Optimal Sludge? The Price of Price of Program Integrity

Volume 70 November 2020
Optimal Sludge? The Price of Price of Program Integrity

Cass R. Sunstein Robert Walmsley University Professor, Harvard University. and Julien L. Gosset†††† Class of 2021, Harvard College. We are grateful to Arevik Avedian for help with the survey here and to Rohit Goyal for help with the statistical analysis. The Program on Behavioral Economics and Public Policy at Harvard Law School provided essential support.

PDF
PDF

Abstract

Public officials often impose eligibility requirements for government programs that have two effects: (1) They screen out ineligible people and (2) they screen out eligible people. Consisting of paperwork requirements, interviews, waiting periods, and administrative burdens, such requirements are sometimes characterized as “sludge,” and for some eligible people, they might prove overwhelming or prohibitive. In these circumstances, there is a pervasive normative issue: What is the optimal tradeoff between the screening out of ineligible people and the screening out of eligible people? It is plausible to think that a great deal depends on numbers. If, for example, the number of ineligible people who are screened out is very large, and if the number of eligible people who are screened out is very small, then there would seem little ground for objection. But if the number of eligible people who are screened out is very large, there is a serious problem, and it might be worthwhile to consider an approach that would not screen out eligible people, even if it would simultaneously fail to screen out, or effectively “screen in,” a small number of ineligible people. We identify competing, plausible positions on the normative question, which we label consequentialist (in the sense that they attempt an assessment of the consequences of alternatives approaches, with a focus on numbers) and legalist (in the sense that they emphasize legal constraints, designed to ensure that benefits go only to those who are actually eligible). We also offer the results of a pilot study, which shows that the overwhelming majority of respondents would favor changes that allow ineligible people to receive benefits, if that is the price of ensuring that eligible people do so as well—unless the number of ineligible recipients is very high. The survey results suggest that most people reject the legalist position and embrace a form of consequentialism. We suggest that in light of the normative analysis and the survey results, a form of consequentialism is best, unless the law explicitly forbids it.

I. A Dilemma

All over the world, governments face a pervasive dilemma with respect to eligibility requirements. A primary purpose of such requirements is to ensure that benefits go only to people who are actually eligible for them. But if the requirements are onerous, then many people who are eligible will not get them, simply because of that onerousness. [1][1]. For examples, see Cass R. Sunstein, Sludge and Ordeals, 68 Duke L.J. 1843, 1849–51 (2019). Here, then, is the dilemma: more onerous requirements will reduce the likelihood that ineligible people will get benefits, while also increasing the likelihood that eligible people will not get benefits.

Such eligibility requirements are sometimes described as “sludge,” defined as frictions, paperwork burdens, waiting time, interview requirements, or other administrative burdens that make it more difficult for people to obtain benefits. [2][2]. See id. (“The term should be taken to refer to the kind of friction, large or small, that people face when they want to go in one or another direction.”); Richard Thaler, Nudge, Not Sludge, 361 Science 431, 431 (Aug. 3, 2018), https://science.sciencemag.org/content/sci/361/6401/431.full.pdf [https://perma.cc/J3VW-JJG4] (“But what about activities that are essentially nudging for evil? This ‘sludge’ just mucks things up and makes wise decision-making and prosocial activity more difficult.”). See generally Cass R. Sunstein, Sludge (forthcoming 2021) (manuscript on file with authors). If, for example, people are required to fill out forms to obtain a visa, to get financial aid, or to obtain an occupational license, they are facing sludge. If they must go to an in-person interview before obtaining aid of one or another kind, sludge is also involved. Many federal and state programs are pervaded by sludge. [3][3]. See Sunstein, supra note 1, at 1862–64 (noting sludge in Obamacare and Medicare, as well as state regulations concerning abortion and voter registration). But how much sludge is too much? On this definition, it should be clear that the optimal level is not zero. Some sludge is good or even essential. Above all, some sludge is necessary to ensure that resources or opportunities do not go to people who are ineligible for them. Indeed, some sludge might be legally mandatory.

For many programs, an underlying goal of administrative burdens is to ensure “program integrity,” [4][4]. See Protecting Americans from Tax Hikes (PATH) Act, Pub. L. No. 114-113, tit. II, 129 Stat. 2242, 3076–86 (2015) (referring to Title II as “Program Integrity” and specifically intending to reduce fraudulent and improper payments in the EITC and other programs); Leslie Book, David Williams & Krista Holub, Insights from Behavioral Economics Can Improve Administration of the EITC, 37 Va. Tax Rev. 177, 180 (2018) (noting that “[p]rogram integrity” of the EITC was an important topic among employees of the IRS because 43 to 50 percent of all EITC returns are incorrect, with most errors benefitting claimants); Program Integrity, Ctrs. for Medicare & Medicaid Servs., https://www.medicaid.gov/medicaid/program-integrity/index.html [https://perma.cc/2ZMC-XTSH] (explaining CMS responsibilities and goals under the Medicaid Integrity Program); Reducing Improper Payments, Soc. Sec. Admin., https://www.ssa.gov/improperpayments [https://perma.cc/T8ZN-XA32] (emphasizing the objective of curbing improper payments). meant to refer to an absence of fraud, abuse, misuse, or access by the ineligible; sludge might be necessary for that purpose. A large amount of sludge—paperwork burdens, delays, information collection—could be an essential way to limit benefits, including taxpayer money, to those who have a legal right to them. Of course it is also true that sludge might be a covert (or at least not readily visible) way to limit expenditures; a reflection of a compromise between those who like a benefits program and those who do not much like it; or a way for the government to acquire information that could be used for multiple purposes. [5][5]. See Sunstein, supra note 1, at 1870–71 (pointing to sludge as a way to ration resources for those willing to tolerate administrative burdens); id. at 1872 (noting how certain administrative burdens help agencies collect useful data on public programs like employment training or infectious disease prevention efforts). But program integrity is often the central motivation. [6][6]. See id. at 1865 (noting agencies impose paperwork burdens to ensure programs comply with the law).

As a case in point, consider the Earned Income Tax Credit (“EITC”), [7][7]. The details and requirements of the EITC are set out in 26 U.S.C. § 32 (2018). a wage subsidy for low-income workers, provided by the Internal Revenue Service (“IRS”). [8][8]. See Pamela Herd & Donald P. Moynihan, Administrative Burden: Policymaking by Other Means 191, 193 (2019). For a general discussion of the EITC and its effects on recipients, see generally Making Work Pay: The Earned Income Tax Credit and Its Impact on America’s Families (Bruce D. Meyer & Douglas Holtz-Eakin eds., 2001). For the EITC, the take-up rate is about 80 percent. [9][9]. Herd & Moynihan, supra note 8, at 191. By the standards of most benefits programs, that number is not especially low. But if 20 percent of eligible people are not receiving a potentially life-changing benefit, and if the reason is the degree of sludge, there is a serious problem. An approach that denies such a benefit to millions of people surely should be rethought. The existing amount of sludge would seem excessive, and it should be reduced.

The IRS almost certainly knows enough to enroll people automatically and send a refund to eligible taxpayers. If so, sludge could be eliminated entirely, and the take-up rate should be close to 100 percent. Is that what the IRS should be doing? [10][10]. Surprisingly, there appears to be a lack of substantial literature on automatic enrollment and the EITC. But see Vada Waters Lindsey, Encouraging Savings Under the Earned Income Tax Credit: A Nudge in the Right Direction, 44 U. Mich. J.L. Reform 83, 112–21 (2010) (arguing that the EITC should have a retirement savings component with automatic enrollment). This area deserves sustained study. Suppose that the IRS indeed decided to send the EITC automatically to apparently eligible taxpayers. If it could do so at low cost, and if the apparently eligible taxpayers are in fact eligible, there would seem to be little ground for objection. The problem, of course, is the phrase “apparently eligible taxpayers.” It is possible that some of the recipients, under automatic enrollment, will not in fact be eligible. Whenever people are automatically enrolled in a program, some of them may not meet the legal criteria.

Or consider the question of voting rights. With respect to both registration and voting, some states impose a degree of sludge, on the articulated ground that it is necessary to ensure that ineligible people do not end up voting. [11][11]. See Herd & Moynihan, supra note 8, at 47–79 (discussing the history of administrative burdens in the exercise of voting rights). But what if the result is also, and simultaneously, to discourage eligible people from voting? [12][12]. See, e.g., id. at 56–57 (noting a survey among registered Texas voters found 6 percent of eligible voters said not having proper identification was “the primary reason they did not vote,” despite the interviewers finding that the “vast majority” did in fact have proper identification). What if sludge operates as a mechanism for voter suppression? That problem is pervasive. [13][13]. Id. at 49; see also Vote Suppression, Brennan Ctr. for Just., https://www.brennancenter.org/issues/ensure-every-american-can-vote/vote-suppression [https://perma.cc/9LBC-W3QD]. For benefits programs of multiple kinds, sludge might be justified as a way of ensuring that ineligible people do not gain access to goods or opportunities; but the consequence might be to screen out the eligible as well. If people have to navigate sludge to exercise their constitutional rights, there seems to be a serious problem.

