Abstract
The current study assesses the relative influence of various individual-level characteristics on the probability of intimate partner violence (IPV) for separated and nonseparated women. While previous studies have found that separated women do in fact have a higher risk for IPV than nonseparated women, these largely bivariate examinations of marital status and risk for IPV have often not considered the effect other characteristics may have on risk estimates. The current study uses the 1995-2010 National Crime Victimization Surveys to examine how separated women’s risk for IPV compares with nonseparated women’s risk for IPV over time, and if separated, women’s risk for IPV is a function of either being separated or possessing characteristics known to be correlated with risk. A key strength of this study is its ability to account for the confounding effects of change in separation status and IPV. Results show that separated women were more likely than nonseparated women to be victims of IPV in most years from 1995 to 2010, and after controlling for the effects of individual-level characteristics, their risk did not change. Age was the only significant predictor of women’s risk for IPV, net of other factors, but had no effect on separated women’s risk for IPV. These results suggest that the status of being separated has the strongest effect on women’s risk for IPV. The importance of understanding how the separation period makes women more likely to be victims of IPV is discussed.
Do women who are separated from their intimate partners experience a greater risk of nonlethal physical and sexual abuse from their partner (i.e., intimate partner violence [IPV]) than nonseparated women? To what extent is separated women’s risk for nonlethal physical and sexual abuse, including assault, robbery, rape, or sexual assault, by their intimate partner contingent on their status of being separated, or a function of their individual-level traits, such as their demographic and socioeconomic characteristics? Unfortunately, previous investigations into the relationship between women’s marital status and risk for IPV have not satisfactorily addressed these questions using robust methodological designs that account for both the temporal ordering of separation and IPV and the confounding nature of separation status and various individual-level characteristics. Without a clear assessment of which occurred first, the act of separation or the IPV incident—temporal details that are necessary to accurately measure postseparation risk for IPV—estimates of IPV risk for separated women may be misleading. This study addresses these issues while also exploring trends in separated women’s risk for IPV from 1995 to 2010 in an attempt to parse out the influence of separation on women’s risk for IPV over time.
Over the last few generations, the status of women in the United States has changed in significant ways. Improvements in the economic well-being of women resulting from, among other things, increases in women’s labor force participation, have provided women greater independence over the past 40 years. This independence has extended itself into areas such as women’s marital relations, where economic independence has been shown to be an important factor in women’s ability to leave their spouses (Goldin, 2006). More than ever before, American women dictate whom they date, whom they marry, whom they separate from, and whom they divorce, without the same degree of stigmatization that previous generations endured (Goldin, 2006).
Consequently, marital separation rates steadily increased throughout the latter half of the 20th century (Ruggles, 1997). These increases were even more significant for individuals of certain demographic backgrounds such as those less educated and non-White (Raley & Bumpass, 2003). With divorce being a viable option, legal or nonlegal separation (i.e., the phase during which one physically leaves one’s spouse due to marital discord, but is not divorced) has become a common life transition for married women looking to divorce their spouse or live apart from them due to marital conflict.
Although the separation period varies in length, it typically follows unresolved marital discord and precedes either divorce or a repaired marriage. Therefore, a significant number of married women likely end up separated at some point prior to finalizing divorce, given that roughly 33% of first marriages and 40% of second marriages end in divorce by their 10th anniversary (Bramlett & Mosher, 2001). In fact, Kitson (1985) found that one in six married couples is likely to separate due to marital discord for at least 2 days at some point in their marriage. While marital discord can result from many factors (e.g., issues with money, infidelity, inconsistent expectations), marital discord resulting from husbands’ physical aggression toward their wives was cited by Kurz (1996) as a significant factor leading to marital dissolution (i.e., separation or divorce) among wives. Similarly, Lawrence and Bradbury (2001) found spousal aggression to be a key predictor of marital discord prompting marital dissolution over the first few years of marriage. In their study of 56 married couples, the authors found that 93% of couples reporting severe physical aggression experienced acute marital distress and eventual marital separation or divorce within the first 4 years of marriage compared with 46% of couples in moderately aggressive marriages and 39% of couples in nonaggressive marriages. Unfortunately, separation from one’s violent spouse does not necessarily end the violence, specifically for women: Evidence shows married women have a heightened risk of IPV after separation (Hotton, 2001; Wilson & Daly, 1993).
To date, there is limited research on the consequences associated with separation and IPV. There are many reasons for the scant research, but one commonly raised is that the literature on marital discord and separation for women has been developed in isolation from the victimization literature and vice versa (Kurz, 1996). Moreover, few of the studies that do exist are based in the United States. These studies and others from Canada tend to offer correlational evidence of the relationship between separation status and risk for IPV. This complex unidirectional or bidirectional relationship between separation and IPV exposes the potential for researchers to confuse the temporal ordering of a woman’s change in her marital status and the timing of her IPV incident, as variations in the causal ordering between separation and IPV are common.
