Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
â1â Introduction T HIRTY- FOUR YEARS AGO, a panel convened by the National Research Council (NRC) began work as requested by the U.S. Department of Justice, reviewing the departmentâs survey-based system to measure criminal victimization. The panelâs work was not preparatory to data col- lection, helping to design the survey from scratch, nor was it a review of a long-established program, as the full survey had been in the ï¬eld for only two years. Instead, the panelâs work was to serve as an early course correctionâa review of the objectives of what was then a suite of related victimization sur- veys, coupled with short- and long-term guidance on retooling the surveys to meet speciï¬ed objectives. Several of the recommendations in the previous panelâs ï¬nal report Surveying Crime (National Research Council, 1976b) were adopted fairly rapidly, including the elimination of victimization surveys of commercial es- tablishments and of selected large cities.1 The National Crime Survey was established as the name of the survey of sampled households. The report also contributed to a discussion of long-term redesign options starting in the early 1980sâinitiated by the Department of Justice and executed by a con- sortium of academic and government survey researchersâthat culminated in a reengineered (and renamed) National Crime Victimization Survey (NCVS) in 1992.2 1 The 1976 panelâs principal recommendations are listed in Appendix B. 2 The nomenclature of the surveyâand that of its sponsoring agencyâhas shifted over time. The most preliminary study plan for the survey in 1970 (Work, 1975) described the new program as the National Victimization Survey. When data collection began, the program was described as the National Crime Surveysâplural, since the program included national and city- 15
16 SURVEYING VICTIMS As presently implemented, the NCVS is a major national household sur- vey using a rotating panel sample of addresses: after an address is chosen for the survey, each person age 12 or older in the household at that address is interviewed seven times at 6-month intervals. The ï¬rst interview with a household is always done by personal visit, but subsequent interviews may be done by telephone if a number is available. The ï¬rst portion of the post- 1992 NCVS is a screening questionnaire, using detailed questions to elicit counts and basic information about crime victimization incidents in the pre- ceding 6 months. An incident report is then prepared for each incident detected in the screener, including a battery of questions on the context of each event. (The operations of the NCVS are described in more detail in Ap- pendix C.) In 2005, the NCVS was administered to approximately 38,600 households, yielding interviews with 67,000 people. Sponsored by the Bu- reau of Justice Statistics (BJS) of the U.S. Department of Justice, the ï¬eld collection of the NCVS is performed by the U.S. Census Bureau.3 Annually, BJS publishes reports and summary tables from the NCVS in two continuing report series, Criminal Victimization and Crime and the Nationâs House- holds (see, e.g., Catalano, 2006; Klaus, 2007). BJS also uses NCVS data as the basis for periodic or one-shot reports on a wide array of topics and victimization types, including carjacking (Klaus, 1999, 2004), ï¬rearm use in crime (Rand, 1994; Zawitz, 1995), perceptions of neighborhood crime (DeFrances and Smith, 1998), victimization of college students (Baum and Klaus, 2005), and workplace violence (Warchol, 1998). Over its 35-year historyâmajor highlights of which are listed in Box 1-1âthe NCVS has been a uniquely valuable source of information on crime. Intended to shed light on the âdark ï¬gure of crimeââthe phrase coined by Biderman and Reiss (1967) to describe criminal incidents that are not reported to policeâit is frequently used in conjunction with data from the Uniform Crime Reporting (UCR) program, through which the Fed- level surveys of businesses as well as national and city-level household surveys. The national household survey componentâoriginally referred to as the National Crime Panel programâ was the only part to survive into the late 1970s and became known as the National Crime Survey (singular), commonly abbreviated as NCS. This remained the standard nomenclature until the 1992 switch to National Crime Victimization Survey or NCVS. In this report, we use NCS in direct quotations when applicable, but generally use NCVS as the descriptor. 3 Another terminological note is in order for references to BJS. The organizational unit re- sponsible for developing and analyzing the survey wasâas of 1970âknown as the National Criminal Justice Statistics Center. By mid-decade, it was dubbed the National Criminal Jus- tice Information and Statistics Service (NCJISS) and was a constituent component of the Law Enforcement Assistance Administration (LEAA) of the U.S. Department of Justice. In 1984, reauthorization legislation dispersed the functions of the LEAA throughout a new Ofï¬ce of Jus- tice Programs (OJP) in the Justice Department, overseen by an assistant attorney general. The statistical and data-gathering functions of the former NCJISS were vested in the new Bureau of Justice Statistics (BJS). As with references to the NCVS, we tend to use âBJSâ to describe the agency, regardless of the exact date in question, except in direct quotations.
