Estimating withdrawal tax rates

[MDM]

The Investment Order sticky has a section for estimating withdrawal tax rates.

The intent is to help people understand if some combination of things such as future pension and growth of existing investment balances, along with retirement issues such as ACA tax credits, etc., make it likely that the answer to questions such as How Much is TOO MUCH in your 401(k)? and Am I putting too much in 401K and other investment? is "yes".

I received a PM suggesting that one should include assumed future year contributions when deciding where (Roth vs. traditional) to direct this year's contributions.  While there are some careers (doctor in residency is the usual example) for which this may make sense, that is a very small fraction of the workforce.  For the rest of us, looking only at the assumed growth of current balances seems best for this year's decision. 

Interested in comments on the specific issue above, and any other thoughts on the thread subject.

[seattlecyclone]

I think there is some logic to this. The withdrawal tax rate will be based largely on your asset mix and withdrawal needs at the time you retire. Whether you contributed to Roth first and traditional later or vice versa doesn't really matter at that point; what matters is the final balances at retirement.

The advice to look at assumed growth of current balances makes a lot of sense in a lot of situations. It nudges you toward making some pre-tax contributions until you're at least pretty sure that you'll exceed the standard deduction in retirement, and then you can make some more fine-grained decisions from there. In this process you're implicitly allocating the first dollars you withdraw each year to the first contributions you made. This is a completely arbitrary decision though! You could just as easily map your first withdrawals to your last contributions, or even to contributions in the middle of your career. For most people this mental ordering doesn't matter too much because they might not experience large increases in their tax bracket over the course of their career, and they especially might not expect that to happen when they're just starting out. If you're predicting to be in a pretty similar tax bracket throughout your career, might as well get the pre-tax contributions out of the way in the beginning just in case your FIRE date comes sooner than you originally expected.

However if you do expect to increase your income dramatically, that can change things a bit. Imagine you're a single person starting out your career. You're earning $50k this year but you're pretty sure you'll be making $300k in a few years. Maybe you're a medical resident, or a software engineer who just graduated midway through the current year and aren't earning your salary for a full tax year yet. That's likely going to bring you from the 12% now to the 35% bracket later in your career. In those 35% years you'll likely find it advantageous to make pre-tax contributions, regardless of what you do this year. Why wouldn't you factor that into this year's plan? Maybe you expect to have enough of those $300k years to fill up the standard deduction in retirement just from those years.

If you look at your current account balance only, you'll see that your $0 pre-tax balance will likely lead to $0 of income in retirement, so the prospect of saving 12% now to pay 0% later seems pretty good. However if you plan to have a decade of definitely making pre-tax contributions because you'll be saving 35%, then you can switch the ordering of your mental mapping between contributions and withdrawals and see that you might instead be saving 12% now to pay 12-22% (plus ACA surcharge) later. Now those pre-tax contributions this year don't look so good anymore.

That said, I'd hesitate to update "standard advice" with this discussion in anything bigger than a tiny footnote. This is something that most people won't need to worry about, and even if you do it "wrong" it won't cost you that much in the end. There's value in having a clear set of instructions that beginners can wrap their heads around, especially in this topic. We regularly need to debunk the idea that your expected effective tax rate in retirement has any relevance to this decision, and the longer we need to make the explanation for what do do instead, the less likely it will sink in.

[secondcor521]

Chiming in to agree with seattlecyclone, but especially the last paragraph.

I see the investment order sticky as a way for beginners to get a good answer quickly and simply as they're getting started out.  As people get more advanced, they'll understand more what's going on and how everything fits together, and make more advanced strategic decisions.  Trying to stuff all of the complexity into a single post or even a single thread will do a disservice to the majority of people who either (a) are OK with good enough (and there are plenty of people like that here even on this forum, and/or (b) are well served by the existing advice.

For those who are ambitious and not well served by the existing advice, they're probably smart enough to figure that out and also hopefully understand that the investment order sticky is good for most people but not for them.  They can then do specialized stuff.

I'd second the notion of a single statement alluding to this - "You may want to do more complicated planning if you expect your earning history to vary greatly over the course of your career" or something like that.

[Cheddar Stacker]

Hi Guys! Been a while. Thanks for the nudge MDM.

