Using the planner to decide on taking social security early

[frp user]

Dear Jim--

Just wanted to thank you for your new Flexible Retirement Planner, which I think is a great tool for those of us doing some retirement planning. I've been using it to try to decide whether to take social security early (age 62) or late (age 70). I've examined this issue with deterministic software (Retirement Savings Planner 2007), and with four Monte Carlo Simulators. Without burdening you with details:

The deterministic software strongly suggested taking social security early.

Two of the Monte Carlo simulators (FIRECALC and Netirement) said it was a wash.

Your planner and that on the Schwab website said the probability of success improved by about 4% if social security were taken late at age 70.

I was puzzled by the qualitative disagreement between the deterministic and stochastic planners, and re-ran the simulations with Felexible Retirement Planner, this time setting the standard deviation equal to zero and otherwise changing some of the settings as suggested in your FAQs section. This time, Flexible Retirement Planner agreed with the deterministic software Retirement Savings Planner 2007, suggesting it was better to take social security early.

I realize I haven't provided you with any of the input parameters, but I have several questions for you:

1. Do you feel that this is an appropriate question to address in this way?
2. Are you surprised (as I was) by the apparent qualitative disagreement between the deterministic and probabilistic approaches?
3. Would you tend to attach greater significance to the Monte Carlo results than to the deterministic results?

With thanks once again,

Sincerely,

[admin]

To concisely answer your questions:

1) Yes, I think this is a reasonable approach. That being said, it may be that the difference between your two choices is smaller than the "error signal" of the tools you're using. All of these tools are making guesses of the future and in the grand scheme of things, a 4% difference in probability is very small. William Bernstein mentions in his famous "Retirement Calculator from Hell" article that above an 80% chance of success, incremental differences in probability are likely meaningless.

http://www.efficientfrontier.com/ef/901/hell3.htm

2) I'm not very surprised by the differences you're seeing for reasons described below.

3) I think I'd attach a roughly equal significance to the Monte Carlo and historical based tools, and a lower significance to the deterministic tools, for reasons described below.


Your analysis sounds pretty thorough to me, but I do have a few additional comments.

1) FireCalc isn't Monte Carlo based. It's a great tool, but it runs a backward looking simulation using historical market returns rather than a Monte Carlo based simulation. This is a bit of a nick-pick, but there's a significant difference in the methodology. I can't remember the exact details, but basically, it assumes you retired in each of the last 80 or so years, then runs though your retirement results using the actual sequence of returns you would have experienced had you retired in that year. That said, your approach of running multiple tools with different methodologies sounds exactly right to me. Get as many answers as you can, then weigh the differences between them.

2) I'm not a big fan of tools that use a purely deterministic approach for this type of planning because the volatility of returns isn't taken into account. So that means I would tend to weigh the results of a stochastic or historical tool above the results of a deterministic tool.

3) You might want to try running flexibleRetirementPlanner with a non-zero std deviation, but set the spending policy to "Stable". This should make the tool behave like other Monte Carlo planners. Also, it's important to know what return/std. deviation parameters the other tools are using for comparison purposes. Finally, there may be differences in the handling of taxes by the different tools.

4) You might be interested in the option that exists to "pay back" your social security benefits after you've started receiving them and thus undo a decision to start benefits at age 62. Here's the link:
http://www.retireearlyhomepage.com/cheap_annuity.html

Let me know if you have any other questions, and good luck with your planning.

Regards,

Jim

[frp user]

Thanks much for your very prompt and insightful reply.

I took your advice, and varied the spending policy --tried all possible configurations. It made a huge difference, as you no doubt are well aware! Of particular interest was what happened when it was set to "stable": the 4% advantage of taking social security at age 70 as opposed to 62 all but disappeared ( I don't think Bill Bernstein would be impressed by a 1% difference <grin>)....

I'm glad you feel that the deterministic solutions are less significant than the stochastic (if they weren't, one would wonder why bother <grin>).

Once again, many thanks for taking the time to set up this impressive piece of freeware, and also for your advice.

Best regards,