INTRODUCTION 17 Box 1-1 Major Steps in the Development of the National Crime Victimization Survey ⢠1967: The Presidentâs Commission on Law Enforcement and Administration of Justice (1967) issues its ï¬nal report. The report contrasts results of its 10,000-household National Survey of Criminal Victims (conducted by the National Opinion Research Center), as well as city-level victim surveys for Washington, Chicago, and Boston (conducted by the Bureau of Social Science Research and the University of Michigan Survey Research Center), with the FBIâs Uniform Crime Reports. Concluding that lack of information on offenses not reported to police makes it difï¬cult to develop useful crime policy, the commission recommends the development of a nationwide crime victimization survey. Procedurally, the commissionâs recommendations also lead to the 1968 establishment of the Law Enforcement Assistance Administration in the U.S. Department of Justice. ⢠1968: Initial discussions between the LEAA and the Census Bureau take place on a national survey of crime victims. The Census Bureau conducts two staff workshops on the scope and basic design of such a survey, resulting in a conference report (U.S. Bureau of the Census, 1968). ⢠1970â1971: Pretestingâsmall-scale tests, most involving reverse record checks (interviews with known victims of crimes reported to police), are conducted in Washington, DC; Akron, Cleveland, and Dayton, Ohio; Baltimore, Maryland; and San Jose, California. Testing in San Jose and Dayton included the ï¬elding of surveys to a sample of about 5,000 households in each site. ⢠1971â1972: Initial implementation as supplementâvictimization questions are added to the Census Bureauâs Quarterly Household Survey. The ï¬rst such nationwide sample, in January 1971, had a sample size of 15,000 housing units and asked respondents to recall events within a 12-month reference period. (The same supplement was used in 1972, using a 6-month reference period.) Subsequent iterations were used to reï¬ne the design of questionnaires and instructions. These surveys-as-supplements were intended only as an experiment and not the basis for published reports. ⢠July 1972: The new National Crime Surveysâjointly developed by the LEAA and the Census Bureauâare ï¬elded. The core National Crime Panel is a sample of 72,000 households; one-sixth of the sample were to be interviewed monthly and then again at 6-month intervals. The suite of surveys also includes a national sample of 15,000 businesses, as well as a sample of 12,000 households and 2,000 businesses in each of 26 major cities, to support local-area estimates of victimization. These âCities Surveysâ begin in 8 âimpact citiesâ and are ï¬elded in sets of the cities over the next 3 years. ⢠1976: The Panel on the Evaluation of Crime Surveys of the National Research Council (1976b) releases its ï¬nal report, Surveying Crime. The major recommendations of the panel are summarized in Appendix B; among the recommendations is that the Cities Surveys be abandoned in order to support the core national household sample. ⢠1986: Redesign phase-inâfollowing the recommendations of a redesign consortium, small-scale changes to the survey and its instruments are implemented, with the larger scale changes originally intended to be implemented in 1989.