The remainder of this Essay is organized as follows. Part II explores the question of optimal sludge as a matter of principle. It shows that if we care about consequences, we might favor sludge reduction even if the consequence is to give benefits or opportunities to people who are ineligible. But if we want to insist on strict adherence to legal requirements, the grant of benefits to ineligible individuals is far worse than the denial of benefits, through sludge, to eligible individuals. Some statutes may, in fact, require sludge even if the overall consequences are plausibly described as very bad. Part III describes our survey, which found that most people embrace a form of consequentialism and firmly reject legalism. Part III also discusses differences along the lines of party affiliation and demographics; surprisingly, the survey found only modest differences along these lines. Part IV, a brief conclusion, argues in favor of sludge reduction on consequentialist grounds, even if the result is to give some benefits or opportunities to ineligible people.

II. Normative Issues

Public officials must often choose between (1) a design ensuring that a certain number of eligible people will not receive a benefit and (2) a design ensuring that a certain number of ineligible people will receive a benefit. That may not be an easy choice, and it may not even be easy to decide on the criteria that officials should use to make it. On one view, we should focus on consequences, which means that we should try to specify relevant numbers: How many eligible people would obtain benefits to which they are entitled? How many ineligible people would obtain benefits to which they are not entitled? On another view, we should focus on the underlying statute and the class of people that it seeks to benefit, certainly if it mandates a particular conclusion, and perhaps even if it does not.

A. Consequences and Consequentialism

If the idea of program integrity is meant to refer to the number of errors, the choice between view (1) and view (2) above might turn on some kind of arithmetic. Which group is larger? If automatic enrollment means that 500,000 eligible people will receive a benefit who otherwise would not, but 499,999 ineligible people will also receive a benefit who otherwise would not, automatic enrollment is justified. Call this a consequentialist position; it is consequentialist in the sense that it assesses the optimal level of sludge by focusing on its consequences.

To be sure, the consequentialist position might be specified in different ways, and a simple count might be too crude. It could be challenged from multiple points of view. Suppose that automatic enrollment gives benefits to 200,000 eligible people but also to 200,100 ineligible people. Some people might think that if the 200,100 people are nearly eligible—if they are relatively poor—it is not so terrible if they receive some economic help. But other people might insist that taxpayer money is accompanied by clear restrictions and argue that if it is given out in violation of those restrictions, a grievous wrong has been committed. They might think that if automatic enrollment gives benefits to 200,000 eligible people but also to 125,000 ineligible people, it is a terrible idea. On consequentialist grounds, reasonable people could differ about the appropriate tradeoff; they may or may not require a significant disparity between the size of the group of benefited eligible persons and the size of the group of benefited ineligible persons.

We might agree that if sludge reduction or automatic enrollment ensures that a very large number of eligible people receive benefits who otherwise would not, and that a very small number of ineligible people also receive benefits who otherwise would not, all is well. Perhaps we could even agree that if sludge reduction or automatic enrollment ensures that a large number of eligible people receive benefits who otherwise would not, and that a significant number of ineligible people receive benefits who otherwise would not, all is still well. For example, in the context of COVID-19, the U.S. government eliminated a great deal of sludge to promote delivery of health and economic benefits. [14][14]. Specifically, the stimulus checks set out under the Coronavirus Aid, Relief, and Economic Security (“CARES”) Act , Pub. L. No. 116-136, § 2201, 134 Stat. 281, 335 (2020), provide a good gauge of public sentiment regarding sludge in emergency scenarios. There were certainly concerned voices regarding the distribution of stimulus checks to deceased individuals as well as the living, but the speed at which the Treasury Department sent out the checks may have justified this oversight. See Erica Werner, Treasury Sent More than 1 Million Coronavirus Stimulus Payments to Dead People, Congressional Watchdog Finds, Wash. Post (June 25, 2020, 8:09 PM), https://www.washingtonpost.com/us-policy/2020/06/25/irs-stimulus-checks-dead-people-gao [https://perma.cc/529D-5XXQ]; Eric Levitz, Treasury Sent $1.4 Billion in Stimulus Checks to Dead People. That’s Good, N.Y. Mag: Intelligencer (June 25, 2020), https://nymag.com/intelligencer/2020/06/irs-coronavirus-stimulus-checks-dead-people-gao.html [https://perma.cc/4AX7-TD3Z]. In fact, greater frustration was likely caused by the fact that while the dead received $1.4 billion in stimulus, tens of millions of living Americans did not see their checks at all. See Savannah Levins, Still No Stimulus Check? You’re Not Alone, WCNC (June 23, 2020, 11:59 AM), https://www.wcnc.com/article/money/still-no-stimulus-check-youre-not-alone/275-11793c79-7fb7-4201-9f42-343b6d602397 [https://perma.cc/CWJ8-J3JZ] (noting that 35 million Americans still had not received their stimulus checks as of June 2020). Even if those steps ultimately rewarded many people who were not in fact eligible, including people who were no longer alive, [15][15]. See Werner, supra note 14 (describing how the Treasury Department and IRS mistakenly disbursed stimulus checks to almost 1.1 million dead people in the rush to quickly deliver payments). perhaps they were nonetheless justified. But reasonable people can disagree about these questions. [16][16]. Compare John Kennedy, Sending Coronavirus Payments to Ghosts Hurts the Living, Hill (May 7, 2020, 2:00 PM), https://thehill.com/blogs/congress-blog/politics/496615-sending-coronavirus-payments-to-ghosts-hurts-the-living [https://perma.cc/6G7C-F5QS] (“Even under normal circumstances, we should not tolerate this level of waste and incompetence, but COVID-19 has landed us far from normal times. . . . Even the most fiscally unfettered politician can’t justify cutting checks to people who are not even alive.”), with Levitz, supra note 14 (“The program had an error rate of roughly 0.4 percent. That is not a scandal. It is a triumph.”), and Cindy Carcamo, Why Are the Dead Getting Coronavirus Stimulus Checks?, L.A. Times (Apr. 24, 2020, 5:00 AM), https://www.latimes.com/california/story/2020-04-24/coronavirus-stimulus-check-death-payment [https://perma.cc/3RPB-UF3D] (quoting a former head of an IRS watchdog group noting that “[t]he IRS is following the law,” as well as a vice president of conservative think tank that “[e]xpediency has been prioritized,” rather than delaying checks for weeks and months by cross-referencing data to eliminate dead recipients).

As usual, consequentialists need to specify what matters, so that we can know what counts as a good or bad consequence, and how good or bad it is. We should be able to agree that it is good, and often very good, to ensure that eligible people receive benefits to which they are entitled (especially if they are poor or sick). But how bad is it, exactly, if ineligible people receive such benefits as well?

That question remains to be answered; consequentialism, in the abstract, cannot do so. In terms of drawing the right conclusion, we will also want to specify the substantive area that is involved. More or less sludge might be imposed and tolerated in such diverse domains as antipoverty programs, voting, abortion, student visas, and occupational licenses. [17][17]. See Sunstein, supra note 1, at 1850–52; see generally Herd & Moynihan, supra note 8 (discussing administrative burdens in a wide variety of policy arenas). In one area, reasonable consequentialists might be willing to accept a grant of a good or opportunity to a large number of ineligibles, without also committing themselves to the same level of toleration in another area.

B. Legalism

On an alternative view, even a modest breach of program integrity, for the advantage of those who are not eligible, is unacceptable. Call this the legalist position. It is legalist in the sense that it sees a violation of legal requirements as a serious breach, one that cannot be justified on consequentialist grounds. An extreme version of the legalist position would be that a grant of benefits to a very large number of eligible people would not outweigh the grant of benefits to a very small number of ineligible people.

We need to make a distinction here. Suppose that the law is conclusive on the question. Suppose that properly read, it mandates a degree of sludge. Or suppose that it forbids officials from taking steps to provide benefits to small numbers of ineligible people, even if those steps are necessary to ensure that benefits go to large numbers of eligible people. If so, the law is conclusive. Officials have no room to maneuver; they are strictly bound. Call this mandatory legalism. If a statute prohibits approaches that would allow ineligible people to receive benefits, the legal question is at an end, and consequentialism is off-limits.

But suppose, alternatively, that the law gives officials discretion to adopt various approaches, such that they can choose an approach that inadvertently but inevitably grants benefits to ineligibles without violating any statutory provision, if that is the price of giving benefits to more eligible individuals. For example, a statute involving voter registration might authorize states to adopt registration procedures that do not entirely screen out people who are ineligible. Or a statute providing free school meals for poor children might come with a general appropriation and give authority to state agencies to ascertain how best to ensure receipt by those who are eligible—authority that might include a license to make some mistakes. [18][18]. For example, the Food, Conservation, and Energy Act of 2008 provides a general appropriation to states for the administration of the Supplemental Nutrition Assistance Program (“SNAP”), but states are allowed to determine the best means of certifying which children are eligible for free school meals. See Food, Conservation, and Energy Act of 2008, Pub. L. No. 110-234, §§ 4111–22, 122 Stat. 923, 1102–13; see generally U.S. Dep’t of Agric., Direct Certification in the National School Lunch Program: : State Implementation Progress Report to Congress—School Year 2015–2016 and 2016–2017 (2018), https://www.fns.usda.gov/sites/default/files/resource-files/NSLPDirectCertification2016.pdf [https://perma.cc/QBV5-24TF] (reporting to Congress “on State progress in direct certification in the National School Lunch Program”). But some people might say that officials should not take this approach, even if they have the legal authority to do so, because it is a grievous wrong to use one’s discretion to give benefits to ineligibles. Call this discretionary legalism.

