Pdxcess wrote: Thu Sep 04, 2025 12:56 am
I still contend, however, that the “Optimize Roth Conversions” feature in Pralana does not isolate the tax arbitrage benefits of Roth conversions,
Can you identify an incorrect calculation it is doing and describe what it should be calculating instead?
and it can produce some whacky, albeit algorithmically correct, results.
Whackiness is subjective. Using a correct algorithm and implementing such algorithm correctly should be objectively testable (thus the question above).
In my case those whacky results are a recommended conversion to the top of the 35% bracket in Year 1, with the result being $385k of conversions being taxed at a fed+ state + local rate of nearly 50%, even though the input effective tax rate for my heirs is only 25%.
Admittedly “unexpected” – but is it incorrect? Consider carefully how “Traditional plus taxable” vs. Roth math works before answering….
And this is for an increase in terminal net worth of less than one half of one percent compared to the result if I never convert above the 32% bracket.
Thus the software is behaving the way any self-respecting constrained optimization package should behave: finding the best results within the allowed space.
My limited point here is that you should not simply rely on the “optimal” conversion feature: you should play around with multiple conversion scenarios within any given set of inputs.
Agreed – notwithstanding the comments above.
Even assuming perfect calculations by Pralana (or, for that matter, any similar tool), that one half of one percent improvement is probably well within the uncertainty surrounding all the assumptions that went into the calculations. Precision is not the same as accuracy.
At least two schools of thought tend to run through threads like this:
– One suggests learning how to use complex models, perhaps with a wide variety of values for those important inputs (investment return, life expectancy, tax law, type of heirs, etc.) and picking a strategy from among the results obtained from all those input values.
– Another suggests taking a single “best guess” set of assumptions, using those to estimate future marginal tax rates, and then converting as much (or as little) to keep the current year’s marginal tax rate about the same as the projected future one.
Both schools suggest revisiting the issue more or less annually. One can argue pros/cons for both, and in the end it’s probably an “eye of the beholder” thing, not unlike “pay down debt vs. invest” arguments in which each side says “why would you ever do that instead of this?”
Having spent a good part of a working career in complex modeling (not the personal finance field) I can see the attraction for those so inclined toward that approach here. On balance, however, and particularly with that “flat optimum” (0.5% difference for very different inputs) in mind, it seems the simpler approach would be more accessible to a wider population. As the saying goes, YMMV.
