Wednesday, May 23, 2018

Asset Location Errors

Deciding how to split your stocks, bonds, and REITs across your RRSPs, TFSAs, and taxable accounts can be confusing. Even well-known authors and web sites such as Gordon Pape, Robert Brown, and 5iResearch make basic mistakes. The best way I know to prevent some fundamental mistakes is to think of your RRSP as partially belonging to the government. Recently, Canadian Couch Potato portrayed this way of thinking as not worth the complexity for a small additional return (see the sixth question here). I disagree. I think it prevents some larger mistakes.


5iResearch claims that you shouldn’t hold “High Growth Equity” in your RRSP, and that TFSAs are the best place for such assets. Their reasoning is that RRSP “gains withdrawn from the account are treated as ordinary income,” and that in TFSAs “Profits on the stock will not incur capital gains tax when realized, nor will there be tax on the appreciated capital when withdrawn from the account.” So, if you compare investing $10,000 in either your RRSP or TFSA in high growth equity, the TFSA will seem better, because you end up with more after-tax money.

This bad advice comes from failing to think of RRSP assets as partially belonging to the government. Suppose you’ll pay an average of 20% tax on all RRSP withdrawals. Then you should be comparing investing $8000 inside a TFSA to investing $10,000 inside an RRSP. In the case of high growth equity, the two investments will work out the same after tax, so there’s no reason to prefer one over the other.

Gordon Pape made a similar mistake in his book Tax-Free Savings Accounts – How TFSAs Can Make You Rich. He prefers to hold U.S. dividend-paying stocks in a TFSA where they will pay only a 15% dividend-withholding tax, instead of having to pay your full marginal rate in an RRSP.

When you understand that 20% of your RRSP belongs to the government, and you compare an $8000 TFSA investment to a $10,000 RRSP investment in U.S. dividend-paying stocks, you’ll come to the correct conclusion that the RRSP is the better place in this case.

Gordon Pape made another error of this type when he answered a question about whether to start withdrawing first from an RRSP or a TFSA. Once again, understanding that your RRSP is not all yours solves the problem.

In his book Wealthing Like Rabbits, Robert R. Brown made a mistake with some advice on whether low-income savers should put their savings in an RRSP or TFSA. Brown prefers RRSPs because “with RRSPs they get the tax savings now, when they are younger, lower income earners and they likely need it more.”

He is comparing equal-sized contributions to an RRSP or a TFSA, which is a mistake. For a 20% marginal tax rate, he should be comparing a $4000 TFSA contribution to a $5000 RRSP contribution. Assuming the low-income saver had $5000 available to save, both scenarios lead to being left with $1000 left, either immediately, or after getting a tax refund. If the low-income saver expects to have higher income in the future, the TFSA is actually the better choice right now.

Another type of mistake comes as we get closer to retirement. People try to calculate their “retirement magic number” based on the 4% rule or some other rule. To know when you have enough for retirement, you must take into account income taxes. This means reducing your RRSP assets by your expected tax rate and taking into account capital gains taxes in your taxable accounts.


It’s certainly more complex to take into account taxes on RRSP withdrawals when managing your portfolio. That’s why I have a spreadsheet that does all the calculations for me. Even when rebalancing my asset classes, the spreadsheet shows me how much to buy and sell in each type of account factoring in taxes. I worked these things out once, and now I don’t have to worry about it.

One simplification I made was to treat my RRSP withdrawals as having a fixed tax rate. It’s possible that if my investments do very well, I’ll pay a higher tax rate. It’s also possible that the government will change tax rates. But I think it’s better to use an estimated tax rate than to implicitly use a 0% rate that is surely wrong.

On the subject of taking into account RRSP taxes, Canadian Couch Potato (CCP) says “we don’t manage portfolios that way because it is hopelessly impractical.” This is an overstatement. It would be difficult to do this properly if you had to do it yourself with a calculator for every client, but anyone running a financial advice business should be able to invest in some software or spreadsheets to get this stuff right. I’ve done it for my own narrow set of requirements. It should be possible for a business to cover a wider range of scenarios correctly with some automated tools.

