The crimes of rape and sexual assault are among the most injurious that perpetrators can inflict on other individuals. These crimes are devastating, extending beyond the initial victimization to such consequences as unwanted pregnancy, sexually transmitted infections, post-traumatic stress disorder, depression, flashbacks, sleep disorders, eating disorders, post-incident substance abuse, self-harm, and even suicide (Rape, Abuse and Incest National Network, n.d.). The effects are often long lasting and can lead to health and work productivity issues for years. Using analyses from three different studies of the cost of crime, Heaton (2010) estimated that the victim-related costs1 for nonlethal rape and sexual assault are between $150,000 and $283,626 per victim.
Understanding the frequency and context under which rape and sexual assault are committed is vital in directing law enforcement and victim-support resources. These data can influence public policy in the areas of public health, mental health, and education. They also can be used to identify and implement interventions that will reduce the risk of future victimizations. Unfortunately, accurate information about the extent of rape and sexual assault is particularly difficult to obtain because these crimes are seriously underreported to law enforcement.
1Total victim-related costs include both tangible and intangible costs. The tangible costs include such things as medical treatment and lost wages. Intangible costs are larger and much harder to estimate. They include such things as lost quality of life resulting from fear and psychological effects of victimization.
In Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics, the National Research Council (NRC) made a strong recommendation to the Bureau of Justice Statistics (BJS) regarding its mission: “[It] must ensure that the nation has quality annual estimates of levels and changes in criminal victimization” (National Research Council, 2009, p. 10). A major tool in accomplishing this mission is the National Crime Victimization Survey (NCVS), an ongoing BJS survey of the noninstitutional population of the United States to estimate victimization rates2 by type of crime, along with details about the victims and the social context of those victimizations. The NCVS provides an independent source of information on criminal victimization—a source of data for the calibrating of police-reported incidents and an indicator of the extent that victimization incidents go unreported to law enforcement. “Since the survey began full-scale data collection in the early 1970s, the NCVS has become a major social indicator for the United States. Serving as a complement to the official measures of crime reported to the police … the NCVS has been the basis for better understanding the cost and context of criminal victimization” (National Research Council, 2008, p. 2).
The NCVS is an omnibus victimization survey. As such, it has a broad mandate and focus to include a wide array of different types of victimizations, including both crimes against people and crimes against property. This is a difficult task, and approaches that may be overall best for a general survey may be less optimal for measuring a specific type of victimization.
This report focuses on a narrow portion of the NCVS: the estimation of incidence rates for rape and sexual assault. There is controversy as to whether the NCVS is providing accurate estimates of the rates of rape and sexual assault, in part because some other sources of statistics over the past two decades have shown higher levels of victimization than those estimated through the NCVS and its predecessor, the National Crime Survey.
A 1992 report by the National Victim Center in collaboration with the Crime Victims Research and Treatment Center, Rape in America, A Report to the Nation (Kilpatrick, Edmunds, and Seymour, 1992), based on the National Women’s Study (NWS), estimated that 683,000 adult women were raped during a 12-month period in 1989 and 1990. That study estimated that only 16 percent of those rapes were reported to the police. For a similar time period, the Uniform Crime Reporting (UCR) Program of the Federal Bureau of Investigation (FBI) reported 102,500 rape victims (a figure that is quite similar to the 16 percent of rapes reported to police in the NWS).3
2See footnote 1 in the Summary.
3The UCR compiles counts of incidents of crimes from participating law enforcement agencies and so is necessarily limited to counting crimes reported to the police.
The NCVS (then known as the National Crime Survey [NCS] prior to its redesign) estimated 130,000 victims for approximately the same time period. Thus, the data from the NWS indicated a level of victimization that was more than five times that reported in the NCVS.4
More recently, the Centers for Disease Control and Prevention launched a major survey on the subject, the National Intimate Partner Violence and Sexual Violence Survey (NISVS). This survey has a number of objectives, one of which is to provide estimates of the extent of rape and sexual assault. In 2010, it counted 1,270,000 victims of rape and attempted rape,5 many times higher than the NCVS estimate of victimizations (188,380) that same year (Bachman, 2012; Black et al., 2011; Bureau of Justice Statistics, 2011).
Thus, an important barrier to understanding the extent of rape and sexual assault in the United States is the existence of multiple sources of information providing different answers. These surveys and programs have somewhat different objectives, are conducted within a different “context,” and use different measurement tools. The end result is that they provide different estimates of the extent of rape and sexual assault. This in turn, creates confusion for the public, for law enforcement, for policy makers, for researchers, and for victim advocacy groups.
