Avoiding future RMDs with Roth conversions as a non-resident alien

Have a question about your personal investments? No matter how simple or complex, you can ask it here.
Post Reply
TedSwippet
Posts: 1632
Joined: Mon Jun 04, 2007 4:19 pm

Avoiding future RMDs with Roth conversions as a non-resident alien

Post by TedSwippet » Mon Jan 08, 2018 6:01 pm

An odd one even by Boglehead standards, but here goes...

The recently passed tax reform has prompted a re-think of my existing Roth conversion plan. I am a US non-resident alien with a 401k from work in the US, and had intended to convert around $4k/year until age 60 and then $40k/year until the 401k is fully converted to a Roth, to mitigate high tax rates on RMDs. I am in an unusually asymmetrical tax situation -- I pay tax on Roth conversions at normal US graduated rates but without any standard or other deductions (and now, no personal exemption either), and at my full local marginal rate on 401k withdrawals (typically 40% but can go up to 60%).

My original plan used the exemption and then the 10/15% brackets to the full, stopping there. But I have now run some projections that suggest that due to fund growth, and exacerbated by the loss of personal exemption, I will not be able to avoid moving solidly into the 22% bracket before the conversion is completed. So it appears that I should be bolder.

I have 13 years before RMDs, and my tax asymmetry seems to strongly suggest converting the entire 401k before then. For more motivation, between ages 70 and 75 I will also be facing my own local country pension 'tax torpedo'.

My revised plan converts 1/N of my 401k each year starting this year, so that it is fully converted by age 70. Assuming 3% inflation and 4% real growth, this will just nudge over the top of 12% tax bracket in the first year, and by the 13th I estimate being around 1/2 way through the 22% bracket. Effective tax rates on the conversions range from 11.5% at the start to perhaps 15.5% in the final year. Not bad, and much better than my estimated 40% on withdrawals. I will have to front up the tax from other savings, at least until I can make penalty-free Roth withdrawals at age 59.5, but this is just a forex and timing nuisance rather than a cashflow problem.

So, questions. I assume nobody else has ever been in a similar position, although if you have then feedback would be great!

1. Is a rolling 1/N conversion a reasonable profile to use? I could either front-load or back-load it a little without changing the final outcome, but skewing too much results in either early under-use of the 12% bracket or a later push beyond 22% and a worse outcome. This might boil down to: which of growth and inflation is the most likely or not to manifest? And who knows?

2. I have assumed AMT is not a threat, but is this right? I cannot claim any deductions for state or other tax, nor any standard deduction, so on my recollection I should never encounter it, but it may have some wrinkle I'm not aware of.

3. I cannot claim medicare or other benefits of this type, so no income effects there. I might at some point receive US social security, but under the US/UK treaty this is taxable only to the UK, so should not affect my Roth conversion tax rates. Might there however be some other externality that would affect my US tax rates?

4. There is an odd effect at the latter part of the 13 years of conversion where stretching out conversion into RMD years can be better than full conversion before RMDs, meaning that extending to 14 or more could be a benefit. For example, with £76k in a 401k it would be better to convert in two lots of $38k even with two RMDs -- the RMDs would be around $4.5k total and taxed at 40%, but this is more than compensated by being able to convert using two 10/12% brackets for the remainder as opposed to a single conversion that extends well into the 22% tax bracket. I have no real idea how to realistically model this to find its breakeven point, but as it is probably more than a decade away I have ignored it for now. But if anyone does have a bright idea on how to project this, I am all ears. (I-ORP certainly cannot handle it!)

5. The potential expiry of the tax cuts in 2025 actually has little to no effect on this plan as far as I can see. If they do expire, the tax I would pay on conversions increases a little, but to nowhere near 40% so the basic principle still holds. Sound right?

6. Anything else I might not have thought of? Obvious dangers include tax treaty issues and 401k provider no longer cooperating. Lesser hassles involve timing estimated taxes and the general difficulties of dealing with the IRS from another country. But other than these?

User avatar
ray.james
Posts: 959
Joined: Tue Jul 19, 2011 4:08 am

Re: Avoiding future RMDs with Roth conversions as a non-resident alien

Post by ray.james » Mon Mar 05, 2018 3:32 am

Hi Ted, some additional thoughts;

Assuming you are able to cover the taxes, will it be profitable to convert all 22%//24% of next 13 years now as long as it fills up to 24% bracket. I think this is one scenario that will help looking at marginal rate than average rate. The advantage is the compounding effect happens in roth rather than pre-tax. The math is tricky though as one has to assume the tax bracket inflation adjustment in 10 years. (ofcourse tax rates can change but something that one cannot control cannot be )

