The scourge of gun violence in our streets, schools, places of worship, workplaces, and entertainment venues around the nation has created a sense of urgency to find prevention and intervention strategies. Research is scarce, however, in part due to decades-long Congressional limits on federal funding to support U.S. Center for Disease Control and Prevention research on firearm-related topics. 
This study was conducted to demonstrate the usefulness of state criminal history records for examining recidivism of specific criminal justice populations, in this case, gun offenders. Besides gathering information on repeat offending through criminal history and prison records, Illinois Criminal Justice Information Authority (ICJIA) researchers obtained state death records of deceased individuals in the study sample. These records provided detail on the cause and manner of death not available in criminal justice administrative data. Together, these findings offer relevant insights into first-time firearm-involved arrestees, their recidivism patterns and mortality rates, and inform policy and practice on the issue of guns and violence.
This research project, funded by the U.S. Department of Justice Bureau of Justice Statistics through a cooperative agreement under the State Justice Statistics grant program, was designed to examine 10-year recidivism rates of firearm offenders using state criminal history records. For more precise measurement of recidivism, ICJIA researchers determined periods of stay in state prison and deaths of individuals in the study, to account for time away from the community during which time individuals did not have the opportunity to re-offend. Subtracting this time from the total follow-up period provided a more accurate picture of how long it took individuals to recidivate. Recidivism was defined in three ways: re-arrest, re-conviction, and re-incarceration following the initial arrest identified in the study.
The study sample was constructed from a pool of 379,275 individuals whose arrests were recorded in the state’s Criminal History Record Information (CHRI) System in 2003. The study design called for identifying individuals arrested for the first time for firearm-related crimes and matching their records to those also arrested in that same year, but never for a firearm-related offense. Every kind firearm-related crime was included, from using a gun in the commission of a violent crime (36 percent of the firearm charges), to the illegal possession, purchase or selling of a firearm (64 percent).
Using Coarsened Exact Matching techniques , 4,323 first-time firearm-involved arrestees were matched on 16 demographic, geographic (county), criminal charge and prior criminal history criteria to 4,323 non-firearm-involved arrestees. This matching ensured that any differences in recidivism could be more confidently attributed to the presence or absence of firearm involvement. Another 1,774 first-time firearm-involved offenders remained without a suitable matching counterpart, as they were more serious offenders than the remaining pool of non-firearm-involved arrestees. This smaller unmatched firearm-involved group was kept in the study to provide additional information on recidivism patterns of persons arrested for the first time for firearm-related charges. The total study sample included 10,420 individuals.
The characteristics of the matched Firearm-Involved group essentially defined the characteristics of the overall study sample, since their attributes were used as the match criteria. Compared to both the Illinois general population and all persons with arrests recorded in 2003, the study sample was highly over-represented by males (90 percent), Blacks (54 percent), arrests from Cook County (56 percent), and teens and young adults, ages 15-24 (56 percent). The study sample also had fewer adults over the age of 35 (19 percent) compared to either the 2003 population of arrested persons or the general Illinois population.
The arrest charges of the matched Firearm-Involved group defined the arrest charge characteristics of the entire study sample, as well. For example, the study sample had a much higher concentration of felony arrests (71 percent) than the larger pool of 2003 arrests from which the sample was drawn (15 percent), and considerably fewer property charges (7 percent) than the larger pool of 2003 arrests (26 percent).
Outcome of the initial 2003 arrest event
Overall, 43 percent of the 10,420 individuals in the study were convicted of any charge related to their 2003 arrest. Proportionally more of the matched Firearm-Involved group were convicted for their firearm-related arrest (48 percent) than the matched Non-Firearm-Involved group for their non-firearm related arrest (33 percent). It is important to note that at least 20 percent of the court disposition records were missing from CHRI, so the percentage of convicted individuals could be higher. A higher proportion of individuals in the smaller Unmatched Firearm-Involved group, who were more serious offenders, were convicted of a firearm-related charge than those in the matched Firearm-Involved group (43 percent compared to 34 percent, respectively).
The research sample was tracked over the next 10 years through their CHRI System records and Illinois Department of Corrections (IDOC) prison files. Recidivism rates were determined for any offense type, and separately, for firearm offenses.