Previous research, for example, has established that violence during a marriage is associated with a couple’s likelihood of separating (see Amato & Rogers, 1997). In addition, research has found that separating from one’s partner is associated with a partner’s risk for future IPV (see Wilson & Daly, 1993). However, there is also support that this relationship, in many cases, may in fact be bidirectional (see Fleury, Sullivan, & Bybee, 2000; Kurz, 1996). For example, a woman who is abused by her husband while married may physically separate from her spouse because of the abuse, yet she may still be victimized by him during the separation period. Thus, victimization leads to separation, which may lead to further victimization.
Unfortunately, many studies of separated women’s risk for IPV do not account for the temporal ordering of separation and victimization (e.g., Bachman & Saltzman, 1995; Siddique, 2016; Spiwak & Brownridge, 2005). Without a clear assessment of which occurred first, the act of separation or the victimization incident, these studies potentially confound separation and victimization so it is unknown if, in fact, women truly have a higher risk for victimization postseparation. In many cases, it may be that the methodological design of the study was such that, even though the women were currently separated, abuse by their spouse was the precipitating factor that led to the couple’s separation. In such cases, the causal process is flipped, but potentially unknown to the researcher. If the marital status of separated women was collected postvictimization, it could be that they were either married or separated at the time of their victimization. Without additional temporal details about whether the change in marital status or the IPV incident came first, estimates of IPV risk by marital status are misleading, especially for separated women who are likely in a transitional phase between marriage and divorce. One strength of this study is that the temporal ordering of marital status and victimization has been resolved. This was possible by limiting the analysis to women who remained in the same marital status category over the entire analysis period (i.e., pre- and post-IPV incident).
Even though marital status and IPV are potentially confounded in many previous studies of separated women’s risk for IPV, the extant literature does suggest that women have both an elevated risk of IPV postseparation and a significantly higher risk for IPV when separated than nonseparated (e.g., Bachman & Saltzman, 1995). Yet, previous research in this area has not successfully established if separated women’s higher risk for IPV is an artifact of compositional differences across women of different marital status groups, specifically those who end up separated or not. This becomes an issue when researchers compare risk for IPV by marital status because selection into a marital status group may share some of the same characteristics that have also been shown to be correlated with IPV (e.g., age, socioeconomic status). Furthermore, it is possible that previous research has overlooked some of the characteristics associated with separated women’s risk for IPV. A second strength of this study lends itself to this issue because several important demographic, socioeconomic, and family structure characteristics of women have been controlled in the model-adjusted risk estimates.
A third strength of this study assesses the potential for change over time in IPV risk estimates for separated women. Because the characteristics of separated women may have changed from 1995 to 2010 (e.g., the average age of separated women may have decreased), this study compares trends in separated women’s risk for IPV over the past 16 years using both baseline model and model-adjusted risk estimates. For example, if separated women’s risk for IPV has decreased from 1995 to 2010, independent of their individual-level characteristics, it may be that increases in both exposure-reducing resources (e.g., the availability of domestic violence prevention services, protection orders, and mandatory arrest policies) and/or gender equality over the past two decades have had an effect on lowering their risk (Dugan, Nagin, & Rosenfeld, 2003; Xie, Heimer, & Lauritsen, 2012). Therefore, this study may produce important recommendations for practitioners related to the timing and delivery of programs aimed to assist women most at risk for IPV, especially if separated women’s risk for IPV, net of individual characteristics, has decreased over time but is still higher than the risk for IPV of women in all other marital status groups.
Measuring Separated Women’s Risk for IPV
A review of the literature in both the United States and Canada suggests that women are at greater risk for IPV when separated from their spouses than at any other time in their marital lives. For example, one survey-based study based in Canada found that roughly one third of separated women experienced violence or threats of violence by their respective former spouses after separation (Ellis & Stuckless, 1992). Using the Canadian Violence Against Women Survey, Johnson and Sacco (1995) reported that among women who were abused in their marriage and who subsequently left their spouses, 19% were victims of postseparation IPV. In one third of those cases, the violence was described as much more severe postseparation. Moreover, Hotton (2001) reported that 39% of women in Canada’s 1999 General Social Survey who left an abusive marriage were later victims of IPV.
In fact, there is evidence that separated women have a greater risk for IPV than nonseparated women, but that evidence has been limited to very few studies that measure separated and nonseparated women’s risk for IPV. For example, Kennedy and Dutton (1989) examined the incidence of female IPV using the Conflict Tactics Scale in a representative sample of residents of Alberta, Canada, and found that a larger proportion of separated women (54.8%) reported experiencing IPV in the year prior to the survey than married women (8.7%), divorced women (39.8%), and single women (43.7%). Using the 1973-1975 U.S. National Crime Survey (NCS), the predecessor of the National Crime Victimization Survey (NCVS), Gaquin (1977) found that separated women in the United States were nearly 3 times more likely to be a victim of IPV than divorced women, and 30 times more likely than married women. Similarly, a study using the 1992-1993 NCVS found that separated women were 3 times more likely to be victims of IPV than divorced women, and roughly 25 times more likely to be victims of IPV than married women (Bachman & Saltzman, 1995). More recently, using the NCVS, Siddique (2016) found that separated women had the highest risk for sexual victimization by an intimate partner compared with women in the other marital status groups. Unfortunately, the analytical designs of these four studies failed to account for the temporal ordering issue described above. Because of the unknown time-ordering of the IPV incident and one’s marital status, it is unclear whether the women in these marital groups, especially those married or separated, were actually in the reported marital status category at the time of their victimization. This limitation makes it impossible for readers to properly evaluate the accuracy of the rates, ratios, or results published in these studies.
