American Journal of Criminal Justice

, Volume 38, Issue 1, pp 141–159

Context Matters: Violence Against Women and Reporting to Police in Rural, Suburban and Urban Areas

Authors

    • School of Public AffairsUniversity of Colorado Denver
  • Molly Dragiewicz
    • Faculty of Social Science and HumanitiesUniversity of Ontario Institute of Technology
  • Walter S. DeKeseredy
    • Faculty of Social Science and HumanitiesUniversity of Ontario Institute of Technology
Article

DOI: 10.1007/s12103-012-9164-4

Cite this article as:
Rennison, C.M., Dragiewicz, M. & DeKeseredy, W.S. Am J Crim Just (2013) 38: 141. doi:10.1007/s12103-012-9164-4

Abstract

Despite plentiful efforts to identify perpetrator, victim, and incident characteristics correlated with reporting violence against women to police, few studies have addressed the contexts that shape such reporting. Even fewer have examined variations in these contexts across geographic areas. Drawing upon National Crime Victimization Survey data from 1992 through 2009, this paper uses conjunctive analysis of case configurations to identify and investigate the dominant situational contexts of reporting of violence against women to police across rural, suburban, and urban areas. Our findings show that context matters and the importance of incident, perpetrator, and victim characteristics vary across geographic areas.

Keywords

Violence against womenPolice reportingRuralUrbanSuburban

Introduction

There is a wealth of social scientific information on criminal justice responses to violence against women in intimate relationships (Buzawa, Buzawa, & Stark, 2012; DeKeseredy, 2011), but more research on reporting to the police is necessary for a variety of reasons. First, reporting may directly facilitate protection of victims through restrictions placed on perpetrators (Klein, 2009). Second, reporting contributes indirectly to victims’ safety by laying a paper trail that can be used as evidence of a pattern of behavior in later legal proceedings (Klein, 2009). Third, reporting can have a deterrent effect (Bachman, 1993; Felson, Ackerman, & Gallagher, 2005; Wiist & McFarlane, 1998; Willson, McFarlane, Lemmey & Malecha 2001). Fourth, reporting to the police often results in referrals to other sources of support for victims of crime, such as counseling or emergency shelter (Kaukinen, 2002a, 2002b; Logan, 2005). And fifth, reporting crime provides clues to theory development by revealing the factors that promote and inhibit crime and responses to it.

Violence against women makes up a large percentage of calls to police. In fact, “Domestic-violence-related police calls have been found to constitute the single largest category of calls received by police, accounting for 15 to more than 50 % of all calls” (Klein, 2009, p. 1). However, this high volume does not necessarily reflect a large proportion of crimes reported to police. For example, findings from the National Crime Victimization Survey (NCVS) continue to illustrate that rape and sexual assault are the least likely of violent crimes to be reported to police (Bureau of Justice Statistics, 2003, 2007, 2011). In addition, crime victimization studies in general suffer from under-reporting of male-to-female violence relative to studies specifically designed to glean rich data on woman abuse (DeKeseredy & Schwartz, 2011; Koss et al. 2007; Schwartz, 2000). Note that Mihalic and Elliot (1997) found that up to 83 % of martial violence incidents reported in surveys of family behavior are not reported in contexts where the emphasis is on criminal assault and victimization.

Despite plentiful efforts to identify perpetrator, victim, and incident characteristics correlated with reporting violence against women to police, few studies have attempted to understand the role that situational contexts play in shaping reporting behavior. Most studies are variable focused leaving out important information regarding the role of situational context. Also poorly examined is variation in reporting of violence against women across geographic areas (Heimer, 2008; Ruback & Ménard, 2001). This paper, then, attempts to fill a research gap by using conjunctive analysis to explore key situational contexts of reporting violence against women1 across geographic areas (i.e., rural, suburban, and urban areas).

