Trends in Intimate Partner Violence

Examining Trends in Lethal and Nonlethal Violence

Given the availability of homicide data from the Supplementary Homicide Reports, research on long-term trends in lethal violence has dominated academic discussions and empirical work. Less often have scholars examined the trends in nonlethal violence against women or explored the factors associated with women’s risk for less than lethal violence. Academic work examining the impact of employment and race in nonlethal intimate partner violence is also limited and the research that has been done is cross-sectional. This study seeks to begin to close both of those gaps by exploring how the trends in intimate partner violence may differ as a function of race and employment. Pooled data from the National Crime Survey (NCS) and the National Crime Victimization Survey (NCVS) for 1980 to 2008 offer the unique opportunity to examine previously unexplored trends in partner violence while also identifying the contributions of race and employment on women’s risk for intimate partner violence.

Lauritsen and Heimer have used data from the National Crime Victimization Survey (1980-2005) to estimate previously unknown long-term trends in violent victimization disaggregated by gender and the victim–offender relationship. Their work first suggests that throughout the time period victimizations at the hands of intimate partners occur at rates lower than either stranger or known/nonintimate offenders. With respect to the trend in female victimization by an intimate partner, their analyses demonstrate an increase between 1980 and 1987, then stability through 1994, at which point rates began to decline and be comparable to rates of female homicide victimization. Disaggregation of these rates by race and ethnicity demonstrate little evidence of a clear gap in levels of violence across White, Black, and Latina women. However, they suggest that these nondiscernable differences in the trends are perhaps due to sample size restrictions and “thus ‘noise’ in the trends for Latinas and non-Latina blacks” (Lauritsen & Heimer, 2009, p. 49). In the analyses presented as follows, we examine trends between White and Non-White women to maximize the sample of minority women. This is particularly important in the later years of the data after 1992 where we have seen declines in intimate partner violence across all racial and ethnic groups.

There are both theoretical and empirical reasons to suspect that there are important and notable trends in intimate partner violence over the past 28 years and that both race and employment shape the risk of intimate partner violence by both employment and race. However, research has yet to specifically explore these long-term trends. Farmer and Tiefenthaler (2003) suggest that there are likely three sets of factors correlated with the decline in rates of male to female intimate partner violence. First, and consistent with work by Dugan et al. (1999) and Browne and Williams (1993) the increased provision of services, in particular legal resources, for victims of intimate partner abuse are likely to have a role in the decline during the 1990s. Second, and consistent with other work on shifting demographic trends, Rosenfeld (1997) argues that those groups at the greatest risk for spousal homicide are the same groups in which marriage rates are declining more rapidly, creating a demographic shift of marriage that impacts risk. This finding, combined with declining rates of marriage among African American women as compared to White women (U.S. Bureau of the Census, 2007), suggests the importance of looking at trends by race. Finally, Farmer and Tiefenthaler suggest that changes in the economic status of women and their educational and employment attainment are likely to offer women the opportunities to end and leave violent and abusive relationships.

Race, Employment, and Intimate Partner Violence

Women’s economic contributions are increasingly important resources for the family’s financial well-being and these are likely to have implications for relationship quality and men’s and women’s satisfaction with marriage (Amato, Johnson, Booth, & Rogers, 2003; Smith, 2008; White & Rogers, 2000). Work within the Family Sociology literature, including research by Nock (2001), points out that approximately one fifth of all marriages and almost a third of dual-income households may be conceptualized as “marriages of equally dependent spouses.” In these relationships, neither the male nor female partner makes a disproportionate contribution to the family’s economic well-being, with each partner contributing approximately 40% to 59% of the total family earnings. He and other scholars (Kaukinen, 2004) suggest that recent changes in the economic roles of women and men have important implications for marital quality, satisfaction with marriage, conflict, and women’s risk for intimate partner violence victimization.

