Asset Location Gets REALLY Complex

May 8th, 2018 by Potato

I’ve long been on record as saying that for many investors, trying to optimize their tax situation is not optimal for their overall strategy — worrying about asset location and moving to US-based funds can add significant levels of complexity for the sake of a few extra basis points. It’s confusing, hard to rebalance, and there’s no agreement on what “optimal” is in the first place (wait, should bonds be the first or last thing sheltered?). When the experts can’t even agree on what optimal is, then you know it’s not worth your time to worry about. If all of that throws you off your base game of saving and investing (e.g., if you defer investing your contributions for a year or two because you have to find a bigger block of time to plan out where everything goes, or if you ignore rebalancing because it’s too hard with everything in separate accounts), then it’s not worth the potential savings (this comes back to execution risk).

Just how many basis points or where to put things to be optimized is tough to say. Ben Felix of PWL recently had a white paper that tried to quantify the boost you can get from being efficient with your asset location. The paper uses a Monte Carlo simulation, which can sometimes be confusing to interpret, but broadly speaking, he found the potential boost from optimization to average about 0.23%/yr. Earlier, Justin Bender and Dan Bortolotti used past returns to estimate the boost from tax efficiency in the same range, about 0.3%/yr. And this is likely at the high end of the effect – this very long post is going to get into some very confusing additional factors, but ultimately I think the gain to be had from optimizing is even less than that for most people.

So if you want to stop here, it’s a fine place to stop: trying to tie yourself in knots with the asset location game is only going to save you at best a few dozen basis points per year and confuse your portfolio and make it harder to rebalance — people pay about the same to save complexity by moving to a robo-advisor or TD e-series and that’s totally worth it. I’m a big believer in just using the same allocation in all your potential locations.

TLDR: optimizing your asset location may only provide modest savings and not be worth the effort and complexity: my general recommendation is not to bother and to simply copy your desired asset allocation in each account.

A big part of the problem is that determining what is tax efficient depends on assumptions about future returns, differences in tax rates, and the behaviour of tax shelters. It all means dealing with added complexity. The Monte Carlo results in Ben Felix’s paper help illustrate this: in certain circumstances, what you predict will be optimal on average can actually work against you in a number of scenarios.

A further problem is that there are differences across the tax brackets: someone in the lowest bracket in Ontario has a negative marginal tax rate on eligible dividends, while at the top tax bracket dividends are taxed at a higher rate than capital gains. Most such analyses only use the top bracket. This helps simplify things for author in a few ways: your tax rates don’t move and you don’t get secondary effects (where higher returns cause the tax rate to increase). But then the “optimal” allocation derived from an analysis for the 1% might not fit the rest of us. Ben Felix is to be commended for including a run with lower brackets (which, TLDR, also found a much lower benefit to optimizing vs. just keeping a balanced portfolio in each account — so if your income is closer to $70k/yr than $200k/yr, then there’s even less potential savings on the table so again, just side-step the thorny issue of asset location entirely).

Another issue is that most such analyses don’t adjust the allocation for the after-tax exposure. So putting bonds in an RRSP comes out as better not because it is necessarily more tax efficient, but because it effectively lowers the after-tax bond allocation — the riskier portfolio will have higher returns in the first place, so the overall outcome is better from that rather than because of any tax savings.

This takes a bit of a conceptual mind shift: remember that the RRSP is filled with pre-tax money, so you can’t spend a dollar in your RRSP the same way you’d spend a dollar in your TFSA. Imagine that your RRSP is not just your money, but rather is jointly owned by you and the government — it is after all, pre-tax money that’s in there. The tax paid when you withdraw is just the government taking their portion back, and then it gets to grow tax-free in the meantime. This doesn’t affect your allocation when your allocation is the same across all your accounts: the government’s portion then neatly mirrors your own.

However, when you start putting parts of your portfolio exclusively in certain accounts, then you may want to factor this in (we talk more about this advanced topic in the course). So if a dollar in your RRSP is really only about half yours (at the highest marginal tax bracket), then you can think of the money you have as being the amount in your taxable and TFSA accounts, and part of your RRSP, with the government owning the rest of your RRSP. So your brokerage statement may say there’s $100 across those accounts, with $40 in your RRSP, but really your money is just $80,000, with $20 of it being the government’s portion of your RRSP.

