Warning: This is one of the posts that I occasionally write to convince myself that views I have developed without careful thought are not completely irrational. Because I have posted this, it passed that low bar. For posts like this one, my blog functions as an archive of my policy notes and thinking; they’re not written for outsiders. Translation: they’re typically plodding and boring. This one is also painfully long. Now that a handful of people inexplicably follow my blog, I feel compelled to advise you to hit delete.
Background note. Since they were adopted by Congress in 1990s, I have had an aversion to Roth IRAs. That’s as a matter of tax policy; as a personal investment vehicle, I like and use them. My instinct has always been that their structure contains inherent policy flaws, even though I have not seen my concerns explicitly discussed in the literature, other than the literature related to general consumption tax options (see the postscript). I will admit, however that I have not diligently and rigorously looked through the literature.
A brief note on why I had not worked through the issue: Because state income taxes cannot deviate much, if at all, from federal income tax retirement rules, I had no cause to engage with the issue while I was working. It is simply not practical for states to do so because (1) of the immense complexity of the federal retirement plan rules, (2) their inherent multi-year effects and participants’ interstate mobility, and (3) the administrators of IRAs and other plans are national and cannot really be expected to comply with state rules that deviate from them as a practical matter. I suppose on the last point, a handful of very big states (California, e.g.) could set their own rules but that’s about it. That sticks most states with the federal rules. When Congress adopted Roths in 1997, some in the Minnesota Senate (especially then Senator Steve Kelley) argued Minnesota should not conform. That might have been a plausible possibility – probably not, in my judgment. But neither the governor nor any involved House member thought it a reasonable option. So, I never worked through the issues or the problems with the Roth structure. Hence, my reason for doing so now.
What caught my attention
The ProPublica stories on Peter Thiel’s $5 billion Roth IRA have led to Congressional hearings and publicity about the problem of big IRAs (mainly Roths but traditional IRAs too). Numbers released by JCT staff in response to Senate requests are eye-popping. I expect Congress to address that problem in some way or another (i.e., if a second reconciliation packages passes with tax payfors). The testimony of Professor Daniel Hemel and TPC’s Steven Rosenthal provides a good primer on the topic. It is available on SSRN or a Cliff Notes version in a TPC blogpost by Rosenthal. Proposals are underway in Congress to address this; I expect some version to pass if Congress passes a second reconciliation bill.
High profile tax proposals in Congress propose expanded use of Roth IRAs. The Securing Strong Retirement Act of 2021 (H.R. 2954 passed by Ways & Means), for example, reduces revenue by delaying the required minimum distribution (RMD) age and by providing auto enrollment in some retirement plans. To offset the revenue loss, it requires catchup and other contributions now made to traditional plans to be made to Roths. See CBO summary. Similarly, the Ways & Means reconciliation tax proposal requires some employers to auto enroll employees in IRAs and makes Roth the default option. See Joint Committee on Taxation summary (p. 13).
I’m not writing about mega-Roths, but why I think the Roth structure generally is bad policy. As I discuss later, I know Congress is unlikely to restrict Roths as they apply generally (restrictions on mega-Roths are likely). They are uber popular both with the masses and on Wall Street. So, this is another one of my exercises of yelling for the wind and rain to stop. Nothing will happen, but I find the primal scream to be mildly therapeutic.
My priors
One’s priors are crucial in shaping policy views. So, I will briefly review my views on the US retirement system to be above board about where I am coming from and to try preventing them from preempting how I judge the Roth structure. I’m not one who thinks the retirement system is a total failure, as some do, but I do think it is poorly designed (it wasn’t obviously designed but has haphazardly evolved).
Since the 401(k) revolution began in the early 1980s, the US retirement system has been converted from a defined benefit (DB) system that provides a stream of (more or less) guaranteed retirement income to a defined contribution (DC) system that provides participants with a tax-advantaged savings accounts to fund their retirements beyond social security. The DC structure heavily relies on individuals to:
- Decide how much to save for retirement in tax-advantaged DC accounts
- Manage investment of the accounts
- Spend down the accounts during retirement
The government provides almost everybody with a starting base of social security benefits, essentially a guaranteed pension benefit. Many employees of governments and some old-line businesses, like utilities and insurance companies, still have DB pensions to supplement social security (or replace it for some government employees who are exempt from social security). But everyone else is expected to provide for and manage their plans with the help of tax incentives and (for some) choices their employers make (matching contributions, default participation, a menus of investment options etc.). So, it is a system that relies heavily on individual choice, management, and discipline. Most people believe social security alone is an insufficient source of retirement income. If so, most folks are subject to a largely, fend-for-yourself system to supplement it.
