The company had adopted automatic enrollment in June 1999, about a year after the concept had received its official blessing from the government. Brigitte compared the behavior of employees who were newly eligible for the plan in 1998, the year before the change, with those hired in the year after the change. Even the most clueless employees eventually figure out that joining the retirement plan is a good idea, especially in a plan like this with an employer match, so automatic enrollment mostly affects the speed with which people join. Before automatic enrollment, only 49% of employees joined the plan during their first year of eligibility; after automatic enrollment, that number jumped to 86%! Only 14% opted out. That is a pretty impressive change in behavior produced by a supposedly irrelevant factor.
Madrian and Shea aptly called the resulting paper “The Power of Suggestion,” and their analyses reveal that the power of default options can have a downside. Any company that adopts automatic enrollment has to choose a default saving rate and a default investment portfolio. Their company had adopted a 3% saving rate as the default, and the money went into a money market fund, an option with little risk but also a very low rate of return, meaning that savings would be slow to accumulate. The government influenced both of these choices. The company had no choice about the selection of the money market account as the default investment because, at that time, it was the only option approved for such use by the U.S. Department of Labor. Since then, the Department of Labor has approved a host of what are called “qualified default investment alternatives,” and most plans now choose a fund that mixes stocks and bonds and gradually reduces the percentage in stocks as the worker approaches retirement.
The choice of the 3% default investment level was also influenced by the government, but not intentionally. In official rulings such as the ones Mark Iwry initiated, there are usually specific facts included, and the June 1998 ruling included language along these lines:‡ “Suppose a firm automatically enrolls employees into a retirement savings plan at a three percent savings rate . . .” Ever since then, the vast majority of firms that use automatic enrollment start people off at that rate. Call it an unintentional default.
Both of these default choices—the money market investment option and the 3% saving rate—were not intended by the employer to be either suggestions or advice. Instead, these options were picked to minimize the chance that the company would be sued. But employees seemed to treat the default options as suggestions. Most ended up saving 3% and investing in a money market fund.
By comparing the choices of people who joined before automatic enrollment with those who came after, Madrian and Shea were able to show that some employees would have selected a higher saving rate if left to their own devices. In particular, many employees had heretofore picked a 6% savings rate—the rate at which the employer stopped matching contributions. After automatic enrollment came in, there were fewer people choosing 6% and more choosing 3%. This is the downside of automatic enrollment. And it is a good reason why any firm that adopts automatic enrollment should deploy the Save More Tomorrow plan as well.
Brigitte’s paper raised awareness about the effectiveness of automatic enrollment, but there were still no takers for Save More Tomorrow. Then, out of the blue, I got a call from Shlomo Benartzi. A financial services consultant, Brian Tarbox, had heard one of us talking about Save More Tomorrow and had implemented it. We had talked to Brian about the implementation plan, but that had been a couple years earlier and I had forgotten all about it. Brian had gotten back in touch with Shlomo and told him that he now had data and was willing to share it with us. Pop the Champagne corks! We finally had a case study to analyze.
The firm that Tarbox had worked with had started out with a problem. In the case of retirement plans, if its lower-paid employees do not join the plan, a firm can be out of compliance with Department of Labor rules that limit the proportion of benefits that can be given to its highest-paid employees. When that happens, the maximum amount any individual can contribute is reduced. Tarbox’s client was desperate to coax their lower-paid employees into saving more, so desperate that they had hired him to meet with each employee for a one-on-one financial planning session. Brian had a laptop loaded with software that could compute how much the employee should be saving, and I think the company hoped he would talk some sense into them. But they needed more than just talking. They needed a plan.
The employees at this firm were currently not saving much, and had not accumulated much in the way of retirement wealth. When Brian would run his program to calculate the employee’s optimal saving rate (the one an Econ would choose), the program would often suggest the maximum allowed at this firm, 15%. If Brian suggested to someone who was now saving 5% that they should increase to 15%, the employee would laugh. Most were struggling to make ends meet. A big increase in saving, meaning a big cut in take-home pay, was not in the cards.
Benartzi and Tarbox worked out a more moderate strategy. Rather than report the recommended saving level from the program, Brian would suggest that employees raise their saving rate by five percentage points. If they were unwilling to take this advice, they were offered a version of Save More Tomorrow.
It was good for Tarbox (and the employees) that we had given him this backup plan. Nearly three-quarters of the employees turned down his advice to increase their saving rate by five percentage points. To these highly reluctant savers, Brian suggested that they agree to raise their saving rate by three percentage points the next time they got a raise, and continue to do so for each subsequent raise for up to four annual raises, after which the increases would stop. To his surprise, 78% of employees who were offered this plan took him up on it. Some of those were people who were not currently participating in the plan but thought that this would be a good opportunity to do so—in a few months.
