what's interesting is to not use longest time frame but to set the start date to 2007, so you're going into retirement at a very bad time. i did it w PRWCX. started with 1 mil and at the end of 2019 you had 780k with a max drawdown of 55%, at the 10th percentile. scary. but at least you still had money. oh -- did retirement of 20 years. thanks for posting this!
I could not figure out where to set a custom start date. I'm pretty sure I clicked all the options. What am I missing?
There is a tendency to use this tool as black box oracle. I'm not faulting the use of a simulator to run models per se. Rather I'm suggesting that people may not fully appreciate what is being modeled.
The Monte Carlo simulation engine of PV does not appear to allow a user to pick the starting date for the simulated runs. When one specifies a range of dates (which one does by setting "Use Full History" to "No"), one is specifying the data set (annual returns) from which the simulator
randomly selects returns. It doesn't mean that the simulated runs start with the 2007 performance.
By selecting 2007 to 2019, you're telling the simulator to use one of 13 annual returns for each year in each run. Which means, among other things, that a run of 20 years must duplicate the returns from some years, since it needs 20 1 year returns and it's got only 13 years to choose from.
See "Historical Returns" in the "Methodology" section of PV's
FAQs.
The simulator does have an option where you can tell it to start with the worst year (or worst two, or worst three, or ...). So if a simulated run of 20 years has returns r1, r2, ..., r20, and r
5 is the worst, the simulator reorders the returns as r
5, r1, r2, r3, r4, r6, ....
Better, but not perfect, because the worst run in your data set may not be in the simulated run. Still, this is much better than nothing.
---------
Numbers:
If you use this option with QQQ over 30 years with an initial withdrawal amount equal to 4% of the portfolio (subsequently inflation adjusted), then the simulations say that about
5% of the time your portfolio doesn't survive 30 years.
2007 was not the worst time to start retirement. The S&P
500 returned 3.
53% that year, and QQQ returned 18.7%.
See graph. If one is looking for a poorly performing data set, one would be better off excluding 2007 and starting with 2008.
Running the model with data from 2008 through 2019,
5% of the time the portfolio doesn't survive. Add in the requirement that the first year in any simulated run is the worst, and 6% or so of runs don't survive. Not a big difference.
For kicks, I ran QQQ, 4% starting withdrawal (inflation adjusted) for the lost decade (2000-2009). Less than 1 in
5 survive for 20 years, barely
5% survive 30. Expand the data set to the past 20 years (2000-2019), and about 7 in 10 survive 20 years; a bit over half survive 30. That doesn't include starting each run with the worst year which would make things worse.
Not comforting figures. Forget about getting the original investment back (perpetual withdrawal rate). According to the model, assuming returns over the next 20 years are like the past 20, there's a good chance that the money won't even last at all.
OTOH, with PRWCX, based on returns over the past 20 years, starting with a 4% initial withdrawal amount (inflation adjusted),
and requiring the worst year to come first, one may have a 98% chance of surviving 30 years.