The following half dozen points are what I think is bad (4 points) and good (2 points) about the general contours of the DC structure:
First, it expects too much of human psychology (bad). Relying on individuals to save, invest, and spend-down is not a realistic expectation to address a core societal expectation. Saving (deferring consumption/gratification) goes against human nature. Most people have little or no interest in or aptitude for investing. That makes them easy prey for financial advisors, many of whom may be little more qualified than their clients and/or may be unscrupulous. Some of these problems have been modestly addressed by the movement to auto-enroll new employees in 401(k) plans, automatically bump up contribution rates, and default enrollees to better investment options like life-cycle index funds. But those are, at best, small improvements and only affect those who work for employers with plans using those options.
Second, it’s too complicated (bad). The tax law is littered with an alphabet soup of alternative retirement plans – traditional versus Roths, 401(k)s, 403(b), 457s, multiple flavors of IRAs (SEPs, SIMPLEs etc.), and so forth. Each of these plans is subject to its own rules, usually modest variations on a common theme. In a job-hopping world, many will taste multiples of these flavors, often accumulating small amounts in several. Yes, balances can be rolled over to an IRA, but then the individual is fully on his/her own without a preselected investment menu and other guidance employer plans provide. The complexity results from Congress enacting someone’s (typically investment houses’) latest idea to encourage more retirement saving – often as a supposed “solution” to data on the broad population’s failure to save adequately. Once enacted, Congress rarely will cut back on tax benefits to make matters clearer, simpler, and more consistent. Complexity is like arteriosclerosis where plaque keeps building up in the system’s arteries as Congress merrily eats more and more cheeseburgers and fries. The net result: a retirement system with angina and shortness of breath.
Third, it is tilted toward high-income folks (bad). Nearly all the supposed congressional solutions to inadequate retirement savings have a common theme of layering on more tax incentives for savings or eliminating restrictions on existing ones often in the name of simplicity. The net results are extraordinarily generous benefits for retirement savings that inevitably favor high-income cohorts and do little to nothing for the real problem, lower and middle-income folks. High-income wage earners or the self-employed can make very large tax-favored retirement contributions and with market returns amass very large accounts. The tables in Hemel’s and Rosenthal’s congressional testimony mentioned above (available here) illustrate that point and the JCT data show that many take advantage of the opportunity. The lost revenue could be better used to target the real problem in a variety of different ways.
Fourth, the system treats large swaths of the population as second-class citizens (bad). While high-income employees of larger employers can make overly big contributions to their work-based plans, many employees of small businesses have no access to plans at all. See this BLS publication for details. They are relegated to setting up their own IRAs, something that requires foresight and initiative. No behavioral nudges of auto-enrollment into good low-cost investments for them. These are typically lower-income or lower-skill workers who work for smaller employers and who have the biggest challenges with making savings and investment decisions and need the most help. That is contrasted with the high-income folks that Congress (both parties) and Wall Street fall all over themselves to help – because they have lots of money to invest (and pay Wall Street fees on) and they are articulate voices who reliably vote and let politicians know about their concerns.
Fifth, DC structures are more neutral across different classes and types of employees than a DB system (good). DB plans inevitably involve cross subsidization, creating “winners” and “losers” relative to their contributions. At one level, this is no different than life insurance (i.e., the luck of when one dies) and should be of little concern. Except that it is not, because longevity is strongly correlated with race and socio-economic status. White and upper-income folks live longer on average. Because actuarial rules must be neutral, DB benefit structures systematically redistribute money from the disadvantaged (poor and minorities) to the advantaged (white and upper income folks) with the same plan. That occurs because, on average, they live longer, contributions are typically a fixed percentage of wages, and the insurance elements of DB plans systematically make them winners. By any account, that is not a good result. By contrast, DC plans avoid that by putting each participant’s employer and employee contributions in their own accounts. There is little cross subsidization.
DB plans typically also include other forms of cross subsidization or redistribution – e.g., from employees with short work histories at an employer to those with longer ones. Some employer contributions under DC plans do not vest immediately and so also across subsidize longer term employees at the expense of short timers. More perniciously, it is also possible for DB sponsors to manipulate benefit structure rules in ways to benefit favored classes of employees. DC plans tend to avoid most of that.
Finally, DC benefits are portable (good). The biggest plus I see with the 401(k) revolution is that retirement contributions and earnings are easily portable as people change jobs. The traditional DB plan model did not work well as employees changed jobs. It often shortchanged employees with short tenures with multiple employers. I remember as a young government employee in my first two jobs being irritated because I felt the contributions that the employer made for me were going to subsidize “lifers,” since I knew I was on terminal appointments. I was very pleased when I took the job with the House of Representatives that I held for most of my career, since I discovered its plan, designed for political appointees, was a DC plan that was fully vested and portable. Of course, I became a lifer. As I approached retirement, I saw the advantages of the DB model and used my DC account to buy a fixed annuity. That option, of course, works generally, although few DC plan participants use it.
Basics: Traditional v. Roth
Traditional IRAs (or other “traditional” DC plans) provide deferred taxation of contributions and investment earnings until they are distributed, typically during retirement. Specifically, contributions are deductible in computing taxable income in the year made and investment earning are exempt, but distributions are includible in taxable income. (Some traditional plans provide for after-tax contributions, but the distributions of their investment income is still taxable unlike the Roth structure’s full exemption.) That model was universally followed until Congress enacted the Roth IRAs in 1997. Roth versions of the other basic DC plans came later (in 2006 under EGTRRA). For a primer on differences between the two plan types, see this TPC report, Leonard E. Burman, William G. Gale, and Aaron Krupkin, How Shifting From Traditional IRAs To Roth IRAs Affects Personal And Government Finances (Sept. 4, 2019).
The Roth model, by contrast, imposes income tax on contributions in the year made. and then, exempts “qualified distributions” from taxation (I won’t repeat the rules which are available in all their detail from the IRS website as well as many of them in the TPC report). Thus, in return for making after-tax contributions, Roth plans typically exempt the account’s investment earnings from taxation.
In a pay-me-now-or-pay-me-later dichotomy, Roths are the pay-me-now option and traditional plans are the pay-me-later option. All else equal (more on that below), both versions provide equivalent tax benefits or incentives; that is, if one holds tax rates across the contribution and distribution years and rates of investment return constant.
Since Roth accounts are an option to traditional plans, individuals will presumably choose one or the other based on what they think will provide the bigger tax benefit. Typically, this means:
- Individuals who expect their tax rates to drop in retirement, compared with the year in which they are contributing, should opt for traditional plans. By contrast, one who expects her tax rate to be higher should opt for a Roth.
- Individuals who want to make bigger contributions should opt for Roths. Both plans have the same nominal dollar limits on contributions, but because Roth contributions are made after-tax, they are implicitly larger (i.e., by 1/(1 – t) where t is the contributor’s tax rate). Another way to understand this is that the Roth contribution includes the tax obligation, while a traditional IRA must pay the ultimate tax (on distribution) out of account assets thereby reducing the amount available for retirement spending.
- Individuals funding an account to leave an inheritance (i.e., not to fund their own retirement) generally should opt for Roths since they are not subject to RMDs. (Note this is not true of Roth 401(k)s and similar but that can be avoided by rolling them over into Roth IRAs. An heir inheriting a Roth IRA is subject to RMDs.) Thus, they allow longer accumulations for those living beyond the age at which RMDs apply to traditional plans, enhancing the ability to pass tax benefits on to one’s heirs.
- Individuals who expect outsized investment returns on their accounts should opt for Roths for those investments. A common strategy would be to put low risk assets, such as fixed income securities, in traditional plans and higher-risk-higher-return assets, such as growth stocks or hedge funds, in Roth plans, if the investment horizon is long. If one’s assumptions about the likely returns are correct, that will maximize the tax benefits of the two structures.
The world is full of uncertainty. Estimating future tax rates requires political prescience (impossible) and predicting investment returns is a fraught exercise (other than at a very general level) without insider information or fudging the rules. So, at least on the margins, there is a bit of a fool’s errand involved in spending a lot of time and effort making the choice based on criteria #1 and #4. Of course, one may have inside information on investment returns or know with high certainty that their tax rates will increase or decrease. A medical resident knows, for example, her income and tax rate will increase when she goes into practice. So, in some cases there is a good basis for choosing between the structures. Even with the uncertainty, there is still good reason for savers and investors to put a lot of effort in selecting which type of account they contribute to, given the scale of the tax consequences (see below). Moreover, this must be an ongoing exercise as one’s circumstances change or new information becomes available, not something that can be done once for all time.
