The purpose of this study was to use data from the National Crime Survey (NCS) and the National Crime Victimization Survey (NCVS) to explore whether the likelihood of police notification by rape victims had increased between 1973-2000. Specifically, the study examined whether the overall probability of reporting by rape victims increased during this period, whether any observed increase in the likelihood of police notification has been more prominent among (or perhaps restricted to) incidents involving acquaintances, and whether differences in the likelihood of reporting between incidents involving strangers and acquaintances have diminished significantly over time.
The researchers used information from two periods of data collection from the U.S. crime victimization survey: National Crime Survey (NCS) data from 1973-1991 and National Crime Victimization Survey (NCVS) data from
1992-2000. These surveys collected data on whether persons and households had experienced a criminal victimization during the previous six months. For each victimization reported, respondents were asked detailed questions about the crime, including whether the incident was reported to the police.
To avoid the ambiguities that could arise in analyses across the two survey periods, the researchers analyzed the
NCS (1973-1991) and NCVS data (1992-2000) separately. They focused on incidents that involved a female victim and one or more male offenders. The sample for 1973-1991 included 1,609 rapes and the corresponding sample for 1992-2000 contained 636 rapes. Supplementary analyses also were performed on nonsexual assaults experienced by women to provide a comparative context within which to assess the results obtained for rape.
In their analyses, the researchers controlled for changes in forms of interviewing used in the NCS and NCVS, contrasting personal interviews with computer assisted phone interviews and traditional phone interviews. Logistic regression was used to estimate effects on the dichotomous measure of police notification and multinomial logistic regression was used to estimate effects on the polytomous measure of police notification. The analyses incorporated the currently best available methods of accounting for design effects in the NCS and NCVS. For the 1992-2000 redesigned data, survey regression models were estimated that directly incorporated NCVS strata and primary sampling unit (PSU) variables and adjusted for design effects. For the 1973-1991 period, all standard errors obtained in regressions using a standard model-based approach were multiplied by the covariate-specific design effects observed in survey regression models based on 1987-1991 NCS data.
The research questions were addressed with data from two periods of data collection from the United States crime victimization survey: National Crime Survey (NCS) data from 1973-1991 and National Crime Victimization Survey (NCVS) data from 1992-2000. The NCS and NCVS used a stratified multistage cluster sample. To avoid the ambiguities that can arise in analyses across the two survey periods, the NCS (1973-1991) and NCVS data (1992-2000) were analyzed separately. Following past research, the main analysis was restricted to incidents that involved a female victim and one or more male offenders. The sample for 1973-1991 included 1,609 rapes, the corresponding sample for 1992-2000 contained 636 rapes. Supplementary analyses also were performed on non-sexual assaults experienced by women to provide a comparative context within which to assess the results obtained for rape.
NATIONAL CRIME SURVEYS: NATIONAL SAMPLE, 1986-1992 [NEAR-TERM DATA] (ICPSR 8864), DS15: 1987-1992 Incident-Level Concatenated File
NATIONAL CRIME SURVEYS: NATIONAL SAMPLE OF RAPE VICTIMS, 1973-1982 (ICPSR 8625), DS1: Rape Victims
NATIONAL CRIME VICTIMIZATION SURVEY, 1992-2000 (ICPSR 3140), DS3: 1992-2000 Incident-Level Concatenated File
NATIONAL CRIME SURVEYS: NATIONAL SAMPLE, 1979-1987 [REVISED QUESTIONNAIRE] (ICPSR 8608), DS16: 1979-1987 Incident-Level Concatenated File, All Victims, Bounded by Calendar Year
program source code
computer-assisted telephone interview (CATI)
Police notification served as the dependent variable in the study and was measured in two ways. First, to account for the possibility that the legal and social reforms that have taken place since the early 1970s may have altered the likelihood that victims or third parties report rape victimizations to the police, the analysis included a polytomous dependent variable that contrasts victim reported incidents and third-party reported (e.g., another member of the household, some nonlaw enforcement official, anybody else) incidents, respectively, with nonreported incidents. Second, to facilitate more direct comparisons with past research, a binary dependent variable, police notified, also was included. This variable was coded 1 if the victim or somebody else reported the incident to the police and 0 for incidents in which the police were not notified.
The primary independent variables in the analysis were the year of occurrence of the incident reported by the victim and the relationship between the victim and offender. Past research on police notification by victims of rape focused primarily on contrasting reporting rates between incidents involving strangers and nonstrangers. Consistent with this past practice, analyses were presented in which victim-offender relationship is measured as a dichotomous variable that contrasts incidents in which the victim identified the offender as a nonstranger with those in which involved an offender who was a stranger. In addition, the effects of more detailed relationship variables also were considered in this study. Specifically, a series of dichotomous variables were constructed that distinguish incidents in which the offender was identified as somebody with whom the victim was acquainted by sight only, was a causal acquaintance, a well-known acquaintance, a spouse or ex-spouse, or other family member.
The regression models included several control variables, including measures of respondents' socioeconomic status, as well as other victim, offender, and incident
characteristics that may be related both to the nature of rape and to the likelihood that victims notified the police.
Response rates for the NCS and NCVS vary by collection. Users are encouraged to refer to the BJS website for more information about response rates.