Thus $40 in bonds in your RRSP and $60 in equities outside is not the same effective allocation as a 60:40 split in each account (for those in the course, recall the bunny infographic). In the first case, the government holds $20 in bonds, and you hold 20/60 (a 75:25 allocation), versus the 60:40 split for both you and the government in the second case.

Let’s see how this plays into asset location decisions. Let’s use a scenario of someone in the highest tax bracket, and assume that bonds will pay 2.7% in interest (fully taxed) and 0.3% in capital gains (deferred and tax-favoured) while stocks will return 2% in dividends and 6% in capital gains. The tax hit in dollars for each $1000 invested in a non-registered would be, using the top Ontario tax bracket:

Bonds: $14.45 + $0.80 (deferrable)
Stocks: $7.87 + $16.06 (deferrable)

I’m clearly paying more tax on the stocks: despite the lower tax rate on dividends and capital gains, the higher expected returns mean I’ll expect to pay more tax (in dollars). Therefore, I should prefer to used my limited tax shelter space to shelter my equities. So it seems weird and counter-intuitive that with basically the same numbers, previous papers on asset location have come to the opposite conclusion and suggest that bonds go into the RRSP first.

So let’s look at what happens when we shift the assets around, with and without adjustments for the government’s portion of the RRSP. Let’s assume there’s a $100,000 portfolio, with only $40,000 of RRSP room available. We’re aiming for a 40% bond/60% stock allocation (for simplicity I’m using just two asset classes with the return assumptions as above and assuming everything is liquidated after 1 year — further allowing things to compound should make the shelter space for the higher-return equities even more valuable).

Taxable

RRSP

After-tax value after 1 year

Effective allocation

Balanced

$36k stocks, $24k bonds

$24k stocks, $16k bonds

$82,076

60:40 stocks:bonds

Bonds in first, no adjustment

$60k stocks

$40k bonds

$82,510

76:24 stocks:bonds

Bonds in first, adjusted

$47,153 stocks, $12,847 bonds

$40k bonds

$81,979

60:40 stocks:bonds

Stocks in first, no adjustment

$20k stocks, $40k bonds

$40k stocks

$81,786

49:51 stocks:bonds

Stocks in first, adjusted

$28,565 stocks, $31,435 bonds

$40k stocks

$82,140

60:40 stocks:bonds

Balanced, 76:24

$45,600 stocks, $14,400 bonds

$30,400 stocks, $9,600 bonds

$82,621

76:24 stocks:bonds

If you imagine that you have a total of $60k nominal in equities and you just re-jig where you put it, then you get the result that bonds in your RRSP makes sense. But really, this is no longer the same after-tax portfolio. Remember, you have just $78,588 of your money — the other $21,412 is the government’s portion of your RRSP. So for a 60:40 portfolio you really have $47,153 in stocks to allocate, and the government has $12,847. If then you pull the government’s stocks out and make them all your stocks, while replacing the government’s share of the portfolio with all bonds, then your tax bill on withdrawal will be lower (the government’s portion will grow less), but your money in the portfolio will be riskier. Your after-tax allocation is now effectively 76:24, a much more aggressive portfolio than the 60:40 you were intending to use.

Now, whether to make this adjustment is a huge area for debate. Because what is the purpose of your asset allocation and risk tolerance in the first place? If it’s to manage your emotional response to nominal declines you might see in your account, then using the nominal amounts might be the right way to go about it even when you split up your asset location. After all, when you check in on your portfolio and are worried about inducing panic, you generally include the government’s portion of the RRSP as your money and look at declines in nominal (vs after-tax) terms. So seeing that the government is down along with you doesn’t really stop the panic attack in a market crash. Indeed, from this point of view it can make sense for an advisor to suggest putting their client’s bonds preferentially in the RRSP: they can get an effectively riskier, higher-returning portfolio without the client really noticing the volatility (because the changes in nominal dollars with volatility is the same). On the other hand, if you recognize that a dollar in your RRSP is not worth the same as a dollar outside, then you may wish to put an adjustment factor on your RRSP balances, which changes what optimal looks like.