5 reasons I dislike the Roth structure
Reason 1: Upside down equity
The basic modus operandi of the US retirement system, aside from social security, is to encourage putting money aside for retirement with income tax incentives for doing so. As noted, traditional and Roth structures do so in different ways:
- Traditional IRAs allow deduction of saved income when it is earned, including its investment earnings, but taxes the money when it is withdrawn, presumably to be consumed in retirement. In other words, the tax benefits are front-loaded, and tax is deferred.
- Roths, by contrast, tax wages or self-employment income when earned but the investment income earned on the contributions is fully exempt. Withdrawals are exempt. The tax benefits are backloaded or, viewed the other way, the tax is prepaid.
These two mechanisms, on the surface, appear to provide quite different incentives or tax benefits. One is a complete exemption, while the other is a deferral. Note that they relate to two different amounts, though. And it has been widely recognized that under simplifying assumptions – as economists like to say all else equal – the two structures provide the same tax benefit. These simplifying assumptions are (1) tax rates remain constant (i.e., neither Congress nor the state legislature change them) and (2) investment return is constant. Because of that equivalence, most public finance economists and other tax policy experts do not express a preference for one or the other of the structures. The two structures’ effects vary depending upon individual circumstances but overall (on average across the whole population and over time), the effects are not much different. This can be shown or proven by a few simple algebraic manipulations (high school level). I won’t go through them; the TPC report (pp. 9 – 10) explains all of this, including showing its work on the algebra.
The problem with this, in my view, is that these simplifying assumptions do not reflect the real world. All else is not equal. Changes in tax rates are not an issue, since that it is a policy choice by Congress and legislatures and/or a life cycle issue for individual taxpayers (i.e., those who make Roth contributions when their incomes and tax rates are low and traditional contributions when their incomes and tax rates are higher). But with regard to investment returns, the Roth structure is pernicious. Everyone does not earn the same investment return. Returns are distributed from very low to very high, likely along a traditional bell curve. Most fall in the fat part of the distributional curve obviously, but substantial numbers of people are on the tails – some with low or very low returns and some with high or very high returns (think Peter Thiel on the latter for the ultimate outlier). Moreover, there is quite lot of variety even in the fat part of the distribution. The Roth structure rewards those on the right side (high returns) of the distribution and penalizes those on left (low returns). Individuals with above average returns are winners; they get bigger relative tax benefits. Individuals on the left side of the curve with below average returns get less; they pay more tax on their lifetime income or consumption.
I see this as a fundamental problem under the ability to pay principle – i.e., if you think taxes should be progressive or at least proportional as a matter of equity. The Roth structure rewards those with higher incomes or consumption (on a lifetime basis) with lower tax rates, while the unlucky investors with low returns pay higher tax rates. Putting it more starkly, individuals who are lucky enough to have above average return investments get a double benefit under the Roth structure – higher investment returns AND lower taxes. These differences in investment return, in my opinion, are almost always a matter of dumb luck (okay, Warren Buffet and for a very few others, skill). In some case, an artifact of when one was born and started investing (e.g., if the Great Depression occurred during a crucial part of your investing cycle) or whether someone advised or defaulted you into consistently investing in stocks. In any case, winners (those with higher lifetime incomes) get systematically rewarded with lower rates or bigger benefits and losers (those with lower lifetime incomes) with higher rates.
Moreover, even if the higher returns were the result of skill, more discipline, higher risk taking or other systematic differences there is no apparent equitable reason for rewarding that with lower tax rates. At least, none I can think of.
These differential effects can be very large, even for modest differences in investment returns. To illustrate, I constructed an example shown in the table below. My simplistic assumptions are:
- An initial salary of $50,000 that increases at 3%/year for 30 years (e.g., ages 35 to 65)
- Spenddown occurs over 20 years of retirement, so the account is presciently emptied at the end of 20th year of retirement (e.g., ages 66 to 85)
- 10% contribution rate applied to pretax wages for the traditional IRA and to after-tax wages for the Roth, so the contributions are equalized for the different tax treatment
- 20% constant tax rate to keep the math simple (it would not matter whether you use a progressive rate structure or added deductions so long as they did not change over the period) – a constant rate is necessary to show that the Roth and traditional plan provide equal tax benefits
- 6% investment return for the basic traditional and Roth example – again, necessary to show the basic equivalence of the tax benefits
- High investment return example assumes a 7% return one percentage point above the average and the low, 5% or one percentage point below the average.
