Problems with Reichenstein's "Asset Location Decis Revisted"

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
less
Posts: 158
Joined: Mon Sep 16, 2013 12:38 pm

Problems with Reichenstein's "Asset Location Decis Revisted"

Post by less » Sun Nov 10, 2013 12:20 pm

This post concerns the paper in the November Journal of Financial Planning http://www.fpanet.org/journal/CurrentIs ... fContents/ . The author claims the paper concerns Asset Location AL but I will argue that it concerns Asset Allocation AA much more, and ignores AL. http://www.fpanet.org/journal/TheAssetL ... Revisited/

The paper restates essentially the same thing Mr Reichestein has been saying for a decade. He separates thinking into two lines -
a) Those who think the objective of Asset Allocation AA is to manage risk, and that the objective of Asset Location is to maximize wealth, and
b) those who think the objectives of both AA and AL are to position a portfolio along the Efficient Frontier to maximize Utility (returns for your risk tolerance).
He is in the second camp and most all the references are back to his own previous papers. I am in the first camp.

1) He argues that "The objective of asset location studies is to determine which strategy is better holding everything else constant". As far as I can see that is pure invention. But it deflects from the huge shortcoming of his argument - that his process makes no effort to maximize the benefits of the tax shelter accounts. He simply ignores all those benefits.

2) His first section (expanded later in an example) clarifies that Tax Deferred Accounts TDAs are partially funded with a loan from the government that they ask you to invest for them. This portion is never really your own money because they take it back when you close the account. So using the diagram below, what is yours at the start is only the $560, not the full $1,000. Therefore $1,000 in a TDA is equivalent (in terms of your own wealth) to $560 in a Taxable account.

Image

3) Past papers have failed to make this adjustment (such as the Vanguard paper http://personal.vanguard.com/pdf/s556.pdf). But he igores the recent "Rethinking Asset Location - Between Tax-Deferred, Tax-Exempt and Taxable Accounts" paper http://papers.ssrn.com/sol3/papers.cfm? ... id=2317970 discussed on another recent thread - which does not make that error. Most people on this site agree that for AA calculations the value of TDAs should be discounted for the portion 'owned' by the government.

4) He also found fault with the math proofs of past papers that modeled no re-balancing between assets. That is a valid objection, but he ignored the "Rethinking Asset Location" paper that showed that yearly-reblancing confirms the same ranking as the zero-reblancing model, given time.

5) Next he addresses Taxable accounts. The taxes of Taxable accounts reduce after-tax profits and also after-tax losses. So the volatility of after-tax returns is dampened. The investor gets reduced returns and reduced risk (as measured by after-tax return volatility). This is the linch-pin of his idea - that investments in Taxable accounts are less risky and therefore Asset Allocations must take this into account.

Note that he does not say that the value of Taxable accounts should be discounted (in the same way TDAs are discounted) in the calculations for AA. That is an idea on this site's Wiki that most probably was derived from Reichenstein, but he does not say that.

He ignores the detractor's arguments in "Rethinking Asset Location" http://papers.ssrn.com/sol3/papers.cfm? ... id=2317970 and chooses to not reference the paper at all. It argues that his ideas are theoretical and intellectual only. Taxes are never paid from investment accounts. There is no mechanism for reducing the value of Taxable accounts for taxes paid from a chequing account. Investors never 'feel' the damping effects of taxes so their gut reactions to price volatility are determined by the before-tax returns, not the after-tax returns. No adjustment in AA is needed because investors' emotional reactions to volatility are not changed by taxes paid from another account 9 months later.

When you separate the different purposes of Asset Allocation vs Asset Location - you are essentially treating ALL income in ALL accounts as if taxed - and AA as if all the assets are the same. The AL decision works backwards from that fully-taxed state - maximizing the reduction in taxes.

6) Then he presents his replacement for traditional AA and AL - mean-variance optimization MVO. This process tries to combine assets in such a way as to increase the portfolio's risk-adjusted return - to position the investor along the Efficient Frontier. For every asset type owned he creates a doppleganger - one with before-tax metrics and the other with after-tax metrics. He inputs these as if different asset-types and allows the MVO to determine how much, and where, of each asset-type.

Image

a) There are a lot of people claiming that MVO is just another academic invention with no value. See the paper http://papers.ssrn.com/sol3/papers.cfm? ... id=2349312 "Optimal Decisions vs. Simple Heuristics".
b) MVO is not something for the DIY. You need lots of data not commonly available. You need the software to manipulate it. So in effect his idea is for everyone to pay an advisor to do this. That probably explains why this magazine of the CFPs has pushed this methodology twice in 4 months.
c) This process ignores all the possible benefits of the tax shelters.
d) It models comparative yearly returns. The paper he ignores deconstructs the benefits of the tax shelters and shows that 'Time' is a factor - that the relative benefits of one year can be reversed over time. (edit moved diagram from f) to here. It was originally wrongly placed)
Image
e) MVO provides no information on which assets to prioritize in TEAs vs TDAs.
f) MVO wrongly gives the same Asset Location regardless if the owner faces a bonus or penalty from a lower/higher tax rate on withdrawal.

7) The only new material in this paper is where he then presents a variety of his black-box results for different input assumptions. If you don't accept his premise then the results are meaningless.
Last edited by less on Mon Nov 11, 2013 11:08 am, edited 3 times in total.

staythecourse
Posts: 6993
Joined: Mon Jan 03, 2011 9:40 am

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by staythecourse » Sun Nov 10, 2013 12:43 pm

Thank you for the link and thoughtful analysis.

Will have to read both and comment after.

Good luck.
"The stock market [fluctuation], therefore, is noise. A giant distraction from the business of investing.” | -Jack Bogle

User avatar
grabiner
Advisory Board
Posts: 27162
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by grabiner » Sun Nov 10, 2013 8:10 pm

less wrote:This post concerns the paper in the November Journal of Financial Planning http://www.fpanet.org/journal/CurrentIs ... fContents/ . The author claims the paper concerns Asset Location AL but I will argue that it concerns Asset Allocation AA much more, and ignores AL. http://www.fpanet.org/journal/TheAssetL ... Revisited/

The paper restates essentially the same thing Mr Reichestein has been saying for a decade. He separates thinking into two lines -
a) Those who think the objective of Asset Allocation AA is to manage risk, and that the objective of Asset Location is to maximize wealth, and
b) those who think the objectives of both AA and AL are to position a portfolio along the Efficient Frontier to maximize Utility (returns for your risk tolerance).
He is in the second camp and most all the references are back to his own previous papers. I am in the first camp.
This doesn't appear to be the distinction he makes. All of the studies consider both risk and return; the first camp looks at after-tax value and pre-tax risk (or treats asset allocation as a given, which is essentially equivalent), and the second camp looks at after-tax value and after-tax risk.

So should you use pre-tax or after-tax risk? If you look only at the raw dollar numbers, then asset location does not affect your risk perception, and you should put your highest-risk assets in a Roth IRA because you get more benefit for no more risk. But if you look at after-tax value, then asset location does affect risk.

And I believe it makes sense to look at after-tax value when determining risk. Your mathematical risk tolerance (the effect of low returns on your standard of living) is determined by after-tax value. Ideally, you can set up your investment view so that even your emotional risk is driven by the after-tax value, either because you understand the principle or because you look at the numbers properly.

I have my taxable balances and retirement plan balances on different forms because I have accounts with several different providers. When I get a statement, I copy its values into the spreadsheet. Therefore, I never get a good look at the total dollar value of my portfolio. The spreadsheet I use doesn't add the total dollar values; if my taxable account has lost $10,000, the bottom line on the spreadsheet decreases by $7200.
5) Next he addresses Taxable accounts. The taxes of Taxable accounts reduce after-tax profits and also after-tax losses. So the volatility of after-tax returns is dampened. The investor gets reduced returns and reduced risk (as measured by after-tax return volatility). This is the linch-pin of his idea - that investments in Taxable accounts are less risky and therefore Asset Allocations must take this into account.

