Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 75
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics –3– Overview of Bureau of Justice Statistics Data Series SOME OF THE DATA COLLECTIONS maintained by the Bureau of Justice Statistics (BJS) are very recent innovations whereas others were developed as the agency took its current form over the course of the 1970s. By comparison, at least one of BJS’s collections can trace its origin to the turn of the 20th century (and, by extension, to the 1850 decennial census). Its portfolio includes those that have been successfully repeated in subsequent years (and hence have developed series continuity), but it also includes one-shot efforts that could have been repeated but were not because of budget or other constraints. In their content and structure, they range from extensively developed population surveys to targeted administrative questionnaires filled out by institution managers to hand-coded summaries of court docket folders—and, accordingly, range considerably in their associated level of expense. As illustration of the varying costs of BJS’s data series, Table 3-1 summarizes BJS’s expected spending in fiscal year 2008. This chapter provides a brief overview of BJS’s major data collection efforts, divided into four major topic areas: victimization (Section 3–A), corrections (3–B), law enforcement (3–C), and adjudications (3–D). (In recent years, BJS has undertaken a series of data collections specifically focusing on justice issues on American Indian reservations and tribal lands; these collections slightly overlap the major topic areas and are separately described in Box 3-1.) Within each of these sections, a table illustrates the years of collection for the various series under that topic heading. As noted in Chapter 1, these summaries are not intended to be full dossiers on the collections, their
OCR for page 76
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics Table 3-1 Estimated Funding for Bureau of Justice Statistics Criminal Justice Statistics Program, Fiscal Year 2008 Program Estimated Funding (thousands of dollars) Victimization National Crime Victimization Survey—Collection 18,700 National Crime Victimization Survey—Redesign Effort 3,900 Law Enforcement Census of State and Local Law Enforcement Agencies 905 Prosecution and Adjudication Statistics Civil Trial Court Cases 340 Court Statistics Project 415 National Judicial Reporting Program (Year 1) 340 State Court Processing Statistics 300 Survey Development, Two New Collections 300 Corrections Annual Probation and Parole Statistics 175 Annual Survey of Jails 230 Capital Punishment Statistics 260 Census of Probation and Parole Agencies 250 Deaths in Custody Reporting Program 553 National Corrections Reporting Program 647 National Prisoner Statistics 130 State Prison Expenditures 300 Federal Justice Statistics Program 800 Criminal Justice Employment and Expenditures 222 Firearm Background Check Statistics 360 Tribal Statistics Tribal Criminal History Improvement Program 704 State Tribal Crime Reports 145 State Justice Statistics Program State Statistical Analysis Centers 2,300 Technical Assistance to SACs/Multi-State Projects 1,400 Publication and Dissemination 2,872 Management, Administration, and Joint Federal Statistics Efforts 2,131 Total 38,679 NOTE: Expenditures for National Criminal History Improvement Program grants and for data collections pursuant to the Prison Rape Elimination Act of 2003 are funded through separate lines in the BJS budget. SOURCE: Adapted from table provided by BJS. uses, and their associated methodological challenges, but rather a general orientation. In particular, our interim report (National Research Council, 2008b) describes the development and protocols of the National Crime Victimization Survey (NCVS) in considerably more detail than we attempt in this more limited treatment.
OCR for page 77
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics We close the chapter in Section 3–F with general assessments of BJS’s portfolio and the structure of BJS collections within that portfolio. 3–A VICTIMIZATION The President’s Commission on Law Enforcement and Administration of Justice (1967) that developed the justice system “funnel” model we use as a framework in this report (Section 2–A) and recommended the creation of what would become BJS also pioneered a new approach to studying crime. The commission sponsored the National Opinion Research Center (NORC) to survey the members of 10,000 households on their experiences as victims of crime and violence, and the commission then compared the results to the reported-to-police estimates from the Uniform Crime Reporting (UCR) program. This prototype National Survey of Criminal Victims demonstrated to the commission that “for the Nation as a whole there is far more crime than ever is reported” (President’s Commission on Law Enforcement and Administration of Justice, 1967:v): Burglaries occur about three times more often than they are reported to police. Aggravated assaults and larcenies over $50 occur twice as often as they are reported. There are 50 percent more robberies than reported. In some areas, only one-tenth of the total number of certain kinds of crimes are reported to the police. As a consequence of the commission’s report, the Omnibus Crime Control and Safe Streets Act of 1968 specifically mandated the new Law Enforcement Assistance Administration (LEAA) to “collect, evaluate, publish, and disseminate statistics and other information on the condition and progress of law enforcement in the several States” (P.L. 93-83 § 515(b); see also U.S. Census Bureau, 2003:A1-5). This data-gathering authority was invoked to begin pilot work and implementation of what would become the National Crime Survey; later, at the culmination of an extensive redesign process, this survey was renamed the National Crime Victimization Survey. Drawn from an original mandate to study the progress of law enforcement, the NCVS was designed to do so by asking respondents about victimization incidents generally, whether or not they were reported to authorities. Accordingly, the NCVS presented the unique ability to shed light on the “dark figure of crime”—the phrase coined by Biderman and Reiss (1967) to describe criminal incidents that are not reported to police. Table 3-2 shows the years of collection of the NCVS and, more specifically, the topic supplements to the NCVS. As we will discuss, there are some instances in which content from a supplement was subsequently integrated into the core NCVS; hence, the table’s suggestion that a supplement was not repeated in later years does not necessarily mean that the topic was dropped.