BJS has committed to a multiyear project to better understand the magnitude and reasons for potential underreporting of rape and sexual assault on the NCVS and the reasons for the differences between NCVS estimates and data on rape and sexual assault from other sources. As part of this effort, BJS asked the Committee on National Statistics of the NRC to convene an expert panel to investigate these issues and recommend best practices for measuring the incidents of rape and sexual assault from BJS household surveys and to assess the quality and relevance of NCVS statistics on rape and sexual assault. The Panel on Measuring Rape and Sexual Assault in Bureau of Justice Statistics Household Surveys was appointed to carry out the task. The panel was asked to examine such issues as legal definitions of rape and sexual assault and best methods for operationalizing these definitions in survey instruments so that they are understood by respondents; to examine the factors that affect underreporting and propose solutions that can minimize that problem; and to look at ways that can ensure that
4The quality, methods, and definitions used in the two surveys differed, and the NCS has since been redesigned as the current NCVS.
5The NISVS has a target population of males and females 18 years or older. It has a broader definition of rape than does the NCVS.
Statement of Task
An ad hoc panel will examine conceptual and methodological issues surrounding survey statistics on rape and sexual assault and recommend to the Bureau of Justice Statistics best methods for obtaining such statistics on an ongoing basis. The panel will assess the quality and relevance of statistics on rape and sexual assault from the National Crime Victimization Survey and other surveys contracted for by other federal agencies as well as surveys conducted by private organizations. Issues to be examined include legal definitions in use by the states for these crimes, best methods for representing the definitions in survey instruments so that their meaning is clear to respondents, and best methods for obtaining as complete reporting as possible of these crimes in surveys, including methods whereby respondents may report anonymously. The panel will organize a workshop and commission papers as principal means of gathering information to support its deliberations. The panel will issue a report with its findings and recommendations at the conclusion of a 21-month study. The panel’s scope of work will not include surveys in nonhousehold, institutional settings, such as prisons.
respondents are able to report privately and anonymously on the NCVS. The formal charge to the panel is shown in Box 1-1.
In addressing the panel at its initial meeting, James P. Lynch, then the director of BJS, provided background for the panel’s work. His statement is presented in Appendix A. He stressed the importance of having crime statistics that are generated independently of, and in addition to, those statistics provided through police reports. He said that he believes that this is particularly important for rape and sexual assault “since there is good evidence that the majority of these offenses are not reported to the police. These offenses remain the darkest of the ‘dark figure’ of crime.” The National Research Council (2008) and others (Baumer and Lauritsen, 2010; Lauritsen and Heimer, 2008) have also highlighted the importance of the NCVS. They have found that no other national data on criminal victimization are collected with this level of detail and published annually.
Addressing rape and sexual assault specifically, Lynch said that two approaches have evolved to measuring these crimes through surveys, with different conceptual definitions and methodologies. Not surprisingly, these two approaches provide different estimates of victimization, which in turn provides confusion for the public. Lynch explained:
One group emphasizes the criminal justice perspective and the other takes a public health approach. The criminal justice school emphasizes crime as a point in time event and employs legal definitions (but plain language
descriptions) of the target behavior…. The public health approach emphasizes victimization as a condition that endures over time and requires treatment to restore the victim. Consequently, there is less concern with identifying point-in-time events that may comprise the condition and legal definitions are of less concern than commonly understood definitions of the behavior.
In broad terms, the NCVS represents the criminal justice perspective, and the NISVS and other surveys described in this report represent the public health perspective.
Lynch expressed several specific expectations for the panel:
• Take a fresh look at the problem, drawing from what the criminal justice and public health schools have done but not being held captive by these traditions. The principal goal of the panel is to consider a wide range of alternative self-report survey designs to measure the incidence and prevalence of the crimes of rape and sexual assault and to recommend an optimum design.
• Recommend whether this optimum design can be incorporated into the ongoing NCVS program and, if so, how.
• Finally, to work closely with Westat in field testing the redesign options.
Thus, Lynch said he hoped for considerable flexibility from the panel in looking for “best practices.” He asked the panel to take a fresh look at this problem, drawing from approaches that have been successful in both the criminal justice and public health approaches. He noted that he does not assume that the optimum design could be incorporated into the ongoing NCVS program.
The panel addressed its charge in two main phases, with the initial phase to gather relevant information. It completed a careful assessment of the design, implementation, and output from various administrative and survey data systems that have been used to produce estimates of rape and sexual assault in the United States. Outside investigators doing work in related areas were commissioned to prepare papers for a public workshop held in June 2012. (Agendas for the workshop and public meetings of the panel are presented in Appendix B.) The panel also consulted Surveying Victims: Options for Conducting the National Crime Victimization Survey (National Research Council, 2008) to enhance its understanding of issues and to identify any recommendations from that study that are relevant today for measuring rape and sexual assault. Beyond a review of available documentation, this phase included a field exercise in which several panel
members participated as respondents in mock interviews on the NCVS (criminal justice perspective) and the NISVS (public health approach).
In phase two, the panel developed a prioritized list of ideas that had potential to improve the quality of survey estimates of rape and sexual assault. These ideas reflected all components of total survey error, including imperfect sampling frames, inefficiency in sample selection, nonparticipation by sampled households and individuals, misspecification and other measurement problems, and processing errors. From these ideas, the panel developed the key elements of an “optimum” design for the measurement of rape and sexual assault. Some features of the recommended design closely resemble the current design of the NCVS, while others would require BJS to move away from the NCVS for implementation. The panel’s recommendations provide BJS with specific guidance on key aspects of this design.