After-tax retirement income varying rates of investment return
| IRA type | Assumed invest return | Amount |
| Traditional | 6% | $913,782 |
| Roth | 6% | $913,782 |
| Roth | 5% | $690.710 |
| Roth | 7% | $1,216,807 |
The table shows the equivalence between the Roth and traditional structures when you assume tax rates and investment returns are constant. That is the basic point economists and others have been making about the two structures for years. For the basic 6% return, both examples have the same after-tax retirement income over the 20-year period, just over $900,000. The Roth participant prepays tax, about $48,000, while the traditional participant pays a lot more in nominal terms (over $200,000) in tax during retirement, but their after-tax retirement income is the same. The tax difference reflects the time value of money between pre- and post-paid tax.
If you modestly vary the investment returns – by just one percentage point – the differences in income available to consume are large. The high-return account has $300,000 more to spend and the low-return account, $200,000 less. (To state the obvious: the after-tax income on a traditional account with those investment returns would be the same, just as it is for the 6% assumption.) If we convert those to rate differences, the basic account pays an effective rate of 5%, the low return account, 7%, and the high return account, 4%, using the pre-paid Roth tax as the numerator and the potential consumption as the denominator. From a life-time consumed income perspective allowing varying investment returns all else equal, the Roth structure is regressive. The account that provides more retirement income to consume (because of higher rates of investment return) pays a lower rate of tax. The Thiel mega-Roth underlines the point. He paid income tax on $2,000; that is the only tax he paid on billions in potential consumption! But the more systematic point is that it is important for differences in very normal variance in investment returns, those in the fat part of the distribution curve.
That is upside down and, in my mind, a strong argument against the Roth structure. If a choice is to be made between the two structures, it should not be the Roth mechanism for the basic reason that it violates ability-to-pay. That is a big deal. It has always perplexed me why the experts appear unconcerned about it. My assumption is that it is an artifact of their focus on the aggregate effects of the two structures and that people have randomly distributed rates of return. If so, to me that ignores the reality that averages mask different effects and that it rewards, all else equal, people with higher investment returns and income. How big an effect occurs will depend upon how much variance there is (e.g., the standard deviation) in rates of investment returns on DC retirement accounts across the population and how much that variance correlates with total income. I assume, without having data, a lot – especially the correlation of investment return with the amount of total income.
It is worth noting that the Roth structure – again on a lifetime basis – violates horizontal equity. Because capital income is taxed differentially (high-return assets paying lower rates than average or low-return assets), two individuals with the same lifetime incomes or consumption amounts can pay different taxes.
As aside, this issue arises in the two competing structures for progressive consumption taxes. In that context there has been more discussion of the issue by experts, but it is generally not considered a big deal. I note this briefly in a postscript.
Reason 2: Don’t reward the lucky or exempt rents
My second reason for disliking the Roth structure, to be fair, is a slightly different take on my first – probably because I think the first is so important. In principle, I would not object as much to the Roth structure if investment returns, which in reality reflect business and factor income, were all normal or standard returns. One way to express that would be that the returns were relatively tightly distributed around the mean or that the returns on the tails of the distribution (extremely high or low) reflected a true risk/reward payoff. I do not have hard evidence for this (translation: I’m operating on intuition), but I think that investors who realize outsized or excess returns (think Peter Thiel as the posterchild) largely do so not because of hard work, skill, risk-taking, or merit. Rather, it is typically a matter of luck and/or stumbling into economic rents. (Economic rents – essentially unearned returns that result for somehow being legally entitled to a market imperfection such as a monopoly – are an ideal tax base that can be taxed without distorting economic behavior.) It makes no sense to reward them with very low effective taxes, which is exactly what the Roth structure does when someone is lucky enough to invest a Roth account in a rent generating asset purchased at a discount (i.e., the contribution on which tax was paid did not reflect the capitalized value of its rents). It’s unnecessary to encourage risk taking and confers extra benefits on those lucky enough to win the lottery.