OCR for page 78
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics Box 3-1 Bureau of Justice Statistics Collections on Tribal Justice Tribal justice agencies are included in several of BJS’s principal data series such as the Census of State and Local Law Enforcement Agencies (Hickman, 2003). However, on tribal lands, “criminal jurisdiction … is divided among the Federal, State, and tribal governments” depending on the “nature of the offense, whether the offender or victim was a tribal member, and the State in which the crime occurred” (Perry, 2005:1). In recent years, BJS has developed one-shot and continuing collections on the justice authorities that operate in American Indian tribal lands. BJS typically has worked with the U.S. Department of the Interior’s Bureau of Indian Affairs on developing address and contact lists; the U.S. Department of Justice also maintains an Office of Tribal Justice that facilitates interactions between the Justice Department and the tribes, and the Federal Bureau of Investigation (FBI) shares law enforcement authority on tribal lands. Conducted on a one-shot basis to date, in 2002, the Census of Tribal Justice Agencies (Perry, 2005) was the U.S. Department of Justice’s first comprehensive effort to document tribal justice agencies as systems in their own right. The major difference between this collection and BJS’s standard series is that it sought to articulate the use of “indigenous forum” arrangements (e.g., councils of elders or “sentencing circles”) that are distinct from court proceedings, and that may combine adjudication and law enforcement functions on reservation lands. For those tribes that have developed specific law enforcement agencies, the census sought staffing and policy information (including the level of interface and cross-deputization with nontribal authorities). Questions on the census were also intended to provide information on tribes’ criminal history record-keeping and ability to provide crime reports to federal efforts such as the FBI’s National Crime Information Center and National Sex Offender Registry. The census achieved participation from 314 of 341 federally recognized tribes; however, “participation by Alaska Native tribes or villages was not extensive enough to enable their inclusion” in the results of the census (Perry, 2005:iii). BJS contracted with Falmouth Institute and Policy Studies, Inc., as the data collection agent for the census. In 1998, BJS began specifically collecting information on tribal jails as a component of the Annual Survey of Jails; the collection has since developed into a separate Survey of Jails in Indian Country that continues on an annual basis. This collection mirrors the content of the Survey of Inmates of Local Jails (Section 3–B.2) but concentrates on those facilities and detention centers operated by the federal Bureau of Indian Affairs or by individual tribal authorities. The Survey of Jails in Indian Country also includes questions on facility programs and services, such as health care and counseling, but these questions typically are asked only in selected years. See, e.g., Minton (2006, 2008) for reports of survey results. BJS’s National Criminal History Improvement Program of grants to assist local law enforcement agencies develop criminal history and other information databases has recently expanded to include a tribal justice–specific component. Perry (2007) summarizes work in the first few years of the program, 2004–2006. Dubbed the Tribal Criminal History Record Improvement Program, the program’s grant solicitation for 2008 puts particular priority on “enhancing automated identification systems, records of protective orders involving domestic violence and stalking, sex offender records, [and] DWI/DUI conviction information,” as well as developing tribal interfaces to federal background check data. The experiences of American Indians living on and off reservation lands in the justice system have been analyzed by BJS staff using the National Crime Victimization Survey and other resources; see, e.g., Greenfeld and Smith (1999) and Perry (2004).
OCR for page 79
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics Table 3-2 Bureau of Justice Statistics Data Collection History and Schedule, Victimization, 1981–2009 Series 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 Total National Crime Victimization Survey • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 29 NCVS Supplements or New Content Areas Crimes Against Disabled · · · · · · · · · · · · · · · · · · · · · · · · · · • • • 3 Cybercrime · · · · · · · · · · · · · · · · · · · · • • • • · · · · · 4 Hate Crimes · · · · · · · · · · · · · · · · · · • • • • • • • • • • • 11 Identity Theft · · · · · · · · · · · · · · · · · · · · · · · • • • • • • 6 Police-Public Contact Survey/Police Use of Force · · · · · · · · · · · · · · · • · · • · · • · · • · · • · 5 School Crime Supplement · · · · · · · · • · · · · · • · · · • · • · • · • · • · • 8 Supplemental Victimization Survey (stalking) · · · · · · · · · · · · · · · · · · · · · · · · · • · · · 1 Vandalism · · · · · · · · · · · • • • • • • • • • • • • • • • • · · 16 Workplace Violence Supplement · · · · · · · · · · · · · · · · · · · · · · • · · · · · · 1 City-Level Victimization Surveys (12) · · · · · · · · · · · · · · · · · · • · · · · · · · · · · 1 NOTES: •, data collected. ·, data not collected. Preliminary versions of NCVS questions on crimes against the disabled were asked as early as 2000, but only those used starting in 2007 were meant for or suitable for producing regular estimates. SOURCE: Bureau of Justice Statistics.