Note that he does not say that the value of Taxable accounts should be discounted (in the same way TDAs are discounted) in the calculations for AA. That is an idea on this site's Wiki that most probably was derived from Reichenstein, but he does not say that.
Wiki reference: Tax-Adjusted Asset Allocation

I don't credit Reichenstein with this particular idea; it is my own calculation, and I got it wrong in the first version of my spreadsheet. The reason I adjust my asset allocation for taxes is that I am attempting to optimize returns. If gaining or losing $10,000 in my taxable account has the same effect on my retirement portfolio as gaining or losing $7200 in my Roth IRA, then I should reduce my taxable account by 28% for that purpose. The final after-tax value of my portfolio requires a different adjustment, because the money in my taxable account and retirement plans will not all by taxed at my marginal rate.
He ignores the detractor's arguments in "Rethinking Asset Location" http://papers.ssrn.com/sol3/papers.cfm? ... id=2317970 and chooses to not reference the paper at all. It argues that his ideas are theoretical and intellectual only. Taxes are never paid from investment accounts. There is no mechanism for reducing the value of Taxable accounts for taxes paid from a chequing account. Investors never 'feel' the damping effects of taxes so their gut reactions to price volatility are determined by the before-tax returns, not the after-tax returns. No adjustment in AA is needed because investors' emotional reactions to volatility are not changed by taxes paid from another account 9 months later.
And this is the key point in the disagreement. Most models are designed to mathematically optimize returns, and that does require a full adjustment, as discussed above.

There is one additional factor needed in order to justify the full adjustment: taxes have to come out of investment money, not spending money. If your investment generates a $5000 tax bill and you reduce your spending by $5000, the investment did not cost you anything in portfolio value (although it reduced your current standard of living). If your investment generates a $5000 tax bill and you keep the same standard of living but reduce your investments the next year by $5000 (or in the current year because you have more tax withheld in order to cover the tax bill), then the $5000 in taxes effectively reduced your investment value.
6) Then he presents his replacement for traditional AA and AL - mean-variance optimization MVO. This process tries to combine assets in such a way as to increase the portfolio's risk-adjusted return - to position the investor along the Efficient Frontier. For every asset type owned he creates a doppleganger - one with before-tax metrics and the other with after-tax metrics. He inputs these as if different asset-types and allows the MVO to determine how much, and where, of each asset-type.

a) There are a lot of people claiming that MVO is just another academic invention with no value. See the paper http://papers.ssrn.com/sol3/papers.cfm? ... id=2349312 "Optimal Decisions vs. Simple Heuristics".
b) MVO is not something for the DIY. You need lots of data not commonly available. You need the software to manipulate it. So in effect his idea is for everyone to pay an advisor to do this. That probably explains why this magazine of the CFPs has pushed this methodology twice in 4 months.
The purpose of this MVO model is not to recommend a particular strategy, but to make a theoretical decision. It is a well-known problem that an MVO is only as good as its inputs, and the inputs in this paper are theoretical. It does make sense to use an MVO to evaluate a model because this is the best way to do the calculations.
c) This process ignores all the possible benefits of the tax shelters.
I agree that this is a flaw in the process. He says, "To conform with an example from a recent study, assume that the ordinary income tax rate is 30 percent, the long-term capital gain tax rate is 15 percent, and pre-tax returns on bonds and stocks are 5 percent and 8 percent, respectively, with all stock returns in the form of capital gains that are realized in one year and one day so that the long-term capital gain tax rate applies." This loses a large part of the benefit of stocks in a taxable account, as it is more likely that a 2% dividend yield will be taxed annually at 15%, and the remainder will not be taxed until withdrawal.
e) MVO provides no information on which assets to prioritize in TEAs vs TDAs.
Here, the issue is the proper marginal tax rate, which can be part of any model. If you will retire in a 15% tax bracket, but you will be in the phase-in range for Social Security taxation, then your marginal tax rate in retirement is 27.75%, not 15%, and you should discount your TDA by 27.75% to get an equivalent risk profile. Once you have made the 27.75% adjustment, it no longer matters what you put in which account.

(It still does matter slightly for other reasons, which I mention occasionally but which I do not normally see in any of the papers. If you have high-returning assets in a TDA, you might be forced to take larger RMDs than you need, increasing the tax cost, and the RMDs might be taxed in a higher bracket than you expected. If you have high-returning assets in a TEA, the extra returns can grow tax-free until your death.)
f) MVO wrongly gives the same Asset Location regardless if the owner faces a bonus or penalty from a lower/higher tax rate on withdrawal.

Image
The bonus/penalty is not an asset location issue. Only the retirement tax rate is relevant for asset location or asset allocation, as the money will grow exactly the same no matter how it was taxed or deducted upon contribution. Two investors who will retire in a 15% tax bracket should use the same asset location even if one is currently in a 25% bracket (and thus has a bonus from tax deferral) and the other is currently in a 15% bracket (and thus has no bonus). In contrast, an investor in a 25% bracket who will retire in a 25% bracket and an investor in a 15% bracket who will retire in a 15% bracket should have different (at least nominal-dollar) allocations, as the investor in the 15% bracket will be able to keep more of his IRA withdrawals.

The bonus/penalty is only relevant to current contributions; the investor in a 15% bracket might prefer a Roth account (particularly if he can max it out), and the investor in a 25% bracket can afford to save more at the same tax cost.
Wiki David Grabiner

Topic Author
less
Posts: 158
Joined: Mon Sep 16, 2013 12:38 pm

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by less » Mon Nov 11, 2013 11:03 am

I really am left with no idea whether Grabiner is pro Reichenstein http://www.fpanet.org/journal/TheAssetL ... Revisited/ , or pro maximizing the account's benefits (either the Wiki way http://www.bogleheads.org/wiki/Principl ... _placement , or the Reed way http://papers.ssrn.com/sol3/papers.cfm? ... id=2317970 ).

1) I really don't think I misrepresent the two different POV on this issue - or disagree with Reichenstein's split. What variables are 'considered' is irrelevant. Everyone 'considers' the tax rate and the rate of return. It is how the variable is used in a decision metric, and most importantly, the different objectives that distinguish us. Unless one clarifies the objective all discussion can be circular.

* Everyone considers risk management an objective. Reichenstein uses both AA and AL to this end. Wiki and Reed use AA to manage risk.
* Reichenstein ignores completely the objective to maximize the benefits of the tax-shelter accounts.
* Wki and Reed both see the objective of AL to maximize wealth. They disagree on how this is accomplished. See point 5c) below

2) In this thread and a previous one Grabiner describes his own personal record keeping set up. It effectively converts all account values and profits to after-tax metrics, so he does not see or react to nominal values. He uses this as an argument that everyone else gets their data in this form. That everyone else reacts to after-tax profits because that is what they see, not nominal profits. That everyone else has set up a system so that income taxes paid on investments come from the investment account (he admits this is necessary for his model).

But 99.9% of other people have not set up their accounts this way. I certainly have not. Nor do I think people 'should' do so. There is a huge gulf between rational theory and what happens in reality. In reality people react to nominal account values and nominal gains/losses.

There is just no reason why traditional AA percentages cannot deal with the risks normal people feel and react to. There is no necessity to use AL in this process - just because of a theory that people should react to after-tax profits/losses.

3) I don't understand the defense of MVO "to evaluate a model" while not recommending it as "a particular strategy". Reichenstein states explicitly in other papers that all advisors should use MVO for their AA/AL decisions. He is not using it as an 'evaluation tool'. He is saying forget AL, forget AA rough estimates of asset-class percentages --- use MVO to replace both.

If Grabiner agrees that MVO doesn't work for AA then how can it be defended for replacing both AA and AL?

4) Grabiner agrees that Reichenstein's use of MVO instead of AL " ignores all the possible benefits of the tax shelters". But then changes the subject. This point is the elephant in the room. All possibility of maximizing those benefits is simply thrown away. A huge opportunity loss. And for what? So that MVO can more exactly place a person on the risk/return spectrum of the Efficient Frontier (MVO which is theoretical and does not work)? A negative sum trade-off.

5a) Grabiner's remaining points revert to disputes with Reed's math proofs.
He does not like the example situation chosen by Reichenstein because it assumes capital gains are triggered yearly. Who cares? You can use whatever variables you like. He clearly thinks that a 30 delay in capital gains results in an effective zero% tax rate for equity - proving it should be in the Taxable account.

But even using the variables Grabiner likes Reed's proof shows equity is better in a tax-shelter than debt. See steps at http://www.bogleheads.org/forum/viewtop ... 4#p1843884
Assume: Debt of 2.5% taxed at 30%, and
Equity with 2% dividend taxed at 15% and 7% capital gain realized after 30 years.
Rebalanced yearly, with equity in the shelter the ending value = $7,017 vs with debt = $6,477
Changing the holding period to 15 years, and that the withdrawal tax rate will be lower by 10%.. That results in a 21% difference between prioritizing equity ($7,851) vs debt ($6,496). In fact using those other assumptions you can reduce the tax rate on equity to Zero% and STILL equity is better in the shelter.

5b) Grabiner says "The bonus/penalty is not an asset location issue. Only the retirement tax rate is relevant for asset location." The math proof that this is wrong was presented by Reed, but ignored by Grabiner in the thread on Reed's paper. The deconstruction of benefits includes the Bonus/Penalty. That bonus//penalty is calculated by multiplying the difference in tax rates by the size of the account.

