Tuesday, March 31, 2009

Tax policy colloquium on Emmanuel Saez's "Details Matter"

Last Thursday, we started our post-spring break Tax Policy Colloquium final stretch run with Emmanuel Saez's "Details Matter: The Impact of Presentation and Information on the Take-Up of Financial Incentives for Retirement Saving." The paper analyzes a large-scale real world field experiment with H&R Block in St. Louis several years back, in which randomly selected but generally lower-income customers were offered a cash incentive for establishing an IRA account with Block. Some got no incentive, a second group got a 50% match (e.g., put in $600 and Block would add $300 to the account), and a third group got a 33% credit (e.g., put in $900 and Block would send you a check for $300).

Take-up of the IRAs was generally low, even though the two incentive plans offered free money given that, despite early withdrawal penalties, one would come out ahead if one closed the account in a year. This point was not emphasized, however, and the low take-up even with incentives reinforced the difficulty of encouraging what we might think is optimal retirement saving by lower-income workers. (Either that, or else they perhaps rightly didn't like this particular savings vehicle, which had high annual fees relative to value for small accounts.)

The main finding of the study was that the match proved more popular than the credit, even though, as the above example shows, they are arithmetically equivalent IF putting down more cash now (typically out of one's refund) and waiting two weeks for a check that one then has to cash is assumed to be cost-free.

In part, people seem to have responded to the nominally larger size of a 50% match compared to a 33% credit, even though they're actually identical since the base for computing the percentages differs as between the two of them. But there's also some reason to think that people like the match structure better than the credit structure even when they're effectively identical. My proposed explanation was that the match looks like free money, while the credit looks simply like a price break, which still leaves the question of whether the price is good enough to justify a purchase. Or, hyperbolic discounting could be doing the work if people think of the match as immediately effective (though in fact it takes a couple of weeks, and one is saving the money in the account anyway) and the cash back as in the future. Neither of these is a rational explanation, of course.

I more generally wonder to what extent incentives (at least relative to applying the regular income tax treatment of saving) can generate the sort of retirement preparation that we believe is in most cases optimal. One is appealing to rational calculation in a setting where it's generally thought not to work so well. Beyond the use of defaults so people have to opt out of retirement saving, I think the better answer lies in a Social Security-style mandatory approach(government takes money from you now and gives it back with interest later, or else simply makes you wait for transfers later). This isn't perfect either (e.g., as Louis Kaplow notes, it can affect work incentives, given the preference for immediate consumption, even if one demonstrably gets fair value back). But I see it as more promising than the income-conditioned saver's credit or match approach that motivated the Saez experiment.


Pat Oglesby said...


That's a big IF. The match looks better than the credit to me, even leaving aside time value. To elaborate, by choosing the match, I wouldn't be out of pocket $300 on the belief that a check were really coming in a deal that seems too good to be true. I wouldn't have to face the remote possibility that the check would be lost in the mail, and that I would have to track it down. I wouldn't lose the check (I've done that plenty). I wouldn't have to deal with the minor hassle of depositing it in person or by mail -- everything I get is deposited directly these days. To get the match, I wouldn't have to lift a finger or even remember (you'll understand this one day) that I'm supposed to be getting a check.

I haven't read the paper, but the choices don't seem effectively identical to me.

Daniel Shaviro said...

That was indeed one of the main criticisms we offered at the session.