This study involved an examination of the relative effects of measures of inmate characteristics, features of facility environments, and managerial practices, including the perceived legitimacy of the correctional staff, on both the prevalence and incidence of violent, drug, and other nonviolent misconduct.
Inmate, officer, and facility data were collected from each of the 33 facilities for adults in Ohio and each of the 13 state operated facilities for adults in Kentucky.
The inmate data (Inmate Level Data, n=5,640) includes information on inmates collected through on-site surveys, and then supplemented with information from administrative records. The methods of administrating the survey varied to some degree across facilities. For most facilities, inmates were passed to designated locations where they were surveyed in groups that differed in size based on the custody level of the inmate populations. For security reasons, some facilities required the surveys to be administrated to inmates in their cells or housing areas. Inmates were not surveyed in areas where surveillance cameras were in operation and precautions were taken to ensure the confidentiality of the inmates' responses.
The officer data (Correctional Officer Data, n=1,841) includes information on correctional officers collected through a mail survey, and then supplemented with information retrieved from administrative records. The administration of the officer survey differed somewhat between states. For both states, an envelope containing a survey, a letter explaining the study, consent to participate, and a postage paid return envelope were placed in each of the sampled officers' employee mail on the date the inmate survey was administered at their facility. In Ohio, this initial survey was followed by a post card one week later and by two follow-up surveys at three and seven weeks after the initial survey was delivered. At the request of the Kentucky Department of Corrections, only one follow-up survey was sent to their officers. Several weeks prior to the administration of the follow-up survey a reminder announcement was made during roll call encouraging officers who were so inclined to complete the survey.
Facility data (Facility Level Data, n=46) were derived from aggregating responses to the inmate survey, from administrative records and from semi-structured interviews conducted with the wardens of each facility.
The inmate (Inmate Level Data, n=5,640) sampling frames were provided by administrative staff and included all the inmates housed within each facility. After the inmates who had served less than six months were removed, the sampling frames for the 11 facilities selected for the longitudinal data collection were further restricted to only those inmates who had at least six months of their sentence remaining at the time of the first survey. Next, the remaining inmates in each facility were stratified by whether they had previously served time. Equal numbers of inmates were then randomly selected from each stratum. These procedures resulted in total sample sizes of 5,094 inmates across the Ohio facilities and 2,200 inmates across the Kentucky facilities. Some inmates were not available on the day of the survey, which reduced the sample sizes to 4,931 inmates across Ohio facilities and 2,066 inmates across Kentucky facilities.
Different procedures were used to select the samples of correctional officers (Correctional Officer Data, n=1,841) in each state. In Ohio, officers were randomly selected from facility-specific lists of all the officers who have direct contact with inmates on a day-to-day basis (correctional officers and sergeants). Sample sizes for each facility were determined by using probabilities proportionate to size. For each facility in Kentucky, 100 officers were randomly selected from the list of all correctional officers and sergeants. For the facilities that did not employ at least 100 officers, all officers were selected.
The facility data (Facility Level Data, n=46) collected information from the prison wardens for each of the 33 facilities for adults in Ohio and each of the 13 state operated facilities for adults in Kentucky.
The inmate level data includes a weight variable, V202. This sample weight was created to adjust for the differences in the odds of selecting inmates based on the strata and between-facility differences in inmate population size. The sample weight reflected the inverse of each inmate's odds of selection.
The officer level data includes a weight variable, V141. This sample weight was created to adjust for the differences in the odds of selecting officers based on the differences in the sampling procedures and on between-facility differences in officer population size and reflected the inverse of each officer's odds of selection.
Mode of Data Collection:
Description of Variables:
The inmate data (Inmate Level Data, n=5,640 202 variables) includes demographic variables ( age, sex, race, marital status, level of education, employment, number of children), prior times in prison, illegal drug use, type of unit and current security classification, number of hours spent in vocational training, other education programs, working, recreation, number of visitors and time in the infirmary, interactions with the correctional officers, incidents in the facility (physical altercations, drug use, disciplinary infractions), and how the incidents were handled by the officers.
The officer data (Correctional Officer Data, n=1,841 141 variables) includes variables on how long the correctional officer has worked for the state's department of corrections, for the specific facility and in the criminal justice system. Variables ask about the types of tasks required, written procedures, and training, types of interactions with co-workers, the warden and inmates. Demographic variables include race, sex, year of birth, level of education, marital status and number of children.
Facility data (Facility Level Data, n=46 41 variables) includes variables on the daily cost per inmate, security level of the facility, design and rated capacity, inmate population count, mean age of inmate population, proportion of inmate population by race, proportion of inmate population by offense type, number of correctional officers, proportion of correctional officers by sex and race, ratings of inmate misconduct, officer morale, written policies and methods to detect and process inmate misconduct.
The inmate level data had an 81 percent participation rate. The officer level data had a response rate of 50 percent in Ohio and 47 percent in Kentucky.
Presence of Common Scales: