On Tuesday, Tom Brennan took the train down for Columbia (where he is visiting for the semester, from Northwestern), to present Smooth Retirement Accounts.
The paper discusses traditional and Roth IRAs from two perspectives: utility to the saver, and federal budgetary optics. It then sketches out an alternative (the "smooth" retirement account) that would have Roth economics but different budgetary optics.
To explain: In a traditional IRA, your contribution is deductible, but your withdrawals are taxable (with tax-free inside build-up in between). A Roth IRA contribution is not deductible, but the withdrawal is tax-free.
These two alternatives are economically equivalent if you simply earn the "normal" rate of return and if the tax rate is fixed. E.g., say the tax rate is 50%, 1 year investment with a 10% return. Under traditional, you might spend $200 out-of-pocket, which costs you $100 after-tax, and the account grows to $220, of which you retain $110 after-tax. In the Roth, you simply deposit $100 of after-tax income and get $110, all of which you keep.
One key assumption that's needed for the two methods to be equivalent is that the investment be "arm's length" rather than, say, in your own business with labor income built in, so that you can only earn the normal rate of return. E.g., if you can turn the first $100 you have into $100M because you're Zuckerberg and it's Facebook (but you can't keep going at that rate), then obviously you do better under Roth (at least, if you couldn't have parlayed $200 into $200M), reflecting that in effect labor income or rents or whatever you want to call it has been layered on top of the normal rate of return.
All this is familiar stuff in the biz. Returning to the more particular analysis in the paper, a big difference in practice is that in Roth, the only tax rate that matters is the current one, so you're locked into it (assuming Congress doesn't renege down the line). But in a traditional IRA, the future tax rate may differ from the current one, causing the investment to be tax-favored vs. Roth (or immediate consumption) if the rate declines, and taxed unfavorably if the rate goes up. This could happen due either to changes in statutory tax rates or because of the progressive rate structure (e.g., you are in a lower rate bracket during your retirement years when you withdraw).
Brennan doesn't like this feature of traditional IRAs for two reasons. It subjects the taxpayer to tax rate risk, and progressive rates can distort portfolio choices by reducing the relative payoff for investment with variability but high upsides relative to those that are relatively fixed. So he prefers the Roth approach of avoiding downstream tax rate variability, though he wouldn't mind having the tax paid at the end (a la traditional IRAs) if it was locked in and equal to the deduction rate up front.
OK, the issue of progressive rates and portfolio choice is a broader one in the income tax (or in a progressive-rate consumption tax), though clearly he has a point. Also a legitimate broader concern is the issue of tax rate variability distorting consumption choice between periods. (A standard model consumption tax such as a VAT or retail sales tax creates this problem, of course, if there may be tax rate changes between years. And suppose you are faced, upon making your tax-free Roth withdrawals, with a newly enacted VAT. Then you have indeed been "taxed twice," although it might not be viewed the same way as explicitly reneging on tax-free Roth withdrawal under the income tax.)
On the subject of tax rate risk, clearly it would increase people's utility if, in response to future tax rate risk, they could purchase, for a suitable price, private insurance locking in the current rate for them. Under such a hypothetical insurance system, I'd pay, say, 35% no matter what, and the government would collect based on the actual future rate no matter what, but the insurance company would pay the extra on my behalf if the rate went up, and take the difference out of my hide if the rate went down. This of course is just a thought experiment, not real life or a practical proposal. Presumably one reason such insurance doesn't exist is that the insurer would have difficulty reinsuring or spreading the risk. And there would also be moral hazard issues if the payments depended in part on how much income I actually had in the later period. But I mention it just to make the point that, while such insurance presumably would be a good thing for consumers if feasible, that doesn't mean that it's optimal for the government to allow people to lock in their future tax rate. That leaves other taxpayers as the "counterparty" without any insurance premium.
Example 1, you lock in the current rate in June 1941, then on December 7 it turns out that taxes will have to go way up to pay for a million-soldier, two-front world war. Taxpayers who locked in the earlier rate have concentrated the revenue risk on everyone else. Case 2, President Romney imposes low rates, then the following year President de Blasio enacts high rates. So it's a change in political preferences, not "objective" needs, but here it's at a minimum unclear whether or not we should view allowing an earlier-year opt-out as socially beneficial.
Anyway. As Brennan agrees, it's a complicated question to what extent we should want tax rates to be locked in, despite the clear benefits at one particular margin to enabling this. The point, of course, is that these are complex and interesting issues, not that the paper is "wrong" to use correct analysis of one of the relevant margins to show an advantage to locking in the rate.
Issue 2 is budgetary accounting. With a short-term budget window, traditional IRAs look costlier than they actually are, since the deductions are counted but the later year inclusions are outside the budget window. Roths look cheaper than they actually are, since Congress is only giving away the out-year revenue (plus the tax on inside build-up within the budget window, but that may be just a small part of the whole). Worse still, as Congress showed through budgetary shenanigans in 2010, you can lose long-term revenue by "bribing" people to switch from traditional to Roth IRAs, yet score it as a revenue gain within the budgetary window that you use to pay for other tax cuts. In that scenario (which, again, actually happened), tax cuts (from the bribe that is offered to prompt Roth conversions) are used to "pay" for other tax cuts.
Brennan has an intricate plan in mind that could work as follows. You use a traditional IRA, in the sense that there is an up-front deduction. But as in a Roth IRA, there is no tax rate risk, because you are guaranteed that the withdrawal will be taxed at the same rate that applied to the deposit. But to improve the budgetary accounting, each year a suitable fraction of the accruing future tax liability would be counted as current revenue gain for budgetary accounting purposes. The deemed tax revenues for budgetary accounting purposes are the source of the adjective "smooth," but, as this is already a long blog entry, I will refer you to the paper itself (see the link above) for the precise mechanics.
The budgetary optics problem has a straightforward solution, which is simply to do infinite horizon budgetary accounting rather than artificially truncating the out years that are deemed to be within the budgetary window. Obviously, this raises issues of its own that a large literature has examined. Next best might be coming up with ad hoc budgetary rules that replace pure cash accounting, for both Roth and traditional IRAs, with something that is the same for both of them and also closer to economic accrual than the cash flow treatment of either way. The paper offers one way of moving in that direction. But once you are not going all the way to accrual, by eliminating the budget window, it comes down to a choice between imperfect alternatives. Brennan, who at all times is extremely and indeed completely fair-minded, agrees with this analysis.