Active vs. Passive Decisions and Crowd-Out in Retirement Savings Accounts: Evidence from Denmark Raj Chetty John N. Friedman Harvard University and NBER and Soren Leth-Petersen Torben Nielsen University of Copenhagen and SFI 14th Annual Joint Conference of the Retirement Research Consortium August 2-3, 2012 Washington, D.C. This research was supported by the Danish Council for Independent Research and by the U.S. Social Security Administration (SSA) through grant #10-M-98363-1-02 to the National Bureau of Economic Research as part of the SSA Retirement Research Consortium (RRC). The findings and conclusions expressed are solely those of the authors and do not represent the views of SSA, any agency of the Federal Government, or the NBER. We thank Tore Olsen, Brigitte Madrian, and James Poterba for helpful comments and discussion. Jessica Laird, Sarah Griffis, and Heather Sarsons provided excellent research assistance. Abstract: Most developed countries use an array of policies to increase retirement savings, including tax-deferred savings accounts, mandatory savings plans, defined benefit pensions, and defined contribution savings accounts. Firms also use defaults and match-rates to encourage saving among workers. While many studies have analyzed the impacts of such policies on savings within retirement accounts, much less is known about the impacts on total retirement savings (i.e., crowd-out). When assessing the effectiveness of these policies, the central question is whether they successfully induce people to raise the total amount saved, or instead simply encourage workers to shift money from one account to another. For instance, a firm that successfully raises pension contributions among its workers may have no affect on their total savings if the workers simply save less in their private stock portfolios. In this paper, we measure the effect of savings policies on total savings. We show that a key determinant of this broader effect is whether policies induce active or passive changes in savings behavior. Our main result is that policies that rely upon individuals to actively respond in order to raise savings -- such as tax subsidies for retirement savings -have smaller impacts on total savings than policies that raise savings passively -- such as defaults or mandatory savings requirements. A secondary result is that policies that raise savings passively are especially effective at raising the savings of individuals saving less for retirement to begin with. We study the impacts of savings policies on individuals' total savings using Danish income tax records. These data provide administrative information on the value of assets and liabilities including stocks, bonds, bank accounts, credit card debt, mortgages, and other secured debt for all Danish citizens from 1994-2008. The Danish data have two advantages over datasets used in prior work. First, they offer considerably more precise information for a larger sample of individuals. The state-of-the-art study in this literature (Gelber 2011) uses 835 observations from the Survey of Income and Program Participation. Our analysis sample includes approximately 45 million observations. Second, Denmark experienced a series of sharp, targeted reforms that provide an ideal opportunity to study the impacts of retirement savings policies on savings behavior. These features of the data yield much more precise estimates of the impacts of saving policies on total savings than prior work, allowing us to uncover how different types of policies vary in their success at increasing total retirement savings. 1 First, we analyze the impacts of savings mandates. These policies generate passive changes in retirement savings, since the policy automatically increases pension contributions without any action by the worker. A growing literature suggests that policies such as mandates or default settings can powerfully affect the amount that workers contribute to their pension plan, even though workers are free to undo the policy if they choose. But this does not imply that such policies improve total savings. For instance, higher pension contributions may cause individuals to run down their bank accounts and run up credit card debt, leaving total savings unchanged. We analyze a Danish government policy introduced in 1998 that required all Danish citizens to contribute 1% of their gross earnings to a pension account that would be managed by the government. This mandatory savings plan continued, with some modifications to its design, until 2003, when it was terminated. Because the plan contribution was proportional to income, individuals earning higher incomes were forced to contribute larger absolute amounts to the account. As in the previous literature, we find that this mandate had large effects on total pension contributions. In order to study the effect of this policy on total savings, we exploit two strategies. First, we analyze those around the income threshold for eligibility. In 1998, individuals earning less than 34,500 Danish kroner were not subject to the program. As a result, individuals who earned just above the threshold were forced to save 1% more than those just below. We then find that individuals just above the threshold save considerably more than those just below. Second, in order to more precisely estimate the impacts of savings mandates on total savings, we examine the savings decisions of individuals who switch between firms and, as a result, experience large increases in employer pension contributions. Since the workers could offset these increases in employer pensions by reducing private savings, they function much like mandates. Here too, we estimate that a large share of increases in employer-sponsored pension savings passes through to total savings. Combining these results, we conclude that approximately 90% of the direct impact of savings mandates remains in total savings. Second, we study tax subsidies for retirement savings. In contrast to mandates and defaults, tax subsidies generate active changes in retirement savings, since workers must themselves increase pension contributions in response to the incentive. Existing 2 research on pension policies suggests that only a small fraction of workers respond to such incentives, but among this group the increase in pension