RaptorD2 wrote:
AAII Computerized Investing has decent article with a spreadsheet for calculating a safe withdrawal rate.
Just a word of caution here. The AAII spreadsheet is, like all such spreadsheets, actually pretty dangerous if you expect to do the SWR thing. For instance, if you have $1Million, start taking out $60K, adjusted for inflation, and assume you make 8% per year with inflation at 3.5%. The math will work out that this will be fine for way more than 30 years.
In reality, however, your return won’t be 8% per year each year, nor will inflation be 3.5% each year. Some years you may lose money. Some years inflation may be sky high. A Black Swan event will probably occur during the 30 years you need this portfolio to last. Any of these can drain a portfolio to a point from which it won’t recover.
rbgibbons wrote:
…That number was calculated using Monte Carlo simulations, and what it really means is that if you only spend that much, there’s a 95% chance (or something like that) that you won’t run out of money before you die.
So, yeah, one way some people went about trying to be more realistic was to introduce randomness into the evaluation. That introduces some variability based on probability, but there are variables like the distribution curves (Gaussian or log or binomial, etc.) that have a big effect on the outcomes. And then this simulation only adds randomness to the values that still average out, but it doesn’t introduce Black Swan type events where more many things are adversely affected at the same time, and do so because of the way things work. Monte Carlo simulations help with volatility, but not with how various things move in tandem.
This article attacks Monte Carlo simulations pretty brutally: http://retirementoptimizer.com/articles/MCArticle.pdf
This article defends Monte Carlo in theory, but admits in practice the tools that actually employ it are lacking: https://www.advisorperspectives.com/articles/2014/08/26/the-…
So, what to do? One proposed solution is to back test your plan against what has really happened in the last 100+ years. Yes, we all agree the past doesn’t predict the future, but testing your plan against how real markets and the correlation between the behavior of stocks, conventional bonds, inflation-indexed bonds and cash have occurred is useful. The idea is to see if your plan would have survived starting at any of the 100+ years for which we have data.
You can check out one such back testing tool at http://www.retirementoptimizer.com/ (or search for Otar Retirement Calculator). There is a free trial version you can try out, and there are some other free resources like a Cash Flow tabulation sheet. The website is geared towards people who advise others on retirement, not for do it yourselfers like us. I have no relationship with the website or its owner and am only just now playing with the trial version. Note that the outcome of this Calculator is not a pass/fail, but it gives you 100 outcomes all graphed on top of each other, with the 90 percentile outcome highlighted. You could, of course, choose to be more conservative and look for a higher percentage of successes, or be more aggressive and go the other way.
This includes things like some combination of equities and bonds and cash as your portfolio. While this is far better than choosing an “average” yearly return and applying some randomness to it, it’s still not the kind of portfolio we here on Saul’s board run for ourselves. So, the best it could give you would be what a portfolio based on the S&P 500 would do for certain withdrawal rates (with optional limits or stop losses) over the past 100+ years. Of course, it’s possible that in our lifetimes markets get whacked in ways not seen in the century. I don’t know how anyone can prepare for that.
If you want, there is a whole website discussion forum for people doing retirement planning for a living at https://www.bogleheads.org. Jim Otar’s book is available from Amazon (https://www.amazon.com/Unveiling-Retirement-Myth-Jim-Otar/dp…) for in PDF form for under $10 from his website.