An information question for me: If two people have 2 ira's. Can they convert each by using married filing separately? Since there is no married filing jointly
When in doubt, http://www.bogleheads.org/forum/viewtopic.php?f=1&t=79939

TedSwippet
Posts: 1632
Joined: Mon Jun 04, 2007 4:19 pm

Re: Avoiding future RMDs with Roth conversions as a non-resident alien

Post by TedSwippet » Mon Mar 05, 2018 12:26 pm

ray.james wrote:
Mon Mar 05, 2018 3:32 am
Assuming you are able to cover the taxes, will it be profitable to convert all 22%//24% of next 13 years now as long as it fills up to 24% bracket. I think this is one scenario that will help looking at marginal rate than average rate. The advantage is the compounding effect happens in roth rather than pre-tax. The math is tricky though as one has to assume the tax bracket inflation adjustment in 10 years.
I have modelled this in my spreadsheet, and the best results definitely come from using the 10% and 12% brackets for as long as possible. For any given single flat tax rate paying it either before compounding or after gives the same result. Say a portfolio value is V, tax is T and growth over 13 years is G, then because multiplication is commutative, taxing at the very end is T * (V * G) and is the same as taxing at the very start (T * V) * G.

V is just what it is, and G is outside my direct control, but I can minimise the tax rate T by converting gradually over time to use multiple lower rate brackets. The jump from 10% to 12% is small, so there is a good case for using those brackets to the full. The next bracket is 22% though, a large step that triples my tax payable if I only double my annual conversion amount. Outside the 'sweet spot', in other words.

My spreadsheet estimates tax bracket inflation, but I only did that to get a sense of possible absolute values. Working in current-year dollars by simply subtracting expected inflation from expected growth functions just as well for projection purposes.
ray.james wrote:
Mon Mar 05, 2018 3:32 am
An information question for me: If two people have 2 ira's. Can they convert each by using married filing separately? Since there is no married filing jointly
I would assume so. Except for residents of Canada, Mexico, and South Korea (and I have no idea why these countries are 'special'), the 1040NR for married filing separately pays no attention to a spouse at all. Except, that is, for asking if you have one so that they can then corral you into the most tax-maximising filing status that they have :-(

User avatar
ray.james
Posts: 959
Joined: Tue Jul 19, 2011 4:08 am

Re: Avoiding future RMDs with Roth conversions as a non-resident alien

Post by ray.james » Mon Mar 05, 2018 2:02 pm

I agree with the multiplicative nature of the tax. I did not do a good job of explaining my idea. However it is not that huge in the grand scheme of things but a good optimization overall. Lets assume one is converting 600K ira in 10 year span.
As of today
12% bracket ends 38700
22% ends at 82500
24% ends at 157000

So if you were to convert in 10 years in 1/N method, you will convert 60K every year. up to 38K in 12% and the the rest in 22%. Lets assume due to inflation the tax brackets of 12% will end at 42K. This implies the median is 40K. So in 10 years 40k * 10 years = 400K converts at 12% or less and the rest at 22%.

Now, what I am proposing is since you have 200K more than 400K(plus FUTURE growth- we will get to it soon...) at 22% why not do that now rather than spread over next 5/10 years. So in this case
Year 1: 82,500 converted
...
year 4: 82,500 converted
year5: 70K converted
year6: 40k converted
...
year 10: 40K converted

This way, the value that would convert at 22% will still be in 22% but its future 8% growth happens in roth. What changed in the above formula is V. V* (1.07 ^10) the first year converted growth is tax free. In this specific case, 82.5k-60K = 22.5 extra converted will lead to 20K tax free withdrawal in future due to 8% growth in next 9 years.

now...if I can improve these numbers for the actual vs future expected growth. I have seen a post by cruncher that did this in his spreadsheet...I will find and hopefully give rest of the math in my next post. I am assuming, even 24% bracket will look good when you are looking at 10 year tax deferral.(ofcourse 400K will always be on 10 year plan like yours now in 10/12 bracket.)
When in doubt, http://www.bogleheads.org/forum/viewtopic.php?f=1&t=79939

TedSwippet
Posts: 1632
Joined: Mon Jun 04, 2007 4:19 pm

Re: Avoiding future RMDs with Roth conversions as a non-resident alien

Post by TedSwippet » Tue Mar 06, 2018 9:57 am

ray.james wrote:
Mon Mar 05, 2018 2:02 pm
This way, the value that would convert at 22% will still be in 22% but its future 8% growth happens in roth. What changed in the above formula is V. V* (1.07 ^10) the first year converted growth is tax free. In this specific case, 82.5k-60K = 22.5 extra converted will lead to 20K tax free withdrawal in future due to 8% growth in next 9 years.
I think I can see what you are saying, and perhaps see a bit of potential benefit, but it is tricky to quantify.

