6


Comparison of Rape and Sexual
Assault Across Data Sources

The sources of data on sexual victimizations discussed in this report have different foci, use different methodologies, and provide different results. At this point, definitive conclusions regarding which data source produced the most accurate estimates of rape and sexual assault would be useful. However, the panel acknowledges that it cannot scientifically make such conclusions in this report. The first barrier to such conclusions: the panel focused on the National Crime Victimization Survey (NCVS) and not the other sources. It examined the NCVS under the structure of total survey error (Chapters 7, 8, and 9), and these analyses allowed the panel to draw conclusions and make recommendations regarding the NCVS. The panel did not provide the same deliberative focus on each of the other sources of data, in part because of limited time and resources. Thus we know a good deal about the potential errors in the NCVS and much less about the potential errors in the other sources. Specifically, this does not mean that the other sources have fewer potential errors, only that these errors are not analyzed in this report. A second barrier: the target populations and definitional constructs (of what was being measured) are inconsistent across sources. It is a case of comparing apples and oranges. Third, only the NCVS and the Uniform Crime Reports (UCR) provided estimates over time, with the other sources providing estimates only for specific (and different) points in time. Thus, the complexity of different concepts, measurement approaches, and timing made definitive comparisons very problematic, and the panel did not have the time and resources available to attempt such a task.

With that said, a better understanding of the differences between these



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



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 91
6 Comparison of Rape and Sexual Assault Across Data Sources T he sources of data on sexual victimizations discussed in this report have different foci, use different methodologies, and provide dif- ferent results. At this point, definitive conclusions regarding which data source produced the most accurate estimates of rape and sexual as- sault would be useful. However, the panel acknowledges that it cannot scientifically make such conclusions in this report. The first barrier to such conclusions: the panel focused on the National Crime Victimization Survey (NCVS) and not the other sources. It examined the NCVS under the struc- ture of total survey error (Chapters 7, 8, and 9), and these analyses allowed the panel to draw conclusions and make recommendations regarding the NCVS. The panel did not provide the same deliberative focus on each of the other sources of data, in part because of limited time and resources. Thus we know a good deal about the potential errors in the NCVS and much less about the potential errors in the other sources. Specifically, this does not mean that the other sources have fewer potential errors, only that these errors are not analyzed in this report. A second barrier: the target populations and definitional constructs (of what was being measured) are inconsistent across sources. It is a case of comparing apples and oranges. Third, only the NCVS and the Uniform Crime Reports (UCR) provided esti- mates over time, with the other sources providing estimates only for specific (and different) points in time. Thus, the complexity of different concepts, measurement approaches, and timing made definitive comparisons very problematic, and the panel did not have the time and resources available to attempt such a task. With that said, a better understanding of the differences between these 91

OCR for page 91
92 ESTIMATING THE INCIDENCE OF RAPE AND SEXUAL ASSAULT sources and their measurement approaches can lead to improvements in the measurement of rape and sexual assault on Bureau of Justice Statistics (BJS) surveys. Therefore, this chapter, drawing on Appendix D and Bachman ­ (2012), summarizes and highlights what the panel learned from the com- parisons among the five surveys and one administrative source covered in this report. 1. UCR summary system (ongoing), 2. NCVS (ongoing), 3. National Women’s Study (NWS) (1989-1990), 4. National Violence Against Women Study (NVAWS) (1995), 5. National College Women Sexual Victimization Study (NCWSV) (1997), and 6. National Intimate Partner and Sexual Violence Survey (NISVS) (2010, and perhaps ongoing). Our comparisons are discussed in terms of five factors: the definitions used for rape and sexual assault; context in which data are collected; target population, sampling frame, and sample size; data collection mode, re- sponse rates, and adjustments for nonresponse; and the resulting measures of victimization. DEFINITIONS USED The definitions used for rape vary, sometimes substantially, among the six data sources (see Table 6-1). The table also shows whether the source collected information on attempted rape and other forms of sexual assaults as well as rape. The UCR definition (used through 2012) is clearly the most restric- tive. It restricts rape counts to male on female attacks with penile-vagina penetration. Attempted rapes are counted, but all other forms of sexual victimizations are included in a general “assault” category. The revised definition, scheduled for implementation in 2013, will provide a broader base for reports of rape and attempted rape. This change should result in a larger number of crimes being counted as rape and fewer crimes being counted in the “assault” category. Importantly, the UCR only measures incidents reported to police. This is an important difference with the other data sources, and the new Federal Bureau of Investigation (FBI) definition will not change this difference. The NCVS has a broader definition of rape. It includes male and female victims and offenders. It includes penetration (vaginal, anal, and oral) by penis, other body parts, and other objects. It also separately measures at- tempted rape and a fairly wide range of sexual assaults, including verbal