18 SURVEYING VICTIMS Box 1-1 (continued) ⢠1988â1989: Redesign pretestâa three-wave national pretest of the redesigned victimization survey is ï¬elded, using a 6-month reference period and bounded data. ⢠January 1989: A 5 percent test group of respondents receive the redesigned survey questionnaire, including a revised set of screener questionnaires. The ï¬rst major supplement to the victimization surveyâthe School Crime Supplement (SCS)âis also implemented and is conducted through June 1989. ⢠1990: Gradual phase-in of new questionnaire continues, with another 5 percent added to the sample. ⢠July 1991: Renamingâthe survey name is changed to National Crime Victimization Survey (NCVS), to refer to the data collected using the new questionnaire. ⢠January 1992: Conversionâthe sample receiving the redesign/NCVS questionnaire is elevated to 50 percent, providing a year of half-samples using the old and new forms. In July 1993, the redesigned NCVS questionnaire becomes the norm and the former NCS questionnaire is discontinued. For calendar year 1992, the Survey of Public Participation in the Arts (funded by the National Endowment for the Arts) is appended to the NCVS as a supplement. ⢠Mid-1990s: Context variables on public housing, college or university housing, and attendance or employment at colleges or universities are added to the screening questions. The SCS is replicated, with joint sponsorship by BJS and the National Center for Education Statistics, in 1995. Another supplementâthe Police-Public Contact Survey (PPCS) is ï¬rst ï¬elded on a pilot basis (for outgoing rotation households) in MayâJuly 1996. ⢠May 1997: BJS elects to suspend production of state-level tabulations (beginning with 1996 data) and preliminary estimates previously published in Crime in the Nationâs Households (beginning with 1997 data). ⢠1998â1999: A parallel, 12-city survey conducted only by random-digit dialing is collected from February through May 1998; the survey is conducted by the Census Bureau and sponsored by BJS and the Ofï¬ce of Community Oriented Policing Services. The SCS and PPCS supplements are repeated in 1999, and hate crime questions are introduced in the main NCVS in the same year. ⢠2001: Questions on cybercrime are added to the NCVS, as are incident-report questions that relate to a Workplace Risk Supplement (requested by the National Institute for Occupational Safety and Health) ï¬elded in 2002. ⢠January 2003: Race and ethnicity questions are revised, pursuant to the 1997 directive issued by the Ofï¬ce of Management and Budget. ⢠2004: Identify theft questions are added to the NCVS. ⢠January 2006: The Supplemental Victimization Survey (SVS), focusing on stalking, is ï¬elded from January to June.
INTRODUCTION 19 eral Bureau of Investigation compiles counts of crimes reported to law en- forcement agencies nationwide (see Appendix D for additional detail on the UCR). However, it has also served a vital function by focusing attention on the characteristics of victims of crime, including the nature of those incidents and how people respond to them. Highly inï¬uential in shaping contemporary knowledge and theories of criminal victimization, NCVS data have informed assessments of victimiza- tion risk, victim-offender relationships, weapon use, and injury in violent crime; the costs of violent and property victimization; and the degree to which crimes are reported to the police (and the reasons for reporting or not reporting). Through the basic survey and topical supplements, NCVS data have also offered insights on new types of criminal behavior such as identity theft (Baum, 2007) as well as violence types of increasing policy importance (e.g., intimate partner violence and crime in school settings). Signiï¬cantly, the NCVS has also served âas a model for victimization surveys implemented throughout the worldâ (as well as victimization studies in individual states or cities) âbecause it incorporates many innovative methodological protocols that enhance its ability to produce reliable estimates of the nature and ex- tent of criminal victimizationâ (Rennison and Rand, 2007:17). (Appendix E describes some of these foreign and state-level victimization surveys.) Although the history of the NCVS has been marked by achievement, it has also been markedâalmost