The difference in tax rates is determined by decisions/opportunities to move $$ in and out of the accounts - and also by the size of the account. The size of the account is determined by the rate of return of the assets inside - most definitely an AL decision.
Image

Grabiner's argument is that the math calculation of the bonus is frozen once the expectation of the withdrawal rate has changed. But the math disproves this. Argue the math.

5c) Both Reed and Grabiner agree (I believe) that the AL objective is to maximize wealth. Grabiner thinks the means to this end is " I am attempting to optimize returns." Or "Most models are designed to mathematically optimize returns". By which he means 'after-tax returns'.

As a result the Wiki uses for a decision metric the difference between the nominal return and the after-tax return (or the $$ tax per $invested). (I am assuming that is what is meant by 'tax efficiency'). Reed gave the math proof why this metric fails - but Grabiner ignored it. Reed math shows the way to maximize wealth is to maximize the accounts' benefits. And that a one-year view does not represent that benefit. That Time changes the benefit.
Image

Conclusion: Reichenstein ignores Reed's paper because it makes valid points with common sense judgements he cannot counter, and because it did not make the errors he chose to use as stray-man arguments against his detractors. Grabiner chooses to ignore Reed's math proofs that show the Wiki wrong.

User avatar
grabiner
Advisory Board
Posts: 27162
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by grabiner » Wed Nov 13, 2013 10:42 pm

less wrote:I really am left with no idea whether Grabiner is pro Reichenstein http://www.fpanet.org/journal/TheAssetL ... Revisited/ , or pro maximizing the account's benefits (either the Wiki way http://www.bogleheads.org/wiki/Principl ... _placement , or the Reed way http://papers.ssrn.com/sol3/papers.cfm? ... id=2317970 ).
I primarily agree with Reichenstein, although not all of his calculations; my contributions to the Wiki articles are based on my own understanding of the issues.
1) I really don't think I misrepresent the two different POV on this issue - or disagree with Reichenstein's split. What variables are 'considered' is irrelevant. Everyone 'considers' the tax rate and the rate of return. It is how the variable is used in a decision metric, and most importantly, the different objectives that distinguish us. Unless one clarifies the objective all discussion can be circular.

* Everyone considers risk management an objective. Reichenstein uses both AA and AL to this end. Wiki and Reed use AA to manage risk.
* Reichenstein ignores completely the objective to maximize the benefits of the tax-shelter accounts.
* Wki and Reed both see the objective of AL to maximize wealth. They disagree on how this is accomplished. See point 5c) below
There isn't a specific objective to maximize the benefits of the tax-sheltered accounts; the benefit is to use the tax-sheltered account to maximize (or, more accurately, to optimize) returns.
2) In this thread and a previous one Grabiner describes his own personal record keeping set up. It effectively converts all account values and profits to after-tax metrics, so he does not see or react to nominal values. He uses this as an argument that everyone else gets their data in this form. That everyone else reacts to after-tax profits because that is what they see, not nominal profits.
I don't argue that they do, but that they can. If you are using a mathematical formula to determine your strategy, it makes sense to use it correctly.

And this is clearer if your risk tolerance is not determined by your emotions, but by the practical consequences. If a portfolio decline will affect your standard of living in retirement, the effect depends on the after-tax loss, as you would not have been able to spend the full pre-tax loss unless the loss was in a tax-exempt account.
3) I don't understand the defense of MVO "to evaluate a model" while not recommending it as "a particular strategy". Reichenstein states explicitly in other papers that all advisors should use MVO for their AA/AL decisions. He is not using it as an 'evaluation tool'. He is saying forget AL, forget AA rough estimates of asset-class percentages --- use MVO to replace both.

If Grabiner agrees that MVO doesn't work for AA then how can it be defended for replacing both AA and AL?
MVOs are models, which are necessarily inaccurate, but you have to work with some type of model to draw any conclusion. A strategy based on a reasonable MVO is more likely to be right than a strategy based on expected value with no consideration of risk.

And as an example of this principle, I will use Reed's model, in which asset A earns 7% taxed at 15%, and asset B earns 3.5% taxed at 30%. The one-year tax cost is the same whichever asset is in a taxable account. The multi-year tax cost of holding A in a taxable account is higher, because putting A in a taxable account causes the taxable account to grow faster and thus reduces the benefit of tax deferral.

But that is based on the assumption that A will earn more than B; if you knew that would happen, you wouldn't hold B at all. A is a risky asset and has a wide range of returns. Putting A in a taxable account gives you some distribution of after-tax values with a lower mean than putting B in a taxable account does, but it isn't at all clear that you prefer the distribution with A in a taxable account.
4) Grabiner agrees that Reichenstein's use of MVO instead of AL " ignores all the possible benefits of the tax shelters". But then changes the subject. This point is the elephant in the room. All possibility of maximizing those benefits is simply thrown away. A huge opportunity loss.
I don't understand this point. The purpose of any model is to get the maximum benefit, and that is what Reichenstein is trying to do. I made an entirely different point, which is that Reichenstein's example computed the beneifts incorrectly.
5a) Grabiner's remaining points revert to disputes with Reed's math proofs.
He does not like the example situation chosen by Reichenstein because it assumes capital gains are triggered yearly. Who cares? You can use whatever variables you like. He clearly thinks that a 30 delay in capital gains results in an effective zero% tax rate for equity - proving it should be in the Taxable account.

But even using the variables Grabiner likes Reed's proof shows equity is better in a tax-shelter than debt. See steps at http://www.bogleheads.org/forum/viewtop ... 4#p1843884
Assume: Debt of 2.5% taxed at 30%, and
Equity with 2% dividend taxed at 15% and 7% capital gain realized after 30 years.
Rebalanced yearly, with equity in the shelter the ending value = $7,017 vs with debt = $6,477
Changing the holding period to 15 years, and that the withdrawal tax rate will be lower by 10%.. That results in a 21% difference between prioritizing equity ($7,851) vs debt ($6,496). In fact using those other assumptions you can reduce the tax rate on equity to Zero% and STILL equity is better in the shelter.
This is partly the result of not tax-adjusting the allocation.

I don't see the calculations from the spreadsheet you linked there, so I'll try the calculation myself. I am assuming no rebalancing, but this is reasonable if the investor also holds other assets in the tax-advantaged account and rebalances there, with the decision just whether to put the last $500 in taxable into bonds or stocks.

$500 bonds in taxable grows at 1.75% for 30 years, which is $883.
$500 stocks in tax-advantaged grows at 9% for 30 years, which is $6634.

$500 bonds in tax-advantaged grows at 2.5% for 30 years, which is $1049. The bond fund lost 16% of its value to taxes in a taxable account.
$500 stocks in taxable grows at 8.7% for 30-years, which is $6107, but 7/8.7 of the $5607 gain is a taxable capital gain of $4512, with tax $677 giving an after-tax value of $5430. The stock fund lost 18% of its value to taxes in a taxable account.

If the tax-advantaged account is tax-free, then adding the numbers gives $7517 with stocks in tax-free, and $6479 with stocks in taxable.

If the tax-advantaged account is taxed at 30% upon withdrawal, that costs $1990 in tax with stocks in tax-deferred, for a net of $5527, and $315 in tax with stocks in taxable, for a net of $6184. Now it appears that tax-sheltering the bonds comes out ahead.

If the tax-advantaged account is taxed at 20% upon withdrawal, that costs $1327 in tax with stocks in tax-deferred, for a net of $6190, and $210 in tax with stocks in taxable, for a net of $6269. This is about break-even.

And the tax adjustment of asset allocation is the whole point here. If you take these numbers at face value, they would indicate that you should hold stocks in taxable if your other account is tax-deferred, but bonds in taxable if your other account is tax-exempt. But if you make a tax adjustment, then you will get the same asset location conclusion either way.

Whether stocks or bonds are better in taxable under these assumptions depends on how much you discount the taxable account for tax purposes. If you do not discount the taxable account at all, so that $500 in taxable is viewed as equal to $500 in a tax-free account, then the first calculation is correct; assuming these returns, you will wind up with significantly more with $500 in stocks in tax-free and $500 in bonds in taxable than the other way around, and the same is true if you have $714 in a tax-deferred account (taxed at 30%) instead. However, I strongly disagree with the lack of an adjustment, because the asset allocation with stocks in tax-free is significantly riskier.

