Tax policy colloquium, week 8: Tatiana Homonoff on SNAP (fka Food Stamps)

Yesterday at the colloquium, Tatiana Homonoff of NYU's Wagner School presented Program Recertification Costs: Evidence from SNAP. As background, when people qualify for benefits under the Supplemental Nutritional Assistance Program (SNAP, formerly known as Food Stamps), they must periodically document their continued eligibility, under a recertification process that includes submitting income documentation and having an interview with a case officer.

This interview can be by phone, and apparently aims more at providing counseling and aid towards completing the process, than at raising obstacles. But I would be unsurprised if SNAP participants who need to recertify, and who still meet the qualifications, were to view it with anxiety.

SNAP has a great deal of "churn." That is, people fail to recertify in time, get thrown off the program, and then succeed in being reinstated. The paper mentions estimates suggesting that this actually costs the SNAP program money, in that the administrative costs of recertification exceed the fiscal savings from the missed benefits. This of course is a lose-lose scenario, since even the fiscal saving comes at the expense of people who needlessly miss benefits that would have helped them. For example, one may have children going hungry due to the missed benefits, at the same time that the program is losing money on net.

The paper studies how randomly assigned interview dates in San Francisco's SNAP program affected rates of successful recertification. San Francisco staggers the interview dates that it assigns people, so that its case workers will have a smoother workload. But if your interview is late in the month and you don't act promptly to reschedule (such as in advance), you may run out of time and get thrown off the program - albeit just temporarily in what turn out to be the "churn" cases.

The effects of being randomly assigned an interview date late in the month turn out to be startlingly large. "We estimate that a one-day delay in the assigned interview date decreases the chance of succesfully recertifying by one-third of a percentage point. In other words, a case that has an initial interview on the 28th day of the month [the last date to which interviews are randomly assigned] is 9.0 percentage points less likely to recertify than a case that has an initial interview scheduled on the first of the month - a 19 perent decrease in recertification success."

The paper finds, moreover, that needier households that would get large benefits are affected on average more than those closer to exceeding SNAP's income thresholds and that would thus get smaller benefits. So the hazard that is effectively  being imposed here, via randomly assigned dates later in the month, worsens rather than improves SNAP's "targeting efficiency," as this typically is defined by SNAP researchers to distinguish between worse-off and better-off households, among those that are sufficiently poor to qualify for the benefits.

Benefit loss here looks less like a function of rational choice (e.g., expending less effort to preserve your benefits if they're smaller) than of reverse screening. That is, needier households are apparently less able to overcome the obstacles created by having less time to reschedule if needed, get everything done, etc.

The most obvious policy takeaway is that interviews should be scheduled earlier in the month when recertification needs to be completed, and/or extra time given to complete the documentation requirements post-interview. One step broader is the question of why recertification is required to be so frequent when, at least in this paper's data set, there appears to be such a high probability of continuing to qualify for the benefits. (And the reverse-screening aspect of recertification makes this even worse.) But broader still is the sociological / fiscal language question of how SNAP's character as a transfer or benefit program appears to influence its structure in telling, and perhaps counterproductive, ways.

A number of my tax colleagues at various law schools have written papers about integrating the tax and transfer systems. One shouldn't necessarily have any particular prior about the extent to which this is likely to be desirable as an administrative matter. But the clearer and perhaps more interesting point for me, which I have written about here, here, and here, is that "taxes" such as the income tax and "transfers" such as SNAP are part of the very same thing. We should think about them in an integrated fashion, but often we don't, and this is partly a function of the optics that result from staring intently at stand-alone transfer programs without a grasp of the broader context. It leads to program maldesign, and bad effects on poor people, even when they are not the deliberate targets of lawmakers' malice. In particular, it encourages

Should we so focus on SNAP's stand-alone "program integrity" that Type 1 errors (mistaken allowance of benefits) get treated as far more regrettable than Type 2 errors (mistaken disallowance)? I don't inherently see why, even though it is a "transfer" rather than a "tax." The direction of cash flow, between the government and a given household, under one program for one period, as considered in isolation from all the rest of the system, is of zero normative interest. Suppose, for example, that Bill Gates mistakenly got $100 in SNAP benefits. Leaving aside the difficulty of imagining Bill at the checkout counter with a SNAP card, this has no more significance than the case in which he ought to have been required to pay $100 more in income taxes.

Transfer programs are often in-kind rather than being paid in cash, although SNAP is one of the more cash-like in that people's total food expenditures typically exceed their SNAP benefits (hence the word "supplemental" in SNAP). Thus, while for behavioral reasons it might increase food outlays relative to giving out straight cash, the difference is probably not that great.

Transfer programs - and SNAP in particular - are also typically more tied to current period available resources, rather than to long-term or even lifetime, material wellbeing levels. Hence, the case for replacing an income tax with a progressive consumption tax would be difficult to apply sensibly here. ("Sorry, we're going to let you starve because you have great expectations for the future, and/or spent too much last year.")

But the broader conceptual interchangeability suggests that we should view under- and over-payment of SNAP benefits, relative to the case where eligibility criteria were correctly applied,more symmetrically than I gather we do. This would likely suggest making recertification less frequent and more easily accomplished. Even people on the high end of SNAP eligibility are presumably below the median economic level in our society. And where hazards are reverse-screening out the neediest, that is worse still. Even the concept of "program size" can involve unduly separating in our minds one intertwined piece of our overall fiscal system from all the rest.