CCP goes on to say he suspects that trying to take into account RRSP taxes in portfolio allocations “will defeat most DIY investors.” This is likely true. I think DIY investors are better off automating their portfolio decisions as much as possible, but few do this. They make essentially active decisions frequently, and adding the complexity of tax calculations increases the odds of making mistakes.


CCP sees little benefit to factoring in RRSP taxes. “If you lost half of the value of your RRSP, would it make you feel better to know that 30% would have gone to taxes in retirement anyway? I doubt it.” He quotes John Robertson as saying “all these optimization games can bring is a few basis points of extra return,” and CCP concludes that “the added complexity is not worth it.”

If these were the only benefits of taking into account RRSP taxes in asset allocation decisions and computing net worth, then I’d agree. But we need to avoid the bigger mistakes described above that are made by many including well-known figures in the financial world.

Note that I’m not talking about completely optimizing asset location decisions as discussed by John Robertson. He’s right that perfect optimization across all scenarios is very complex. However, each advisor or DIY investor uses some process to make “good enough” asset location decisions. Taking into account RRSP taxes with this good enough process adds only a modest amount of complexity to automated tools.

CCP says “For professionals managing portfolios for multiple clients [accounting for RRSP taxes] is close to impossible, and I am not aware of any firm that does so.” I don’t believe this is anywhere near impossible. All that is required is a one-time effort to include it in some automated tools. I think it’s about time for those selling financial advice to calculate portfolio allocations and net worth after subtracting estimated taxes.


  1. Consider someone who has $10,000 in bonds in their TFSA and $10,000 in stocks in their RRSP.

    If they sell the bonds and buy stocks in their TFSA, and then sell the stocks and buy bonds in their RRSP, they have not changed their portfolio allocation or the portion that belongs to the government.

    And yet if the stocks go on to return more than the bonds, have they not changed the amount of taxes they will pay on withdrawal from the RRSP?

    1. @Richard: It's not true that this person didn't change portfolio allocation. Assuming 20% taxes on RRSP/RRIF withdrawals, their portion of the RRSP is only $8000, and the other $2000 belongs to the government. The transaction you described changed the stock allocation from 8/18 (45%) to 10/18 (56%).

      This is the same error made by Pape, Brown, 5iResearch, and dozens of others I've read. A portion of RRSPs belongs to the government and it makes a difference in working out asset location strategies.

    2. It is true that if you adjust the allocations for the eventual RRSP taxes then the total after-tax long-term return looks the same. Interestingly it does lead to paying a lot less taxes (the same as you would if you didn't adjust the allocations). Usually paying less taxes will leave you better off so that may affect the way it's perceived.

      Luckily I don't have to think about it that much since I don't own bonds.

    3. @Richard: Asset location choices are simpler when you only own stocks. But note that one of Pape's errors I described in the article applied to an all-stock portfolio. He decided U.S. dividend-paying stocks are better in a TFSA when they're actually better in an RRSP. The 5iResearch error also applies to an all-stock portfolio.

    4. I'll continue to make plans that result in paying the least expected taxes. It's possible that in some scenarios it makes no difference to me. But it's also possible that the tax rate is not flat, and paying more taxes results in paying taxes at a higher rate. The idea of building a large portfolio in a TFSA has a very strong appeal.

    5. @Richard: You'll pay the least taxes if you go to cash in all your accounts. That's not what you meant, but it shows that minimizing taxes cannot be the only rule to follow.

      Personally, I seek to maximize my after-tax returns. I'm indifferent to how much the government gets (subject to the constraint that I get as much as possible).

    6. Maximize returns first, minimize taxes second :)

      I wonder if the RRSP rules might be changed. The TFSA was cut back because the higher limits can (presumably) only be used by people with higher incomes. With an RRSP, having a higher income not only gives you a higher limit but also saves you a higher marginal tax rate.

    7. @Michael, I disagree with your first reply. You have not changed the asset allocation, because the government doesn't care which asset you use to pay them. Your response implies somehow that they lay claim to specifically what's in your RRSP, and they do not. Put another way, the day before you intend to withdraw, you could simply change you asset allocation to anything you want, with no implications other than a trading commission. So I think it is a mistake to look at it the way you are. The govt has an overall claim to some percentage of your assets (the value of which is equal to the taxable portion of your RRSP). This is true. But it does not impact your asset allocation, because you can choose at any time to pay them with whatever assets you wish, via rebalancing.

