Before describing his strategy, I need to make a technical point. The game played on Golden Balls differs from the standard Prisoner’s Dilemma in one way: if you split and the other player steals, you are no worse off than if you had stolen as well. You get zero either way, whereas in the traditional example, if one prisoner stays silent and the other confesses, the silent one gets a severe punishment.‡ Nick exploited this small difference in devising his plan.
As soon as the discussion period started, Nick jumped in and made a surprising pitch: “Ibrahim, I want you to trust me. I promise you that I am going to steal, but then I will share my winnings with you.” Both Ibrahim and the host had a lot of difficulty understanding the logic behind this offer. As Ibrahim pointed out, there would be a much easier way to arrange an even split. Both players could choose the “split” ball. But Nick said no, he was going to steal. The host, who had never heard anything like this, stepped in to clarify that any such promise was not authorized or guaranteed by the show, and the only way to assure that both players got half was for them both to split. Apparently, the discussion went on for much longer than the usual time allotted, and most of it was edited out of the televised version, which had a strict time limit. You might consider what you would do in Ibrahim’s shoes.
Poor Ibrahim was clearly under great stress, and could not fathom what Nick was up to. At one point he asked Nick, in total exasperation, “Where do you keep your brains?” Nick smiled and pointed to his head. When the host finally ended the banter and demanded that the two players choose which ball to play, Ibrahim, who had appeared to be highly skeptical of Nick’s pitch, suddenly switched from the ball he had originally selected to the other one, giving every indication that he had decided to play along and choose the “split” ball, perhaps feeling that he had no choice. Or perhaps it was one final feint.
Then came the reveal. Ibrahim had indeed selected the “split” ball, but what about Nick? Nick opened his ball, which also read “split.”
The National Public Radio show Radiolab devoted an episode to this particular show. The hosts asked Ibrahim what he had been planning to do, and he said he was planning to steal right up until the last minute. The hosts reminded him that he had given an impassioned speech about his father telling him that a man is only as good as his word. “What about that?” the hosts asked, somewhat aghast at this revelation. “Oh, that,” Ibrahim said. “Actually, I never met my father. I just thought it would be an effective story.”
People are interesting.
________________
* I agreed but with some warnings. I said that a collaboration might be unwise on at least two counts. First, I am notoriously slow. (I didn’t mention the lazy part.) Second, I worried about the “Matthew effect,” a term coined by sociologist Robert K. Merton, which states that excessive credit for any idea will be attributed to the most well-recognized person who is associated with it. Stephen Stigler, a statistician at the University of Chicago, called his alternative version of this effect Stigler’s Law (irony intended): “No scientific discovery is named after its original discoverer.” The joke, of course, is that Stigler’s Law was just a restatement of Merton’s proposition. Thierry and the group decided that we would collaborate, with the proviso that if I did not think I was adding anything, I would withdraw.
† Prospect theory was the clear winner.
‡ In the game theory literature this is called a “weak” Prisoner’s Dilemma (Rapoport, 1988).
VIII.
HELPING OUT:
2004–PRESENT
By the mid-1990s, behavioral economists had two primary goals. The first was empirical: finding and documenting anomalies, both in individual and firm behavior and in market prices. The second was developing theory. Economists were not going to take the field seriously until it had formal mathematical models that could incorporate the additional findings from psychology. With talented new behavioral economists entering the field, and even some well-established theorists such as Jean Tirole (the winner of the 2014 Nobel Prize) dabbling with behavioral models, there was continual progress on both fronts. But there was a third goal lurking in the background: could we use behavioral economics to make the world a better place? And could we do so without confirming the deeply held suspicions of our biggest critics: that we were closet socialists, if not communists, who wanted to replace markets with bureaucrats? The time was right to take this on.
31
Save More Tomorrow
Given the attention the behavioral economics community had collectively devoted to the problems of self-control, a natural place to start was ways to help people save for retirement. Designing better retirement savings programs is a task for which standard economic theory is ill-equipped. First of all, the standard theory starts with the assumption that people are saving the exact right amount (not to mention investing intelligently). If people are already doing something perfectly, how can you help? Furthermore, even if an economist did want to help out on such a project, she would only have one policy tool to play with, namely the after-tax financial return on savings. The standard theories of saving such as those offered by Milton Friedman or Franco Modigliani implicitly make the very strong prediction that no other policy variable can matter, since the other factors that determine a household’s saving—such as age, income, life expectancy, and so forth—are not controlled by the government. The government cannot change how old you are, but it can change the after-tax return to your saving, for example, by creating tax-free retirement savings plans. And yet, there is a basic problem with the use of this policy tool—economic theory does not tell us how responsive savers will be to such a change. In fact, we cannot even be sure that making saving tax-free will increase or decrease the total amount of money people put aside for retirement.
