The problem is not upfront taxes paid, but an apples to oranges assumption you are making about asset allocation and growth.
I didn't make any assumption about asset allocation and growth. You established the parameters of the example and the two cases, with a completely unspecified set of background circumstances. I simply clarified the background circumstances, which changed the conclusion entirely. With that clarified, it seems you want to change the original parameters. OK... Yes, when all the growth assumptions are aligned, there is a breakeven point at age 87 (not 85). Either way, I'm sure that sends a warm and fuzzy message to Roth converters everywhere.
On to more interesting things...
Yes, please do. Asset allocation drift is undoubtedly the primary reason, and perhaps the sole reason, that your spreadsheet has convinced you that Roth IRA conversions will never pay off, even at a favorable tax rate.
On the one hand, although his spreadsheet appears to be doing the calculations correctly, it seems clear that his methodology is quite suspect. It seems to share with the I-ORP calculator a tendency to produce long term asset allocation shifts as a result of the Roth conversions that bias the result either for or against Roth conversions.
You are quite presumptuous for a person who has never actually peeked under the hood of my spreadsheets.
First, I have not studied i-orp to the degree that you have, so I cannot comment on its tendencies at that level of granularity. My own spreadsheet model assumes a static AA and associated growth rate for each account class: taxable, tax-deferred, Roth. The spreadsheet also calculates AA at the total portfolio level at every age to ensure there is no unintended drift resulting from these static assumptions. Ordinarily, with such a static model, one would expect some drift toward higher equities in the overall AA since the taxable account (mostly equities) has a higher growth assumption than tax-deferred (50/50). But, in my base case, this drift effect is almost perfectly compensated by withdrawals from the taxable account prior to age 70. After 70, the overall AA drifts modestly heavier toward equities. I have purposefully not compensated for that, as SS would join the mix at around the same time, which, as a fixed-income equivalent, balances the overall AA picture back to my target.
In any case, that's the base case, and I'm comfortable with it. Your question is... What happens when I make Roth conversions? You suggest my methodology is suspect because Roth conversions produce long-term AA shifts that bias the conclusion. It has always been my assumption that Roth conversions would come out of the stock allocation in the tax-deferred accounts, and go into stock in the Roth account. Thus the Roth would be 100% stock while tax-deferred would gradually drift heavier toward fixed income during the conversion period. The taxable account would remain all equities, so again, the overall portfolio AA would be unchanged throughout the conversion process. Just a shift of stock holdings from tax-deferred to Roth.
Now, as Roth conversions occur, the static growth assumptions for taxable and Roth are not an issue, since they don't change with conversions (all equity). However, the model cannot dynamically change AA and associated growth rate each year for the tax-deferred account, which is the only account that changes AA during the conversion period. So, for the conversion scenario, my solution was to assign a growth rate to the tax-deferred account based on the average AA over the course of these periods, as follows:
41/59 average AA from 53-70 (starts at 50/50, ends at 33/67)
33/67 AA from age 70 onward
37/63 overall average
Using the average AA from 53-70 seemed reasonable since the conversions occur in a fairly linear profile. Using the average of that period (41/59) plus 33/67 for the period after 70 is accurate for death at 87. Since I had to make an assumption, that sounded as good as any. For death prior to 87, that average is slightly understated, and for death after 87, it's slightly overstated. I also tested the model's sensitivity to this assumption. In order to change the conclusion in favor of converting, the tax-deferred AA would need to be higher than 42% equity, which seems unreasonable, given the profile above, which is fixed at 33% from age 70 on.
Sorry for the long explanation. But I made every attempt possible within the constraints of the model to avoid bias one way or the other. The overall AA is constant across time in both scenarios, and it's consistent between the two scenarios. I welcome suggestions to improve this.