OCR for page 80
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics In its role as an indicator of crime levels in the United States, the NCVS is often compared to data from the UCR program maintained by the Federal Bureau of Investigation (FBI); we discuss the UCR and related programs in more detail in Section 4–C. 3–A.1 National Crime Victimization Survey Since the outset, BJS has commissioned the U.S. Census Bureau as the data collection agent for the NCVS; the NCVS is one of the major federal surveys administered by the Census Bureau’s Demographic Surveys Division. The NCVS is a household survey using a rotating panel sample of addresses, meaning that addresses are chosen to be eligible for interviewing for a certain number of interviews over a fixed period of time. Currently, contacts are made for interviews at sample addresses for 3 years, seven interviews at 6-month intervals.1 As sample addresses complete their time in sample, they are replaced with new ones. When the NCVS began in 1972, the NCVS was administered to 72,000 households—a large sample, meant to produce reliable estimates of year-to-year change in victimization as well as information on relatively rare crime types. However, sample size reductions (for purposes of cost savings) since the 1980s have reduced the sample size of the NCVS by almost half—in 2005, the NCVS was administered to about 38,600 households or 67,000 people. The sample sizes and response rates for the NCVS between 1996 and 2006 are shown in Table 3-3. Although the current sample size qualifies the NCVS as a large data collection program, occurrences of victimization are essentially a rare event relative to the whole population: many respondents to the survey do not have incidents to report when they are contacted by the survey. Consequently, as we described in Chapter 1, the reduced sample size (combined with generally low and decreasing estimated overall victimization rates) is such that only a large percentage change in violent crime victimization rates—at least 8 percent—is a statistically significant year-to-year change. Indeed, as noted in the U.S. Office of Management and Budget (2007:8) annual review of statistical program funding, “cost cutting measures applied to the NCVS continue to have significant effects on the precision of the estimates—year-to-year change estimates are no longer feasible and have been replaced with two-year rolling averages” in BJS reports on victimization. 1 The sample is further divided into six “rotation groups,” and each of these into six “panels.” One panel from each of the rotation groups is designated for interviewing each month, hence the “rotating panel” nomenclature. In large part, the use of this rotating panel structure derives directly from the choice and retention of the Census Bureau as data collector for the NCVS; to achieve some efficiencies in collection, such as sharing a pool of interviewers, some design features of the NCVS were chosen to emulate those of the Census Bureau’s Current Population Survey (CPS), then the Census Bureau’s largest intercensal survey (National Research Council, 2008b:123; see also Cantor and Lynch, 2000:107).
OCR for page 81
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics Table 3-3 Number of Households and Persons Interviewed by Year, 1996–2006 Year Households Persons Sample Size Response Rate Sample Size Response Rate 1996 45,000 93 85,330 91 1997 42,910 95 79,470 90 1998 43,000 94 78,900 89 1999 43,000 93 77,750 89 2000 43,000 93 79,710 90 2001 44,000 93 79,950 89 2002 42,000 92 76,050 87 2003 42,000 92 74,520 86 2004 42,000 91 74,500 86 2005 38,600 91 67,000 84 2006 38,000 91 67,650 86 NOTE: These sample sizes correspond to the number of separate households and persons designated for contact in a particular year. Participation rates for a particular year would be roughly double these, accounting for two interviews with sample addresses in the same year. SOURCE: Bureau of Justice Statistics (2006b, 2008c). A multiyear redesign effort, culminating in 1992 with the first collection using all of the new procedures (and the renaming of the survey to NCVS), focused principally on implementing a screening procedure. The first part of an NCVS interview is the screening questionnaire, which uses a series of carefully constructed questions to elicit counts—but not yet full information—about crime victimization incidents in the past 6 months. After this screener has been completed, the NCVS interviewer guides the respondent through the completion of a detailed incident report on the circumstances of each incident counted by the screener. This interviewing process is repeated for each person age 12 or older in the household at the sample address (although only the first respondent is asked about general characteristics of the household). Because the number of victimizations experienced by respondents varies—and, with it, the number of incident reports that must be completed—the length of time that it can take to complete an NCVS interview can also vary. BJS estimates that the average face-to-face interview lasts 26 minutes (Bureau of Justice Statistics, 2008d:10). Over time, the mode of NCVS interviewing and the use of interviews to calculate estimates has shifted. BJS still insists that the first interview with a sample household be conducted in person by a Census Bureau enumerator. Beginning in 1980, BJS began to permit every other interview (after the first contact) to be conducted via phone; by 2003, NCVS interviews were
OCR for page 82
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics being advised to complete their interviews by phone “whenever possible” (U.S. Census Bureau, 2003:A1-11). In the late 1990s and early 2000s, BJS also invested in the use of survey automation procedures, so that the NCVS and its supplements are fully electronic; face-to-face interviews are done by computer-assisted personal interviewing (CAPI) with the interviewer using a laptop computer, and telephone interviews can be completed using the same computer interface.2 In terms of how NCVS interviews result in final estimates, BJS currently produces “collection year” estimates from the NCVS, accumulating the data from all interviews completed in a particular year t. Because of the 6-month reference period of the survey, this means that a year t estimate from the NCVS includes events that may have occurred in the last half of the year t − 1 and does not include all incidents that actually occurred in year t (since late year t incidents would only be picked up in interviews in year t + 1). Until recently, a key feature of the NCVS was that the first interview with a sample household was not included in the estimates. Instead, it was withheld and used as a bounding interview: counts and incidents reported in the second interview could be checked against the bounding interview to correct for the same incidents being reported multiple times. As one of a bundle of cost-cutting measures, BJS and the Census Bureau began to include these “bounding interviews” in the production of estimates, beginning with estimates from the 2006 administration of the survey. As we discuss in Section 3–A.3, these changes produced anomalous results that led BJS to declare a “break in series” that prevents comparison with previous years’ data. NCVS results are annually described in two BJS report series, Criminal Victimization (e.g., Rand and Catalano, 2007) and Crime and the Nation’s Households (e.g., Klaus, 2007). BJS staff have issued a number of “Special Reports” dedicated to analysis of particular content from the NCVS, such as the involvement of weapons and firearms in victimization incidents (Perkins, 2003), violence in the workplace (Duhart, 2001), reporting of rape or attempted rape to the police (Rennison, 2002b),3 and specific reports on victimization experiences by racial and ethnic groups (e.g. Rennison, 2002a; Harrell, 2007). A complete list of references to publications that have used and analyzed NCVS data is beyond the scope of this report, but some recent references suggest the breadth of application of the data. NCVS data have been probed to study the issue of violence and police response to incidents in disadvantaged areas and neighborhoods (Baumer, 2002; Baumer et al., 2 As mentioned in the next section, BJS and the Census Bureau attempted to do many of the NCVS telephone interviews from centralized Census Bureau call centers, but this practice was eliminated in the most recent set of cost-cutting measures. 3 Rennison (2002b) is officially designated a “Selected Findings” report rather than a “Special Report.”