Work by Smith (2008) has highlighted these changes over the past 40 years, particularly difference in employment by race. Data from the Bureau of Labor Statistics demonstrate that since 1970, women’s share of the labor market has risen from 35% to 45% (women’s overall labor force participation during that period rose from 50% to more than 75% of women being involved in the paid labor market). Researchers have also examined specific trends in African American women’s involvement in the labor market. Offner and Holzer (2002) have shown that the employment rate among African American women lagged behind that of White women during the 1980s. They then note that in the 1990s African American women experienced greater gains in employment as compared to white women—from a 37% employment rate during the early 1990s to a 52% employment rate in 2000. In addition, work by Lerman (1997) notes that the percent change in real wages for African American women has grown 20% and that the wage differential between white and black women declined from 8.7% to 6.0% between 1984 and 1995. More recently, Smith has shown that after 2001, women’s employment rates began to decline. She notes that these employment declines were particularly large among women with low education levels, young women, African American women, single women, and single mothers. More important, Smith notes is the growing importance of women’s employment for the economic health of families. Whereas in 1970 less than 17% of women would be conceptualized as being in a dual-income or female-headed household, by 1980 this had risen to almost 25% of couples, and by the late 2000s close to half of all marriages, women make substantial contributions to the family’s economic well-being. At the same time there has been a substantial drop in the role of men as sole providers and modest drops in men’s contributions as the primary breadwinner.

Fox, Benson, DeMaris, and Van Wyk (2008) have highlighted the two sides of women’s involvement in the paid labor market on their risk for intimate partner violence. They suggest there is reason to expect that women who are employed outside the home are at less risk of intimate partner violence than their unemployed counterparts. They point out that, in addition to increasing economic opportunities, employment shapes women’s daily routines outside the home thereby increasing interaction and connections outside the home and reducing isolation. In contrast, they also hypothesize that employment may increase the risk for partner violence when her male partner is unemployed, reflecting both economic stress in the household and the potential use of violence by under- and un-employed partners to gain control.

The idea of female employment having implications for the power dynamics in the relationship and thus impacting women’s risk of violence has been the focus of feminist theories examining this relationship. In examining the factors associated with female homicide victimization, Vieraitis, Kovandzic and Britto (2008) have outlined the feminist perspectives’ operationalization of the factors associated with women’s risk for violence by their male partners. In particular, they point to the importance of women’s status relative to that of men with respect to a number of economic, legal, educational, and occupational factors. They highlight the way in which status plays a role in risk for violence. They point out that Marxist feminists highlight absolute measures of economic, legal, educational, and occupation status as predictive of violence against women, whereas Liberal feminists draw on gender inequality in understanding the context in which women are at risk. They contrast these two positions with radical feminism in which greater equality is argued to lead to a backlash effect in which violence against women will be greatest after initial gains by women with respect to economic status.

Gartner and McCarthy (1995) have suggested the importance of an historical examination that recognizes that the relationship between social statuses, such as women’s employment, and women’s risk for violent victimization may have changed over time. They suggest that employment for women may heighten the risk for intimate partner violence when female employment is less normative, less common, or during the initial movement of women into the labor force. Over time, employment may serve as a mechanism by which women will be able to protect themselves from the risk for violence while also offering avenues for ending violent relationships. Taken together, these perspectives suggest that both race and employment may place women at a differential risk of victimization. Research has yet to explore race and employment as two potential sources for disparity in the trends of intimate partner violence. Our research examines how employment modifies the overall trends in women’s risk of intimate partner violence. Furthermore, we examine the way in which race conditions the trends of employment on the risk for violence and whether this relationship has changed over the past 30 years.

Method

Analysis

Before discussing how estimates of victimization were calculated, two methodological considerations should be discussed. In 1992, the NCS/NCVS underwent a redesign, which, among other goals, was aimed at facilitating more accurate recall of victimization. The effects of the redesign were successful in that the redesigned survey reports of victimization were higher than the reports of victimization prior to the redesign (Kindermann, Lynch, & Cantor, 1997; Rand, Lynch, & Cantor, 1997). Fortunately, the design was phased in over an 18-month period, which allows for reporting differences between the two designs to be assessed. These differences can then be used to factor upward the pre-redesign estimates to make them comparable to the redesign estimates (Rand et al., 1997). Although reasonable, it is important to note that this strategy hinges on the assumption that any differences between the pre-redesign and redesign estimates are solely a function of the survey redesign and that these differences are uniform for the entire length of the survey prior to the redesign.

However, the increase in reporting was not uniform for all crime types. Therefore, crime-specific redesign adjustments are applied to the pre-redesign estimates of victimization rates (Lynch, 2002; Rand et al., 1997). Each “redesign weight” is simply the ratio of the weighted redesign victimization rate to the pre-redesign victimization rate using the entire 18-month phase-in period. Crime-specific redesign weights are employed instead of subgroup- and crime-specific redesign weights because previous research assessing the impact of the redesign has suggested that subgroup-specific redesign weights are unnecessary (Cantor & Lynch, 2005). Therefore, all estimates prior to 1992 are weighted upward as follows: sexual assault by 2.57, robbery by 1.00, aggravated assault by 1.23, and simple assault by 1.75.

