clever Market Timing Withdrawal Strategy) and back tested them using the data present in ERN’s Toolbox: the monthly inflation adjusted returns of stocks, bonds, cash, and gold since 1871. I’ve (poorly) programmed them in Python.
The results are quite good: this strategy provides 0%, 2%, and about 5% failure rates for 30, 40, and 50 years retirement respectively, for a 4% withdrawal rate. It considers inflation and does not reduce the withdrawals in any circumstances.
The withdrawals are actually increased in good times leading to a median total withdrawal equal to 2.27 times the value of the original portfolio, for 50 years retirement (neglecting the failed portfolios). Only in 10% of the surviving cases are the withdrawals strictly nominal (2 times original portfolio value for 50 years retirement). The value of the surviving portfolio after 50 years is larger than the original value in 75% of the simulations and only lower than 40% of the original value in 5% of the cases (again, only considering the surviving portfolios for 50 years retirement). The median value is around 1.9 times the original portfolio value after 50 years.
August 1929 almost makes it to 40 years while September 1929 survives almost 49 years. The longest stretch of time during which the strategy fails before 50 years is almost 2 years (from February 1905 to January 1907). The worst month to retire with this strategy would have been September 1906 (failure after 34.67 years).
Higher SWR don’t produce terrible results either: 12% failure for 4.25% and 17% failure for 4.5% over 50 years if you renounce the increase of withdrawals in good times.
The original asset allocation is: 2 years worth of withdrawals in bonds, 2 years in cash, and 2 years in gold (for a 4% SWR, that’s 8% each) and the rest in stocks (76% for a 4% SWR). I don’t like cash or gold as “investments” but boy do they do well in terrible sequences…
The main idea is to withdraw from stocks only, unless the value of your stock portfolio drops below 90% of its original value adjusted for inflation. You then sell other assets until the value of your stock portfolio goes back to its original value, adjusted for inflation. Assuming that at the beginning of your retirement, you have $1m, you would sell 40k worth of stocks, leaving bonds, cash, and gold untouched. That’s assuming yearly withdrawals but I have used monthly withdrawals in my simulations.
So here are the rules:
- If the value of your stock portfolio is above or equal 90% and less than 120% of its value adjusted for inflation, sell stocks.
- If the value of your stock portfolio is below 90% of its original value (adjusted for inflation), you only sell bonds. In our example, assuming that the value of your stock portfolio has dropped below 684k, you would sell 40k worth of bonds.
- If bonds are not enough, use cash. If there isn’t enough cash, sell gold.
- You switch back to selling stocks either when the value of your stock portfolio is above its the original value adjusted for inflation, or when there is nothing else to sell.
- If the value of your stock portfolio is larger that 1.2 times its original adjusted for inflation, then increase your withdrawal by 100% (yes, double it). Being conservative with a 4% SWR and not increase the withdrawal in good times doesn’t change the results much.
What I’m actually looking for is someone to check this algorithm because I can’t be 100% sure that I’ve not made a mistake somewhere, especially since it seems a bit too good to be true. I don’t want someone to check my code, rather, someone to program these rules independently and perform the same simulations to compare results. Honestly, I’m expecting someone to point out something obvious that I’ve missed…
Thanks a lot for your support!