OCR for page 83
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics 2003); the NCVS has also been used in lines of research on the reporting of crime by women, particularly of rape, to authorities (Baumer et al., 2003; Felson et al., 2005; Xie et al., 2006; Addington and Rennison, 2008) and the effect of victimization on residential turnover (Dugan, 1999; Xie and McDowall, 2008). Longer-term analyses have considered the comparability of NCVS with other sources of data on crime such as the UCR (Lynch and Addington, 2007) and the National Violence Against Women Survey (Rand and Rennison, 2005), and differential trends in violence by gender (Lauritsen and Heimer, 2008). 3–A.2 NCVS Supplements Over time, NCVS’s flexibility as a survey vehicle has been exploited to gather occasional or one-time data through survey supplements. These supplements have been supported by contributions from partner agencies or grants from other organizations; supplements have also been directly developed to respond to new mandates from Congress. A battery of questions on crime against the disabled was developed in direct response to P.L. 105-301, the Crime Victims with Disabilities Awareness Act of 1998, which directed that the NCVS be used to measure “the nature of crimes against individuals with developmental disabilities” and “the specific characteristics of the victims of those crimes.” The module of questions is meant to assess whether victims of crime were in poor health, had any physical or mental impairments, or had disabilities that affected their everyday life. They are also asked to judge if any of these provided an opportunity for their victimization. Similarly, since 1999, the NCVS has collected data on hate crimes, a content area motivated by the Hate Crime Statistics Act of 1990 (P.L. 101-275). The hate crime questions were not a standalone module but rather a section added directly to the NCVS questionnaire. BJS collaborated with its fellow Office of Justice Programs agency, the National Institute of Justice, to include a module of questions on crime in schools in the 1989 version of the survey (Bureau of Justice Statistics, 2008e:5). After a repeat administration in the mid-1990s, the National Center for Education Statistics (NCES) of the U.S. Department of Education has paid for BJS to include the School Crime Supplement (SCS) to the NCVS on a biennial basis. As described by the U.S. Census Bureau, Demographic Surveys Division (2007:51): The supplement contains questions on preventative measures employed by the school to deter crime; students’ participation in extracurricular activities; transportation to and from school; students’ perception of rules and equality in school; bullying and hate crime in school; the presence of street gangs in school; availability of drugs and alcohol in
OCR for page 84
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics the school; attitudinal questions relating to the fear of victimization in school; access to firearms; and student characteristics such as grades received in school and postgraduate plans. The school crime questions were administered to all individual respondents in NCVS sample households who were between ages 12 and 18, “who were enrolled in primary or secondary education programs leading to a high school diploma, and who were enrolled in school sometime during the six months prior to the interview” (U.S. Census Bureau, Demographic Surveys Division, 2007:51).4 The SCS is administered to NCVS respondents (maintaining the age-12-and-older restriction of the main survey) who attend schools. In the 2005 administration of the SCS (running from January through June), 61.7 percent of the 11,525 eligible SCS respondents completed the questions (Bureau of Justice Statistics, 2008e:6). SCS data are distinct from, and provide a different vantage on crimes in school from, the Schools Survey on Crime and Justice that NCES contracts directly with the Census Bureau to construct; that survey is essentially an establishment survey, meant to be completed by principals of a nationally representative sample of public elementary and secondary schools. Results from the various administrations of the SCS, and related data resources, are described by Dinkes et al. (2007), and the SCS data have been used in analyses by, for example, Addington (2003). Questions on experiences with identity theft were first added to the NCVS in July 2004. Whereas other NCVS supplements such as the SCS function as a true supplement to the NCVS interviewing experience—a standalone questionnaire that is meant to be answered by everyone meeting certain eligibility requirements—the identity theft “supplement” was built into the NCVS screening interview. The version of the identity theft questions used in 2004 is illustrated in Figures 3-1 and 3-2. As noted in Section 2–C.1, this set of identity theft questions is one of few available quantitative measures of the levels of some types of fraud. With additional sponsorship from the Federal Trade Commission and other bureaus within the Office of Justice Programs, BJS planned to field a more comprehensive identity theft supplement to the NCVS beginning in 2008 (Baum, 2007:4). Another important supplement, the Police-Public Contact Survey (PPCS), stems from a brief provision in the Violent Crime Control and Law Enforcement Act of 1994 (P.L. 103-322 § 210402): “the Attorney General shall, through appropriate means, acquire data about the use of excessive force by law enforcement officers.” As a direct result of this mandate, BJS developed the PPCS to measure the extent of all types of interactions between the police and members of the public (of which those involving 4 However, “students who were home schooled were not included past the screening questions since it was determined that many of the questions in the SCS were not relevant to their situation” (Bureau of Justice Statistics, 2008e:5).