In 2006, the NCS/NCVS survey underwent a second redesign that modified the sampling frame based on the decennial census, and which resulted in a drastic increase in victim reporting, especially in rural areas (Rand & Catalano, 2007). However, the increase was the result of the change in sampling methodology and not the result of an actual increase in victimization (or a more accurate recall of victimization), and as a consequence, the estimates for 2006 are not comparable to prior years. Unfortunately, this redesign was not phased in; therefore, the differences in reporting from the second survey redesign cannot be obtained to properly adjust the annual victimization rates. In 2007, the new redesign was terminated, making the estimates from 2007 onward comparable to estimates obtained prior to 2006. Consequently, this study omits the year 2006 from the analysis.

Aside from the differences noted, our rate estimation procedure mimics the methodology used by the Bureau of Justice Statistics (for a tutorial on post-redesign rate estimation see, NACJD). The numerator was obtained by weighting the annual number of reported victimizations that comprise intimate partner violence in the sample by the sampling weight included in the data set, followed by the “redesign weight.” The denominator was obtained by weighting the demographic subgroup sample size by the same weight included in the numerator. However, because the survey is implemented every 6 months, the weight in the denominator of the rate formula is halved to obtain annual estimates of the population at risk.

The equation for domestic victimization rate is as follows:

I P V k t = ∑ j [ D W j × ∑ ( W i k t × N V I C S i j k t ) ] N U M k t × 1 2 ∑ W i k t

where IPV_kt is the victimization rate of subgroup k in calendar year t, NVics_ijkt is the number of crime-specific victimization incidents for the subgroup summing across sampled individuals, W_ikt is the person weight, DW_j is the survey redesign weight that corresponds to the crime-specific weights discussed above prior to 1992 and a value of 1 from 1992 onward, and Num_kt is the number of respondents for the corresponding subgroup interviewed in that year.

The Bureau of Justice Statistics considers victimization rate estimates based on victimization counts more than 10 to be stable. Using this criterion, a few estimates fell below this cutoff for unemployed non-White women. Therefore, to increase the stability of the results and smooth out the trends, the victimization trends shown in the results reflect 3-year moving averages, with the first and last year of the series averaged with the next adjacent year. Because the 2006 estimates are invalid, to obtain the average for 2005 and 2007, the mean of those 2 years was applied as the upper and lower value for the estimate of the moving average, respectively.

Results

This section proceeds as follows. First a discussion of the trends in intimate partner violence disaggregated by race will be discussed. Second, the results for the trends in victimization and employment status will be presented. Finally, the intersection between race and employment as it relates to domestic violence victimization will be examined.

With regard to race, there are notable differences in the risk of intimate partner violence across White and non-White women (Figure 1). White women’s risk for partner violence increased steadily between 1980 and 1992 from a rate of 7 to 9 per 1,000 women. After 1992 the rate for White women declined until 2004 where an apparent reversal in the risk occurred and a modest increase is evident between 2004 and 2008. With respect to the trend for non-White women there is some up-and-down fluctuations ranging from 8 to 9.5 per 1,000 non-White women between 1980 and 2001 after which the rate of victimization appears relatively flat with approximately 4 per 1,000 women at risk among non-White women. Overall, the race differences identified in Figure 1 point to non-White women being at a slightly increased risk of intimate partner assault, particularly before the late 1990s. With the exception of a brief 4-year period in the early 1990s, between 1980 and 1998, non-White women were at greater risk of partner violence as compared to White women. Beginning around 1997, the trends for Whites and non-Whites began to converge, making them virtually indistinguishable until the mid-2000s. This convergence appears to be due in large part to a more dramatic decrease in the rate of non-White intimate partner victimization. In fact, the rate of non-White victimization decreased by more than half between 1996 and 2001 (10.54 to 4.68).

OPEN IN VIEWER

Figure 1.

