Supporting Statement - SSV - Part B 06.24.21

SUPPORTING STATEMENT—PART B

Survey of Sexual Victimization, 2020, 2021, 2022

2. Statistical Methods

1. Universe and Respondent Selection

The Survey of Sexual Victimization (SSV) collects information on allegations and substantiated incidents of sexual victimization that occur in adult correctional and juvenile justice facilities. BJS uses a series of sampling frames to identify the universe of facilities covered by PREA. This universe is fluid due to changes in the operational status of facilities, including openings, closings, new contracts (i.e., privately operated facilities), and ended contracts (i.e., no longer privately operated).

Approximately 7,050 facilities are covered by PREA. This includes about 5,200 adult prisons and jails and 1,850 juvenile facilities. PREA requires that BJS collect data from a sample of at least 10% of these facilities. (See Table 3 at the end of this section for the number of facilities and reporting units, by facility type covered by PREA and sampled in 2019.)

Because of the low numbers of reported sexual victimizations to correctional authorities and the centralized authority at the system level that governs responses to the surveys, BJS elected to conduct a complete enumeration at the system level – including the Federal Bureau of Prisons (BOP), all state department of corrections (DOC), all state juvenile justice systems, and each branch of the U.S. military. In each of these cases, both the information systems and the authority to report are centralized. Moreover, this annual enumeration minimizes burden on the respondents (rather than selecting a sample from the more than 1,600 facilities operated by these systems). About 2 months prior to launching the survey each year, BJS receives an update of the listing of state and federal prison administrators from the Correctional Leaders Association (formerly the Association of State Correctional Administrators) and state administrators from the Council of Juvenile Justice Administrators, who serve as the current year’s data collection respondents. Each year, the four military branches provide BJS with a single point-of-contact for their centralized reporting.

BJS also conducts a complete enumeration of all facilities operated by or under contract with the U.S. Immigration and Customs Enforcement (ICE) primarily housing detainees for ICE, called dedicated ICE facilities, and juvenile justice facilities in Indian country housing exclusively juveniles. Unlike the full enumeration of facilities with a central reporter, a respondent at each of the dedicated ICE adult facilities and Indian country juvenile facilities provide BJS data. BJS annually obtains a list of the dedicated ICE facilities from a point-of contact at ICE and identifies the full list of Indian country juvenile facilities from the Annual Survey of Jails in Indian Country (ASJIC) (described below under “Other jails.”).

Finally, private prisons, local jail jurisdictions, private jails, Indian country jails, and local and private juvenile facilities are completely decentralized. Sample designs have been relatively stable from year to year. For each type of sampled facility or jurisdiction, a brief description of the sampling frame and sample design used for SSV 2019 has been provided below (see Attachment 4 for more detail, including sampling frames, coefficients of variation, and variance estimates.) The sampling design used for the SSV 2019 will be used to draw the SSV 2020 sample. Annually, BJS will review and evaluate the precision of each of these designs. If changes are needed, BJS will submit a request for a substantive change for SSV 2021 and 2022.

Adult Correctional Facilities

Private prisons

The private prison sample is drawn from the full enumeration of facilities identified in the Census of State and Federal Adult Correctional Facilities (CCF, OMB Control No. 1121-0147), which is conducted every 5-7 years by BJS. The most recent CCF was conducted in 2019 and will be used for the 2020, 2021, and 2022 SSV samples.

A sample of 155 (29%) privately operated state and federal prison facilities was drawn from the 530 private prisons identified in the 2019 CCF. The facilities were first sorted by average daily population (ADP). To determine the size cut-off for certainty facilities, the total ADP for all eligible facilities was divided by the number of facilities in the sample. Facilities with an ADP greater than or equal to this cut-off were sampled with certainty. The remaining total ADP was then divided by the remaining number of facilities in the sample to determine whether any facilities meet the subsequent certainty cut-off size. This process was repeated until no remaining facilities have an ADP greater than or equal to the cut-off, resulting in 75 facilities with an ADP of 321 or more being sampled with certainty. The facilities remaining in the frame at this point were sorted by region (i.e., Northeast, Midwest, South, or West), state, and ADP, and sampled systematically with probability proportional to size.

Public jails

Historically, the public jail sample has been drawn from the Mortality in Correctional Institutions (MCI, formerly the Deaths in Custody Reporting Program (DCRP)) universe file. SSV used the Annual Summary of Inmates under Jail Supervision portion of this frame. The MCI file for the year preceding the SSV collection has been used as the sampling frame for SSV since 2007. BJS ceased collection of mortality data in state and local correctional facilities after the 2019 reference year.¹

BJS will now rely on the Census of Jails (COJ, OMB Control No. 1121-0100), a periodic data collection of all local jail facilities conducted every 5-6 years, for the sampling frame of public jails. Most recently, a full enumeration of facilities was identified in the 2019 COJ. The public jail sample for 2020, 2021 and 2022 will be drawn from the universe identified through this collection.