from the surveyâs inceptionâby a certain level of unanticipated changes in design through budget cuts. The survey has been subject to a lengthy string of sample size reductions and changes in pro- cedure, typically in the interest of reducing costs. Many of these are itemized in Box 1-2. In recent years, the problems of keeping the NCVS operational have been particularly acute because BJSâthe surveyâs sponsorâhas been effectively ï¬at-funded since 2000. As shown in Figure 1-1, BJSâlike many federal agenciesâhas consistently received less in total appropriations than requested, but appropriated funds have been particularly tight over most of the past decade. Howeverâeven with all of the cost-cutting measures de- scribed in Box 1-2âthe cost of implementing the NCVS has continued to grow. In part, this is attributable to the ï¬xed costs of converting the survey to electronic forms to facilitate computer-assisted interviews and to increased labor costs, but it is also due to the increasing difï¬culty of contacting survey respondents and gaining their cooperation. For BJS, the net consequence of these ï¬scal constraints has been the painful trade-off decisions required to support a full suite of justice-related data collections while a single program has consumed the bulk of the agencyâs limited resources. Aside from victimization, BJS maintains major, core data collection efforts in the areas of corrections (including censuses and surveys of jails and prisons as well as data on probation and parole pro- grams), law enforcement management and administration, courts and sen-
20 SURVEYING VICTIMS Box 1-2 Notable Reductions in the National Crime Victimization Survey ⢠1977: The commercial component of the National Crime Surveysâincluding measurement of commercial burglary and robberyâis eliminated, principally for budgetary reasons but also due to the inadequacy of the business sampling frame. ⢠January 1980: Personal visit interviewing is reduced and telephone interviewing introduced. Interview 1 with a household is still required to be conducted in person, but interviews 2, 4, and 6 are directed to be conducted by telephone to the extent possible. ⢠June 1984: A sample size cut is made, resulting in the reduction of nonself- representing strata from 220 to 153 and a 20 percent reduction in the larger of the 153 self-representing primary selection units. According to the Demographic Surveys Division, U.S. Census Bureau (2007b:25), this cut was imposed âso that funds could be redirected to pay for redesign research.â ⢠February 1986: Due to budget problems, reinterview activities are suspended for one month and reduced by half for seven additional months. ⢠March 1986: To reduce costs, the protocol for personal visit interviewing rather than by telephone is revisited. Telephone interviewing for rotation groups 3 and 7 is increased, with only interviews 1 and 5 slated for personal visit. ⢠Second half, 1992: Due to funding constraints, discussions begin with the Census Bureau on possible ways to reduce the costs of the NCVS. To cut costs, reinterviews were stopped for one month in August 1992, and a 10 percent sample size reduction was implemented in October 1992 (affecting both the groups receiving the NCVS and NCS questionnaires). ⢠October 1996: A 12 percent sample cut is imposed, due to budget constraints. The cut came after a summer of some computer-assisted telephone interviewing (CATI) follow-up work (for noninterviews) being suspended due to a budget impasse and was suggested by the Census Bureau (along with two other options) as a âtwo-year plan.â ⢠July 2000: Lifestyle and home protection questions are dropped from the screener portion of the questionnaire. ⢠April 2002: The NCVS sample is reduced by 4 percent. ⢠January 2006: The NCVS sample reduced by 16 percent. ⢠July 2007: Due to budget constraints, three major changes are slated to occur in summer 2007: (1) use data from the ï¬rst interviewâpreviously withheld as a âboundingâ caseâin annual estimates, (2) implement a 14 percent sample cut, as a balance for using the bounding ï¬rst interviews, and (3) suspend all computer-assisted telephone interviewing from Census Bureau call centers (however, ï¬eld interviewers may still use the telephone to conduct their scheduled interviews).