In principle, mandatory legalism is right, if the statute really is conclusive. But in principle, extreme versions of discretionary legalism—where, it will be recalled, the statute is not conclusive—would seem hard to defend: a grant of benefits to ten people who are almost (but not) eligible would appear to be a price worth paying in exchange for a grant of benefits to two million people who are in fact eligible. But granting benefits in violation of the law is no light thing, even if officials do have some discretion under the statute, and the correct tradeoff is not self-evident. Public officials encounter the problem frequently, and so do companies, universities, nonprofits, homeless shelters, hospitals, and others. It is a general truth that as the burdens of eligibility requirements increase, the class of people who receive a benefit may well include fewer ineligibles; and as those burdens loosen, there is an ever-growing risk that a benefit will go to people who should not be getting it.

III. A Pilot Survey

We could imagine theoretical debates about these questions, and they would undoubtedly allow us to make progress. But let us now turn to an empirical question: What do people actually think about these issues? In a democracy, public officials are attentive to what members of the public think, and if most people turn out to favor an identifiable form of consequentialism, or instead embrace a form of legalism, their judgment would certainly be of interest. And for those who believe in “the wisdom of crowds,” a widespread public judgment might also have epistemic value. It might tell us something about what is right.

We do not mean to say that the results of surveys should be taken to be decisive. Suppose that under the law, agencies have no statutory authority to give out benefits to people who are not eligible, even if a sludge-reduction strategy, producing that result, is the only way to ensure that benefits are received by a very large number of eligible people. Even if people approve of that result on consequentialist grounds, it is by hypothesis unlawful. Or suppose that some form of discretionary legalism is correct; those who embrace it would be unlikely to change their stance upon learning that, in surveys, most people disagree with them. Or suppose that some form of consequentialism is consistent with law and also can be shown to benefit a very large number of people. Should we reject it because most people do so in surveys? Committed consequentialists would not be willing to do so. They might insist that survey answers do not show informed or reflective thinking, and that their own approach is preferable in principle.

For these reasons, we do not suggest or believe that, for democratic reasons, the views elicited in surveys should be taken as decisive. But they are of unquestionable interest. At a minimum, public officials, when deciding what to do, will be attuned to whether the public embraces or rejects one or another position. Surveys are informative about that question.

With that point in mind, we conducted a pilot survey in which we attempted to elicit people’s views on the underlying questions in one substantive domain. We emphasize the word “pilot”; this was meant as a preliminary effort to obtain a sense of what people think. Using Amazon’s Mechanical Turk, we asked 407 people to consider four different scenarios, specifying different numbers of ineligible people who would receive EITC benefits under automatic enrollment. The survey was not nationally representative, but it did have a high degree of diversity. The participants were 60 percent male, 40 percent female; 30 percent Republican, 43 percent Democratic, 27 percent independent (that is, unassociated voters); 68 percent White, 14 percent Asian, 7 percent Black, 4 percent Hispanic. We chose a within-subjects design, in which all subjects saw all four scenarios, with the goal of promoting comparative assessments.

We described four scenarios, reprinted below, and then asked the survey participants a simple question: whether they would approve of what we described as a “new approach” to the distribution of the EITC, which would have varying consequences. (Participants were asked a binary question: approve or disapprove.) The actual text read as follows:

(1) Under the new approach, the IRS would give the Earned Income Tax Credit only to the 4 million people who are actually eligible; no one would receive it who is not eligible.
(2) The IRS would inevitably make mistakes, and it would send out some money to people who are not, in fact, eligible for the Earned Income Tax Credit. Under the new approach, 4 million people would receive the credit who would otherwise not get it. Of those, 3.98 million are eligible, but 20,000 are not.
(3) The IRS would inevitably make mistakes, and it would send out some money to people who are not, in fact, eligible for the Earned Income Tax Credit. Under the new approach, 4 million people would receive the credit who would otherwise not get it. Of those, 3.9 million are eligible, but 100,000 are not.
(4) The IRS would inevitably make mistakes, and it would send out some money to people who are not, in fact, eligible for the Earned Income Tax Credit. Under the new approach, 4 million people would receive the credit who would otherwise not get it. Of those, 3.5 million are eligible, but 500,000 are not.

The aggregate results were straightforward. Very strong majorities approved of automatic enrollment in scenarios (1) and (2); a solid majority approved of automatic enrollment in scenario (3); and a solid majority rejected it in scenario (4). In short, most people rejected a legalist approach to these questions, in favor of a form of consequentialism. Even if 100,000 people would receive benefits to which they were not entitled, people would favor automatic enrollment. And when as many as 500,000 ineligible people would receive benefits, a substantial minority would nonetheless approve. It is noteworthy, however, that with those numbers, the majority disapproved, demonstrating that in scenario (4), many consequentialists would draw the line.

More specifically: For scenario (1), the approval rate was 89 percent, with Democrats at 91 percent, independents at 90 percent, and Republicans at 85 percent. For scenario (2), the approval rate was 75 percent, with Democrats at 84 percent, independents at 74 percent, and Republicans at 62 percent. For scenario (3), the approval rate was 61 percent, with Democrats at 74 percent, independents at 54 percent, and Republicans at 50 percent. Lastly, for scenario (4), the approval rate was 44 percent, with Democrats at 49 percent, independents at 37 percent, and Republicans at 42 percent.

We expected to find significant differences across various lines, but in testing for statistical significance, we were surprised to find few differences along almost every line that we tested. [19][19]. For further details, see infra app. In particular, we anticipated much higher approval ratings from Democrats rather than from Republicans, but even there, the picture is complicated, for the differences were less than stark. To be sure, there were significant differences between Republicans and Democrats in scenarios (2) and (3). Note, however, that the differences between the two parties is not statistically significant for scenarios (1) and (4), which makes it hazardous to say that we have found a consistent difference, on these issues, along party lines. In these circumstances, the relatively high level of agreement among Democrats, independents, and Republicans is the main story.

Analysis of a range of demographic differences also failed to find significant differences, perhaps because of the relatively small sample sizes. Surprisingly, we were unable to find differences between men and women or along lines of race or educational attainment. The few significant differences that we did find tell no clear or obvious tale, and we would not make much of them. For example, we found a statistically significant result for those in the $120,000–$149,999 salary range in scenarios (3) and (4): those in this income category (the second highest) showed lower approval ratings than did people in other income categories. With respect to age, we found that 50–64-year-olds showed a significantly lower approval rate in scenario (3), and those 65 and over showed a significantly lower approval rate in scenario (1).

Because these findings seem a bit random, and do not show differences across all four scenarios, it would be hazardous to draw general lessons from them. The important story is shown by the aggregate data, and by the significant, but relatively modest, differences between Republicans and Democrats in scenarios (2) and (3).

Conclusion

There has been growing attention to administrative burdens and sludge, with a mounting awareness that many eligible people are deterred or prevented from obtaining benefits or opportunities to which they are legally entitled. The strongest justification for such burdens is that they are necessary to screen out ineligible people. Whether this is an adequate justification depends on whether we embrace consequentialism, mandatory legalism, or discretionary legalism, and how we specify each of these approaches.

In principle, and if the law permits it, we would argue in favor of a form of consequentialism. Suppose that a sludge-reduction approach ensures that a very large number of eligible people receive an economic benefit who otherwise would not, while also ensuring that a very small number of ineligible people also end up with that benefit. The details matter, but in general, we would favor that approach. We would also be willing to embrace consequentialism even if the numbers were not so overwhelmingly supportive of sludge reduction. But we do not mean here to take a specific stand on what consequentialism entails, or about the appropriate conclusions when particular numbers are presented. The substantive area undoubtedly matters. Reasonable consequentialists might offer different conclusions in the domains of antipoverty programs, voting, abortion, student visas, and occupational licenses. In one area, they might tolerate a grant of a good or opportunity to ineligibles, without also committing themselves to the same level of toleration in another area.

In the particular domain of the EITC, a pilot survey suggests that most people embrace consequentialism, and that they are willing to accept the grant of benefits to numerous ineligible people if that is the price of granting benefits to far more numerous eligible people. It appears that people will accept automatic enrollment unless it means that benefits are also granted to an extraordinarily high number of ineligible people.

A great deal of research remains to be done. In the future, it would be valuable to probe the normative and legal issues in more detail. Recall mandatory legalism and suppose that a form of consequentialism would result in a violation of relevant statutes—perhaps because it would unlawfully grant economic benefits to ineligible people. If so, that approach should not be adopted. But how many and which statutes are best understood in this way? How many and which statutes should alternatively be understood to give agencies discretion to take sludge-reducing steps that admittedly give out benefits to people who are not eligible? It would be most valuable to obtain answers to these questions.

