Data and Methods: Release Cohort

Recidivism rates are calculated based on a release cohort—for this study, a group of prison inmates—released during a specified time period. The release cohort used to measure recidivism rates may vary. For example, a recidivism rate may be calculated for inmates released during a fiscal year or for all inmates released since a certain date. Because recidivism rates of state prison inmates do not vary much by year of release, this report relies on the latter method, calculating recidivism rates for inmates released from Department of Corrections custody from July 1995 through June 2001.

Data was collected on 106,167 prison releases involving 96,415 inmates during the six year study period. This number of releases is lower than those reported elsewhere by the Department, mainly because some inmates are released more than once on the same prison commitment. For example, an inmate who leaves prison on conditional release—a form of post-release supervision—can return to prison for a violation of supervision conditions (other than committing a new offense) and later leave prison again. For recidivism analysis, only the first such release is counted. Counting more than one of these releases would artificially lower the recidivism rates. This is because an inmate released twice who only recidivated after the second release would be counted both as a success (non-recidivist) and a failure.

Of these 106,167 releases, 2,643 (2.8%) were excluded from the analysis because they involved death, release out of state or to another prison system, escape, or a vacated (court-removed) sentence. Including these releases would also improperly lower the recidivism rates. Inmates released due to death cannot recidivate, and those released out of state are unlikely to recidivate in Florida. Releases to other prison systems must also be excluded because the true follow-up time can not be measured since the Department has no data on when the inmate was subsequently released. Those released by court-vacated sentences and the few who escape are excluded because they are unusual releases.

This analysis retained 85.7% of the 103,524 remaining releases (94,272 inmates), including 88,678 releases (81,737 inmates) with no missing data on important inmate characteristics relevant to recidivism. Almost all cases excluded for missing data have no score on the Test for Adult Basic Education because they did not take the test. Excluding these cases does not lower recidivism rate estimates in this study because the excluded cases have rates somewhat lower than those retained.

The study analyzed 80,919 releases of 74,467 male inmates and 7,759 releases of 7,270 female inmates. This study analyzes male and female releases separately, due to differences in their fundamental recidivism rates and because factors that influence recidivism affect males and females differently. Table 1 shows the final male and female cohorts broken down by the number of releases during each fiscal year (July - June) in the study period.

For more information on recidivism rates based on fiscal year release cohorts, see Recidivism Rate Curves.

Table 2 describes the male and female release cohorts analyzed for this report. The characteristics presented are those found to predict recidivism rates best when used simultaneously. These characteristics were selected from 35 variables (counted in parentheses below) that cover 17 aspects of inmate characteristics in four general areas:

Demographic:	gender (1), age (1), race (1), ethnicity (1)
Post release status:	supervision (1), time not at risk (1)
Criminal History:	onset / length (3), density (1), dispersion - geographic (1), diversity - offense specialization (2), severity offense volume, type, degree (11), prior supervision failure (2), prior recidivism (1)
Prison Experience:	length of stay (1), behavior / confinement (5), custody level (1), education level (1).

Where possible, multiple ways of measuring each variable were tested. These measures are those that met three qualifications:

• correlated most with recidivism among multiple measures for each variable,
• correlated least with measures used for other variables, and
• survived appropriately in tests of a variety of statistical models of recidivism.

This study only analyzed characteristics for which the Department has data at the time an inmate is released from prison. This data is reliable, readily available, and should remain so over time. Data is either unavailable or inadequate to analyze post-release factors such as employment, drug and alcohol use, and housing stability, etc., which may also affect recidivism rates according to other research.

For more information about how these cohort characteristics were measured and selected, see the Technical Appendix.