One reason that there is scant literature on separated women’s risk for IPV is that many studies of risk for IPV by marital status do not isolate separated women from other marital status groups of nonseparated women. For example, the incidence and prevalence of IPV for separated women alone is often not reported in studies because separated and divorced women are frequently pooled together as one group (e.g., Lauritsen & Heimer, 2009). In addition, official government reports either focus specifically on women’s risk for IPV without disaggregating by marital status (e.g., Catalano, Snyder, & Rand, 2009) or disaggregated annual rates of IPV by marital status groups are limited to rates for males and females combined (e.g., Truman, Langton, & Planty, 2013). Because these studies report simple rate estimates without controlling for potential confounding individual-level characteristics that previous research has found may affect women’s marital status and their risk for IPV, a more complex model of risk that adjusts for these covariates is necessary to generate a more thorough understanding of separated women’s risk for IPV.
Factors Affecting Separated Women’s Risk for IPV
Numerous studies have examined risk factors associated with IPV, with many finding overlap in the demographic and socioeconomic victim characteristics that increase a woman’s likelihood of IPV. However, in the overwhelming majority of these studies, marital status was either not controlled in the models (e.g., Kramer, Lorenzon, & Mueller, 2004), or separated women were not isolated from other women, specifically divorced women (e.g., Romans, Forte, Cohen, Du Mont, & Hyman, 2007; Smith, 1990; Vest, Catlin, Chen, & Brownson, 2002). And, even when separated women are singled out for analysis, it is unclear whether these characteristics are related to a woman’s selection into the “separated” marital status, or that these characteristics genuinely increase a woman’s risk for IPV as a result of her experience being separated. Studies that examined risk factors that increase a separated woman’s likelihood for IPV, but may not have considered the temporal ordering of separation and IPV found that separated women who were both young in age (Brownridge et al., 2008; Spiwak & Brownridge, 2005), and who lived in rural areas (Rennison, DeKeseredy, & Dragiewicz, 2013) had an increased risk for IPV. In addition, separated women with a previous history of IPV while living with their spouses were more likely to experience IPV postseparation (Spiwak & Brownridge, 2005). Research has also found that it is often difficult for separated women who have been abused to escape from their violent partners. Therefore, IPV postseparation continues. In some cases, it may be nearly impossible for women to leave if they rely heavily on their abuser for resources or they share custody of their children with their abuser (McMurray, Froyland, Bell, & Curnow, 2000).
Studies of women’s risk for IPV, regardless of a women’s marital status, are plentiful. These studies found that women who are younger (Bachman & Saltzman, 1995; Kramer et al., 2004; Siddique, 2016; Vest et al., 2002), less educated (Bachman & Saltzman, 1995; Kramer et al., 2004), less affluent (Bachman & Saltzman, 1995; Cho, 2012), living in an urban area (Rennison & Welchans, 2000), or living with children in their home (Lauritsen & Schaum, 2004) reported experiencing higher risk for IPV. These results held up in several of these studies even after controlling for various individual, family, or community-level factors (Cho, 2012; Kramer et al., 2004; Lauritsen & Schaum, 2004; Vest et al., 2002). Women of color have also been found to have a higher risk for IPV than White women in some studies (Rennison & Welchans, 2000; Tjaden & Thoennes, 2000); however, race differences have been dismissed by other studies either overall (Kramer et al., 2004) or after controlling for socioeconomic status (Cho, 2012; Rennison & Planty, 2003).
Trends in IPV
The cross-sectional design of the majority of the studies described above is limited in that all of these studies examine separated women’s risk for IPV at one moment in time. Previous research has not yet determined whether the relationship between separation status and IPV persists over time in the United States, especially given that the characteristics of separated women may have also changed over the past few decades. Notably, criminologists have found that both lethal and nonlethal violence against women have declined over the past 30 years in the United States (Dugan, Nagin, & Rosenfeld, 1999; Lauritsen & Heimer, 2009). Using the NCS and NCVS from 1973 to 2005, Lauritsen and Heimer (2009) found that female rates of victimization, including IPV, decreased even when disaggregated by various correlates of victimization. More recently, Powers and Kaukinen (2012) produced trends in risk for female IPV by race/ethnicity and employment status using the NCVS and also found declines across race and employment categories. However, it remains unknown if trends in separated women’s risk for IPV using the NCVS mirror the declines found in risk for IPV for all women in the United States.