Review of Relevant Literature

Violence Rates

Few social scientific areas of inquiry have progressed as quickly as the study of violence against women. Forty years ago, a comprehensive bibliography of North American sources on male violence in marriage and cohabiting relationships would have fit on an index card (DeKeseredy & Schwartz, 2009). Today, hundreds of journal articles, books, and several important journals specifically address violence against women in a variety of contexts. Nevertheless, we still see variance in incidence and prevalence rates across studies, even when they use similar measures. This variance is due to numerous methodological issues, such as the use of broad or narrow definitions, recall bias, underreporting, question order, sampling differences, directness of questioning, framing of the study, external validity, and even the sex and ethnicity of interviewers (DeKeseredy, 2000; Koss et al., 2007; Koss, Gidycz & Wisniewski, 1987; Schwartz, 2000). The resulting confusion is well described by Schwartz (2005) who states “Each year, we have more conflicting information, which leads to more disputes on how to interpret the information we do have” (p. 7). Furthermore, many studies analyze sexual and nonsexual violence separately, and violence by intimates and non-intimates separately, rendering estimates of violence against women difficult to compare across studies. Still, there have been multiple large-scale efforts to estimate rates of violence against women.

For instance, the National Violence Against Women Study (NVAWS) using data collected in 1995 and 1996 found an annual rape victimization rate of 8.7 per 1000 women, an annual physical assault victimization rate of 58.9 per 1000, and an annual stalking victimization rate of 10 per 1000. More than half (55.9 %) of the women reported experiencing victimization in their lifetime. This number includes 17.6 % of women who reported completed or attempted rape, 51.9 % who reported physical assault, and 8.1 % who reported stalking (Tjaden & Thoennes, 2000). NVAWS elicits higher nominal rates than the NCVS due to methodological issues that have been discussed extensively elsewhere (Bachman, 1998, 2000; however, see Rand & Rennison, 2005).

The NCVS is another major representative sample survey and it uncovered violent victimization rates of 14.2 per 1000 women, a total of 1,854,980 violent victimizations, in 2010. Of this number, 64 % involved non-stranger perpetrators. Among these non-strangers, friends and acquaintances were the most frequent perpetrators (33 %), followed by intimates (22 %), and other relatives (9 %). Thirty percent of women’s violent victimizations were perpetrated by strangers. In 6 % of violent victimizations, the relationship to the perpetrator was unknown (Truman, 2011).

Reporting to Police

For most crimes, victimization surveys find less under-reporting to police than other studies of violence. This is not surprising given that such surveys evoke disclosure of incidents that respondents identify as crimes. The NCVS found that 53.3 % of all violent victimizations of women were reported to police in 2010 (Truman, 2011). Bachman investigated rape reporting rates in the NCVS from 1987 to 1990 and discovered that 51 % of respondents had reported to police (1993). As well, Hart and Rennison (2003) found that 57 % of robberies, 55 % of aggravated assaults, and 31 % of rapes were reported to police from 1992 to 2001.

While sexual assault is underreported in all contexts, it is especially so in mainstream crime surveys (DeKeseredy, 2000; DeKeseredy & Schwartz, 2011). Lizotte’s (1985) study of the “uniqueness of rape” was the first to use national representative crime survey data to answer questions about reporting to police. He noted that the research consensus then was that rape was “vastly underreported and that this underreporting has the effect of minimizing our awareness of rape as a social problem” (p. 169), a view that persists today. Recent studies using a variety of samples confirm that rape and sexual assault are reported to police at rates ranging from 5 % to 50 % (DuMont, Miller & Myhr, 2003; Kaukinen 2002a, 2002b; Koss et al., 1987; Wolitzky-Taylor et al., 2010). This is markedly less than the percentage of non-sexual assaults reported to the police.

Reporting in Rural, Urban, and Suburban Areas

Recent research, especially studies of violence against women (e.g., DeKeseredy & Schwartz, 2009; Rennison, DeKeseredy & Dragiewicz in press), challenges conventional wisdom assuming lower crime rates in rural areas (Donnermeyer, 2012). Nevertheless, the NCVS reveals that for many crimes, reporting rates vary across urban, suburban, and rural areas. It is speculated that lower reporting of crimes in rural areas is a function of stronger means of informal social control (Carrington, 2007; Hogg & Carrington, 2006). Rapes are particularly less likely to be reported to the police in rural versus urban areas (Weisheit, Falcone, & Wells, 2006), which reflects the powerful influence of the “good ol’ boy network” (Websdale, 1998). Findings by DeKeseredy and Schwartz (2009) soundly illustrate this. They found that while rural residents are able to count on their neighbors to help prevent public crimes such as vandalism, many rural men rely on their male friends and neighbors, including those who are police officers, to support a violent patriarchal status quo, which these neighbors may interpret as acting on behalf of the common good. In rural sections of many states, widespread acceptance of woman abuse exists alongside related community norms that prohibit victims from publicly talking about their experiences and seeking social support (DeKeseredy & Schwartz, 2008; Lewis, 2003).

Key Factors Associated with Reporting

Stereotypes and victim blaming affect the reporting of rape (Caringella, 2009). For example, rapes that comply with stereotypes about “real rape” (Estrich, 1987)—rapes perpetrated outdoors by a stranger and that involve beatings—are more likely to be reported than those that do not fit the stereotypes (DuMont et al., 2003; Felson & Paré, 2005; Fisher, Daigle, Cullen & Turner, 2003; Lizotte, 1985; Rennison, 2002). While Bachman (1993) argued that that the victim’s relationship to the perpetrator is no longer a significant factor in reporting rape, critics state that this finding is anomalous. Despite increases in reporting of rapes by intimates to police, they are still less likely to be reported than those perpetrated by strangers (DeKeseredy, 2011).

Findings on underreporting of non-sexual violence against women by intimate partners are mixed, with some studies finding it under-reported to police (Block,1974; Gartner & Macmillan, 1995) and others concluding it is as likely to be reported as other crimes (Felson, Messner, Hoskin, & Deane, 2002; Felson & Paré, 2005). It is, though, important to note that claims that domestic violence is not “special” with regard to reporting (Felson & Paré, 2005) camouflage significant sex differences in reasons for reporting and non-reporting. Such claims also arbitrarily excise sexual assaults from domestic violence. Since violence by intimates and other known persons comprises the majority of violence against women, these types of violence have a disproportionate impact on reporting.

Research Questions

The extant literature offers much information on the reporting of violence to police, but many questions remain. The neglect of investigating situational contexts in research on violence against women contributes to a lack of understanding of the interaction of factors affecting reporting rates. That is to say that rather than being of fixed importance, particular variables may be more or less influential in different combinations that create different situational contexts. To understand whether this is the case, attention must be given to situational context. Given the dearth of knowledge in this area, the present study seeks to provide an initial understanding of the situational contexts most and least associated with police reporting across rural, suburban, and urban areas.

To address this objective, we focus on several research questions. The research questions are as follows: 1. What are the dominant situational contexts associated with police reporting among urban female victims of violence? 2. What are the dominant situational contexts associated with police reporting among suburban female victims? 3. What are the dominant situational contexts associated with police reporting among rural females victims of violence? 4. Are there differences and/or similarities among contexts most and least likely to be reported in urban, suburban and rural areas?

Data, Sample, Measures, and Analytic Strategy

Data

This study uses National Crime Victimization Survey (NCVS) data from 1992 through 2009 to address our research questions. NCVS data are collected using a stratified, multistage cluster design and are publicly available through the National Archive of Criminal Justice Data (Hubble, 1995; Rennison & Rand, 2007). The data are gathered at a sample of housing units and group quarters (such as college dormitories) in the United States and the District of Columbia. In each sampled unit, all persons age 12 or older are interviewed, either in person or on the phone. Interviews are repeated at 6 month intervals for a 3 year period. On average, from 1992 to 2009, 184,000 persons were interviewed in approximately 99,000 households annually for the survey. During the analytic period used here, the NCVS was characterized by response rates in the 90+ percentiles for households and from about the mid-80 to mid-90-percentiles for individuals.

Despite criticisms of crime surveys for underestimating violence against women (e.g., DeKeseredy & Schwartz, 2011), the NCVS is one of the few large probability sampled data sets available and it contains much useful information. The NCVS is particularly well-suited for the study of situational factors affecting reporting to police. First, the data offer a large sample size enabling investigation into specific groups of victims. Second, NCVS data include an extensive range of situational characteristics of violent victimizations, including weapon presence, type of violence, and injuries sustained that can be used to create situational contexts. Third, NCVS data include both reported and unreported violent victimizations. Still, the data are subject to limitations. First, the NCVS covers a limited set of violent victimizations: rape/sexual assault, robbery, and assault (aggravated and simple). Other types of violence such as homicide, kidnapping, emotional or psychological terror, and stalking are not included in these analyses. Second, though the data includes violence by cohabitors, it does not provide a variable specifically identifying these individuals as perpetrators. Third, marital status of the victim is measured at the time of the interview, and it is not necessarily the same as it was at the time of the victimization. While imperfect, the NCVS provides an excellent opportunity to examine violent victimization against female victims of violent crime.