I agree with the discussion thus far but just wanted to bring up a few points.

1) On the exact subject of the thread, there are just too many variables to reliably and accurately estimate withdrawal tax rates over the course of a long retirement. Aside from knowing your spending levels, annual withdrawals or investment sales, and other personal factors far into the future, you have to think about tax law changes. We can't predict those, so just plan on big changes at some point.

2) Since you can expect tax law changes over your 40-50 year retirement, you should build flexibility into your plan and your accounts. Maxing out your Traditional 401K and IRA but doing nothing else could result in higher than necessary taxes in retirement. Use them, but build some Roth accounts into the mix as well at some point. Open an after tax brokerage account. Max out an HSA. Buy some real estate. All the stuff you already said in the investment order thread.

3) Holding non tax-deferred assets will allow you to draw down some funds for unplanned costs without jumping to a new tax bracket. Holding non tax-deferred assets will allow you to react to a major increase in tax rates during a new administration.

Nothing really new or exciting here for this group. However, for the people who are still learning or seeking out further understanding, my message would be to just "do a little bit of everything so you have tax flexibility on the back end".

[SeattleCPA]

Hi Guys! Been a while. Thanks for the nudge MDM.

I agree with the discussion thus far but just wanted to bring up a few points.

1) On the exact subject of the thread, there are just too many variables to reliably and accurately estimate withdrawal tax rates over the course of a long retirement. Aside from knowing your spending levels, annual withdrawals or investment sales, and other personal factors far into the future, you have to think about tax law changes. We can't predict those, so just plan on big changes at some point.

2) Since you can expect tax law changes over your 40-50 year retirement, you should build flexibility into your plan and your accounts. Maxing out your Traditional 401K and IRA but doing nothing else could result in higher than necessary taxes in retirement. Use them, but build some Roth accounts into the mix as well at some point. Open an after tax brokerage account. Max out an HSA. Buy some real estate. All the stuff you already said in the investment order thread.

3) Holding non tax-deferred assets will allow you to draw down some funds for unplanned costs without jumping to a new tax bracket. Holding non tax-deferred assets will allow you to react to a major increase in tax rates during a new administration.

Nothing really new or exciting here for this group. However, for the people who are still learning or seeking out further understanding, my message would be to just "do a little bit of everything so you have tax flexibility on the back end".

Late to this discussion, sorry @MDM ... but for what it's worth, I agree with @Cheddar Stacker here.

I don't have anything directly on point to add (other than to "+1" Cheddar's comments)... but two tangential thoughts:

1. I think it's enormously difficult to predict tax law changes including tax rates. And that bias makes me nervous about any analysis that goes beyond a simple assumption like, "Let's just assume rate schedules look like they do now..." (Again, not really on point...a tangential remark and admission of bias...)

2. Other factors may overwhelm the impact of tax rates. As some might know, I'm personally operating/planning based on assumption that long-term interest rates and equity market returns are low-ish for my decades-long planning horizon. That seems far, far off topic... except that it ripples through things like Roth vs traditional decisions and impacts RMD worries.

3. Oh, oh, a third thought to add. I also worry in any discussion like this that folks apply a general rule of thumb to their decisions that doesn't apply materially to their specific situation. The roth vs traditional area seems to be an area where this risk may be relevant. E.g., someone who accumulates $500K (so 90th percentile) probably doesn't get that much reward (or any?) for optimizing in this area in comparison to someone who accumulates $2M or $3M (which is probably 99th percentile).

[MDM]

Thanks all.  Updated the Investment Order post to summarize this, and linked back to this thread for those interested in the detailed discussion.

[fyre4ce]

Hi there,

I realize I'm late to the party, but thought I should chime in here anyway. I'm the one who suggested some updates to the CSS that included adding an option for future contributions in estimating future tax rates, and I want to explain why I think this is a good idea.

In general, my approach to financial planning (including tax planning) is to write down a simple but comprehensive plan, and base my decisions off that plan. The plan should include income, expected expenses, location, savings rates, investment performance, etc. These plans should be really simple; in most cases you should assume the future will be like the present, except if you have some reason to think otherwise. As mentioned, certain careers (eg. doctors) have planned changes income, maybe someone is planning to buy a house or move to another state, and these should be included. We all know plans change frequently, so it's foolish to make it too detailed, but it's equally foolish not to plan at all.