A sample of 700 (24%) publicly operated jail jurisdictions was selected from the 2019 COJ. The largest jail jurisdictions in 45 states and the District of Columbia were selected to meet the PREA requirement that at least one jail per state is selected each year.² All jail jurisdictions with ADPs greater than or equal to 1,000 inmates were selected with certainty. The remaining sample of jail jurisdictions were selected using a stratified systematic random sample. The remaining jail jurisdictions on the frame were grouped into three strata. The cumulative sqrt(f(y)) method was used to determine the boundaries of these three strata based on ADP as the measure of size (Cochran, Sampling Techniques, 1997 edition, p. 129). The jail jurisdictions in each of these three strata were then sorted by region, state, and ADP, and selected systematically with probability proportional to their size.

Private jails

The private jail sample was also historically drawn from the MCI frame. Because the MCI is no longer being conducted, the private jail sample for 2020, 2021, and 2022 will rely on the universe from the 2019 COJ (described above under “Public jails”), which includes an indicator that identifies privately operated facilities.

A sample of 15 (38%) privately operated jails was selected from the private jails in the 2019 MCI file. These private facilities were sorted by region, state, and ADP, and 9 jails were systematically sampled with probability proportional to size. Given the large standard errors, estimates for private jails are combined with public jails. The separate sample is used to ensure inclusion of private jails in the SSV.

Other jails

The Indian country jail sample is drawn from the ASJIC. BJS maintains the ASJIC, which is updated annually and includes all known Indian country adult correctional and juvenile justice facilities operated by tribal authorities or by the U.S. Department of the Interior, Bureau of Indian Affairs. The prior year of ASJIC has been used for the SSV sample each year. BJS will continue this for the next three years of SSV.

For jails in Indian country, 25 (42%) of the approximately 60 jails in Indian country housing adult inmates were selected. Using the same method to determine certainty size cut-offs as used for private prisons, 7 facilities were sampled with certainty due to their size (ADP of 104 or more). The remaining facilities on the frame were sorted by state and ADP, and 18 facilities were sampled with probability proportionate to size. Facilities in Indian country housing exclusively juveniles were excluded from the adult sample.

Juvenile Justice Facilities

Local and private juvenile justice facilities

The local and private juvenile facility sample is drawn from two frames which are used in alternating years. The Census of Juveniles in Residential Placement (CJRP), which collects population data, is conducted by the Census Bureau for the Office of Juvenile Justice and Delinquency Prevention (OJJDP) in even-numbered years. The Juvenile Residential Facility Census (JRFC), which collects data on facility characteristics, is conducted, again by the Census Bureau, for OJJDP in odd-numbered years. Again, the most recent frame is from the year preceding the SSV collection. That is, BJS will use CJRP 2019 for SSV 2020; JRFC 2020 for SSV 2021; and CJRP 2021 for SSV 2022.

A sample of 530 (46%) local and private juvenile facilities was drawn from the JRFC, as 2019 is an odd year. The facilities in the frame were separated into three groups: local, private, and detention facilities. The local and private facility groups were sorted by state and size and the largest locally and privately operated facility in each state were sampled with certainty. The remaining facilities in each of the three groups were sorted by size with the large facilities in each group being sampled with certainty due to their size. The remaining facilities were serpentine sorted by geographic region (Midwest, Northeast, South, and West), state, facility type (detention, local, or private), and size. The remaining detention facilities were sampled with probabilities proportionate to size from 4 strata based on geographic region, the remaining local facilities were divided into 2 strata based on commitment status (commitment and non-commitment), and sampled with probabilities proportionate to size, and the remaining private facilities were sorted by region and state, then sampled with probabilities proportionate to size.

Tribal juvenile justice facilities

The Indian country juvenile justice facilities are identified from the ASJIC. The prior year of ASJIC has been used for the SSV sample each year. BJS will continue this for the next three years of SSV. The 2019 SSV data collection included all 19 Indian country juvenile justice facilities identified in the ASJIC.