Sample

To examine the situational contexts associated with reporting violence against females, this study uses attempted and completed non-fatal violent victimizations, which include rape, sexual assault, robbery, and assault (aggravated and simple). Standard NCVS definitions are utilized (BJS, 2011). Because the focus of this research is violence against women and whether it was reported to the police (by anyone), the working data are restricted to female victims of violence. In the NCVS, this means females victims age 12 or older only. Victimizations in which the victim did not know if the violence was reported were excluded from the analyses. The first year, 1992, is selected as the starting point as it represents the first year of NCVS data following a major redesign which improved measurement of issues associated with violence against women. The significant changes in pre- and post-redesign data indicate a break-in-series of the NCVS and combining the two is not generally recommended. The final year in our analytic sample is 2009. This year is selected because it is the most recent data available at the time of these analyses.

The year 2006 is excluded from our analyses because several changes were implemented in the NCVS at that time, including the elimination of centralized CATI, a reduction in sample, the inclusion of unbounded surveys, and modifications to the sampling frame. The revisions to the sampling frame had an extreme effect on 2006 estimates that were not due to changes in victimization or sampling variation (Truman & Rand, 2010). Unfortunately, these issues were especially problematic for rural areas. These issues were corrected in the 2007 data (and beyond). In the end, these criteria result in 18,638 unweighted violent victimizations against females that form the analytic data set used in this research.

Measures

Dependent Variable

The dependent variable, or outcome variable, of interest in this research is whether or not the violence was reported to the police. This variable is coded “1” if the violence was brought to the attention of the police by anyone, and “0” if it was not. Given this coding scheme, results presented in the tables describe the proportion of that particular situational context that was reported to the police. For example, a 0.87 indicates that 87 % of those situational contexts were reported to the police. This would indicate a particular context that is highly reported.

Control Variable

The conjunctive analyses are performed separately for the three geographic areas: urban, suburban, and rural. This facilitates comparisons between the situational contexts and associated probabilities of reporting for each of the geographic area. While one could add this variable into the creation of situational contexts and present one very large table, by presenting separate tables for each geographic area, it is easier to understand the findings and address the research questions. Urban, suburban, and rural areas are based on categories determined by the U.S. Office of Management and Budget (OMB), which defines central city, outside central city, and nonmetropolitan areas. For purposes of the present research, we utilize the more common language of urban, suburban, and rural areas.

Variables Used to Create the Situational Contexts

The situational contexts of violence against women are created using the conjunctive distribution of the response categories within each of the following five situational variable: victim/offender relationship (intimate, family, friends, stranger, don’t know); type of crime (rape/sexual assault, robbery, assault); marital status (never married, married, widowed, divorced, separated, missing data); weapon presence (yes, no, don’t know); and injury to the victim (yes, no).2 A simultaneous consideration of the attributes associated with these five variables results in 540 possible situational contexts of violence against women for each geographic area. This total is calculated by multiplying the number of categories within each situational variable (i.e., five [victim/offender relationship] x three [type of crime] x six [marital status] x three [weapon presence] x two [injury to the victim]. While 540 situational contexts are theoretically possible for each geographic area, all 540 may not all actually appear in the data.3

As is standard practice in conjunctive analysis, results presented in tables are based on the application of a minimum frequency “rule of ten” (Hart & Miethe, 2008; Miethe & Regoeczi, 2004; Ragin, 1987, Rennison, 2010). That is, only those situational contexts of violence against women actually appearing in the data ten or more times are considered. This approach minimizes attention to rare situational contexts. Situational contexts that are observed in the data ten or more times are referred to as dominant situational contexts, or simply dominant contexts. By employing the minimum frequency rule and examining only dominant situational contexts, the investigation focuses on empirical identification of the predominant situational contexts of violence against women and the proportion at which each context is reported to the police.

Analytic Strategy

Conjunctive analysis of case configurations is based on comparative approaches utilized in qualitative and quantitative multivariate analysis of categorical data which identifies the proportion of dominant situational contexts for which a particular outcome is present (Miethe, Hart, & Regoeczi, 2008). In the present research, the outcome of interest is whether the violence was reported to the police. Using this technique, situational attributes of variables contributing to reporting violence against women to the police can be rank-ordered and situational contexts where violence against women is more or less likely to be reported can be identified.

