Medicaid Expansion’s Spillover to the Criminal Justice System: Evidence from Six Urban Counties

Carrie E. Fry; Thomas G. McGuire; Richard G. Frank

doi:10.7758/RSF.2020.6.2.11

Article Figures & Data

Figures

Tables
Additional Files

Download figure
Open in new tab
Download powerpoint
Figure 1.
Study Counties
Source: Authors’ analysis of county-level booking data.
Note: Medicaid expansion counties are represented in the lighter shade; non-expansion counties are in the darker shade. Counties were chosen to provide geographic diversity across the United States and based on demographic similarity. Counties also had to have adequate volume in the county jail to provide adequate power for the analyses and had to be able to track recidivist outcomes for the entire study period. Counties in the Midwest are Hennepin County, Minnesota (expansion), and Dane County, Wisconsin (non-expansion). Counties in the Southwest are Pima County, Arizona (expansion), and El Paso County, Texas (non-expansion). Counties in the Southeast are East Baton Rouge Parish, Louisiana (expansion), and Hinds County, Mississippi (non-expansion).
Download figure
Open in new tab
Download powerpoint
Figure 2.
Probability of Arrest in Each County over the Study Period
Source: Authors’ analysis of arrest data from six county jails.
Note: In the Midwest and Southwest, the pre-period observation window is July 1, 2012, through September 30, 2013. In the Southeast, the pre-period observation window is January 1, 2015, through March 31, 2016. In the Midwest and Southwest, the post-period observation window is July 1, 2014, through December 31, 2015. In the Southeast, the post-period observation window is January 1, 2017, through June 30, 2018. Medicaid expansion in the Midwest and Southwest county pairs occurred on January 1, 2014. Medicaid expansion in the Southeast county pair occurred on June 1, 2016.
Download figure
Open in new tab
Download powerpoint
Figure 3.
Changes in the Probability of Rearrest Between Medicaid Expansion and Non-Expansion Counties
Source: Authors’ analyses of arrest data from county jails.
Note: The study period for the Midwest and Southwest is from July 1, 2012, through December 31, 2015. The pre-expansion period for the Southeast is from January 1, 2015, through June 30, 2018. Observations are at the person-month level. Estimates are from comparative interrupted time series regressions. Regressions are linear probability models. Each full sample regression is adjusted with gender and prior contact with the criminal justice system. The Midwest pair also adjusts for whether the arrest was a felony or misdemeanor and the interaction of this variable with the monthly counter. The Southwest county pair also adjusts for whether the arrest was for a parole violation and for whether the arrestee was Hispanic-Latino plus the interactions of these two variables with the monthly counter. Regressions using the Southeast county pair also adjust for whether the arrestee was African American and the interaction of this variable with the monthly time trend. Stratified regression analyses in each county pair adjust for these same covariates except for the variable that the sample was stratified on.
Download figure
Open in new tab
Download powerpoint
Figure 4.
Changes in the Number of Rearrests Between Medicaid Expansion and Non-Expansion Counties
Source: Authors’ analyses of arrest data from county jails.
Note: The pre-expansion period for the Midwest and Southwest is from July 1, 2012, through December 31, 2015. The pre-expansion period for the Southeast is from January 1, 2015, through June 30, 2018. Observations are at the person-month level. Estimates are from comparative interrupted time series regressions. Each full sample regression is adjusted with gender and prior contact with the criminal justice system. The Midwest pair also adjusts for whether the arrest was a felony or misdemeanor and the interaction of this variable with the monthly counter. The Southwest county pair also adjusts for whether the arrest was for a parole violation and for whether the arrestee was Hispanic-Latino plus the interactions of these two variables with the monthly counter. Regressions using the Southeast county pair also adjust for whether the arrestee was African-American and the interaction of this variable with the monthly time trend. Stratified regression analyses in each county pair adjust for these same covariates except for the variable that the sample was stratified on.