Teacher Pensions Blog

As a country, we spend hundreds of billions of dollars a year on tax incentives for individuals to save for retirement--including employers that contribute toward those accounts. But do those incentives actually work? Do they change people’s behaviors? Are people satisfied with the results?

A new paper from William G. Gale, Benjamin H. Harris, and Claire Haldeman surveys the landscape and finds that we don’t have good answers to these questions. They write, “policymakers do not have access to robust empirical consensus when making decisions that affect the retirement security of tens of millions of families.”

I spoke with Harris about these findings and his ideas for a research agenda to begin to answer questions about how Americans experience retirement. What follows is a lightly edited transcript of our conversation:

Aldeman: Can you give us a brief background of the paper. What were your goals in writing it?  

Harris: I think that Americans hear a lot about retirement saving, and there’s a lot of anxiety about people being poorly prepared for retirement. Part of this has to do with some of the information that’s coming out of the financial services sector, part of it is the information coming from policymakers, part of it has to do with concerns about Social Security solvency, and part of it has to do with rising health care costs. For whatever reason, there’s a lot of anxiety and misinformation about Americans’ preparation for retirement.

The point of this paper was to lay out a case for ways we can learn more about retirement and try to provide a pathway for policymakers to know more about ways to improve the retirement experience. We’ll get into the details later, but our idea was to say, “there’s a lot we don’t know about retirement, here’s how we can get to place where we know more.”

On that front, you write, “almost no retirement policymaking is rooted in evidence; programs simply continue indefinitely with little or no Congressional oversight.” That’s a striking statement. Can you say more about what you mean by that, and what the implications are?

Sure. The predominant form of subsidizing retirement saving in this country is through 401k-type plans, also known as defined contribution plans. These are plans where people get a tax exemption when they put the money in during their working years, the money grows tax-free, and then is taxed when people take out the money in retirement. There has been very little attention by policymakers to whether or not these types of plans are actually working. By “working,” I mean whether or not they’re boosting households’ savings.

Now, we know there’s money in these plans, trillions of dollars. But the key question is whether or not they’re actually inducing Americans to save more. Someone could ask, “why does this even matter?” Well, it matters because we’re foregoing hundreds of billions in tax revenue every year in order to subsidize this type of saving. So the first big question is whether these plans are working, if they’re leading to higher rates of saving by Americans.

The second general question is about how Americans spend down their retirement savings. Even thought someone might accumulate a lot of money in these plans, we don’t know if it leads to a better retirement. Does it allow them to retire when they want to retire? Does it allow them to take care of the heath costs that they need to take care of? Does it allow them to not be a burden on their children, if that’s what they want?

The basic point is we’re giving hundreds of billions of dollars in tax breaks every year. Maybe we should pause to ask, “are these really working?”

When we’re talking about retirement savings policies in this country, other than Social Security, we’re mainly talking about tax policies. But is the tax code the right lever to encourage retirement savings?

That’s a great question. With retirement, we often talk about a three-legged stool, which refers to Social Security, private savings, and employer-provided retirement accounts. The tax code is the major lever for all of these. Workers get big tax breaks for saving in their pensions or 401k-type plans. Returns on saving outside of retirement accounts are taxed at lower rates, and the tax rate is zero on gains held until death. Even when we’re talking about Social Security, there’s a tax element as well, because retirees are taxed differently on their Social Security benefits than they are on other types of income.

The tax code is the lever for much of our retirement policy. I think it’s a good question to ask whether we should think this through better, and evaluate whether it’s working as well as it could. It doesn’t have to be this way. Medicare is an example of a different way. Medicare does not have a big tax aspect to it, and it seems to be working well. So this tax issue is an open question and one we think worth considering.

Could you talk a bit about the power of defaults? There’s a lot of work showing that “nudging” people into better retirement savings habits materially changes their behavior. Should we just create more nudges, or how should we think about nudging in this context?

This is the big innovation in retirement saving, or in saving more broadly, over the past 10 or 15 years. Thirty years ago we thought it all came down to incentives. If we gave people a plus-up, some financial benefit for saving, that was really all we needed.

There are a couple problems with that. The first is that the structure for savings are “upside-down.” By that I mean that if you’re in a higher tax bracket, your incentive for saving is higher. Each dollar you put into a retirement account is subsidized more than someone in a lower tax bracket. And if you’re not paying much in income taxes, there’s also an open question of whether you should be explicitly saving for retirement at all.

The research on nudges has helped us realize that incentives can be unequal, and that incentives sometimes just don’t work. There’s a study we reference in the paper by Raj Chetty and his co-authors where they looked at Denmark {Note: see the paper here}. The reason that Denmark was a good country to study is that they collect data on every retirement saver in the country. They also had big changes to their retirement policy that made it possible to study people’s behavior. One of the interesting findings that came from the study was that it classified people as either being active savers or passive savers. They defined an active saver as someone who was paying attention to incentives and then responding to them. They found roughly 15 percent of people to be active savers. But if only 15 percent are really paying attention—and this is Denmark, so it could be different than the United States—then incentives don’t matter that much, because people don’t know about them and can’t respond to them even if they wanted to.

Automatic saving is a way to get around that. The idea is that if you automatically put people on a sound savings path, people largely stick to that. There are some problems with that, of course, which we can talk about. But in general there’s been a shift in the thinking around retirement due to this type of research. If it’s the case that incentives don’t work that well, then automatically nudging people onto the right path is a more effective strategy.

You mention the Chetty study on Denmark, and it’s clear in reading your paper that much of what we know about retirement savings come from European countries. Can you comment about why that is, and whether the studies on Europe are applicable to the United States?

