In a recent article, we described how a hypothetical couple, Barney and Betty, would enter their paycheck information into the Retirement Budget Calculator (RBC) so that RBC could help them accurately estimate taxes in retirement, arguably one of the most difficult but important aspects of retirement planning.
Background
Barney is 61 years old and has been reading a lot about how Roth retirement assets can improve retirement outlooks. He and Betty have already accumulated $500k in Roth assets, along with $500k in a taxable brokerage account and $300k in 401k and 457b accounts. He is wondering if contributing the maximum amount to his Roth 401k is a good idea, so he decides to model that in RBC and see how increased Roth assets might affect retirement withdrawals.
Barney’s current gross monthly income is $10,100 and he plans to stop working at the end of 2025. Starting in 2031, Barney will have a social security benefit of $3000 per month and Betty will have a spousal benefit of $1400 (ends in 2050 when the model has her passing away). The taxable brokerage account is generating a bit more than $17k of income each year (dividends and capital gains distributions). Planning for a worst-case scenario, Barney assumes that their investment return is a flat 3% in RBC, just matching assumed inflation.
Barney wants to wait to collect social security at age 70 because he has a history of longevity in his family, and their plan has Barney passing away at age 95 (2056) and Betty at age 90 (2050). Their current monthly expenses are $7000, adjusted annually by 3% for inflation and we assume that expenses don’t change in retirement.
Scenario 1: Default Withdrawal Order
When withdrawing funds, the conventional wisdom is to use taxable brokerage assets first, then qualified investment assets (traditional IRA or 401k) and save Roth assets for last. This is also the default withdrawal order in RBC if you don’t enter an explicit withdrawal order, so Barney decides to start with that withdrawal strategy as a baseline.
The first scenario has them contributing the maximum amount to Barney’s Roth 401k. To do that, they will need $2371 from their brokerage account to meet their monthly budget needs.
Figure 1 shows the bottom part of the RBC dashboard for this plan if they choose to use the default withdrawal order to meet their budget needs in retirement.
Barney would like the plan to go to 2056, which is the last year of his life expectancy. However, the future view shows the money only lasting until 2052. The lifetime taxes are projected to be about $150k but concentrated in the initial working years and in the first few years after social security starts. What is going on?
Brokerage assets are being used to pay for the budget shortfall while Barney is still working, and then they are used to pay for all expenses for the years before starting social security. In 2031, when social security starts, the brokerage funds run out. This means that fully taxable withdrawals from 401k and 457b accounts add to the highly taxed social security income so federal taxes are high again. A few years later, those assets run out. As a result, the taxes on social security drop and Barney and Betty owe almost no tax for the last half of their retirement.
In this scenario, the Roth contributions were during the relatively high tax working years and the Roth withdrawals were when Barney and Betty had very low taxes. This is exactly the opposite of how Roth assets are supposed to work, so blindly choosing to withdraw according to conventional wisdom didn’t work out very well.
Scenario 2: Equal Withdrawals
Let’s see what happens if we fund the budget shortfalls using equal amounts of taxable brokerage assets, 401k assets, and 457b assets. There is no reason for choosing this other than the default withdrawal scheme seems rather ill-suited for Barney’s situation. The results for this scenario are shown in Figure 2.
The plan now runs out of money in 2055, which is closer to when they wanted the plan to end. It also has a dramatic effect on the lifetime tax burden, lowering it from $150k to about $88k. During the working years, the annual tax bill is higher because of the 401k and 457b withdrawals to cover the spending deficit. This is somewhat equivalent to doing Roth conversions during the last few working years since Barney is trading 401k and 457b assets for Roth assets. The 401k and 457b accounts run out of assets in 2034 (you can see taxes drop after that) and the taxable brokerage assets last until 2042 (taxable social security drops at that point).
Scenario 3: Traditional 401k
There are still a lot of years where they are withdrawing Roth assets while in a very low tax bracket, so Barney starts wondering if Roth contributions are a good idea. He modifies the plan to contribute the maximum to his traditional 401k (instead of the Roth 401k) and uses the default withdrawal order to fund budget shortfalls.
As shown in Figure 3, the lifetime tax burden is $171k, which is higher than scenario 1. Their assets run out in 2054, which is longer than scenario 1, but one year short of scenario 2. With the default withdrawal order, Roth contributions during the working years give an overall lower tax burden but the assets run out more quickly. Like scenario 1, the taxable brokerage runs out of assets in 2032 and the gap years are tax-free. The required minimum distributions are not very large but using up the 401k and 457b assets increases the taxability of the social security benefits until those assets run out in 2041.
Scenario 4: Traditional 401k with Equal Withdrawals
Since funding budget shortfalls with equal amounts from taxable brokerage, 401k and 457b accounts worked so well for scenario 2, Barney decides to try that strategy with the traditional 401k contributions during the working years.