If you discount the taxable account by 20%, the third calculation is correct, and it is just about break-even. (And given the very low bond yields, that is a reasonable conclusion; these low-yielding bonds have about the same annualized tax cost as the stock fund.)
5b) Grabiner says "The bonus/penalty is not an asset location issue. Only the retirement tax rate is relevant for asset location." The math proof that this is wrong was presented by Reed, but ignored by Grabiner in the thread on Reed's paper. The deconstruction of benefits includes the Bonus/Penalty. That bonus//penalty is calculated by multiplying the difference in tax rates by the size of the account.
The math proof is a valid proof, but not of what you want to prove. The math correctly indicates the tax benefit (or cost) of putting assets in a tax-deferred account. If you expect to retire in a lower tax bracket, you should invest in tax-deferred accounts in preference to tax-exempt accounts.
Grabiner's argument is that the math calculation of the bonus is frozen once the expectation of the withdrawal rate has changed. But the math disproves this. Argue the math.
The bonus itself changes; if you retire in a lower tax bracket, then you gained a larger bonus from the tax deferral. But the bonus isn't the goal, because you don't spend the bonus; you spend the account balance.

Consider this example:

Alice and Bob have equal balances in their Roth, traditional, and taxable accounts. Both will retire in a 25% tax bracket. Alice is in a 25% bracket now, while Bob is in a 33% bracket.

Bob gains more from tax deferral than Alice does, but their portfolios will have the same after-tax value if invested identically. (This is true even in the taxable accounts, as long as they hold either municipal bonds taxed at 0%, or stocks which pay only qualified dividends and long-term gains taxed at 15% for both.) If they have the same risk tolerance, they should have the same asset location preferences.
5c) Both Reed and Grabiner agree (I believe) that the AL objective is to maximize wealth. Grabiner thinks the means to this end is " I am attempting to optimize returns." Or "Most models are designed to mathematically optimize returns". By which he means 'after-tax returns'. As a result the Wiki uses for a decision metric the difference between the nominal return and the after-tax return (or the $$ tax per $invested). (I am assuming that is what is meant by 'tax efficiency').
This is correct; the Wiki estimates the difference between pre-tax and after-tax returns for various asset classes.
Reed gave the math proof why this metric fails - but Grabiner ignored it. Reed math shows the way to maximize wealth is to maximize the accounts' benefits.
As I mentioned above, this is not quite correct even by Reed's metric; the way to maximize wealth is to maximize after-tax value, not tax savings.
And that a one-year view does not represent that benefit. That Time changes the benefit.
Image
And I disagree with the metric because it ignores risk. If it happens that bonds and stocks have only equal returns over your investing career, then you should prefer to hold the asset in a taxable account which loses the lowest percentage to taxes. And the difference in returns is likely to be more significant in this scenario than in a scenario with more normal returns.
Conclusion: Reichenstein ignores Reed's paper because it makes valid points with common sense judgements he cannot counter, and because it did not make the errors he chose to use as stray-man arguments against his detractors
This is an unfair criticism of Reichenstein. Given the realities of writing papers, Reichenstein may have submitted his paper before Reed's preprint was even released, and certainly did most of his work before the preprint. If Reed's paper is published, I would expect Reichenstein and others to discuss it in subsequent papers.
Wiki David Grabiner

Topic Author
less
Posts: 158
Joined: Mon Sep 16, 2013 12:38 pm

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by less » Thu Nov 14, 2013 11:07 am

grabiner wrote:I don't see the calculations from the spreadsheet you linked there.
I threw up my hands in disbelief when I read that. After all these posts in two threads he has still not read Reed's paper or opened the spreadsheets linked. Yet he has opinion on it. Unbelievable.
grabiner wrote: The math proof is a valid proof, but not of what you want to prove..... I disagree with the metric because it ignores risk ..
Etc, Etc, All through the he post above Grabiner treats AA and AL as one single process, with one single objective, much like Reichenstein, and completely at variance with the Wiki and Reed. If you don't agree with the objective then there is no point in further discussion.

* I support Reed's idea that AA is perfectly capable of managing risk all by itself. At best one's placement on the risk spectrum is a ball-park estimate. No fine-tuning is going to make that estimate 'more accurate'. Both Reed and the Wiki seem to agree. They both see the objective of AL as 'minimizing taxes', or 'maximizing wealth by minimizing taxes' (given the AA). Reed clarifies that 'maximizing wealth' is accomplished by 'maximizing the accounts' benefits'.

* Grabiner and Reichenstein see that both AA and AL processes are necessary to position one on the risk/return spectrum. They feel AA cannot do this by itself. AL must ignore 'maximizing the account's benefits' in order to join with AA in risk management.

User avatar
Doc
Posts: 9614
Joined: Sat Feb 24, 2007 1:10 pm
Location: Two left turns from Larry

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Doc » Thu Nov 14, 2013 12:51 pm

grabiner wrote:This doesn't appear to be the distinction he makes. All of the studies consider both risk and return; the first camp looks at after-tax value and pre-tax risk (or treats asset allocation as a given, which is essentially equivalent), and the second camp looks at after-tax value and after-tax risk.

So should you use pre-tax or after-tax risk? If you look only at the raw dollar numbers, then asset location does not affect your risk perception, and you should put your highest-risk assets in a Roth IRA because you get more benefit for no more risk. But if you look at after-tax value, then asset location does affect risk.

And I believe it makes sense to look at after-tax value when determining risk. Your mathematical risk tolerance (the effect of low returns on your standard of living) is determined by after-tax value. Ideally, you can set up your investment view so that even your emotional risk is driven by the after-tax value, either because you understand the principle or because you look at the numbers properly.
David hits on some points here upon which there is little agreement. When we talk about stock/bond allocations or TBM/Junk bond or small cap/large cap/domestic/foreign "risks" we are usually thinking in terms of standard definition and covarience. But when we get to asset allocation among different types of taxable and tax advantage accounts somehow "risk" becomes only how much money we have to use in retirement.

I think a lot of the difficulty we have had on this subject over the past several weeks is due to this lack of agreement on what "risk" means.

For myself I think one should adjust their various accounts on a consistent tax basis. ( I think an after tax basis based on accrued taxes is easiest to understand and use.) If after you have chosen your AA on this basis and then determine that you will not have enough money for retirement because of taxes you have to either take on more risk, save more or change your expectations for your future lifestyle. The idea of decreasing your risk of retirement deficiencies by somehow changing your effective tax rate plays only a small part in this overall picture.
A scientist looks for THE answer to a problem, an engineer looks for AN answer and lawyers ONLY have opinions. Investing is not a science.

Topic Author
less
Posts: 158
Joined: Mon Sep 16, 2013 12:38 pm

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by less » Thu Nov 14, 2013 2:15 pm

@ LadyGeek

The following seems to be about the only thing I agree with Grabiner about. The metric used for decisions on the Wiki does not accomplish what I see as the objective of AL, what the Wiki states as being the objective, or for whatever objective Grabiner wants. I don't understand Grabiner's argument, but we both think the Wiki metric wrong. Could I suggest the Wiki authors chime in with their justification?
less wrote: The Wiki uses for a decision metric the difference between the nominal return and the after-tax return (or the $$ tax per $invested). (I am assuming that is what is meant by 'tax efficiency'). A one-year view does not represent that benefit. That Time changes the benefit.
Image
grabiner wrote:This is correct; the Wiki estimates the difference between pre-tax and after-tax returns for various asset classes. I disagree with the metric because it ignores risk. If it happens that bonds and stocks have only equal returns over your investing career, then you should prefer to hold the asset in a taxable account which loses the lowest percentage to taxes. And the difference in returns is likely to be more significant in this scenario than in a scenario with more normal returns.

User avatar
grabiner
Advisory Board
Posts: 27162
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by grabiner » Thu Nov 14, 2013 9:22 pm

less wrote:
grabiner wrote:I don't see the calculations from the spreadsheet you linked there.
I threw up my hands in disbelief when I read that. After all these posts in two threads he has still not read Reed's paper or opened the spreadsheets linked. Yet he has opinion on it. Unbelievable.
I opened the spreadsheet which you had linked; it had one-year calculations, not the multi-year figures, and I didn't see an easy way to expand the calculations. Therefore, I did the calculations myself, and while my numbers don't match yours, they are close.

In any case, my disagreement is with the interpretation of the numbers. If you look at the raw dollar amounts, you will conclude that it is better to put stocks in a taxable account if your other account is tax-deferred, but to put stocks in a tax-deferred account if your other account is tax-free.
grabiner wrote: The math proof is a valid proof, but not of what you want to prove..... I disagree with the metric because it ignores risk ..
Etc, Etc, All through the he post above Grabiner treats AA and AL as one single process, with one single objective, much like Reichenstein, and completely at variance with the Wiki and Reed. If you don't agree with the objective then there is no point in further discussion.
I see the two as two parts of the process. Once you have decided on the proper tax-adjusted asset allocation for your risk level, you can then make an asset location decision.