Table 2. Characteristics of Inmate Releases
FACTOR	CATEGORY	MALES		FEMALES
		N	%	N	%
Gender		80,919	91.3*	7,759	8.7*
Age at Release	UNDER 18	347	0.4	22	0.3
	18 TO 24	20,046	24.8	930	12.0
	25 TO 34	29,954	37.0	3,240	41.8
	35 TO 49	27,334	33.8	3,336	43.0
	50 TO 59	2,620	3.2	193	2.5
	60+	618	0.8	38	0.5
	Mean / Std. Dev.	32.1 / 9.2		33.9 / 7.9
	Max / Min	91 / 15		79 / 15
BLACK		47,343	58.5	4,449	57.3
HISPANIC		4,469	5.5	240	3.1
Supervision - Any After Release		29,982	37.1	1,998	25.8
Months in Prison	1-12	20,323	25.1	3,005	38.7
	13-24	24,134	29.8	2,606	33.6
	25-36	13,965	17.3	1,100	14.2
	37-48	8,319	10.3	490	6.3
	49-60	4,837	6.0	244	3.1
	61+	9,341	11.5	314	4.0
	Mean / Std. Dev.	30.4 / 26.6		21.2 / 18.9
	Max / Min	370 / 1		305 / 1
Disciplinary Reports - Total	0	36,396	45.0	4,302	55.4
	1	13,491	16.7	1,294	16.7
	2	7,218	8.9	586	7.6
	3-6	12,059	14.9	861	11.1
	7+	11,755	14.5	716	9.2
	Mean / Std. Dev.	3.2 / 6.9		2.3 / 6.4
	Max / Min	208 / 0		113 / 0
Custody Level at Release	COMMUNITY	2,650	3.3	337	4.3
	MINIMUM	38,102	47.1	5,423	69.9
	MEDIUM	26,719	33.0	1,705	22.0
	CLOSE	13,448	16.6	294	3.8
Test for Adult Basic Education Grade Level - last before release	1-3.9	14,438	17.8	1,356	17.5
	4-8.9	42,364	52.4	4,406	56.8
	9-11.9	13,684	16.9	1,135	14.6
	12-12.9	10,433	12.9	862	11.1
	Mean / Std. Dev.	7.3 / 3.2		7.1 / 3.1
	Max / Min	12.9 / 1		12.9 / 1
Prior Recidivism Events	0	41,631	51.4	4,606	59.4
	1	17,291	21.4	1,599	20.6
	2	10,459	12.9	856	11.0
	3	6,228	7.7	444	5.7
	4	3,176	3.9	171	2.2
	5+	2,134	2.6	83	1.1
	Mean / Std. Dev.	1.0 / 1.4		0.7 / 1.1
	Max / Min	11 / 0		7 / 0
Most Serious Offense Type- Career	HOMICIDE	3,935	4.9	349	4.5
	SEX/LEWDNESS	6,216	7.7	74	1.0
	ROBBERY	17,227	21.3	906	11.7
	OTHER VIOLENT	20,265	25.0	2,081	26.8
	BURGLARY	17,322	21.4	882	11.4
	PROPERTY	7,043	8.7	1,882	24.3
	DRUGS	8,059	10.0	1,513	19.5
	WEAPONS	338	0.4	17	0.2
	OTHER	514	0.6	55	0.7
Property Offenses - Total	0	39,925	49.3	3,727	48.0
	1	14,004	17.3	1,138	14.7
	2	9,303	11.5	782	10.1
	3+	17,687	21.9	2,112	27.2
	Mean / Std. Dev.	1.7 / 3.0		2.4 / 5.2
	Max / Min	111 / 0		120 / 0
Drug Offenses - Total	0	41,163	50.9	3,103	40.0
	1	11,686	14.4	1,085	14.0
	2	8,822	10.9	954	12.3
	3+	19,248	23.8	2,617	33.7
	Mean / Std. Dev.	1.6 / 2.5		2.2 / 2.9
	Max / Min	35 / 0		41 / 0
Weapons Offenses - Total	0	69,543	85.9	7,304	94.1
	1	7,879	9.7	359	4.6
	2+	3,497	4.3	96	1.2
	Mean / Std. Dev.	0.2 / 0.6		0.1 / 0.3
	Max / Min	8 / 0		4 / 0

These characteristics of the release cohorts are factors that influence recidivism rates: either raising or lowering an inmate's risk of recidivism. For information about these factors, see Statistical Analysis; and for how they affect recidivism rates, see Factors Affecting Rates.