OCR for page 91
COMPARISON OF RAPE AND SEXUAL ASSAULT ACROSS DATA SOURCES 93 TABLE 6-1  Definitions of Rape and Sexual Assault, by Data Source Data Source Basic Description Estimates Uniform Crime Rape includes only male offenders and Rape and attempted Reports (UCR) female victims, with penile penetration of a rape. Does not Summary vagina. include other forms Reporting of sexual assault. Attempted rape is counted separately. System Other forms of sexual assault are included in a general category of assault and not summarized separately or with rape. An updated definition has been developed and was scheduled to be used beginning January 2013. The updated definition covers male and female victims and penetration with other (than penis) body parts and objects. It covers anal penetration and oral penetration by a sex organ. The definitional change does not change sexual assault, which is still included in the general category of assault. National Crime Rape includes psychological coercion Measures rape, Victimization as well as physical force. Forced sexual attempted rape, and Survey (NCVS) intercourse means vaginal, anal, or oral a wide category of penetration by the offender(s). It also sexual assault. includes incidents in which the penetration is by a foreign object. It includes attempted rapes, male as well as female victims, and both heterosexual and homosexual rape. Sexual assault included in this category includes a wide range of victimizations that are separate from rape or attempted rape. These crimes include attacks or attempted attacks generally involving unwanted sexual contact between victim and offender. Sexual assaults may or may not involve force and include such behavior as grabbing or fondling. Sexual assault also includes verbal threats. National Women’s Rape was defined as an event that occurred Measures rape. Study (NWS) without the woman’s consent, involved use Does not measure of force or threat of force, and involved attempted rape sexual penetration of victim’s vagina, mouth, or other forms of or rectum. The NWS results included only sexual assault. female victims and measured prevalence rather than the number of incidents. continued

OCR for page 91
94 ESTIMATING THE INCIDENCE OF RAPE AND SEXUAL ASSAULT TABLE 6-1  Continued Data Source Basic Description Estimates National Violence Rape was defined as an event that occurred Measures rape and Against Women without the victim’s consent and that attempted rape. Study (NVAWS) involved the use or threat of force to Does not measure penetrate the victim’s vagina or anus by other forms of penis, tongue, fingers, or object, or the sexual assault. victim’s mouth by penis. The definition included both attempted and completed rape. National College Rape is unwanted completed penetration Measures rape and Women Sexual by physical force or the threat of physical attempted rape, Victimization Study force. Penetration includes penile-vaginal, as well as various (NCWSV) mouth on the victim’s genitals, mouth on forms of sexual someone else’s genitals, penile-anal, digital- assault. vaginal, digital-anal, object-vaginal, and object-anal. Attempted rape is the unwanted attempted penetration by force or the threat of force. Threat of rape is the threat of unwanted penetration with force and threat of force. The NCWSV results include only female college students as victims. National Intimate Rape is defined as any completed or Measures rape and Partner and Sexual attempted unwanted vaginal (for women), attempted rape, Violence Survey oral, or anal penetration through the use of as well as various (NISVS) physical force (such as being pinned or held forms of sexual down, or by the use of violence) or threats assault. to physically harm. It includes times when the victim was drunk, high, drugged, or passed out and unable to consent. Rape is separated into three types: completed forced penetration, attempted forced penetration, and completed alcohol- or drug-facilitated penetration. Among women, rape includes vaginal, oral, or anal penetration by a male using his penis. It also includes vaginal or anal penetration by a male or female using his or her fingers or an object. Among men, rape includes oral or anal penetration by a male using his penis. It also includes anal penetration by a male or female using his or her fingers or an object. SOURCES: Data from Black et al. (2011); Bureau of Justice Statistics (n.d.-b); Federal Bureau of Investigation (2004); Fisher, Cullen, and Turner (2000); Kilpatrick, Edmunds, and Seymour (1992); Tjaden and Thoennes (2000).