The wiki is not a single article; it has articles on multiple issues, and it does discuss this issue.

Tax-adjusted asset allocation (section on Consequences: Asset Location).
Wiki David Grabiner

User avatar
LadyGeek
Site Admin
Posts: 63963
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by LadyGeek » Fri Nov 15, 2013 3:59 pm

Doc and less - I don't have the background to offer an experienced opinion.

I think we are discussing a contention with: Tax-adjusted asset allocation (Asset location) and Principles of tax-efficient fund placement

Grabiner has stated his position, which is reflected in the wiki articles. The wiki is intended to reflect a forum consensus. If you (or anyone else) disagrees with the content, what is the wording that you think should be revised?
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.

User avatar
abuss368
Posts: 20108
Joined: Mon Aug 03, 2009 2:33 pm
Location: Where the water is warm, the drinks are cold, and I don't know the names of the players!
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by abuss368 » Fri Nov 15, 2013 4:51 pm

We use the equal location appraoch as recommended by Rick Ferri rather than the asset location approach. Over a lifetime of investing, I am convinced it is the easiest to rebalance and possibly the most tax advantaged.

Perhaps Rick will add his valuable perspective to this discussion.
John C. Bogle: Two Fund Portfolio - Total Stock & Total Bond - “Simplicity is the master key to financial success."

steve_14
Posts: 1507
Joined: Wed Jun 20, 2012 12:05 am

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by steve_14 » Fri Nov 15, 2013 5:32 pm

Is this paper going to tell me to make my portfolio less tax efficient based on a historical efficient frontier I know isn't going to repeat? Mean/variance is fleeting; taxes are forever.

User avatar
Rick Ferri
Posts: 9112
Joined: Mon Feb 26, 2007 11:40 am
Location: Georgetown, TX. Twitter: @Rick_Ferri
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Rick Ferri » Fri Nov 15, 2013 6:04 pm

The problem with asset location is that nothing is static. Tax rates change, tax brackets change, tax preferences change, and on and on. What was a logical tax location one year may turn out to be a lousy one a few years later, but you're stuck in the one you have.

Rebalancing is also a problem. You could end up with all equity in your taxable account and need to rebalance out of equity. This could creates more capital gains tax than if you had a balanced account in both taxable and your non-taxable.

The long-term tax benefits are just not clear-cut, so do what is most convenient.

Rick Ferri
The Education of an Index Investor: born in darkness, finds indexing enlightenment, overcomplicates everything, embraces simplicity.

User avatar
Doc
Posts: 9614
Joined: Sat Feb 24, 2007 1:10 pm
Location: Two left turns from Larry

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Doc » Fri Nov 15, 2013 6:11 pm

LadyGeek wrote:Doc and less - I don't have the background to offer an experienced opinion.

I think we are discussing a contention with: [1] Tax-adjusted asset allocation (Asset location) and [2] Principles of tax-efficient fund placement

Grabiner has stated his position, which is reflected in the wiki articles. The wiki is intended to reflect a forum consensus. If you (or anyone else) disagrees with the content, what is the wording that you think should be revised?
Less and David are much more up to date on this subject then I am. However a very quick review of the two Wiki articles show glaring problems.
[1] wrote:You should also reduce the value of your taxable account by about 25% if you are early in your career, and 15% if you are near or in retirement, plus state taxes; estimated values are necessary because the tax cost depends on how much you will have in capital gains.
This statement makes the assumption that it is final value of the portfolio that is the measurement of risk. There is no qualification or even statement of the assumptions .
[1] wrote:Therefore, it makes sense to assume that the IRS owns 25% of the [taxable] account; whatever the gain or loss in nominal dollars, 75% will go to you and 25% to the IRS. In contrast, the IRS owns none of your Roth, which is tax-free
Flat wrong. The government owns 25% of the unrealized gains. What percentage that will be 30 years from now is going to depend on how the individual handles his investments. If he uses tax exempt bonds or reduces his unrealized gains through tax loss harvesting the unrealized gains and therefore the amount owed to the government could be as low as zero.
[2] wrote:Moderately inefficient
Balanced funds
Most bonds
Active stock funds
Most bonds are simply not tax inefficient. Tax efficiency is the return times the effective tax rate. If the return on a bond segment is very low than the tax is very low no matter what the assumptions are about the tax rates. This single factor is probably more important than all the other subtleties in the rest of the article. Yet it is almost universally ignored because people incorrectly equate "most bonds" with Total Bond Market.

It is a disservice to the general readership to make generalizations on tax efficiency as if they came from writings on some gold tablets. It is my opinion that this is the main take away from the Reed paper. I think many of the other aspect of the paper are controversial because of the varying definition of risk or are almost insignificant in importance compared to the other factors in the calculus.
A scientist looks for THE answer to a problem, an engineer looks for AN answer and lawyers ONLY have opinions. Investing is not a science.

User avatar
abuss368
Posts: 20108
Joined: Mon Aug 03, 2009 2:33 pm
Location: Where the water is warm, the drinks are cold, and I don't know the names of the players!
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by abuss368 » Fri Nov 15, 2013 6:29 pm

Rick Ferri wrote:The problem with asset location is that nothing is static. Tax rates change, tax brackets change, tax preferences change, and on and on. What was a logical tax location one year may turn out to be a lousy one a few years later, but you're stuck in the one you have.

Rebalancing is also a problem. You could end up with all equity in your taxable account and need to rebalance out of equity. This could creates more capital gains tax than if you had a balanced account in both taxable and your non-taxable.

The long-term tax benefits are just not clear-cut, so do what is most convenient.

Rick Ferri
Thank you Rick for proving that excellent perspective.

Best.
John C. Bogle: Two Fund Portfolio - Total Stock & Total Bond - “Simplicity is the master key to financial success."

User avatar
Ketawa
Posts: 2283
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Ketawa » Fri Nov 15, 2013 7:39 pm

Doc wrote:
[1] wrote:Therefore, it makes sense to assume that the IRS owns 25% of the [taxable] account; whatever the gain or loss in nominal dollars, 75% will go to you and 25% to the IRS. In contrast, the IRS owns none of your Roth, which is tax-free
Flat wrong. The government owns 25% of the unrealized gains. What percentage that will be 30 years from now is going to depend on how the individual handles his investments. If he uses tax exempt bonds or reduces his unrealized gains through tax loss harvesting the unrealized gains and therefore the amount owed to the government could be as low as zero.
This section is talking about a Traditional account, not a taxable account. It is true for a Traditional account. Full quote:
[1] wrote:If all of your IRA or 401(k) withdrawals will be in a 25% tax bracket when you retire, then 25% of your withdrawals from a traditional IRA or 401(k) will be lost to taxes. Therefore, it makes sense to assume that the IRS owns 25% of the account; whatever the gain or loss in nominal dollars, 75% will go to you and 25% to the IRS. In contrast, the IRS owns none of your Roth, which is tax-free.

User avatar
Doc
Posts: 9614
Joined: Sat Feb 24, 2007 1:10 pm
Location: Two left turns from Larry

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Doc » Fri Nov 15, 2013 8:02 pm

Ketawa wrote:
Doc wrote:
[1] wrote:Therefore, it makes sense to assume that the IRS owns 25% of the [taxable] account; whatever the gain or loss in nominal dollars, 75% will go to you and 25% to the IRS. In contrast, the IRS owns none of your Roth, which is tax-free
Flat wrong. The government owns 25% of the unrealized gains. What percentage that will be 30 years from now is going to depend on how the individual handles his investments. If he uses tax exempt bonds or reduces his unrealized gains through tax loss harvesting the unrealized gains and therefore the amount owed to the government could be as low as zero.
This section is talking about a Traditional account, not a taxable account. It is true for a Traditional account. Full quote:
[1] wrote:If all of your IRA or 401(k) withdrawals will be in a 25% tax bracket when you retire, then 25% of your withdrawals from a traditional IRA or 401(k) will be lost to taxes. Therefore, it makes sense to assume that the IRS owns 25% of the account; whatever the gain or loss in nominal dollars, 75% will go to you and 25% to the IRS. In contrast, the IRS owns none of your Roth, which is tax-free.
Maybe I got my paragraphs confused in my quotes. I just skimmed the article. Is this more appropriate?
[1] wrote:You should also reduce the value of your taxable account by about 25% if you are early in your career, and 15% if you are near or in retirement, plus state taxes; estimated values are necessary because the tax cost depends on how much you will have in capital gains.
A scientist looks for THE answer to a problem, an engineer looks for AN answer and lawyers ONLY have opinions. Investing is not a science.

