The Qualified Case for Quasi-Rothification
The Republican proposal to cap contributions to traditional tax-deferred 401(k) plans isn’t dead yet. Republicans will need to find revenue from somewhere if they want to enact broad tax cuts while still complying with the terms of the budget resolution that the Senate already passed and the House is set to approve today. The resolution allows for tax cuts that add $1.5 trillion to the national debt over the next 10 years, but the GOP tax plan in its current form loses somewhere in the range of $2.4 trillion, and Republicans need to fill that gap. One way they might do so is by capping contributions to traditional 401(k) plans at $2,400 per year and requiring that contributions above that amount go into Roth accounts.
This post argues that Republicans can include Rothification as part of their tax reform package while preserving one of the more attractive features of traditional 401(k) plans: taxation of withdrawals at the taxpayer’s marginal rate in the year of withdrawal. That might on first glance seem impossible: After all, a core feature of Roth plans is that taxpayers pay at their regular rates in the year of contribution and then can withdraw years later tax-free. By contrast, in a traditional plan, taxpayers pay no tax on contributions when made and then include withdrawals in their taxable income for the withdrawal year. But as I’ll argue, Republicans can achieve the basic benefits of Rothification while also maintaining the rate attribute of traditional plans — albeit with an element of complexity that might make the game not worth the candle. [Disclaimer: None of this is to suggest that 401(k) plans — Roth or traditional — are a first-best policy for encouraging retirement savings, as I think they’re almost certainly not.]
Most tax experts think that mandatory Rothification is a budget gimmick. In a post here last month, I argued that it might not be. The argument, in a nutshell, is that Rothification will reduce the sum of assets under management in 401(k) plans, and that if fund fees are based on assets under management, the result of Rothification might be that asset managers end up with less while the Treasury ends up with more. Since then, Mattia Landoni at Southern Methodist University and Stephen Zeldes at Columbia Business School have posted a really neat working paper on SSRN making this same argument. (They had presented the paper at various conferences months before my post, though I didn’t see it until it went up on SSRN. Suffice it to say that Landoni and Zeldes hit on this insight long before I did and their paper does a much better job of fleshing out the argument than my brief intervention. The whole paper is essential reading for folks thinking seriously about retirement savings policy.)
According to the authors’ back-of-the-envelope estimate:
[T]he government could achieve savings equivalent to $15 billion per year by forcing the conversion of all existing tax-deferred retirement accounts into Roth accounts. This $15 billion, a cost for the government, is an annual subsidy to the asset management industry.
The authors go on to develop a general equilibrium model in which switching from backloaded to frontloaded taxation of retirement savings (i.e., Rothification) reduces the size of the asset management industry and increases social welfare.
Why Not Rothify?
But while there’s a case for Rothification on these grounds, there’s a serious cost as well. As William Birdthistle and I wrote in The Atlantic earlier this week:
For workers whose marginal tax rate remains roughly the same throughout their lifetimes, a shift from traditional to Roth plans would have at most a modest effect. But such a shift in saving would harm people whose incomes decline significantly in their later years — precisely the people who need the most help saving for retirement. For example, a worker who is in the 25 percent tax bracket today but who will drop to the 10 percent bracket in retirement would much prefer to be taxed at a lower rate later rather than a higher rate now. In those circumstances, the worker will have materially more to spend in retirement if she can save through a traditional plan rather than a Roth.
To illustrate: Let’s say that A is in the 25% bracket at time 1, contributes $100 to a traditional 401(k) plan, and withdraws from her 401(k) at time 2 when she is in the 10% bracket. Let’s also say that the assets in the plan double between the time of contribution and withdrawal. Thus, A’s $100 becomes $200, she pays a tax at time 2 of 10% x $200 = $20, and she ends up with $180. If she saves through a Roth plan, then she pays a 25% tax at time 1 (leaving $75 that she can contribute and invest), her $75 becomes $150, and she can withdraw that amount. A is $30 (20%) better off in a traditional plan than a Roth plan.