Incorporating temporal trends in risk for IPV into the current study is critical to understanding separated women’s risk for IPV for two reasons. First, because prevalence rates reported by Gaquin (1977) from the early years of the NCS show that separated women’s risk for IPV was at least 3 times higher than risk found in other marital status groups, it is important to assess potential changes in their risk for IPV over a more recent period. Second, using the NCS and NCVS, Lauritsen and Heimer (2009) produced incident rates of female IPV disaggregated by marital status (never married, married, and divorced/separated) that revealed unique trends over the past three decades, demonstrating the need for further disaggregation of the separated and divorced female population in the United States.
The Current Study
This study addresses the following questions: (a) To what degree is the risk for IPV a function of individual-level characteristics of separated women? (b) How is the model-adjusted risk for IPV (which controls for potential confounding individual-level characteristics) different between separated and nonseparated women? and (c) Has separated women’s model-adjusted risk for IPV remained stable from 1995 to 2010?
While previous research has already established a bivariate association between separation and IPV, this research looks at whether that relationship persists after individual-level covariates of risk have been controlled for. This study also extends previous research by accounting for the temporal ordering of separation and IPV before estimating risk for IPV for separated women, as well as controlling for the potentially confounding effects of women’s individual-level characteristics on their risk for IPV by marital status. Additional analyses explore whether separated women’s risk for IPV changed from 1995 to 2010. Therefore, this analysis provides fundamental information about the differences in the level of risk for IPV for separated women in the United States compared with nonseparated women, net of many individual-level factors, and a thorough assessment of the trend in separated women’s risk for IPV from 1995 to 2010, accounting for the potential confounding effect of various individual-level factors.
Data
This analysis uses data from the National Crime Victimization Survey (NCVS), a nationally representative household survey conducted by the U.S. Census Bureau on behalf of the Bureau of Justice Statistics. Over the last few decades, the NCVS has been the nation’s primary source of information on the frequency, characteristics, and consequences of criminal victimization in the United States. NCVS data are collected annually using a stratified, multistage cluster sampling design from a nationally representative sample of roughly 50,000 households consisting of more than 100,000 individuals age 12 or older. In the NCVS, respondents are asked twice a year to report personal victimizations during the last 6 months. A series of cues and questions by the survey interviewer is designed to help elicit respondents’ victimization experiences. Because violent victimization is relatively rare in random samples of the population, the large sample size of the NCVS makes it ideal for capturing IPV.
Data from the 1995 to 2010 NCVS were used in this study. Annual NCVS person-level and incident-level files were merged to create annual prevalence files, where the unit of observation is the respondent’s interview, and victimizations (up to four) in the previous 6 months are linked to the respondent. A victim in this analysis is any female respondent age 18 or older who reported experiencing one or more rapes, sexual assaults, robberies, aggravated assaults, or simple assaults by an intimate partner during a 6-month period in a given year. Analyses of prevalence risk for IPV does not take into account the number of IPV victimizations a victim experiences, only whether she experienced any IPV victimizations. IPV victimizations occurring outside of the United States are excluded. Analyses considered women aged 18 and above in one of the following marital status categories: never married, married, separated, divorced, and widowed.
Measures
Dependent variable
The outcome measure is 6-month risk for IPV. IPV includes any rapes, sexual assaults, robberies, aggravated assaults, and simple assaults that were perpetrated by a woman’s intimate partner in the past 6 months and reported in the NCVS. An intimate partner in the NCVS is defined as a current or former spouse, boyfriend, or girlfriend (i.e., including same sex relationships). Risk for IPV is a dichotomous variable representing 1 if the female respondent reported being a victim of any of these violent crime types during the 6 months prior to her interview and 0 otherwise.
Independent variables
The key independent variable is the marital status of a woman, as recorded by the NCVS interviewers at the time of the interview. In the NCVS, a respondent’s marital status can be one of five classifications: never married, married, separated, divorced, or widowed. Women’s marital status is included in this analysis as a series of dichotomous measures representing these five categories. Married women serve as the comparison group in the regression analysis because previous research suggests married women have lower rates of IPV compared with other women. Therefore, using them as the reference category allows for easier interpretation of the coefficients for women of other marital status groups.
Unfortunately, the NCVS is not capable of determining the marital status of a woman at the exact time of her victimization. However, during each semiannual interview, NCVS interviewers elicit from respondents which marital status group they currently belong to (referred to as their ‘current’ marital status), as well as their marital status group during their previous interview 6 months before (referred to as their “previous” marital status). Because the “current” and “previous” marital status categories of most women are known, all analyses have been restricted to women who were in the same marital status group during their “previous” and “current” interview within a given 6-month period. This restriction omits approximately one quarter of women from the analyses each year. 1 Such a restriction is necessary because it is the only way of capturing a woman’s marital status at the time of her IPV victimization.