So with that mind-set in place, the second row (the traditional optimum allocation) comes out ahead of a default balanced everywhere option not because of the tax inefficiency of bonds and the sheltering of the RRSP, but because the effective 76:24 portfolio has a higher expected return in the first place. We see that in the last row: if we just move to that same effective allocation but balance the location across the accounts, we actually get more money than in the so-called optimal, nominally 60:40 case. Also in the third row: keeping the same after-tax allocation to bonds and sheltering them first leads to a lower return than the balanced case. You also see it in examples like this post from Justin Bender, where sheltering bonds in the RRSP is optimal, but not in the TFSA (i.e., scenario 2 and 4).

This is a simplified case (there’s no international equity, no US withholding tax issues, no multi-year compounding), but we see that the savings on offer is about 50 basis points when putting bonds in the RRSP without remembering to adjust the allocation — not too far off what the other papers get. Of course, going to an effective 76:24 allocation but without asset location would give an extra 66 bp. If you’re adjusting the allocation for the government’s portion of your RRSP, then putting stocks in the tax shelter first was the more optimal move, but all that added worry and complexity only saved you 8 basis points. Adding compounding over time and the withholding tax issue for US equities should further help make sheltering equities first the more optimal strategy when you expect low bond returns, and increase the potential benefit (which I still have yet to estimate well), but being in a lower tax bracket will likely reduce it. Again, the amounts involved are not huge, especially in comparison to the added complexity.

Interim Conclusion:

Adjusting your effective after-tax asset allocation can change your results of what an optimal asset location strategy looks like. Note that in any case, all these optimization games can bring is a few basis points of extra return — that may be worth the headaches of reading all these whitepapers for a few of you, but bear in mind the complexity involved versus the suggested default of just using the same portfolio in each account.

Optimizing in any case requires estimating future returns and tax rates, which can be difficult — post hoc, you may be worse off than a simple balanced approach if your estimates are wrong.

Engaging in asset location adds a lot of complexity to your portfolio for potentially minor benefit, and this benefit is likely even lower for people in middle tax brackets vs. those in the top bracket where such analyses usually focus.

Take-Home Message:

This is all very confusing and my default suggestion is that it’s best to just replicate the same allocation in all your accounts.

The Pros & Cons of Each Approach:

No asset location: repeat your asset allocation in each of your accounts (taxable, TFSA, RRSP).

  • Pros: You don’t even have to enter the argument about adjusting allocations for after-tax effects, and it’s very simple to do. Plus, you won’t ever run into problems with rebalancing as each account rebalances on its own, and you don’t have to predict the future in your optimization attempts.
  • Cons: However, you may not be able to shake the feeling that somehow you’re leaving a few basis points on the table, and you may pay more in commissions to repeat trades in each account.
  • Operating quick rule-of-thumb: decide on your asset allocation, then just do that in each of your accounts.

Asset location without adjusting for the tax effects of your RRSP: split your asset allocation up, using only nominal values (i.e. treat a dollar of bonds in your RRSP the same as a dollar of bonds in your TFSA or non-registered).

  • Pros: It’s much easier conceptually to see all the money as being equal no matter where it sits, and there are lots of resources to read that use this method so you’ll feel like you’re in good company and won’t get confused when you try to refresh your memory in a few years. You can trick yourself into accepting a higher risk portfolio, and asset allocation is only approximate at best (the difference may be behind the extra return seen, but really, is 60:40 meaningfully different from 76:24? …and the skew is even lower at lower tax brackets). You can save a bit on commissions by making fewer transactions because some asset classes won’t appear in all accounts.
  • Cons: However, with asset classes sitting in separate accounts rebalancing will be a pain, and it adds a lot of complexity and cognitive load to your investing. Plus if you put your equities in your non-registered account, your portfolio isn’t really set up to help add resiliency to your life, as you’d likely look to sell your bonds to cover any emergency spending that was larger than your cash emergency fund. You’ll treat your RRSP and TFSA as having quite different tax sheltering effects, which is odd given all the other reading you’ve done on how they both shelter gains from taxes.
  • Operating quick rule-of-thumb: put the asset with the lowest expected return in your RRSP. Put the one with the highest tax hit (expected return * tax rate) in your TFSA, which will likely be US & International equities. If your RRSP overflows with bonds, those go out to a non-registered account rather than TFSA. If your TFSA overflows with equities, then those go out to the non-registered (Canadian first).

