Could Employer Contributions Actually Lead to Leakage?

I recently stumbled across an academic study that claimed to find a correlation between higher employer contribution rates and leakage.

I will confess to a certain skepticism at that finding. There are, after all, a well-established series of things that contribute to leakage, broadly defined as distribution of retirement savings prior to retirement – but employer matching contributions – and certainly more generous matching contributions – have never been on that list.

The study – innocuously titled “Cashing Out Retirement Savings at Job Separation” – spends most of its 20-odd pages talking about leakage, its impacts on retirement security, and some possible solutions.  That said, one needs read no further than the abstract of this paper to find its surprising conclusion regarding one such underlying cause; its authors “estimate that a 50% increase in employer/employee match rate increases leakage probability by 6.3% at job termination.” More specifically, “The higher the proportion of one’s 401(k) balance contributed by the employer, the more likely employees are to cash out, holding constant balance and covariates.”

Proportion ‘Ate?’

That latter part is significant, since we know that participants with lower balances are more likely to have their balances distributed at job separation (so-called “force-outs” being typical at $1,000 or less). In fact, the paper acknowledges that “A higher balance discourages leakage holding all else constant.” Even so, a 6.3% increased probability might be “statistically significant,” but it most assuredly isn’t significant in economic terms. But to see any kind of connection between a more generous employer match and leakage just seemed – unusual. Particularly since – and as the study’s authors acknowledge – “Employers with more generous matches care about their employees’ well-being in retirement, but unintentionally nudge employees to cash out when they change jobs.”

The research cites a relatively robust sample (162,360 employees terminating from 28 retirement plans form 2014-2016 from a recordkeeper “that covers 15% of the U.S. workforce”), from a variety of industries. They acknowledge that the cash-out percentage (41.4% of employees cashing out at job separation) in this sampling is “strikingly high,” although in this group[i] – though interestingly “only 27.4% of terminating employees ever carried a loan, and only 3% of those defaulted.” The latter data point stands out because previous studies have found that outstanding loans defaulted at job separation are a significant cause of leakage. And – while averages are notoriously unreliable datapoints, the terminating participants in this sample had an average account balance of $46,556.[ii]

Reasons Able?

Of course, these researchers were looking for a connection between employer contributions and leakage – and, having found one – held out four possible rationales for that connection. First, they considered a scenario where workers, cognizant of the higher match actively planned to “leak” – basically “over-saving” to obtain the match, cutting into the income they actually needed for current expenses, and then needing the leakage to fill that hole. Secondly, they opined that a higher employer contribution rate during employment might engender a higher level of job security, and a correspondingly higher spending rate by the worker – that, upon termination, might then need to be funded by higher rate of withdrawal/leakage. Thirdly, they thought that workers might retain a sense of mental accounting that compartmentalized the employer match as “free” money, rather than sums set aside specifically for retirement (though the leakage impacted more than that account). Finally – and this is the rationale they landed upon to explain this “account composition” effect – that individuals who contributed a smaller proportion of their 401(k) balance (relative to the match) may be prone to think of their accounts at job separation as a readily spendable pile of cash (less so if one contributed more).

All of this felt to me like they were trying (too hard?) to rationalize behavior that wasn’t “rational.” That said, the researchers nonetheless conclude that “exiting one’s firm and being told that a sum is available can transform a perceptually illiquid source of long-term retirement security into a psychologically liquid pile of cash. Terminating employees spend the money when, arguably even for the minority of employees involuntarily terminated, there are good options of reducing household spending, adding gig forms of employment, or leveraging home equity lines of credit to supplement unemployment benefits until back in the workforce.”

Ultimately, it was impossible to really get inside the numbers and assumptions presented to ascertain how much of this conclusion was data-based versus “extrapolation.” The contributions labeled as matching looked to be more than just standard matching, perhaps including QNECs or safe harbor contributions as well, but there wasn’t enough detail in the paper’s tables to confirm that. As noted above, the withdrawal rates were high, and the “average” account balance presented clearly covered a wide variety of possibilities. And let’s not forget that, even with those considerations, the additional rate of leakage attributed to these generous employer contributions was pretty small.

