Leaving benefits on the table: Evidence from SNAP

https://doi.org/10.1016/j.jpubeco.2019.104054Get rights and content

Abstract

Many social insurance programs require that beneficiaries periodically re-verify eligibility, allowing officials to identify ineligible applicants but risking the loss of eligible applicants who do not complete these requirements. While many states have invested in simplified verification procedures to reduce that risk, the extent of exit among eligible beneficiaries and the efficacy of these simplifications is not well established. This paper uses administrative data from the Supplemental Nutrition Assistance Program (SNAP) to establish three facts. First, retention in SNAP is low, with approximately one-half of entering cases leaving in the first year. Second, qualitative evidence and quantitative simulations suggest that approximately half of those who exit in the first year remain eligible. Third, using the staggered roll-out of an online case management tool in Michigan, I find that this simplification reduced the rate of long-term exit at key verification dates by almost 2pp (12%). These facts suggest that eligible retention is very incomplete, and that ongoing simplification efforts increase retention among eligible beneficiaries.

Introduction

While researchers have documented low take-up in many public assistance programs (Ribar, 2014), few have distinguished between initial take up and retention in transfer programs. There is reason to believe that retention is an important channel of incomplete take-up in it's own right: government agencies often require periodic verifications of eligibility, and may incidentally filter out eligible beneficiaries who fail to complete these requirements. The purpose of this paper is to illustrate the importance of understanding retention in at least one major transfer program – the Supplemental Nutrition Assistance Program (SNAP) – and to test the extent to which simplifying verification procedures increases program retention.

I make three distinct contributions. First, I document that retention in SNAP is quite low. Using administrative SNAP enrollment data from seven states, I find that about one-half of new SNAP cases are not receiving benefits one year later. Exits are concentrated at points when verification is required – especially at so-called recertification months – and usually last for an extended period of time. Second, I provide evidence that many of these exits occur among households that remain eligible. While I cannot directly observe eligibility for households that lose contact with SNAP officials, I use three distinct facts to support this conclusion:

  • Administrative data from two states report the reason for program exit, which is overwhelmingly the result of a failure to submit paperwork rather than confirmed ineligibility.

  • Combining administrative SNAP enrollment data with Unemployment Insurance (UI) records for the state of Michigan, I find that households with and without UI-covered earned income have similar rates of program exit at recertification.

  • I simulate unobserved changes in income and household size for cases in Michigan by following households with similar demographics and earnings paths in the Survey of Income and Program Participation (SIPP). Simulations suggest that approximately half of the cases that leave the program by their one-year anniversary are still eligible.

Third, I provide quasi-experimental evidence that administrative barriers meaningfully affect retention. In 2009, Michigan introduced a website that helped beneficiaries understand program requirements, communicate with SNAP officials, and track their benefits online. The website was rolled out incrementally across different groups of counties at different times through the year. The new interface coincides with a sharp reduction in long-term exit at recertification of almost 2 percentage points (12%), with larger effects for childless adults and for cases with earned income.

This paper builds on an existing body of work that studies retention in public assistance programs, but alleviates multiple barriers that have limited previous empirical progress. One existing literature finds low retention in public health insurance programs (Dick et al., 2002, Pei, 2017, Sommers, 2007, Sommers and Rosenbaum, 2011). These results may not apply to other transfer programs, since beneficiaries may not know that their health insurance has lapsed and can often re-enroll retroactively. In contrast, those who exit SNAP lose immediate and often substantial flow benefits, and discover this quickly when they try to buy groceries. A different strand of literature finds moderate effects of SNAP policy changes using state-level difference-in-difference methods over extended time horizons (Ganong and Liebman, 2018, Heflin and Mueser, 2010, Pomerleau, 2013, Schwabish, 2012). This paper complements that work and corroborates its broad conclusions by studying the within-state roll-out of a program over a short time horizon, thereby alleviating possible concerns over differential trends or endogenous state policy decisions. Moreover, this paper draws from a small literature documenting a large amount of non-random under-reporting among SNAP beneficiaries in surveys (Meyer and Goerge, 2010, Meyer and Mittag, 2015). I rely almost exclusively on administrative panel data collected by state agencies, allowing for reliable and representative findings.

Perhaps the most closely related paper is Homonoff and Somerville (2019), who show that recertification rates are substantially lower when recertification interviews are scheduled for dates later in the calendar month. This leaves less time for beneficiaries to reschedule a missed interview before benefits expire at the end of the calendar month, leading to substantial short-term churn from the program. This work and my analysis are complementary: while they demonstrate that a certain class of administrative barriers (one-time interview deadlines) increases short-term churn disproportionately for a certain subset of beneficiaries (lower-resource cases with children), I demonstrate that a different class of administrative barriers (paperwork burdens) increases long-term dropout disproportionately for a different subset of beneficiaries (higher-income cases without children). In theory, the distinct impacts of these and other policy tools could be combined to more precisely direct benefits towards beneficiaries with specific observable characteristics.