After three and a half years and four annual raises, the Save More Tomorrow employees had nearly quadrupled their savings rate, from a meager 3.5% to 13.6%. Meanwhile, those who had accepted Brian’s advice to increase their savings rate by 5% had their saving rates jump by that amount in the first year, but they got stuck there as inertia set in. Brian later told us that after the fact he realized that he should have offered everyone the Save More Tomorrow option initially (see figure 25).
FIGURE 25
Armed with these results, we tried to get other firms to try the idea. Shlomo and I offered to help any way we could, as long as firms would agree to give us the data to analyze. This yielded a few more implementations to study. A key lesson we learned, which confirmed a strongly held suspicion, was that participation rates depended strongly on the ease with which employees could learn about the program and sign up. Brian’s setup for this was ideal. He showed each employee how dire his savings situation was, offered him an easy plan for starting down a better path, and then, crucially, helped him fill out and return the necessary forms. Unfortunately, this kind of hands-on implementation is expensive. Some companies have tried group educational seminars, which can be helpful, but unless these are accompanied by a chance to sign up on the spot, their effectiveness is limited. And simply making the option available in some hard-to-find location on the plan administrator’s website is not going to attract the lazy procrastinators (a.k.a., most of us) for whom the program was designed. One practical solution to this problem is to make a program like Save More Tomorrow the default (of course with the option to opt out). Certainly, any firms that are still using a 3% initial default saving rate owe it to their employees to ramp them up to a saving ceiling that provides some chance of a decent retirement income. I would say saving 10% of income would be a bare minimum for those without other sources of wealth, and 15% would be better.
Both automatic enrollment and Save More Tomorrow are now finally spreading. Many firms have adopted a simpler version of Save More Tomorrow, called automatic escalation, which delinks saving increases from pay increases. It turns out that many payroll departments are not capable (or willing) to do the computer program
ming to combine the two. (Fortunately, this does not appear to be a vital feature of the program.) According to a survey conducted by Aon Hewitt that focuses on the largest employers, by 2011, 56% of employers were using automatic enrollment and 51% were offering automatic escalation or Save More Tomorrow. These high participation numbers are in part the result of a law passed in 2006 called the Pension Protection Act, which gave firms a small incentive to adopt these features.
In a recent paper published in Science, Shlomo and I estimate that by 2011 about 4.1 million people in the U.S. were using some kind of automatic escalation plan, and by 2013 they were collectively saving an additional $7.6 billion per year as a result. The United Kingdom has recently launched a national personal saving plan that utilizes automatic enrollment, and so far the opt-out rate for those employees who were subject to automatic enrollment has been about 12%. There is talk of adding automatic escalation later on. Similar programs also exist in Australia and New Zealand.
One question we were often asked and were unable to answer was whether this sort of automatic saving really increases a household’s net worth. Perhaps, some argued, once enrolled, participants reduce their savings elsewhere or take on more debt. There are no American data sets that have adequate information about household wealth to allow this question to be answered. But a team of American and Danish economists led by Harvard’s Raj Chetty, a rising star in economics, have used Danish data to provide a definitive answer to this question, as well as to the more general one discussed earlier about whether the tax-free savings aspect of retirement plans is effective in increasing savings. They were able to do so because the Danes keep meticulous records on household wealth as well as income.
There are two principal conclusions from the Danish study. The first is that the bulk of the saving generated by automatic saving plans is “new.” When someone moves to a company with a more generous retirement saving plan and automatically starts saving more via that plan, there is neither a discernible decrease in savings in other categories nor an increase in debt. In a world of Econs this result would be surprising because Econs treat money as fungible and are already saving just the right amount, so if an employee is forced or nudged into saving more in one place, she would just save less or borrow more somewhere else. The second conclusion compares the relative contributions of two factors that are combined in these plans: their automatic features and the tax break gained by saving in a tax-free account. In allocating the source of the new saving that comes from these programs, the authors attribute only 1% of the increase to the tax breaks. The other 99% comes from the automatic features. They conclude: “In sum, the findings of our study call into question whether tax subsidies are the most effective policy to increase retirement savings. Automatic enrollment or default policies that nudge individuals to save more could have larger impacts on national saving at lower fiscal cost.”
In 2004, several years after Brian Tarbox ran that first experiment, Shlomo and I wrote a paper about the findings. The first time I presented the research at the University of Chicago, it was at a conference in honor of my thesis advisor, Sherwin Rosen, who had recently died prematurely at age sixty-two. The discussant of our paper was Casey Mulligan, one of the several remaining hard-core Chicago School economists in the university’s economics department.
