One adjustment which this paper makes and which I have long advocated is that tax-deferred (traditional 401(k) or IRA) and tax-free (Roth) accounts are equivalent after adjustment; the tax-deferred account must be adjusted based on the percentage that the government owns. If you will pay 25% taxes on your IRA withdrawals, then an IRA with a nominal balance of $4000 represents $3000 owned by you tax-free, and $1000 owned by the government, and should be invested almost the same (see below) as if it were a Roth IRA with a balance of $3000. This is an important point which is often missed; some investors prefer to put stocks in Roth IRAs because the growth is tax-free and thus the expected final value is higher, not recognizing that this higher expected value comes with a greater risk and a larger loss if the stock market loses value.
I disagree with the paper's decision not to tax-adjust taxable accounts as well; while the nature of the tax is different (a tax on gains, rather than on the full value), it is still a concern which affects the money you will have to spend.
Key points of the paper:
This paper argues against AL practices that a) phrase general rules in terms of asset-types, instead of in terms of the metrics that decide the issue, b) produce rules for AL that are shown to fail to maximize wealth and c) ignore the impact of a change in tax rates between contribution and withdrawal, and d) make the objective of AL to maximize utility, when investors assume AL decisions maximize wealth.
Point (a) is correct by definition but hard to work with in practice. The key metrics for tax efficiency are return and tax rate. For example, while fixed income is normally tax-inefficient, low-yielding short-term bonds do not lose much to taxes and therefore belong in taxable accounts. But it's best for many investors to start with general rules, particularly when the actual answers are dependent on the exact estimates made about future tax rates and returns. And the main rules of thumb are fairly clear: stocks are tax-efficient, bonds become less tax-efficient as the interest rate rises, and it usually works out that stocks are better in taxable accounts. You do need to work out the numbers yourself; for example, Treasury bonds have a low yield and are exempt from state tax, so if you pay high state tax, you may want to hold Treasury bonds in taxable even if yu hold your other bonds in tax-deferred. (At current yields, I recommend I-Bonds, and even TIPS, in taxable accounts for investors who pay high state tax.)
I disagree with point (d) and therefore with point (b). An investor's goal should be to maximize expected utility, not expected wealth; this is why most investors do not hold 100% stock. Holding stocks in a taxable account, rather than an equal dollar amount in a tax-free account, reduces the risk. If the tax-free account loses $10,000 the year before withdrawal, that is $10,000 less to spend; if the taxable account loses $10,000 the year before withdrawal, that is $8500 less to spend (assuming 15% tax on a long-term capital gain). And if the tax-free account loses $10,000 long before withdrawal, that is a loss of retirement spending equal to the future tax-free returns on $10,000; if the taxable account loses $10,000 long before withdrawal, that is a loss of retirement spending equal to the future returns on a $10,000 taxable investment, reduced by a further $1500 of capital-gains tax that will no longer be due (and an even lesser loss if part of the loss is on recently-bought shares and can be recovered immediately from the IRS by tax loss harvesting).
My disagreement is based on the assumption that asset allocation is used to set mathematical risk levels. The author writes, "It is the investors’ emotional reaction that determines their tolerance for portfolios’ greater or lesser risk." Risk should be determined by three criteria: need, ability, and willingness to take risk. Only the willingness criterion is emotional; if you are likely to panic or lose sleep when your portfolio loses 30% of its value, then you should choose an allocation which is unlikely to lose 30%, however you measure the losses. Your need to take risk is based on the after-tax value of your portfolio, since that is what you will have available for spending. And your ability to take risk is based on the after-tax cost of any potential losses, since these are the losses you will either suffer or make up. I will update the wiki article Tax-adjusted asset allocation based on the three views of risk.
Theoretically, the proper way to deal with emotional risk is to view your accounts as having already lost any expected future tax loss, but that is difficult for most investors. A decrease of $10,000 in an account balance looks the same to most investors whether it is $10,000 in a Roth, $10,000 in a taxable account which can be immediately harvested as a loss to offset capital gains for a $1500 savings or to save $2500 by offsetting ordinary income for four years, $10,000 in a taxable account which still has gains and thus will reduce taxes by $1500 upon withdrawal (plus any taxes due on future appreciation), or $10,000 in a 401(k) which will be taxed at 25% and is thus a cost of $7500. (My own asset allocation spreadsheet is designed for this adjustment. It never adds the dollar balance of all the accounts, only the adjusted value, so a $10,000 decrease on an account taxed at 25% shows up as only a $7500 loss on the bottom line.)
Viewing risk mathematically also justifies making asset location decisions based primarily on percentage tax costs. If a particular stock and bond investment lose the same percentage of their value to taxes, putting the stock in a tax-free account will give higher after-tax wealth if the stock has higher returns. But those higher returns are not certain, which is why you held the bond in the first place rather than 100% stock. Putting the stock in a tax-free account will give the same after-tax wealth if stock and bond returns are equal (and the tax costs are still the same), and a lower after-tax wealth if the stock market loses money.
And I agree with point (c); the tax rates on contributions are only relevant in your decision whether to invest in a tax-deferred or tax-free account. Once you have made the decision, tax-deferred and tax-free accounts are equivalent.
The article suggests putting lower-returning assets into a tax-deferred rather than tax-free account, to reduce the probability that higher-returning assets will be taxed in a higher bracket than expected. The article only considered risk in one direction, but the risk-return trade-off actually works both ways to make this more desirable. If you expect to retire in a 25% tax bracket and have stocks in your tax-deferred account (with its asset allocation value reduce by 25%), and the stock market does well, you might pay 28% tax on some of your stock withdrawals, so you get less than the full value. Conversely, if the stock market does badly, you might retire in the 15% bracket and get only 15% tax savings on some of your stock losses. (However, given current tax laws, the phase-in of Social Security taxation comes at about the top of the 15% tax bracket, so if the stock market does badly, you might actually save tax at 27.75% or 46.25% rather than 15%.)