IRAs – SALT Speak – Joel Michael's Blog

Caveat: This post is about public policy not investment or personal finance tactics or strategies. As a personal finance matter, Roth accounts are a great way to diversify the tax risk of a retirement portfolio – especially for someone who is sophisticated enough (or has a sophisticated advisor) to use them intelligently. I have a significant share of our retirement portfolios in Roths for precisely that reason.

One of my pet peeves is the misperception¹ by sophisticated people of a what I see as a crucial flaw in the Roth structure in how we incentivize retirement savings: the Roth structure piles tax benefits on those lucky enough to be winners in the investment race, while punishing losers. The net result likely makes the structure regressive on a permanent or lifetime income basis. I have written about this twice before in an overly long and obtuse ways here and here (see “The other Roth problems” section of the latter).

A recent X post on an unrelated topic (sequencing risk in retirement savings) and a striking graph in it (see below) provided what I think is strong, indirect evidence for this effect – investment returns can systematically vary a lot by the randomness of when you start and stop retirement investing (largely when you’re born and start working). So, it caused me to once again mount one of my old hobby horses and write this post. Apologies. You can hit delete now, if you aren’t interested in hearing me drone on about this relatively esoteric topic.

Roth versus traditional structure

The incentive or subsidy provided by Roth versus traditional IRAs varies based on whether the subsidy is provided at the point of (1) the contribution or savings (traditional or frontloaded tax benefits) or (2) the distribution or consumption (Roth or backloaded tax benefits). It is well known (pp. 9-10) – based on simple algebra – that the amount of the subsidy is mathematically identical if/when two elements are the same for the entire period of saving and retirement:

The tax rate on contributions and distributions; and
The rate of return on the investment

Of course, in the real world we know that those two things are never constant. Legislation changes effective tax rates, as well as individuals’ incomes varying over their lifetimes causes their tax rates to rise and fall. Similarly, we know that investment returns vary widely. This variance results mostly from random stuff: investment skill, luck, or pure timing.²

This post relates to the last of these, timing: even if everyone invested the same amount or percentage of their retirement savings in index funds (eliminating skill and luck as a factor), investment returns will systematically vary based on when they started saving and when they started drawing down their accounts. That variance occurs because of the business cycle, animal spirits, changing expectations, yada, yada.

Why it matters

In my view, this matters crucially on a policy basis because providing the subsidy at the point of contribution is fairer or more equitable.³ That is so because the Roth structure, all else equal, provides larger incentives to individuals who realize above average investment returns. (This gets back to the assumption of a constant rate of return to make the two mechanisms equivalent.) Thus, these lucky ducks (winners in the investment lottery, so to speak) get bigger subsidies, magnifying their higher investments returns. Put another way, the portion of their return that exceeds the average return in the assumption goes untaxed. They get bigger incentives; those with lower returns get smaller incentives on average. Second, one has to assume that folks who realize systematically better investment returns typically have higher incomes, because their accounts end up being larger on average, all else equal. So, the Roth structure makes the tax and retirement system more regressive, compared to the traditional structure.

Birth lottery

A recent interesting X post on a tangentially related subject – sequencing risk in retirement savings – provides an illustration of why this assumption is likely correct and that it is not trivial. (Note: the post is about the variance in investment returns and the effects on retirement savings, NOT taxation.) The post is by Jesús Fernández-Villaverde, a U Penn economics professor, who I had never heard of but assume is a reliable source.

Investment returns vary because of the cyclical ups and downs of the stock and bond markets. There can be long periods of good markets (e.g., the 1990s) and bad markets (e.g., the 1970s). When you start investing/saving for retirement and when you stop and begin drawing down your account matters – potentially by a lot. Fernández-Villaverde did some extensive calculations of those effects and the results are quite striking. The entire X post is well worth reading for those interested in the general topic. It’s short. Here’s his graph that summarizes his results, which show a lot of variation:

Methods and results. His calculations are for a pure stock portfolio (S&P 500 index) for the period 1945-2024. He calculated real returns (i.e., he took inflation out of them). Here’s his description of the results (my emphasis added):

Over this 80-year period, the S&P 500 delivered a geometric mean real total return of about 7.5% per year. That is an impressive number. But this average return masks a lot.

