Quote:
Originally Posted by kramer
. . . An 80% probability means that in 80% of the cases the portfolio would survive to the end of my life.

Well, yeah  that much is pretty clear. Thanks for replying to my question, by the way. As I said, my questions were more about what does it mean from a
prescriptive point of view. If the simulation indicates your current lifestyle and portfolio provides an 80% probability of success, is that good enough? Or do you still worry? Would 85% be good enough to reduce that worry? Or do you need 95%? . . . These questions have to be dealt with whether the simulation is historical or monte carlo, but they are more worrisome for monte carlo simulations since monte carlo simulations cannot address the correlations between the various variables (stock returns, bond returns, inflation, yeartoyear, . . .). Adding a longevity distribution to the simulations will tend to spread the distributions out even further. If 85% probability makes you feel good with a 40 year longevity simulation, how does that relate to the probability calculated with a longevity distribution added? . . . There is some probability that you die tomorrow. When the simulator chooses that longevity solution, any withdrawal rate is probably acceptable. But there is some age (110?, 115?) where it is probably safe to assume you will not live beyond. Does this fact tend to increase probability of success? Of does the long life end of the distribution tend to reduce the success rate?
Quote:
Except for the well known shortcomings of Monte Carlo and its limitations in regards to financial modeling, I see no reason why this approach is invalid.

I don't question the validity. It is a valid way to estimate how many people of a given age, with a given portfolio and a given spending plan will live in retirement without sacrifice or bankruptsy. What I don't understand is how to convert the results into greater understanding or prescriptive actions for me.
Quote:
I model the life expectancy by an average and a standard deviation.

There is a problem with using a normal distribution. At any moment in life, there is a finite probability that our remaining life is measured in seconds  even if you are only 18 years old. But there is really no chance that you live to 150, for example. Yet a normal distribution will predict some real probability that you live to 594 years . . . No matter how old you are or how much money you have accumulated, you probably won't survive a 500+ year retirement.
Quote:
Yes, running the simulations as a function of longevity might give interesting results. Perhaps I should try that (and thanks for the suggestion). Although the number that I am really after is the SWR for my life expectancy based on the information that I have available now.

It would really be interesting to compare the two sets of results. Take the default unit retiree  30 years retirement, $1M, 50/50 stock/bond split, 4% withdrawal rate. Run a simulation with your longevity distribution. Then run one that forces the longevity standard deviation to zero and compare. Then increase or decrease the longevity assumption ( using zero standard deviation) until you match the probability of success for your longevity inclusion. The results might tell us something about how much we should weigh longevity risk vs withdrawal rate risk in our plans.