User avatar
LadyGeek
Site Admin
Posts: 63963
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by LadyGeek » Fri Nov 15, 2013 8:23 pm

Here's what I get so far in Tax-adjusted asset allocation:
Wiki - Tax Adjust wrote:Adjusting by the marginal rate

You should value your Roth at its full value, and reduce the value of your traditional IRA or 401(k) by your expected tax bracket at retirement (including state taxes if your state taxes these withdrawals, which varies by state). You should also reduce the value of your taxable account by about 25% if you are early in your career, and 15% if you are near or in retirement, plus state taxes; estimated values are necessary because the tax cost depends on how much you will have in capital gains.

If all of your IRA or 401(k) withdrawals will be in a 25% tax bracket when you retire, then 25% of your withdrawals from a traditional IRA or 401(k) will be lost to taxes. Therefore, it makes sense to assume that the IRS owns 25% of the account; whatever the gain or loss in nominal dollars, 75% will go to you and 25% to the IRS. In contrast, the IRS owns none of your Roth, which is tax-free.
Underlined from Doc: This statement makes the assumption that it is final value of the portfolio that is the measurement of risk. There is no qualification or even statement of the assumptions.
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.

User avatar
LadyGeek
Site Admin
Posts: 63963
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by LadyGeek » Fri Nov 15, 2013 8:27 pm

Here's what I get so far in Principles of tax-efficient fund placement:
Wiki - Tax placement wrote:Approximate Tax Efficiency Ranking for Major Asset Classes
Moderately inefficient
  • Balanced funds
  • Most bonds
  • Active stock funds
From Doc:
Doc wrote:Most bonds are simply not tax inefficient. Tax efficiency is the return times the effective tax rate. If the return on a bond segment is very low than the tax is very low no matter what the assumptions are about the tax rates. This single factor is probably more important than all the other subtleties in the rest of the article. Yet it is almost universally ignored because people incorrectly equate "most bonds" with Total Bond Market.

It is a disservice to the general readership to make generalizations on tax efficiency as if they came from writings on some gold tablets. It is my opinion that this is the main take away from the Reed paper. I think many of the other aspect of the paper are controversial because of the varying definition of risk or are almost insignificant in importance compared to the other factors in the calculus.
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.

User avatar
grabiner
Advisory Board
Posts: 27162
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by grabiner » Fri Nov 15, 2013 9:20 pm

LadyGeek wrote:Here's what I get so far in Tax-adjusted asset allocation:
Wiki - Tax Adjust wrote:Adjusting by the marginal rate

You should value your Roth at its full value, and reduce the value of your traditional IRA or 401(k) by your expected tax bracket at retirement (including state taxes if your state taxes these withdrawals, which varies by state). You should also reduce the value of your taxable account by about 25% if you are early in your career, and 15% if you are near or in retirement, plus state taxes; estimated values are necessary because the tax cost depends on how much you will have in capital gains.

If all of your IRA or 401(k) withdrawals will be in a 25% tax bracket when you retire, then 25% of your withdrawals from a traditional IRA or 401(k) will be lost to taxes. Therefore, it makes sense to assume that the IRS owns 25% of the account; whatever the gain or loss in nominal dollars, 75% will go to you and 25% to the IRS. In contrast, the IRS owns none of your Roth, which is tax-free.
Underlined from Doc: This statement makes the assumption that it is final value of the portfolio that is the measurement of risk. There is no qualification or even statement of the assumptions.
The following paragraph explains the logic behind this adjustment:
Wiki article on tax-adjusted asset allocation wrote:If your IRA or 401(k) is large enough that the withdrawals will span multiple tax brackets each year, the marginal tax rate is the rate which affects the risk and return of any marginal decisions you might make, such as a change in your asset allocation. For example, if you expect that half your withdrawals will be taxed at 15% and half at 25%, you will lose only 20% of the total to taxes. However, if you increase the risk of your IRA investments and earn an extra $1,000, you will pay an extra $250 in taxes, not $200; if the risk works out badly and you lose $1,000, you will pay $250 less in taxes. Thus, in this situation, the IRS owns 25% of the part of the IRA in which your decision will make a difference.
This is implicitly referring to the effect on risk; if asset allocation is done to manage risk, then the marginal tax rate should be used for adjustment. I will clarify it and move it to the introduction to the section.
Wiki David Grabiner

User avatar
grabiner
Advisory Board
Posts: 27162
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by grabiner » Fri Nov 15, 2013 9:32 pm

LadyGeek wrote:Here's what I get so far in Principles of tax-efficient fund placement:
Wiki - Tax placement wrote:Approximate Tax Efficiency Ranking for Major Asset Classes
Moderately inefficient
  • Balanced funds
  • Most bonds
  • Active stock funds
From Doc:
Doc wrote:Most bonds are simply not tax inefficient. Tax efficiency is the return times the effective tax rate. If the return on a bond segment is very low than the tax is very low no matter what the assumptions are about the tax rates. This single factor is probably more important than all the other subtleties in the rest of the article. Yet it is almost universally ignored because people incorrectly equate "most bonds" with Total Bond Market.
The Wiki says that "most bonds" are tax-inefficient, but has an exception in the category above; "Low-yielding bonds or cash" are tax-efficient. It is rare that bond yields are so low that "most bonds" are less tax-efficient than stocks. Even now, the 2.16% yield on Total Bond Market Index Admiral gives a 0.54% tax cost in the 25% bracket, and the 2.42% yield on Intermediate-Term Tax-Exempt Admiral (the most likely holding if you choose to hold bonds in a taxable account) gives an implicit 0.81% tax cost if you assume that munis are priced to have the same after-tax yield as corporate bonds of comparable risk in a 25% bracket.

It would be reasonable to say something like, "Bonds or cash yielding less after-tax than stocks are efficient; bonds yielding more after-tax than stocks are inefficient." This is not the exact break-even (which depends on a more complex calculation, and on state tax rates), but it's a good estimate.
Wiki David Grabiner

User avatar
grabiner
Advisory Board
Posts: 27162
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by grabiner » Fri Nov 15, 2013 9:51 pm

Doc wrote:
Wiki: Tax-Adjusted Asset Allocation wrote:You should also reduce the value of your taxable account by about 25% if you are early in your career, and 15% if you are near or in retirement, plus state taxes; estimated values are necessary because the tax cost depends on how much you will have in capital gains.

(two paragraphs about tax-deferred accounts deleted)

The IRS also owns part of your taxable account, representing the difference between what you earn on it and what you could have earned in a tax-free account. If your taxable account is invested in stocks, you will lose some money every year you hold it to the tax on the dividends, and more to the capital gains tax when you sell. If it is invested in municipal bonds, you effectively lose the difference between yields on municipal bonds and corporate bonds of comparable risk every year. The calculation is more complicated (and depends on assumptions about future returns and tax rates), but the adjustment usually comes out about the same for a taxable account and an IRA; this spreadsheet can make an estimated calculation. The spreadsheet calculates the marginal tax rate, for the same reason as above.
Flat wrong. The government owns 25% of the unrealized gains. What percentage that will be 30 years from now is going to depend on how the individual handles his investments. If he uses tax exempt bonds or reduces his unrealized gains through tax loss harvesting the unrealized gains and therefore the amount owed to the government could be as low as zero.
The government does effectively own a portion of investments in municipal bonds, as noted above; the IRS doesn't actually take the money, but municipal bonds earn less than corporate bonds, so a municipal-bond investment in a taxable account will give you less money than a corporate-bond investment of the same dollar value in a tax-free account.

Tax loss harvesting cannot reduce the final tax cost on capital gains to zero if the shares are sold. Any harvested losses which are deducted against gains will increase future gains, and thus postpone the tax, but tax is due on any gain upon sale. Any harvested losses which are used to offset ordinary income produce a more valuable tax savings, but the offset amount will eventually be taxed.


Thus the 25% and 15% numbers are estimates for the marginal tax rate on gains in a taxable account; as noted, the cost depends on returns.

There are some situations in which the capital-gains cost can be reduced to zero, with only the dividends effectively taxed. If the stock is not needed for spending, you may donate it to charity or leave it to heirs, and pay no tax on the capital gain. If you can sell the stock when in a 15% tax bracket, and are not in the Social Security phase-in (which would give a 12.25% marginal tax rate), the tax on capital gains which keep you in the 15% bracket is zero, and you also pay no tax on qualified dividends; this is not common for investors who have significant taxable accounts, as their retirement savings is likely to all be in IRAs and 401(k)s.
Wiki David Grabiner

User avatar
LadyGeek
Site Admin
Posts: 63963
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by LadyGeek » Fri Nov 15, 2013 10:03 pm

Doc wrote:
[2] wrote:Moderately inefficient
Balanced funds
Most bonds
Active stock funds
Most bonds are simply not tax inefficient. Tax efficiency is the return times the effective tax rate. If the return on a bond segment is very low than the tax is very low no matter what the assumptions are about the tax rates. This single factor is probably more important than all the other subtleties in the rest of the article. Yet it is almost universally ignored because people incorrectly equate "most bonds" with Total Bond Market.
To Doc's point - I would also equate "most bonds" with the Total Bond Market. There are actually two exceptions: High-yield bonds and low-yielding bonds. How should "most bonds" be described? Does it apply to munis, Treasuries, or TIPS? I'm just looking at the wording, so perhaps it's unclear. I believe the descriptions are in the text, it's the table that needs some help.
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.