    8. @Anonymous: The facts that assets are liquid and money is fungible doesn't change anything. By similar reasoning, I could say that the government isn't taking away 1/3 of my income because I could pay my income taxes by selling my house.

      In Richard's example, the swap of bonds and stocks between RRSP and TFSA makes a real change in the exposure of my future after-tax income to the returns of stocks vs. bonds. Therefore, my asset allocation changed.

    9. @Michael: I don't agree here. The fact that assets are liquid and money is fungible is exactly my point. The fact that you owe money does not impact your ability to independently control asset allocation**

      Your analogy is off the mark. I am not suggesting that the government isn't owed a portion. Of course they are. I am asserting, exactly contrary to your example, that they don't care how you pay them - so long as you do.

      They are owed an amount, which is equal to the taxable portion of the value of your RRSP account, but they are totally indifferent to your asset allocation. You are asserting above, in contrast, that they are somehow enforcing an asset allocation after the deregistration, or otherwise laying specific claim to certain asset CLASSES held in a certain account.

      Lets use the example above. 10k RRSP, 10k TFSA, 20% tax bracket. We all agree that $2k tax is owed. You are claiming that it should be factored into your asset allocation prior to sale. I am saying it should not.

      So now we deregister the RRSP completely. We owe 2k in tax. So now we have $18k left, consisting of 10k TFSA and 8k outside the RRSP. At the moment of deregistration, you are right that my asset allocation is 8/18 (45%) to 10/18 (56%). But within seconds, for a cost of $10 or so, I can put that that back to 9/18 each.

      So while your view was perhaps correct looking only at the moment, it's ignoring the obvious next step, since you can adjust at any time (and for all practical means)

      I am not sure the advantage of ignoring this step. It seems to add complexity to how you view the portfolio for no reason.

      One needs to know and understand that you "own" less than the nominal value of your RRSP, and factor that into retirement calculations. But there's no reason to factor the amount you owe into your asset allocation for the years prior, those things are independent.**

      **The exception to this may be someone with a very large unregistered account, or complex portfolio with many holdings, such that the post-deregistration rebalance above triggers a large cap gain or leads to massive trading fees. But I don't think this is the average situation.

    10. @Anonymous: In your example, if your tax rate is 20%, your economic interest in the RRSP is 80% of its value today, tomorrow, and any other day in the future regardless of how much it grows or shrinks. To treat it as though you own the whole thing makes no sense. Only 80% of it should be accounted for in your asset allocation.

      It's certainly possible to account for taxes differently and make correct choices. But I offer the endless list of online commentators who get these choices wrong as evidence that it's better to look at it the way I do.

    11. @Anonymous money is fungible, and you are certainly correct that the tax bill could be paid from any source of capital. The tricky thing is that the tax bill will scale with whatever asset is held in the RRSP account. If the intention is to stick to an asst allocation it is only sensible to treat the RRSP assets as after-tax. Justin Bender recently did a proof of this

  2. RRSP withdrawals getting taxed at a high rate would be a good problem to have.

    1. @aB: True. That's why I don't worry about it much. I assume a tax rate on RRSP/RRIF withdrawals based on slightly conservative investment outcomes, and if I end up paying a little more, it's because I'm richer than I expected to be and there's no problem, or as you say "a good problem to have."

  3. Good post Michael.

    I started typing up a big explanation of why the 5iResearch claim actually had some validity. Then as I worked through it, I realized you were correct that it is false claim.

    Only recently I had come to the conclusion that in order for a multi-account portfolio to be balanced, you had to increase the amount of investments in the RRSP to account for the average tax rate on withdrawals before comparing to amounts in TFSA's. When I did come to this conclusion, I was shocked this wasn't more widely "advertised" in books, newspaper articles and blogs. Then I was glad to stumble on the John Robertson explanation you linked to, which I wholeheartedly agree with.

    But since coming to this conclusion, I had forgotten to re-evaluate my previous misconception that it makes sense to hold lower expected return investments in my RRSP and higher expected return in my TFSA to maximize my whole portfolio's expected return. Without understanding the need to adjust account values based on withdrawal tax rates, the 5iResearch claim does hold water.