At first blush, it would seem that increasing the returns to saving by creating tax-free accounts should increase saving, since the rewards for saving have gone up. But upon further reflection, one can see that the higher rates of return mean that it takes less saving to achieve a given retirement savings goal. Someone who is trying to accumulate a specific nest egg can achieve that goal with less saving if rates of return go up.* So economic theory offers only one policy tool, the after-tax rate of return, but we don’t know whether to raise or lower it to induce more saving. This is not much help. Of course, empirical tests could tell us what effect changing tax rates will have, but until recently, it was difficult to provide definitive results. As Stanford economist Douglas Bernheim put it in his very thorough review of this literature published in 2002: “As an economist, one cannot review the voluminous literature on taxation and saving without being somewhat humbled by the enormous difficulty of learning anything useful about even the most basic empirical questions.”
One of the problems in determining the effect of a change in the tax law is that to qualify for the low tax rate investors have to satisfy other rules, such as putting the money in a special account, possibly with penalties for withdrawals before retirement. The special account may facilitate saving in two ways. First, the penalty for withdrawal acts as an inducement to leave the money invested. Second, a mental account that is designated as “retirement saving” is less tempting to dip into than a simple savings account. In fact, following the introduction of tax-sheltered retirement savings plans in the U.S., there was a heated debate in the economics literature about whether such plans were increasing saving or just shifting money from taxable accounts to tax-free accounts. Only very recently has there been what I consider to be a definitive test, which we will get to later in this chapter.
Behavioral economics offers more potential in this and many other policy domains because more stuff matters, namely, all those SIFs. I first dipped my toe into these waters in 1994 with a short paper entitled “Psychology and Savings Policies.” In it I made three policy proposals that drew on behavioral insights. The first two were aimed at the then-popular savings vehicle called the Individual Retirement Acco
unt or IRA. (They became less important when income limits for eligibility were tightened and employer-based retirement savings plans, such as 401(k)s, became more common.) At the time I was writing, individuals could contribute up to $2,000 a year ($4,000 for a married couple) into these tax-sheltered accounts. Since contributions were tax deductible, an individual with a marginal tax rate of 30% who contributed the maximum of $2,000 would reduce his tax bill by $600.
One problem with the IRA design is that a taxpayer must have already made the contribution before filing the tax return. This design is problematic in that for many taxpayers, it is only after they have filed their tax return and settled up with the government that they have the cash to invest in an IRA. American taxpayers are more likely to be flush with money after they file their tax return, because 90% of them get a refund that averages about $3,000 per household, and it takes a while for the refund to arrive.
So my first suggestion was to allow taxpayers to use their income tax refund to make a contribution that counts on the return currently being filed (for the previous year’s income). Under my proposal, the taxpayer would only have to create an IRA account before filing their taxes, and then could just ask the IRS to send some portion of the refund to that account and repeat this in following years using the same account.
The second proposal was designed to reinforce the first. I suggested that the government adjust the formula used to determine how much money is withheld from workers’ paychecks by the Treasury Department as a prepayment of taxes. This formula could be tweaked so that taxpayers would get somewhat larger refunds at the end of the year unless they actively reduced their withholding rates, which anyone can do. The evidence suggests that when people get a windfall—and this seems to be the way people think about their tax refund, despite it being expected—they tend to save a larger proportion from it than they do from regular income, especially if the windfall is sizable. So, my thinking was that if we gave people bigger refunds, we would generate more savings, whether or not we figured out a way to make it easier to funnel those refunds into IRA saving. These two proposals would ideally have been combined.
I suspected that increasing the withholding rates would likely have another beneficial side effect: better tax compliance. My sense was that many taxpayers consider a refund as a gain and an underpayment as a loss, and when faced with a loss, they might get “creative” in filing their tax return. Recall that people tend to be risk-seeking in the domain of losses when they have a chance to break even. A recent study of 4 million tax returns in Sweden has confirmed my hunch. The authors find that taxpayers sharply increase their claimed deductions for “other expenses for earning employment income” if they would otherwise have to write a check to the government. Claims for small amounts in this category studied by the authors (less than 20,000 Swedish kroner or about $2,600) are known to be mostly bogus. When taxpayers are audited (which is rare) such claims are rejected over 90% of the time.
My third proposal involved a simple change to the way in which people sign up for defined contribution savings plans offered by their employers, such as the 401(k) plans offered in the United States. Quite basically, I asked: why not change the default? Under the usual rules, in order to sign up for the plan the employee had to fill out a bunch of forms, choose a saving rate, and decide how to invest the money. Why not make joining the plan the default and tell people that if they do not opt out, they will be enrolled in the plan at some default saving rate and in some default investment product?
Economics makes a clear prediction about this last proposal: it will have no effect. The designation of a particular option as the default is a SIF. The benefits of joining a 401(k) plan can add up to large amounts of money, tens if not hundreds of thousands of dollars, especially if, as is common, the employer matches some portion of the contributions. No Econ would let the minor inconvenience of filling out a couple forms prevent her from cashing in on so much money. To do so would be like buying a winning lottery ticket and not bothering to turn it in because it would require a five-minute stop at the convenience store. But for Humans, for whom filling out forms can be daunting and choosing investment strategies can be frightening, making enrollment in the plan the default option could have a big effect.