YDNAL
Posts: 13774
Joined: Tue Apr 10, 2007 4:04 pm
Location: Biscayne Bay

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by YDNAL » Sat Nov 16, 2013 8:32 am

less wrote:This post concerns the paper in the November Journal of Financial Planning http://www.fpanet.org/journal/CurrentIs ... fContents/ . The author claims the paper concerns Asset Location AL but I will argue that it concerns Asset Allocation AA much more, and ignores AL. http://www.fpanet.org/journal/TheAssetL ... Revisited/
This is what the authors write in the Executive Summary (2nd link).
The Asset Location Decision Revisited, by William Reichenstein, Ph.D., CFA; and William Meyer wrote:
This paper concludes by indicating that except in rare cases, investors should hold stocks in taxable accounts and bonds in retirement accounts.
Admittedly not reading this entire thread of veeeeery lengthy posts, my take on this is:
  • 1. Asset Allocation is to allocate your savings over multiple choices of investment -- primary asset classes being Equity and Fixed Income (stocks and bonds, per authors).
    2. Asset Location is to allocate each choice above MOST efficiently based on personal circumstances -- including all costs, which by default include *potential* taxation.
The future is unknown - and that includes the U.S. tax code and its effect on our personal circumstances. Any and all attempts to generalize are failures, IMO.

2¢ (short and sweet)
Landy | Be yourself, everyone else is already taken -- Oscar Wilde

Topic Author
less
Posts: 158
Joined: Mon Sep 16, 2013 12:38 pm

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by less » Sat Nov 16, 2013 9:37 am

None of the Wiki authors of Principles of tax-efficient fund placement has so far spoken up to justify the use of 'tax-efficiency" (= rate of return * tax rate = difference between nominal return and after tax return) for the decision metric for AsssetLocation.

* Reichenstein http://www.fpanet.org/journal/TheAssetL ... Revisited/ obviously does not agree because he uses AL to further the objectives most people feel satisfied by Asset Allocation. He uses no metric for AL because he ignores the tax benefits of the accounts. He uses MVO Mean Variance Optimization for AA, and the fall out from that process for AL.
* Grabiner (above) also disagrees with the metric (without stating his alternate) even though his objective, like Reichenstein, is not to maximize the benefits of the tax shelter accounts. So obviously he is not responsible for the Wiki's definition of the objective, or the metric to use.
* Reed http://papers.ssrn.com/sol3/papers.cfm? ... id=2317970 proves the 'tax-efficiency' metric is only valid for year 1 of an account, and then fails to maximize the accounts' benefits. Proved using both an assumption of never rebalancing, and an assumption of yearly rebalancing.
* Lady Geek (above) doesn't feel qualified to have an opinion.
* R Ferri (above) won't say what metric he uses for a 'logical' AL, and thinks the issue is a waste of time. I tend to agree, but if advice is being given it had better be correct. And (added later) The advice should not be used as a selling tool for advisors and their industry that pushes the MVO process.
* Doc (above) disputes the details of the application of the 'tax-efficiency' metric in the Wiki, thereby seeming to support its use. But does not argue why/how it is valid - and seems to dismiss Reed's proof it is wrong.
* My position http://www.bogleheads.org/forum/viewtop ... 6#p1810509 is from Reed's conclusion with added pictures. Even using extreme assumptions like "holding for 30 years without realizing capital gains' the advice to prioritize debt over equity in the tax-shelters is proven to be the exact opposite. Anyone can use the spreadsheets linked in Reed's paper using the step by step process shown in http://www.bogleheads.org/forum/viewtop ... 4#p1843884 .

----------------------------------------------------------------

Regarding Asset Allocation. Everyone loves to redirect the discussion to the issue ,,,"whether you should reduce either/both the TDA and Taxable account for the calculation". But the AA issue is only of relevance to the AL discussion as far as it impacts the stated objective of AL. If the objective of AL is to maximize the benefits of the tax-shelters (I do. The Wiki does), then AA is a 'given' and irrelevant - no matter how you chose to do its calculations.

Grabiner and Reichenstein do not agree with that objective because they feel AL is necessary to get an exact AA. The issue is "Do you think AA can ever be so exact that you give up the benefits of the tax-shelters for that exactitude?" In the default spreadsheet calculation below, exactly what benefit is gained by putting Asset B in the tax-shelter? what benefit that must offset the lost benefit from AL ($5,825 - $5,472) ?
Image

----------------------------------------------------------------------------
I think the Wiki page Principles of tax-efficient fund placement should be taken down in total.
a) The authors responsible for the decision-metric used refuse to justify it.
b) The metric used is proven wrong
c) Everyone who pins their colours to the mast disagrees with it.
Last edited by less on Sat Nov 16, 2013 10:11 am, edited 1 time in total.

User avatar
Ketawa
Posts: 2283
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Ketawa » Sat Nov 16, 2013 9:57 am

less wrote:If the objective of AL is to maximize the benefits of the tax-shelters (I do. The Wiki does), then AA is a 'given' and irrelevant - no matter how you chose to do its calculations.
This statement is obviously false. You have been deflecting any criticisms of the methods in the Reed paper by claiming that they're about AA, not AL. Now you're claiming that the process for AA doesn't matter? I already showed in this post that it absolutely does, and used Reed's own spreadsheet to show how.
Ketawa wrote:Now I will demonstrate that the Asset Location decision is not changed by a bonus/penalty from differences in contribution/withdrawal tax rate. The scenario is that an investor has $1000 in pre-tax income to put in either a Roth or Traditional account. The investor faces a 20% contribution tax rate and 40% withdrawal tax rate. The investor also has $700 to invest in a taxable account. We're still using the same Asset A and Asset B.

Go to [Reed's] spreadsheet in the YearByYear tab. Set the following variables to simulate an investor who uses a Roth account. This investor starts with $800 in the Roth account.

Tax Rate on Contributions: 0%
Tax Rate on Withdrawals: 0%
RRSP Balance @ Start: $800
Taxable Balance @ Start: $700

The final ending wealth numbers match the numbers from your image earlier in this thread. Clearly, the spreadsheet is calculating the final wealth correctly using these different variables.

Ending Wealth with Asset A Prioritized in RRSP (Roth Account): 6289
Ending Wealth with Asset B Prioritized in RRSP (Roth Account): 5915

Now, set the following variables to simulate an investor who decides to use a Traditional account. This investor can look at the Traditional account balance and see $1000 in the account, but knows that 40% of the balance and 40% of all future earnings belong to the government. This means that we can simplify the situation to one where tax rates are 0%, and the investor has $600 to invest in the Traditional account.

Tax Rate on Contributions: 0%
Tax Rate on Withdrawals: 0%
RRSP Balance @ Start: $600
Taxable Balance @ Start: $700

Ending Wealth with Asset A Prioritized in RRSP (Traditional Account): 5357
Ending Wealth with Asset B Prioritized in RRSP (Traditional Account): 5034

Conclusion: an investor who tax adjusts properly does not face a bonus/penalty that affects the Asset Location decision. In both simulations, the investor is better off with Asset A in tax-advantaged.

User avatar
Doc
Posts: 9614
Joined: Sat Feb 24, 2007 1:10 pm
Location: Two left turns from Larry

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Doc » Sat Nov 16, 2013 10:22 am

grabiner wrote:
Doc wrote:
Wiki: Tax-Adjusted Asset Allocation wrote:You should also reduce the value of your taxable account by about 25% if you are early in your career, and 15% if you are near or in retirement, plus state taxes; estimated values are necessary because the tax cost depends on how much you will have in capital gains.