    While considering the (unknown) withdrawal tax rate adds a form of complexity, it's nice that it removes the complexity of considering expected return when deciding which account to hold the investment.

    Thanks for getting me to think!

    1. I agree that failing to account for the government owned part of your RRSP is a common error and can results in asset location errors. Understanding this issue also helps to appreciate the need to consider both tax efficiency and expected returns when it comes to asset location decisions. However when it comes to retirement and the 4% rule, does it really matter if you do not tax adjust your portfolio, as long as you taxes into account in your retirement cash flow projections? Eg. If you have a $2M portfolio, including an RRSP, you can withdraw $80K a year from the portfolio. Knowing the amount in the RRSP etc. you calculate $X of taxes will be payable. If your cash flow needs are $80K-$x per year, why not do it that way?
      Btw, I hope you are enjoying your retirement, and intend to continue with this blog as part of your post retirement activities.

    2. @Return Reaper and @Grant: No matter how you approach answering these questions, it's possible to come to the right answer. I find that the easiest way to get the answers right consistently is to think of part of your RRSP belonging to the government. But Grant's way of thinking about the 4% rule can work. It's just that I see others answering different questions incorrectly when they don't factor in taxes properly.

  4. I agree that income taxes must be taken into account. This applies to all aspects of investing, not just RRSP withdrawals. Having said that, I have a couple of comments about this post.

    First, assuming a 20% average tax rate on withdrawals seems like it would be too high for many people (of course, everyone has unique circumstances). If you assume that a person in Ontario has no source of income other than RRSP withdrawals (and no deductions), you don't hit a 20% average tax rate until around $70,000 is withdrawn. I'm guessing that most people will not find themselves in that position. If you reduce the withdrawal to $30,000 then the average tax rate drops to 12.4%. If this is supplemented with eligible dividends received in a non-registered account, the average tax rate could drop even lower since the dividends would have a negative tax rate. All this to say, everyone will have unique circumstances and assuming a 20% average tax rate may make the issue appear larger than it really is.

    Second, on the issue of asset location - does it make sense to distinguish between the accumulation phase and the drawdown phase? The government does not own any percentage of my RRSP until I make a withdrawal and (potentially) have to pay tax. In the meantime, I seek to maximize total return while also maintaining the desried risk profile. Again, the government does not own a percentage of my investments during this phase. When I start to make withdrawals, I can see the issue that you raise. However, I can adjust my asset location at that time (i.e., after accumulation). Admittedly, I have not tried to do any calculations to test this - it's what strikes me as "common sense", which can be dangerous.

    1. Regarding the average tax rate, all forms of taxation should be considered, including things such as GIS and OAS claw-backs. The GIS clawback makes your first income dollars in retirement effectively taxed at 50%. In fact, I think when you consider all forms of taxation on income, the marginal tax rate is never below 20%, which implies 20% is too low for an average tax rate.

      One blog post on this subject is:

      It is rather old and may not be entirely up to date. Also, IIRC there may have been a mistake in it -- I think it over estimates the actual effect of the age amount, but I'm not certain (it's been a long time since I've reviewed it). Nevertheless, I think the concept is important and should be considered.

      @Michael, do you recall if you've ever done a post on effect tax rates in retirement?

      Also, while it is true you don't pay tax until you withdraw it, the fact that in order to actually get the money until you pay tax on it means that it isn't all yours. So I think considering that some portion of it isn't yours for the entire investment time horizon is valid.

    2. @Mark B: As @Returns Reaper explained, calculating effective tax rates at low incomes can get very tricky when you start factoring in the loss of income-tested benefits like GIS. For people with other types of income (like dividends, OAS, CPP) that gets into the $20k or so range, RRSP/RRIF withdrawals bringing income up to about $43k in Ontario are taxed at 20.05%. You'll see other articles using example tax rates of 30% or even 40%. I assume this is because the article is written by some sort of financial advisor who works primarily with high net worth clients.

