Prior Police Contact and Subsequent Victim Reporting: Results from the NCVS

Abstract

This study investigated the association between victim reporting and the police response to past victimizations with data from the National Crime Victimization Survey from 1998–2000. The findings include: (1) investigatory effort by police when an individual had been victimized in the past increased the likelihood that the individual would report an ensuing victimization to the police; (2) however, this relationship only held when the victim, rather than someone else, reported the prior victimization to the police; (3) whether the police made an arrest after an individual was victimized in the past had no effect on whether the individual reported an ensuing victimization to the police; (4) the probability of victim reporting was unaffected by investigatory effort or whether an arrest was made after a prior victimization of a member of the victim’s household.

Keywords:

• victim report
• procedural justice
• social learning
• rational choice
• NCVS

Acknowledgments

The authors thank Tom Zelenock, for assistance in constructing the longitudinal data set, and Chris Dunn and other staff members, for the 2003 ICPSR Summer Workshop, Quantitative Analysis of Crime and Criminal Justice.

Notes

1. Hickman and Simpson (2003, pp. 615–616) reported: “Of 907 eligible cases, 594 (65 percent) victims completed the initial interview. A similar share (385 of 594, or 65 percent) of these victims completed the six‐month follow‐up interview, accounting for 42 percent of the 907 originally eligible cases.”

2. Crime victims may not always want the perpetrator arrested. For this reason, later analyses controlled for known correlates of victims’ preferences for arrest, such as whether the offender was previously known to the victim and the gravity of the victimization.

3. Clearly, victim reporting may also be influenced by the prior victimization experiences of individuals who are not members of the victim’s household. The NCVS does not permit us to investigate this proposition, and our conceptual model reflects this practical reality.

4. Personal records were matched using linkage variables provided by the Census Bureau. Based on a comparison of demographic characteristics, 5 percent of cases were incorrectly matched and thus excluded, a percentage comparable to that in prior research (cf. Ybarra & Lohr, 2002).

5. This study excluded 3.8 percent of victimized households either because they reported series crimes or because a victimization occurred outside the United States (cf. Bachman, 1993, 1998; Conaway & Lohr, 1994; Ybarra & Lohr, 2002). Beginning in January of 1993, the NCVS has defined series crimes as six or more related incidents, with the respondent unable to distinguish the details of each incident (US Department of Justice, 2002).

6. An incident was considered a prior victimization if the police were summoned.

7. In 4.6 percent of cases, the two latest victimizations occurred in the same month. For these cases, the outcome variable was based on the last physical record. In unreported analyses that excluded these cases, the results were unchanged.

8. For 247 victims with prior personal victimization and 229 victims with prior household victimization, someone other than the victim (for example, a witness or household member) reported the most recent victimization. In these cases, it is unknown why the third party became involved. On the one hand, the victim may have been reluctant to notify police, thus prompting the third party’s involvement. On the other hand, it is unknowable whether the victim would have eventually reported the crime, had the third party remained uninvolved. In unreported analyses, we recoded the dependent variable more liberally by assigning a 1 to each most recent victimization that was reported to the police, either by the victim or someone else. In these analyses, the results were unchanged.

9. We also re‐estimated the ensuing models after substituting the mean values of the explanatory variables for all prior victimizations and the results were unchanged.

10. The NCVS also includes other potential indicators of police effort, such as whether they took a report, promised surveillance, and promised to investigate. When these additional measures were included in the models presented later, none predicted victim reporting, and the remaining parameters were unchanged.

11. For 69 of the 1,635 victims with prior personal victimization, the two most recent prior victimizations occurred in the same month. In 45 of these cases, Self Reporter was either coded 1 because the victim reported both incidents or 0 because all incidents were reported to the police by someone other than the victim. For the remaining 24 cases, Self Reporter was coded as missing.

12. The model chi‐square statistic is –2(log likelihood_{reduced model} – log likelihood_{full model}), where the reduced model excludes the corresponding set of explanatory variables. The degrees of freedom for each model chi‐square test is the difference in the number of parameters between the full and reduced models.

13. Unreported analyses tested the role of similarities between the prior and current victimization. The findings were unchanged when model (1) was limited to cases where the prior and current victimization were “similar” in any of the following respects: whether the victim was injured, whether a weapon was involved, the value of any property loss, the victim–offender relationship, the victim’s presence during the crime, or the distance between the crime location and the victim’s home.

14. The interpretation of this negative coefficient is unclear, since family size may have several counterbalancing effects on reporting. On the one hand, more household members means that a victimization incident within the household potentially jeopardizes more individuals. On the other hand, more household members also means that there are more individuals to potentially report the victimization, perhaps reducing pressure on the victim to do so.

15. The logistic regressions in Table 3 take the form, ln(p / 1 – p) = α + βX, where p is the probability of reporting, and β and X are vectors of regression coefficients and variables, respectively. Solving for p yields

This formula was used to calculate the probabilities in Figure 1.

16. In the household victimization group, ((1,177/1,371) × 100 = 86 percent) had no prior personal victimization experiences to draw upon, perhaps leaving room for the influence of victim characteristics to emerge.

17. We were unable to directly test the other primary finding of Hickman and Simpson (2003) that whether the police acted consistently with the victim’s preference for arresting the perpetrator related positively to the probability of subsequent reporting; the NCVS does not measure the victim’s preference for arrest.

Additional information

Notes on contributors

Min Xie

Min Xie is a doctoral student in the School of Criminal Justice at the University at Albany. Her current research interests include theories of criminal victimization, victim decision making, multilevel and longitudinal models of victimization, and spatial data analysis.

Greg Pogarsky

Greg Pogarsky is an Associate Professor in the School of Criminal Justice at the University at Albany. His research focuses on refining models of crime decision‐making by revisiting their assumptions about human nature and by appealing to recent advancements in economics and psychology.

James P. Lynch

James P. Lynch received his PhD at the University of Chicago in Sociology and is a Presidential Scholar at John Jay College. His interests include crime statistics, victimization surveys, and theories of victimization risk as well as the role of punishment in social control with emphasis on cross‐national comparisons.

David McDowall

David McDowall is a professor in the School of Criminal Justice at The University at Albany and codirector of the Violence Research Group. His research interests include crime trends and the causes and consequences of interpersonal violence.