(two paragraphs about tax-deferred accounts deleted)

The IRS also owns part of your taxable account, representing the difference between what you earn on it and what you could have earned in a tax-free account. If your taxable account is invested in stocks, you will lose some money every year you hold it to the tax on the dividends, and more to the capital gains tax when you sell. If it is invested in municipal bonds, you effectively lose the difference between yields on municipal bonds and corporate bonds of comparable risk every year. The calculation is more complicated (and depends on assumptions about future returns and tax rates), but the adjustment usually comes out about the same for a taxable account and an IRA; this spreadsheet can make an estimated calculation. The spreadsheet calculates the marginal tax rate, for the same reason as above.
Flat wrong. The government owns 25% of the unrealized gains. What percentage that will be 30 years from now is going to depend on how the individual handles his investments. If he uses tax exempt bonds or reduces his unrealized gains through tax loss harvesting the unrealized gains and therefore the amount owed to the government could be as low as zero.
The government does effectively own a portion of investments in municipal bonds, as noted above; the IRS doesn't actually take the money, but municipal bonds earn less than corporate bonds, so a municipal-bond investment in a taxable account will give you less money than a corporate-bond investment of the same dollar value in a tax-free account.

Tax loss harvesting cannot reduce the final tax cost on capital gains to zero if the shares are sold. Any harvested losses which are deducted against gains will increase future gains, and thus postpone the tax, but tax is due on any gain upon sale. Any harvested losses which are used to offset ordinary income produce a more valuable tax savings, but the offset amount will eventually be taxed.


Thus the 25% and 15% numbers are estimates for the marginal tax rate on gains in a taxable account; as noted, the cost depends on returns.

There are some situations in which the capital-gains cost can be reduced to zero, with only the dividends effectively taxed. If the stock is not needed for spending, you may donate it to charity or leave it to heirs, and pay no tax on the capital gain. If you can sell the stock when in a 15% tax bracket, and are not in the Social Security phase-in (which would give a 12.25% marginal tax rate), the tax on capital gains which keep you in the 15% bracket is zero, and you also pay no tax on qualified dividends; this is not common for investors who have significant taxable accounts, as their retirement savings is likely to all be in IRAs and 401(k)s.
Maybe I am addressing only a typo. The Wiki says that you should reduce the value of the taxable account. Here David is addressing the gains on the account.

I agree on the idea of marginal tax rates being used. Whether or not you use current rates or future rates depends on which definition of risk you are addressing. While David and I don't agree here I don't think he is wrong. At least for someone in or near retirement risk defined in the traditional MPT/CAPM/Efficincy Frontier approach makes more sense to me. but you get slightly different answers depending on the approach and that adds confusion to the basic idea that both tIRA's and taxable accounts should be adjusted for taxes.

The idea that the government "owns" part of one's muni's is one way of getting at the muni/taxable decision. I have enough trouble convincing people that the government "owns" part of your 401k/tIRA. applying the same thought process to muni's may be a bridge too far. :D

Reducing ones gains to zero is of course not desirable. I was just trying to emphasize the point.

I will address the "most bonds" though later.
A scientist looks for THE answer to a problem, an engineer looks for AN answer and lawyers ONLY have opinions. Investing is not a science.

User avatar
Ketawa
Posts: 2283
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Ketawa » Sat Nov 16, 2013 10:47 am

Doc wrote:
[2] wrote:Moderately inefficient
Balanced funds
Most bonds
Active stock funds
Most bonds are simply not tax inefficient. Tax efficiency is the return times the effective tax rate. If the return on a bond segment is very low than the tax is very low no matter what the assumptions are about the tax rates. This single factor is probably more important than all the other subtleties in the rest of the article. Yet it is almost universally ignored because people incorrectly equate "most bonds" with Total Bond Market.
One issue with this is that Bogleheads typically advocate simplifying and holding one core bond fund. When yields were lower, we did see some prominent posters advocate holding bonds in taxable, and they were typically looking at Total Bond Market. You can improve tax efficiency by separating Total Bond Market into separate Treasury and corporate funds, or by holding separate short/intermediate/long term funds. However, most people aren't doing this in favor of simplicity.

You could also improve tax efficiency by separating Total Stock Market into separate Growth (VUG), Value (VTV), Small Cap Growth (VBK), and Small Cap Value (VBR) funds. You could find ETFs to do the same thing for international funds. Sometimes, people split Total International into VEA, VWO, and VSS, then held VEA in taxable for tax efficiency. There is a balancing act between simplicity and holding more funds for tax efficiency.

As long as the main Bogleheads portfolio is the Three Fund Portfolio, I don't think there is anything wrong with the current Wiki page on tax efficient fund placement. Maybe there could be a separate section on advanced strategies to improve tax efficiency.

User avatar
Doc
Posts: 9614
Joined: Sat Feb 24, 2007 1:10 pm
Location: Two left turns from Larry

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by Doc » Sat Nov 16, 2013 11:08 am

LadyGeek wrote:To Doc's point - I would also equate "most bonds" with the Total Bond Market. There are actually two exceptions: High-yield bonds and low-yielding bonds. How should "most bonds" be described? Does it apply to munis, Treasuries, or TIPS? I'm just looking at the wording, so perhaps it's unclear. I believe the descriptions are in the text, it's the table that needs some help.
And that's what makes so much of the discussion on asset location almost meaningless. The tax efficiency difference between the S&P and small cap or the qualified dividend difference between developed and EM is small compared with the difference in the bond segments.

As an example:

The effective annualized tax rate for an 8% pretax annual return with LTCG at 15% paid all at the end, range from 6.5% at 30 years to 11.3% at 10 years. (This represent an idealized S&P 500 Index fund.) The average annual amount lost to taxes is therefore .065*8% -> .52% to .113*8% -> 0.9%. At a 25% tax rate on interest this means that any bond segment with yields less than .52/.25=2.08% to .9/.25=3.6% will be more tax efficient than the pure 8% all capital gains equity . According to Rick Ferri's Portfolio Solutions 30 year forecast almost all investment grade bond segments with less than ten years maturity would be more tax efficient than the equity investment for the ten year time frame. And even the thirty year Treasury is yielding less for the thirty year time frame.

http://www.portfoliosolutions.com/2013marketforecast/

http://www.treasury.gov/resource-center ... =realyield


Thirty years probably exceeds the average investment lifetime for most of us. If you start investing for retirement at age 35 and die at 95 the average holding period would be thirty years if you ignore rebalancing. A 5% FI investment would have a future value of 1.05^30=4.32 at 30 years. It would take an equity investment only 19 years to reach the same future value. I would suggest that 20 years is a more likely maximum lifetime investment horizon when rebalancing is considered.
Ketawa wrote:One issue with this is that Bogleheads typically advocate simplifying and holding one core bond fund
Well maybe they shouldn't.


(Aside: Was anyone aware the "text calculator" feature in Microsoft's One Note can even do logarithms?
Microsoft One Note wrote:Ln(4.32)/ln(1.08)=19.01293668400117
)
A scientist looks for THE answer to a problem, an engineer looks for AN answer and lawyers ONLY have opinions. Investing is not a science.

User avatar
LadyGeek
Site Admin
Posts: 63963
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by LadyGeek » Sat Nov 16, 2013 2:11 pm

Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.

User avatar
LadyGeek
Site Admin
Posts: 63963
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Problems with Reichenstein's "Asset Location Decis Revis

Post by LadyGeek » Sun Nov 17, 2013 9:51 pm

I added a caveat to the top of the wiki article, which was based on Rick Ferri's comment: Tax-adjusted asset allocation
Adjusting your asset allocation based on taxation may have significant unintended consequences. Tax rates change, tax brackets change, or your tax preferences may change. What was a logical tax location one year may turn out to be a poor choice a few years later. Consider the implications carefully.
Also, I added this thread to the wiki.
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.

User avatar
Taylor Larimore
Advisory Board
Posts: 29545
Joined: Tue Feb 27, 2007 8:09 pm
Location: Miami FL

Asset Location Study (2013)

Post by Taylor Larimore » Sat Jun 20, 2015 4:28 pm

Bogleheads:

As a former Internal Revenue Officer, I have long been interested in the taxation of mutual funds. One of the pioneers of fund location (taxable vs. tax-advantaged) is professor William Reichenstein, Ph.D., CFA. Another authority is William Meyer, a developer of software that calculates the most favorable asset location. In 2013 these two experts undertook a joint study of fund location which was published in the Journal of Financial Planning. This is the study's conclusion:
Results from this analysis indicate that the usual asset location advice is for an individual investor to locate stocks in taxable accounts and bonds in retirement accounts (TDAs and TEAs) to the extent possible, while attaining the optimal asset allocation. Even with today’s historically low interest rates, this usual asset location advice prevails for most investors. However, exceptionally low yields can reverse the usual asset location advice, especially for individuals with above-average risk aversion. Finally, the usual asset location advice prevails for investors who select a risk-free asset for the fixed-income portion of their portfolios.
https://www.onefpa.org/journal/Pages/No ... sited.aspx

Best wishes.
Taylor
"Simplicity is the master key to financial success." -- Jack Bogle

Post Reply