The Prison Rape Elimination Act of 2003 (P.L. 108-79) requires the Bureau of Justice Statistics (BJS) to carry out a comprehensive statistical review and analysis of the incidence and effects of prison rape for each calendar year.
The National Survey of Youth in Custody, 2012 provides facility-level estimates of youth reporting sexual victimization in juvenile facilities.
The second National Survey of Youth in Custody (NSYC-2)
was conducted in all 50 states and the District of Columbia
by Westat (Rockville, MD), under a cooperative agreement
with the Bureau of Justice Statistics (BJS). Data collection
was conducted in 326 juvenile facilities between February
and September 2012.
Interviewing juveniles in residential facilities on such
sensitive topics required extensive preparations with agency
and facility administrators prior to the interview. These
preparations ranged from methods to obtain consent,
procedures to file mandatory reports of child abuse or
neglect, arrangements for counseling in case a youth
became upset, and logistical support to physically carry out
the interviewing. The specific procedures that had to be
negotiated with state and local authorities were:
- Consent to interview minors - 22 states and the District
of Columbia provided consent in loco parentis (ILP), in
which the state agency acting as the guardian provided
consent; 20 states required written consent and 3 states
required either verbal or written parental or guardian
consent (PGC); and 5 states allowed for a combination of
ILP and PGC.
- Mandatory reporting of abuse or neglect - all survey
staff in direct contact with youth had to comply with
state and local reporting requirements when a youth
made a verbal statement suggesting abuse or neglect.
Jurisdictions provided contact information and
instructions for submitting reports to an agency outside
of the facility (e.g., local Child Protective Services).
- Counseling services - jurisdictions were asked to
identify both facility-based and external resources for
counseling services in the event a youth would become
emotionally upset during the interview or make a
specific request to the interviewer for such services.
The NSYC-2 comprised two questionnaires - a survey of
sexual victimization and a survey of past drug and alcohol
use and treatment. Youth were randomly assigned one of
the questionnaires so that, at the time of the interview, the
content of the survey remained unknown to facility staff
and the survey interviewers.
The interviews, which averaged approximately 30 minutes
in length, used audio computer-assisted self-interviewing
(ACASI) data collection methods. Youth interacted with the
computer-administered questionnaires using a touchscreen
and synchronized audio instructions delivered through
headphones. Youth could choose to take the interview in
either English or Spanish. Youth completed the interview in
private, with the interviewer remaining in the room but in a
position that did not offer a view of the computer screen.
A total of 9,703 youth participated in the NSYC-2. Of these,
8,707 youth completed the survey on sexual victimization
and passed editing and consistency checks. A total of
996 completed the survey on drug and alcohol use and
The NSYC-2 sampling frame included contract facilities in states where contract facilities held at least 20% of all
state-adjudicated youth or where fewer than 80 completed interviews were expected from youth held in state facilities.
Since locally and privately operated facilities were more difficult to enroll and less likely to participate in surveys
related to the Prison Rape Elimination Act (PREA), the NSYC-2 excluded contract facilities in states not needed for
state-level estimation. Given these parameters, the NSYC-2 collected data from contract facilities in 15 states.
A multistage stratified sample design was used. At the first stage of selection, 446 facilities were selected from
503 eligible facilities in the United States. Facilities were selected using the 2010 Census of Juveniles in Residential
Placement (CJRP), conducted by the Office of Juvenile Justice and Delinquency Prevention.
All facilities in the frame with 20 or more adjudicated youth were sampled with certainty. This threshold
yielded at least one sample facility in each state except Vermont, which had one state facility that housed fewer
than 10 adjudicated youth. (This facility was selected to meet the PREA mandate of including at least one facility
in every state.)
Facilities with 10 to 19 adjudicated youth were sampled with probability proportional to size. For state facilities,
the measure of size was the number of adjudicated youth reported in the 2010 CJRP. For the contract facilities, it was
the number of state-adjudicated placed youth. The selection probability of these facilities was their measure
of size divided by 20. This number corresponded to the measure of size for the smallest certainty facility.
A supplemental sample was taken to include additional contract facilities that were misclassified during the
initial sample selection. An additional 10 facilities were selected from among 24 reclassified facilities.
Subsequent state-level and facility-level enrollment efforts determined 113 of these 446 facilities to be out-of-scope.
Facilities were out-of-scope under any of the following conditions:
- closed or were schedule to close prior to data collection (35)
- did not house youth for more than 90 days (49)
- did not house state-placed youth (13) or adjudicated youth (6)
- merged with another enrolled facility (6) or was split into two separate facilities (1)
- housed only youth with a limited cognitive capacity who were unable to self-consent or assent or complete the survey (2)
- no longer a juvenile corrections facility (1)
Of the remaining 333 eligible juvenile facilities, 4 lacked consent for a sufficient number of youth to permit data
collection, and 3 were not visited due to issues related to scheduling and burden.
