CJP OMB Supporting Statement B 10-16-12

OMB Paperwork

Reduction Act Submission

Census of Juveniles on Probation

Consisting of 1 Data Collection Form

OMB No. 1121-0291

Department of Justice Office of Justice Programs Office of Juvenile Justice and Delinquency Prevention

810 7^th Street, NW

Washington, DC 20531

Brecht Donoghue Research Coordinator (202) 305-1270

brecht.donoghue@usdoj.gov

_{Section 2:}

Supporting Statement for

PRA Submission

PAPERWORK REDUCTION ACT SUBMISSION SUPPORTING STATEMENT B

SEPTEMBER 2012

Agency: Office of Juvenile Justice and Delinquency Prevention (OJJDP) Title: Census of Juveniles on Probation

Form: 1121-0291

OMB No.: (current approval expired 11/31/2011)

This document presents the Supporting Statement B for the 2012¹ Census of Juveniles on Probation (CJP). The primary goal of the CJP is to develop accurate and reliable statistics regarding the numbers and characteristics of youth on juvenile probation in the United States. OJJDP requests reinstatement, with change, of OMB collection #1121-0291. The CJP (Form CJ-17) was approved in 2009 and expired on 3/31/2011. OJJDP has spent the past year planning improvements to the collection.

Part B. Statistical Methods

1. Universe and Respondent Selection

The organization of juvenile probation in the United States is complex. In some states, probation is administered at the state level, in some at the county or judicial district level. Some states have a mixed organizational structure, such as Ohio where half the counties are administered at the state level and the other half at sub-state levels. Some jurisdictions administer probation supervision through the executive branch and some through the courts. Some jurisdictions administer adult and juvenile probation services separately and some as unified systems. In unified systems, officers may have mixed adult and juvenile or unmixed caseloads. The Bureau of Justice Statistics (BJS) has learned that the complexity of organization may result in state-level estimates undercounting significant portions of the population.

The study contractor used BJS rosters, association membership lists, directories, and websites to develop a comprehensive universe file. It is important that probation data be collected at the right level to ensure that the counts are complete and that they do not include duplication. We have found that the office level is not the best reporting level for probation counts. First, as we discovered in discussion with our American Probation and Parole Association (APPA) workgroup, there is wide variation across jurisdictions in the definition of “office,” with its meaning ranging from entire agencies to jurisdictional sub-offices to field offices. The better approach is to identify the highest-level entity for each jurisdictional unit in which the state is administered (e.g., a county, a group of counties, a judicial circuit, etc.), understanding that many states comprise multiple types of organizational units. That entity is the best reporter for its jurisdiction, whether it is an entire state or a municipality. This is the level at which policies and procedures are set and at which record- keeping is consolidated. Not only is it an inefficient use of resources to collect data from every entity that calls itself an “office,” it risks duplication in counts.

As this is a census, no sampling or respondent selection methods will be used. OJJDP seeks to provide precise national and state-level estimates on juvenile probationers and the agencies that supervise them. Ideally, a census would not require estimation because all units would respond, yielding a complete count, but in most censuses there is at least some item nonresponse. Data collection and follow-up methods have been designed to minimize nonresponse (see item 3 below for details), so the expected response rate for the data collection is 95 percent or higher.

_{2. Procedures for Collecting Information}

The CJP collection will occur only once every 2 years, rather than annually, in order to reduce the reporting burden on agencies. As this is a census, the following information is not relevant to the data collection:

• Stratification and sample selection

• Estimation

• Degree of accuracy needed for the purpose described in the justification

Information on juveniles’ race and ethnicity will not be collected using the official U.S. Census definitions. This is because OJJDP requests such information in the format most frequently used by respondents to reduce their time burden; currently more than half of data providers use definitions that combine race and ethnicity (i.e., they include “Hispanic/Latino” as a race option instead of listing this separately). Once state and local reporting systems have adopted the current U.S. Census definitions of race and ethnicity, the CJP collection method will be changed to this format.

3. Methods to Maximize Response

The study contractor will send an advance letter on behalf of OJJDP to agency directors to arrive approximately 3 weeks prior to data collection (see Attachment G). The letter will explain the survey and list whom our records indicate is the appropriate agency respondent. (Note that there are four different versions of the prenotification letter, each with slightly different wording depending on the audience: (1) a state-level agency that has not been contacted previously to provide data; (2) a city- or county-level agency that has not been contacted previously to provide data; (3) a state-level agency that has provided data previously; and (4) a city- or county-level agency that has provided data previously.)

The letter (see Attachment G) will contain a return postcard for use in updating respondent information. Staff will follow up by telephone with any agency from which we have not received a returned acknowledgement card within 1 week of fielding. At the beginning of data collection (March 2013), we will email survey invitations to all respondents identified in the frame development process as in-scope. The invitation will contain the URL and user password needed to access and complete the web survey, the toll-free Help Desk telephone number, and instructions for data upload. The field period will be 12 weeks, including nonresponse follow-up.