Once the plan is written down, then decisions can be made based on it. When can I afford a down payment? When can I retire? Am I likely to owe estate tax? etc. If you assume that future tax rates stay the same as today in real terms (which is what I suggest), then you can also estimate your future tax rate. Future contributions are, mathematically, an important part of that equation. I fully concede that if someone has a very uncertain income, or a terminal cancer diagnosis or something, it may not make sense to include any future contributions. And, of course, if someone is planning on retiring early, then they would only include future contributions for as many years as they plan on working. But many, maybe most, people work normal-ish length careers and make pre-tax contributions along the way. At 20 years out and 5% growth, about one third of someone's future pre-tax balance will come from future contributions. If someone has confidence that they'll work and make these contributions, it should be included for planning. If/when the plans change, then the predicted future tax rate may change too, but at least they made the best decision they could with the information that was available at the time.

Another way to look at it: I have some formal training in analysis, and predictive power is one of the measures of a good analysis method. In other words, if your analysis says that if A, B, C, and D happen, your future tax rate will be X, and then A, B, C, and D do happen, then your future tax rate should be X. So, if you plan to work and contribute to your pre-tax account for another 20 years, and you do, and your investment returns and future tax brackets are as you predict, then your tax rate will be what the analysis predicts. That won't happen unless you include planned future contributions. Excluding them is a bit like deliberately skewing the scope on your rifle 5 degrees to one side- you'll always be on one side of your target, until it's so close that you can hit it without aiming.

I've heard several proposed reasons to exclude future contributions, here and elsewhere. Here are some, and why I don't agree:

Future tax rates are too hard to estimate anyway, so it doesn't matter - I definitely agree future tax rates are very hard to predict. It depends on future tax laws (impossible to predict decades in advance), investment performance (almost as uncertain), and future contributions. But just because it's a difficult problem, doesn't mean we shouldn't try to be as accurate as possible where we can. If someone's plan is to work and contribute for X more years, then those contributions should be included, until the plan changes. And while the uncertainty is very high decades out, the analysis should be repeated every few years, and it will get much more accurate closer to retirement. Estimates ~5 years away from retirement can be made with pretty good accuracy, I'd argue.

As a side note, for purposes of choosing between traditional and Roth contributions, the high uncertainty doesn't necessarily make the problem intractable, even several decades out. Future balances are hard to predict, but it also takes huge changes in future balances to make small changes in tax rates. The difference between a 12% tax rate and 35% is ~10x difference in pre-tax balance. Most schoolteachers can be confident they won't retire with a $15M inflation-adjusted balance in their 403b. And, at the end of the analysis, the tax rate gets compared in a binary fashion to the current tax rate. So, while there is a lot of uncertainty, there's also a lot of tolerance for uncertainty and still getting to the right answer.

Including future contributions is counting your chickens before they hatch - The whole of financial planning is counting your chickens before they hatch, figuring out where you want to be in XX years, and the steps necessary to get there. It's necessarily forward-looking, and this is just one question under that umbrella. In cases where someone's income is uncertain for some reason, it could very well make sense to exclude future contributions, but not generally. Also, while there are plenty of cases where someone's income or contributions unexpectedly goes to 0 (eg. disability), there are also cases where people's income rises dramatically, and prior pre-tax contributions would have been better off made as Roth in hindsight. I don't have data on which case happens more often, but either way I still think making a plan, making decisions based off that plan, then updating the plan as necessary, is the best way to make these decisions.

Most careers don't have big changes in income - This is true, and I wouldn't advise someone to plan on big changes in their income unless they had good reason. But even with completely stable income, excluding future income will introduce a bias. Someone with $500k in pre-tax savings planning to retire in 25 years will underestimate their future balance by 35% (FV(5%,25,-19500,-500000) = $2.62M, FV(5%,25,0,-500000) = $1.69M). That's a big difference that could introduce errors in tax planning going on today.