Overall, 59 percent of the entire study sample was arrested again at least once during the 10-year period studied. Proportionally more of the matched Firearm-Involved group were re-arrested for any charge (67 percent) than the matched Non-Firearm-Involved group (41 percent). The Unmatched Firearm-Involved group was re-arrested at the highest rate (81 percent).
Of the 6,112 individuals (in all groups) re-arrested at least once, 14 percent were re-arrested for another firearm-related charge. The two Firearm-Involved groups were rearrested at about the same rate - 18 percent for the matched Firearm-Involved group, and 21 percent for the Unmatched Firearm-Involved group. Of these new firearm arrests, 45 percent were for the commission of a violent offense, an increase from the 36 percent recorded for the initial 2003 arrest.
A small number (3 percent) of the matched Non-Firearm-Involved group were later arrested for a first firearm arrest over the next 10 years, including about half for violent use of a gun. Approximately 97 percent of the matched Non-Firearm-Involved group remained gun arrest-free over the next decade.
Of the 4,476 individuals convicted for their 2003 arrest, 39 percent were convicted of another offense during the next 10 years. More of the matched Firearm-Involved group were re-convicted for any charge (43 percent) than the matched Non-Firearm-Involved group (20 percent). Of the Unmatched Firearm-Involved group, 57 percent were re-convicted at least once during the follow-up period. These rates are likely undercounts, due to missing court disposition information in the CHRI System data.
**Of the 863 individuals initially convicted of a firearm charge, 77 percent were convicted again for a new firearm offense. More of the matched Firearm-Involved group (74 percent) were re-convicted for a firearm-related offense than the few (n=46) matched Non-Firearm-Involved individuals arrested for a firearm offense for the first time during the follow-up period (65 percent). Proportionally more of the Unmatched Firearm-Involved group were re-convicted (83 percent) for a firearm-related offense than the two matched groups.
Re-incarceration in state prison
Few individuals in the sample (15 percent) were sentenced to state prison upon conviction for their 2003 arrest. Of these 1,552 incarcerated individuals, almost half (46 percent) were sentenced to prison again for a new crime during the follow-up period. The rank ordering of re-incarceration by group was consistent with the other two recidivism measures. Proportionally more of the matched Firearm-Involved group was re-incarcerated (53 percent) than the matched Non-Firearm-Involved group (43 percent). Fifty-nine percent of the Unmatched Firearm-Involved group was re-incarcerated.
It was not possible to accurately determine the re-incarceration rates for solely firearm charges, as the prison files from the Illinois Department of Corrections contain only the most serious charge in each case; any lesser-included firearm charges would be missed in the analysis.
Summary of recidivism findings
For every measure of recidivism - re-arrest, re-conviction, re-incarceration - the matched Firearm-Involved group recidivated at a higher rate than the matched Non-Firearm-Involved group. The more seriously criminal justice-involved Unmatched Firearm-Involved group recidivated at the highest rate on all measures, for all types of charges, including firearm-related charges.
Figure 1 shows the recidivism rates of the three groups on each measure, over the 10-year follow-up period. The largest difference in recidivism rates between the two matched groups was the degree to which they were ever arrested again for a new charge after the initial arrest incident in 2003.
Summary of recidivism rates within 10 years, by recidivism measure and group membership
Time to first re-arrest
The risk of a first re-arrest was determined using survival analysis techniques for the matched Firearm-Involved and Non-Firearm-Involved groups only. Time-to-event was defined as days from the 2003 arrest date to the end of the observation period (3,653 days, or 10 year). Here, times of incapacitation due to a prison stay and death were taken into account in the statistical models constructed for the analysis. No individual stayed in prison during the entire length of the 10-year follow-up period.
As anticipated, adding both measures of time away from the community reduced the time to the first re-arrest from an overall median of 5 years (1,852 days) to 3.6 years (1,320 days). That is, subtracting time during which the individual was not at risk of re-offending provided a more accurate picture of the pace at which a re-arrest occurred after the initial 2003 arrest incident.
Statistical tests on the resulting survival curves confirmed significant differences between the matched Firearm-Involved and Non-Firearm-Involved groups for both first re-arrest for any charge and first re-arrest for a firearm-related charge only. For re-arrests for any type of charge, the time to first re-arrest occurred at the same rate for both groups during the first year after the 2003 arrest incident, but continued at a faster pace for the Firearm-Involved group in the second and subsequent follow-up years. The first re-arrest for a firearm-related charge for a few individuals in the Non-Firearm-Involved group (n=46) occurred primarily during the first year after the initial 2003 arrest.