There is, however, at least one conclusion worth drawing from this – and that’s that if the worker considers these accounts “free” money – and goodness knows, the employer match has long been positioned as such – they might well not realize the price they will pay, both at the point of distribution (taxes and penalties) – and ultimately at retirement – for spending those retirement savings…now.

- Nevin E. Adams, JD  

---

[i] Another aspect of this group that struck me as odd – only about two-thirds of this group took a one-time total cashout, whereas another 21% depleted their 401(k)balances in two or more withdrawals within eight months.  One would normally expect traditional leakage patterns to be tied to a single withdrawal, rather than a series.

[ii] With an understandably large standard deviation of more than $97,000 – I say understandably because individuals with that size account balance tend to stay with the plan (an easy default) or rollover to an IRA or other plan). As the authors acknowledge, “A higher balance discourages leakage holding all else constant.”

Has the 401(k) Passed its ‘Expiration Date’?

That’s the premise behind a recent column by Morningstar’s John Rekenthaler, who writes that “the plans are as good as they can be under the current framework – and that's not good enough.”

I had the pleasure of meeting John a number of years back – and I’ve been keeping up with his writing ever since. His columns are thoughtful and thought-provoking, his perspectives rational and well-reasoned, his commentary nearly always not only interesting, but entertaining. But on this one – well, let’s just say we disagree.

John acknowledges that his views on the 401(k) have “evolved,” and that while he has long been in the camp that called for improvements in the current system, a “defender” of the 401(k) – but now, apparently, he’s calling for an “overhaul.”

‘Leaky’ Assumptions

He doesn’t fault the current system for its perceived shortcomings; he notes that the 401(k) wasn’t designed to be a solution for the general public’s retirement, saw the growth in the 1980s and 1990s as “modest,” and while he apparently saw the advent of automated design solutions as “solving half the country’s problems,” that confidence seems to have been shaken by a couple of academic studies. And that’s where things begin to get shaky.

“Whether auto-enrolled or not, leakage from 401(k) accounts, caused by early withdrawals, is substantial,” he writes, going on to cite one paper that “shows for households under the age of 55, average 401(k) outflows equal 40% of the inflows. Another study, by Laibson's co-researcher Beshears, estimates that, within the first four years, 25% of the benefits of automated-enrollment programs are consumed by early withdrawals.”


Now, John’s not the first to have logic waylaid by respected academics. One of these “studies” got picked up in the Wall Street Journal back in 2018. I encourage you to revisit my analysis of that survey at your leisure, but here’s the salient part(s): this study is based on activity at a single firm. One. Granted, it’s described as a large (approximately 7,500-participant), Fortune 500 financial services firm – but it’s one that even the researchers concede has high turnover. It’s also, based on the salary information provided, one with relatively modest income workers. However, even with those impediments, automatic enrollment did “work” – transforming the plan’s participation rate from 62% to 98%.

That said, you can imagine what happened when the employer in question adopted automatic enrollment in a plan of modest income workers; you get more, albeit arguably smaller, account balances (also their contribution default was just 2%). And with a workforce that has high turnover – those smaller balances are more likely to be cashed out – and then create “leakage.” And apparently in this isolated circumstance, with contribution rates (and amounts) so low and turnover so high you can actually get a result that allows you to conclude with a straight face that the withdrawals add up to 40% of the inflows. A conclusion that might well be transformed by the casual reader into an assumption that the result could rationally be imputed to the 401(k) system at large  And then (with a similarly straight face) hold that  out as though that is in any way representative of the 401(k) system overall.

All in all, Rekenthaler’s issues with the 401(k) appear to be two-fold: the aforementioned issue with leakage[i] – and coverage. The former he apparently wants to rectify by “replacing early withdrawal tax penalties with a stronger deterrent.” However, since he appears to want to preserve a participant’s right to “opt out” of this new design, it seems fair to worry that restricting access to those funds will almost certainly diminish the amount(s) that workers are willing to set aside in those accounts. In other words, there may be less coming out – but then, there might well be (much?) less going in.  

Access Able?

As for coverage, Rekenthaler apparently wants to solve with some kind of program such that “Every worker at every company, in every state, in every industry, will have access to a New Retirement Plan.” Apparently he’s not referring to Social Security, though of course, it already extends that far.

However, when it comes to coverage, I share Rekentaler’s concern. As we’ve commented before, with all of its success, and expansion, with all the trillions of dollars now set aside for retirement, coverage remains an issue. We’ve estimated that nearly 5 million employers nationwide do not provide a retirement plan to employees, and as a result some 28,280,000 full-time employees still do not have access to an employer provided retirement plan.

Now, Rekenthaler’s hinted at a solution (he promises to unveil it today) – that involves “removing the employer from the system.”
And that’s where I think he makes a (really) big mistake.

There’s no doubt that a system that relies both on voluntary action on the part employers and an active response by workers will inevitably miss some – and indeed there’s no getting around the reality that, 40 years on, retirement plan coverage has, basically, “flat-lined”.

That said, there’s already a version of “removing the employer from the system” in place, of course – it’s the state-run programs for private sector workers in Oregon, and more recently in California. Proponents tout the contribution and participation rates of those programs as a success and – relative to their non-participation outside those programs–,that’s a fair assessment.

Of course, those contribution (5.5%) and participation (approximately 70%) rates pale in comparison to private sector plan experience, where participation rates in automatic enrollment plans typically exceed 90% and where the 62^nd annual Plan Sponsor Council of America (PSCA) 401(k) survey recently found a combined employer/employee contribution rate of 12.2%.

The involvement of the employer is even more compelling when you consider that even workers of relatively modest means (those earning $30,000-$50,000/year) are 12 times as likely to save for retirement ina tax-advantaged account if they have access to a plan at work.

There’s something to be said for making it easier to transport those accounts from one employer to another, for making it easier to roll them over into an IRA than to simply take the distribution in cash. That said, the problem seems not to be the 401(k) plan, but the lack of 401(k) plans – plans that thrive because of the support and encouragement of employers – not just in enrollment and education, but in the matching contributions which often accompany such undertakings. 

It’s been said that there’s no use crying over spilled milk. But it seems to me that discarding the 401(k) as “expired” would be like throwing out milk weeks ahead of its “best if served by” date.

- Nevin E. Adams, JD

---

[i]With regard to leakage, while it’s an issue of some concern, none other than the non-partisan Employee Benefit Research Institute (EBRI) has previously considered the issue, and found that cashouts at job change were found to have a much more serious impact on 401(k) accumulation than either plan loan defaults or hardship withdrawals (even with the impact of a six-month suspension of contributions included, and though thanks to recent legislation, that should be less of an issue going forward. How much more? Well, cashouts at termination were approximately two-thirds of the leakage impact. 

Easy Come, Easy Go?

Earlier this year, I commented that it would be interesting to see how expanded access to hardship withdrawals might impact that activity. Now we have some answers.

There’s more than a little irony in a legislative body that has long bemoaned both the paucity of retirement savings and the nefarious impact of “leakage” (pre-retirement withdrawal of retirement savings) opening those floodgates a little wider – but mostly the new law, and clarifying regulations[i] seemed to provide plan sponsors a bit more flexibility in administering these programs, some welcome latitude in helping their workforce navigate choppy financial waters.

What remained unknown was – would participants take advantage – or, more precisely, would they abuse the privilege.

The first sign – and it was a bit of an eye-opener – came from Fidelity who, in a white paper, claimed to have seen a shift in participant behavior. Not in the percentage of participants taking loans and hardships overall, but - at least among Fidelity-recordkept plans, there were fewer loans and more hardships. Less than 6 months after the law took effect, based on the evidenced trends, they predicted that the annual loan rate for 2019 would dip slightly to 9.2%, while the annual hardship rate would rise to 4.4% – up from about 3% in 2018 and an average rate of 2.2% since 2009, according to their data.