Intimate partner victimization trends disaggregated by race

As shown in Figure 2, there are notable differences in the risk of intimate partner violence across employed and unemployed women. Between 1980 and 1997, with the exception of a few increases, the risk for employed women fluctuated around 10 per 1,000 women. After 1997, the risk for employed women declined until 2005 to a low of 4 per 1,000 women after which we have seen a modest reversal and increase in the risk of partner violence for employed women. Among unemployed women, the risk increased between 1980 and 1994 from 5 to 7 per 1,000 women until a decline began in 1994. This decline continued until 2001 for unemployed women in which a reversal occurred with a modest increase in the risk of partner violence until 2008. With regard to the differences across employed and unemployed women, employed women are consistently at a higher risk for intimate partner victimization compared to unemployed women until the mid-2000s. Although the level of victimization is higher for employed women, temporally, both groups follow relatively the same pattern. The rate increased until the mid-1990s before drastically decreasing for the next decade. As with the trends in race, both groups appear to converge towards the end of the series, becoming closest in 2007 when both groups begin increasing together.

OPEN IN VIEWER

Figure 2.

Intimate partner victimization trends disaggregated by employment

The conclusion that employed women are at a higher risk of victimization appears to be partly contingent on the race of the victim (Figure 3). As shown, employed non-White and White women are comparably at risk for intimate partner victimization. In fact, their trends are virtually indistinguishable for the entire duration of the series. For both White and non-White employed women their risk throughout the series fluctuated between 1980 until the decline in the late 1990s. For employed White women the unaveraged rate ranged from 9.85 per 1,000 women in 1980 to a high of 12.35 per 1,000 women in 1992, then declining until 2003 to a low of 3.31 per 1,000 women after which the rate remained relatively constant. For employed non-White women the rate in 1980 was 10.46 per 1,000 women to a high of 15.33 per 1,000 women in 1989 after which the rate peaked again in 1992 (14.08) and then declined and flattened out after 2004. However, unemployed women are not equally at risk with regard to race. Although there is substantial variation in the trend for non-White unemployed women (due to small samples), this trend appears to be relatively comparable to employed women, at least for the first 15 years of the series. Notable, is the trend for unemployed White women who throughout the series had the lowest risk for intimate partner violence. The risk for unemployed White women increased between 1980 and 1990 after which it declined until 2001 where a modest reversal in the rate and increase is evident through 2008. Unemployed White women’s risk throughout the series began at a rate of 4.80 per 1,000 women in 1980 and slowly increased to a high of 6.91 per 1,000 women in 1991 to a low of 2.50 per 1,000 women in 2005. Temporally, for all women the rates of victimization began decreasing in the mid-1990s with the most drastic decreases in the groups that historically were at the highest at risk for violence.

OPEN IN VIEWER

Figure 3.

Intimate partner victimization disaggregated by race and employment

Discussion

This article has contributed to research on intimate partner violence by highlighting the importance of looking at long-term trends in nonlethal intimate partner violence that disaggregate victimization data to examine theoretically relevant variables that are likely correlated with the risk for violence by an intimate partner, including race and employment status. The findings point to the need for scholars to continue to engage in research that examines long-term trends in violence against women and conduct empirical analyses of the correlates and etiologies of violence by intimate partners. The findings have important implications for identifying those women most at risk, designing interventions that are meaningful, effective, and address the needs of high risk women, and ultimately prevention efforts that reduce violence against women. Our research points to four key findings for which we can draw important conclusions regarding trends in intimate partner violence.

First, and consistent with other research on long-term trends in women’s nonlethal violent victimization (see Lauritsen & Heimer, 2009), we find that particularly after the mid- to late-1990s there has been a substantial decline in the risk for intimate partner violence victimization among women. The importance of women’s economic contributions and greater equality within intimate relationships is likely to have shaped working women’s reduced risk for intimate partner violence after the mid-1990s. Our research is consistent with previous work of other scholars who point to the importance of domestic violence services, resources, and legal remedies. These have shaped women’s risk for violence and their help-seeking decisions over time, and ultimately, the decline in both nonlethal and lethal intimate partner violence (Dugan et al., 1999). Although we have seen a substantial and noteworthy drop in the risk for intimate partner violence, it remains crucial that researchers, service providers, and legislators continue to work to eradicate violence against women.

Second, our research highlights the disparity minority women continue to face with respect to the risk of life-threatening violence. Although there has been a decline in the risk for both White and non-White women in the risk of intimate partner violence, for much of the period between 1980 and 1998, non-White women were at greater risk of intimate partner violence. More important, after a convergence in the risk for White and non-White women in the late 1990s and early 2000s, the risk for non-White women has remained higher than for White women. Our findings contrast with work on lethal intimate partner violence including work by Fox and Zawitz (2010) who point to trends in the murder of women in which the risk is substantially higher for White women. The trends for lethal violence are not indicative of the trends in nonlethal violence, which highlights the importance of expanding beyond homicide in the study of the long-term trends in intimate partner violence against women.

