Among many commentators it’s taken for granted that most Americans – in particularly lower-income households – need to save more for retirement. In response, more than half of U.S. states are establishing so-called “auto-IRA” programs that would automatically sign up workers without a 401(k) for a government-run retirement savings plan. There are two problems with these proposals. First, it’s not clear that Americans, including low earners, really do need to save more. And even if they do, new research raises doubts whether auto-IRA plans would succeed in raising household saving. Policymakers need to think carefully about how such plans are designed.
To start, I’ll come clean: I don’t think there’s a “retirement crisis,” either in general or among lower-income Americans. Average retirement plan contributions are about one-quarter higher as a percentage of wages today than in the 1970s; retirement plan assets are about six times higher relative to wages than in the 70s; and more retirees are collecting private retirement plan benefits than ever before.
Things are similar for lower-income retirees. For instance, the Federal Reserve’s Survey of Consumer Finances shows that from 1989 to 2016, inflation-adjusted incomes for households aged 55 to 61 with a high school education or less rose by less than 5%. But for less-educated households aged 62 to 68, who are mostly retired, real average incomes rose by 25%. That’s a big difference. Likewise, research by economists at the Investment Company Institute and the IRS shows that the poorest 20% of retiree households typically had retirement incomes that exceeded their incomes immediately prior to retirement. Other research published by two Census Bureau economists shows that the poverty rate in retirement is significantly lower than for working-age households, which should raise the question of why low-income workers would feel the need to save more for retirement.
Yes, we’d like everyone to be richer. But that’s not what a retirement system does. A retirement system makes you poorer during your working years – more accurately, it reduces the income you have available to spend – in exchange for raising your income in retirement. The goal is to be able to maintain the same standard of living between work and retirement. There really isn’t a ton of evidence that the U.S. retirement system isn’t currently doing that for the great majority of retirees, including the poor.
Nevertheless, much of the retirement policy community – along with many states – believe the poor should save more. The means to do that is by automatically enrolling employees who aren’t offered a retirement plan at work into Individual Retirement Accounts administered by state agencies. To be clear, for many Americans this will be a good thing. Many employees aren’t currently offered a retirement plan at work and many Americans of all income levels would benefit by saving more for retirement.
But if we push low-income workers who don’t need to save more for retirement to do so, one reaction might be to borrow more elsewhere. Sure, state auto-IRA plans would let you withdraw if you chose, but inertia is strong. Many households might remain in the IRA plan and have contributions automatically deducted from their paycheck, but then borrow in order to make up for the lost take-home pay.
A new research paper by John Beshears, David Laibson, Brigitte C. Madrian (all of Harvard), James J. Choi (Yale) and William L. Skimmyhorn (United States Military Academy) provides some useful new information on the effects of automatic enrollment in retirement accounts. This group of prominent behavioral economists explored the effects of the U.S. Army’s 2010 decision to automatically enroll its civilian employees in the federal government's version of a 401(k), called the Thrift Savings Plan (TSP). What makes the paper really interesting is that the authors had access not only to the employees’ TSP contribution records, but also to credit reports showing employees’ levels of consumer, auto and mortgage debt. While it’s not the full financial picture, it’s more information than previous studies have had access to.
The authors find, as expected, that automatic enrollment increased employees’ retirement plan contributions. Over the first four years of auto-enrollment, employee contributions increased by an amount equal to 2.6% of their first-year pay. (Since this is over four years, average annual contributions would rise by roughly 0.65% of salary). Along with employer matches equal to 3.2% of pay, total retirement contributions over four years rose by 5.8% of first-year salary. That’s about a 14% increase in contributions versus employees who weren’t automatically enrolled.
But household debt also increased following automatic enrollment. For the average federal employee there was no statistically significant change in consumer debt, such as credit card balances. But auto loan debt increased by an amount equal to 2.0% of employee wages relative to workers who weren't auto-enrolled. Mortgage debt increased by 7.4% of pay. Thus, automatic enrollment appears to have increased total debt by 9.4% of pay, which is 62% more than retirement plan contributions increased.
Things are even more worrying when we look at less-educated employees, who are a target audience for state auto-IRA plans. For employees with a high school diploma or less, total retirement plan contributions rose by 7.1% of first-year pay, of which 3.3% were new employee contributions. Since this group has lower voluntary opt-in participation rates in retirement accounts, auto-enrollment has a larger positive effect on contributions. But debt rose substantially more, with consumer debt rising by 2.1% of pay, auto loans by 3.7% and mortgage debt by 17.0% of pay, for a total debt increase of 22.8% of first-year salary. In other words, debt rose by three times more than total retirement contributions and 7 times more than employee contributions.
| Impact of Auto-Enrollment on Retirement Plan Contributions and Debt | ||
| All | <HS education | |
| Employee contribution | 2.6% | 3.3% |
| Employer contribution | 3.2% | 3.8% |
| Total contribution | 5.8% | 7.1% |
| Consumer loans | 0.0% | 2.1% |
| Auto loans | 2.0% | 3.7% |
| Home mortgage loans | 7.4% | 17.0% |
| Total debt | 9.4% | 22.8% |
| From Beshears, et al. (2017). All figured measured over 48 months and expressed as increase in contributions/debt as percent of first-year salary. | ||
Such results should give proponents of state auto-IRA plans and automatic 401(k) enrollment pause. I’ve myself argued for making 401(k) auto-enrollment mandatory – until now, it seemed like a no-brainer. Of all the retirement experts I’ve spoken with, I can think of only one – Michael Hurd of RAND – who has raised this issue. To their great credit, the authors of the report – who gained prominence both within and outside the academic community for their earlier work showing the effectiveness of auto-enrolment in increasing 401(k) participation rates – treat these concerns seriously.
The authors are clear that ordinary consumer debt that’s used for day-to-day spending should be netted out against new retirement savings. Consumer debt didn’t increase for the average federal employee, but for less-educated federal workers higher consumer debt offset nearly two-thirds of employee contributions.
The authors also clearly see auto loans a problematic. While an auto is technically an asset, it’s also a rapidly depreciating one. Borrowing to purchase a more expensive car is hard to paint as building household wealth.
Higher mortgage loan balances, as the authors put it (with my emphasis), “may be more benign.” A home depreciates less rapidly than a car and can in theory be turned back into cash, either by selling the home or by taking out a home equity loan. On the other hand, a larger, more expensive home is as much consumption as it is an investment. Moreover, it’s not clear that auto-enrolled employees are actually purchasing more expensive homes. What the study shows is that these employees have bigger home loans, which is also consistent with the employees having made a smaller down payments, perhaps because they had less disposable cash on hand. Auto-enrolled employees might also have felt richer, which they were due to the federal government's generous TSP match. Even then, however, rising mortgage debt threatened to overwhelm total TSP contributions even including the match.
In any case, this isn’t something to ignore. As recent research has shown, “Baby boomers are carrying more debt into retirement — most of it in home mortgages.” That’s especially true given that the average federal employee subject to TSP auto-enrollment was nearly 40 years old, meaning that larger mortgage balances may last into retirement. It troubles me that some proponents of auto-IRAs are instead calling large mortgage balances “the sort [of debt] correlated with financial health.”
What should policymakers do? Despite these findings, giving up on automatic 401(k) enrollment seems almost inconceivable. It's the most effective way short of coercion to sign people up for retirement plans. However, policymakers and employers might consider omitting the very lowest-income workers from automatic enrollment. For instance, in the United Kingdom’s new nationwide auto-enrollment retirement plan, workers with annual earnings under approximately $12,000 aren’t automatically signed up. Both the U.K. and the U.S. have progressive Social Security-type government pensions that provide high replacement rates to the very poor. Once workers' earnings rise above that level, they would automatically be enrolled. This doesn't fix the issue of rising household debt, but might reduce it. At the end of the day, it's very hard to make someone save if they don't really want to do so.
Regardless, this new information on the broader effects of automatic 401(k) enrollment on household saving is extremely useful, as it forces the retirement policy community to think about automatic enrollment in a more comprehensive way. There’s still a lot more to be learned about how households do and should save for retirement.
