Postscript on my spending variance post...

On Dec 30 I posted about spending variance in "More Than I (Or You) Ever Wanted To Know About Spending Variance Vs. Fail Rates."  The conclusion for generic assumptions was that spending variance has almost no impact and that for an unholy mix of assumptions slightly closer to my own there might be some effects but pretty small.

I went back and used assumptions for me as close as I could and then added a new function that creates a custom skewed distribution for spending that looks as close as I could get to what little I know about my own spending variance over the last seven years. When I re-run a sim with and without the custom spending variance (keep in mind it still varies by random inflation either way its just that I add an extra mini-variance "shock" each year in one case) it looks like this (I hate too much personal data out in public so I deleted out the actual fail rates but I'm somewhere less that 10%...down from 80-100% in 2010):

Conclusions:

• If my longevity expectation were to be the median age of the mortality tables (let's say 81) then the difference is more or less meaningless
  
• If, on the other hand, my longevity expectation were to be later, like 95 (the third quartile expectation if I survive to age 85 so reasonable) or 105 (or about as far as I want to go planning-wise) then it probably is material.  The difference is more than 10%
  
• If my longevity expectation were to be 85 or 87 (closer to mode-levels for a lot of retirement start ages) and I were to aggressively retire early then who knows how big the effect is but one can assume there is an effect due to the length of time involved
  
• The main conclusion of the parent article was that spending process control for early retirees is probably wise no matter what. That conclusion still stands
  
• The first three conclusions mostly apply to me. I don't think they are all that universalizable at this point.