I am wondering if Pralana's Roth Conversion and Withdrawal Priority optimizers are valid with the Actuarial and with the Consumption Smoothed withdrawal methods. Given that both aim for zero (or legacy) as the final effective savings. I'm not sure they do.
With the Roth Optimizer, the manual tells us,
A 3-pass algorithm is used to determine the optimum marginal brackets for every year in the conversion period. In the first pass, it uses the same bracket in every year and determines which bracket yields the highest effective savings at the end of the modeling period. In the second pass, it applies the best overall bracket established in the first pass to every year in the modeling period and then seeks improvements on a year-by-year basis. It starts with year 1 and tests all marginal brackets for a possible improvement in the final savings balance while all future years are using the best overall bracket established in the first pass. It then modifies the bracket for year 1 with the bracket yielding the best long-term result. Then it tests all marginal brackets in year 2 for a possible improvement in the final savings balance while all future years are using the best overall bracket established in the first pass... . [my emphasis]
With the Withdrawal Priority Optimizer, the manual explains,
For each optimization time period, OWP will calculate the scenario’s final effective savings using 24 permutations of withdrawal order, plus the Proportional withdrawal option, from these four accounts: taxable investment, your and your spouse’s tax-deferred, and Roth. [my emphasis]
Since these two lifecycle methods aim for zero (or legacy) final savings balance, the optimizer is left chasing its tail, preferring a plan with a few dollars more rounding error. This doesn't seem valid. Rather, Actuarial and Smoothed Consumption should be optimizing based on having a higher total consumption during the life of the plan (=total spending + legacy amount, or =total discretionary spending + legacy)--not the end of plan value.
I'm pretty sure the the roth optimizer doesn't work with actuarial withdrawal method. I am not positive about consumption smoothing. My fear is that it wouldn't be easy to fix things because it might require a more complicated (resources heavy) iteration than the current routine. At a minimum the program should warn users or not allow optmization.
One possible "work around" with Consumption Smoothing is to enter into the Build sections your essential spending, but leave your discretionary "Phased Expenses" at 0. Then run "Consumption Smoothing". Copy things over to Scenerio 2, enter that figure into Build> Phases Expenses (non-essential), and run the Roth Optimizer and Withdrawal Optimizer, and then Run Consumption Smoothing. If this consumption smoothed amount for Scenario 2 is higher than Scenario 1, then Roth conversions make sense. If not, they don't. (I haven't tested this, I am just speculating.)
I am not sure about how to fix withdrawal priority optimization with Actuarial and Consumption Smoothing, or even if it needs it.
Thoughts?
Adding @ricke and @bhines, @golich428 to see if I'm on the right track.
The Roth Optimizer and the Consumption Smoother have to be run iteratively, the Roth Optimizer does not re-smooth and the Consumption Smoother does not re-check the Roth plan. So run Consumption Smoothing first as that sets the general plan to run the accounts down, see what your allowed spending is. Then run the Roth Optimizer, it will try to increase the end of plan value. Then re-run Consumption smoothing and it will spend whatever the Roth plan came up with. While this converges pretty fast, I would repeat the back and forth the cycle at least one at more time. Once there is no further improvement, that's the answer.
Actuarial does not require iteration because it withdraws a fixed percentage of accounts based on the life expectancies you entered. However, it is harder to judge the results as not only does the non-essential spending vary a bit each year due to the method, but also, the early drawdown in the account balance due to pre-paying taxes causes the method to reduce your early spend rate. When the tax savings from Roth Conversions eventually manifest themselves, then the method will ease up on th spending reins, so you can really party when you're 94. That's inherent in the method, not really a Pralana problem, but you are right that it isn't a very useful analysis for Roth Conversions.
If you have no legacy concerns, then I would use Consumption Smoothing as the way to judge Roth Conversions, Actuarial doesn't play well with the delayed payoff that Roth Conversions give.
