TopicsTopics

Stay Connected

Follow ERS on Twitter
Subscribe to RSS feeds
Subscribe to ERS e-Newsletters.aspx
Listen to ERS podcasts
Read ERS blogs at USDA

Food and Nutrition Assistance Research Database

The RIDGE Program summarizes research findings of projects that were awarded 1-year grants through its partner institutions. All projects were conducted under research grants from ERS, and the views expressed are those of the authors and not necessarily those of ERS or USDA. For more information about publications or other project outputs for a specific RIDGE study, contact the investigator or research center that awarded the grant. For a customized list of RIDGE projects and summaries, search by keyword(s), project, research center, investigator, or year:

Project:
The Dynamics of SNAP Participation and the Increase in SNAP Caseloads during the Recovery of 2003–2007

Year: 2011

Research Center: Institute for Research on Poverty, University of Wisconsin-Madison

Investigator: Johnson, Janna

Institution: Harris Graduate School of Public Policy Studies, University of Chicago

Project Contact:
Janna Johnson
Harris School of Public Policy
University of Chicago
1155 E 60th St
Chicago, IL 60637
Phone: 218-929-7497
E-mail: jannaj@uchicago.edu

Summary:

The recent sharp rise in Supplemental Nutrition Assistance Program (SNAP) participants has received much attention in the press and from policymakers. Since the start of the Great Recession in December 2007, SNAP participation has increased to its highest level ever, serving 40.3 million Americans each month in fiscal year 2010, more than 13 percent of the population. Less attention has been given to the fact that SNAP participation also increased during the preceding economic expansion. Between fiscal years 2003 and 2007, total SNAP participation increased from 21 million to 27 million, an increase of almost 30 percent. This rise marked the first time in the program’s history that participation increased during a period of economic recovery and growth. Fiscal years run from October 1 to September 30 of the year named.

The fact that participation in SNAP behaved contrary to expectations based on the economic conditions during this time period has prompted others to seek alternative explanations for the surprising increase, using State-level data. This study looks for the cause of the increase in SNAP participation at its underlying source: the determinants of the participation decision at the individual level, including the dynamics of SNAP entry and exit, using panel data from the Survey of Income and Program Participation (SIPP). While others have studied the dynamics of program entry and exit, this study is the first to analyze them during the economic recovery of 2003 to 2007, as well as the more recent Great Recession.

The study is divided into two sections. The first conducts a descriptive analysis of the dynamics of SNAP participation during the first decade of the 2000s to find the cause for the increase in SNAP participation during the recovery of 2003 to 2007, focusing on the population eligibility rate for the program, the entry rate into SNAP, and the rate at which current participants exit the program. The second section is an empirical hazard analysis of the determinants of the SNAP exit rate. Both sections use restricted-use data from the 2001, 2004, and 2008 panels of the SIPP, which contain more detailed geography information than the public-use SIPP.

Results based on descriptive analysis of SNAP participation spells using SIPP data indicate that a fall in the rate at which participants left the program was likely the primary cause of the increase in SNAP participation during the economic recovery period of 2003 to 2007. Over this period, the entry rate into SNAP as well as the proportion of the population eligible for the program did not significantly change. The Great Recession, in contrast, saw increases in the entry rate and eligibility rate but no further change in the SNAP exit rate.

Comparing Kaplan-Meier hazard functions across the 2001 recession, the 2003 to 2007 recovery, and the Great Recession shows that the fall in the exit rate during the recovery period occurred primarily among short fresh spells (less than 12 months in length) and long left-censored spells (those lasting longer than 29 months). The hazard functions, which control for spell duration in contrast to the overall exit rate calculated in the descriptive analysis, also show that exit rates continued to decline among all spell lengths and types during the Great Recession.

A basic hazard regression investigation into the causes of the exit-rate decline provided inconclusive results. While many SNAP policy changes—such as expanded simplified reporting, looser reporting requirements, and expanded categorical eligibility—were predicted to have a negative impact on the exit rate, the study revealed that they had a negative and statistically significant relationship with the exit rate for the entire time period (2001 to 2010). Once the three economic time periods were examined separately, the relationships were not as strong, nor always of the expected sign. Results using improved policy data from an outside source were largely consistent, although not always, with those using the author-constructed Food Stamp Program State Rules (FSPSR)/SNAP State Options Report (SOR) dataset, but variables indicating the adoption of different vehicle and electronic application policies also had significant coefficients. These policies were not expected to have an impact on the exit rate, and therefore these results do not bode well for the results for the other policies. The estimated correlations could be appearing because they are in fact true causal impacts of the policies on the exit rate, or due to some other unobserved mechanism affecting the exit rate at the same time these policies were introduced.

Therefore, the hazard regression results do not provide evidence one way or another that the SNAP policy changes that occurred during the 2003 to 2007 recovery were responsible for the decline in the exit rate. The explanation must lie elsewhere, or the policy variables currently used are not accurately measuring true policy implementation. Future research is required to provide a definitive answer to this question.

Last updated: Monday, August 18, 2014

For more information contact: Alex Majchrowicz