OCR for page 91
COMPARISON OF RAPE AND SEXUAL ASSAULT ACROSS DATA SOURCES 95 threats. It reports separate estimates of rape and sexual assault, and also reports the two categories together. When comparing data from the NCVS with other sources, one has to decide which of the NCVS’s numbers to use. The NWS used the term “forcible rape.” It included much of the broader set of penetration victimizations by force or threat of force in- cluded in the NCVS definition of rape. The study’s estimates did not include attempted rape or other forms of sexual assault. The estimates it produced were more narrowly focused than the NCVS because its target population was only adult women. The NVAWS measured victimizations of both male and female adults. The definition of rape included penetration and attempted penetration (vaginal, oral, and anal) by force or threat of force. It did not cover other forms of sexual assault. The NCWSV targeted only college women. Thus, sexual victimizations against men and against women not in college were not included. The study measured 12 different types of victimization (see Table 5-2 in Chapter 5). Completed rape included penetration (vaginal, oral, and anal) by force or threat of force. The NISVS measures both completed and attempted rape as defined by penetration with use or threat of physical force. It attempts to measure victimizations of both adult males and females. However, because of limited sample size in the first (2010) survey, estimates were published only for females. It extends the definition of rape to include penetration when the victim was unable to consent by being drunk, high, drugged, or passed out. The panel next compared the above definitions with the common- alities of legal definitions we found across jurisdictions and presented in Chapter 2. Table 6-2 provides a summary of this comparison. Across the sources, there was less uniformity among the data sources regarding the inclusion of nonpenetration sexual assault and in gender restriction. The NISVS was the only source that specifically included alcohol- and drug-facilitated penetration as part of forced sexual activities. The panel identified this as a missing component to the NCVS definition (see Recom- mendation 10-7). SURVEY CONTEXT The context of a survey is very important to both response rates and to the quality of responses that are received. A simple change in context can make a big difference. For example, when the National Survey of Drug Abuse changed its name to the National Survey of Drug Use and Health, reported drug use increased (Office of Applied Studies, 2003). Context can be established in a number of ways, including the prior questions in a questionnaire (Holyk, 2008, p. 42):

OCR for page 91
96 ESTIMATING THE INCIDENCE OF RAPE AND SEXUAL ASSAULT The term context effect refers to a process in which prior questions affect responses to later questions in surveys. Any survey that contains multiple questions is susceptible to context effects. Context effects have the poten- tial to bias the thinking and answers of survey respondents, which reduces the accuracy of answers and increases the error in survey measurement. In comparing the overall context of data sources for measurements of rape and sexual assault, two sources focus specifically on crimes: the UCR and the NCVS. The UCR summarizes “crimes known to police”—those that are both reported to and recorded by police. The NCVS’s goal is to measure the victimization rate by type of crime. The NCVS is a national crime survey, and the questionnaire asks many questions about different types of crimes as well as the well-being of victims. The other four sources (NWS, NVAWS, NCWSV, and NISVS) do not frame the survey around/on criminal victimization. They focus instead on the situations in which the respondent may have experienced nonconsensual or unwanted sexual contact. These surveys also collect additional informa- tion about the respondent’s well-being. The panel believes that survey context is likely a major contributor of differences in the estimates of rape and sexual assault between the several sources. TARGET POPULATION, SAMPLING FRAMES, AND SAMPLE SIZE Target Populations and Sampling Frames The target populations for the six surveys are different, with resulting effects on the estimates. The NWS targeted adult (18+ years) women and made no estimates for men. The NCWSV targeted a narrower group of women, only those attending college. The NVAWS and the NISVS both targeted adult (18+ years) men and women. However, the NISVS only published estimates (12-month prevalence) of rape and attempted rape for women for its first implementation (in 2010). The NCVS targets both men and women with a broader age range (12+ years). The sampling frames were different for different sources. Three surveys (NWS, NVAWS, and NISVS) are based on geographically spread random digit dialing (RDD) frames. The RDD frames cover only U.S. households that have telephones. It is important to consider whether this undercoverage is serious. The NWS and the NVAWS were conducted between 1989 and 1995. The 1990 census showed that 5.2 percent of U.S. households had no telephones. The percentage was above 10 percent in five states.1 The next 1  Arkansas, 10.9 percent; Kentucky, 10.2 percent; Mississippi, 12.6 percent; New Mexico, 12.4 percent; West Virginia, 10.3 percent.

OCR for page 91
COMPARISON OF RAPE AND SEXUAL ASSAULT ACROSS DATA SOURCES 97 TABLE 6-2  Comparing Definitions Used for Rape with Commonalities Found Across Jurisdictional Legal Codes Commonalities Comparison with Definitions Used by Data Sources Victimization not restricted by The UCR definition is currently restricted by gender. The gender. Both male and female updated definition is not. victims and offenders. The NCVS is not restricted by gender. The NWS and the NCWSV targeted only women victims but did not restrict the gender of the offender. The NVAWS and the NISVS were not gender restricted. Rape involves a broad range The UCR definition is currently restricted to penile of penetrations. penetration of a vagina. The updated definition is not. The NCVS, NWS, NCWSV, NVAWS, and NISVS include a broad range of penetrations. Purpose is for sexual arousal This is not a specifically stated component of the or degradation. definitions of any of the sources but probably does not need to be. In legal statutes, this is used to distinguish assaults from such things as medical exams. Use of force or threat of force Consistent for all sources. against the victim or another person. Lack of consent or lack of Lack of consent is consistent for all sources. The NISVS capacity to consent. specifically includes questions related to the lack of capacity to consent due to alcohol and/or drug use. Sexual assault includes a fairly The UCR includes sexual assault with other types of wide range of victimizations assault and does not have a separate category. that involve unwanted non- The NCVS has a definition of sexual assault that is penetration sexual contact. consistent with the commonality. It summarizes it separately and also combines it with rape and attempted rape. The NWS does not measure attempted rape or sexual assault. The NVAWS includes attempted rape but not other types of sexual assault. The NCWSV and the NISVS have definitions of sexual assault consistent with this commonality. NOTES: Commonalities described in Chapter 2. NCVS = National Crime Victimization Survey, NCWSV = National College Women Sexual Victimization Study, NISVS = National Intimate Partner and Sexual Violence Survey, NVAWS = National Violence Against Women Study, NWS = National Women’s Study, UCR = Uniform Crime Reports. SOURCES: Data from Black et al. (2011); Bureau of Justice Statistics (n.d.-b); Federal Bureau of Investigation (2004); Fisher, Cullen, and Turner (2000); Kilpatrick, Edmunds, and Seymour (1992); Tjaden and Thoennes (2000); Tracy et al. (2012).