A lot of time we need a change in policy to study whether a policy works. In Denmark, it was helpful because we could see how people responded to a change in incentives. In the United States, we don’t have those same natural experiments to study the impact of savings incentives.

The data here in the U.S. is another problem, and it’s becoming worse. People are getting saturated with surveys, they’re distrustful of people asking them about their personal information. I think that’s a perfectly reasonable response to developments in privacy, but the drawback is that the public surveys that we rely on are deteriorating in quality over time. It can be very difficult to find a quality dataset to study retirement security. If we had data like Denmark’s, then we would know much more about the effectiveness of incentives here. We don’t even have to go as far as Denmark, where they collected data on every single person in the country. Even if we just had a high-quality sample of people showing how their saving evolves over their life path, that would be tremendously effective in our understanding of retirement.

In the paper you lay out a couple different paths for developing a research framework to evaluate our retirement policies. What sort of systems would that entail? Or how would we get to the vision you’re putting forward?

There are three things we lay out in the paper, and they all seem pretty achievable. They’re not all that expensive, in the millions and not the billions, which is relatively small on the federal level.

The first step is that we should evaluate tax expenditures. In the context of retirement savings, tax expenditures are mainly the tax breaks we give for putting money in a 401k or, for an employer, offering employees a pension. These have never been formally evaluated, which may be surprising given that we’re spending over $200 billion a year on them. Maybe we should pause for a second and evaluate whether they’re actually working. Private companies regularly take a step back and ask whether their strategy is working, whether that’s in retail or tech or finance. We think the government should do the same thing. There are a couple agencies within the federal government that are well-suited to evaluate tax expenditures. What we should have is a formal, ongoing process whereby government evaluators periodically study whether our policies are working.

The second thing we call for is a 1 percent carve-out for program evaluation. This has long been a priority of Results for America, where I am affiliated and which was a sponsor on our paper. The idea is, for programmatic funding, we should carve out 1 percent for evaluation. That could be everything from randomized control trials (RCTs), where you assign a treatment and control groups and evaluate whether a given intervention had an effect on the treatment group. But it doesn’t have to be RCTs. The fundamental idea is to set aside $1 for every $100 in spending towards formal evaluation of programs to see whether or not they’re working.

We specifically called out the Social Security Administration. To be clear, we’re not talking about 1 percent of all the dollars spent on Social Security. We’re talking about 1 percent of all the dollars spent on administering the program, which comes down to $1 out of every $7,000 spent on the program overall, including both the disability and the old-age program. Devoting that 1 percent toward program evaluation might help us understand more about how the program is impacting retirement. 

The last thing we thought would be a good idea would be to put together a dataset that approaches what Denmark has. It would be a panel dataset that would follow households over time. It would link administrative data that the government already has on people’s participation in different government programs, as well as tax data and survey data that people report. The survey questions might include answers to questions like, “when do you want to retire?” “how do you feel about your retirement?” or “how healthy are you?” Having a dataset combining all those elements would allow us to know so much more about retirement, and that would allow policymakers to make reforms that ultimately make people better off in the long run.

It doesn’t take a lot of money to do this. If we siphon off just a little bit, 1 percent from each program, we can make the programs much better.

What are the biggest barriers to implementing this vision? Is it cost, inertia, fear of privacy, or what are the biggest obstacles?

This is a great question. If I had to rank the problems, I would say the biggest one is what do you do when you find out the program isn’t working? What do we do if we find out that 401ks aren’t as successful as we thought? It’s tough to make policy decisions in the face of finding out that a program is ineffective. People like most programs, even if they may not be working all that well. That’s a big challenge.

A second challenge comes down to funding. Even though 1 percent seems like a small number, we are budget constrained these days, and policymakers are always looking for cuts. Even a 1 percent add-on could feel like a tall order.

The last thing I’ll say is on the data. We’re fighting this sea change on survey data. If you look at some of the work by Bruce Meyer and others, who have found that the rates of response are plummeting over time {Note: see the paper here}. It used to be that we’d see response rates of 60 or 70 percent. Those have been cut in half, and there are some questions some people are just no longer comfortable answering. People are reluctant to answer questions about their personal finances, and people don’t love that the government has this information and they’re sharing it with researchers.

Everyone understands the importance of paying taxes, but there’s sometimes an expectation that that information stays with the IRS. When the IRS does share information, it’s incredibly powerful from a research perspective. Some of the best papers we’ve seen over the past 10 years, including papers written by Raj Chetty’s team, have come out because they had access to IRS data. But the IRS is so careful and cautious with the data it can be a barrier to implementing our third pillar.

Are there any other points you haven’t mentioned?

The last thing I’ll say is that in addition to a shift in our understanding of incentives versus nudges, there’s also been a big shift between the saving process and how we spend down assets in retirement. We’ve shifted from a world in which everyone got a pension, or at least most people who got a retirement benefit through the workplace had a pension, to one where we’re mostly just saving on our own in our individual accounts. The implied expectation seems to be if you saved enough you’d be okay. With expanding longevity and high out-of-pocket health care costs and other sources of uncertainty in retirement, we need to focus not just on how much you have saved in your 401k, but also on strategies for spending down those assets. Economists call this the “decumulation” process.

That means retirement is not just about understanding how people save, but also the strategies people use to spend their nest eggs that they spent their whole life accumulating. Do strategies like reverse mortgages make sense? Should people buy annuities? Should we think about a better market for long-term care insurance? Those types of questions are only going to get more and more important as Baby Boomers get deeper and deeper into retirement.