Author Topic: Retirement Planning Simulators and Spending Models  (Read 1440 times)

hugin7

  • 5 O'Clock Shadow
  • *
  • Posts: 2
Retirement Planning Simulators and Spending Models
« on: May 18, 2017, 05:10:55 PM »
I investigated a number of retirement planning simulators, including FIRECalc, cFIREsim, and Vanguard's retirement nest egg calculator, and ran into discrepancies and limitations with each of them.  Can someone please recommend a calculator, assuming one exists, that supports the following spending model?

I am looking for a hybrid of "inflation-adjusted" and "% of remaining portfolio" models, both of which are commonly available in online calculators.  If portfolio returns for current year < inflation for current year, use the "% of remaining portfolio" model.  Else, if returns > inflation, limit your spending to (portfolio value + inflation) * % of remaining portfolio, rather than (portfolio value + returns) * percentage of remaining portfolio.  More succinctly, spending = floor(value from "% of remaining portfolio model", value from "inflation-adjusted" model).

I am not sure if any studies have been done on the model I describe above, but I suspect that it should allow for considerably higher average annual spending than either of the two popular models that it is based on.  During years in which your portfolio drops, one can cut spending to avoid early depletion, as with the "% of remaining portfolio" model, to avoid early depletion, while during good years when returns > inflation, one can limit their "pay raise" to a cost of living adjustment, which should mitigate the risk of poor future returns.

In addition, I have a couple of specific questions/observations regarding online calculators that I would appreciate feedback on:

1.  Does the "% of Portfolio" spending plan work for anyone in cFIREsim?  Even when I enter a ridiculous yearly spending amount, such as 20%, I still get a 100% success rate result... wouldn't that be nice?
2.  Both FIRECalc and cFIREsim yield much better results than Vanguard, often by a factor of 50-100%, despite similar input settings.  Based on my understanding of their methodologies, FIRECalc and cFIREsim limit their analysis to actual historical sliding time windows of, say, 50 years, while Vanguard randomly picks a string of return/inflation values for specific years from their database.  The latter approach results in some economic cycles that are far less favorable than anything ever experienced during our history (20-year Great Depression, anyone?), which yields a rather grim bottom line compared to what's produced by FIRECalc and cFIREsim. 

Am I correctly interpreting the differences among these tools?  Also, is the general opinion that Vanguard's tool is too conservative or are the other calculators too aggressive?  As a point of reference, in order to achieve a 100% success rate, Vanguard recommends a meager withdrawal rate of 2.2% for a 50-year retirement, while, FIRECalc's rate is a far more palatable 3.77%.

secondcor521

  • Handlebar Stache
  • *****
  • Posts: 2056
  • Age: 50
  • Location: Boise, Idaho
  • Big cattle, no hat.
    • Age of Eon - Overwatch player videos
Re: Retirement Planning Simulators and Spending Models
« Reply #1 on: May 18, 2017, 06:43:42 PM »
I can't recommend a calculator to meet your needs because I doubt whether one exists.

Regarding your point 1, the percentage of remaining portfolio will always result in a success rate of 100%.  However, the withdrawals can get very small:  Suppose you start with $1M, taking 20%, and the market gets cut in half each year.  You take $20K the first year, then $10K the second year, then $5K the third year, $2.5K the fourth year...etc.  You're still taking 20% of your remaining portfolio, but your withdrawal amounts have been decimated.  Eventually you are taking a withdrawal of 20% of $1, meaning your withdrawal doesn't even fund a modern day gumball machine.  But you technically haven't run out of money, because you still have 80 cents left.

Regarding point 2, it sounds like Vanguard uses what's called Monte Carlo simulation, where it picks numbers randomly out of a range.  And yes, historical analyses like FIRECalc and cFIREsim will produce almost twice the annual withdrawal rate versus a Monte Carlo simulation.  As you note, Monte Carlo simulations can come up with situations that are far worse than what has happened so far, which is why they are more conservative.

Whether one is too aggressive or too conservative really is a reflection of your opinion about where we are in history, your acceptance of the premises of each of the tools, and your opinion about the future.  As well as how much you like or hate your job, what options you have, how old you are, your risk tolerance, and a lot of other variables.  There are people who are retired with negative withdrawal rates (that is, they don't even spend their portfolio incomes) and people who are withdrawing at rates as high as 6.5% and possibly higher.  In the end, "you pays your money and takes your chances".

Good luck!

hugin7

  • 5 O'Clock Shadow
  • *
  • Posts: 2
Re: Retirement Planning Simulators and Spending Models
« Reply #2 on: May 20, 2017, 09:15:13 AM »
Thank you for your feedback.  Further investigation revealed that cFIREsim's "% of remaining portfolio" floor and ceiling parameters, defined as constants, are actually inflation-adjusted, so setting ceiling = starting withdrawal amount does exactly what I want.

As for the guaranteed 100% success rate of the "% of remaining portfolio" model, you are absolutely right - as retirement duration approaches infinity, your nest egg converges to 0, but never truly runs out.  IMO, cFIREsim should dynamically disable the UI option to investigate the maximum starting withdrawal rate for this withdrawal model, since its results are misleading.