The findings of our paper fly in the face of much of what Mulligan believes. We were able to get people to save more simply by using supposedly irrelevant plan design features. An Econ would not enroll in Save More Tomorrow because he would already be saving the right amount, and if he did enroll, it would not affect his saving rate, because he would make adjustments elsewhere to get himself back to saving the optimal amount that he had previously chosen. Mulligan grudgingly admitted that we did seem to be able to perform this black magic, but he worried that we were up to some kind of mischief. He thought we might be tricking them into saving too much. Of course, I was thinking to myself that if people are as clever as rational choice adherents like Mulligan typically assume, they would not be so easily tricked, but I left this unsaid. Instead, I conceded it was possible that we could induce people to save more than the optimal amount an Econ would choose, though this seemed unlikely given the low rates of personal saving in the United States. Still, as a precaution, we built in a maximum saving rate after which the automatic saving increases would cease.
Furthermore, if a household is going to miss its ideal saving target, it seems better to overshoot the desired retirement nest egg than to save too little. I am not taking a position on how people should allocate their consumption over their lifetimes, and surely there are many misers who have lived appropriately miserable lives. Instead, I am concerned with the difficulty of forecasting the rate of return on savings, and the ease of making adjustments later in life. Someone turning sixty who finds herself flush with surplus savings has numerous remedies, from taking an early retirement, to going on lavish vacations, to spoiling the grandchildren. But someone who learns at sixty that she has not saved enough has very little time to make up lost ground, and may find that retirement must be postponed indefinitely.
Casey Mulligan ended his discussion with a question. “Yeah,” he said, “it seems like you can get people to save more. But, isn’t this ‘paternalism’?”
At the University of Chicago, you can call someone a Marxist, an anarchist, or even a Green Bay Packers fan (the archrival of the Chicago Bears, the local NFL team), but calling a colleague a paternalist is the cruelest cut of all. I was genuinely puzzled by this accusation. Normally we think that paternalism involves coercion, as when people are required to contribute to Social Security or forbidden to buy alcohol or drugs. But Save More Tomorrow is a voluntary program. I said as much and went on to say that if this is paternalism, then it must be some different variety of paternalism. Struggling for the right words, I blurted out: “Maybe we should call it, I don’t know, libertarian paternalism.”
I made a mental note to discuss this new phrase with Cass Sunstein the next time I saw him.
________________
* Economic theory does predict that the total nest egg people accumulate will go up if saving is made tax-free; it just does not say whether saving contributions will go up or down, and as a society we care about both. Here is an analogy. Suppose you trade your old car for a new one that is twice as fuel-efficient. If you are an Econ you will drive more miles since the cost of driving has gone down, but it is unlikely that you will buy more fuel.
† Brigitte did not remain a skeptic for long. She soon partnered with David Laibson and a rotating group of coauthors to replicate and extend her original findings. She and David are now prominent experts in the field of retirement saving design.
‡ Iwry and his team landed on 3% merely because a low level would be less likely to arouse opposition and at the very least would establish the guiding principle. By 2000, his team tried to recalibrate to a higher level through various other rulings, but the initial anchor stuck.
32
Going Public
When I next saw Cass, I told him about my new term, “libertarian paternalism.” The phrase was not beautiful, but he had to admit it was more constructive than his term, “anti-anti-paternalism,” and he was intrigued.
The notion of paternalism was very much on the minds of behavioral economists at the time. Colin Camerer, George Loewenstein, and Matthew Rabin had collaborated with Ted O’Donoghue and law professor Sam Issacaroff on a paper with a similar idea and an equally forbidding title: “Asymmetric Paternalism.” They defined their concept this way: “A regulation is asymmetrically paternalistic if it creates large benefits for those who make errors, while imposing little or no harm on those who are fully rational.” Rabin and O’Donoghue had earlier coined the phrase “cautious paternalism” but then raised their ambitions to “optimal paternalism.” We were all trying to dig into the question that had been the elephant in the room for decades: if people make systematic mistakes, how should that affect
government policy, if at all?
Peter Diamond happened to be serving as president-elect of the American Economic Association in 2002 and was in charge of organizing the annual meeting, to be held in January of 2003. Peter was an early fan of and contributor to behavioral economics, and he took the opportunity to organize a few sessions at the meeting on behavioral topics and invited a session on paternalism. Cass and I wrote a short paper that introduced the idea of libertarian paternalism. With the five published pages we were allotted, Cass was barely getting warmed up, so he took that piece and developed it into a proper law review article, over forty pages. We called it “Libertarian Paternalism Is Not an Oxymoron.”
When I printed a draft of the law review version of the paper it looked quite long to me. One day I asked Cass whether he thought there might be a book in it. It would be an understatement to say that Cass loved the idea. There is nothing Cass relishes more than writing a book.
The premise of the article, and later the book, is that in our increasingly complicated world people cannot be expected to have the expertise to make anything close to optimal decisions in all the domains in which they are forced to choose. But we all enjoy having the right to choose for ourselves, even if we sometimes make mistakes. Are there ways to make it easier for people to make what they will deem to be good decisions, both before and after the fact, without explicitly forcing anyone to do anything? In other words, what can we achieve by limiting ourselves to libertarian paternalism?
Misbehaving: The Making of Behavioral Economics Page 34