Asset location with adjustments for the tax effects of your RRSP:

  • Pros: You more consistently apply your thinking of how RRSPs work, can maximize your tax savings, and will treat your TFSA and RRSP more consistently in asset location decisions. You can also feel like a pioneer at the forefront of your field, and can carefully explain why you do the crazy thing you do and watch as people’s brains ooze out of their ears at parties. You can save a bit on commissions by making fewer transactions because some asset classes won’t appear in all accounts. Plus, if you use similar low-interest-rate assumptions and bonds are the first thing in your taxable account, this allows your bonds to act as your second-tier emergency fund: already non-registered and ready to be spent if needed.
  • Cons: However, it is the most confusing option of all, adding the most complexity (though to be fair, asset location is already one of the most complex things you’ll inflict upon yourself as a DIY investor). Furthermore, the complexity keeps going: after adjusting your RRSP for the government’s portion, you may want to try to adjust your non-registered account, too, versus your TFSA (which is harder, though the factor will be much smaller). And with asset classes sitting in separate accounts rebalancing will be a pain. Also, while your portfolio may have the same real, after-tax risk as one you targeted when balancing across all accounts, the nominal amounts will look riskier — above, you had $68.6k nominal dollars in stocks when locating them in your RRSP with the pre-tax adjustment, versus just $60k nominal dollars in the all-balanced default approach and the non-adjusted asset location approach. In a market downturn, it will look and feel like you’re losing more money, and will take an extra effort off will to remember that you really only had $60k on the line and it’s the government that’s losing on their stock portion. In short, it adds a lot of complexity and cognitive load to your investing.
  • Operating quick rule-of-thumb: Bonds will be the first thing kicked out to non-registered, then likely Canadian equities. For most assets there isn’t a difference between the treatment of the RRSP and TFSA, except US equities held through US vehicles, which can save the withholding tax in the RRSP (and international equities gets complicated but let’s say a weak RRSP preference as well).

This seemingly pedantic debate about whether to adjust the RRSP amounts for the pre-tax nature of the tax shelter ends up changing the answer as to what you prioritize putting in your shelter when optimizing (given certain expectations about returns and tax rates), so decide and execute carefully.

Bonus Round:

There’s a legitimate debate for whether or not to adjust the amounts in your RRSP, and I just gave you some pros and cons for each way. Here are a few examples to help reinforce why I think you should use the method with adjustments (even though it’s complicated).

First, running an asset location optimization without adjusting the amounts in the RRSP tells you not to use your limited tax shelter space to shield the thing with the highest marginal tax hit, but rather to stick the thing with the lowest potential for growth in there. We can see this if instead of a 60:40 portfolio of stocks and bonds, we use an example of a portfolio composed of zero-yielding cash and stocks. It should be intuitively obvious that there is no benefit to putting something with zero yield (and thus zero tax) in your RRSP tax shelter, yet:

Taxable

RRSP

After-tax value after 1 year

Your effective allocation

Balanced

$36k stocks, $24k cash

$24k stocks, $16k cash

$81,499

60:40 stocks:cash

Cash in first, no adjustment

$60k stocks

$40k cash

$81,953

76:24 stocks:cash

Cash in first, adjusted

$47,153 stocks, $12,847 bonds

$40k cash

$81,232

60:40 stocks:cash

Stocks in first, no adjustment

$20k stocks, $40k cash

$40k stocks

$81,197

49:51 stocks:cash

Stocks in first, adjusted

$28,565 stocks, $31,435 bonds

$40k stocks

$81,677

60:40 stocks:cash

Yes, with no adjustment to the allocation, you would get the result that putting the zero-yielding cash in your RRSP was still the optimal thing to do (and would save you nearly the same 55 bp benefit as in the base case above with bonds). This strange result suggests to me that adjusting the allocation for tax concerns is the correct way to think about your RRSP in asset location exercises.

