Control Variables

The variables used in this study to establish equivalency between public and private prison comparison groups through statistical controls are ones the FDOC identified, from an extensive list of over 100 initial variables, as having meaningful and unique influences on the likelihood of recidivism. The 17 control variables selected were captured directly from FDOC’s OBIS or were created from these data.⁸ By controlling for this larger set of statistically independent variables that influence recidivism, the present study improves substantially over Lanza-Kaduce et al. (1999) and moderately over Farabee and Knight (2002).⁹

Control variables used in this study met two criteria: 1) factors for which FDOC has reliable data, and 2) factors found to be significant recidivism predictors in prior studies. They include both individual and institutional characteristics and time factors. At the individual level, characteristics include race,¹⁰ gender,¹¹ age,¹² and the inmate’s level of education.¹³ Other factors found to be associated with recidivism include offense history,¹⁴ inmate custody level at the time of release,¹⁵ disciplinary history,¹⁶ prior recidivism,¹⁷ and post-release supervision.¹⁸ Time factors include the amount of time served in prison¹⁹ and the amount of time since release. However, time since release is not used as a recidivism predictor in this analysis because it forms the recidivism measures used.

The control variables and how they were created are as follows:
Age at Release – the age of the inmate at prison release in years based on the offender’s date of birth and release date.
Race – dichotomized as Black (1) or Non-Black (0).
Ethnicity – dichotomized as Hispanic (1) or Non-Hispanic (0).
Prior Recidivism – the number of times an inmate has been released from Florida’s prisons in the past and subsequently committed a new offense leading to a state prison or supervision commitment, measured by comparing release dates and offense dates.
Custody Level – the level of custody assigned to the inmate at the time of their prison release (close, medium, minimum, and community). Custody level is based on several measures relating to the inmate’s current and prior offenses, sentence length, and institutional conduct yet appears to have some effect on recidivism independent of offense history, length of prison stay, and disciplinary reports received. The FDOC’s analysis groups custody levels for recidivism analysis: High – if they were close custody at release (0=No, 1=Yes); and Low – if they were community or minimum custody at release (0=No, 1=Yes).
Months in Prison – the number of months between prison admission and release dates.
Total Disciplinary Reports – The number of disciplinary actions against inmates that occurred between their admission and release.
Last TABE Grade –The last total battery TABE score received prior to prison release was used in this study.
Supervision – This variable originates from the FDOC supervision movement data, where inmates were admitted to community supervision about the time they were released from prison. This was quantified as a dichotomous variable (0 = No post-prison supervision, 1 = Post-prison supervision).
Criminal History – Three control variables quantify certain offenses in an inmate’s history. These count the total number of convicted offenses in the inmate’s record within each category: drug, property, and weapons offenses. Five control variables classify an inmate’s offense history: if the most serious prior convicted offense was for a homicide, sex/lewdness offense, robbery, burglary, or other violent offense (five dichotomous variables with 0 = not most serious type, 1 = most serious type).

8. FDOC, 2003.
   
   
9. Farabee and Knight also relied on FDOC data for their analysis. However, since that research (FDOC, 2001), the department has expanded its recidivism predictors, including more complex measures of offense history and new measures of ethnicity and post-release supervision (FDOC, 2003).
   
   
10. Anderson et al., 1991; Beck and Shipley, 1989; FDOC, 2003; Harer, 1995a; Langan and Levin, 2002; Maguire et al., 1988; WSDOC, 2002.
    
    
11. Beck and Shipley, 1989; FDOC, 2003; Kim et al., 1993; Langan and Levin, 2002; WSDOC, 2002.
    
    
12. Batiuk et al., 1997; Beck and Shipley, 1989; FDOC, 2003; Harer, 1995a; Kim et al., 1993; Langan and Levin, 2002; Smith and Polsenburg, 1992; Uggen, 2000; WSDOC, 2002.
    
    
13. Beck and Shipley, 1989; FDOC, 2003; Harer, 1995b; Maguire et al., 1988; Ulmer, 2001.
    
    
14. Beck and Shipley, 1989; Harer, 1995a; Langan and Levin, 2002; Maguire et al., 1988.
    
    
15. Farabee and Knight, 2002; FDOC, 2003.
    
    
16. FDOC, 2003; Harer, 1995a; Maguire et al., 1988.
    
    
17. FDOC, 2003; Harer, 1995a; Schmidt and Witte, 1988; Smith and Polsenberg, 1992; Uggen, 2000.
    
    
18. FDOC, 2003; Harer, 1995a; Jernigan and Krosnick, 1992.
    
    
19. FDOC, 2003; Maguire et al., 1988; WSDOC, 2002.