Examining the Externalities of Welfare Reform: TANF and Crime

Abstract

This study explores the relationship between welfare policy variation in the United States following the introduction of Temporary Assistance for Needy Families (TANF) and its relationship with various types of crime. While early studies of the effects of welfare assistance on crime consistently found a negative association, more recent examinations have complicated these findings. Nearly all prior research focuses on Aid for Families with Dependent Children or early years of TANF. Examining a longer time-series and using propensity score weighting to model the tendencies of states to select into more stringent welfare regimes, we find a strong association between states with greater levels of welfare restrictiveness and higher rates of violent crime. There is mixed evidence that this relationship also exists with property crimes.

Keywords:

• property crime
• violent crime
• welfare
• TANF

Notes

¹ Oregon is an uncommon state that actually liberalized eligibility rules in 2008 in light of the economic crash.

² These states were Delaware, Georgia, Nebraska, Oklahoma, South Carolina, Tennessee, and Wyoming.

³ Average assistance levels for a family of three accounts for income equivalent to less than 50% of the poverty threshold.

⁴ Harsher sanctions for compliance failures affect a full-family (i.e. 100%); that is, all assistance is ended. Moderate sanctions denote a partial-family penalty; benefits are reduced but not terminated. Weak sanctions allow for several cases of noncompliance before enacting any penalty.

⁵ Models using the percentage of single-parent and female-led households did not yield substantively different results.

⁶ This variable could theoretically be endogenous to the relationship between welfare and crime (perhaps acting as a mediating variable), but they have a low correlation coefficient (0.09). Running the models excluding this variable produces similar results, which are available upon request.

⁷ To account for the fact that while TANF was leading to restrictions during the period under study, other types of social assistance were expanding, including child healthcare via the Children’s Health Insurance Program, income subsidies via the Earned Income Tax Credit, and food subsidies via the Supplemental Nutrition Assistance Program, we do robustness checks that include state level data on these variables as well, labeled SCHIP, EITC, and SNAP respectively. SCHIP data come from annual Medicaid Financial Management Reports, and EITC and SNAP data were accessed from the University of Kentucky Center for Poverty Research. Including these covariates does not substantively alter the findings. Results are available upon request (Table 2).

⁸ Random Effects models allow us to control for second order devolution of welfare in the individual states. These models do not produce meaningfully different results and are available upon request.

⁹ A test of covariate balance using the STATA tebalance command indicates variance ratios close to zero, with the exception of unemployment.

Additional information

Notes on contributors

Scott Liebertz

Scott Liebertz is an assistant professor of Political Science, Criminal Justice, and Public Administration at the University of South Alabama. His research interests include comparative criminal justice, public policy, and political behavior.

Jaclyn Bunch

Jaclyn Bunch is an assistant professor of Political Science and Public Administration at the University of South Alabama. Her research interests include state and local relations, federalism, hierarchical dilemmas, political media, and motivation and reasoning.