It would also be valuable to obtain a more comprehensive understanding of what people think in different areas, and to explore whether there are differences along political or demographic lines. Our own study is a mere pilot in this domain, with a relatively small sample. We could do far more to probe what people actually think. Even now, however, we believe that a heavy burden of justification lies on those who would reject effective sludge-reduction efforts, including automatic enrollment, even if the unintended but inevitable consequence would be to benefit some or even many people who are not in fact eligible. So long as sludge-reduction efforts are consistent with law and basic principle, the survey results here would suggest that those efforts deserve support, at least if they are necessary to ensure that large numbers of eligible people receive potentially life-changing or life-saving benefits that they would not otherwise receive.

Appendix


Table 1.
This table reports the results of the survey discussed in the text. More specifically, it reports the approval ratings for each demographic group with respect to the four scenarios. Recall that in these scenarios, the IRS gives the EITC to 4 million people:

(1) All of whom are actually eligible for the EITC;

(2) 3.98 million of whom are eligible, but 20,000 are not eligible for the EITC;

(3) 3.9 million of whom are eligible, but 100,000 are not eligible for the EITC; or

(4) 3.5 million of whom are eligible, but 500,000 are not eligible for the EITC.


Figure 2. Significance testing

Exponentiated coefficients; p-values in parentheses



Copyright © 2020 Cass R. Sunstein and Julien L. Gosset.

Robert Walmsley University Professor, Harvard University.
††Class of 2021, Harvard College. We are grateful to Arevik Avedian for help with the survey here and to Rohit Goyal for help with the statistical analysis. The Program on Behavioral Economics and Public Policy at Harvard Law School provided essential support.

[1] For examples, see Cass R. Sunstein, Sludge and Ordeals, 68 Duke L.J. 1843, 1849–51 (2019).

[2] See id. (“The term should be taken to refer to the kind of friction, large or small, that people face when they want to go in one or another direction.”); Richard Thaler, Nudge, Not Sludge, 361 Science 431, 431 (Aug. 3, 2018), https://science.sciencemag.org/content/sci/361/6401/431.full.pdf [https://perma.cc/J3VW-JJG4] (“But what about activities that are essentially nudging for evil? This ‘sludge’ just mucks things up and makes wise decision-making and prosocial activity more difficult.”). See generally Cass R. Sunstein, Sludge (forthcoming 2021) (manuscript on file with authors).

[3] See Sunstein, supra note 1, at 1862–64 (noting sludge in Obamacare and Medicare, as well as state regulations concerning abortion and voter registration).

[4] See Protecting Americans from Tax Hikes (PATH) Act, Pub. L. No. 114-113, tit. II, 129 Stat. 2242, 3076–86 (2015) (referring to Title II as “Program Integrity” and specifically intending to reduce fraudulent and improper payments in the EITC and other programs); Leslie Book, David Williams & Krista Holub, Insights from Behavioral Economics Can Improve Administration of the EITC, 37 Va. Tax Rev. 177, 180 (2018) (noting that “

rogram integrity” of the EITC was an important topic among employees of the IRS because 43 to 50 percent of all EITC returns are incorrect, with most errors benefitting claimants); Program Integrity, Ctrs. for Medicare & Medicaid Servs., https://www.medicaid.gov/medicaid/program-integrity/index.html [https://perma.cc/2ZMC-XTSH] (explaining CMS responsibilities and goals under the Medicaid Integrity Program); Reducing Improper Payments, Soc. Sec. Admin., https://www.ssa.gov/improperpayments [https://perma.cc/T8ZN-XA32] (emphasizing the objective of curbing improper payments).

[5] See Sunstein, supra note 1, at 1870–71 (pointing to sludge as a way to ration resources for those willing to tolerate administrative burdens); id. at 1872 (noting how certain administrative burdens help agencies collect useful data on public programs like employment training or infectious disease prevention efforts).

[6] See id. at 1865 (noting agencies impose paperwork burdens to ensure programs comply with the law).

[7] The details and requirements of the EITC are set out in 26 U.S.C. § 32 (2018).

[8] See Pamela Herd & Donald P. Moynihan, Administrative Burden: Policymaking by Other Means 191, 193 (2019). For a general discussion of the EITC and its effects on recipients, see generally Making Work Pay: The Earned Income Tax Credit and Its Impact on America’s Families (Bruce D. Meyer & Douglas Holtz-Eakin eds., 2001).

[9] Herd & Moynihan, supra note 8, at 191.

[10] Surprisingly, there appears to be a lack of substantial literature on automatic enrollment and the EITC. But see Vada Waters Lindsey, Encouraging Savings Under the Earned Income Tax Credit: A Nudge in the Right Direction, 44 U. Mich. J.L. Reform 83, 112–21 (2010) (arguing that the EITC should have a retirement savings component with automatic enrollment). This area deserves sustained study.

[11] See Herd & Moynihan, supra note 8, at 47–79 (discussing the history of administrative burdens in the exercise of voting rights).

[12] See, e.g., id. at 56–57 (noting a survey among registered Texas voters found 6 percent of eligible voters said not having proper identification was “the primary reason they did not vote,” despite the interviewers finding that the “vast majority” did in fact have proper identification).

[13] Id. at 49; see also Vote Suppression, Brennan Ctr. for Just., https://www.brennancenter.org/issues/ensure-every-american-can-vote/vote-suppression [https://perma.cc/9LBC-W3QD].

[14] Specifically, the stimulus checks set out under the Coronavirus Aid, Relief, and Economic Security (“CARES”) Act , Pub. L. No. 116-136, § 2201, 134 Stat. 281, 335 (2020), provide a good gauge of public sentiment regarding sludge in emergency scenarios. There were certainly concerned voices regarding the distribution of stimulus checks to deceased individuals as well as the living, but the speed at which the Treasury Department sent out the checks may have justified this oversight. See Erica Werner, Treasury Sent More than 1 Million Coronavirus Stimulus Payments to Dead People, Congressional Watchdog Finds, Wash. Post (June 25, 2020, 8:09 PM), https://www.washingtonpost.com/us-policy/2020/06/25/irs-stimulus-checks-dead-people-gao [https://perma.cc/529D-5XXQ]; Eric Levitz, Treasury Sent $1.4 Billion in Stimulus Checks to Dead People. That’s Good, N.Y. Mag: Intelligencer (June 25, 2020), https://nymag.com/intelligencer/2020/06/irs-coronavirus-stimulus-checks-dead-people-gao.html [https://perma.cc/4AX7-TD3Z]. In fact, greater frustration was likely caused by the fact that while the dead received $1.4 billion in stimulus, tens of millions of living Americans did not see their checks at all. See Savannah Levins, Still No Stimulus Check? You’re Not Alone, WCNC (June 23, 2020, 11:59 AM), https://www.wcnc.com/article/money/still-no-stimulus-check-youre-not-alone/275-11793c79-7fb7-4201-9f42-343b6d602397 [https://perma.cc/CWJ8-J3JZ] (noting that 35 million Americans still had not received their stimulus checks as of June 2020).

[15] See Werner, supra note 14 (describing how the Treasury Department and IRS mistakenly disbursed stimulus checks to almost 1.1 million dead people in the rush to quickly deliver payments).

[16] Compare John Kennedy, Sending Coronavirus Payments to Ghosts Hurts the Living, Hill (May 7, 2020, 2:00 PM), https://thehill.com/blogs/congress-blog/politics/496615-sending-coronavirus-payments-to-ghosts-hurts-the-living [https://perma.cc/6G7C-F5QS] (“Even under normal circumstances, we should not tolerate this level of waste and incompetence, but COVID-19 has landed us far from normal times. . . . Even the most fiscally unfettered politician can’t justify cutting checks to people who are not even alive.”), with Levitz, supra note 14 (“The program had an error rate of roughly 0.4 percent. That is not a scandal. It is a triumph.”), and Cindy Carcamo, Why Are the Dead Getting Coronavirus Stimulus Checks?, L.A. Times (Apr. 24, 2020, 5:00 AM), https://www.latimes.com/california/story/2020-04-24/coronavirus-stimulus-check-death-payment [https://perma.cc/3RPB-UF3D] (quoting a former head of an IRS watchdog group noting that “[t]he IRS is following the law,” as well as a vice president of conservative think tank that “[e]xpediency has been prioritized,” rather than delaying checks for weeks and months by cross-referencing data to eliminate dead recipients).

[17] See Sunstein, supra note 1, at 1850–52; see generally Herd & Moynihan, supra note 8 (discussing administrative burdens in a wide variety of policy arenas).

[18] For example, the Food, Conservation, and Energy Act of 2008 provides a general appropriation to states for the administration of the Supplemental Nutrition Assistance Program (“SNAP”), but states are allowed to determine the best means of certifying which children are eligible for free school meals. See Food, Conservation, and Energy Act of 2008, Pub. L. No. 110-234, §§ 4111–22, 122 Stat. 923, 1102–13; see generally U.S. Dep’t of Agric., Direct Certification in the National School Lunch Program: : State Implementation Progress Report to Congress—School Year 2015–2016 and 2016–2017 (2018), https://www.fns.usda.gov/sites/default/files/resource-files/NSLPDirectCertification2016.pdf [https://perma.cc/QBV5-24TF] (reporting to Congress “on State progress in direct certification in the National School Lunch Program”).

[19] For further details, see infra app.