Reason 3: Discourage chicanery
An important element of tax design is to make it easy for taxpayers to comply and difficult to cheat. Traditional IRAs score better on the latter criterion (difficult to cheat) than Roths do. That is so because of the way in which the two deliver their tax benefits. The simplest way to cheat or stretch the rules for either type of IRA is to put undervalued investments (e.g., non-publicly traded stock) in the accounts. That puts more money into the accounts than the law allows, increasing the amount of tax benefits that the owner receives. Both accounts have equal dollar limits for contributing wages and salaries to the account, so on the surface they are on equal footing (beyond the Roth’s implicitly larger amount as noted above). But the consequences of the undervaluing of contributions differ. The table shows the relevant structure.
| IRA element | Traditional | Roth |
| Contribution | $ limited deduction | $ limit no deduction |
| Investment earnings | Tax deferred | Tax exempt |
| Distribution | Taxable | Tax exempt |
Putting an undervalued asset (e.g., founder stock) in a traditional IRA results only in tax deferral, while for the Roth it provides full exemption. The Roth structure’s theory is that tax is paid when the income (wages, salary, or self-employment income) is earned. If founder stock or private equity holdings are undervalued, that income is not reflected in the Roth account holder’s wages or salary and so the undervaluation amount avoids tax then and again when it generates investment income (capital gain or dividends), because Roth distributions are tax free. By contrast, for a traditional IRA the contribution limit (a fixed dollar amount) also limits much of the tax benefit (i.e., the upfront deduction). Yes, the investment income would benefit from deferral, but tax would be paid on distribution of the undervaluation. Effectively, the miscreant gets only the equivalent of making a after-tax contribution to a traditional IRA, still a benefit but less.
The Peter Thiel situation reported by ProPublica illustrates why this is important. I obviously do not know if the PayPal stock he invested his account was intentionally undervalued or not. But if it was, because he used a Roth, the investment income generated will never be taxed. By contrast, if he had used a traditional IRA the amounts in the account would eventually be subject to tax at ordinary income rates. This is an extreme example, but the tax involved for Thiel and his heirs (and the fisc) are huge. Upfront, he only paid tax on $2,000 of income to get the exemption.
Reason 4: Roths distort Congressional budgeting
Because they front-load government revenues, Roth’s distort congressional budgeting practices. The TPC Report (pp. 18ff) implies this is a problem if government budgeting is “myopic.” Of course, we all know that politicians live in the here and now (not the long-run or some future world), so I’m willing to stipulate that it is a problem. Political decision making and, hence, most government budgeting is almost always myopic. A little discussion shows why the Roth structure leads to this problem. By contrast, the traditional structure counters the natural political impulse to spend now and worry about paying later, because more of its costs are incurred up-front.
Budgeting is typical done for fixed periods of time using nominal dollars and government budgets follow this pattern. Minnesota state budgets are done on a biennial or 2-year basis (looser rules apply for a 4-year period), which is very short run. Congress, in a somewhat better model, budgets on a 10-year basis. Because Roth structures provide front-loaded revenues (i.e., the tax is “pre-paid”), their revenues appear immediately or within the budget period. Much of their reduced tax revenues or costs occur later when the tax exemption applies to distributions, i.e., outside the budget window (10 years for the feds). By contrast, traditional IRAs back-load their tax revenues, which are collected when the accounts make distributions and front-load more of their costs, the reduced revenues from the deduction and deferral. The “all-else-equal” rule, as discussed above, says that the long-term revenue effect of the two types of structures should be about the same – i.e., over the life of the average taxpayer. In a perfect world, the revenue effects of both types would be reduced to present values and budgeted on that basis (like budgeting for government pensions is supposed to work). But for government budgeting purposes they are very different. The immediate or front-loaded revenues from the Roth structure allow Congress to fund current spending or tax cuts, ignoring the reduced revenues outside the window (TPC’s myopia).
The TPC Report illustrates this effect using former Ways & Means chair Dave Camp’s tax reform proposal that used $200 billion in (illusory) revenues from converting traditional plans to Roths to cover other features that permanently reduced revenues. The proposals now making their way through Congress that I noted above in the reconciliation bill or the Secure Strong Retirement Act are additional examples of how Congress deviously uses revenues from taxing Roth contributions to fund current obligations. I would be willing to bet a small sum that the architects of those plans opting for the Roth structure had nothing to do with a policy preference for Roths but was all about hitting 10-year budget targets.
The net result of all of this is to bias congressional budgeting to long-term deficit financing and incurring more debt. Politicians are inherently myopic; they focus on the next election. The traditional structure puts a natural brake on that impulse and to my lights should be preferred on that basis. And it creates an inherent bias for choosing the Roth structure with all the resulting policy problems.
Reason 5: Too much choice = complexity = excess burden
My final major reason for disliking the Roth structure is not a criticism of the structure itself but rather of providing a choice between Roths and traditional IRAs. Moreover, it is a more general criticism (expressed in the section on my priors) of the retirement system rules.