Even though the NCVS questionnaire is not designed to elicit the marital status of a woman at the time of her victimization, it is reasonable to assume that women who have not altered their marital status across the two survey periods in a given year (i.e., 6 months apart) were likely in that marital status category at the time of their victimization. In situations where the “current” or “previous” marital status of a woman could not be determined at the time of the interview it was coded as missing. Any woman with a missing “current” or “previous” marital status has also been excluded from the analysis. This includes women in households that were new to the NCVS or missed one of the two interviews that year.
NCVS interviewers record a respondent’s marital status as separated when they are separated from their spouse due to marital discord; married, but have a legal separation; married, but have parted from their spouse because of marital discord; married and expect to obtain a divorce in the future; or separated, and intend to remain separated from their spouse permanently and never get a divorce. 2
Control variables
A number of control variables available in the NCVS that have been shown in the literature to potentially have an association with women’s risk for IPV are also included. To control for some of the individual-level characteristics that could possibly account for the relationship between separation and IPV, and to allow for a more complete understanding of how marital status affects variation in the risk for IPV, the following demographic, socioeconomic, location of residence, and family structure control variables are considered.
Because separated women could be demographically different from nonseparated women, a woman’s self-reported age and race/ethnicity are included. In this study, age is measured with a continuous variable that reflects the age reported by the woman at the time of her interview. A woman’s “race and ethnicity” was self-reported and combined into the following response categories: non-Hispanic White, non-Hispanic Black, Hispanic, and other race/Hispanic unknown (with non-Hispanic White serving as the reference category).
Measures of socioeconomic risk in this study include whether the woman had obtained a high school degree (yes vs. no), whether she was not employed in the week prior to the interview (yes vs. no), and her household’s poverty status (one of the following three dichotomous measures: at or below the poverty line, above the poverty line, or missing data on the poverty measure). Households above the poverty line serve as the reference category.
The location of the female respondent’s residence (i.e., urban, suburban, or rural) was coded at the time of the interview. These dichotomous measures are based on the location of the household and are one of three categories: the city of a Metropolitan Statistical Area (i.e., urban), a noncity area of a Metropolitan Statistical Area (i.e., suburb), and not a Metropolitan Statistical Area (i.e., rural). Urban areas serve as the reference category.
A woman’s family structure is operationalized with two dichotomous measures: whether any children under age 12 are present in her household (yes vs. no), and whether she lives alone (yes vs. no). If a woman does not live alone, she could live with either adults or children.
Analytical Strategy
Because most previous research on women’s risk for IPV fails to disentangle separated women from women of other marital statuses, a descriptive overview of the women in the NCVS from 1995 to 2010 is presented to observe differences across, and changes over time in, the individual-level characteristics of women by their marital status group, with an emphasis on how separated women compare with nonseparated women. For ease of interpretation, the percentages in 1995 and 2010, as well as the average of percentages across all 16 years of data are reported for all measures.
Second, survey-weighted logistic regression models—a type of binary outcome model—are used to estimate the baseline 6-month risk for IPV for women of different marital status groups each year from 1995 to 2010. Survey-weighted logistic regression models are necessary to account for the methodological design of the NCVS. Survey-weighted modeling takes into account the stratified, multistage cluster sampling design. When NCVS data are weighted, sample estimates are generalizable to the entire U.S. population of women age 18 or older residing in households. Adjusted Wald tests were performed for each year of data in all regression models to test whether separated women’s risk for IPV was statistically greater than the risk for IPV for divorced and never married women. 3
Third, a number of control variables are added to survey-weighted multiple logistic regression models to examine the influence these individual-level characteristics have on women’s 6-month risk for IPV by marital status for each year from 1995 to 2010. Multiple regression analyses attempt to generate more precise estimates (i.e., full model or model-adjusted risk estimates) of female risk for IPV by holding constant other potential correlates of victimization that may confound the estimation of risk for IPV (see Lohr & Liu, 1994, and Long & Freese, 2006, for a detailed discussion of this regression technique). By including demographic, socioeconomic, and family structure measures as control variables, 6-month risks for IPV are adjusted for differences in the composition of women across marital status groups.
Finally, predicted probabilities of separated women’s 6-month risk for IPV are estimated and graphed using estimates from both baseline and full regression models. By observing the nature of temporal trends in separated women’s risk for IPV in both the baseline and full models, this study is not only able to assess the level of risk for separated women from 1995 to 2010 but also gauge the degree to which the control variables may have confounded risk estimates for separated women (for a discussion of the utility of predicted probabilities see Long & Freese, 2006 and Williams, 2012). Risk estimates in the figure have been smoothed using 3-year moving averages to reduce fluctuations associated with sampling error.