contributions is very large. Once again, this does not imply that total savings have increased, since workers may actively shift savings from bank accounts or stock portfolios into tax-deferred accounts in response to the incentives. In 1999, the Danish government reduced the tax subsidy for contributing to capital pension retirement savings accounts -- analogous to 401(k)'s -- by approximately 15 cents per dollar for individuals in the top income tax bracket. Individuals below the top income tax bracket were unaffected by the reform. This reform allows us to identify the impacts of changes in the tax subsidy on savings, by comparing the responses of affected individuals just above the top tax bracket to unaffected individuals just below it. As in the previous literature, we find that a small fraction of individuals responded sharply to the policy change, and as a result capital pension contributions fell sharply for individuals in the top income tax bracket but remained virtually unchanged for individuals just below that bracket. We find that the impacts of this reform on total savings are very limited. Most of this reduction in capital pension contributions is offset by increases in contributions to other types of retirement savings accounts and increases in non-retirement savings. We find that active savers follow a "pecking order" model of savings in that they substitute for capital pensions first using other retirement savings accounts and then using taxable accounts that are poorer substitutes. Because they must be attentive to the tax subsidy to response in the first place, they do not increase pension contributions in isolation but rather rebalance their savings across all accounts. On net, we estimate that only 10% of the reduction in capital pension savings due to the elimination of the tax subsidy remains in total savings. These results suggest that savings mandates and tax subsidies for savings affect two very different groups of people. Mandates affect the majority of savers who do not pay close attention to savings policies. Tax subsidies, on the other hand, affect those few savers who do pay attention. Our final results show that these two groups are not randomly scattered throughout the population but are in fact two quite different types of people. "Active" savers who respond to tax subsidies and undo savings mandates are 3 wealthier, more likely to be college educated, and contribute more to their pension accounts each year to begin with. Therefore savings mandates, which affect less educated and less wealthy savers who do not pay attention, may better target individuals who are at risk of undersaving for retirement. The main lesson of these results is that policies such as tax subsidies that rely on active changes by attentive individuals may be less effective in increasing total savings than policies like mandates that shift individuals' behavior passively. As with any empirical study of this form, our estimates apply to the specific Danish institutional context that we study. It is plausible that the specific estimates we obtain would differ in other countries, e.g. those with less generous pension systems such as the U.S. Nevertheless, these results raise important questions about the development of savings policy in the U.S. Over the last 25 years, savings policies in the United States have focused increasingly on tax-deferred accounts such as 401(k) plans and defined-contribution pensions. These tax subsidies cost the federal government $125 billion in 2010. These large subsidies were intended to increase personal savings, but since price subsidies rely on active decisions, our findings suggest instead that tax-deferred accounts may be ineffective at increasing total savings. Much of the increase in savings in tax-deferred accounts over the past 25 years, therefore, may have come directly out of other accounts rather than from reduced consumption. A number of questions remain open for future research. Our study finds sharp differences in the short-term effects of the savings policies studied. These policies may not have as starkly different effects in the long-term, however. Another fertile area for future work is characterizing how savings policies differ in their impacts on individuals' overall welfare. Although mandated savings policies may be most effective at increasing total savings, they also cause inattentive workers to misperceive how much they are saving, and how much wealth they have. Believing themselves to be poorer than they actually are, individuals alter their consumption; the effect on their overall welfare depends on how the benefits of their additional savings balance with the benefits or harms of their changed consumption decisions. Additionally, policies that forcibly increase savings are only of interest when many individuals are below the level of 4 savings that is socially optimal; future research might attempt to estimate the social value created by each additional dollar saved. Given the policy objective of raising the savings of individuals who would otherwise save too little for retirement, our work suggests that policies which raise savings passively are preferable to those which rely on active responses. The effectiveness of tax subsidies at raising savings levels is limited by three important factors. First, only individuals who are attentively optimizing their savings portfolio will actively respond to the changed incentive. Second, these attentive individuals are generally aware of the substitution possibilities between their savings accounts, and respond by shifting money across their accounts rather than raising the overall amount they save. Finally, the attentive individuals affected by the subsidy are those who are saving more for retirement to begin with -- the benefits of the subsidy are therefore missed by those who need it most. In contrast, policies which change individuals' savings behavior passively, such as mandatory savings requirements, have a much larger impact on total savings levels, and their impact is also better targeted at individuals with low initial retirement savings. We therefore find that policies which passively raise savings will likely be more effective, better targeted, and achieve the objective of raising total savings at a lower cost. 5