OCR for page 91
98 ESTIMATING THE INCIDENCE OF RAPE AND SEXUAL ASSAULT census, in 2000, showed the percentage of households without telephone was only 2.4 percent of households, but the same five states still lagged in coverage (U.S. Census Bureau, 2000).2 Thus, there is some concern about coverage in these two surveys for poorer households (the ones that generally lack telephones) in those states, but the potential coverage error is small enough that it was unlikely to have much effect on the estimates. In more recent years, there has been an increase in the percentage of cell-phone-only households (with cell phones but no landline), from approxi­ ately 3 percent in 2003 to 34 percent in 2012 (Blumburg and m Luke, 2012). The overall trend to cell phones instead of landlines is signifi- cant (Hall, Carlson, and CyBulski, 2011, p. 2): This drastic change in cell phone usage has significantly affected the cover- age of surveys that use random digit dialing sampling. Because of this trend in cell phone usage over the last decade, using only a landline-based RDD sample results in reducing the coverage of the population. Responding to these changes, the NISVS augmented its RDD sample of landline phone numbers to include a cell phone sample. In 2011, only about 2 percent of U.S. households had no telephone, landline, or cell phone (Blumburg and Luke, 2012). Two of the studies reviewed by the panel relied on cluster sampling. Sampling for the NCWSV involved two stages of selection. The first stage frame was a list of academic institutions stratified by total student enroll- ment and institution location, and the second stage was a list, for the se- lected institutions, of women enrolled in the fall of 1996. The sample design for the NCVS begins with a selection of primary sampling units (PSUs) from the Census Bureau, and then uses the Master Address File supplemented with the New Building Permits frame and the Group Quarters frame (see Chapter 4 for more details). The UCR data come from an administrative source (voluntary reports from law enforcement agencies); they are not based on a random sample. Sample Size Table 6-3 displays information on sampling frames and sample sizes. The sample size for the NCVS is substantially larger than for the other surveys. Its sample size has fluctuated with annual budget changes over the past years: the smallest number of interviews was 134,041 in 2005, and the largest number was 181,205 in 1994. The NWS interviewed 4,008 women, approximately 4 percent of the 2  Arkansas, 5.4 percent; Kentucky, 4.7 percent; Mississippi, 6.5 percent; New Mexico, 5.7 percent; West Virginia, 4.7 percent.

OCR for page 91
COMPARISON OF RAPE AND SEXUAL ASSAULT ACROSS DATA SOURCES 99 people interviewed in the 1990 NCVS (95,000 interviews). The sample size for the NCWSV was approximately the same as the NWS, with 4,446 inter­ views. However, the NCWSV was targeting a much smaller population. The sample size for the NVAWS was 16,000 interviews, and the sample size for the NISVS was 18,049 interviews. Thus, the sample size for these two surveys were only about 11 percent of the sample size for the 2011 NCVS. Smaller sample sizes are a particular problem when measuring a low inci- dence event, such as rape. For example, the NVAWS, with a sample size of about 16,000, found only 24 women and 8 men who reported having been raped. And the NISVS was unable to publish estimates of male victimiza- tions because of the small number of reported victimizations. The UCR is essentially a census of all police reports from approxi- mately 18,000 participating jurisdictions, which covers approximately 90 percent of all jurisdictions. It has no incident-level or individual-level re- cords. Coverage in metropolitan areas is slightly higher than in rural areas. These police reports are widely believed to be missing a substantial percent- age of the rapes and sexual assaults that occur, as much as 65-80 percent. DATA COLLECTION MODE AND RESPONSE RATES Data collections for all five surveys are interviewer administered and rely heavily on telephone interviewing. Beyond that generality, however, there are differences (see Table 6-3). The NCVS is an ongoing panel survey, with selected households in the survey for 3 years. This is different from the other surveys. The NCVS begins with a presurvey letter and an in- person visit for wave 1. It uses telephone interviews, conducted by the field representatives, for other waves if feasible (see Chapter 4). The NCWSV also began with a presurvey letter so that the telephone interview that fol- lowed was not based on a cold contact. The others (NWS, NVAWS, and NISVS) were all RDD surveys with a cold initial contact from a central- ized telephone facility.3 All interviews used computer-assisted interview- ing technology. This discussion is not relevant for the UCR, which uses administrative data. Nonresponse can affect both survey estimates and their estimated vari- ances. For a survey, nonresponse bias is dependent on both the size of the nonresponse and to the extent differences exist between respondents and nonrespondents regarding important variables being measured in the sur- vey. Because of these potential effects, response rate has been one indicator used to assess survey quality. The rates for the data sources discussed in this chapter varied consider- 3  The NISVS was able to match 50 percent of its landline sample with addresses. These households were sent an advance letter.