Results
As shown in Table 1, on average, the majority (57.3%) of women age 18 or older from 1995 to 2010 were married, with nearly 20% never married, 2% separated, 11% divorced, and 11% widowed. Overall, the median age for separated women was 41 to 46 years old, while on average, never married women were younger and divorced women were older. Roughly three quarters of women were non-Hispanic White, although this percentage declined from 1995 to 2010 for women in all marital status categories. Separated women were significantly less likely to report being non-Hispanic White (compared with never married, t = −18.34; married, t = −42.81; divorced, t = −49.66; widowed, t = −52.40), while women of all other marital status groups were predominately non-Hispanic White. Bivariate associations across all years revealed that an average .69% of separated women were victims of IPV during the 6 months prior to their interview (i.e., nearly 7 per 1,000 separated women). On average, separated women were significantly more likely than divorced (t = 4.03) and never married women (t = 3.91) to be victims of IPV. In contrast, widowed (compared with never married, t = −11.67; separated, t = −6.47; divorced, t = −8.09) and married women (compared with never married, t = −9.72; separated, t = −6.08; divorced, t = −6.68) were the least likely to be victims, a reasonable finding for widowed women considering one of their main potentially abusive partners is deceased.
Descriptive Statistics for Women Age 18 or Older by Marital Status, NCVS 1995-2010.
Note. Statistics are based on weighted data unless otherwise noted. Mean is the average of the 16 annual percentages from 1995 to 2010. NCVS = National Crime Victimization Survey; IPV = intimate partner violence.
Bivariate associations between socioeconomic measures of risk and marital status revealed that widowed women were significantly more likely to have never received a high school degree (compared with never married, t = 13.31; married, t = 18.81; separated, t = 6.62; divorced, t = 18.11), a finding likely because widowed women tend to be older and less likely to have had the opportunity to further their education when they were younger. Of the remaining marital status groups, separated women had significantly higher rates of high school noncompletion than never married (t = 11.48), married (t = 27.66), and divorced women (t = 24.53). On average, 44% of women did not have employment in the past week. However, the inclusion of married and widowed women in that figure may have upwardly influenced the percentage of women not employed. Because there is a greater chance that never married, separated, and divorced women do not live with someone else contributing to their household income, their average percentages not employed in the previous week are more alarming (32.7%, 37.5%, and 31.9%, respectively). The average percentage of all women who lived at or below the poverty line declined approximately 39% from 1995 to 2010, with comparable declines found for all marital status groups. Separated women, however, were significantly more likely to be living in poverty than women of all other marital status groups over the years (compared with never married, t = 6.27; married, t = 13.80; divorced, t = 7.47; widowed, t = 4.33).
While separated women were significantly more likely than married (t = 19.56), divorced (t = 10.25), and widowed women (t = 12.49) to live in an urban area, separated and never married women had similar rates of urban residence. On average, one quarter of separated women lived alone, compared with less than 1% of married women, and approximately 40% of divorced women. A significantly higher rate of separated women than never married (t = 15.42), married (t = 6.38), divorced (t = 16.47), and widowed women (t = 36.36) lived with at least one child under the age 12. Thus, separated women differed from nonseparated women in a number of ways: They were more likely to be minorities, less educated (excluding widowed women), living at or below the poverty line, living in an urban area (similar to never married women), and living with at least one young child. These are all characteristics or circumstances that previous research suggests predisposes women to greater risk for IPV.
Table 2 reports the coefficients and standard errors for the baseline logistic regression models predicting female risk for IPV by marital status. Compared with married and widowed women, separated, divorced, and never married women were significantly more likely to be victims of IPV during the 6 months prior to their interview across the entire series from 1995 to 2010. Adjusted Wald test results showed that separated women were significantly more likely than divorced and never married women to be victims of IPV in half of the years from 1995 to 2010. No significant differences were found in risk for IPV between separated women and divorced or never married women in all other years. 4
Coefficients and Standard Errors of Women’s 6-Month Risk for Intimate Partner Violence, NCVS 1995-2010.
Note. (n/a)—There were no incidents of intimate partner violence for widowed women captured in the 1998, 2004, 2005, 2006, and 2010 NCVS. Reference category is married. NCVS = National Crime Victimization Survey.
p < .05. **p < .01. ***p < .001.
The next step in the analysis determined whether this finding persists when various demographic, socioeconomic, and family structure measures are included in the models. The inclusion of these measures adjusts for differences in the compositional attributes of women across these marital status groups over the 16 years of the study.
The results shown in Table 3 revealed that even after controlling for these characteristics, separated and divorced women were still significantly more likely than married women to be victims of IPV. Adjusted Wald tests found that separated women’s risk for IPV in the multiple regression models was significantly higher than divorced women’s risk in only 4 of the 8 years found in the baseline regression model (1998, 1999, 2005, and 2010). Separated women were also significantly more likely than never married women to be victims of IPV in the same years reported in the baseline regression models, and 2 additional years in the full models. Aside from marital status, the only characteristic consistently associated with risk for IPV in the full models was age, where it was negatively associated with risk, net of all other factors.
Coefficients and Standard Errors of Women’s 6-Month Risk for Intimate Partner Violence, NCVS 1995-2010.
Note. There were no incidents of intimate partner violence for Other Race/Hispanic Unknown women in 2008. (—) Nonsignificant coefficients are not shown. Variables with nonsignificant coefficients across all years have been deleted from table for visual purposes. These are widowed vs. married, non-Hispanic Black vs. non-Hispanic White, suburban residence vs. urban residence, and rural residence vs. urban residence. Reference categories include married, non-Hispanic White, high school diploma, employed, above the poverty line, no child present, and lives with ≥1 other person. NCVS = National Crime Victimization Survey.
p < .05. **p < .01. ***p < .001.
