Q & A's

Subject: Responses to OMB Passback Questions on the NCRP Clearance Package

Date: September 30, 2009

OMB requested responses to the following questions:

1. Please provide a corrected “abstract” as the one in ROCIS ends mid sentence.

2. Please provide dates bounding the research on investigating alternative methods for using NCRP data to generate national-level estimates.

3. Please provide (if something is fairly readily available) more information on the record linkage and recidivism studies referenced in SS A16, preferably including key dates and status. 

4. Please see below for a brief summary of the different collections mentioned in this clearance package for which OMB has issued terms of clearance related to reporting of race and ethnicity.  We are struck by the differences across collections particularly given the concern that some are from the same respondents.  We would like BJS to provide a summary of how it is addressing this set of terms of clearance related to the prison collections overall.

BJS responses are as follows:

1. Below is the corrected abstract to be uploaded into ROCIS:

Abstract: The National Corrections Reporting Program (NCRP) is the only national data collection furnishing annual individual‑level information for state prisoners admitted or released during the year, those in custody at year‑end, and persons discharged from parole supervision. The NCRP collects data on sentencing, time served in prison and on parole, offense, admission/release type and demographic information. BJS, the Congress, researchers and criminal justice practitioners use these data to describe annual movements of adult offenders through state correctional systems. Providers of the data are personnel in state Departments of Corrections and Parole.

2. Regarding the dates bounding the research on investigating alternative methods for using NCRP data to generate national-level estimates:

Pending availability of funds, BJS intends to issue a solicitation in January 2010 to fund research on this issue (and the one of the record linkage projects, see #3 below). The solicitation issuance/review/funding cycle takes about 6 months to complete, and the research project, expected to take about 12 months to complete. Hence, the project would likely start during the late summer/early fall of 2010 and be completed by summer 2011.

3. Regarding the record linkage projects:

Two record linkage projects were mentioned: (a) linking NCRP cohorts to each other; and (b) linking criminal history records to NCRP records. First, the project to link existing cohorts of NCRP record (say, from 2000 through 2008, prior to the 2009 collection in which state inmate id numbers will be collected and can be used for linkage purposes) involves using probabilistic matching methods, based on information such as type of charge, county of sentencing, and dates of release from prison that contained within, say, the NCRP prison release record (NCRP-B) and the NCRP parole discharge record (NCRP-C). This research to be undertaken to test and verify the linkage methods will be done as part of the methodological research project outlined in #2. Linking county-level labor market data to NCRP records is fairly simple, as NCRP includes a variable with the FIPS code for the county of sentencing. The FIPS code facilitates linking county-level data from other sources that use FIPS codes (or other county code that can be related to FIPS codes) to NCRP records.

Second, the project that aims to link NCRP to criminal history records is part of the 2005 prisoner recidivism study. The dates for that study are as follows:

• 9/30/2009 - Award cooperative agreement to develop the software to restructure raw RAP sheets into a research database with standardized codes and to create a database summarizing the releasees’ criminal histories.

• 2/1/2010 - Request the raw RAP sheets for a sample of 60,000 released prisoners using State Identification Numbers found on their NCRP records.

• 12/1/2010 – Receive a completed standardized RAP sheet database and the database summarizing the releasees’ criminal histories.

• 9/30/2011 – Publish a report detailing the recidivism patterns of prisoners released in 2005. 

4. Regarding a summary of how BJS is addressing the issues of the clearances for the race and ethnicity questions:

BJS is moving towards eliminating differences in formats for measuring race and ethnicity across its data collections in corrections. Some of these differences arose from clearances being approved at different times, and some of the differences highlighted in the examples that OMB provided no longer exist. For example, the clearance for National Prisoner Statistics—2005, #1121-0078 has expired, and the format for race and ethnicity used in current clearance (#1121-0102) is consistent with the format used in the Probation and Parole Surveys (#1121-0064) with minor modifications (see the next paragraph). The format for Deaths in Custody collection (#1121-0249) that was cleared in 2006 was modified to address OMB’s request for changes. A new clearance package for Deaths in Custody is being prepared currently, and it will also use a format that is consistent with other collections.

To bring existing formats in line with each other, BJS will modify the labels on the race categories in the current National Prisoner Statistics (#1121-0102) to make them consistent with the labels for categories used in the Probation and Parole Surveys format by (1) adding the phrase “not of Hispanic origin” to the following categories: American Indian/Alaska Native, Asian, Native Hawaiian or other Pacific Islander, and Two or more races; (2) providing an insert that provides the definitions for race, as is done with the Annual Probation and Parole Surveys (#1121-0064); and (3) adding the phrase “See insert for race/ethnicity definitions”) after the item that requests the race/ethnicity data. BJS will make these modifications beginning with the next cycle of these annual surveys (NPS-1B, December 2009, and NPS-1A June 2010) and maintain them throughout the clearance cycle. In the Deaths in Custody submission for clearance that is being sought, a one-question format is proposed that uses the same categories that are used in the Probation and Parole Surveys.