Factors associated with risk of re-arrest
Statistical models were also fit to determine which, if any, of the demographic, geographic, and arrest charge variables available in the 2003 arrest records were predictive of a subsequent re-arrest for any charge and separately, for a firearm charge. For first re-arrests for any charge, the results indicated a 70 percent greater risk of re-arrest for the Firearm-Involved group than the Non-Firearm-Involved group. Being initially firearm-involved was such a strong factor predictive factor of re-arrest that other variables, such as being male, Black, and younger at the initial arrest added little explanatory power for the different rates of recidivism between the two groups.
The effect of group membership (Firearm-Involved vs. Non-Firearm-Involved) was even more evident when first re-arrest for a firearm charge was considered. The daily risk of a first re-arrest being for a firearm-related charge was 800 percent higher for the Firearm-Involved group compared to the Non-Firearm-Involved group. Further, being male, Black, and younger at the time of initial arrest were positively correlated factors with the daily risk of a first re-arrest for a firearm charge.
Interestingly, the place of initial arrest (Cook County vs. rest of the state) was not a statistically significant explanatory variable for the daily risk of re-arrest, for either a firearm arrest charge or any arrest charge generally, even though the study sample included proportionally more arrests from Cook County than all arrests recorded in the CHRI System in 2003.
Death certificate records indicated that 448 individuals died during the study period, or 4 percent of the total study sample. These deceased individuals represented each of the three groups in equal proportions (an average of 4 percent of each group). Death records provided information on the cause and manner of death, information unavailable in the other criminal justice administrative datasets.
Causes of death are classified as natural (disease) and injury (non-natural). In Illinois from 2003 to 2013, unintentional accidents (including motor vehicle accidents, drug overdoses, and falls) accounted for 68 percent of all injury-related deaths, suicides accounted for 19 percent, and homicides accounted for 12 percent.
Individuals in the two Firearm-Involved groups died at a higher rate from homicide (41 percent combined) - almost exclusively by gunshot wound - than the Non-Firearm-Involved group (21 percent). Conversely, proportionally more individuals in the Non-Firearm-Involved group died from natural causes (48 percent) than the two Firearm-Involved groups combined (34 percent) (Figure 2).
Proportion of homicide deaths during the 10-year follow-up period, firearm-involved groups compared to non-firearm-involved
The most common age range at time of death by homicide was 15 to 24. This age group accounted for 61 percent of all deaths by homicide, followed by 25 to 34 year-olds (34 percent). In Illinois from 2003 to 2013, motor vehicle accidents were the leading cause of injury-related death among 15 to 24-year-olds, with firearm-related homicides ranked second. In this study of criminal justice-involved individuals, the reverse was true. Of the 15 to 24-year-olds in all three groups who died, 73 percent were killed in firearm-related homicides, and 22 percent were involved in motor vehicle accidents.
Suicide was the least common cause of death in every group, accounting for fewer than 10 deaths per group (7 percent of all deaths). However, a firearm was involved in 80 percent of the suicides in the Unmatched Firearm-Involved group, 50 percent of the suicides in the matched Firearm-Involved group, and 36 percent of the Non-Firearm-Involved group.
Implications for policy and practice
The results of this study have implications for policy and practice regarding firearm-related crime and its consequences, as well as recommendations for further research.
Consider first-time firearm-related arrests as important predictors of recidivism and mortality risk of those arrested
During the period studied, first arrests for gun-related offenses were highly predictive of future arrests, especially new firearm arrests. Further, this group of arrestees was at greater risk for homicidal death from a firearm than other criminal-justice involved individuals not arrested for firearm-related charges. A better understanding of the risk factors associated with initial gun involvement is needed to develop the most appropriate and effective intervention and prevention strategies. This study could not measure motivational factors for either firearm involvement or avoidance of firearms, except to document that these behaviors persisted over years. In the short term, those dealing with firearm-involved individuals should be aware that even minor initial illegal firearm involvement can signal risk of serious long-term consequences for both public safety and the involved individual, and should not be minimized as a risk factor.