Interestingly, when the same models are estimated on a sample of women by their “current” marital status (which does not control for potential changes in a woman’s marital status over the 6-month survey period), women at or below the poverty line had higher odds of becoming a victim of IPV during the 6 months prior to their interview compared with those above the poverty line for one third of the years from 1995 to 2010. In addition, using women’s “current” marital status, women with a child present in their home and those who lived alone were consistently more likely to be victims of IPV in the 6-months prior to their interview. This finding suggests that there may be certain circumstances in which a woman is more likely to change her marital status pre- and post-IPV, and one of those situations may be that she has limited resources due to living in poverty, living alone, and/or having a child present in her household.
As a way of estimating the level of risk that separated women face after controlling for the effects of all individual-level covariates, baseline, and model-adjusted risks for separated women were calculated and graphed using the logistic regression results from Tables 2 and 3. Figure 1 displays the smoothed 6-month risk estimates for IPV for separated women in both baseline and model-adjusted regression models. As shown, separated women’s model-adjusted 6-month prevalence estimates hardly deviated from their baseline estimates, indicating that these other variables were not as influential as previous research has suggested. However, separated women’s risk for IPV during the 6 months prior to their interview did fluctuate from 1995 to 2010. While these fluctuations may not seem to be substantively significant, it is important to note that for a few years in the series, separated women’s 6-month prevalence of IPV is 2 to 3 times higher than in other years. In general, however, women’s postseparation risk for IPV declined over time.

Smoothed baseline and model-adjusted estimates (6-month prevalence estimates) of separated women’s risk for intimate partner violence, NCVS 1995-2010.
Discussion
The robustness of previous research suggesting there is an increased risk for IPV during separation is hard to discount. The goal of this analysis was to extend prior research by teasing out the independent effect that separation has on risk for IPV from 1995 to 2010 while ensuring that marital status at the time of the victimization is accurately measured using a nationally representative sample of women. While the results presented describe the 6-month risk for IPV for all women by their marital status, the focus of this discussion is on the risk for IPV that separated women face, independent of other factors, and relative to nonseparated women.
These findings show that separated women were in fact more likely than women of all other marital status groups to be victims of IPV during many years from 1995 to 2010, even after controlling for the effects of characteristics that may be associated with their likelihood of victimization. This finding suggests that the status of being separated was associated with greater risk for IPV. These results held for most of the years from 1995 to 2010. The only control variable consistently associated with female risk for IPV from 1995 to 2010 was age, which was negatively associated with risk, net of all other factors. This finding is consistent with previous research that found younger women had higher risk of IPV (e.g., Bachman & Saltzman, 1995; Siddique, 2016). While age was a significant predictor of female risk for IPV independent of other factors in the model, its effect was additive and did not account for separated women’s higher risk for IPV. In fact, none of the control variables mediated the effect of separation on risk for IPV in any given year, or across the years.
One reason for this result may be that important covariates of risk for IPV were excluded from the models. For example, some scholars have argued that offender characteristics, such as men’s demographic features, mental health status, attitudes, and beliefs (e.g., related to their acceptance of gender equality) may have greater efficacy in assessing IPV risk than victim characteristics (Fleury et al., 2000; Hotaling & Sugarman, 1986). In addition, victim characteristics such as psychological difficulties and childhood trauma may also be important risk factors for IPV (for more detailed discussions of these and other risk factors for IPV, see Capaldi, Knoble, Shortt, & Kim, 2012 and Kuijpers, van der Knaap, & Lodewijks, 2011). Unfortunately, the NCVS does not provide the measures necessary to examine the role that these other important characteristics have on women’s risk for IPV.
By studying IPV across 16 separate years, these results showed that separated women’s risk for IPV varied from 1995 to 2010. However, some of the year-over-year differences may not be statistically significant. Their risk increased some from 1995 to 1998 and then generally declined from 1998 to 2010, although there appears to be some increase from 2003 to 2005. Overall, however, their prevalence rate in 2010 was somewhat lower than in 1995. Because the changes that occurred in separated women’s characteristics over these 16 years were mostly in a direction that makes them less prone to IPV, the reason why their risk for IPV increased from 1995 to 1998 and 2003 to 2005 is unclear. For example, from 1995 to 2010, separated women became older, more likely to have a high school degree, and less likely to be living in cities, living at or below the poverty line, and living with a child, all characteristics that may make a women less vulnerable to victimization. In addition, many scholars have argued that risk for IPV has decreased due to increases in both exposure-reducing domestic violence resources and gender equality over the past two decades (Dugan et al., 2003; Xie et al., 2012), yet while separated women’s 6-month risk did decrease slightly from 1995 to 2010, declines were not consistent over the entire period.