Using traditional statistical measures of central tendency and dispersion (i.e., the mean and standard deviation), conjunctive analyses establish empirical boundaries of normative and deviant situational contexts associated with violence reported to the police (Hart & Miethe, 2008). Normative situational contexts are those observed within one standard deviation from the mean situational context. These situational contexts are not shown in an effort to save space. Deviant situational contexts are those contexts observed more than one standard deviation above (i.e., situational contexts more often reported to the police) or below (i.e., situational contexts less likely to be reported to the police) the mean dominant situational context. Aside from the mean context (the context in the middle of each table with space above and below) these are the only contexts presented in the tables. Finally, the tables also indicate those situational contexts (if present) that are found two or more standard deviations from the mean are identified using a double thin lines.4

Findings

Before presenting the results of the conjunctive analysis, the descriptives of the variables used in the analyses are presented in Table 1. Almost half (49 %) of violence against the females was reported to the police. Nearly half (45 %) of the victims lived in suburban areas, 38 % lived in urban areas, and the lowest percentage in rural locations (16 %). Violence was most often committed by a friend (35 %); however, 32 % of the victimizations considered were committed by a stranger. Almost one-quarter (22 %) of the violence was perpetrated by an intimate partner. The violence most often took the form of an assault (82 %) and most often did not involve a weapon (73 %). Victims were generally not injured as a result of their victimization (70 %), and violence was most often experienced by those who have never been married (51 %).
Table 1

Violence against women descriptives of variables, 1992–2009 NCVS

 

Percent

 

Percent

Reported to the Police

 

Victim’s Marital Status

 

 Yes

49.3

Never Married

51.1

 No

50.7

Married

22.7

  

Widowed

2.4

MSA/Geographic area

 

Divorced

15.1

 Urban

38.2

Separated

8.3

 Suburban

45.2

Unknown

0.4

 Rural

16.2

  
  

Weapon Presence

 

Victim/Offender Relationship

 

No Weapon

73.2

 Intimate

21.7

Weapon

19.7

 Family

7.5

Don’t know

7.1

 Friend

34.9

  

 Stranger

32.2

Victim Injury

 

 Unknown

3.7

Not injured

69.8

  

Injured

30.2

Type of Violence

   

 Rape/Sexual assault

8.8

  

 Robbery

9.1

  

 Assault

82.2

  
The first research question is what are the dominant situational contexts related to reporting violence against urban women to the police? Only 346 of the theoretically possible 540 combinations of attributes used to create the situational contexts of violence against urban women were empirically observed in the data. Following the rule of ten (i.e., removing rare situational contexts found fewer than ten times in the data), 117 urban situational contexts associated with violence against women remained for the analysis. Because a table of 346 situational contexts is not possible to present, only deviant and the mean situational contexts of urban violence against women are presented in Table 2. These deviant situational contexts are ranked high to low based on the proportion of that context that was reported to the police.5
Table 2

Deviant situational contexts of violence against urban women and reporting to the police

https://static-content.springer.com/image/art%3A10.1007%2Fs12103-012-9164-4/MediaObjects/12103_2012_9164_Tab2_HTML.gif

The likelihood of reporting violence against urban females varies across the situational contexts. Table 2 presents the distinctive situational characteristics associated with high proportions of police reporting. Only two characteristics acted as main effects: Injury and weapon presence. Specifically, when the victim was injured, the police were likely to be contacted (12 of the 18 deviant contexts involved an injured victim). And when a weapon is brandished, (11 of the 18 deviant contexts), the police are likely to be contacted. Thus, when a female victim is injured or is victimized by an armed offender, regardless of the other characteristics of the situational contest, the police are likely to be contacted. In contrast, the other factors that combine in a violent victimization against an urban female are highly contextual. Stated differently, victim/offender relationship, the type of crime, and marital status each fail to demonstrate a consistent pattern among the contexts most likely to be reported. This indicates that victim/offender relationship, the type of crime, and marital status are highly contextual and their influence on reporting violence depends on the particular combination of those factors. Ten of the 18 deviant contexts include a stranger, but the other eight involve intimates (6) family (1) and friends (1) as offenders. Type of crime is contextual as ten of the 18 deviant contexts involve a robbery, while eight involve assault. Also important to note is that not one of these contexts include a rape/sexual assault in the most likely to be reported group of contexts. Marital status is also highly contextual. Of the 18 deviant contexts, five involve never married victims, six involve married victims, two include divorced victims, four include separated victims and one involves a widowed victim.