INTRODUCTION 21 Figure 1-1 Bureau of Justice Statistics budget requests and ï¬nal total appropriations, 1981â2006, and National Crime Victimization Survey data collection costs, 1990â2006 NOTE: Budget request and appropriations ï¬gures include both base and program costs. Final appropriations reï¬ect amounts after any applicable budget recission or across-the-board cut. NCVS data collection costs exclude additional costs for developing and respecifying sample based on new decennial censuses, as well as costs associated with the automation (conversion to computer-based administration) of the survey. SOURCE: Data provided by the Bureau of Justice Statistics as brieï¬ng for the panelâs ï¬rst meeting in February 2007. tencing (including series on both civil and criminal cases at the state level, as well as data on all steps of processing in federal criminal cases), and expen- diture and employment in the justice system. However, during ï¬scal years 2001 through 2006, the NCVS consumed at least 51.2 percent (in both 2001 and 2002) and as much as 64.0 percent (in 2004) of the total BJS appropri- ations. For the American public, the consequence has been the degradation of a key social indicator: rising data collection costs have led BJS to re- duce the sample size of the survey and induce other less visible cost-cutting measures. Over the past decade, two trendsâthe diminishing NCVS sample size and generally low and decreasing estimated overall victimization ratesâ have combined, with the result that only large percentage changes in violent crime victimization ratesâat least 8 percentâwould be a statistically signif- icant year-to-year change.4 As noted in the U.S. Ofï¬ce of Management and 4 This ï¬gure is from a presentation by Michael Rand, BJS, at the panelâs ï¬rst meeting; for 2005, the violent crime victimization rate was estimated as 21.2 per 1,000 population, and the 95 percent conï¬dence interval of this rate is ±8.1 percent. Since 2000, the estimated annual
22 SURVEYING VICTIMS Budget (2007:8) annual review of statistical program funding, âcost cutting measures applied to the NCVS continue to have signiï¬cant effects on the precision of the estimatesâyear-to-year change estimates are no longer fea- sible and have been replaced with two-year rolling averagesâ in BJS reports on victimization. The predecessor panel offered an early course correction to the still-new National Crime Survey, noting that (National Research Council, 1976b:9â 10): A surveyâor any measurement processâshould be designed to meet stated objectives in so far as feasible. Evaluation of the survey or the measuring instrument should in large part be an assessment of the de- gree to which it meets or can meet objectives. That panelâs work was furthered by the NCS redesign consortium in the 1980s, culminating in a phased-in implementation of a redesigned NCVS in 1992. The NCVS and its design has beneï¬ted from a steady stream of methodological research and reï¬nement since its inception; now, some 35 years later and recognizing the challenges facing the NCVS and other survey data collections, the question of whether the survey is meeting its goals and user base needs is ripe for review. 1âA CHARGE TO THE PANEL The Bureau of Justice Statistics requested that the National Academiesâ Committee on National Statistics (in cooperation with the Committee on Law and Justice) convene this Panel to Review the Programs of the Bureau of Justice Statistics with a broad charter to review the full suite of BJS pro- grams. The panel has been tasked to review all of the agencyâs varied data collections, with an eye toward gaps in substantive coverage that should be ï¬lled as well as collections that should be suspended or dropped. The full charge to the panel is to: examine the full range of programs of the Bureau of Justice Statistics (BJS) in order to assess and make recommendations for BJSâ priorities for data collection. The review will examine the ways in which BJS statistics are used by Congress, executive agencies, the courts, state and local agencies, and researchers in order to determine the impact of BJS programs and the means to enhance that impact. The review will assess the organization of BJS and its relationships with other data gathering entities in the Department of Justice, as well as with state and local violent crime victimization rates have dipped from 27.9 to 21.2 per 1,000, and the 95 percent intervals have been in excess of ±7 percent for each of those annual estimates. See Figure C-1 in Appendix C.