Intrigued, I posed the question to NAPA-Net readers in early October – and the responses indicated that while the move to embrace the option was underway, it was – well, it was still underway. Just over a third (35%) said the rules were already in place at most of their clients, and nearly a quarter (24%) said they were already in place at all of their clients, while another 23% said they were in place at “some” of their clients. The rest were in “not yet” territory. The most surprising aspect (to me, anyway) was the lack of plan sponsor response to the impact of the changes (at least in earshot of their advisors).

Now we have a direct response from plan sponsors, courtesy of a “snapshot” survey on the subject by the Plan Sponsor Council of America. The PSCA found that nearly two-thirds (65%) of respondents already have adopted the new hardship provisions. However, most (73%) reported no change in the number of hardship withdrawals since the new provisions were implemented. Fewer than one in five (17.8%) noted an uptick in hardship withdrawals in 2019 – but even among those that did, the vast majority (92%) are not considering any changes to their provisions at this time. Indeed, most plan sponsors amended their plans even where change was not mandatory; 65.1% eliminated the requirement to take plan loans before taking any hardship withdrawal, and 59.6% expanded the assets available for hardship withdrawals to include earnings on 401(k) contributions. More than half (52.3%) voluntarily expanded the list of reasons that qualify for a hardship distribution.

It's still early to assess the long-term impact of these changes, of course, though plan sponsors appear to have embraced them without much hesitation. And if the recent take up rates have been pretty much status quo – well, the markets and unemployment numbers have been good, and the natural disasters that often produce upticks in financial hardships have been muted of late. Yes, it’s early to evaluate the impact – to see if greater access will mean more access, to see if removing barriers does, indeed open floodgates, or if knowing that it will be easier to access the money, individuals are less inclined to do so.

In sum, to see if addressing the hardships of today wind up creating hardships down the road.

- Nevin E. Adams, JD

---

[i]The expanded access, of course, came courtesy of the Bipartisan Budget Act of 2018, which, among other things, set aside or made optional several of the penalties and restrictions that had long been in place to discourage usage, and/or to ensure that the request was “serious.” This included aspects like the requirement to take a plan loan first (it’s now optional), and more significantly, the suspension of contributions. The changes not only broadened not only the categories of contributions eligible for hardship (it now includes matching contributions and non-elective contributions, as well as earnings on those accounts), but also included changes in the ability to qualify for a hardship distribution in the case of casualty losses and losses associated with federal disaster areas. The IRS also loosened the rules for determining the status of a hardship – arguably lessening the burden of both requesting and approving these distributions (final regulations were published in September).

Multiplication ‘Fables’

Did you hear the one about loan defaults adding up to $2.5 trillion in potential retirement savings shortfalls over the next 10 years? How about the “$210 Billion Risk in Your 401(k)”?

Those reports were based on an “analysis” by Deloitte that claims to find that “…more than $2 trillion in potential future account balances will be lost due to loan defaults from 401(k) accounts over the next 10 years…” That’s right, $2 trillion lost “due to loan defaults” (the Wall Street Journal apparently picked the figure that matched the impact on a “typical 401(k) borrower” – more on that in a minute).

Now, when you see headlines putting a really big number on what you already suspect is a problem – in this case “leakage” – roughly defined as a pre-retirement withdrawal of retirement savings – well, you could hardly be blamed for simply accepting at face value the most recent attempt to quantify the impact of the problem.

A closer look at the assumptions behind that analysis, however, puts things in a different light.
The Deloitte paper leads with the question “How can we keep loan defaults from draining $2 trillion from America’s 401(k) accounts?” – but despite the positioning of the premise that this is a leakage issue, the loan default turns out to be only a small part of the problem.

The Deloitte authors outline the assumptions underlying their conclusions in the footnote of the 12-page document. Specifically, they draw many of their starting points from a 2014 study, “An Empirical Analysis of 401(k) Loan Defaults,” which found that with 86% of the participants that terminated employment with loans outstanding defaulted on those loans – and took their entire account balance out at the time of loan default. That report also noted that the average age¹ of the loan defaulter was 42. The Deloitte authors also draw from Vanguard’s “How America Saves 2018” that the average loan default was approximately 10.1% of the total account balance.