As requested by Director Lynch, the panel worked publicly with investigators at Westat, which had been contracted by BJS to develop a pilot project to test two alternative survey designs to measure rape and sexual assault. Westat staff presented the status of their work at each of the panel’s open meetings and participated in open discussion at those meetings with panel members and other participants. Following the June 5-6, 2012, public workshop, several panel members provided individual informal comments to Westat on the draft plans that Westat presented at that public workshop. They are provided for the purpose of full disclosure in the Public Access File (see http://www8.nationalacademies.org/cp/information.aspx?key=Internet_FAQ[March2014]). The Westat team and the panel kept each other advised of their project timelines for various activities throughout the process. Following NRC policy, there was no sharing of the panel’s deliberations, conclusions, and recommendations with Westat or BJS during this study.
Note that it was beyond the charge of this panel to compare estimates from the NCVS for other types (beyond rape and sexual assault) of victimizations to examine whether the NCVS possibly underestimated or overestimated these victimizations. The NRC has completed other studies of the broader NCVS (National Research Council, 2008, 2009).
The panel is charged with recommending best practices for measuring the incidences of criminal victimization of rape and sexual assault through BJS household surveys. This charge was not intended to nor did it restrict the panel to only considering solutions within the structure of the NCVS. A critical first step is to establish a broad understanding of the conceptual definitions for the terms rape and sexual assault. A second step is to look at how these concepts are measured operationally in existing data systems
(including, but not limited to, the NCVS)—the methods used and the results obtained.
In pursuit of these goals, Chapters 2 through 6 of the report first consider definitions of rape and sexual assault and their legal histories and then detail several important sources of data on these victimizations.
In order to design a national survey on rape and sexual assault, consistent definitions for these criminal victimizations have to be defined. This task is complicated by the fact that these crimes are generally based on state, rather than federal, statutes. Chapter 2 explores rape and sexual assault in a legal context, analyzing the components of existing legal definitions and their differences and commonalities across jurisdictions. It also covers the historical context from which modern laws against rape and sexual assault have evolved and the changes in those statutes over time. The purpose of this chapter is to look for commonalities across jurisdictions that would be important to include in operational definitions.
The next four chapters detail the data that are available and the methods used to obtain them. Chapter 3 describes the statistical information about crimes available from law enforcement agencies. It looks at police incident reports and the FBI’s UCR. This system provides the official measure of crimes reported to the police, and it thus provides an important baseline for comparing other sources of survey-based data.
Addressing the estimation of victimization from population surveys, Chapter 4 provides a description of the NCVS. It includes a review of the survey’s history, methodology, and implementation, as well as the survey’s resulting estimates of rape and sexual assault. Chapter 5 looks at four other important surveys of rape and sexual assault that have been conducted over the past 25 years. They have used different methods and produced different results.
Chapter 6 compares and contrasts the data discussed in the previous three chapters, focusing on methods and results. It offers the panel’s conclusions that inform the central part of our charge to propose improvements to the design of BJS household surveys that measure rape and sexual assault.
In the second half (Chapters 7 through 10) of the report the panel turns to in-depth analyses of the NCVS and its adequacy for the goal of accurately measuring rape and sexual assault. It is important to note that the report does not provide the same in-depth evaluation of the other sources of data described in Chapter 5. Chapters 7, 8, and 9 focus on the NCVS as the current vehicle through which BJS measures victimization rates for rape and sexual assault. This focus on the NCVS reflects a prioritization of the panel’s time and resources: it does not imply that the panel believes that the other sources have better measures of rape and sexual assault or are subject to fewer errors.
error, specification error, imperfect sampling frames, nonresponse, measurement error, and error in data processing. For each of these error types, the panel has evaluated the potential to generate errors in estimates of rape and sexual assault so that solutions could be identified.
Chapter 9 summarizes the analyses from Chapters 7 and 8, clarifying which potential errors may have the largest effects on the reported estimates. The chapter specifically identifies four major obstacles for accurately estimating incidents of rape and sexual assault in the current NCVS, which are the basis for the panel’s conclusions and recommendations in Chapter 10. The chapter also includes four recommendations for BJS.
Chapter 10 details the panel’s conclusion that the NCVS is not an adequate vehicle for the goal of accurate measurement and presents the panel’s recommendations for best practices, including a recommendation for a separate survey to measure rape and sexual assault victimizations. It provides guidelines on the optimum design of this new survey, as well as lower-cost variations. It also includes recommendations for specialized training and monitoring, research, and enhanced communication with data users.
Of special note, this report uses the terms “low incidence” and “statistically rare” to describe the criminal victimizations of rape and sexual assault because the frequency with which they occur makes them difficult to measure in a household population survey. The report explores a number of statistical practices that better measure rare attributes in a population. The panel’s position is that these victimizations are critically important to measure accurately and the terms “low incidence” or “statistically rare” do not diminish that importance.