Although the NCVS is arguably the best nationally representative survey available to study trends in risk for IPV for separated women, and the analytical approach used in this study accounted for both the temporal ordering of marital status and victimization and various measures that previous research suggested may have altered their risk for IPV, some limitations exist that suggest caution should be taken when interpreting the level and trends presented in this analysis.
As it stands, NCVS interviews do not capture whether a separated woman’s spouse or boyfriend/girlfriend is the intimate partner that victimized her during the previous 6 months, which is assumed in most previous research. For example, it could be that a woman is legally or nonlegally separated from her spouse due to marital discord, but dating a partner who assaulted her during the prior 6-month period. Because current and former spouses and boyfriends/girlfriends are all considered intimate partners, it is difficult to know who committed the IPV incident because the victimization is rarely coupled with the specific offender. Moreover, the definition of IPV used in this study is based solely on an aggregation of nonlethal violent crimes (i.e., rape, sexual assault, robbery, aggravated assault, and simple assault) available in the NCVS. These limitations make the findings of this study less comparable with studies that identified the intimate partner who committed the IPV, and those that included other forms of IPV (e.g., homicide) or a wider range of abusive behaviors by one’s intimate partner (e.g., emotional or psychological abuse).
There is also the potential that the methodological design of the NCVS lends itself to underestimate rates of IPV due to its sampling frame and potential survey nonresponse. While this study measures the risk of IPV postseparation for separated women living in households in the United States, unfortunately, there are women not included in NCVS (e.g., the homeless or women in institutional settings such as jails) that may have an increased risk for IPV. Prior research has even suggested that homeless women and women living in shelters are at greater risk for IPV than other women, yet the NCVS does not sample these populations and, therefore, likely underestimates rates of IPV for women (Rennison & Welchans, 2000). In addition, a study by Coker and Stasny (1994) found that because NCVS interviews occurred in a respondent’s home where an intimate partner may have been present, some women may have withheld IPV victimizations from NCVS interviewers. 5 Therefore, the survey limitations and lack of information on nonrespondents and survey nonresponse suggests that caution should be taken when interpreting the results of this study.
Furthermore, the restriction of data to women who remained in the same marital status group across the two survey periods dropped a significant number of women from the analyses. By excluding women with different marital statuses in their “previous” or “current” survey periods or missing marital statuses in either survey period, roughly one in four women were omitted from the analyses each year. With only 2.2% of women classified as separated from their “previous” to “current” survey period, and 0.69% of separated women reporting any incidents of IPV in the 6 months prior to their interview, the unweighted number of separated women in any given time period who were victims of IPV was relatively small. This may have been the reason separated women’s risk when compared with divorced or never married women’s risk failed to be statistically different from one another across a few years of this study. It is important though that this finding not distract from the bigger pattern in the majority of the years.
Conclusion
The results of this study suggest that separated women’s postseparation risk for IPV from 1995 to 2010 is not only higher than women in other marital status groups, but it is the status of being separated that is associated with their increased risk, not the fact that separated women tend to possess different individual-level characteristics than women of other marital status groups. This study is a contribution to the literature on separation and IPV because it is the first of its kind to control for the temporal ordering of separation status and IPV using the NCVS and present temporal trends in separated women’s risk for IPV during a more recent time period. However, an ideal methodology to address questions related to the temporality of marital discord, separation, and IPV that have been left unanswered would be longitudinal in nature, following women over time to see whether changes in marital status are associated with changes in rates of IPV. Future research should attempt to garner this kind of information, precisely detailing at what point, and for how long, separated women are at the greatest risk for postseparation IPV, and whether it is the same offender responsible for each victimization.
Specific recommendations for future research and practice can be gleaned from this study related to the delivery of social service programs meant to assist women most at risk for IPV. Because separated women were found to have a higher 6-month risk for IPV than women of all other marital status groups independent of several covariates of risk in most years from 1995 to 2010, this study highlights how essential it is that domestic violence prevention resources such as crisis hotlines, shelters, and legal services are made available for separated women. It is noteworthy that separated women’s risk for postseparation violence has in fact decreased from a high in 1998, suggesting that some of these preventive efforts may have been successful at protecting separated women from their abusers. Furthermore, these findings suggest that separated women are at a higher risk solely because of their state of being separated from their intimate partner. Therefore, if service practitioners (e.g., social, legal, and health care workers) were more aware of the heightened risk for IPV of separated women by virtue of their separation status, efforts could be made to quickly screen for IPV when confronted with a separated woman as part of their daily interactions with the public. Because some women experience IPV for the first time during separation, while others find that separation only exacerbates the violence they were experiencing prior to separation, the results of this study indicate that further research is needed to better understand how to protect a woman from IPV during these critical periods in her life. Furthermore, other groups besides separated women may have an increased risk for IPV, but their numbers in the U.S. population do not provide enough statistical power to be studied using the NCVS (e.g., disabled individuals, military veterans and their spouses, immigrants). Future research should also attempt to understand how separation plays a part in the lives of these individuals as well.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