B, by contrast, is in the 10% bracket at time 1, contributes $100 to a traditional 401(k), and withdraws at time 2 when he is in the 25% bracket. If B’s investment also doubles, then he pays a tax of 25% x $200 = $50 upon withdrawal and ends up with $150. If he contributes to a Roth instead, then he pays a 10% tax at time 1 (leaving $90 that he can invest); his $90 becomes $180; and he can withdraw that amount tax-free. B is $30 (20%) better off in a Roth plan than in a traditional plan.
More generally: People like A whose incomes and rates will fall over their lifetimes should prefer traditional plans to Roths. People like B whose incomes and rates will rise over their lifetimes should prefer Roths to traditional plans. There are a few exceptions to this rule — including that if tax rates change significantly, one’s income might move in a different direction than one’s marginal rate. But to a first approximation, it holds up.
And from a policy perspective, we probably want to do more to help A save for retirement than to help B save for retirement. Insofar as the goal is to help individuals smooth consumption over their lifetimes, then if anything we want to help B borrow from his future self in order to finance consumption in his younger years. It makes sense to save if you’ll earn less tomorrow than you do today. The rate feature of traditional 401(k) plans targets the biggest tax benefits at precisely those individuals (though even folks like B whose incomes will rise in their later years are usually better off saving inside traditional 401(k) plans than in taxable accounts because of the benefit of tax-free accumulation).
The Quasi-Roth Solution?
So how to achieve the benefit of Roth plans that Landoni and Zeldes highlight (i.e., the frontloading of taxation and the corresponding reduction in AUM and management fees) with the rate feature of traditional plans that William and I emphasize (i.e., the added subsidy to individuals whose incomes and rates fall over their lifetimes)? One way to do so would be to true up at the time of withdrawal. That is, we could require taxpayers to pay tax as they would under a Roth plan and then impose an additional tax or give them a refund upon withdrawal so that they are just as well off as they would be under a traditional plan. Let c be your marginal rate in the year of contribution; let w be your marginal rate in the year of withdrawal, and let X be the amount you withdraw. The taxpayer would pay an additional tax (or receive a refund) in the year of withdrawal equal to (X/(1 — c))*(w — c).
To illustrate: A pays a tax of 25% x $100 at time 1, leaving her $75 to invest; her $75 becomes $150, and she wants to withdraw that at time 2. Her c is 25% and her w is 10%. (X/(1 — c))*(w — c) = ($150/(1–0.25))*(0.1–0.25) = -$30. So A gets a refund of $30, leaving her in the same position as she would have been under a traditional plan ($180 after taxes).
Meanwhile, B pays a tax of 10% x $100 at time 1, leaving him $90 to invest; his $90 becomes $180, and he wants to withdraw that at time 2. His c is 10% and his w is 25%. (X/(1 — c))*(w — c) = ($180/(1–0.1)*(0.25–0.1) = $30. So B pays an additional tax of $30, also leaving him the same position as he would have been under a traditional plan ($150 after taxes).
The upside of this rule is clear enough: It preserves the rate feature of traditional plans and thus delivers the largest benefit to individuals with falling incomes — precisely the folks whose saving we want to encourage. The downside is additional complexity. First, there are a bunch of design details one would need to iron out. E.g.: What c to use when the taxpayer makes contributions in different years? (Something like first-in-first-out, or FIFO, would probably work.) And: What to do when a taxpayer is close to the cut point between two brackets in the year of contribution, such that part of the contribution amount was taxed at higher rate than the rest? (The answer might be: Break the contribution into two pieces and calculate separately.)
Beyond just thinking through these problems (which smart lawyers and accountants at the Joint Committee on Taxation and Treasury could probably solve conceptually in an afternoon), there is the additional challenge of recordkeeping. Taxpayers (or their 401(k) plan custodians) would have to record their rate in the year of contribution in order to calculate the tax or refund due at the time of withdrawal. That might not be too tough if your money is with, say, Vanguard throughout your life, but it might get messy with custodian changes and rollovers.
The key point is that the frontloading benefit of Rothification could indeed be paired with a system that yields the same after-tax results for taxpayers as a traditional plan — even if individuals’ rates change over the course of their lifetimes. Whether this is a good idea depends upon a weighing of the benefits of targeting under traditional plans against the costs of making the already-too-complicated section 401 that much more so.