OCR for page 91
100 ESTIMATING THE INCIDENCE OF RAPE AND SEXUAL ASSAULT TABLE 6-3  Sampling Information, Data Collection Mode and Response Rates, by Data Source Data Source Year Sampling Frame Uniform Crime Reports Since 1929 Police reports (UCR) Summary Reporting System National Crime Annually since 1991 2-stage cluster sample, with area- Victimization Survey Predecessor survey based PSUs. Addresses sampled from (NCVS) since 1972 the Census Bureau’s Master Address File, the new building permits frame, and the group quarters frame Target—males and females 12+ years National Women’s 1989-1990 RDD: 2-stage sample, with primary Study (NWS) stage geographic areas within the United States; 2nd stage a random digit dialing using landline frame within selected geographic areas National Violence Nov 1995 through RDD: National random digit dialing Against Women Study May 1996 (landline) sample, selected within (NVAWS) census regions National College 1996-1997 2-stage sample, with primary stage Women Sexual stratified list of 2- and 4-year Victimization Study colleges; 2nd stage a sample of (NCWSV) women enrolled in fall 1996 National Intimate 2010 RDD: Random digit dialing using Partner and Sexual both landline and cell phone frames Violence Survey (NISVS) aSample size—individuals interviewed. bResponse rate calculated based on AAPOR Standard 4. cThe participation rate for the NVAWS was calculated by dividing the number of completed interviews (including those that were screened out because they were ineligible) by the total number of completed interviews, screened-out interviews, refusals, and terminated interviews (Tjaden and Thoennes, 2000, p. 4).

OCR for page 91
COMPARISON OF RAPE AND SEXUAL ASSAULT ACROSS DATA SOURCES 101 Sample Sizea Data Collection Mode Response Rates 100% from participating Administrative records NA 18,000 jurisdictions n = 143,122 individuals Interviewer administered. 2011—88% for interviewed, male and female, In-person interview on first individual respondents 12+ years, from 79,802 and last waves. Telephone households (2011) interview on other waves, to the extent feasible n = 4,008, female only, 18+ Interviewer administered 34% response rateb years with cold telephone contact 85% participation rate on wave 1, with subsequent waves also administered with telephone interview n = 16,000, male and female, Interviewer administered with 34% response rateb 18+ years cold telephone contact 72% participation rate for females and 69% participation rate for malesc n = 233 institutions Interviewer administered. 67% response rate,b n = 4,446 female college Presurvey letter followed by with 86% participation students telephone interview rate n = 18,049 (2010), male and Interviewer administered with 2010—34% response female, 18+ years cold telephone contact rate NOTE: AAPOR = American Association for Public Opinion Research, PSUs = primary sam- pling units. SOURCES: Data from Black et al. (2011); Bureau of Justice Statistics (n.d.-b); Federal Bureau of Investigation (2004); Fisher, Cullen, and Turner (2000); Kilpatrick, Edmunds, and Seymour (1992); Tjaden and Thoennes (2000).

OCR for page 91
102 ESTIMATING THE INCIDENCE OF RAPE AND SEXUAL ASSAULT ably. The NCVS has both the largest sample size and the highest response rate, 88 percent. The NCWSV achieved a modest response rate of 67 per- cent. RDD surveys invariably have lower response rates because of “screen outs” needed to obtain a qualified respondent and the effect of cold calling. The three RDD surveys (NWS, NVAWS, and NISVS) had similar response rates—NWS, 34 percent; NVAWS, 34 percent; NISVS, 34 percent—all substantially lower than that achieved by the NCVS and the NCWSV. Coopera­ion rates (which are calculated once a qualified respondent is t reached) were higher in each of these three surveys. MEASURES OF RAPE AND SEXUAL ASSAULT The number of rapes and sexual assaults from the four surveys and the UCR are shown in Figure 6-1.4 One only has to look at the graphic to understand that these data sources are both measuring different things and measuring things differently. Table 6-4 shows the estimates of number of rapes from all six sources. Table 6-5 shows estimates for females aged 18+ only. The lowest 12-month estimate for rape is measured by the UCR. This is not surprising because the UCR has the most restrictive definition of rape and only measures rapes (and attempted rapes) that are known to the police. Therefore, the UCR is assembled in ways that make it vulnerable to major undercounting. The estimates from the NISVS are the largest. It is important to note that the NISVS published 12-month prevalence numbers and only for women: if the 12-month estimates were of incidents and for all adults, then the numbers would be even larger. The number of females raped or sexually assaulted (adult females only) estimated by the NISVS is 5 times larger than the number of incidents measured by the NCVS (including series victim- ization) for rape and sexual assault for females (age 12+) in 2010, twice as large as the prevalence number estimated by the NWS (adult females, completed rape only in 1990),5 and 30 percent greater than measured by the NVAWS (both male and female adults, completed and attempted rape but no other forms of sexual assault in 1995). This differential between the NISVS and the other surveys is surprising. The definitions are not identi- cal, but they are roughly consistent. The NISVS, along with the NWS and NVAWS, used RDD survey designs. All three of these surveys had the same response rate, 34 percent. The panel attempted to look at confidence intervals for comparisons 4  The rates from the NCWSV are excluded because the survey covered only a small at-risk population—college women. 5  However, the NISVS prevalence estimate for completed rape only (620,000—see Table 5-4) is close to the NWS estimate of incidents of completed rape for adult women (683,000).