Selection of Youth
Rosters of adjudicated youth were provided by facilities granting in loco parentis (ILP) consent 5 weeks prior to
data collection. Facilities granting other forms of consent (either PGC or some combination of PGC and ILP)
provided a roster 9 weeks prior to data collection. Rosters were updated weekly, up to 2 weeks prior to the collection,
to reflect youth who were subsequently admitted to or discharged from each facility.
Interviewing capacity at each facility was assessed based on the number of available days, interviewing rooms, and
interviewers. In facilities determined to have sufficient capacity, all eligible youth were selected for the survey. In
other facilities, youth were randomly subsampled so the number of youth did not exceed interviewing capacity.
A total of 22,944 youth were initially selected. Among these individuals, 5,402 left prior to the interview team arriving
at the facility and 940 were excluded based on subsampling within the facility. Once the discharges and excluded cases
were removed from the pool of selected youth, 16,602 youth remained eligible for the NSYC-2.
Approximately 26 percent of youth did not participate because consent from the parent or guardian could not be
obtained, 8 percent refused to complete the interview, and 6 percent were nonrespondents for other reasons (e.g., they did not
complete the entire interview, they were not at the facility at the time of visit, the facility denied access, or they were
excluded due to extreme or inconsistent response patterns).
Weighting and nonresponse adjustments for facility and national estimates
To generate facility estimates, each youth was assigned an initial weight that corresponded to the inverse of the probability of selection within each facility. A series of adjustments was applied to the initial weight to compensate for nonresponse. These adjustments were completed in three steps:
1. Adjustment cells were constructed based on each youth's most serious offense, race or Hispanic origin, age, sex, and the number of days held in the facility.
2. An adjustment required a minimum nonresponse cell size of 10 responding youth. In many facilities, this resulted in no nonresponse adjustment, as either the facility had too few interviews (fewer than 20) to create multiple cells or the differences between respondents and nonrespondents were not significant. In facilities where significant differences were observed, 2 to 4 nonresponse cells were created.
3. After an initial nonresponse adjustment, the weights within a facility were examined and trimmed to reduce undue
influence from a small number of respondents with very large
weights. If the highest weight was 4 times greater than the
lowest weight in the facility, the highest weights were
trimmed and the difference in weighted counts was
distributed to the remaining youth. After trimming, the
high-to-low ratio in the final weight would equal to 4.
To generate national estimates, each sampled facility was assigned a weight that corresponded to the inverse of the facility's probability of selection into the sample, and the weight was adjusted for facility nonresponse. The adjusted facility weights were then multiplied by the youth weights that resulted from the three-step process outlined above, thereby producing a national-level youth weight.
Calculating response rates
A total of 8,845 youth completed the survey on sexual
victimization, and 996 completed the survey on drug and alcohol use and treatment. After excluding 138 youth whose interviews were deleted due to extreme or inconsistent responses in the sexual victimization survey, the NSYC-2 achieved a weighted overall response rate of 59 percent for all sampled youth.
Separate response rates were calculated for each participating facility. Within each facility, a base weight was created for each youth in the sexual victimization survey by taking the inverse of each youth's probability of selection. In most facilities, youth selection probabilities were the same; however, in facilities in which youth were subsampled or where rosters contained duplicate records, selection probabilities varied.
An initial facility response rate was calculated by summing the base weights for all youth who completed the sexual victimization survey and dividing it by the sum of the base weights for all sampled youth. Ineligible youth in each facility were excluded.
A final response rate was calculated to account for deleted interviews that contained extreme or inconsistent responses. This was achieved by multiplying the initial facility response rate by an adjustment ratio. In each facility, this ratio represented the sum of final weights for all interviewed youth (excluding those with extreme or inconsistent responses) divided by the sum of final weights for all interviewed youth (including those with extreme or inconsistent responses). This final adjusted response rate was then multiplied by 100.
Calculations for Adobe Mountain School (Arizona) illustrate the measurement of these weighted facility-level response rates. This facility listed 284 youth on its roster. Among those listed, 124 were subsampled out, and no interview was attempted with them. Among the remaining 160 sampled youth, 144 were sampled for the sexual victimization survey and 16 for the survey of past drug and alcohol use and treatment. Of the 144 eligible youth, 105 completed the NSYC-2 sexual victimization survey (72.9 percent). After adjusting for the probability of selection for each youth, the 105 youth who completed the sexual victimization survey represented the 284 youth in this facility. Five of the interviewed youth (4.8 percent) provided extreme or three or more inconsistent responses and were excluded. After adjusting for these cases, the resulting facility response rate was 69.4 percent (0.729 times 0.952 times 100).
Extent of Processing: ICPSR data undergo a confidentiality review and are altered when necessary to limit the risk of
disclosure. ICPSR also routinely creates ready-to-go data files along with setups in the major
statistical software formats as well as standard codebooks to accompany the data. In addition to
these procedures, ICPSR performed the following processing steps for this data collection:
Standardized missing values.
Checked for undocumented or out-of-range codes.