And then, with that as a foundation, the Deloitte authors begin to build.

The Deloitte modeling assumption relies on the notion that the vast majority of participants who default on these loans take their whole balance out of the plan. Rather than using the 86% assumption from the 2014 study, they scaled it back to a more conservative assumption that (only) 66% of participants that defaulted on their loan also took their entire account balance. They then back into the total account balance that those individuals would ostensibly have ($47.8 billion, assuming that their loan is 10% of their total balance) that they claim, based on the earlier assumptions, would be withdrawn in addition to the outstanding loan amounts. They then project growth in those totals assuming that everyone in that group is 42 (the average age of a loan defaulter) all the way out to age 65, assuming a 6% return over that period – and wind up with $2.5 trillion!

Now, there are so many assumptions imbedded in that calculation – and so many that act as multipliers on the original assumptions – it’s hard to know where to start.

To put it in individual terms, the Deloitte analysis assumes that every borrower is 42 years old, that two-thirds of these that default not only do so on a loan of $7,081 (the average outstanding loan amount), but go on to cash out their remaining account balance of $70,106 (assuming, of course, that the loan amount is approximately 10% of the balance). They then assume – and this is the assumption based on complete speculation – that there was no subsequent rollover, nor any renewal of contributions anywhere over the rest of their working lives – while also assuming that they could have attained a 6% return on those monies over that extraordinary period of time if only they hadn’t cashed out.

Still with me?

What’s obvious is that the bulk – indeed, the vast majority – of that projected impact comes not from the highlighted loan defaults – which are, in fact, a mere fraction of the terminating participants’ 401(k) balance – but from what the Deloitte authors term the “leakage opportunity cost” – basically it’s the “magic” of compounding applied to both the defaulted loan and the other 90% of the participant account balance… at a 6% rate of return over nearly a quarter century.

They say that two “wrongs”² don’t make a right – and that’s particularly true when multiplication is involved.

It’s not that the math isn’t accurate. It’s just that the answer doesn’t “figure.”

Nevin E. Adams, JD

1. There are always potential problems with extrapolating conclusions from averages. That said, the 2014 study from which those averages are drawn note that participants who defaulted on their loan were more likely to have larger loan balances than those who repaid, and to have lower household incomes and smaller 401(k) balances – in sum, not exactly “average.” ↩
2. The paper also makes some curious comments about loans and fiduciary liability – but we’ll deal with those in a future post. ↩

‘Automatic’ Transmission: Does Auto-Enrollment Create Leakage?

Most people view automatic enrollment in a 401(k) as a good thing¹ – but apparently it has a heretofore unappreciated “dark” side.

At least that was the focus of a headline in a recent Wall Street Journal article (subscription required) that asked (and answered) the provocative question: “401(k) or ATM? Automated Retirement Savings Prove Easy to Pluck Prematurely.”

That is at least how the Journal chose to position its coverage of a study based on a single firm’s experience with automatic enrollment. That study, according to the Journal, serves to “answer a question that has long concerned employers that put workers into 401(k) plans and give them the option to drop out, rather than requiring them to sign up on their own: Will auto-enrolled workers treat their 401(k)s like automated-teller machines?”

Now, I’ve heard a lot of questions over the years from plan sponsors about automatic enrollment – but never that one.

Regardless, the Journal says that the study provides an affirmative response to the question – but (and you can almost hear the disappointed sigh) “…not to the extent that the workers spend all their gains from auto-enrollment.” Nor is it the first time that the Journal has taken aim at automatic enrollment.²

Now the researchers themselves – John Beshears, David Laibson and Bridget Madrian of Harvard, and James Choi of Yale – have solid retirement plan researcher “cred.” And, comparing employees hired in the 12 months after the introduction of automatic enrollment to those hired in the 12 months prior, they find that automatic enrollment increases total potential retirement system balances by 7% of starting pay eight years after hire.

