It has been a while since I ran Firecalc, so I decided to have a look. I started with a $1,000,000 portfolio, a $30,000 (3%) withdrawal and a 50 year time frame. I used the default market portfolio with equity percentages varying between 30% and 75% in 5% steps. All other inputs were at the default. Then I did it all over again with a 4% withdrawal. The results were surprising.
At 3% withdrawal, everything is hunky-dory. Firecalc predicts 100% success for all portfolios with more than 40% equities, and the success rate is still 92% at 30% equities.
The picture is drastically different at 4% withdrawal. Even at 75% equities, the success rate was only 85.4%. At 30% equities, the success rate is only 41%.
This is very disturbing to an old retired engineer. I want solutions to be stable and relatively insensitive to input parameters. If the solution is on a part of the curve with a very steep slope, like it seems to be here, I get nervous.
Here it seems like the difference between 3% and 4% withdrawal is like the difference between easy street and eating cat food.
Have I screwed up (again)? Any Firecalc gurus out there?
I know that modern portfolio theory and all that is not well accepted here (with some justification IMHO), but it is worth remembering that the standard deviation of the endpoint of a random walk is proportional to the square root of the time frame.
yuk: I wish I could make that sound less geeky, but I can't.) In other words, if you annual portfolio standard deviation is say, 10%, then projecting out 50 years with a random walk model means that you are looking at a standard deviation of the final portfolio value of about 70%. Even if you don't believe in the random walk model, it does seem reasonable that the amount of "fuzziness" in results will grow with time.
One last problem. For a 50 year timeframe, Firecalc only ran 89 simulations. I'm not going to bother calculating the Student-T distribution values for that (I can sense eyes glazing over just at the mention of it), but suffice it to say that I wouldn't base my retirement solely on statistical results derived from 89 simulations. It just ain't enough. This isn't a slam against Firecalc. There just isn't enough market history to go on.