Hi Tom,

I think what you are seeing, when using MC along with being constrained in regular assets and "forcing" special withdrawals, is a magnification effect playing out over 500 MC runs. I'll try to explain what I think is happening.

Depending on how close you are to being constrained, the variation in MC returns will lead to at least a few runs hitting negative assets. The earlier this happens (such as a bad sequence of returns in the next few years) and the larger the negative returns, causes the worst results.

Since multiple factors come into play, it's hard to tell for certain. But...there are variations in MC return randomness, return sequencing, amounts of assets in regular assets in different years, different special withdrawals in certain years, etc.

ESPlanner's MC calcs should take the special withdrawals exactly as you input. Normally, the smoothing algorithm adjusts these as needed. Since you are "forcing" the program to withdraw various amounts of money over ~10 years, you can get in trouble at any number of points in the MC calculations. My guess is that this "forcing" of withdrawals is the key factor. Your results seem (to me at least) to show the interplay of these constraints and MC return randomness during the first 10-15 years or so.

My guess is that over 500 runs, during the years when you have special withdrawals, the interaction of multiple factors causes the wide dispersion that you see. As you describe, once you get past this "Special WD era" (SWD), everything seems to start behaving "normally". Basically, the smoothing algorithm has an easier time once you get past the constrained years, but is forced to contend with the combination of constraints and randomness which causes the more chaotic results early on.

You can think of the interim MC report values at the end of 2024 (SWD era) as being the initial conditions for a much less constrained era until end of life. Once you get past 2024 and stop the special withdrawals along with reduced constraints in regular assets due to SS and no mortgage, the wide initial spread of results (and associated asset values) starts to get back to "normal" by late in life.

Regarding John's case and the "fake annuity" method, it's harder for me to speculate since he says he talked to Larry about "using too many years of Special Withdrawals can have odd effects because of the way they are handled during the Monte Carlo runs". Perhaps this is just what we are talking about? My guess is the "forcing" of the program to do special withdrawals is the issue here as well, but that's just speculation. Given my experience with MC, I'm not shocked at all with the wide variation in his charts.

Anyway...I don't know that we've resolved anything, but the speculation seems to be somewhat reasonable that the program's results are coping with constraints and MC randomness. But...perhaps this isn't 100% accurate given John's conversation with Larry about his profile?

If the ESPlanner team can add/clarify anything on this, that would be great.

One suggestion that I've used to cope with the uncertainty (without MC), is to develop a "range of reasonable estimates" using some "what if" assumptions. This way you can use conventional planning and still model low returns, losing $XX assets, high inflation, tax increases, Social Security cuts individually or everything in combination. At the same time I use a "safety factor" approach based on the “Standard of Living” (SOL) index in the Assumptions tab. To do this, create a copy of your profile and set your last few years to a high SOL. That will force the program to reduce consumption in earlier years and preserves assets for late in life. The approach is very flexible and is sort of a general purpose risk cushion that can be used to essentially self-insure to some extent against various risks.

Best,
Brian