Optimal Sludge? The Price of Price of Program Integrity

Volume 70 November 2020
Optimal Sludge? The Price of Price of Program Integrity

Cass R. Sunstein Robert Walmsley University Professor, Harvard University. and Julien L. Gosset†††† Class of 2021, Harvard College. We are grateful to Arevik Avedian for help with the survey here and to Rohit Goyal for help with the statistical analysis. The Program on Behavioral Economics and Public Policy at Harvard Law School provided essential support.

PDF
PDF

Abstract

Public officials often impose eligibility requirements for government programs that have two effects: (1) They screen out ineligible people and (2) they screen out eligible people. Consisting of paperwork requirements, interviews, waiting periods, and administrative burdens, such requirements are sometimes characterized as “sludge,” and for some eligible people, they might prove overwhelming or prohibitive. In these circumstances, there is a pervasive normative issue: What is the optimal tradeoff between the screening out of ineligible people and the screening out of eligible people? It is plausible to think that a great deal depends on numbers. If, for example, the number of ineligible people who are screened out is very large, and if the number of eligible people who are screened out is very small, then there would seem little ground for objection. But if the number of eligible people who are screened out is very large, there is a serious problem, and it might be worthwhile to consider an approach that would not screen out eligible people, even if it would simultaneously fail to screen out, or effectively “screen in,” a small number of ineligible people. We identify competing, plausible positions on the normative question, which we label consequentialist (in the sense that they attempt an assessment of the consequences of alternatives approaches, with a focus on numbers) and legalist (in the sense that they emphasize legal constraints, designed to ensure that benefits go only to those who are actually eligible). We also offer the results of a pilot study, which shows that the overwhelming majority of respondents would favor changes that allow ineligible people to receive benefits, if that is the price of ensuring that eligible people do so as well—unless the number of ineligible recipients is very high. The survey results suggest that most people reject the legalist position and embrace a form of consequentialism. We suggest that in light of the normative analysis and the survey results, a form of consequentialism is best, unless the law explicitly forbids it.

I. A Dilemma

All over the world, governments face a pervasive dilemma with respect to eligibility requirements. A primary purpose of such requirements is to ensure that benefits go only to people who are actually eligible for them. But if the requirements are onerous, then many people who are eligible will not get them, simply because of that onerousness. [1][1]. For examples, see Cass R. Sunstein, Sludge and Ordeals, 68 Duke L.J. 1843, 1849–51 (2019). Here, then, is the dilemma: more onerous requirements will reduce the likelihood that ineligible people will get benefits, while also increasing the likelihood that eligible people will not get benefits.

Such eligibility requirements are sometimes described as “sludge,” defined as frictions, paperwork burdens, waiting time, interview requirements, or other administrative burdens that make it more difficult for people to obtain benefits. [2][2]. See id. (“The term should be taken to refer to the kind of friction, large or small, that people face when they want to go in one or another direction.”); Richard Thaler, Nudge, Not Sludge, 361 Science 431, 431 (Aug. 3, 2018), https://science.sciencemag.org/content/sci/361/6401/431.full.pdf [https://perma.cc/J3VW-JJG4] (“But what about activities that are essentially nudging for evil? This ‘sludge’ just mucks things up and makes wise decision-making and prosocial activity more difficult.”). See generally Cass R. Sunstein, Sludge (forthcoming 2021) (manuscript on file with authors). If, for example, people are required to fill out forms to obtain a visa, to get financial aid, or to obtain an occupational license, they are facing sludge. If they must go to an in-person interview before obtaining aid of one or another kind, sludge is also involved. Many federal and state programs are pervaded by sludge. [3][3]. See Sunstein, supra note 1, at 1862–64 (noting sludge in Obamacare and Medicare, as well as state regulations concerning abortion and voter registration). But how much sludge is too much? On this definition, it should be clear that the optimal level is not zero. Some sludge is good or even essential. Above all, some sludge is necessary to ensure that resources or opportunities do not go to people who are ineligible for them. Indeed, some sludge might be legally mandatory.

For many programs, an underlying goal of administrative burdens is to ensure “program integrity,” [4][4]. See Protecting Americans from Tax Hikes (PATH) Act, Pub. L. No. 114-113, tit. II, 129 Stat. 2242, 3076–86 (2015) (referring to Title II as “Program Integrity” and specifically intending to reduce fraudulent and improper payments in the EITC and other programs); Leslie Book, David Williams & Krista Holub, Insights from Behavioral Economics Can Improve Administration of the EITC, 37 Va. Tax Rev. 177, 180 (2018) (noting that “[p]rogram integrity” of the EITC was an important topic among employees of the IRS because 43 to 50 percent of all EITC returns are incorrect, with most errors benefitting claimants); Program Integrity, Ctrs. for Medicare & Medicaid Servs., https://www.medicaid.gov/medicaid/program-integrity/index.html [https://perma.cc/2ZMC-XTSH] (explaining CMS responsibilities and goals under the Medicaid Integrity Program); Reducing Improper Payments, Soc. Sec. Admin., https://www.ssa.gov/improperpayments [https://perma.cc/T8ZN-XA32] (emphasizing the objective of curbing improper payments). meant to refer to an absence of fraud, abuse, misuse, or access by the ineligible; sludge might be necessary for that purpose. A large amount of sludge—paperwork burdens, delays, information collection—could be an essential way to limit benefits, including taxpayer money, to those who have a legal right to them. Of course it is also true that sludge might be a covert (or at least not readily visible) way to limit expenditures; a reflection of a compromise between those who like a benefits program and those who do not much like it; or a way for the government to acquire information that could be used for multiple purposes. [5][5]. See Sunstein, supra note 1, at 1870–71 (pointing to sludge as a way to ration resources for those willing to tolerate administrative burdens); id. at 1872 (noting how certain administrative burdens help agencies collect useful data on public programs like employment training or infectious disease prevention efforts). But program integrity is often the central motivation. [6][6]. See id. at 1865 (noting agencies impose paperwork burdens to ensure programs comply with the law).

As a case in point, consider the Earned Income Tax Credit (“EITC”), [7][7]. The details and requirements of the EITC are set out in 26 U.S.C. § 32 (2018). a wage subsidy for low-income workers, provided by the Internal Revenue Service (“IRS”). [8][8]. See Pamela Herd & Donald P. Moynihan, Administrative Burden: Policymaking by Other Means 191, 193 (2019). For a general discussion of the EITC and its effects on recipients, see generally Making Work Pay: The Earned Income Tax Credit and Its Impact on America’s Families (Bruce D. Meyer & Douglas Holtz-Eakin eds., 2001). For the EITC, the take-up rate is about 80 percent. [9][9]. Herd & Moynihan, supra note 8, at 191. By the standards of most benefits programs, that number is not especially low. But if 20 percent of eligible people are not receiving a potentially life-changing benefit, and if the reason is the degree of sludge, there is a serious problem. An approach that denies such a benefit to millions of people surely should be rethought. The existing amount of sludge would seem excessive, and it should be reduced.

The IRS almost certainly knows enough to enroll people automatically and send a refund to eligible taxpayers. If so, sludge could be eliminated entirely, and the take-up rate should be close to 100 percent. Is that what the IRS should be doing? [10][10]. Surprisingly, there appears to be a lack of substantial literature on automatic enrollment and the EITC. But see Vada Waters Lindsey, Encouraging Savings Under the Earned Income Tax Credit: A Nudge in the Right Direction, 44 U. Mich. J.L. Reform 83, 112–21 (2010) (arguing that the EITC should have a retirement savings component with automatic enrollment). This area deserves sustained study. Suppose that the IRS indeed decided to send the EITC automatically to apparently eligible taxpayers. If it could do so at low cost, and if the apparently eligible taxpayers are in fact eligible, there would seem to be little ground for objection. The problem, of course, is the phrase “apparently eligible taxpayers.” It is possible that some of the recipients, under automatic enrollment, will not in fact be eligible. Whenever people are automatically enrolled in a program, some of them may not meet the legal criteria.

Or consider the question of voting rights. With respect to both registration and voting, some states impose a degree of sludge, on the articulated ground that it is necessary to ensure that ineligible people do not end up voting. [11][11]. See Herd & Moynihan, supra note 8, at 47–79 (discussing the history of administrative burdens in the exercise of voting rights). But what if the result is also, and simultaneously, to discourage eligible people from voting? [12][12]. See, e.g., id. at 56–57 (noting a survey among registered Texas voters found 6 percent of eligible voters said not having proper identification was “the primary reason they did not vote,” despite the interviewers finding that the “vast majority” did in fact have proper identification). What if sludge operates as a mechanism for voter suppression? That problem is pervasive. [13][13]. Id. at 49; see also Vote Suppression, Brennan Ctr. for Just., https://www.brennancenter.org/issues/ensure-every-american-can-vote/vote-suppression [https://perma.cc/9LBC-W3QD]. For benefits programs of multiple kinds, sludge might be justified as a way of ensuring that ineligible people do not gain access to goods or opportunities; but the consequence might be to screen out the eligible as well. If people have to navigate sludge to exercise their constitutional rights, there seems to be a serious problem.