OCR for page 91
COMPARISON OF RAPE AND SEXUAL ASSAULT ACROSS DATA SOURCES 103 NISVS Female Prevalence FIGURE 6-1 Number of rapes, by data source. NOTES: • NWS and NISVS estimates are for adult (18+) women only. • NWS and NISVS estimates are prevalence rates and not incidence rates. • NWS estimates do not include attempted rape or other sexual assaults. • NCVS estimates are for rape and sexual assault, ages 12+ years. •  CR estimates are for rape and attempted rape that are known to police U (no age limit). Estimates from the NCWSV are specifically for college women and thus not com- parable to the others and not included in this figure. NCS = National Crime Survey, NCVS = National Crime Victimization Survey, NCWSV = National College Women Sexual Victimization Study, NISVS = National Intimate Partner and Sexual Vio- lence Study, NVAWS = National Violence Against Women Study, NWS = National Women’s Study, UCR = Uniform Crime Reports. SOURCES: Panel-developed graphic using data from Black et al. (2011); Bureau of Justice Statistics (1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002a, 2003, 2004, 2005, 2006, 2007, 2008a, 2009, 2010, 2011, 2012a, n.d.-a); Federal Bureau of Investigation (n.d.-c); Fisher, Cullen, and Turner (2000); Kilpatrick, Edmunds and Seymour (1992); and Tjaden and Thoennes (2000). as well as the point estimates described above. This attempt was somewhat unsatisfying because we were unable to obtain standard errors from the NWS or the NISVS.6 Rand and Rennison (2005) compared the NVAWS with the NCVS, and we extracted the NVAWS standard errors from that journal article. For 1995, the NCVS estimate of the number of rapes and sexual assaults for adult women was 403,735, including series victimization 6  The NISVS published standard errors for their lifetime prevalence rates but not for the 12-month prevalence rates, which were needed for our comparisons.

OCR for page 91
TABLE 6-4 Estimates of Number of Rapes and Sexual Assault by Year, by Source 104 Source Gender of NCVS Year Victim UCR No Seriesa Seriesb NWS NVAWS NISVS 1990 Fc 102,555 106,000 NA 683,000d All 130,260 NA 1991 All 106,593 174,000 NA 1992 All 109,062 NA NA 1993 All 106,014 521,223 898,239 1994 All 102,216 443,509 674,291 1995 All 97,470 363,527 563,249 987,362 1996 All 96,252 307,100 437,198 1997 All 96,153 311,110 553,523 1998 All 93,144 332,500 391,101 1999 All 89,411 383,170 591,460 2000 All 90,178 260,950 366,747 2001 All 90,863 248,250 476,578 2002 All 95,235 247,730 349,805 2003 All 93,883 198,850 325,311 2004 All 95,089 209,880 255,769 2005 All 94,347 190,592 207,760 2006 All 92,757 260,940 463,598 2007 All 90,427 248,277 248,277 2008 All 90,479 203,830 349,691 2009 All 89,241 125,910 305,574 2010 Fc 253,555 1,270,000e All 84,767 188,380 268,574 2011 All 243,803