OCR for page 85
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics Figure 3-1 Module of questions on identity theft in the 2004 National Crime Victimization Survey (part 1) NOTE: The questions are part of NCVS-1, the screening questionnaire part of the NCVS interview.
OCR for page 154
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics Against this backdrop, the two recidivism studies conducted by BJS in 1983 and 1994 have been influential in structuring research on the relationship between crime and incarceration—and hinting at a major looming challenge for policy makers. The studies, which followed two cohorts of prison releasees in selected states for 3 years and recorded their subsequent patterns of arrest and reincarceration, demonstrated that around 60 percent of those coming out of state prison were rearrested within 3 years of prison release. These special studies of recidivism yielded several widely cited BJS reports. The recidivism microdata were also made publicly available, and have been widely studied by researchers and policy analysts (e.g. Solomon et al., 2005; Travis and Visher, 2005). About 700,000 people are now released annually from state and federal prison (Sabol et al., 2007:1). Another 5 million are currently under some kind community supervision either on parole (800,000) or on probation (4.2 million) (Glaze and Bonczar, 2007:2). The most recent estimates of recidivism, from 15 states in 1994, suggest that about two-thirds of prison releasees will be rearrested within 3 years, and a quarter will be reincarcerated with a new sentence (Langan and Levin, 2002). The significant growth of imprisonment rates over the past several decades has highlighted the policy and social science challenges presented by historically large cohorts of released prisoners and has increased the urgency of developing techniques for community reintegration in order to deter recidivism. Between the 1994 recidivism study and BJS’s recent reorganization to elevate “recidivism, reentry, and special projects” as a program priority, BJS’s direct role in reentry issues has been outpaced by research efforts mounted by other units in the Justice Department and external researchers. For example, NIJ funded a major evaluation of the Serious and Violent Offender Reentry Initiative (SVORI), a collaborative grant program funded by five cabinet departments that instituted reentry programs in 69 sites around the country.17 SVORI programs included in-prison training programs prior to release as well as postrelease programs; programs included substance abuse and mental health treatment, housing assistance, and faith-based programs. The multisite evaluation is intended to identify effective approaches; findings from the evaluation in progress are described by Lattimore et al. (2005) and Lattimore et al. (2004). Other researchers have begun to intensively study the consequences of incarceration for the employment, family, and health outcomes of men and women released from prison; see, for example, Pager (2003), Lopoo and Western (2005), Kling (2006), and Johnson and Raphael (2006b). 17 Specifically, the Departments of Education, Health and Human Services, Housing and Urban Development, Justice, and Labor contributed funding to SVORI programs. The NIJ-funded Multi-site Evaluation of SVORI Programs was administered by RTI International and the Urban Institute.
OCR for page 155
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics Though its direct work in the field has been limited to date, BJS can and should be a major source of quantitative information on prisoner reentry, recidivism, and community-based supervision issues. Nearly 80 percent of prisoners are released to community supervision, and so data collections on prisoner reentry would significantly extend the coverage of parolees. More than this, regular data collections on reentry and recidivism would advance the core charge of compiling statistics on crime and analyzing its correlates. In the current period of historically unprecedented incarceration rates, reentry and recidivism data would also offer valuable information about the social impacts of incarceration, an area now regarded as of pressing policy significance. Data collection efforts should evolve in response to social trends and their policy context; developing a program on reentering prisoners reflects the new reality of large cohorts of releases, and the significance of these cohorts for crime and social cohesion in the general population. Recommendation 3.14: BJS should mount a feasibility study of the flow of individuals between correctional supervision and community settings. Repeated interviews of samples of about-to-be-released prisoners that track their successes and failures in reintegrating with the community would enhance understanding of this critical policy issue. In early 2008, the Second Chance Act became law, instituting and authorizing a wide variety of reentry programs. Included in its provisions is a section defining a role for BJS data collections; see Box 3-5. The act permits BJS to routinize its earlier recividism studies, calling for BJS to conduct them on a triennial basis. In the legislative context, of course, authorization is different from appropriation, and the act’s language that BJS “may conduct research” rather than “shall” stops short of a direct mandate. Still, the act is an important signal that correctional programming will likely gain renewed support over the coming decade, and that a BJS role is both expected and required. The act also authorizes NIJ to carry out research studies along these lines; exactly how large efforts such as a major recidivism study would be divided between and administered by a statistical agency (BJS) and a research agency (NIJ) would need to be carefully determined. Although BJS’s coverage of custodial correctional populations is strong and it has a fairly complete picture of annual flows of persons moving in and out of prison, its current coverage of the population released from incarceration is seriously incomplete at present. Hence, higher priority to these issues and congressional authorization are both welcome developments. As BJS approaches the problems of recidivism and prisoner reentry—significant frontier issues that should weigh heavily in the agency’s strategic planning—we suggest some possible topics that form part of this planning and shape expanded data collections:
OCR for page 156
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics Box 3-5 The Second Chance Act of 2007 Signed into law on April 9, 2008, and codified as 42 USC § 17551, the Second Chance Act of 2007 directs that the Bureau of Justice Statistics (BJS): may conduct research on offender reentry, including— an analysis of special populations (including prisoners with mental illness or substance abuse disorders, female offenders, juvenile offenders, offenders with limited English proficiency, and the elderly) that present unique reentry challenges; studies to determine which offenders are returning to prison, jail, or a juvenile facility and which of those returning offenders represent the greatest risk to victims and community safety; annual reports on the demographic characteristics of the population reentering society from prisons, jails, and juvenile facilities; a national recidivism study every 3 years; a study of parole, probation, or post-incarceration supervision populations and revocations; and a study concerning the most appropriate measure to be used when reporting recidivism rates (whether rearrest, reincarceration, or any other valid, evidence-based measure). In