We can start with a simple example. Suppose I have a 401k valued at $120k, two years in which to convert, and growth is 4% real (that is, after inflation, which we can now ignore). Two years at the top of the 12% bracket gives me $77,400 at 12% or lower, so the rest (including any growth) has to go into the 22% brackets somewhere in years one and two.

Using a 1/N profile, I convert $60,000 in year one. Tax payable is $9,139.50, the Roth balance is $50,860.50, and the 401k is now $60000. One year later, the Roth has grown by 4% to $52,894.92 and the 401k to $62,400. I convert the $62,400 in year two. Tax payable is $9,667.50, the Roth increases by $52,732.50 to $105,627.42, and the 401k is empty.

If I convert to the top of the 22% bracket, I would instead convert $82,500 in year one. Tax payable is $14,089.50, the Roth balance is $68,410.50, and the 401k is now $37,500. One year later, the Roth has grown by 4% to $71,146.92 and the 401k to $39,000.00. I convert the $39,000.00 in year two. Tax payable is $4,519.50, the Roth increases by $34,480.50 to $105,627.42, and the 401k is empty.

So the final Roth balance is the same, but the cumulative tax bill is lower. And where the conversion tax is being paid from other taxable savings, by paying it earlier through a bit of front-loading I have effectively taken money out of taxable savings and put it into a Roth where it can now grow tax-free sooner than otherwise. Had I paid the tax later, I would have had the growth in my taxable savings account, but that growth would itself have been taxable when realised to pay the later tax bill.

Is that the benefit you are thinking of here?

If yes, there may be some mileage to be had out of it. I will however have to be very careful not to waste any of the 12% and lower bracket space in later years, and much of that comes down to how much larger (or smaller) growth is compared to inflation. It would also be a problem if a large market crash appears in later years. And if or when I start to pay conversion tax out of the Roth balance itself after age 59.5 any benefit from this front-loading strategy will be lost because I will then no longer be effectively moving money from taxable accounts into a Roth.

Complicated then, and with lots of unknowns.

User avatar
ray.james
Posts: 959
Joined: Tue Jul 19, 2011 4:08 am

Re: Avoiding future RMDs with Roth conversions as a non-resident alien

Post by ray.james » Tue Mar 06, 2018 2:07 pm

TedSwippet wrote:
Tue Mar 06, 2018 9:57 am

So the final Roth balance is the same, but the cumulative tax bill is lower. And where the conversion tax is being paid from other taxable savings, by paying it earlier through a bit of front-loading I have effectively taken money out of taxable savings and put it into a Roth where it can now grow tax-free sooner than otherwise. Had I paid the tax later, I would have had the growth in my taxable savings account, but that growth would itself have been taxable when realised to pay the later tax bill.

Is that the benefit you are thinking of here?
Exactly. You have laid it out much better than I intended. I agree in this scenario it is tight and overall not a big deal. But when there is 13 years like your original plan the compounding effect is substantial. It also depends on how much is the spillover amount from 12 to 22%. If it is a few'K it does not matter.
When in doubt, http://www.bogleheads.org/forum/viewtopic.php?f=1&t=79939

TedSwippet
Posts: 1632
Joined: Mon Jun 04, 2007 4:19 pm

Re: Avoiding future RMDs with Roth conversions as a non-resident alien

Post by TedSwippet » Tue Mar 06, 2018 3:54 pm

ray.james wrote:
Tue Mar 06, 2018 2:07 pm
Exactly. You have laid it out much better than I intended. I agree in this scenario it is tight and overall not a big deal. But when there is 13 years like your original plan the compounding effect is substantial. It also depends on how much is the spillover amount from 12 to 22%. If it is a few'K it does not matter.
Ah, right. Thank you for raising this.

On current estimates -- 'guesses' might be a better word! -- I will be just into the 22% bracket on year one, and around half way through it by the time I do the last conversion. That suggests that starting half way through it on year one instead might be a good first step. It all really depends on how far growth runs ahead of inflation over the period.

... Assuming it runs in front at all, that is. It turns out that I do not have much wiggle-room here. If growth were to exactly match inflation I could convert my entire 401k in 13 years and stay below the 22% bracket every year by just $100-$200 per year with a flat 1/N. So to make best use of this strategy I need a good guess at the differential between growth and inflation over the remaining conversion window. The numbers will start to look more definite as the window shortens, but then there is little or no opportunity for a 'course-correction'.

Vexing. Perhaps I will start with a gloomy view of growth, say 2% real rather than 4%. The loss from missing out on some of the 12% bracket in later years looks to be much larger than the gain from front-loading into the 22% bracket earlier, so underestimating growth rather than overestimating it seems safest. I think.

I also need to get past my visceral dislike of paying taxes any earlier than the latest date possible.

Post Reply