Question for dory36: how do you explain that intercst got even lower figures than I did. If we are both using the same input, why should the results not be exactly the same?
There is still my original question of why there is any difference between FIRECalc and spread sheet results for the same inputs?
I can't answer for Intercst - but he is the author of the spreadsheet, so he certainly knows how it works.
Now, about why they differ:
They both used the same data when written, but they aren't the same program. In writing Firecalc, I worked from scratch in developing the algorithms, rather than trying to reverse-engineer Intercst's spreadsheet. There are probably a few differences in such things as when the expenses are deducted from the portfolio (before or after inflation is applied that year?), for example.
In addition, there were some deliberate differences, such as applying the first withdrawal immediately upon retirement rather than 364 days later. (I was looking at survival of a portfolio when SEPPs were being used, and planned an immediate withdrawal.) Also, I felt that the bad years of the late 60s were being left out of too many analyses for the 40 year runs many people were doing, so my analysis includes partial cycles, such as 1969-2001, when Intercst's spreadsheet stopped the analysis at 1961 for the 40 year runs. This lets me catch more failures for those years, but also changes the numbers used in the success rate equation.
When Firecalc was written, it was pretty thoroughly tested against the spreadsheet by lots of folks. Intercst researched the differences, determined they were insignificant, and posted his analysis at http://rehphome.tripod.com/captbill.html
. (He posted this on May 1, 2001, the day I joined the ranks of the early retired!
If you continue to get significant
differences, I would certainly suggest you try to see why. At a minimum, see if you are getting the same results as Intercst shows on the page describing the spreadsheet, for example. You can also run the detailed results option on Firecalc, and then look year by year at the outcomes for each. (We did
find a typo a couple of years ago this way, when one user saw a divergence in one particular run.)
If you are getting differences that aren't too big, ignore them. No sense measuring with a micrometer when you have to cut with a chainsaw!