Our survey tracking system allows us to monitor responses and identify nonrespondents for follow-up. Two weeks after sending the survey invitation, we will send a thank-you email or postcard to all respondents. It will thank those who have responded and encourage those who have not to do so promptly. We will send a second reminder to those who have still not responded 1 week later and then shift to all-telephone follow-up. Help Desk staff will call any respondent who has not completed the survey after receiving two reminders, in order to determine the reason for the nonresponse and try to facilitate a response. They will encourage the nonrespondent to complete the survey on the web and will have the person’s unique password and ID number available to get them started. They will offer to provide other secure response modes if the respondent prefers. The Help Desk staff can also enter the respondent’s data during the telephone call or schedule a later date to do so. Our experience with a wide variety of web surveys (including those with judges, attorneys, and probation and police officers) is that few will request to mail in paper forms or complete the survey over the telephone. However, staff will accommodate the response mode needs of all respondents. We want it to be clear to all agencies that their response is important, and we will accommodate their needs to facilitate their participation in the survey. If we are unable to speak to the designated respondent in our telephone follow-up, continually reaching a “gatekeeper” (usually an assistant), we will email a survey PDF to the assistant and ask them to print it out, leave it for the contact person to complete, and return it to us via fax. Help Desk staff will enter surveys received via fax or PDF into the web instrument.

The 2009 CJP allowed respondents to provide data online through a web-based form, in an Excel spreadsheet template, via an uploaded data file (CSV, Excel or XML files) conforming to required formatting instructions and field definitions, and via paper. However, respondents may have found the reporting burden more significant because it required manipulation of data to meet the reporting requirements, so the current study also will offer respondents the option of providing non-uniform submissions, including complete “data dumps.” Allowing respondents to provide data in the formats of their choice will both increase response rates and result in a more timely completion of the analytic data file. We also provide a toll-free phone number for respondent questions.

As previously noted, the CJP provides multiple avenues for data submission. One option for respondents without electronic data will be to submit their data via an online web-based form. At the end of the 2005 and 2007 CJPSO, respondents were asked if they would use web-based reporting options; 84 percent of 2007 respondents said that they would. Further options include downloading an Excel template, completing the form, and then uploading the spreadsheet via a secure file transfer site or sending it via encrypted email; or completing the paper questionnaire by hand and either mailing or faxing responses. Respondents will choose their response mode. The surveys, whether paper or electronic, will contain the same questions and follow the same logical sequence of questions.

OJJDP requires accurate and reliable statistics on the number and characteristics of juveniles on probation. The study contractor will produce complete and fully-imputed analytic files within 12 months of the reference date (October 24, 2012) for the survey. As described above, we will make every effort to increase response rates as the first step in improving data quality. Despite our efforts, however, we expect at least some item nonresponse. The study contractor statisticians are internationally known experts in nonresponse adjustment and estimation.

Although weighting is the most common method to address unit nonresponse, imputation can be used for unit nonresponse provided that: (1) the unit nonresponse rate is low (less than 5%); and (2) there is background information available for each unit in the frame (e.g., the past year’s data for similar variables, state published data, or data available in sources such as the 2010-2011

Probation and Parole Directory, ACA, 2011). We expect that our extensive follow-up will minimize unit nonresponse, so imputation will be straightforward. Although weight adjustment can be used for item nonresponse, imputation is preferable. This is because weight adjustment is equivalent to mean imputation for estimation of the population means and totals and mean imputation is usually much less efficient than other imputation methods commonly used (e.g., ratio imputation). Ratio and regression imputation methods are good choices for continuous variables using both historical information and currently observed data, taking account of regional differences in trends.

Imputation methods such as those described above enhance the quality of data and data analysis for both continuous and categorical variables. The study contractor has extensive experience in imputation and powerful proprietary software packages (AutoImpute and WesDeck) that perform imputation for both types of variables. Through imputation, we can create a quality data set without item-missing values. We will deliver complete and documented data files capable of supporting national and state-level estimates.

4. Testing of Procedures

OJJDP is committed to creating the least burdensome, highest quality and utility data collection on juvenile probation. The CJP form was previously pre-tested, pilot-tested, subjected to focus group scrutiny, and administered in 2009. It required only minor modification for the current data collection. The study contractor has improved the roster list and modified the instructions in order to increase clarity. In addition, we created separate instructions for the paper form respondents and the web respondents so that instructions for each instrument is not cluttered with information not pertinent to the respondent.

5. Contacts for Statistical Aspects and Data Collection

The study team consists of a collaboration between Westat (the prime contractor) and Westat’s subcontractor, the National Center for Juvenile Justice (NCJJ). Westat will be managing data collection activities for the CJP. NCJJ will conduct the initial data processing and produce complete unweighted files, returning them to Westat for analysis.

The individual consulted on statistical aspects of the design was Dr. Hyunshik Lee, a senior statistician at Westat.