Table 3. Number of facilities and reporting units in 2019 covered by the Prison Rape Elimination Act

			Facilities		Reporting Units
Facility type		Form	Universe		Universe	Sampled	Percent
	Total		6,427		5,126	1,581	33.8%
Prisons
	Public – federal system	SSV-1	114		1	1	100%
	Public – state systems	SSV-2	1,155		50	50	100%
	Private	SSV-4	406		406	155	38.2%
Jails
	Public – local jurisdictions	SSV-3	3,119		2,882	700	24.3%
	Private	SSV-4	34		29	15	51.7%
Other adult facilities
	Indian country jails	SSV-4	60		60	25	41.7%
	Military systems	SSV-4	16		4	4	100%
	ICE	SSV-4	31		31	31	100%
Juvenile facilities
	Public – state/DC systems	SSV-5	329		51	51	100%
	Public – local	SSV-6	559		559	266	47.6%
	Private	SSV-6	585		585	264	45.1%
	Indian country	SSV-6	19		19	19	100%

2. Procedures for Information Collection

Collection Procedure

BJS will administer the 2020-22 SSV as a web survey. The website developed for previous iterations of the survey will continue to be used (see Attachment 5 for website screenshots). The website is hosted by the Census Bureau, and is located on a secure server. Each respondent has an individual username and a password to enter the website. To improve data quality and lower cost, all respondents are encouraged to submit data online. Paper copies of the survey will not be mailed to respondents at the onset of data collection, but will be made available upon request. If necessary, respondents will be able to submit data by email, fax, or mail.

The SSV collects data on sexual victimizations occurring in a calendar year. The implementation of the SSV 2020 will begin in September 2021. Data collection, nonresponse follow-up, data quality follow-up will last 7 months and will include a variety of mailings and telephone contacts. A brief description of all steps in the data collection protocol are provided below.

• Pre-notification letter. In early September 2021, a pre-notification letter will be mailed to each central respondent and respondents at selected facilities announcing the impending data collection and describing PREA and the importance of the survey (see Attachment 6). PDF survey forms will be made available on the SSV website for preview at that time.

• Invitation letter. In late September, each respondent will receive a survey invitation letter (see Attachment 7) announcing the start of the SSV data collection and directing them to the SSV website with login information.

• Nonresponse reminder letters and telephone contact. Two weeks prior to the survey due date a reminder email will be sent to respondents who have not yet submitted their data as notification of the impending due date. Personalized nonresponse follow-up via email and phone will be repeated about every two weeks after the submission due data and continue until the data collection closes (see Attachment 8 for a sample script for these contacts).

• Collection closeout letter/final reminder. Three weeks prior to the close of data collection, a letter, signed by the BJS Director and requiring receipt signature, will be mailed to non-respondents stressing the importance of the collection and encouraging participation (see Attachment 9).

• Telephone and e-mail data quality follow up. Data review will begin as soon as data are submitted and will continue through data collection closeout. During the review, discrepancies or missing data values or incident forms are discovered, outreach via e-mail or phone will be conducted to clarify responses or obtain missing forms (see Attachment 10 for call specifications for these contacts).

• Thank you email. A personal email is sent to respondents from the Census Bureau project manager thanking them for their participation in the SSV if communication ensued to resolve data quality issues.

3. Methods to Maximize Response Rates

While BJS is mandated to collect data under PREA (P.L. 108-79), overall, participation in the SSV has declined since 2016 when 95% of sampled systems and facilities responded. The overall response rate in 2017 was 85%. The decline in the response rate in 2017 was due to delays in launching the collection which led to the shortening of the survey cycle. While BJS anticipated improving response rates in the 2018 collection, nonresponse follow-up was hindered by COVID-19. As a result, the response rate remained at 85%. Response rates improved in the 2019 collection and we expect further improvement for the 2020-2023 collection years (see table 4 below for more detail on response rates for 2019).

Table 4. Sampled reporting units by type of facility and response status, 2019

Facility type	Form	Sampled	Active/Eligible	Responded	Response rate
Prisons
Public – federal system	SSV-1	1	1	1	100%
Public – state systems	SSV-2	50	50	50	100
Private	SSV-4	155	147	133	90.5
Jails
Public – local* jurisdictions	SSV-3	700	695	627	90.2%
Private	SSV-4	15	13	12	92.3
Other adult facilities
Indian country jails	SSV-4	25	25	21	84.0%
Military systems	SSV-4	4	4	4	100
ICE	SSV-4	31	31	31	100
Juvenile facilities
Public – state/DC systems*	SSV-5	51	47	47	100%
Public – local	SSV-6	265	262	246	93.9
Private	SSV-6	265	242	208	86.0
Indian country	SSV-6	19	17	17	100

Note: Arkansas, Connecticut, Montana, and South Dakota did not have state-operated juvenile facilities in 2019.

*5 Public Jails were out of scope, 4 of which are still open were state-operated. Their data were captured on their state’s SSV-2 form.

Over time, significant effort has been made to make the SSV data collection materials clear and straightforward, in an effort to yield high response rates. The SSV questionnaires have been designed to make collection of the data as concise and easy for the respondents as possible. Uniform definitions of terms and concepts, as well as counting rules for items to be reported are included on the forms. Item response rates for the SSV summary forms is near or at 100%. BJS receives incident- based reports on all substantiated allegations of sexual victimization.