Third, we find that across much of the time series, employment is associated with a higher risk of intimate partner violence. Employment has socioeconomic and symbolic implications for women’s decision making, and conflict and satisfaction within intimate relationships are associated with women’s risk for nonlethal intimate partner violence. This is consistent with the suggestion that women’s employment has important consequences for the power dynamics within heterosexual intimate relationship and thus impacting women’s risk of violence (Anderson, 1997; Kaukinen, 2004).

Fourth, the finding that employed women are at a higher risk of intimate partner violence victimization is at least partly contingent on the race of the victim. For White women, employment clearly heightens the risk for victimization, whereas for non-White women the effects of employment are less evident. For non-White women they are at greater risk regardless of employment as compared to White women. It is likely that a larger set of economic disadvantage measures shape the risk for victimization of non-White women as well as White women. Other research suggests that, in addition to employment, there is a larger group of economic measures including income, education, and primary responsibility for children that are likely to shape dependence on male partners and potentially the risk for intimate partner violence (Hotaling & Sugarman, 1986, 1990; Straus, Gelles, & Steinmetz, 1980). Future research will need to explore long-term trends in intimate partner violence that take account of partner power differences, but these analyses are not possible with data from the National Crime Victimization Survey.

Although our research has offered a number of important contributions to the literature on trends in intimate partner violence and the role of race and employment on the risk for violence, some limitations temper our conclusions. First, it is important to note that the types of intimate partner violence tapped by the National Crime Victimization Survey may not reflect the entire spectrum of violence against women and in particular may underrepresent the most severe forms of partner violence. Johnson (1995) and Johnson and Ferraro (2000) have argued that the more severe forms of domestic violence and coercive control, which they refer to as intimate patriarchal terrorism, are likely underrepresented in community and nationally representative surveys. The sampling design of the National Crime Victimization Survey and the measurement of intimate partner violence within the survey may underestimate the frequency of victimizations by the patriarchal terrorist. We therefore concede that the effect of employment on the most severe forms of violence may be much more complex and not easily tapped in our analyses. It might be that among the small number of women victimized by a patriarchal terrorist, unemployment heightens the risk for intimate partner violence and isolates women from help-seeking resources. In addition, it may be that violent and controlling men may limit women’s access to employment.

A second limitation of our research is that, to calculate the trends in light of the redesign of the survey, we have made an assumption of a stable understanding of the nature of intimate partner violence, for the women at risk, over time. That is, we assume that women in 1980 have interpreted their male partner’s behavior in the same way as women in 2008. In addition, we have made the assumption that the weighted data prior to 1992 assume that the relationship is the same between 1980 and 1991. Again, we are cognizant of these factors and recognize that after the 1990s with the expansion of domestic violence legislation and services, women in 2008 are potentially more likely than women in the 1980s to view abusive behaviors as violent, criminal, and reflecting intimate partner violence. These differences over time in the definitions of violence and perceptions of violence by women at risk are therefore likely to underestimate the decline that occurred over time, particularly between 1980 and 1991. We therefore suggest that the decline we have documented in our analyses is a modest estimation and that it is likely to be even more substantial and noteworthy.

Our examination of the long-term trends in male-on-female intimate partner violence and identification of differential trends with respect to race and employment may offer direction for interventions to address violence against women. The complex role of economic factors in shaping the risk of intimate partner violence for women points to the need to decrease the prevalence of intimate partner violence among all women, while addressing the severity of violence among women within the lowest socioeconomic statuses who may be less able leave violent relationships. Consistent with Moore (1997) this suggests the need to implement interventions that are consistent with the life situations of the victims most at risk for life-threatening violence and least likely to have access to avenues to ending violence. This includes the expansion of shelter services, personal development curricula, psychological services, and employment and educational programs for those women at risk for violence and least able to leave. In addition, for women whose economic resources are not a barrier to leaving a violent relationship, policy initiatives need to be directed to the provision of outreach activities implemented in the workforce to target employed and higher-income women to provide these women with emotional support and other services for them and their children. Given these women are less likely to access the resources offered through shelters and community service providers, outreach and advocacy services need to find avenues to providing these women with the tangible and emotional resources needed to address violence in the home. This may also include the integration of health services with violence prevention in medical and health care settings.