The deviant contexts found in the lower portion of Table 2 indicate those least likely to be reported to the police. Urban contexts least likely to be reported to police involved a rape/sexual assault, unarmed perpetrators, and uninjured victims. Thirteen of the 19 deviant contexts associated with low reporting involved an uninjured victim. Eighteen of the 19 deviant contexts involved an unarmed offender. And 11 of the 19 involved a rape/sexual assault. Victim and offender relationship and marital status failed to demonstrate consistent patterns among these deviant contexts suggesting they are highly contextual and that their influence on not reporting depends on the particular combination of factors that define the situational context.

The second research question focuses on reporting violence against suburban women. Findings show 331 of the 540 theoretically possible contexts were observed in the data. Following the application of the rule of ten, 109 suburban situational contexts remained. Table 3 presents the deviant situational contexts for violent crime against suburban women, ranked high to low in order of the relative proportion of each context being reported to the police.
Table 3

Deviant situational contexts of violence against suburban women and reporting to the police

https://static-content.springer.com/image/art%3A10.1007%2Fs12103-012-9164-4/MediaObjects/12103_2012_9164_Tab3_HTML.gif

Findings indicate that all five variables are contextual in a suburban area. Victim/offender relationship, the type of crime, marital status, weapon presence and injury do not demonstrate a consistent pattern among the contexts most likely to be reported. For example, eight of the 22 deviant contexts associated with high proportion of reporting involved a stranger as an offender. Twelve of these contexts involved a robbery, ten involved an assault, and one involved a rape/sexual assault. No pattern is evident for marital status, and ten of the 22 included unarmed assailants, and ten involved uninjured victims. Thus when considering contexts of violence against suburban women that are most likely to be reported, victim/offender relationship, the type of crime, marital status, weapon presence and injury are highly contextual and their influence on reporting violence depends on the particular combination of factors.

In contrast, situational factors associated with unlikely police reporting are evident when considering deviant contexts found at the bottom of Table 3. Rape/sexual assaults, an unarmed perpetrator, and an uninjured suburban female victim are associated with low likelihood of police reporting. Thirteen of the18 deviant contexts involved an uninjured victim, 17 involved an unarmed offender, and 13 involved rape/sexual assaults. Victim and offender relationship and marital status failed to demonstrate consistent patterns of “main effects” indicating their role in choosing not to report are highly contextual and depend on the particular combination of factors that define the situational context.

Turning next to rural areas, results demonstrate 243 of the 540 possible combinations of characteristics that define the situational context for police reporting of violence against rural women were empirically found in the data. Following the application of the rule of ten, 55 rural situational contexts remained. Table 4 presents the deviant situational contexts of violent crime against rural women, ranked high to low in order of the relative proportion of each context being reported to the police.
Table 4

Deviant situational contexts of violence against rural women and reporting to the police

https://static-content.springer.com/image/art%3A10.1007%2Fs12103-012-9164-4/MediaObjects/12103_2012_9164_Tab4_HTML.gif

The probability of violence against rural women to be reported to the police varies across situational contexts. The highest proportion of police reporting is found in situational contexts involving an assault. Every deviant context associated with high probability of reporting (nine of nine) involved an assault. The other four variables making up the contexts did not demonstrate main effects, suggesting they are highly contextual and depend on the particular combination of factors. Thus, the role of victim and offender relationship, marital status, weapon presence and injury to the victim in high proportions of police reporting are contingent on the context in which they are found.

Turning to contexts in rural areas unlikely to be reported, findings show that when a victim faces an unarmed offender, the police are unlikely to be contacted. Every deviant context associated with low reporting involved an unarmed offender. With only eight deviant contexts to consider, it is difficult to make definitive statements. However, with five of the eight contexts involving never married victims, findings suggest that marital status is important. Similarly, injury appears significant with five of the eight contexts included an uninjured victim. And finally, the deviant contexts indicate that rape/sexual assault is a main effect associated with non-reporting. While four of the eight involve rape/sexual assault, it is important to recognize that rape/sexual assault make up a very small proportion of non-fatal violence. Thus, its presence in 50 % of these contexts is indicative of a main effect. Victim/offender relationship appears contextual as no pattern is apparent.