The Capital Punishment collection (clearance #1121-0030) obtains individual-level data for using a two-question format. Many respondents complete each form individually by going to a variety of data sources and are therefore able to provide responses to the separate race and ethnicity questions. While it is not evident that the two-question format is necessary for capital punishment, prior to deciding to make a change from a two-question format to a one-question format for race/ethnicity, we plan during the fielding of the 2009 reference year capital punishment collection (fielded during 2010) to talk with several of our data providers who also provide either National Prisoner Statistics or Parole data about issues that using a two-question format for capital punishment poses for them when we are also requesting aggregate count data in the National Prisoner Statistics and Parole data collections in a one-question format. We will do this work with respondents during the course of regular followup work associated with the collection. We want to learn about burden associated with providing data in two formats and efforts required to reprogram from a two-question format into a one-question format.

For NCRP, we plan to retain the two-question format because our respondents have been providing data in the format and reprogramming data into a single-question format would impose costs and burden. However, we will modify the response categories on the race question to bring them more in line with current practices. Specifically, we will omit the “Other” category and replace it with “Two or more races (not of Hispanic origin)” and add the “Additional categories in your information system” response option. For the Hispanic origin question, we will modify the response categories to add “or Latino” after the word “Hispanic.”

While we plan to keep the two-question format for the NCRP data during this clearance cycle, we will also discuss with NCRP respondents the issues associated with changing their data extraction programs to submit the data in a one-question format that is consistent with other corrections collections. Our current understanding is that once respondents have data extraction programs in place to prepare an NCRP data submission, the amount of effort required to prepare an additional year of data is relatively small compared to the effort associated with preparing the data extraction programs in the first place. The same holds for other data submissions. Thus, if the programs are in place to provide data in different formats across submissions, then the inconsistency between formats may be relatively unimportant to respondents. Moreover, asking respondents to make a change to a program (e.g., NCRP) to prepare a submission in a one-question format may impose additional burden with no gain in information and no loss in BJS ability to classify offenders into the OMB standard race/ethnicity categories. As indicated, we will use collection year 2010 (reference year 2009) of the NCRP to discuss these issues during the normal course of our followup work. If need be, we will select 9 respondents who provide data for more than one collection with whom to discuss these issues.

From this effort, we will improve our understanding of burden and inconsistency issues and be able to approach these issues systematically. If it is feasible to request all data in a one-question format, we will develop a plan in which we first notify respondents of when we expect to implement the changes, attempt to secure resources to provide technical assistance and computer programming support as needed, closely assess the data we receive in new formats to determine if there are unexpected shifts in race/ethnicity distributions, and take appropriate corrective action.

Additional actions under way to enhance consistency

Each of these collections retains the items “Additional categories in your information system” and “Not known.” BJS plans to retain these items during their current clearance cycles and to work with data providers to minimize the use of them. However, given information system constraints under which our data providers operate, we conclude that at this point, omitting them from the collection instrument would lead to degradation in the quality of the data.

Analysis of the 2008 National Prisoner Statistics data show that 21 respondents (out of 51) reported data in the “Additional categories…” option; the responses included categories such as Polish, Russian, Middle Eastern, Croatian, Macedonian, Arab, No Data, and Other (which was the modal category, as 15 of 21 used the “other” response category). In addition, 26 respondents provided data in the “Not Known” category. Combined, the counts of prisoners who were classified as “other” or for which race/ethnicity was “Not Known” was about 13,000 (out of a total of about 1.6 million for 2008). These 13,000 prisoners are not evenly distributed among state respondents. Hence while the less than 1% of prisoners nationwide fell into this “other” category, within states providing this response, the percentages ranged as high as 5% of prisoners classified as “other.”

Similarly, in the probation data for 2008, 11 respondents (out of 464) provided additional categories for race that could be reclassified into the OMB standards. The responses included categories such as Middle Eastern, Arabic, East Indian, Arabic/Chaldean. An additional 27 provided an “other” response that described roughly 12,700 probationers. These 12,700 were not uniformly distributed among these respondents. Hence, while in the aggregate, 12,700 probationers out of roughly 4 million on probation is small percentage, within the agencies whose respondents provided the “other” responses, the percentages of all probationers are much higher.