Imagine a worker who starts investing at age 22 and retires at age 68. That gives them 46 years of contributions. In their first year, they contribute $1. Each subsequent year, they increase their contribution by 1% (roughly keeping pace with real wage growth). Every dollar is invested in the S&P 500. They never touch the money until retirement. No panic selling, no market timing, no strategy switching (and no management fees!). Textbook investing and waiting.

I ran this exercise for every possible cohort for which the data allow. The first cohort starts investing in 1945 and retires in 1991. The second starts in 1946 and retires in 1992. And so on, all the way to the last cohort, which starts in 1978 and retires in 2024. This yields 34 cohorts, each investing for 46 years, making the same contributions and investing in the same index. The only difference among them is which 46-year slice of historical returns they happen to live through.

The most fortunate cohort, the one that started investing in 1954 and retired in 2000, had $607 on the day of retirement (remember, all in real terms), with a real annual return of 8.82%. The unluckiest cohort, the one that started in 1963 and retired in 2009, accumulated $210, with a real annual return of 4.83%. Same contributions. Same index. Same strategy. Same investment horizon. Yet the luckiest retiree ended up with 2.9 times more wealth than the unluckiest. ⁴

The typical retirement account is not invested 100% in stocks, which provide a higher rate of return than bonds or cash. Conventional wisdom warns against an all-stock portfolio, especially as retirement approaches. So, his assumption exaggerates the difference between the highest and lowest returns. He redid it using a target date fund. This reduced the 2.9X difference to 1.6X.⁵ That is still a large difference.

He also points out that in the real world, the ones with unlucky timing relative to the business cycle will suffer more than just subpar investment returns:

In fact, sequencing risk is even worse because poor returns in the stock market are correlated with weak labor markets: you have a much higher probability of losing your job (or seeing your wage income fall) precisely when the market is doing poorly, preventing you from saving when prices are low and equities are most attractive. However, let me set that point aside today to simplify the exposition.

Policy implications. To me, this illustrates one policy reason why the Roth structure is flawed as a retirement incentive: it piles larger tax benefits on those lucky enough to realize above-average investment returns (investment skill, pure luck, or as here when they started and ended their retirement savings) who need it least. Higher returns are almost certainly correlated with higher retirement incomes, probably strongly correlated.⁶ This undoubtedly makes the tax and retirement systems less progressive. It persists across age cohorts, if you look at the pattern of the graph.

I can think of no countervailing policy reason for giving more to investment winners and less to investment losers. It does not incent desirable behavior (like savings or careful investing), rather random luck. Moreover, Fernández-Villaverde’s calculations make it clear what we’re often rewarding is simply timing, mainly the timing of one’s birth which no one controls. If anything, a better policy is to offset the effects of bad luck. The traditional IRA structure does that by the government assuming some of the risk of random variance in investment returns, while the Roth structure works in the opposite direction.

We’d have been much better off if Senator Roth and President Clinton had never started us down path of using the Roth structure. Unfortunately, it’s now too late. Because the Roth structure allows Congress to game the measurement of fiscal cost, it makes further movement to Roths fiscally irresistible as Congress almost surely enacts future editions of the Great American Retirement Fraud. (See my summary here.)

Of course, there are a host of other reasons for criticizing the retirement incentive system. It’s too complex, too expensive, gives a lot of help to those who don’t need it, is tilted to the affluent, etc.