It is well-known that the American retirement system rules are very complex. A myriad of different types of plans are available, each with separate rules. Many employers can choose among multiple options. It is also well-known reality of human behavior that too much choice can be mentally debilitating. A quote from the final report of the Bush Tax Advisory Panel on Tax Reform (p. 91) captured the problem well: “One professional financial advisor described to the Panel how investors become paralyzed by the range of tax-preferred savings choices.”
A major decisional complexity for a retirement saver is whether to use a traditional or Roth structure. In some cases, the best choice is easy, but in many it is not. Financial advisors and worker/savers spend untold hours deliberating over which to use (year after year). The number of analyses, columns, papers, and so forth written on the topic are legion. All you need to do is Google variants on the question “Should I use a Roth or traditional IRA” to see this. You will get millions of hits. Because the consequences can be large (in some circumstances), it is completely rational for people to spend time on this decision and for financial advisors to run scenarios, write more papers and columns, and so forth.
The net effect of this decisional complexity is that society wastes a huge amount of resources making this choice. Those resources are like a tax that yields no government revenues. In economic terms, it is all excess burden – pure loss. In short, dumb, the worst side effect of a tax. Of course, this loss could be avoided by opting for one or the other structure, just not giving taxpayers a choice. So, to circle back, it is not strictly a Roth criticism. Because the Roth option came much later as an option to the “traditional” approach (hence, the name), it seems like it to me.
Bottom line
The retirement and tax systems would be better off if Clinton and the 1997 Congress had not started us down the Roth path. The Roth structure has equity problems in my view, both vertical and horizontal. It is more susceptible to chiseling by those so inclined than traditional IRAs, while making an overly complex and confusing system even more so. But national experts who I respect do not appear to share my view, as far as I can tell, making me wary.
I see no possibility that Congress will makes matters better beyond a mild response to mega-Roths, likely imposing some sort of accelerated RMD rules, while leaving accounts with $10 million alone (accounts 95% of the population can only dream about). Were I czar, I would freeze all new contributions and impose regular RMDs on existing accounts until they completely go away. That will never happen.
This little exercise has slightly modified my priors, causing me to think more favorably of DB options or, even better, DC plans that pool investments over large groups of people, so that investment returns are averaged over many participants in either case. See this TPC publication (Collective Defined Contribution Plans) on an option of that nature.
Postscript – consumption tax
My equity issues with Roth IRAs versus traditional IRAs are paralleled in the two basic structures for a progressive consumption tax:
- The X Tax (developed by the late Princeton economist David Bradford), a close analogue of a Roth IRA. It is essentially a wage and self-employment tax coupled with an exemption for investment income that imposes a flat profits tax on business firms with expensing of capital investment.
- Personal consumed income tax (this got as far as an introduced bill by former Senators Nunn and Domenici in 1995) which looks more like a traditional income tax with unlimited deductible contributions to and withdrawals from an account like a traditional IRA (i.e., spending/consumption occurs out of the account).
Robert Carrol and Alan D. Viard, Progressive Consumption Tax The X Tax Revisited (AEI 2012) provides a detailed and accessible, book-length discussion of the issues involved with the two taxes. The authors favor the X Tax structure.
The X Tax is slightly different from a Roth under the current income tax in this more theoretical context because it includes an explicit tax on capital income (imposed on firms) to offset consumption financed with the return on capital. That is the way it taxes consumption from capital income. Because the tax is imposed at the firm level, it also is not adjusted for difference in rates of return realized by individual investors. As a result, the effects are somewhat like what I discuss. Carroll and Viard dismiss the concern somewhat summarily because they are attracted by other advantages of the X Tax, primarily its simplicity and resulting ease of administration and compliance (very important advantages in my view). See pages 33 to 39 for the discussion of the comparison of the progressivity of the two structures. This quote summarizes the authors’ thinking (they refer to the consumed income tax by the acronym PET):
The advantage of using the front-loaded approach at the household level is that it achieves a finer calibration of progressivity for households that have highly variable wages but relatively smooth consumption. This advantage does not seem to warrant the complexity of tracking financial transactions, particularly because, as we will discuss in chapter 3, the X tax can partially attain this advantage by allowing households with highly variable wages to average their wages across different years.