As shown in Figure 4, the lifetime tax burden is now $88.5k and the plan runs out of liquid assets in 2055, which is almost the same as scenario 2. During the working years, taxes are reduced because of the traditional 401k contributions, but the gap years are like scenario 2. When they start taking social security benefits, their 401k and 457b balances are higher than scenario 2 and those assets last until 2040. That means that this scenario is trading lower taxes now for increased taxes on their social security benefits later.
Summary
In hindsight, we could have guessed that Roth contributions were not going to make much of a difference. Barney and Betty already have 77% of their assets in Roth and taxable accounts. Roth assets are not taxed at all, and taxable accounts are preferentially taxed, so it’s obvious they are not going to pay a lot of taxes in retirement. You could even argue that they went a bit overboard amassing Roth assets and that they should have had a more balanced portfolio distribution.
Figure 5 compares the liquid asset balances over time for all four scenarios. Comparing scenarios 1 and 3 shows that deciding to make more Roth contributions was not really a significant issue. Doing traditional 401k contributions increases the liquid asset balance at retirement, but the lifetime tax burden is increased so the assets only last a little bit longer. Therefore, looking only at the lifetime tax burden is not the best way to optimize the plan.
Comparing scenarios 2 and 4 confirms that Roth or traditional 401k contributions have very little effect on the longevity of the plan. In this case, the longevity and lifetime tax burden are almost identical. While the liquid asset balance is higher at retirement, the decreased tax burden in the final working years is balanced by an increased tax burden in retirement.
Comparing all four scenarios, whatever Betty and Barney choose for their retirement contributions for the last four years has very little impact on the longevity of the plan. However, their choice of withdrawal strategy can have a significant impact on the longevity of the plan.
You will have to do your own modeling or get help from a financial planner to figure out what will be best for your situation.
Future work
This exercise only looked at the worst-case scenario where investment returns are very low and only match inflation. If investment returns are higher, the taxable brokerage and 401k accounts will last longer and that will affect the longevity of the plan and the tax burden. In other words, the best withdrawal plan for a low-return environment may not be the best withdrawal plan for a normal (median) or high-return environment.
It would be good to repeat this exercise with different proportions of Roth, taxable brokerage, and qualified assets. For someone with lower amounts of Roth assets, making Roth contributions and/or doing Roth conversions could be more beneficial. Another factor that is not considered in the RBC models is the possibility of going over the IRMAA thresholds for Medicare premiums.
Additional Thoughts and Suggestions for Improved Modeling
Modeling taxes in retirement is hard, but it is arguably one of the most important aspects of retirement planning. Here are a few things learned along the way, along with suggestions for how to work around them, or how RBC might improve in the future.
The future view shows the total liquid assets over time, but without clicking on each year for account details, you can’t see when the taxable brokerage account runs out of money, for example. It took some time for me to realize that the first scenario was running out of brokerage money in 10 years and that I should stop that investment income to avoid modeling investment income when there were no investments. It would be very nice to have an option to show account balances over time (e.g., stacked line chart with different colors for each account) to show when each account was running out of money and what the total balance is.
Taxable brokerage accounts generate taxable income (e.g., dividends), so you probably want to model that in RBC. That taxable income is going to go up or down with the balance of the account, and it will stop entirely if/when the taxable account balance runs out. The first scenario withdrew the entire brokerage balance in about 10 years, so I just modeled that as 0% inflation on the income amounts, and then stopped the income in the year the account ran out. The other scenarios took longer to run out, so I used 1% inflation, but again stopping when the account ran out. Given the uncertainties involved in projecting retirement cash flows and investment returns, it’s probably a reasonable approximation. However, it would be nice if RBC could model the investment income as a percentage of account balance, which would then allow the investment income to fluctuate with the account balance.
Having just one choice for withdrawal order can be limiting. During the working years, covering budget shortfalls with taxable brokerage money is probably better. However, in the gap years, a combination of 401k/457b assets and brokerage assets would allow users to manage their tax bracket during those years and later (because of taxation of social security). After starting social security, withdrawing some Roth assets might allow you to better taxation of social security benefits.
Looking at the total tax burden in nominal dollars might be a bit misleading. For example, if you have consistent income and your average tax burden is constant each year, the lifetime taxes chart would show increasing taxes (income and taxes are going up with inflation). It might be better to have an option to show the percentage tax burden instead of the nominal tax burden because better tax plans are generally going to be the ones with lower average percentage of taxes paid, not necessarily the ones with the lower lifetime taxes.
You might have noticed that state taxes were not considered in this article. Currently, RBC doesn’t have a good way to model state or local income taxes. You can enter withholding and RBC assumes that the withholding is correct. For federal taxes, RBC uses the withholding to determine if you get a refund or not and adjusts the year-end balances accordingly. It might be reasonable to approximate state taxes as a percentage of federal taxes or as a flat percentage of income.