Notes

Maybe systematic ignoring would be a fairer way to characterize it. ↩︎
See the classic book, Burton Malkiel, A Random Walk Down Wall Street for chapter and versus. Reading it in the early 1980s transformed my approach to investing and put me on the path to prosperity in retirement FWIW. ↩︎
There are IMO a host of other flaws with the tax incentive system for retirement savings. Among them: Requiring people to choose between Roth and traditional structures creates immense complexity and adds a lottery effect to the amount of the benefit received. The Roth structure encourages policy makers to undercount the fiscal cost of providing retirement incentives compared to the traditional structure, because it pushes cost outside the budget window. Because the size of incentives rise with tax rates, both types give bigger incentives to individuals with higher incomes and tax rates, those who are less likely to need them. Affluent people typically save without incentives. That only scratches the surface of the policy problems with our system of retirement incentives. ↩︎
Of course, comparing the worst and best case provides an exaggerated view of the difference. A more balanced approach would look at the statistics measuring the dispersion or variance in the return. Because this is a post on X, he didn’t provide that type of detail. The graph appears to show a lot of variance, and the pattern of the bars show that some age cohorts are systematic losers or winners. I personally happen to be in age cohort winners FWIW. ↩︎
My observation is that his glide path allocation much more aggressively shifts to bonds than the typical one used by Fidelity, Vanguard, etc. He has the stock allocation at 20% at the time of retirement. Vanguard and Fidelity’s target 2025 funds have their stock allocations set between 50% and 60%. So, the actual reduction is probably smaller than he suggests. ↩︎
This is what I would like to see data analysis on to potentially validate my inferences. ↩︎

I regularly read John Rekenthaler’s Morningstar column, largely because he is very good writer and often has interesting insights into investing and the financial industry (mainly mutual funds). He and other Morningstar commentators occasionally lapse into writing about public policy – mainly retirement, financial regulation, and taxation of mutual funds. I find that their policy prescriptions miss the mark (in my view) as often as they hit, mainly I think because they come at it from a different frame of reference than I do.

Rekenthaler’s last two columns address a public policy issue he has been visiting over the years – problems with America’s retirement system reliance on 401(k) and similar plans (e.g., 403(b)s) as its core source of retirement income (beyond Social Security, of course) – but now he actually is proposing a number of fixes or solutions, rather than just pointing out problems and flaws. I have generally agreed with his criticisms and now find his proposed solutions to be sensible, as far as they go (not very). Unfortunately, the hyper-partisan gridlock that defines Congress is likely to prevent any measured consideration and discussion of his or similar improvements. So, we are likely to continue to muddle along with accidental policy (401(k)’s, as Rek points out were a policy accident that employers hopped onto) that is a mess, at best.

The two columns are here: “401(k)s Have Reached Their Expiration Date” and “The New American Retirement Plan” and are both worth reading.

In my view, using voluntary defined contribution plans, like 401(k)s and 403(b)s, as a basic building block of a national retirement system has two huge flaws:

They depend upon voluntary, regular saving for the distant future – consistently setting aside a portion of earnings to be used often decades later. Many/most humans are not psychologically set up to do that; deferred gratification is not easy even if the payoff is in the near term.
They require individuals themselves to do the investing in a way that realizes a reasonable rate of return. That is an immensely complex task that the vast majority of people do not want to undertake, even if they were intellectually capable of doing so (and many/most are not). Of course, a vast pool of people/entities is out there and more than happy to do that for a fee. Unfortunately many of them are either incompetent or charge high fees or both, making it much more difficult to earn the necessary reasonable rate of return on those savings.

Thus, in my view, a voluntary, defined contribution system faces the basic problem that it must swim against strong currents of human psychology and behavior that make it difficult to save consistently and invest intelligently. The expected result is what we have for the generation that is dependent on these plans – many people have saved little or nothing and those that have saved often realize below market returns, either because they were lured into high-fee plans or bought high and sold low.

These are only the big problems; there are plethora of smaller ones with the current system. However, the mobility of employees more or less dictates a defined contribution structure. Even if employers wanted to provide defined benefit plans and they obviously don’t, that model only works for long-term employees. The practice of working for one employer for your whole career is now the exception, not the rule. And as a society, we should encourage or make it easy for people to change employers and careers. Multi-employer plans have been an abject failure – witness the problems with the plans covering miners and truckers. (Watching The Irishman gives you one idea of the problem, but not the real one.) So, we should to fix the 401(k) system.