I later learned that I was not the first to think of changing the default option for retirement savings plans. A few firms had tried it, most notably McDonald’s, the fast food giant. But the name that was commonly used for this plan at that time was unfortunate. In the industry, it was called “negative election.” It is hard to get people very excited about a plan that is called negative election.
A few years after publishing this paper, I was asked to give a talk to the retirement plan clients of Fidelity, the American mutual fund giant. Fidelity, of course, had a pecuniary interest in this topic. Firms across the United States had quickly been switching over from the old-style pension plans, in which the employer made all the decisions, to the new defined contribution plans. In response, Fidelity and many other large financial service companies had started new lines of business to administer the plans for employers, and their mutual funds were also offered as potential investment vehicles for the employees. Increasing the account balances would be good for the employees and for Fidelity.
If I could think of something that might put more money into retirement savings accounts, I would have the representatives of several hundred large employers in the audience, and they might be willing to try it. Of course, I would advocate changing the default to automatic enrollment, but it would be good to also come up with something new.
After some brainstorming with Shlomo Benartzi, by that time a regular collaborator, the approach I took was to make a list of the most important behavioral reasons why someone might fail to save enough for retirement, and then design a program that could overcome each of these obstacles. This is an approach I now often use when trying to dream up a behavioral intervention for some problem. For my list, I came up with three factors.
The first obstacle is inertia. Surveys reveal that most people in retirement savings plans think they should be saving more, and plan to take action, uh, soon. But then they procrastinate, and never get around to changing their saving rate. In fact, most plan participants rarely make any changes to their saving options unless they change jobs and are confronted with a new set of forms they have to fill out. Overcoming inertia is the problem that automatic enrollment magically solves. The same concept should be included in a plan to increase saving rates. If we could somehow get people started on a plan to increase their saving rates and let that kick in automatically, inertia could work for us instead of against us.
The second obstacle is loss aversion. We know that people hate losing and, in particular, hate to see their paychecks go down. Based on the findings from our fairness study, we also know that in this domain, loss aversion is measured in nominal dollars, that is, without adjusting for inflation. So, if we could figure out a way that employees would not feel any cuts to their paychecks, there would be less resistance to saving more.
The third behavioral insight was related to self-control. A key finding from the research on this topic is that we have more self-control when it comes to the future than the present. Even the kids in Walter Mischel’s marshmallow experiments would have no trouble if today they were given the choice between one marshmallow at 2 p.m. tomorrow or three marshmallows at 2:15 p.m. tomorrow. Yet, we know that if we give them that same choice tomorrow at 2 p.m., few would be able to wait until 2:15. They are present-biased.
The proposal I eventually presented at the Fidelity conference was called “Save More Tomorrow.” The idea was to offer people the option of deciding now to increase their saving rates later, specifically, when they get their next raise. Then, keep them enrolled in the program until they opt out or hit some cap. By tying the increases in saving rates to pay increases, loss aversion would be averted. By asking them to make a decision that would take effect sometime in the future
, present bias would be mitigated. And by leaving the plan in place unless the person opts out, inertia would work for us. Everything I knew about behavioral economics suggested that such a plan would work. Naively, I was also confident that one of the hundreds of companies that were represented at that conference would soon be getting in touch about how to try out this great new idea. And I was happy to give it away and offer free consulting to anyone who was willing to try it, as long as they let Benartzi and me evaluate what happened.
Boy, was I wrong. Not one company got in touch. And automatic enrollment was not doing much better, even with its improved name.
One thing that was slowing down the adoption of automatic enrollment was that companies were not sure it was legal. Here a lawyer and pensions expert, Mark Iwry, intervened to help. Iwry, then a Treasury Department official in charge of national pension policy, led the Department of the Treasury and the IRS to issue to issue a series of rulings and pronouncements that defined, approved, and promoted the use of what they referred to as automatic enrollment in 401(k) and other retirement savings plans. So Mark Iwry really paved the way for firms to try out this new idea, not only giving it a better name but also giving it a legal stamp of approval. (He did this quite independently, though we later got to know each other and have worked together on other initiatives.)
Yet it remained hard to encourage take-up of the idea without proof that it actually worked. This problem was solved by a colleague at Chicago, Brigitte Madrian, who now teaches at the Kennedy School of Government at Harvard. Brigitte wandered into my office one day to show me some interesting results she had obtained that were so strong she could not quite believe them, even though she had crunched the numbers herself. A company that had tried automatic enrollment asked Brigitte if she would analyze the data. She worked with an employee of the company, Dennis Shea, to see whether automatic enrollment was effective. The results were stunning, at least to Brigitte, who had received traditional training as an economist. She knew that the default option was an SIF and therefore should not matter. But she could see that it did.†
Misbehaving: The Making of Behavioral Economics Page 33