During each iteration of the SSV, the materials are reviewed to determine if further improvement can be made to increase response rates, reduce respondent burden, save cost, and enhance data quality. To address the decline in response rates, BJS has added additional correspondence to the usual respondent contact materials. New to the 2020 collection correspondence is a pre-notification letter and the BJS Director close-out letter. The pre-notification letter will notify the respondent of the upcoming data collection and provide a survey preview, while the BJS Director close-out letter, requiring receipt signature, will be used as a last attempt to convert non-responders. BJS will continue to encourage online data submission for best data quality, but will also provide other modes of data submission (fax, e-mail, or telephone) to accommodate respondents that have limited access to the internet or simply prefer to work with hardcopy survey forms.

Throughout the data collection period, the data collection agent will monitor data submission and track survey para-data, e.g., respondent log-in timestamps, contacts made with each respondent, and comments respondents left on the survey form. The para-data will be used for tailored nonresponse follow-up and data quality follow-up. For these outreach activities, respondents receive emails from collection team from a project emails and provided the contact information of the SSV project manager at the Census Bureau. This project manager serves as a consistent “agency liaison” throughout data collection.

Nonresponse adjustments

BJS implements nonresponse weight adjustment procedures that account for unit nonresponse in order to produce national estimates of sexual victimizations. Data from BOP and all state prison systems, and U.S. military facilities, are given a weight of 1.00 because they were all selected with certainty and all respond to the survey. Additionally, data from all dedicated ICE facilities are given a weight of 1.00 as it is a full enumeration of such facilities.

Public jails, private jails, Indian country jails, and private prisons facilities are assigned an initial sampling weight equal to the inverse probability of selection. Nonresponse adjustment calculations differ for public jails and the other sampled facility types because they have different sampling designs. In each survey year, weights for responding public jail jurisdictions are adjusted for nonresponse by multiplying initial sample weights by the ratio of the sum of initial weights of active jurisdictions in each stratum to the sum of weights for participating jurisdictions. After applying the nonresponse adjustment, the sum of the final weights in each stratum equal the sum of weights for active jails in each stratum.

Nonresponse adjustments for samples of private jails, private prisons, and jails in Indian country are based on the ratio of the sum of weights times the measure of size for each affected stratum. Within each stratum the number of active jails or prisons is multiplied by the measure of size of each facility, and then summed. The ratio of the first sum to the latter sum equals the nonresponse adjustment factor for the affected stratum. Overall, after adjusting for nonresponse and summing across all strata, multiplying the adjusted final weight by the sum of the measure of size equals the total number of inmates held in private jails, private prisons, and jails in Indian country.

4. Final Testing of Procedures

BJS proposes no changes to the measures in the summary forms (SSV-1, 2, 3, 4, 5, 6). In response to a GAO request for clarification on whether attempted sexual victimizations are reported by respondents, a change was made in the introduction to definitions of inmate-on-inmate sexual victimization types on all SSV Summary Forms. BJS added, “Attempted nonconsensual sexual acts are included if recorded by the facility” to the definition.

4. Contacts for Statistical Aspects of Data Collection

BJS takes responsibility for the overall design and management of the survey, including sampling procedures, development of the questionnaires, and the analysis and publication of the data. The BJS contact is—

Laura Maruschak, Statistician

Bureau of Justice Statistics

810 Seventh St., N.W.

Washington, DC 20531

(202) 307-5986

[email protected]

The Economic Reimbursable Surveys Division at the Census Bureau is the collection agent and is responsible for the collection of all data. The Economic Statistical Methods Division is responsible for drawing the samples. The Census Bureau contact is—

Greta Clark, Survey Statistician Criminal Justice Statistics Branch Economic Reimbursable Surveys Div.

U.S. Census Bureau

4600 Silver Hill Road

Washington, DC 20233-6800

(301) 763-2586

[email protected]

1 When the Death in Custody Reporting Act was reauthorized in 2014 (P.L. 113-242), it included additional enforcement and reporting compliance requirements that are incompatible with BJS’s authorizing statute as a federal statistical agency. The U.S. Department of Justice (DOJ) determined that it would be more appropriate for the Bureau of Justice Assistance (BJA) to administer the program and collect mortality data for DOJ starting with quarter 1 (Q1) of fiscal year (FY) 2020 (October to December 2019). State DOCs and local jails are now asked to report their death information on a quarterly basis to centralized state agencies, which compile and submit the information to BJA to comply with all applicable requirements in P.L. 113-242.

2 Five states with combined jail/prison systems had no public jails: Connecticut, Delaware, Hawaii, Rhode Island, and Vermont.

7