Learn from the persistently non-firearm-involved individuals
This study used statistical techniques that matched firearm-involved and non-firearm-involved individuals on a one-to-one basis using 16 demographic, geographic (county), arrest charge type, and prior criminal criteria. A surprising finding was the stability of the non-firearm group membership over time. That is, few, if any, of this group went on to be arrested for firearm-related charges over the next 10 years, although some became victims themselves of fatal firearm violence. This group was composed of an equal number of individuals with characteristics significantly associated with firearm recidivism (being male, Black, and at a young age at the time of arrest) as the matched firearm-involved group, yet remained not involved with firearms for subsequent re-arrests. While county-level geography (Cook County vs. rest of the state) was not found to be related to recidivism in this study, other community factors not measured here, such as differences in gun availability, and gang or drug market presence, may help to explain this finding. The non-gun offenders also could possess more protective factors not measured in this study, such as better employment opportunities or access to social services. Further study of this group could produce valuable information for the development of new prevention strategies.
Apply a multidisciplinary approach to the state’s social problems by leveraging agency collaboration and the capacity to match criminal justice administrative data with other data sources.
In this study, researchers combined criminal history and prison records with public health data to provide a unique perspective on firearm recidivism, an issue relevant to both criminal justice and public health systems. Success was achieved not only from a technical record matching standpoint, but from the broader perspective of a collaborative approach among the state agencies involved in this research.
Often, other agencies are not aware of the impact that criminal justice-involved sub-populations have on their specific areas of responsibility, and have no way to measure or evaluate this impact. For example, it could not be known how many of the study subjects were wounded by gun shots and hospitalized, only the number that did not survive. If hospitals and health care providers could learn of the likelihood of gunshot wound patients’ future criminal justice-related incapacitation, such as imprisonment, that information could assist with more effective coordination of follow-up care, and expand knowledge about persons involved with firearms.
Success in understanding and tailoring solutions to important social issues will occur at a faster pace when policymakers bring multidisciplinary perspectives to data gathering and sharing initiatives. Currently, the state is planning the development and implementation of the new Incident-based Reporting / Uniform Crime Reporting (IBR/UCR) program, which will allow law enforcement agencies to report contextual information on each crime incident. Incorporation of information on the outcome of the incident, such as criminal justice identifiers of persons arrested for the crime, or the location of the hospital to which a victim was transported, will provide authorized users with enhanced capability to match records across systems, overcoming administrative and technical barriers to information sharing, for statewide benefit.
This study successfully combined CHRI records with other criminal justice and public health administrative records to improve knowledge about gun offender recidivism patterns and more precisely estimate recidivism rates. It was possible to demonstrate that, holding other characteristics constant, those facing the criminal justice system for the first time as firearm offenders persisted in criminal justice involvement for firearms at a much higher rate and for a longer period than their justice system-involved peers who were not engaged with firearms.
Finally, and most tragically, the mortality records showed that firearm-involved arrestees were themselves killed twice as often as non-firearm involved arrestees, most often because of a firearm injury. These homicide victims were predominately within the age range of 15 to 24-years old, a finding that urgently calls for implementation of prevention and intervention strategies tailored specifically to this age group to avert this premature loss of life.
Erin Sheridan, Illinois Student Assistance Commission, and John Specker, Sentencing Policy Advisory Council, contributed to this study.
Suggested citation: Devitt Westley, C., Kang, B., Sheridan, E., & Specker, J. (2018). Examining the recidivism of firearm offenders using state criminal history and mortality data. Chicago, IL: Illinois Criminal Justice Information Authority.
Sofer, D. (2017). Gun violence and children. American Journal of Nursing. 117(9), 14. ↩︎
Iacus, S., King, G., & Porro, G. (2012). Causal inference without balance checking: coarsened exact matching. Political analysis. 20(1), 1-24. ↩︎
Wintermute, G. (2015). The Epidemiology of Firearm Violence in the Twenty-First Century United States. Annual Review of Public Health. 36, 5-19. ↩︎
Uniform Crime Reporting Program, National Incident-Based Reporting System. (2011). A Guide to Understanding NIBRS. Washington, D.C. United States Bureau of Justice, Federal Bureau of Investigation. Retrieved from: https://ucr.fbi.gov/nibrs/2011/resources/a-guide-to-understanding-nibrs ↩︎