Tax policy colloquium, week 14: Raj Chetty's "Active vs. Passive Decisions and Crowd-Out in Retirement Savings Accounts: Evidence from Denmark"

Earlier this week we completed the 18th (!) annual edition of the NYU Tax Policy Colloquium.  Thanks to my colleague for the semester, Bill Gale, for being great to work with both in general and with our constituencies, the students and the PM attendees.  I have now co-led the Colloquium with David Bradford, Alan Auerbach, Mihir Desai, Rosanne Altshuler, and Kevin Hassett, in addition to Bill.  I look forward to co-leading it again next year with Alan.

Our final speaker of the semester was Raj Chetty, presenting the above-titled paper, which has already (deservedly) received enormous attention, both within the field and from policymakers, even though it hasn't been officially published yet.  (Of course, it's not as if all that many people, at least in the field, really do read much in the way of published versions of papers, since the working paper versions may appear years earlier.)

Chetty's paper takes advantage of some stunningly complete data from Denmark, providing information about people's income, savings within retirement accounts, and overall wealth-holdings (at least of financial assets), and extending over a long period when Danish policy changes created the equivalent of large-scale natural experiments.  The idea is to test the effects on people's overall financial wealth - not just that which they hold in employers' retirement savings accounts - when the rules for the accounts change in either of two respects.  The first is either (i) requiring mandatory contributions to the accounts or (ii) changing the amount that one invests through them, out of one's paycheck, as a default (i.e., unless one specifically opts to change the contribution amount).  The second is changing the amount of any tax incentives for investing in the accounts.

A couple of preliminary notes: First, mandating contributions can be a lot more like merely defaulting them in than one might initially think.  If you are required to contribute more, thus reducing your currently available cash flow and increasing your retirement savings, then, so long as you are not in a corner, you can respond by reducing your outside saving.  In such cases, the only real difference between a mandate and a default is that reversing out the latter may be easier (as well as more precise, if the mandated saving has different characteristics than the outside saving you can reduce).

Second, when we say that retirement saving through employer accounts is tax-subsidized, we are describing it relative to an income tax framework.  From a consumption tax perspective, exempting returns to saving (or the equivalent thereto) is "normal" treatment, not a tax subsidy.  This can be important for certain purposes when we are thinking about the significance of the Chetty paper's findings - something that I may be writing about this summer, although I will not address it here and now.

A theoretical apparatus that the paper uses to make predictions about the data, which then are strongly confirmed, posits that there are two types of savers: "active" savers and "passive" savers.  The former act like neoclassical utility maximizers.  In other words, they have some sort of preference for the split between current consumption and saving, and if you change where they stand through mandates or defaults they will simply steer back towards where they wanted to be to begin with.  If you give them tax incentives for retirement saving, a key response will be simply to redirect saving that they were already doing to the tax-favored accounts.  They will only respond by saving more insofar as the incentive sufficiently affects the relative payoffs to create a slight change in the overall balance.

Passive savers, by contrast, are inattentive and easily steered.  There actually are several different explanations of what they might be doing and why, and the paper doesn't try to choose between these explanations, although for some purposes it is quite important.  They are posited not to respond to tax incentives because they aren't paying attention.  They also don't deliberately offset any changes in retirement saving from their facing mandates or a change in the default.

It is theoretically indeterminate, in the paper's model, whether mandates and defaults will actually cause them to save more.  Suppose that they start getting smaller paychecks due to larger contributions to their employer retirement accounts.  Possibility (a) is that they go on consuming just as much as before, only they run their bank balances lower, perhaps have to borrow more, etc.  So there is no change in their net saving, not because they are deliberately undoing the induced saving, but because their behavior is so entirely fixed.  Possibility (b) is that they consume less in response to having smaller paychecks.  E.g., suppose they are rule of thumb consumers who don't just go blindly on with the same consumption activities no matter what, but rather they treat their current checking account balances as telling them how much they have to spend for the month.

The Denmark data finds that the population was about 15% active savers and 85% passive savers.  The latter used possibility (b) above - that is, they did reduce consumption when mandatory or default saving reduced the size of their paychecks.  Unsurprisingly, the active savers tended to be people who are wealthier, older, save more in general, etc.

The paper finds that each dollar of increased tax incentives for saving produced only about 1 cent (!) of additional saving.  This reflected the passive savers' inattentiveness and the active savers' mainly shifting saving from outside to tax-favored accounts.

The paper also funds that each dollar of increased mandated or default saving produced about 90 cents of additional saving.  This reflected the predominance of passive savers in the population.

Lots more to say about this in due course.  But for now I'll just say that it's interesting and has important implications for both theory and policy.  (Not just this paper, of course, but also others in the same literature that reach similar findings albeit generally less far-reaching and conclusive ones because the U.S. data, say, simply isn't as rich.)