Although the household wage tax in an X tax system adopts a Roth back-loaded approach to saving, the overall X tax system, no less than the PET, adopts the conventional front-loaded approach, with a deduction for savings and a tax on outflows. Rather than implementing the deduction and the tax at the household level as the PET does, the X tax implements them at the firm level through the business cash-flow tax as firms deduct their business investments and are taxed on the proceeds of the investments. The business cash-flow tax is simple because it is imposed at a flat rate and does not track financial flows to individual recipients. The choice between the X tax and the PET is a close call. But for the reasons described above, we judge the X tax to offer the better combination of simplicity and progressivity.
Robert Carrol and Alan D. Viard, Progressive Consumption Tax The X Tax Revisited p. 37 (AEI 2012)
A contrary view is provided by Professor McCaffery and Hines who prefer the consumed income tax. Edward J. Mccaffery and James R. Hines Jr., The Last Best Hope For Progressivity In Tax, 83 S. Calif. L. Rev. 1031 (2010). This passage (p. 1039, footnotes omitted) summarizes their view on progressivity:
What about the comparison between a spending tax and a wage tax? While these tax systems are commonly lumped together in the economics and tax policy literatures as roughly equivalent variants of consumption taxes, and in fact have clear analytic similarities, there are nonetheless important practical differences. One especially important difference is that a wage tax does not attempt to tax unexpectedly high returns on capital investments, sometimes known as ― windfalls. It is not only the case that the rich save; it is also the case that the ― ex post rich — those who turn out to be rich, looking backward through time—tend to be those whose capital investments earned extremely high, indeed unexpectedly high, returns. Most of Warren Buffett‘s or Bill Gates‘s income has come from capital — specifically, the capital appreciation in the price of Berkshire Hathaway or Microsoft stock — and hence would be exempt from wage taxation. Yet the same income would be taxed under a spending tax when consumed. A spending tax alone among feasible alternatives can collect significant tax revenue from the propertied, capitalist classes in a way that does not necessarily undercut their incentives to become propertied in the first place.
Edward J. Mccaffery and James R. Hines Jr., The Last Best Hope For Progressivity In Tax, 83 S. Calif. L. Rev. 1031, 1037 (2011).
I think they have the better argument, since the tax legislative process is likely to blow away the simplicity advantages that Carroll and Viard see and the equity disadvantages are likely larger than they recognize.
- A firm level federal profits tax is sure to be defective in taxing capital income because of the problems with profit shifting to tax haven foreign countries. I simply do not see that problem being fixed anytime soon, if ever, despite current efforts appearing to make progress.
- The Carroll and Viard view overlooks the huge disconnect between capital gains income realized by holders of company stock and either the firm’s taxable income or accounting earnings (assuming the theoretical tax would work better than the actual corporate tax). That would not be the case if the stock market presciently capitalized future income and investors held the same stocks throughout their lives. Neither is the case. Investors make a lot of money in stocks that never have taxable income (think We Work, Uber, etc. for recent high profile examples). The stock market often misperceives companies’ prospects (capitalizing future earnings that never materialize) and investors make and lose a lot of money on those misperceptions. Probably even bigger problems are caused by very successful companies, like Apple and Google, that make above normal returns and pay abnormally low firm level taxes, while enjoying above average stock market valuations of their accounting earnings. The fact that there is almost no discussion of any of that in the literature (e.g., by Carroll and Viard in their book) is perplexing to me. Probably because economists tend to live in assumed worlds (the old joke about the economist stranded on a desert island with a case of unopen canned food who solves the problem by assuming he has a can opener).
Update 3/23/2022
When I wrote this post in December 2021, I assumed that Senator Roth, who was the Senate Finance Committee chair and for whom the accounts are named, had proposed Roth IRAs and compelled the Clinton administration to accept them. But Michael Doran, The Great American Retirement Fraud, was posted shortly after that. When I read his article, I discovered that the idea originated with the Clinton administration (see pp. 5, 53), Roth like the idea so much that he appended his name to the accounts in statute.
This update deletes the erroneous paragraph of the original post. I assume the Clinton administration proposed Roths as a way to partially pay for their tax proposal. It just shows, in my opinion, how very well-intended budget rules can have unintended and perverse effects. As Doran notes in his article, their effects under the budget scoring rules (discussed above in the original post and by TPC) drive their congressional popularity:
[B]oth Democrats and Republicans have long been entranced by the Roth IRA’s ability to turn revenue losses into revenue gains when scoring the federal budget. Since 1997, both parties have aggressively pushed the expansion of Roth IRAs and the extension of what is now generically called “Roth treatment” to retirement plans.
Doran, p. 5
SALT Speak - Joel Michael's Blog 