Tax policy colloquium, week 9: Damon Jones on responses to IRA early withdrawal penalties

Yesterday at the colloquium, Damon Jones of the University of Chicago Harris School of Public Policy presented "How Do Distributions From Retirement Accounts Respond to Early Withdrawal Penalties?," an empirical study using IRS data. (Damon's coauthors were Gopi Shah Goda and Shanthi Ramnath.) But before getting to the paper, a bit of background:

"Is it hard to make arrangements for yourself / When you're old enough to repay, but young enough to sell?"

So asked Neil Young at age 25. But by the time you're 59-1/2, while one hopes you've repaid and built up some equity (if you're a homeowner), are you still young enough to sell? It can be a transitional age, as I know from recent personal experience - rather late to start a new business or career or start living in a new place (except for a few relatively privileged and successful people), rather early to be thinking about retirement, and - at least for significantly older age cohorts than my own, when people tended to marry and start raising families (if that was their path) by their early to mid twenties - a bit late to be putting one's kids through school.

It's thus always seemed a bit odd to me that individual retirement account (IRA) early withdrawal penalties - discouraging withdrawals that undo the IRA provisions' aim of encouraging retirement saving - cease to bite when one reaches the particular age of 59-1/2.  That's awfully early from a retirement standpoint, yet a bit late for some possible uses of the funds - e.g., handling major life cycle expenses or charting a new course in life.

I gather that the use of this age dates from 1974 ERISA legislation, which included an IRA provision although prior to the full-fledged IRA boom. This was a time when "normal" Social Security retirement started at age 65, with "early" starting at age 62. I'm told that someone or other who was guiding the ERISA legislation, possibly at the staff level, apparently figured that age 60 was about right for permitting people to withdraw retirement savings without penalty, perhaps on the ground that voluntary retirement savings had to be more leniently structured than the mandatory kind that Social Security offers, or else people wouldn't opt in sufficiently.  The reason for then picking 59-1/2, rather than 60, was to make it seem more appealing still to the prospective participants, just as retail stores offer 99-cent pricing.

Anyway, IRAs to this day have early withdrawal penalties. For traditional IRAs, you're taxed on the withdrawal (in effect, under normal income tax rules), but with the addition of a tax penalty equaling 10% of the amount withdrawn. There is a hardship exception to owing the penalty, but it's fairly narrow - covering, for example, death or disability, unreimbursed medical expenses, and health insurance premiums while unemployed.

Evaluating the withdrawal penalty requires a deep dive into theories of lifecycle optimization - how would people generally be expected to optimize the allocation between periods of their lifetime resources?  And secondarily, why would the government seek to influence what people do? Here the main rationale is behavioral - e.g., if we think that people are prone to irrationally under-saving for retirement - but there are also aspects of possible market failure (e.g., difficulty in life-annuitizing sufficiently given adverse selection, or on the other side difficulty of borrowing against future earnings) and moral hazard (e.g., expectation of being supported at retirement if one under-saves).

It's often said (as a convenient, if over-simplified, shorthand) that the goal is to increase consumption-smoothing, which at the limit would mean equal consumption in all periods, if all periods are otherwise the same and one has period-specific declining marginal utility of consumption. But of course there can be rational reasons, unrelated to market failure or moral hazard, for favoring higher or lower spending in different periods. These might range, for example, from one's taste for consumption when young versus old, or alternatively for a particular pattern such as rising consumption or for periodic "binge" years. One also might have periods with especially high needs (e.g., to launch one's children or pay uninsured medical costs).

But the standard conclusion, which I certainly accept, is that the main aim calling for policy intervention with regard to saving - leaving aside the question of borrowing against future anticipated resources - is to push people towards greater retirement saving. Only, one should think about the possibility that they will want to take back some of the savings sooner than anticipated. Now, just to make voluntary retirement saving more attractive than it would otherwise be, one might want to allow some of this. If the saving is voluntary, rather than mandatory like Social Security (assuming effective barriers to borrowing against expected future Social Security benefits), then one reason for allowing early withdrawal - even where it might undo the policy to a degree - is to reduce people's reluctance to participate. But in addition, unanticipated shocks may contribute to early withdrawal's being apparently optimal in some cases, and not just "leakage" that reflects the reasons for expecting too-low retirement saving.