To address the fourth research question, Table 5 offers a summary of the findings allowing a comparison across geographic area. Using this table, several conclusions are clear. First, factors that act as main effects or in a contextual manner vary across geographic areas. For example, while an injured female victim and an armed offense acts as a main effect of contexts most likely to be reported in urban areas, they do not in suburban or rural areas. Second, findings indicate that a variable may act as a main effect in contexts least likely to be reported, however, the variable may act contextually among contexts most likely to be reported (or vice versa). For example, weapon presence behaves differently contingent on whether one is considering most likely or least likely contexts to be reported to the police in suburban and rural areas. Third, just because a variable acts as a main effect associated with a low proportion of reporting (e.g., unarmed offender in rural areas), does not mean the opposite attribute (i.e., armed offender) acts as a main effect associated with a high proportion of reporting. Third, it appears that much of what is known about factors associated with police reporting best describes urban violence. These findings do not explain high levels of reporting in suburban or rural areas well. In some cases however, no differences were measured among geographic areas. For example, violence against a female by an unarmed offender, resulting in an uninjured victim, and a rape/sexual assault acts as main effects associated with low levels of reporting regardless of the geographic area considered. Finally, one factor proved itself to be contextual across all geographic areas and across high and low probability reporting: victim and offender relationship. The role of victim and offender relationship on the decision to report or not report violence against females is consistently contextual and contingent on the other factors that coalesce to create a specific context.
Table 5

Summary of main effects and contextual variables in violence against women by geographic areas

 

Most reported

Least reported

Main effect

Contextual

Main effect

Contextual

Urban

Injury

VO Relationship

Rape/Sexual Assault

VO Relationship

 

Weapon

Marital Status

No Weapon

Marital Status

  

Type of Crime

No Injury

 

Suburban

 

VO Relationship

Rape/Sexual Assault

VO Relationship

  

Marital Status

No Weapon

Marital Status

  

Type of Crime

No Injury

 
  

Weapon

  
  

Injury

  

Rural

Assault

VO Relationship

Rape/Sexual Assault

VO Relationship

  

Marital Status

No Weapon

 
  

Weapon

No Injury

 
  

Injury

Never Married

 

Conclusion

The results of this study highlight the importance of understanding the role of situational contexts of violence against women in reporting to the police. Factors that influence reporting, whether to police or social services, affect the provision of resources as well as our understanding of the nature of violent crime. Our findings indicate that research focused on identifying factors correlated with reporting of violence against women to police obscures significant differences across a variety of situational contexts and geographic areas. Conjunctive analysis sheds some light on the reasons that previous studies have found mixed results on factors related to reporting. The current study indicates that in addition to the individual influence of certain factors on the likelihood of reporting, combinations of factors forming varied situational contexts influence reporting.

Conjunctive analysis provides a new perspective on reporting of violence against women to police. By combining information about multiple contextual factors already available in extant data sources like the NCVS, this type of analysis can illuminate the dynamics of crime reporting more effectively than has been done in other analyses. This research demonstrates that factors associated with a high probability of reporting are not necessarily the same factors that are related to a low probability of reporting. It identifies how factors present as main effects in some geographic areas may not present as main effects in others. It also identifies new areas for research.

The fact that differing contexts promote and inhibit reporting across rural, urban, and suburban areas indicates that community factors are also important. While these findings do not as decisively and consistently identify particular contexts associated with reporting, they do so for non-reporting. This is even more remarkable given that the NCVS is a crime study, and incidents reported necessarily rise to the level of identification as a crime by respondents.

Our study responds to calls to consider violence against all victimized women, as well as looking closely at rape and other forms of woman abuse by intimate partners. Contextualizing these types of violence within violent crime broadly provides a macro-level perspective that is important for understanding violence and abuse. The assumption that theories of violence which pay attention to gender differences simply “imply that domestic and sexual assaults against women are particularly likely to go unreported” is flawed (Felson & Paré, 2005, p. 4). Certainly, scholars who study violence and gender have repeatedly argued that women experience violence in their lives as a continuum (Stanko, 2006), and that violence by strangers and intimates may be more similar than criminologists have traditionally assumed (Dobash, Dobash, Cavanagh & Lewis, 2004). Our results shed some light on the mixed findings of research on the reporting of violence against women to police. Previous studies have focused on incident, victim, perpetrator, seriousness, and regional characteristics. Results reported here indicate that the particular combination of these factors matters as well. The lack of consistent patterns across geographic areas challenges assumptions about the factors related to reporting to police. Understanding the combination of factors appears to be at least as important as incident related factors to reporting.