This paper will proceed as follows. In Section 2, I explain the institutional background of the SNAP program and describe my data sources. Section 3 documents rates of SNAP retention using multiple measures across seven states. In Section 4, I present evidence that many of the program exits are among cases that would still be eligible, with reasonable simulations suggesting a fraction around one-half. Section 5 demonstrates that states' efforts to simplify the recertification process can meaningfully reduce the number of eligible beneficiaries exiting the program. I show this by documenting the effects of a new website in Michigan, which I can credibly study due to its staggered roll-out. I emphasize that this particular policy does not fully eliminate eligible exits. Rather, the fact that this modest simplification appears to have a meaningful impact suggests that the suite of similar efforts that states have undertaken are likely to increase eligible retention. Section 6 concludes.

Section snippets

The SNAP program

The Supplemental Nutrition Assistance Program (SNAP) is among the most widespread social insurance programs in the United States: in July 2011, SNAP served about 15% of the U.S. population (Ganong and Liebman, 2018). The core aspects of the SNAP program are the same across all U.S. states. Each month, households enrolled in the program get money loaded onto an Electronic Benefits Transfer (EBT) card, which they can use to buy most food items at grocery or convenience stores. Households of a

Retention in SNAP

I first show that retention is low across many states' SNAP programs. Fig. 1a shows survival curves for SNAP enrollment in Michigan by following each household for 24 months after their first enrollment in my sample window. The solid line shows the percent of households that are enrolled at each month relative to their entry month, while the dotted line shows the percent that have continuously remained enrolled in the program since entry. In Michigan, over half of SNAP entrants exit at some

Eligibility of exiting cases

I next provide evidence that a high fraction of exits in SNAP are among eligible beneficiaries. Reliably measuring eligibility at the case level would require accurate, linked data on many disparate sources of income and expenses. To my knowledge, no such data source exists. However, using three pieces of evidence, I estimate that approximately one-half of those who have exited the program by the one-year mark are still eligible.

I first show that exits are typically associated with missed

Simplifying administrative requirements

The evidence suggests that state SNAP agencies face a trade-off in setting recertification policies: more frequent recertifications allow the program to shed ineligible applicants, but as many as half of the cases that exit are actually still eligible. These households overwhelmingly exit because they fail to complete the requirements of recertification, which include collecting pay stubs and receipts, attending an interview, and communicating with the state agency when paperwork is missing or

Conclusion

This paper illustrates the importance of understanding retention in at least one major social insurance program, and suggests that simplifying recertification procedures meaningfully reduces exits among eligible beneficiaries. I establish three facts. First, retention in SNAP is low, with about one-half of newly entering cases exiting within the year. Second, a substantial fraction of those who exit remain eligible. Simulations using Michigan administrative data and the SIPP suggest that about

References (16)

  • C.A. Bee et al.

    Do older Americans have more income than we think?

    SSRN

    (2017)
  • A. Belloni et al.

    High-dimensional methods and inference on structural and treatment effects

    J. Econ. Perspect.

    (2014)
  • A.W. Dick et al.

    Consequences of states' policies for SCHIP disenrollment

    Health Care Financ. Rev.

    (2002)
  • A. Finkelstein et al.

    Take-up and targeting: experimental evidence from SNAP

    Q. J. Econ.

    (2019)
  • P. Ganong et al.

    The decline, rebound, and further rise in SNAP enrollment: disentangling business cycle fluctuations and policy changes

    Am. Econ. J. Econ. Pol.

    (2018)
  • C. Heflin et al.

    Assessing the Impact of a Modernized Application Process on Florida's Food Stamp Caseload

  • T. Homonoff et al.

    Program Recertification Costs: Evidence from SNAP

    (2019)
  • B. Meyer et al.

    The Analysis of Food Stamp Program Participation with Matched Administrative and Survey Data

    (2010)
There are more references available in the full text version of this article.

Cited by (32)

  • Understanding SNAP: An overview of recent research

    2023, Food Policy
    Citation Excerpt :

    Not only does churn interrupt benefit receipt for participants, but it also imposes substantial additional administrative costs on agencies. Gray (2019) investigates exit rates among SNAP participants using administrative data from several states. Across states in the study, 40 to 50 percent of SNAP participants stay on the program less than a year.

View all citing articles on Scopus

This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. 1122374. I performed much of this analysis as an employee of the W.E. Upjohn Institute for Employment Research. I am extremely grateful to Chris O’Leary and Ken Kline for making this project possible. I would also like to thank David Autor, Amy Finkelstein and Jon Gruber for their comments; Kathryn Law at USDA for access to supplemental data sources; and Jason Page at Michigan DHHS for generously sharing his institutional expertise.

View full text