The Jones et al paper makes an ingenious use of IRS data to examine the effect of the 10% early withdrawal penalty on behavior around traditional IRAs. It examines withdrawals during a 5-year window for people, around the year in which they turned 59-1/2. As it happens, given the period covered, these were people born between 1941 and 1951. The data includes people's birthdays, how much they withdrew in a given taxable year, and what they paid in penalties. Since the actual withdrawal dates aren't known (other than whether they led to a penalty), the key distinction is that between people who turned 59-1/2 early versus late in the middle year. If this date was early in the year, then (a) any penalty paid could have been avoided by waiting not all that long, and (b) there was more time in which withdrawals could be made after the penalty had ceased to apply (a distinction that one could imagine not mattering all that much, but in the data appears to have mattered).

A central finding was that people very much did respond to the early withdrawal penalty. In other words, it discouraged prior withdrawals, as it was meant to. This would have been an obvious result if we could assume that people are well-informed and acting rationally, but given questions about that it was worth establishing empirically. One imagines that the prominence of this date, perhaps including in brokers' solicitations and the like, would help to fix it in people's minds, along with the strong evidence from behavioral literature that people hate "penalties."

But a secondary finding in the paper is that a distressingly high percentage of individuals, many of them low-income, incur the early withdrawal penalty when it seems relatively irrational to have done so - e.g., when one's turning age 59-1/2 must have been relatively close, rather than being far enough in the future (such as December) that even a 10% penalty might have been preferable to, say, multiple months of high interest rate credit card debt.

Especially as applied to lower-income individuals, the reason for the penalty is so that it won't be incurred and people will retain their retirement saving. But it uses a cliff, and if it's being incurred too frequently that might indicate a need to rethink the discouragement design.

A related issue concerns the main reasons for favoring both greater retirement saving and an ability to access funds pre-retirement (and also potentially to borrow) in response to great needs. Just as we want people to have adequate retirement saving, so we want them to be able to meet major medical needs, smooth consumption when they lose their jobs, handle disability expenses, etcetera. And both choice failures and market failures (along with having low lifetime income) may undermine their doing this adequately.

When one is thinking about these various needs, whether incurred later or sooner, there are two complementary perspectives that one could have in mind. The first is optimization. Is one making the best use of one's lifetime income, treating that as fixed? In principle, improving someone's lifetime optimization, such as by adjusting for choice failures or solving market failures that the government is better equipped than private firms to address, can make her better-off at zero cost to everyone else. (If this sounds paternalistic, it is - but if one wholly rejects it then one should question the existence of anything like Social Security. A key mitigating idea is that, much of the time, one will only be forcing people to do things that they wanted to do anyway. E.g. I personally am not being forced by Social Security to over-save for my retirement. It's in effect just a floor on my retirement saving.)

The second approach I'll call adequacy, for want of a better word. We may want to make sure that people can meet basic needs of retirement, sickness, disability, etcetera.

The optimization perspective arguably supports having Social Security as a forced retirement saving program that could in principle be actuarially fair as to everyone (although of course in practice it transfers resources between age cohorts, different types of households, etc.). The adequacy perspective might call for, say, paying demogrants to retirees, and then separately deciding on the financing for this benefit as just one more input to one's overall distribution policy. Likewise, it might call for a general safety net approach to earlier needs arising from unemployment, accident, sickness, etcetera.

While we employ aspects of both approaches in the U.S. fiscal system, a more generous social safety net and approach to adequacy might ease (although not eliminating) our concerns about optimization, by mitigating both the worst failures of optimization and our conviction that there are failures (given the connection between consumer choice and reasonably presumed utility). Thus, for example, returning to IRAs and early withdrawal penalties, the case for allowing hardship withdrawals without penalty would be eased if the needs that most obviously might trigger this approach were better approached by our fiscal system from the adequacy perspective.