The remainder of this Essay is organized as follows. Part II explores the question of optimal sludge as a matter of principle. It shows that if we care about consequences, we might favor sludge reduction even if the consequence is to give benefits or opportunities to people who are ineligible. But if we want to insist on strict adherence to legal requirements, the grant of benefits to ineligible individuals is far worse than the denial of benefits, through sludge, to eligible individuals. Some statutes may, in fact, require sludge even if the overall consequences are plausibly described as very bad. Part III describes our survey, which found that most people embrace a form of consequentialism and firmly reject legalism. Part III also discusses differences along the lines of party affiliation and demographics; surprisingly, the survey found only modest differences along these lines. Part IV, a brief conclusion, argues in favor of sludge reduction on consequentialist grounds, even if the result is to give some benefits or opportunities to ineligible people.

II. Normative Issues

Public officials must often choose between (1) a design ensuring that a certain number of eligible people will not receive a benefit and (2) a design ensuring that a certain number of ineligible people will receive a benefit. That may not be an easy choice, and it may not even be easy to decide on the criteria that officials should use to make it. On one view, we should focus on consequences, which means that we should try to specify relevant numbers: How many eligible people would obtain benefits to which they are entitled? How many ineligible people would obtain benefits to which they are not entitled? On another view, we should focus on the underlying statute and the class of people that it seeks to benefit, certainly if it mandates a particular conclusion, and perhaps even if it does not.

A. Consequences and Consequentialism

If the idea of program integrity is meant to refer to the number of errors, the choice between view (1) and view (2) above might turn on some kind of arithmetic. Which group is larger? If automatic enrollment means that 500,000 eligible people will receive a benefit who otherwise would not, but 499,999 ineligible people will also receive a benefit who otherwise would not, automatic enrollment is justified. Call this a consequentialist position; it is consequentialist in the sense that it assesses the optimal level of sludge by focusing on its consequences.

To be sure, the consequentialist position might be specified in different ways, and a simple count might be too crude. It could be challenged from multiple points of view. Suppose that automatic enrollment gives benefits to 200,000 eligible people but also to 200,100 ineligible people. Some people might think that if the 200,100 people are nearly eligible—if they are relatively poor—it is not so terrible if they receive some economic help. But other people might insist that taxpayer money is accompanied by clear restrictions and argue that if it is given out in violation of those restrictions, a grievous wrong has been committed. They might think that if automatic enrollment gives benefits to 200,000 eligible people but also to 125,000 ineligible people, it is a terrible idea. On consequentialist grounds, reasonable people could differ about the appropriate tradeoff; they may or may not require a significant disparity between the size of the group of benefited eligible persons and the size of the group of benefited ineligible persons.

We might agree that if sludge reduction or automatic enrollment ensures that a very large number of eligible people receive benefits who otherwise would not, and that a very small number of ineligible people also receive benefits who otherwise would not, all is well. Perhaps we could even agree that if sludge reduction or automatic enrollment ensures that a large number of eligible people receive benefits who otherwise would not, and that a significant number of ineligible people receive benefits who otherwise would not, all is still well. For example, in the context of COVID-19, the U.S. government eliminated a great deal of sludge to promote delivery of health and economic benefits. [14][14]. Specifically, the stimulus checks set out under the Coronavirus Aid, Relief, and Economic Security (“CARES”) Act , Pub. L. No. 116-136, § 2201, 134 Stat. 281, 335 (2020), provide a good gauge of public sentiment regarding sludge in emergency scenarios. There were certainly concerned voices regarding the distribution of stimulus checks to deceased individuals as well as the living, but the speed at which the Treasury Department sent out the checks may have justified this oversight. See Erica Werner, Treasury Sent More than 1 Million Coronavirus Stimulus Payments to Dead People, Congressional Watchdog Finds, Wash. Post (June 25, 2020, 8:09 PM), https://www.washingtonpost.com/us-policy/2020/06/25/irs-stimulus-checks-dead-people-gao [https://perma.cc/529D-5XXQ]; Eric Levitz, Treasury Sent $1.4 Billion in Stimulus Checks to Dead People. That’s Good, N.Y. Mag: Intelligencer (June 25, 2020), https://nymag.com/intelligencer/2020/06/irs-coronavirus-stimulus-checks-dead-people-gao.html [https://perma.cc/4AX7-TD3Z]. In fact, greater frustration was likely caused by the fact that while the dead received $1.4 billion in stimulus, tens of millions of living Americans did not see their checks at all. See Savannah Levins, Still No Stimulus Check? You’re Not Alone, WCNC (June 23, 2020, 11:59 AM), https://www.wcnc.com/article/money/still-no-stimulus-check-youre-not-alone/275-11793c79-7fb7-4201-9f42-343b6d602397 [https://perma.cc/CWJ8-J3JZ] (noting that 35 million Americans still had not received their stimulus checks as of June 2020). Even if those steps ultimately rewarded many people who were not in fact eligible, including people who were no longer alive, [15][15]. See Werner, supra note 14 (describing how the Treasury Department and IRS mistakenly disbursed stimulus checks to almost 1.1 million dead people in the rush to quickly deliver payments). perhaps they were nonetheless justified. But reasonable people can disagree about these questions. [16][16]. Compare John Kennedy, Sending Coronavirus Payments to Ghosts Hurts the Living, Hill (May 7, 2020, 2:00 PM), https://thehill.com/blogs/congress-blog/politics/496615-sending-coronavirus-payments-to-ghosts-hurts-the-living [https://perma.cc/6G7C-F5QS] (“Even under normal circumstances, we should not tolerate this level of waste and incompetence, but COVID-19 has landed us far from normal times. . . . Even the most fiscally unfettered politician can’t justify cutting checks to people who are not even alive.”), with Levitz, supra note 14 (“The program had an error rate of roughly 0.4 percent. That is not a scandal. It is a triumph.”), and Cindy Carcamo, Why Are the Dead Getting Coronavirus Stimulus Checks?, L.A. Times (Apr. 24, 2020, 5:00 AM), https://www.latimes.com/california/story/2020-04-24/coronavirus-stimulus-check-death-payment [https://perma.cc/3RPB-UF3D] (quoting a former head of an IRS watchdog group noting that “[t]he IRS is following the law,” as well as a vice president of conservative think tank that “[e]xpediency has been prioritized,” rather than delaying checks for weeks and months by cross-referencing data to eliminate dead recipients).

As usual, consequentialists need to specify what matters, so that we can know what counts as a good or bad consequence, and how good or bad it is. We should be able to agree that it is good, and often very good, to ensure that eligible people receive benefits to which they are entitled (especially if they are poor or sick). But how bad is it, exactly, if ineligible people receive such benefits as well?

That question remains to be answered; consequentialism, in the abstract, cannot do so. In terms of drawing the right conclusion, we will also want to specify the substantive area that is involved. More or less sludge might be imposed and tolerated in such diverse domains as antipoverty programs, voting, abortion, student visas, and occupational licenses. [17][17]. See Sunstein, supra note 1, at 1850–52; see generally Herd & Moynihan, supra note 8 (discussing administrative burdens in a wide variety of policy arenas). In one area, reasonable consequentialists might be willing to accept a grant of a good or opportunity to a large number of ineligibles, without also committing themselves to the same level of toleration in another area.

B. Legalism

On an alternative view, even a modest breach of program integrity, for the advantage of those who are not eligible, is unacceptable. Call this the legalist position. It is legalist in the sense that it sees a violation of legal requirements as a serious breach, one that cannot be justified on consequentialist grounds. An extreme version of the legalist position would be that a grant of benefits to a very large number of eligible people would not outweigh the grant of benefits to a very small number of ineligible people.

We need to make a distinction here. Suppose that the law is conclusive on the question. Suppose that properly read, it mandates a degree of sludge. Or suppose that it forbids officials from taking steps to provide benefits to small numbers of ineligible people, even if those steps are necessary to ensure that benefits go to large numbers of eligible people. If so, the law is conclusive. Officials have no room to maneuver; they are strictly bound. Call this mandatory legalism. If a statute prohibits approaches that would allow ineligible people to receive benefits, the legal question is at an end, and consequentialism is off-limits.

But suppose, alternatively, that the law gives officials discretion to adopt various approaches, such that they can choose an approach that inadvertently but inevitably grants benefits to ineligibles without violating any statutory provision, if that is the price of giving benefits to more eligible individuals. For example, a statute involving voter registration might authorize states to adopt registration procedures that do not entirely screen out people who are ineligible. Or a statute providing free school meals for poor children might come with a general appropriation and give authority to state agencies to ascertain how best to ensure receipt by those who are eligible—authority that might include a license to make some mistakes. [18][18]. For example, the Food, Conservation, and Energy Act of 2008 provides a general appropriation to states for the administration of the Supplemental Nutrition Assistance Program (“SNAP”), but states are allowed to determine the best means of certifying which children are eligible for free school meals. See Food, Conservation, and Energy Act of 2008, Pub. L. No. 110-234, §§ 4111–22, 122 Stat. 923, 1102–13; see generally U.S. Dep’t of Agric., Direct Certification in the National School Lunch Program: : State Implementation Progress Report to Congress—School Year 2015–2016 and 2016–2017 (2018), https://www.fns.usda.gov/sites/default/files/resource-files/NSLPDirectCertification2016.pdf [https://perma.cc/QBV5-24TF] (reporting to Congress “on State progress in direct certification in the National School Lunch Program”). But some people might say that officials should not take this approach, even if they have the legal authority to do so, because it is a grievous wrong to use one’s discretion to give benefits to ineligibles. Call this discretionary legalism.