Second, we remember to make this adjustment in every other instance where we compare amounts in a non-registered account, or comparing TFSAs and RRSPs head-to-head. For example, one frequently asked question is whether it’s better to hold dividend-paying stocks in a non-registered account than an RRSP, because the withdrawals will be taxed at your full marginal rate while the dividends/capital gains will be taxed at a lower rate (e.g., the second myth in this column by John Heinzl in the Globe or this one). A naïve analysis would compare $1000 in a non-registered account and show that after say 200% capital growth you have $2465 left after paying a relatively light capital gains tax, while in a RRSP all the withdrawals (including the principal) are taxed at your full marginal rate, so you’d only have $1394 of your $3000 after paying 53.53% in tax. Of course, the flaw in that analysis is that you ignored the fact that pre-tax money goes into the RRSP — in a fair comparison, you’d really start with $2152 in your RRSP, which would then experience the 200% growth, and even after taxes on withdrawal would leave you with $3000, showing that the RRSP does indeed completely shelter gains from taxes when you account for the pre-tax nature properly.

Every time, the answer is that to fairly compare the RRSP, you have to recall that it’s pre-tax money so there should be more in there in the first place to be comparable to a TFSA/non-registered investment.

So really, I think we should be continuing to account for the fact that it’s pre-tax money in the RRSP (that is to say, part of what’s in there is the “government’s portion”) and adjust your asset allocation accordingly when you get into fancy asset location footwork. This is consistent with the way we treat amounts in the RRSP in any other analysis. Plus the idea of asset location is to try to get a bonus return through optimization with little added risk — the results from not adjusting (being worse off ~20% of the time) look like what happens when you add more risk.

Implementation: How do you figure out what is the government’s portion of your RRSP? First, you have to figure out what the government’s share is — what the average tax rate of your withdrawal will be. For the case of the ultra-wealthy investor who is stuck fully in the highest tax bracket, it’ll be the highest tax bracket with no uncertainty. For someone more middle class, it may be your current marginal tax bracket, or a bit more or less than that. Your future RRSP withdrawals may span a few brackets.

Remember that you will never be able to peg this exactly. I suggest that using a round number that’s close enough — 30%, say — is likely a better way to go than to get lost in an exercise in trying to project it any more accurately than that, and certainly better than no correction at all. While the difference in effective portfolio ratios can be meaningfully far off if you don’t adjust, they likely won’t be if you use 30% as your correction factor and it ends up that you pay say 33% tax on withdrawal.

So you take that tax rate, t, and you say that much belongs to the government. Amount in RRSP (R) * t = government’s portion. To adjust your allocation, you can focus on just your part: (1-t).

So your portfolio may have a non-registered component (X) and a TFSA (Y). The total of your money (assuming we’re not going to go down the rabbit hole of also adjusting the non-registered amount) is X + Y + (1-t)*R.

Allocate based on that amount, then gross-up the RRSP with the government’s portion by dividing your after-tax RRSP amount by your tax adjustment factor (1-t) so you know how much to buy in the RRSP for each asset class.

So for example, if you’re sitting with $100,000 in your RRSP, $50,000 in your TFSA, and $35,000 in your non-registered, and you’ll use 30% as your close-enough tax rate:

Government’s portion of RRSP = 100,000 * 30% = $30,000
Your portion of RRSP = $100,000 * (1 – 0.3) = $70,000

Your portion of your portfolio = $70,000 + $50,000 + $35,000 = $155,000

Then if you want to aim for a 60:40 portfolio, you’d have 40% of $155,000 = $62,000 in bonds, the other $93,000 in stocks.

You’d start to fill your buckets:

Stocks in the TFSA = $50,000. Still have $43,000 of stocks to allocate, which go in the RRSP.

So now you have $43,000 of your money in stocks in the RRSP, and $70,000 of your money as total room in there. $27,000 of bonds go in the RRSP. Your RRSP is now 38.5% bonds — we’ll keep that in mind when we go to add back the government’s portion.

Finally, you have $35,000 of bonds for your non-registered account.

Before you can go and make the trades, you’ll have to gross up your RRSP again with the government’s portion (their $30,000). You’ll allocate it the same as your money in the RRSP, so 38.5% * $30,000 = $11,550 to bonds, $18,450 to stocks for the government portion. Or, same math a different way: RRSP_stocks_total = $43,000 [your money] / (1 – 0.3) = $61,450, RRSP_bonds_total = $27,000 / (1 – 0.3) = $38,550 [rounded to the nearest $50].