These responses suggest at least three problems, only one of which is readily solvable. First, for the categories such Russian, Polish, Arab, etc., we will provide guidance to our data providers on how to reclassify these categories into the OMB categories. Adding the insert to the National Prisoner Statistics will help with this, and with the probation collection, we will ask our data collection agent to followup with respondents and assist them with the reclassification. Moreover, if the respondents who provide aggregate count data to the National Prisoner Statistics and parole and probation data can reclassify these types of race categories into the OMB standard categories, then with the individual-level data obtained from the Deaths in Custody, Capital Punishment, and NCRP, BJS’ data collection agents can also accomplish this type of reclassification task, as well as assist data providers with their efforts to do this prior to submitting data. BJS prefers, however, to have its data collection agents do the reclassification of these types of race categories so as to minimize the response burden on our data providers.

Second, respondents who use the “other” response in the “Additional categories…” option pose a bigger challenge. Based on conversations with several of them, we have learned that the root of the issue is that the databases that they use to generate the data for these collections contain either a field or a response option within the race variables for “other,” and they have no information specific to individual records that allows them to re-classify persons who fall into the “other” category into one of the OMB standard categories. Hence, even if the respondents were able to reclassify the Poles, Croats, Arabs, and other groups into standard OMB categories, they would not be able to reclassify the “other” races or those for whom race is “Not Known.” If we removed the “other” and “Not Known” options, respondents would not be able to make the sum of those for whom they are able to categorize race/ethnicity equal to the total number they report. Eliminating this condition would potentially lower data quality, as an internal check on sums of categories is a standard practice used to enhance the reliability of data. If we combined them into a single “Not Known” category, we would confound responses in which race was knowable but classified as other with responses in which race was not reported and therefore not known. We have no direct solution to this problem other than to retain the “Additional categories…” response, but see below for more on the challenges facing our respondents in providing data that meet the OMB standards.

Third, we could try to work with our data providers to change the information systems that they use to provide data to bring them into compliance with OMB standards for race/ethnicity. However, we believe that this is not a wise course of action. A problem that many of our respondents face is that they are not the ones who designed or necessarily operate and manage the information systems that they use to generate reports to respond to our data requests. Nor are they are not the ones who design intake forms or enter data obtained at intake into the information system. Rather, most of our respondents are located within the research rather than the operations units of corrections departments, and they are viewed internally as users of the data that are designed for operational purposes, They are not necessarily in a position within the organization to exert a large amount of influence over the design of the databases that are designed primarily for corrections’ operational purposes (not necessarily research purposes or for purposes of meeting an external requestor’s definitions). This occurs in large and small corrections departments. For example, in the federal Bureau of Prisons, the research unit (our respondent) obtains “downloads” or extracts from BOP’s SENTRY database but the research unit has little input into the design of SENTRY. A similar situation exists in Ohio and many other large systems. Keeping in mind that all of these data collections are voluntary collections in which we rely on the good will of our respondents to provide data, BJS has been successful in its efforts by working with respondents to minimize burden and providing reports that they find useful. Asking them to take on intra-organizational matters—such as changing a department’s database to measure race in a manner that is consistent with OMB’s definitions—would impose a large burden on them that would not offer them a large payoff and it could threaten their continued participation in the voluntary efforts to provide data.

At more senior levels, BJS has and will continue to work with associations of corrections executives to stress the importance of providing data on race and ethnicity in a manner that is consistent with federal standards. For example, BJS regularly attends the business meetings of the Association of State Correctional Administrators and the American Correctional Association. During 2010, BJS will use available opportunities at twice a year meetings of these executives to discuss these issues. More immediately, this October BJS staff plan to attend a meeting of state probation and parole executives organized by the National Institute of Corrections. Some of these executives operate within unified prison, probation, and parole systems. We plan to present to these executives about our probation and parole data collection efforts, and among other matters will discuss issues related to reporting data on race and ethnicity. By addressing senior leadership, we believe that we can affect changes in how data on race and ethnicity are recorded and reported by state-level corrections agencies. Also, twice during 2010, BJS will attend the American Probation and Parole Association meetings. BJS attends several business meetings of the Association and reports on its data collections on community corrections. We will use the available opportunities at these meetings to discuss with state leaders in community corrections the importance of reporting race and ethnicity in a manner that is consistent with federal standards.

Thus, during the current clearance cycles, BJS will continue to work with respondents on the reclassification problem associated with race categories such as Russian, Polish, Arab, etc. BJS will work to gain more knowledge about the design and operations of the respondents’ information systems, explore the opportunities that might exist to make changes to them to bring them into compliance with OMB standards, and assess the costs associated with making these changes. To the extent that the costs are minimal (e.g., providing some technical assistance), BJS will work through its data collection agents to provide this. If there are dollar costs (e.g., hiring a computer programmer to make a one-time grant for programming. On the larger issue of affecting change in corrections information systems, our approach will continue to be to work with the executives of corrections departments to

3