Using FRP During Retirement

[rpstillman]

Hi Jim,

I do have a follow up question:

FRP creates a probability of success (having a positive bank balance at plan's end) based on the worst case analysis - that is, the the plan balance in the Bottom 10% column. Because you're creating that bottom 10% each year, that means every year for, let's say 30 years, the plan's return is essentially in the bottom 10%.

To me, the odds of that happening are exceedingly low. What are the odds of the market return being in the tank every year for that long? It's never happened.

I assume I'm missing something, and appreciate your enlightening me.

Thanks much,

Rich

[ochotona]

FRP creates a probability of success (having a positive bank balance at plan's end) based on the worst case analysis - that is, the the plan balance in the Bottom 10% column. Because you're creating that bottom 10% each year, that means every year for, let's say 30 years, the plan's return is essentially in the bottom 10%.

To me, the odds of that happening are exceedingly low. What are the odds of the market return being in the tank every year for that long? It's never happened.

Rich, of course I'm not the expert about FRP, but I do know about Monte Carlo simulation generally. I think you're maybe not understanding how to interpret the outputs.

The bottom 10% estimate of a portfolio balance doesn't result from the worst-case analysis. That's not what it means. It means that if there are 1000 simulation runs created by the simulator, it means that 10% of them, or 100, will have a plan end portfolio balance below the bottom 10% value. That isn't worst-case, that happens 10% of the time. It's frequent.

Conversely, it means that 90% of the time the end balance will be above that. 90% of the time sounds secure... but if only 90% of planes flying in the US everyday landed successfully, air travel would stop really quickly, for obvious reasons.

Are you planning for a target end of plan balance, or do you plan for the probability of meeting your spending needs?

Personally, I like to see a 95% probability of meeting my income needs with the Stable income plan, knowing that if things are bad I could flip to the Conservative plan.

[rpstillman]

Hi Ochotona,

Thanks very much for your reply. I was using hyperbolic language so let me clarify my question. You correctly state that my description of the 10th percentile as worst case is wrong.

However, my question still stands. The probability produced by the model is predicated on a return that is at the 10th percentile EVERY YEAR. So, something that only occurs 10% of the time would occur every year for the life of the plan ( in my case, 30 years). What are the odds that something that occurs only 10% of time would happen every time, 30 times in a row?

Does that help clarify my question?

Also, like you I've studied Monte Carlo simulations a lot, and (just my two cents) your 95% goal at a steady spending rate is extremely conservative. Current research is pointing to probability in the low 80's, or 75% will produce a successful outcome. My investment manager (running his firm's proprietary model) reassures me on this point constantly when I get nervous. In my plan for instance, I have a 100% success rate until the final three years, and then end up at 82% in the final (30th year).

And again, if I have a 90% chance of returning more than the amount in the 10th percentile column at least some of the years in the first 27 years, what are the odds that there will be enough money to see me through the final 3?

I appreciate your thoughts -- look forward to hearing back.

[ochotona]

However, my question still stands. The probability produced by the model is predicated on a return that is at the 10th percentile EVERY YEAR. So, something that only occurs 10% of the time would occur every year for the life of the plan ( in my case, 30 years). What are the odds that something that occurs only 10% of time would happen every time, 30 times in a row?

That is a really good question. It gets to the heart of whether it is reasonable to model a portfolio using what is called a "normal" or "Gaussian" or bell-shaped curve probability distribution, with each year being an independent experiment, just like flipping a coin.

Clearly, that's not how markets behave. After a number of bad years, you run out of bad years to have... from 1929 to 1933 the US stock market just wasn't going to go any lower. Similarly on the upside... trees don't grow into outer space. You can have a blow-off exuberant top, and then you run out of good years for a while.

I do not know if FRV does something sophisticated like considering the performance of a block of years, rather than one year at a time. I think it's called "block bootstrapping". Some simulators do sequence of returns stress testing, where you front end load the bad years. You could argue with the market where it is, it's more likely we have a set of bad years in the near future, which is the opposite of 2009.

I think you're just up against the limitation of simple tools against a phenomenon (global financial markets) which are not at all simple and won't ever be. None of us can afford a supercomputer like a weather or climate forecasting machine every time we want to look at our retirement.

But I would take heart, we know that the worst cohort for retirement to belong to in history was the Class of 1966. The Monte Carlo planning would still have been of value for that cohort. And cutting back the spending budget after bad market years would help even more.

I think it's a science and an art. If we have some bad years, I think we will all reflexively cut back spending (I will), and I think it would be useful to re-run the planner after the future bear market is over and get a new view on a budget.