Your final, after-tax-adjusted asset-located portfolio is:
RRSP: $61,450 in stocks, $38,550 in bonds [$100,000 total]
TFSA: $50,000 in stocks
Non-reg: $35,000 in bonds.

Go make some trades and you’re done. (Or, you know, throw your hands in the air and keep each account as an undifferentiated copy of the whole, like a perfect atom because you’re not going to mess around with the subatomic physics nonsense that this involves: particle accelerators are for nerds and supervillains).

References:

I wrote the first draft of this before stumbling across these references (h/t Justin Bender), which have already trod the same ground:

W Reichenstein, Asset Allocation and Asset Location Decisions Revisited, The Journal of Wealth Management, 2001

“We argue that the traditional approach to calculating an individual’s asset allocation is wrong. The traditional approach fails to distinguish between the before-tax funds in deductible pension accounts and the generally after-tax funds in taxable accounts. Goods and services must be purchased with after-tax funds. Yet, the traditional approach treats $1,000 in deductible pension accounts as equivalent to $1,000 of after-tax funds in taxable accounts. I believe that the profession must first convert before-tax funds to after-tax funds, and then calculate the asset allocation based on after-tax funds… In Portfolios A and B, he could withdraw, say, $1,000 from the stock fund and buy $1,000 of goods and services. In Portfolio C, he must withdraw $1,538 from the stock fund held in a deductible pension account to buy $1,000 of goods and services; taxes consume the other $538… we can convert the $153,800 of before-tax funds to after-tax funds by multiplying by (1 – t), where t is his expected tax rate during retirement. The deductible pension account represents $153,800 (1 – 0.35) or $100,000 of after-tax funds.”

C Reed, Rethinking Asset Location – Between Tax-Deferred, Tax-Exempt and Taxable Accounts (2015). Available at SSRN.

Also, I deep-linked some PWL white papers above, because I don’t know how to link to a specific paper otherwise. Their collection of whitepapers is available here.

Future work: Using more realistic methods (portfolio with 4 components, withholding taxes, and compounding/deferring over multiple years), what’s the tax savings to be had with a stocks-sheltered-first, adjusted-for-taxes approach. And, how does it change if you’re not in the highest tax bracket?


7 Responses to “Asset Location Gets REALLY Complex”

  1. Steve Bridge Says:

    John, this is absolutely superb stuff, thank you very much for taking the time and effort to do this.

    I really liked the part on how small the savings were from an ‘optimal’ location of investments- it is easy to bang your head on your desk trying to figure this out.

    Steve

  2. Gerry Says:

    John,

    This is truly super stuff — thanks for sharing your insights. I had to read through your post a few times over a period of a couple of weeks but I think I mostly get it now. (Much easier to work through than the Retail Investor’s site.) I’m not going to sell everything and restructure my accounts accordingly (and perhaps having HBB in the non registered account complicates the story a little), but you’ve convinced me that my current system isn’t TOO terribly far from optimal, and optimal is blurry anyway, so I can sleep peacefully. There’s something seductive about trying to create the perfect asset location scheme, but at some point you just have to give it a rest, hey?

    Gerry

  3. Grant Says:

    To add another wrinkle to this, I like Justin Bender’s idea of putting Canadian/US/International equities in equal portions in your TFSA, because expected returns overwhelms the foreign dividend withholding tax issue for foreign equities, and you cannot know which of the 3 will do better going forward, so hence 1/3 each.

  4. Mark Says:

    This article is fantastic, thanks for posting.

  5. Potato Says:

    Thanks everyone!

  6. Potato Says:

    Justin Bender just went through this, and it can sometimes be helpful to see the same concept expressed in different words from a different voice, so be sure to head over there.

    On that note, here’s another reminder that having a higher nominal exposure to risky assets in your RRSP will have a psychology impact:

    “Although I would like to think that all investors are rational (i.e. they realize that Option #1 (a) is identical to Option #1 (b)), the next market downturn may cause Investor #1 (b) to abandon their investment plan, while Investor #1 (a) may keep a much more level head during the market drop.”

  7. Podcasts Spring 2018 | The Value of Simple Says:

    […] Over on the blog, I had a post on how complex asset location decisions can get. Justin Bender also covered the same topic. […]