debating the act on April 12, 2007, bill sponsor Edward Kennedy (D-Mass.) cited existing BJS corrections data series in arguing for the bill’s merits (Congressional Record, pp. 4430–4431): “Large prison populations and high recidivism rates place heavy burdens on prisons, communities, and taxpayers. Of the 2.2 million persons housed in prisons today—an average annual increase of 3 percent in the past decade—97 percent will be released into the community. Overcrowding continues to plague the system. State prisons are operating at full capacity and sometimes as much as 14 percent above capacity, and Federal prisons are 34 percent above capacity. In 2005, prison populations in 14 States rose at least 5 percent. Recidivism and inadequate reentry programs add to the problem. Over 600,000 prisoners are released each year, but two-thirds of them are arrested again within 3 years.” “According to a recent Bureau of Justice Statistics report, of the approximately 50 percent of prisoners who met the criteria for drug dependence or abuse, less than half participated in drug treatment programs since their admission to prison.” “The Bureau of Justice Statistics reports that only 46 percent of incarcerated individuals have a high school diploma or its equivalent. The limited availability of education and vocational training programs exacerbates the problem. Only 5 percent of jail jurisdictions offer vocational training, and 33 percent of jurisdictions offer no educational or vocational training at all.” Reinstitute sample survey of probationers and parolees: As described in Section 3–B.6, BJS’s current coverage of persons under community supervision is limited to administrative data collected through the Annual Probation and Parole Surveys administered to supervising agencies. These surveys provide counts of probationers and parolees disaggregated by race, sex, and offense category. However, these administrative data are extremely limited for studying reentry and recidivism; they provide little information about the conditions of supervision or
OCR for page 157
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics the circumstances of success or failure in individual cases, and they yield no information about past criminal history. Actual person interviewing with a sample of probationers has only been conducted once by BJS, in 1995, yet it is this kind of rich information on the experiences of those who have already reentered the community that would be most valuable in shaping emerging reentry strategies. Occasional surveys of the parole and probation populations would improve understanding of the process of reentry, recidivism, and successful reintegration into the community. A reinstituted representative sample of probationers and parolees should elicit data on risk factors (schooling, social background, health status, and criminal history, for example), the conditions of community supervision and its intensity, and participation in assistance programs. Information on the spatial distribution of parole and probation populations (e.g., distance from previous “home” communities prior to incarceration and limitations on geographic mobility) would also likely advance understanding of recidivism and reentry. The current administrative data also do not speak to the circumstances of arrest, revocation, conviction, or incarceration of parolees and probationers. Regular statistics are collected on arrest and recommitment to prison, but there is little detail on technical violations or the administrative procedure of parole and probation revocation; a survey program, complemented by revision of the content of the administrative survey questionnaire, could help fill these gaps. Survey costs for this population are likely to be substantial, and meeting these costs will likely require long-range planning and partnering with other agencies. Still, because the population of released prisoners is now so large, agencies within the Departments of Health and Human Services, Education, Housing and Urban Development, Labor, or Veterans Affairs (as well as other statistical agencies such as the Census Bureau and the Bureau of Labor Statistics) may have shared interests in the probation and parole populations and be a source of input and funding. Routinize the national recidivism studies: As acknowledged in the debate on the Second Chance Act and exemplified by a direct request in the act’s language, the BJS recidivism studies of 1983 and 1994 have made major contributions to understanding of postprison experiences of state prisoners and should be conducted on a more regular basis. Recidivism studies are also important by significantly expanding the empirical scope of the BJS data collections by including those who have completed sentences and are no longer under any kind of supervision at the time of their return to incarceration. The previous recidi-
OCR for page 158
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics vism studies were based on a large and complex record linkage effort that joined correctional records to arrest and court data. Although exemplary, the paradigm could be pushed further with either survey interviews or linked records for several terms of prison incarceration. (This was typically infeasible in the recidivism study because of the relatively short 3-year follow-up period.) Survey data from released prisoners would be particularly informative about the social context of recidivism, describing in greater detail the economic and social situation of those coming out of prison. Measuring jail flows: A point inherent in our suggestion to improve measurement of transitions in BJS’s existing corrections data (Section 3–F.1) is worth reemphasizing here. Not as much is known about those persons passing through the nation’s jail system as BJS’s data reveal about the federal and state prisons. While about 700,000 people annually enter and exit prison, some 10 million people are estimated to pass through local jails in a given year. Although jail incarceration is likely common for released prisoners and releasees might cycle in and out of jail before returning to prison, there are no national statistics to document the pattern. Likewise, relatively little is known about the frequency with which the same individuals go in and out of the jail system—for instance, whether frequent contacts with law enforcement and numerous short spells spent in jails or police lock-ups constitute a de facto form of community supervision. Alternative approaches to studying the unsupervised population: Just as prison sentences expire, so too do sentences of probation or other community supervision. As rules for probation supervision change (in part due to state efforts to grapple with growing costs of corrections), the size of the released and unsupervised population is growing. Some may argue that those who have “maxed out” of prison or corrections supervision fall outside the statistical jurisdiction of BJS, but the point remains that the unsupervised population may be at high risk of rearrest, and their criminal histories place them at risk of an array of diminished life chances. If BJS data collections are going to be significantly informative about recidivism and reentry, the released unsupervised population should be contemplated as targets for new data collection. Unlike parolees, the unsupervised present acute difficulties for data collections. Here, linking criminal justice to noncriminal justice (say, social welfare) administrative records, may provide one promising path for data collection. Studying the demographic significance of the penal system: While the topics for BJS strategic planning might speak to the challenges of understanding recidivism and reentry, they also speak to the broader de-