It's too complex to include future contributions - The Future Value function in Excel ("=FV") and other tools includes an argument for contributions. It's not adding any complexity to set that value to something other than zero. In most cases the value isn't that hard to figure out anyway; it's just the total amount you've been contributing to your pre-tax accounts ($19,500 for many people). A tool that estimates future tax rates should at least have the option for users to put in a non-zero value, in my opinion.

Happy to read any more thoughts on the subject.

[secondcor521]

I agree that multi-year planning tools are nice and useful.  I have some of my own and find them useful for making informed decisions.

I disagree that the CSS is intended to be a multi-year planning tool.  I think it is mostly a single year evaluation tool.

I think future contributions logically belong in a multi-year planning tool, not a single year evaluation tool.

I'd suggest creating your own multi-year planning tool as a separate tool from the CSS.  There are ways to have two spreadsheets interact in a manner that would support this sort of thing.

[fyre4ce]

I agree that multi-year planning tools are nice and useful.  I have some of my own and find them useful for making informed decisions.

I disagree that the CSS is intended to be a multi-year planning tool.  I think it is mostly a single year evaluation tool.

I think future contributions logically belong in a multi-year planning tool, not a single year evaluation tool.

I'd suggest creating your own multi-year planning tool as a separate tool from the CSS.  There are ways to have two spreadsheets interact in a manner that would support this sort of thing.

Fair enough, but aren't there many features of the CSS that are multi-year? I'd argue that "Estimating withdrawal tax rates" itself is one example, but there are others too: 401k vs. Taxable, Non-deductible IRA, TGH or Roth-convert, many of the Misc calcs, etc. I'm not suggesting adding a true multi-year capability where parameters can be varied for each year in the future, just a fixed value to represent the expected average annual contribution between now and withdrawal. Along the lines of the last point in my post, I think the accuracy/complexity trade-off for adding this in is really good.

As I said in the CSS thread, I think a more robust future rate estimator would pair nicely with the current CSS. The CSS is the gold standard for current year tax calculations, no doubt, but any future value calculations involving accounts that aren't Roth (or HSA for qualified expenses) will require an estimated future tax rate. An example worksheet I proposed in that thread had the option to read in the marginal tax rate in the current year for contributions of a certain size, and had a more robust future rate estimator that had an option for future contributions, then ran that taxable income value through a tax rate table to give an actual estimated future tax rate. That can be used all sorts of places in the current CSS tool. I'm open to making my own tool rather than add this to the CSS, but it seemed worth at least suggesting.

[fyre4ce]

The advice to look at assumed growth of current balances makes a lot of sense in a lot of situations. It nudges you toward making some pre-tax contributions until you're at least pretty sure that you'll exceed the standard deduction in retirement, and then you can make some more fine-grained decisions from there. In this process you're implicitly allocating the first dollars you withdraw each year to the first contributions you made. This is a completely arbitrary decision though! You could just as easily map your first withdrawals to your last contributions, or even to contributions in the middle of your career. For most people this mental ordering doesn't matter too much because they might not experience large increases in their tax bracket over the course of their career, and they especially might not expect that to happen when they're just starting out. If you're predicting to be in a pretty similar tax bracket throughout your career, might as well get the pre-tax contributions out of the way in the beginning just in case your FIRE date comes sooner than you originally expected.

As you point out, money is fungible, so it's not really meaningful to try to distinguish which dollars of contribution will be withdrawn when in retirement. What is significant is that tax rates on both ends (today, and at withdrawal) can depend on how much pre-tax you contribute today. If you're near a bracket boundary, contributing more pre-tax after you drop below that boundary will save you less in taxes. What's less widely appreciated is that contributing more pre-tax this year, and especially year-after-year, can push you up into a higher bracket at withdrawal, due to the larger expected balance. There are situations where it's best to split contributions this year, usually at the bracket boundary (here is a worked example).

In these situations, including future contributions in the analysis can save you taxes. If you don't, pre-tax contributions will look artificially more attractive during the first half of your career, so you'll over-contribute, partly at a lower rate than you should. Then, once your pre-tax balance is really big, then you'll make smaller contributions for the second half of your career. Due to progressive tax rates, this is less optimal than splitting contributions more evenly every year throughout your career, and making best use of that top bracket.