OCR for page 91
NOTES: NCVS = National Crime Victimization Survey, NISVS = National Intimate Partner and Sexual Violence Survey, NVAWS = National Violence Against Women Study, NWS = National Women’s Study, UCR = Uniform Crime Reports. aSeries victimizations are excluded. bIncludes series victimizations up to 10 incidents. cFemale adults. dPrevalence estimate for completed rapes only; does not include attempted rape or other forms of sexual assault. ePrevalence estimate (number of victims) rather than the number of victimizations. SOURCES: Panel-developed table using data from Black et al. (2011, 2012a, n.d.-a); Bureau of Justice Statistics (1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002a, 2003, 2004, 2005, 2006, 2007, 2008a, 2009, 2010, 2011, 2012a, n.d.-a); Federal Bureau of Investiga- tion (n.d.-c); Fisher, Cullen, and Turner (2000); Kilpatrick, Edmunds and Seymour (1992); and Tjaden and Thoennes (2000). 105

OCR for page 91
106 ESTIMATING THE INCIDENCE OF RAPE AND SEXUAL ASSAULT TABLE 6-5  Estimates of Number of Rapes and Attempted Rapes of Females, 18 Years of Age and Older for Selected Years, by Source Source NCVSb Year UCRa No Series Series NWSc NVAWSd NISVSe 1990 102,555 106,000 NA 683,000 1995  97,470 NA 403,735 876,064 2010  85,593 NA 209,740 1,270,000 NOTES: These numbers represent the most consistent comparison that the panel was able to identify. Even so, they have differences, which are listed below. NCVS = National Crime Victimization Survey, NISVS = National Intimate Partner and Sexual Violence Survey, NVAWS = National Violence Against Women Study, NWS = National Women’s Study, UCR = Uniform Crime Reports. aUCR Forcible rape and attempted rapes known to police. Includes females, all ages. bNCVS Rape, attempted rape, and sexual assault, for 1990 only, includes females ages 12+ years. For 1996 and 2010, includes females, 18+ years. cNWS Forcible rape only. Does not include attempted rape. Count of victims rather than victimizations. dNVWS Forcible rape and attempted rape. eNISVS Rape and attempted rape. Count of victims rather than victimizations. SOURCES: Data from Black et al. (2011); Bureau of Justice Statistics (1991, 1996, 2011, n.d.-a); Federal Bureau of Investigation (n.d.-c); Kilpatrick, Edmunds, and Seymour (1992); Tjaden and Thoennes (2000). (see Table 6-5). The standard error was 41,643.7 The 95 percent confidence interval was [322,115, 485,355]. The NVAWS estimate of the number of forcible rapes and attempted rapes was 876,064. The NVAWS had a much smaller sample size than did the NCVS, and as expected, its standard er- ror was much larger: 467,098. The 95 percent confidence interval for the NVAWS estimate was [–39,448, 1,791,576]. Thus the 95 percent confidence interval for the NCVS was contained completely within the 95 percent con- fidence interval for the NVAWS. Rand and Rennison (2005) made a similar finding, with no statistical difference in those estimates. In reviewing all of this material, the panel thinks that it is highly likely that the NCVS is underestimating rape and sexual assault. The panel, with limited resources, was not able to measure the extent of such an under- count, but the pattern is one that shows lower estimates of rape and sexual 7  This is the standard error for estimate of rape and sexual assault for all women. The panel did not have the actual standard error for women 18 years of age and older.

OCR for page 91
COMPARISON OF RAPE AND SEXUAL ASSAULT ACROSS DATA SOURCES 107 assault in the NCVS than the estimates published from the other surveys.8 Thus, the panel looked in more detail at the error profile of the NCVS to better understand procedures that might be contributing to this undercount. This analysis is contained in Chapters 7, 8, and 9. 8  This was the case even though these other surveys were often more restrictive in what they measured, such as focusing on adults only, or women only, or not including attempted rape or other forms of sexual assault.

OCR for page 91