Heimer (2008) observes that:

research on violence against women using the NCVS has not reached its potential. There have been sustained efforts by the Bureau of Justice Statistics and some academic researchers to assess the levels of violence against women and describe distributions of victimization across subgroups of women over time. This descriptive research is a necessary first step. Although important questions remain to be answered in this first step, the major gap in research using the NCVS is in “explanatory” research. That is, research using rigorous analytic techniques to tease apart the complex patterns of individual characteristics and social contexts associated with violence against women. (p. 1)

Conjunctive analysis provides an opportunity to do just this, identifying the situational contexts that we will need to investigate to build and test theories about the causes of violence as well as responses to it. Indeed, this type of analysis answers Heimer’s call for researchers to begin to identify pertinent social contexts to consider across studies. We can learn as much from the contexts that fell out of the results as the contexts associated with greatest and least likelihood of reporting.

Implications

Although this is an exploratory study, our findings have implications for policy and practice. First, the results strongly confirm the disproportionate underreporting of rape and sexual assault to police. And, the data indicate that other factors in the event fail to motivate reporting in urban and suburban areas. Regardless of perpetrator characteristics, sexual assault is a highly gendered crime that must be handled differently than other crimes. Reporting rates cannot be used to determine the need for resources for responding to sexual assault. Second, looking at violence against women as a whole reminds us that crimes perpetrated by friends and strangers comprise a significant portion of crimes against women. Programs for crime victims should address the spectrum of crimes against women rather than simply that by strangers and intimates. Third, since the contexts associated with high and low reporting vary across urban, rural, and suburban areas, it is important to tailor resources to the area in which they will be delivered. Practices that show evidence of effectiveness in one geographic area may not be the most appropriate in other areas.

Future Research

The results of this study raise several questions for future research. For example, why does the role of weapon presence depend on the other factors in the incident among high level reporting contexts when it acts as a main effect for low level reporting contexts for suburban and rural areas? Why are the reporting rates for marital status almost always contextual? Does this indicate that it is not a significant factor in reporting? Or does it suggest that it is important, but only in conjunction with some other factor? Why are more factors contextual in rural and suburban settings than in urban settings when it comes to unlikely contexts to be reported? Is there something about suburban and rural areas that inhibit reporting to the police? Are these structural in nature and do they influence reporting of victimization against males in similar fashions? Future studies using NCVS data to compare reasons for not reporting in urban, rural, and suburban areas would be a logical next step to understanding the situational contexts. Following further excavation of the NCVS data, the difference in contexts across geographic areas calls for in-depth and qualitative research to help explain why and how these contexts matter.

The results also point generally to the need for additional research on the situational contexts that affect reporting of violence in general and violence perpetrated by intimates in particular. Many of the victimizations reported to the NCVS are perpetrated by friends and strangers, yet there has been little research on violence against women by friends and strangers outside of sexual assault. It would also be interesting to investigate the situational contexts affecting reporting to police by sex. Does the importance of contexts differ by sex for perpetrators or victims of the same type of crime? It would also be desirable to compare the situational factors affecting reporting of violence against women in the NCVS with other large data sets, especially those collected in a non-crime victimization context, in order to ascertain their applicability to a wider range of victimizations. This and other future research can enhance our understanding of reporting violence to the police by identifying the complex interplay of a variety of variables. Through better understanding, useful policy may be implemented to better serve all female victims of violence regardless of the geographic area they find themselves.

Footnotes
1

‘Violence against women,’ in this research refers to violent victimization experienced by females by any perpetrator. Perpetrators may be strangers, family members, friends or intimate partners.

 
2

Conjunctive analysis can be conducted using standard software packages such as SPSS and SAS. The analyses presented here were conducted using SPSS. For a full and detailed description of how to conduct conjunctive analysis, as well as the actual code used for a variety of software packages, see Miethe, Hart and Regoeczi (2008).

 
3

540 situational contexts are possible for each of the areas considered (urban, suburban and rural). Taking into consideration the area variable, 1,620 possible situational contexts are theoretically possible as 540 * 3 = 1,620.

 
4

If present in the data, these contexts appear above the double thin lines at the top of the table, or below the double thin lines at the bottom of the table.

 
5

Only the deviant (more than one standard deviation above and below the mean) and the mean situational contexts are shown in the following tables due to space considerations.

 

Copyright information

© Southern Criminal Justice Association 2012