      My initial reaction to the question of whether it makes sense to distinguish between the accumulation and draw-down phases is no. If I'm destined to pay 20% tax on RRSP/RRIF withdrawals in the future, then this is the same as 20% of my RRSP belonging to the government and growing alongside my 80% share (that will be untaxed). However, ways of thinking about these things are neither right nor wrong. We're only right or wrong when we use our way of thinking to make a choice. I'd have to see how your way of thinking translates into the choices you'd make to have an opinion on whether you're making good choices.

      @Returns Reaper: Tax rates in retirement can get tricky. I think I've written a few things about tax rates in retirement, but nothing thorough. Try searching for GIS in my blog's search.

  5. Another thoughtful post that I really enjoyed reading (as usual for your blog!). Though I'm not sure I totally agree with this one.

    I think in your post, as well as some of the others you linked, there are some assumptions being made, but not acknowledged. I think there are three important things that we need to agree to here.

    1) One cannot say for certain what his tax rate will be at the time of withdrawing RRSPs. We may guess or estimate or plan - zero is a possibility - but at the end of the day, we don't know. However...

    2) We may assume that we will have SOME ability to control that rate (by timing withdrawals, when we leave the workforce, taking other income, etc). This ability will not be limitless, but it will likely not be zero. And..

    3) All else being the same, someone with a higher value inside his RRSP is almost certain to pay the same or more tax on withdrawal than someone with a smaller value (and say, more in a TFSA).

    With those in mind, I think the way to look at this is:

    a) Whether to contribute to an RRSP or a TFSA should be based on an educated guess as to whether you think you'll be able to stay in a lower tax bracket on withdrawal. 30 years out, that's pretty tough to say. But if you're in the max (>50% bracket) today, it doesn't seem like you have a lot to lose. Conversely, if you're in the low (<10% bracket) today, not much to gain.

    b) On the question of locating growth assets in a TFSA, I disagree with you and would favor locating growth assets inside a TFSA. I understand your tax argument, but the way to look at this is, when I retire, would I rather end up with a bigger RRSP and smaller TFSA, or smaller RRSP and bigger TFSA. The second is almost guaranteed to leave you with more flexibility at the time, and should be the preferred outcome.

    c) On the idea of whether to analyze RRSP returns pre- or post-tax, I'm kind of agnostic here. I get CCP's argument, it is valid. Your spreadsheet approach is good too, though it assumes no flexibility & ability to control, which may not be right either.

    1. @Anonymous: I disagree on a few points.

      1) We may not be certain of the RRSP/RRIF withdrawal tax rate, but zero is almost certainly wrong. It's better to use a guess than to use zero implicitly.

      2) Yes, people have some ability to affect their RRSP/RRIF withdrawal tax rate. They should take this into account when they guess the tax rate they'll end up with.

      3) The focus should be on your after-tax gains, not how much you pay in taxes. These two ways of thinking don't end up with the same results.

      a) Agree.

      b) If your RRSP/RRIF withdrawal tax rate is fixed, then you shouldn't care which one grows larger. If a larger RRSP would increase your overall RRSP/RRIF withdrawal tax rate, then you'd prefer to grow your TFSA more than grow your RRSP. But in the analyses I've done, this effect is small.

      c) You seem to be taking both sides here. The 3 possibilities you're considering are i) treat the tax rate as zero, ii) treat the tax rate as a fixed percentage, and iii) treat tax rates as the complex graduated system we have. These are on a continuum from least accurate to most accurate (but also most complex). How can we accept i) as valid but reject ii) in favour of iii)?

  6. The exchange below is reproduced to remove broken links.

    ----- BHCh May 23, 2018 at 8:45 PM

    I tend to use RRSP for US ETFs to minimize withholding taxes. Also use non registered for fixed income, also to minimize taxes (lower expected growth). Canadian ETFs go into a TFSA or non-reg, again to lower tax.

    Long term ETFs have similar expected return, regardless of the region. For this reason I don’t really care about subtracting future taxes when rebalancing.

    ----- Michael James May 23, 2018 at 10:04 PM

    @BHCh: My approach is very similar. I do discount RRSPs for future taxes and non-registered for future capital gains taxes, but as long as you aren't making active market-timing choices about whether to hold one region or another in RRSP instead of TFSA, I doubt you're hurting your expected returns.