On the other hand, they also find that leakage “in the form of outstanding loans and withdrawals that are not rolled over into another qualified savings plan” (more on that in a minute) also increase – by 3% of starting pay, which they say offsets approximately 40% of the potential increase in savings from automatic enrollment. Of course, even then, they acknowledge that the “net effect is that automatic enrollment increases retirement system balances by 4-5% of first year pay eight years after hire.”

‘Ample’ Turnover?

Here’s the thing: This is the experience at a single firm. Granted, it’s described as a large (approximately 7,500-participant), Fortune 500 financial services firm – but it’s one that the researchers concede has high turnover. It’s also, based on the salary information provided, one with relatively modest income workers. And automatic enrollment did “work” – transforming the plan’s participation rate from 62% to 98%.

So, this employer adopts automatic enrollment in a plan of modest income workers – creating more, albeit arguably smaller account balances – for a workforce that has high turnover – and since they have smaller balances, more likely to be below the cashout thresholds, and thus resulting in a higher number of termination payout “leakage.”

Cash Out ‘Cache’?

The Journal tries to craft a picture that automatically enrolled participants are more likely to cash out than other participants – and arguably they may well be less committed to the savings proposition. In fact, the study found (and the Journal article notes) that more than half of the auto-enrolled participants – 59% – cashed out their savings (largely driven by terminations), while among those who signed up for the plan on their own, the figure was 43%.

But was it automatic enrollment – or the smaller balances that resulted from a 3% default contribution rate at termination – that produced that result? I think we can all sense what the answer is – but the researchers note that a greater proportion of the automatically enrolled workers had balances below the $1,000 cashout threshold. Lest we need any further affirmation, even then, the researchers note that those net results “mask substantial differences across those who remain employed at the firm versus those who separate” – with the former seeing relatively little impact from leakage.

Change ‘Ranges’

And while the researchers don’t make much distinction in the impact of different types of leakage, the nonpartisan Employee Benefit Research Institute (EBRI) has previously considered the issue, and found that cashouts at job change were found to have a much more serious impact on 401(k) accumulation than either plan loan defaults or hardship withdrawals (even with the impact of a six-month suspension of contributions included). How much more? Well, cashouts at termination were approximately two-thirds of the leakage impact.

Even the best researchers are limited by the amount and quality of the data. While they are careful to outline those factors in the study, the coverage of such things tends to gloss over the details that might impact the broad-based applicability of the results – things like the fact that these results are from a single employer, with high turnover and modest-incomes, a combination that may well mean that the amount – and impact – of the leakage is exacerbated.

Ultimately, the point of the study might well be a call to make it easier for individuals with smaller balances to leave their employer without cashing out their savings.

That’s hardly the fault of a plan design that helped a third more workers who weren’t saving do better. But it seems that if there’s the slightest possibility of a potential downside, the negative coverage is… automatic.

- Nevin E. Adams, JD

Footnotes

1. There are some administrative issues with automatic enrollment – see Why Your Recordkeeper Might Not Be an Automatic Enrollment Fan. ↩
2. Earlier this year the Journal, in an article plainly titled, “Downside of Automatic 401(k) Savings: More Debt” (subscription required), cited another academic study noting that automatic enrollment has “pushed” millions of people who weren’t previously saving for retirement into those plans – but quickly cautioned that “many of these workers appear to be offsetting those savings over the long term by taking on more auto and mortgage debt than they otherwise would have.” And back in 2013, the Journal had an article titled, “Mixed Bag for Auto-enrollment,”  claiming that “employees who are automatically enrolled in their workplace savings plans save less than those who sign up on their own initiative.” That article, in turn, built on – and cited – a 2011 article that suthoir Anne Tergesen jaw-droppingly titled, “401(k) Law Suppresses Saving for Retirement.” In the case of the latter, Tergesen glommed on to one of 16 possible scenarios, and focused on the notion that some workers would simply rely on the mechanics of automatic enrollment’s 3% default, rather than picking the higher rate that they might if they filled out an enrollment form (encouraged by things like education meetings and incentivized by things like a company match). ↩