If you want to leave a legacy but want smooth non-essential spending, you may have to go old-school. Start with Actuarial to get a range of spending that hits the target legacy, and then go back to a fixed expenses only spending plan. Enter that non-essential spend rate you found in the Actuarial analysis as an expense stream. You may have to vary the non-essential spend somewhat to make the legacy amount reach your target. Only then would you run the Roth Optimizer. You may then have to repeat the exercise of varying the non-essential spend rate to get back to your desired legacy amount, repeating the whole procedure until the benefit of the Roth on spending becomes clear.
Rick, thanks for validating my thoughts that the optimizers work for Consumption Smoothing but not its cousin the Actuarial withdrawal method.
run Consumption Smoothing first as that sets the general plan to run the accounts down, see what your allowed spending is. Then run the Roth Optimizer, it will try to increase the end of plan value. Then re-run Consumption smoothing and it will spend whatever the Roth plan came up with. While this converges pretty fast, I would repeat the back and forth the cycle at least one at more time. Once there is no further improvement, that's the answer.
This is exactly the process I described in my post which I've been doing. (Call it a "Maxifi Planner Imitator".) The assumption, of course, is that optimizing for best end-of-life plan (ie what the Roth optimizer does) will translate in the end into best consumption-during-life also. (That is, smoothing the end benefit across time doesn't negate its creation.) That assumption is not strictly true, since the extra consumption smoothed income will affect the headspace for Roth and taxes--but iteration even a couple times should make it "good enough". For quick decisions one might even get away with just looking at the red and blue lines without consumption smoothing; with less confidence if all the conversion benefit happens only at the end. And when you needed to get into the weeds you could iterate things to find your base discretionary consumption generated by the Roth vs not or Withdrawal account priority x vs y. You could even see the "cost" of consumption smoothing by comparing the iterative result vs not smoothing at all.
I hope this post is useful to the group of us who value Lifetime Consumption in addition to Legacy. Pralana's consumption smoothing is an underappreciated power-tool, with multiple uses. (To leverage its power, I've requested additions columns on the Monte Carlo data table results next to the charts that show total average plan consumption as well as the end-of-life-values which are currently shown.)
Actuarial does not require iteration because it withdraws a fixed percentage of accounts based on the life expectancies you entered. However, it is harder to judge the results as not only does the non-essential spending vary a bit each year due to the method, but also, the early drawdown in the account balance due to pre-paying taxes causes the method to reduce your early spend rate. When the tax savings from Roth Conversions eventually manifest themselves, then the method will ease up on the spending reins, so you can really party when you're 94. That's inherent in the method, not really a Pralana problem, but you are right that it isn't a very useful analysis for Roth Conversions. If you have no legacy concerns, then I would use Consumption Smoothing as the way to judge Roth Conversions, Actuarial doesn't play well with the delayed payoff that Roth Conversions give.
It sounds like you agree with me that the Roth optimizer doesn't work well with the Actuarial Method. I hadn't thought about the pre-paying of taxes, but that just adds to the misfit between the two. I was thinking more of the mechanics/workflow of how the two work: If you try actually interating the Roth optimizer with Actuarial withdrawal you'll see what I mean about "chasing its tail" because you end up with the optimizer recommending conversions that give you like $7.00 advantage over no conversions, since the actuarial keeps working its way down to zero at the end, the opposite of the optimizers maximum at the end. Iteration thus doesn't seem to be valid even aside from the theoretical issues.
I am still not sure if the same reasoning affects the Withdrawal Order Priority optimizer also; need to do more tests.
If you want to leave a legacy but want smooth non-essential spending, you may have to go old-school. Start with Actuarial to get a range of spending that hits the target legacy, and then go back to a fixed expenses only spending plan. Enter that non-essential spend rate you found in the Actuarial analysis as an expense stream. You may have to vary the non-essential spend somewhat to make the legacy amount reach your target. Only then would you run the Roth Optimizer. You may then have to repeat the exercise of varying the non-essential spend rate to get back to your desired legacy amount, repeating the whole procedure until the benefit of the Roth on spending becomes clear.
FWIW, I've been adding "Total Discretionary Consumption + Legacy" in evaluating different things like conversions. In theory I could weight the two desiderata, but simply adding the two outcomes gives me an objective measure of plan effectiveness, other things being equal.
Thanks again for helping me think through all this.