Rek’s proposed improvements all make sense to me. They are largely based on systems in other countries, so he concludes (correctly, I think) that they are workable. His proposed changes include (read the column for the full view):

Delinking access from your employer’s choice to offer a plan – currently many employers (mostly smaller ones) do not offer 401(k) plans forcing their employees to use IRAs. Rek would take employers out of the picture altogether and create one national plan that everyone would have access to. That makes sense to me for many of the reasons he outlines.
Basic participation would be involuntary – you would be auto enrolled but could opt out of increases in contributions. He allows as how he’s not stuck on the involuntary element. Making participation mandatory would ensure that all workers accrue some retirement savings beyond social security; makes sense to me but I don’t see any Congress I can imagine going for it.
Stop leakage – do not allow assets to be withdrawn for anything other than retirement or disability. There is evidence of a lot of leakage – when people change jobs, to finance home purchases, etc. It is worth noting that preventing leakage seems to run counter to the thinking of some Republicans in Congress who are proposing to allow people to tap their social security benefits to pay for child care! I really don’t get that.
Mandating some sort of default, diversified investment options (undefined) that would insure a basic market return at low fee levels. Other countries apparently manage to do this.

You can debate whether America has a retirement crisis or not. I think it does, but it’s a slow decay, not a rapid burn. As a result, Congress is unlikely to address it. Congress should at least have the discussion Rek’s columns suggest but it likely won’t because it will inflame opposition from many quarters, I’m sure. It always does and that apparently freezes any meaningful congressional action. I expect that members of Congress will content themselves with doing more meaningless stuff like the SECURE Act. They won’t even discuss small bore issues like:

Why do we have both Roth and traditional IRAs, 401(k)s, etc.? The amount of needless time and energy spent by savers trying to divine which flavor IRA to use is staggering. Just Google “Should I contribute to a traditional or Roth IRA?” I may write a separate post that explains why I think Roths are the version that should go.
Along similar lines, why don’t we have one plan for the self-employed? Why should they need to figure out whether a SEP IRA, Simple IRA, a Keogh, etc. is better for them?
Are the tax benefits more generous than necessary to encourage funding a basic retirement? Does Mitt Romney really need to be allowed to have a $50+ million IRA? (How did he ever accumulate that much? He never released the relevant tax returns to allow determining that; a fact that now pales in the light of the current occupant of the office refusal to release any returns at all.) Or should there be a cap on contributions or IRA assets, just as there are limits on the benefits that can be paid under defined benefit plans? Obama proposed a $3.5 million cap, I think. That might be too low, but the concept sure seems reasonable.
There is a long litany of similar and even smaller-bore issue that should be addressed or at least debated.

After I initially posted this, Rekenthaler followed up with a third column/post that responds to criticisms or other feedback that he received about his proposal. This followup contains a useful summary of what he is proposing and made me think about another implication of his proposal, because he obliquely addresses it:

By going to one national plan (i.e., employers no longer would voluntarily decide whether or not offer 401(k)s to their employees), that opens up the possibility – indeed, it seems like a certain implication – that employers would no longer be constrained by nondiscrimination rules. Nondiscrimination rules likely drive many employers who do participate (particularly smaller one) to provide employee matches, because doing so is necessary to satisfy the nondiscrimination rules that allow participation and contributions by highly compensated employees and owners.
If that occurs and I’m sure it would, the system could tilt a little bit even more to favoring high income individuals. At least, I’m guessing that employer matches drive a lot of contributions. This offsets a bit the benefits of making access easier for employees of firms that don’t offer 401(k)s.
It’s unclear to me how to fix that in a sensible way, unless you mandate matches if owners and highly compensated employees participated at some level in the previous tax year. This could be done similar to the current nondiscrimination rules, I would suppose. The linkage would be less clear and harder for elected officials to justify I would think.