In principle, mandatory legalism is right, if the statute really is conclusive. But in principle, extreme versions of discretionary legalism—where, it will be recalled, the statute is not conclusive—would seem hard to defend: a grant of benefits to ten people who are almost (but not) eligible would appear to be a price worth paying in exchange for a grant of benefits to two million people who are in fact eligible. But granting benefits in violation of the law is no light thing, even if officials do have some discretion under the statute, and the correct tradeoff is not self-evident. Public officials encounter the problem frequently, and so do companies, universities, nonprofits, homeless shelters, hospitals, and others. It is a general truth that as the burdens of eligibility requirements increase, the class of people who receive a benefit may well include fewer ineligibles; and as those burdens loosen, there is an ever-growing risk that a benefit will go to people who should not be getting it.

III. A Pilot Survey

We could imagine theoretical debates about these questions, and they would undoubtedly allow us to make progress. But let us now turn to an empirical question: What do people actually think about these issues? In a democracy, public officials are attentive to what members of the public think, and if most people turn out to favor an identifiable form of consequentialism, or instead embrace a form of legalism, their judgment would certainly be of interest. And for those who believe in “the wisdom of crowds,” a widespread public judgment might also have epistemic value. It might tell us something about what is right.

We do not mean to say that the results of surveys should be taken to be decisive. Suppose that under the law, agencies have no statutory authority to give out benefits to people who are not eligible, even if a sludge-reduction strategy, producing that result, is the only way to ensure that benefits are received by a very large number of eligible people. Even if people approve of that result on consequentialist grounds, it is by hypothesis unlawful. Or suppose that some form of discretionary legalism is correct; those who embrace it would be unlikely to change their stance upon learning that, in surveys, most people disagree with them. Or suppose that some form of consequentialism is consistent with law and also can be shown to benefit a very large number of people. Should we reject it because most people do so in surveys? Committed consequentialists would not be willing to do so. They might insist that survey answers do not show informed or reflective thinking, and that their own approach is preferable in principle.

For these reasons, we do not suggest or believe that, for democratic reasons, the views elicited in surveys should be taken as decisive. But they are of unquestionable interest. At a minimum, public officials, when deciding what to do, will be attuned to whether the public embraces or rejects one or another position. Surveys are informative about that question.

With that point in mind, we conducted a pilot survey in which we attempted to elicit people’s views on the underlying questions in one substantive domain. We emphasize the word “pilot”; this was meant as a preliminary effort to obtain a sense of what people think. Using Amazon’s Mechanical Turk, we asked 407 people to consider four different scenarios, specifying different numbers of ineligible people who would receive EITC benefits under automatic enrollment. The survey was not nationally representative, but it did have a high degree of diversity. The participants were 60 percent male, 40 percent female; 30 percent Republican, 43 percent Democratic, 27 percent independent (that is, unassociated voters); 68 percent White, 14 percent Asian, 7 percent Black, 4 percent Hispanic. We chose a within-subjects design, in which all subjects saw all four scenarios, with the goal of promoting comparative assessments.

We described four scenarios, reprinted below, and then asked the survey participants a simple question: whether they would approve of what we described as a “new approach” to the distribution of the EITC, which would have varying consequences. (Participants were asked a binary question: approve or disapprove.) The actual text read as follows:

(1) Under the new approach, the IRS would give the Earned Income Tax Credit only to the 4 million people who are actually eligible; no one would receive it who is not eligible.
(2) The IRS would inevitably make mistakes, and it would send out some money to people who are not, in fact, eligible for the Earned Income Tax Credit. Under the new approach, 4 million people would receive the credit who would otherwise not get it. Of those, 3.98 million are eligible, but 20,000 are not.
(3) The IRS would inevitably make mistakes, and it would send out some money to people who are not, in fact, eligible for the Earned Income Tax Credit. Under the new approach, 4 million people would receive the credit who would otherwise not get it. Of those, 3.9 million are eligible, but 100,000 are not.
(4) The IRS would inevitably make mistakes, and it would send out some money to people who are not, in fact, eligible for the Earned Income Tax Credit. Under the new approach, 4 million people would receive the credit who would otherwise not get it. Of those, 3.5 million are eligible, but 500,000 are not.

The aggregate results were straightforward. Very strong majorities approved of automatic enrollment in scenarios (1) and (2); a solid majority approved of automatic enrollment in scenario (3); and a solid majority rejected it in scenario (4). In short, most people rejected a legalist approach to these questions, in favor of a form of consequentialism. Even if 100,000 people would receive benefits to which they were not entitled, people would favor automatic enrollment. And when as many as 500,000 ineligible people would receive benefits, a substantial minority would nonetheless approve. It is noteworthy, however, that with those numbers, the majority disapproved, demonstrating that in scenario (4), many consequentialists would draw the line.

More specifically: For scenario (1), the approval rate was 89 percent, with Democrats at 91 percent, independents at 90 percent, and Republicans at 85 percent. For scenario (2), the approval rate was 75 percent, with Democrats at 84 percent, independents at 74 percent, and Republicans at 62 percent. For scenario (3), the approval rate was 61 percent, with Democrats at 74 percent, independents at 54 percent, and Republicans at 50 percent. Lastly, for scenario (4), the approval rate was 44 percent, with Democrats at 49 percent, independents at 37 percent, and Republicans at 42 percent.

We expected to find significant differences across various lines, but in testing for statistical significance, we were surprised to find few differences along almost every line that we tested. [19][19]. For further details, see infra app. In particular, we anticipated much higher approval ratings from Democrats rather than from Republicans, but even there, the picture is complicated, for the differences were less than stark. To be sure, there were significant differences between Republicans and Democrats in scenarios (2) and (3). Note, however, that the differences between the two parties is not statistically significant for scenarios (1) and (4), which makes it hazardous to say that we have found a consistent difference, on these issues, along party lines. In these circumstances, the relatively high level of agreement among Democrats, independents, and Republicans is the main story.

Analysis of a range of demographic differences also failed to find significant differences, perhaps because of the relatively small sample sizes. Surprisingly, we were unable to find differences between men and women or along lines of race or educational attainment. The few significant differences that we did find tell no clear or obvious tale, and we would not make much of them. For example, we found a statistically significant result for those in the $120,000–$149,999 salary range in scenarios (3) and (4): those in this income category (the second highest) showed lower approval ratings than did people in other income categories. With respect to age, we found that 50–64-year-olds showed a significantly lower approval rate in scenario (3), and those 65 and over showed a significantly lower approval rate in scenario (1).

Because these findings seem a bit random, and do not show differences across all four scenarios, it would be hazardous to draw general lessons from them. The important story is shown by the aggregate data, and by the significant, but relatively modest, differences between Republicans and Democrats in scenarios (2) and (3).

Conclusion

There has been growing attention to administrative burdens and sludge, with a mounting awareness that many eligible people are deterred or prevented from obtaining benefits or opportunities to which they are legally entitled. The strongest justification for such burdens is that they are necessary to screen out ineligible people. Whether this is an adequate justification depends on whether we embrace consequentialism, mandatory legalism, or discretionary legalism, and how we specify each of these approaches.

In principle, and if the law permits it, we would argue in favor of a form of consequentialism. Suppose that a sludge-reduction approach ensures that a very large number of eligible people receive an economic benefit who otherwise would not, while also ensuring that a very small number of ineligible people also end up with that benefit. The details matter, but in general, we would favor that approach. We would also be willing to embrace consequentialism even if the numbers were not so overwhelmingly supportive of sludge reduction. But we do not mean here to take a specific stand on what consequentialism entails, or about the appropriate conclusions when particular numbers are presented. The substantive area undoubtedly matters. Reasonable consequentialists might offer different conclusions in the domains of antipoverty programs, voting, abortion, student visas, and occupational licenses. In one area, they might tolerate a grant of a good or opportunity to ineligibles, without also committing themselves to the same level of toleration in another area.

In the particular domain of the EITC, a pilot survey suggests that most people embrace consequentialism, and that they are willing to accept the grant of benefits to numerous ineligible people if that is the price of granting benefits to far more numerous eligible people. It appears that people will accept automatic enrollment unless it means that benefits are also granted to an extraordinarily high number of ineligible people.

A great deal of research remains to be done. In the future, it would be valuable to probe the normative and legal issues in more detail. Recall mandatory legalism and suppose that a form of consequentialism would result in a violation of relevant statutes—perhaps because it would unlawfully grant economic benefits to ineligible people. If so, that approach should not be adopted. But how many and which statutes are best understood in this way? How many and which statutes should alternatively be understood to give agencies discretion to take sludge-reducing steps that admittedly give out benefits to people who are not eligible? It would be most valuable to obtain answers to these questions.