INTRODUCTION 23 governments, to determine ways to improve the relevance, quality, and cost-effectiveness of justice statistics. The review will consider priority uses for additional funding that may be obtained through budget ini- tiatives or reallocation of resources within the agency. A focus of the panelâs work will be to consider alternative options for conducting the National Crime Victimization Survey, which is the largest BJS program. The goal of the panelâs work will be to assist BJS to reï¬ne its priorities and goals, as embodied in its strategic plan, both in the short and longer terms. The panelâs recommendations will address ways to improve the impact and cost-effectiveness of the agencyâs statistics on crime and the criminal justice system. [emphasis added] Given the prominence of the NCVS in BJS operationsâand its domi- nance of BJS budget resourcesâthe panel was speciï¬cally asked to evaluate options for conducting the NCVS in our ï¬rst year of work,5 before turn- ing to the agencyâs data collections related to other areas, like corrections and judicial processing. Consistent with this principal task of the report, it is important to make clear that this report is not intended to be a com- plete sourcebook on the NCVS; it is neither a full procedural history of the survey, a complete literature review of its uses, nor a detailed operational plan for any speciï¬c alternative design.6 This report is also not meant to revisit in full detail the comprehensive redesign efforts that culminated in the ï¬elding of the new NCVS instrument in 1992. That redesign effort is re- ported thoroughly by Biderman et al. (1986) and Bureau of Justice Statistics (1989) and major issues faced in the redesign are also summarized by Sko- gan (1990). It generated a large body of valuable methodological research, much of which exists in technical BJS and Census Bureau memoranda. We do not reprise that work in detail in this report, but the recommendations we make do suggest the need for an ongoing evaluation system to produce a similar, rich body of updated methodological work to inform future NCVS design changes. Due to the nature of our panelâs charge, it is also essential to under- score that this is a ï¬rst-stage or interim report. As we discuss further in Section 3âD, the methodological focus of this report means that we do not attempt as exhaustive a listing of constituencies and uses for the NCVS as our overall charge suggests; we intend to consider a more complete assessment of the user base in our ï¬nal report. Furthermore, this report does not and is not intended to provide comprehensive treatment of all BJS programs, nor 5 In âStrengthening Federal Statistics,â Appendix 4 of the presidentâs proposed budget for ï¬scal year 2008, one of two initiatives speciï¬cally referenced in the BJS budget request is âa redesign of the National Crime Victimization Survey based on anticipated recommendations from the Committee on National Statistics of the National Research Council,â i.e., this panel. 6 The two volumes edited by Lehnen and Skogan (1981, 1984) are an important resource on the early history of the survey.
24 SURVEYING VICTIMS does it completely reconcile NCVS components with other BJS data collec- tion activities. As this panel continues the work of suggesting data collection priorities based on a fuller review of the suite of BJS programs, it is possible that some NCVS-related issues will have to be revisited in the ï¬nal report. 1âB REPORT CONTENTS Following this introduction, Chapter 2 reviews the current and historical goals of the NCVS, describing the various roles that the survey has been ex- pected to fulï¬ll. Chapter 3 then uses the goals of the NCVS to describe the major technical and operational issues surrounding the current conduct of the survey. In Chapter 4, we lay out several sets of goals for a reconï¬gured NCVS and describe the design choices that follow directly from particular sets of goals and priorities. We close in Chapter 5 with a look forward, mak- ing recommendations on how best to calibrate a victimization measurement system for the 21st century. Appendix A reproduces the panelâs recommendations for ease of refer- ence andâas previously notedâAppendix B recounts the principal ï¬ndings and recommendations of National Research Council (1976b). As a reference for the reader and to enhance the ï¬ow of the main text, we have placed detailed descriptions of the NCVS and related programs in appendixes. Ap- pendix C describes the sample design of the NCVS, as well as the interview- ing procedures for the survey and the content of the survey instrument. In Appendix D, we describe other data resources that are available for studying aspects of criminal victimization; a major portion of that appendix describes the Uniform Crime Reporting program of the Federal Bureau of Investiga- tion whichâas the national inventory of crimes reported to policeâis the data source to which the NCVS is most commonly compared for the pur- pose of assessing crime trends in the United States. Finally, in Appendix E, we describe other existing victimization surveys, such as those conducted in foreign countries (particularly the British Crime Survey) and those that have periodically been ï¬elded in American states and cities.