OCR for page 159
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics mographic significance of the penal system. As an institutionalized influence on the life course and spatial mobility, the penal system now commands a large demographic influence that is mostly hidden from the nation’s statistical system. Though prisoners and other institutionalized populations were counted in the 2000 decennial census and are included in the group quarters component of the Census Bureau’s American Community Survey, nothing is known of the geographic areas from which they originate or much about their lives immediately before institutionalization. By trying to capture the flow of people into and out of prison, and their return, the reentry and recidivism perspective highlights the significant influence of the prison on basic population processes. The ideas discussed here suggest a variety of new BJS activities focused on formerly-incarcerated men and women. In the current climate of tight budget constraints, and short of the authorizations in the Second Chance Act actually yielding significant appropriations of new funds, the birth of extensive new data collections seems unlikely. Still, BJS should think opportunistically about (a) partnering with other agencies, (b) linking records across databases, and (c) augmenting existing administrative surveys of probation and parole agencies. First, BJS should study the possibility of partnering with other statistical agencies. Such partnerships would provide two kinds of benefits. Items about involvement in the criminal justice system could be added to household and other surveys, expanding understanding of the reach of the justice system in the noninstitutionalized population. For example, questions about prior arrests or incarceration could be asked of the large samples in the American Community Survey, the Current Population Survey, or the NCVS. Because of the missions (and already large scope) of these surveys, such additional queries would not be highly detailed; however, they would provide general indicators at the national level and present the opportunity for some disaggregation by geography and demographic subgroups. In return, the BJS is also uniquely placed to provide detailed information about the institutionalized population. Questions about education, health status, aging, or demography, for example, could all be of interest to other agencies that have largely focused on the noninstitutionalized population. We return to this point in Section 5–B.11. Second, record linkage holds great promise for expanding understanding of how people move in and out of institutional settings, pass through the formal labor market, and use social services. Those released from custodial supervision may be easier to track through their contacts with criminal justice and other public agencies than through surveys. Thus linking administrative records may yield special benefits in the study of recidivism and
OCR for page 160
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics reentry. Such a system could connect, for example, records from different files of the NCRP, providing longitudinal records of movements from prison to parole, and return to prison. In contrast to the special studies of recidivism among release cohorts in 1983 and 1994, linked records of NCRP would provide an automatic and ongoing measure of reimprisonment and successful parole completion. In its most ambitious implementation, a system of linked records could join criminal justice to social service data such as unemployment insurance records or welfare enrollment. A broad linked system of administrative records would provide help to place released prisoners in a much broader social context, providing measures of their legitimate earnings, poverty status, and use of social services. There are two main obstacles to exploiting the potential of record linkage for gathering data on recidivism and reentry. Severe practical difficulties are associated with matching records from different databases for the same individual. Because identifiers differ across databases, record linkage is expensive and prone to error. A unified system of identifiers for a range of databases—say, all BJS correctional microdata collections—would unlock the potential of record linkage. The practical challenges to a unified system are substantial but we urge BJS to explore concrete steps in this direction. The other obstacle to large-scale record linkage, particularly linking criminal justice to social service records, relates to privacy protections. Because of the sensitivity of the linked data, BJS and cooperating agencies would need to take special steps to protect the confidentiality of records. Some kind of institutional review, monitoring data security and research data centers, may provide a process for ensuring the privacy of individuals recorded in a linked system of administrative records. Finally, administrative surveys of correctional, probation, and parole agencies could be expanded to explicitly incorporate policy interest in recidivism and reentry. Administrative surveys have been relatively inexpensive and accurate sources of counts of different correctional populations. The surveys could speak more directly to interests in recidivism and reentry by obtaining more detailed information about program participation and conditions of supervision. Enrollment counts of correctional populations in specific programs, and program spending and staffing information would help measure the resources applied to reintegration and criminal desistance. Some of this information is already reflected in surveys of correctional institutions, though spending on different categories of correctional programs is largely unmeasured. To advance understanding of the criminal justice system as an institutionalized influence on population processes (births, deaths, and migration), the administrative surveys could usefully collect more detailed demographic data. In addition to information about race and sex, which is currently reported for the prison, probation, and parole populations, a more detailed survey could obtain population counts by race and sex for given age
OCR for page 161
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics and education groups. Spatial data describing, say, counties of origin and destination for entering and released prisoners would also help map the spatial distribution of the criminal justice system’s reach into the community; the accuracy of those data would, of course, have to be evaluated. Understanding Prosecution and Declination Decisions The adjudication phases of the crime funnel model are areas of major “leaks” that are not well understood or measured. The percentage of cases that actually go to trial (enter the formal court system, where they might be more readily followed and tracked) can be small, and that percentage can vary strongly by jurisdiction and by type of court. Mechanisms for resolutions and alterations through “bargaining” at various stages are not well understood or studied: “charge bargaining” between attorneys prior to filing, “plea bargaining” as alternative to trial, “sentence bargaining” during or after trial. The prosecution function or component in the funnel includes the charging decision (including plea bargaining), filing decision, the pretrial custody decision, tests of evidentiary strength, and all pretrial motions (e.g. discovery). In addition to