It would also be valuable to obtain a more comprehensive understanding of what people think in different areas, and to explore whether there are differences along political or demographic lines. Our own study is a mere pilot in this domain, with a relatively small sample. We could do far more to probe what people actually think. Even now, however, we believe that a heavy burden of justification lies on those who would reject effective sludge-reduction efforts, including automatic enrollment, even if the unintended but inevitable consequence would be to benefit some or even many people who are not in fact eligible. So long as sludge-reduction efforts are consistent with law and basic principle, the survey results here would suggest that those efforts deserve support, at least if they are necessary to ensure that large numbers of eligible people receive potentially life-changing or life-saving benefits that they would not otherwise receive.

Appendix


Table 1.
This table reports the results of the survey discussed in the text. More specifically, it reports the approval ratings for each demographic group with respect to the four scenarios. Recall that in these scenarios, the IRS gives the EITC to 4 million people:

(1) All of whom are actually eligible for the EITC;

(2) 3.98 million of whom are eligible, but 20,000 are not eligible for the EITC;

(3) 3.9 million of whom are eligible, but 100,000 are not eligible for the EITC; or

(4) 3.5 million of whom are eligible, but 500,000 are not eligible for the EITC.


Figure 2. Significance testing

Exponentiated coefficients; p-values in parentheses



Copyright © 2020 Cass R. Sunstein and Julien L. Gosset.

Robert Walmsley University Professor, Harvard University.
††Class of 2021, Harvard College. We are grateful to Arevik Avedian for help with the survey here and to Rohit Goyal for help with the statistical analysis. The Program on Behavioral Economics and Public Policy at Harvard Law School provided essential support.

[1] For examples, see Cass R. Sunstein, Sludge and Ordeals, 68 Duke L.J. 1843, 1849–51 (2019).

[2] See id. (“The term should be taken to refer to the kind of friction, large or small, that people face when they want to go in one or another direction.”); Richard Thaler, Nudge, Not Sludge, 361 Science 431, 431 (Aug. 3, 2018), https://science.sciencemag.org/content/sci/361/6401/431.full.pdf [https://perma.cc/J3VW-JJG4] (“But what about activities that are essentially nudging for evil? This ‘sludge’ just mucks things up and makes wise decision-making and prosocial activity more difficult.”). See generally Cass R. Sunstein, Sludge (forthcoming 2021) (manuscript on file with authors).

[3] See Sunstein, supra note 1, at 1862–64 (noting sludge in Obamacare and Medicare, as well as state regulations concerning abortion and voter registration).

[4] See Protecting Americans from Tax Hikes (PATH) Act, Pub. L. No. 114-113, tit. II, 129 Stat. 2242, 3076–86 (2015) (referring to Title II as “Program Integrity” and specifically intending to reduce fraudulent and improper payments in the EITC and other programs); Leslie Book, David Williams & Krista Holub, Insights from Behavioral Economics Can Improve Administration of the EITC, 37 Va. Tax Rev. 177, 180 (2018) (noting that “

rogram integrity” of the EITC was an important topic among employees of the IRS because 43 to 50 percent of all EITC returns are incorrect, with most errors benefitting claimants); Program Integrity, Ctrs. for Medicare & Medicaid Servs., https://www.medicaid.gov/medicaid/program-integrity/index.html [https://perma.cc/2ZMC-XTSH] (explaining CMS responsibilities and goals under the Medicaid Integrity Program); Reducing Improper Payments, Soc. Sec. Admin., https://www.ssa.gov/improperpayments [https://perma.cc/T8ZN-XA32] (emphasizing the objective of curbing improper payments).

[5] See Sunstein, supra note 1, at 1870–71 (pointing to sludge as a way to ration resources for those willing to tolerate administrative burdens); id. at 1872 (noting how certain administrative burdens help agencies collect useful data on public programs like employment training or infectious disease prevention efforts).

[6] See id. at 1865 (noting agencies impose paperwork burdens to ensure programs comply with the law).

[7] The details and requirements of the EITC are set out in 26 U.S.C. § 32 (2018).

[8] See Pamela Herd & Donald P. Moynihan, Administrative Burden: Policymaking by Other Means 191, 193 (2019). For a general discussion of the EITC and its effects on recipients, see generally Making Work Pay: The Earned Income Tax Credit and Its Impact on America’s Families (Bruce D. Meyer & Douglas Holtz-Eakin eds., 2001).

[9] Herd & Moynihan, supra note 8, at 191.

[10] Surprisingly, there appears to be a lack of substantial literature on automatic enrollment and the EITC. But see Vada Waters Lindsey, Encouraging Savings Under the Earned Income Tax Credit: A Nudge in the Right Direction, 44 U. Mich. J.L. Reform 83, 112–21 (2010) (arguing that the EITC should have a retirement savings component with automatic enrollment). This area deserves sustained study.

[11] See Herd & Moynihan, supra note 8, at 47–79 (discussing the history of administrative burdens in the exercise of voting rights).

[12] See, e.g., id. at 56–57 (noting a survey among registered Texas voters found 6 percent of eligible voters said not having proper identification was “the primary reason they did not vote,” despite the interviewers finding that the “vast majority” did in fact have proper identification).

[13] Id. at 49; see also Vote Suppression, Brennan Ctr. for Just., https://www.brennancenter.org/issues/ensure-every-american-can-vote/vote-suppression [https://perma.cc/9LBC-W3QD].

[14] Specifically, the stimulus checks set out under the Coronavirus Aid, Relief, and Economic Security (“CARES”) Act , Pub. L. No. 116-136, § 2201, 134 Stat. 281, 335 (2020), provide a good gauge of public sentiment regarding sludge in emergency scenarios. There were certainly concerned voices regarding the distribution of stimulus checks to deceased individuals as well as the living, but the speed at which the Treasury Department sent out the checks may have justified this oversight. See Erica Werner, Treasury Sent More than 1 Million Coronavirus Stimulus Payments to Dead People, Congressional Watchdog Finds, Wash. Post (June 25, 2020, 8:09 PM), https://www.washingtonpost.com/us-policy/2020/06/25/irs-stimulus-checks-dead-people-gao [https://perma.cc/529D-5XXQ]; Eric Levitz, Treasury Sent $1.4 Billion in Stimulus Checks to Dead People. That’s Good, N.Y. Mag: Intelligencer (June 25, 2020), https://nymag.com/intelligencer/2020/06/irs-coronavirus-stimulus-checks-dead-people-gao.html [https://perma.cc/4AX7-TD3Z]. In fact, greater frustration was likely caused by the fact that while the dead received $1.4 billion in stimulus, tens of millions of living Americans did not see their checks at all. See Savannah Levins, Still No Stimulus Check? You’re Not Alone, WCNC (June 23, 2020, 11:59 AM), https://www.wcnc.com/article/money/still-no-stimulus-check-youre-not-alone/275-11793c79-7fb7-4201-9f42-343b6d602397 [https://perma.cc/CWJ8-J3JZ] (noting that 35 million Americans still had not received their stimulus checks as of June 2020).

[15] See Werner, supra note 14 (describing how the Treasury Department and IRS mistakenly disbursed stimulus checks to almost 1.1 million dead people in the rush to quickly deliver payments).

[16] Compare John Kennedy, Sending Coronavirus Payments to Ghosts Hurts the Living, Hill (May 7, 2020, 2:00 PM), https://thehill.com/blogs/congress-blog/politics/496615-sending-coronavirus-payments-to-ghosts-hurts-the-living [https://perma.cc/6G7C-F5QS] (“Even under normal circumstances, we should not tolerate this level of waste and incompetence, but COVID-19 has landed us far from normal times. . . . Even the most fiscally unfettered politician can’t justify cutting checks to people who are not even alive.”), with Levitz, supra note 14 (“The program had an error rate of roughly 0.4 percent. That is not a scandal. It is a triumph.”), and Cindy Carcamo, Why Are the Dead Getting Coronavirus Stimulus Checks?, L.A. Times (Apr. 24, 2020, 5:00 AM), https://www.latimes.com/california/story/2020-04-24/coronavirus-stimulus-check-death-payment [https://perma.cc/3RPB-UF3D] (quoting a former head of an IRS watchdog group noting that “[t]he IRS is following the law,” as well as a vice president of conservative think tank that “[e]xpediency has been prioritized,” rather than delaying checks for weeks and months by cross-referencing data to eliminate dead recipients).

[17] See Sunstein, supra note 1, at 1850–52; see generally Herd & Moynihan, supra note 8 (discussing administrative burdens in a wide variety of policy arenas).

[18] For example, the Food, Conservation, and Energy Act of 2008 provides a general appropriation to states for the administration of the Supplemental Nutrition Assistance Program (“SNAP”), but states are allowed to determine the best means of certifying which children are eligible for free school meals. See Food, Conservation, and Energy Act of 2008, Pub. L. No. 110-234, §§ 4111–22, 122 Stat. 923, 1102–13; see generally U.S. Dep’t of Agric., Direct Certification in the National School Lunch Program: : State Implementation Progress Report to Congress—School Year 2015–2016 and 2016–2017 (2018), https://www.fns.usda.gov/sites/default/files/resource-files/NSLPDirectCertification2016.pdf [https://perma.cc/QBV5-24TF] (reporting to Congress “on State progress in direct certification in the National School Lunch Program”).

[19] For further details, see infra app.