information on these decisions, a useful statistical system describing the prosecutorial function would contain data on the social organization of the prosecutorial and the defense process. This would include resources, such as the number of staff, but also the way in which those staff are assigned and organized. Is the chief prosecutor elected or appointed? Is the staff specialized by stage of litigation or crime type? Is the indigent defense bar staffed by public defenders or court-appointed attorneys? Since the nation has a state and federal justice system, statistical systems describing the prosecutorial function would address both levels. Currently BJS describes the state-level prosecution function with two different but related data collections—SCPS and NPS—which we summarized in Section 3–D; federal prosecution is described by the Federal Justice Statistics Program. There was a one-time survey of indigent defense in 1999. Some observations on the ways in which these collections cover (and do not cover) important parts of the prosecutorial function follow: The SCPS series does not cover a number of the decisions included in the prosecutorial function. The most important omission is the declination decision wherein the prosecutor decides not to file charges on arrests brought by the police or from some other source. This decision is not reviewable by anyone, and it is the single greatest source of prosecutorial discretion. Commenting on 1996 data from the NJRP that preceded SCPS, Forst (2000:26) lamented that: [The program] gives no information about cases rejected or dropped by prosecutors; more than a few people might like to
OCR for page 162
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics know why over 80% of all arrests for motor vehicle theft fail to end in conviction, and why about 60% of all arrests for robbery and burglary fail as well. On one hand, since the data collection in SCPS begins at filing, all of the decisions made prior to filing are lost. On the other hand, many important decisions made after filing are captured in this data collection. For those cases filed, it is possible to assess the amount of charge mobility that occurs from arrest to filing, through adjudication and sentencing. There is also extensive information on pretrial custody decisions and status, time that it takes to complete stages of processing, as well as some criminal history data. The key limitation of the NPS in understanding trends in prosecution is that it is to prosecution what the LEMAS survey is to law enforcement: a strictly partial look at basic administrative and management information. As an establishment survey, it has the same potential respondent selection problems (effects) as other surveys; there is also a certain inherent amount of noncomparability across prosecutorial units, because one can serve a single county and another an entire state. For what information it does provide on the dynamics of personnel and workload in prosecutor’s offices, the NPS has suffered as a measurement device because of its unstable periodicity (shifting from a 2-year to a 5-year cycle). It has also been unstable in the degree to which it has been conducted and treated as a sample or a census, as we described in Section 3–D; in the “census” years, content is particularly pared back, excluding all but the basic administrative questions. The Federal Justice Statistics Program operated by the Urban Institute with BJS sponsorship links administrative records across decision point in the federal justice system. It does provide the “flow” data that the “funnel” promises, and it does cover more of the decisions made by prosecutors than the SCPS series (including the declination decision). However, it is strictly limited to the federal justice system. There are senses in which prosecution and prosecutorial decisions should be amenable to data collection efforts, among them the fact that basic concepts and definitions are relatively invariant. In broad strokes, Forst (2000:22) observes that changes in prosecution “have mostly followed rather than led developments outside the prosecution domain. The basic nature and goals of prosecution, the role of the victim as witness in a matter between the state and defendant, the essential steps in processing cases through the courts and systems of public accountability have all remained fundamentally unchanged over the past 30 years.” However, a major reason for this resistance to procedural change—the relative insularity of prosecutors, as opposed to the police and elected officials whose work has a larger
OCR for page 163
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics profile—also serves to create a culture that works against openness in providing data. By and large, “the prosecutor’s work is invisible to the public at large” (Forst, 2000:23), and the prevailing inclination is to keep things that way. Another reason for prosecutorial insulation is adherence to a basic maxim of their adversarial culture: “Do not divulge the particulars of your case to anyone who is not in a position to help you win it.” Accordingly, prosecutors “typically see little to gain and considerable risk in divulging any information that is not required by law” (Forst, 2000:25). As a means to “improve the systems by which prosecutors are held accountable” and to make the operations of their offices more transparent, Forst (2000:42) argues for: the annual publication of uniform office performance statistics and a formal periodic survey of all who depend on prosecutors: victims, witnesses, judges, police, defense bar, and the general public. Private sector organizations have long used surveys to obtain systematic feedback about the effectiveness of service delivery, including measures of consumer satisfaction about specific elements of service. Police departments and other public agencies are turning increasingly to such assessment systems, and so can prosecutors. An effort along these lines should be coordinated by the Bureau of Justice Statistics to minimize political adulteration of the system, perhaps in collaboration with national associations of district attorneys and state attorneys general. Such an idea seems no more farfetched than that of a uniform crime reporting system with the cooperation of virtually all 20,000 independent police departments in the United States, a program that has been operating for most of the twentieth century. That vision remains far off. However, as noted above in Section 3–D, BJS’s investment in redesigning the SCPS program raises interesting possibilities. In working with local jurisdictions to participate in SCPS-type collections, it will also be important to assess whether prosecutor’s offices may be able to provide similar types of information. This is particularly the case if methods to work with electronic submissions from court and prosecutor databases continue to develop. As a short-term measure—and consonant with our advice to expand the concept of “law enforcement” data beyond the strict management focus of LEMAS (Section 3–F.2)—BJS should consider low-cost means to gather at least some procedural information in its existing NPS. A question or set of questions asking for basic counts of resolutions reached by the prosecutor’s office within some time window—ideally, broken down to include cases handled through alternative dispute resolution techniques such as mediation—would provide a partial picture of prosecutorial activity, but a fuller one than currently exists.
OCR for page 164
Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics This page intentionally left blank.