Cohabitation Is No Longer Associated With Elevated Spousal Homicide Rates in the United States

Introduction

Margo Wilson, with her partner Martin Daly, pioneered the epidemiological analysis of relationship-specific homicide rates. One of their early findings (Daly & Wilson, 1988) was a huge difference between registered marriages and cohabiting unions (also called “de facto” or “common law” marriages). Combining data from 1974 through 1990, Wilson, Daly, and Wright (1993) reported that the rate of spousal homicide victimization for women in common law marriages in Canada was 8.4 times higher than that for women in registered marriages, while the corresponding ratio for men was 15.0.

A similar pattern has since been found elsewhere. In the United States, Shackelford (2001a, 2001b) estimated that cohabiting women were slain at a rate 8.9 times greater than married women in 1976-1994, while the corresponding ratio for men was 10.4. For Australia in the 1990s, these ratios have been estimated as 9.6 for women (Shackelford & Mouzos, 2005) and 16.2 for men (Mouzos & Shackelford, 2004). These nations differ greatly in overall homicide rates, but the striking contrast between married and cohabiting couples was remarkably consistent. Surprisingly, however, it has not persisted. We report here that in the United States, the elevation in homicide risk associated with cohabitation had vanished by 2005.

We first describe our data sources and the trends in spousal homicide in the United States between 1990 and 2005, showing not only that the rates have been decreasing, as is already well known (e.g., Dugan, Nagin, & Rosenfeld, 1999; Fox & Zawitz, 2007; Puzone, Saltzman, Kresnow, Thompson, & Mercy, 2000), but also that rates in married and cohabiting couples have been converging. We then attempt to determine whether the difference has evaporated because the two sorts of unions have become more alike in other ways. By and large, it appears that they have not, and no satisfactory explanation for the homicide rate convergence emerges as a result of this exercise. Finally, we review proposed explanations for the homicide rate difference that formerly prevailed, and suggest future research directions that may shed light on the convergence.

Homicide Rate Estimation

The numerators for the rates in this study of U.S. spousal homicides are derived from the Supplementary Homicide Reports (SHR) file (Fox & Swatt, 2009), as accessed via the Interuniversity Consortium for Political & Social Research website (www.icpsr.umich.edu). Analyses are restricted to cases of murder and nonnegligent manslaughter in 1990-2005, in which the SHR codes indicate that there was a lone perpetrator and the victim was the killer’s “wife,” “husband,” “common law wife,” or “common law husband.” These 13,619 cases were further reduced by excluding incidents in which the victim’s or offender’s age was either missing or coded as less than 15 years, leaving 13,083 cases. Victims were 8,820 wives, 2,637 husbands, 1,008 common law wives, and 618 common law husbands. The SHR file is estimated to have captured 94% of homicides known to U.S. police forces (Federal Bureau of Investigation, 2009), but for present purposes we make no adjustments to the above counts.

Our denominators are estimates of the numbers of persons cohabiting with opposite-sex partners or living in registered marriages in the U.S. population-at-large, which we derived from the Decennial Census of 1990 and 2000 (U.S. Census Bureau, 1992, 2005) and the American Community Survey (ACS: Ruggles et al., 2010) for the years 2001-2005. For 1991-1999, we estimated population numbers by linear interpolation between the two censuses (see James, 2011, for further details and justification). The ACS provides annual estimates beginning in 2000. The two sources yield very similar estimates for 2000, the only year for which both are available, differing by 1.6% or less in the demographic groups of interest.

U.S. Spousal Homicide Trends 1990-2005

Figure 1 depicts annual rates of spousal homicide victimization in the United States, with the sexes and married versus cohabiting couples distinguished. Victimization rates declined substantially in all groups, but especially in cohabiting couples, such that by 2005 cohabitation no longer entailed a higher homicide risk than registered marriage. Figure 2 highlights this change in relative rates of victimization.

OPEN IN VIEWER

Figure 1.

Spousal homicide rates by sex and cohabiting versus registered marriage status, United States, 1990-2005

OPEN IN VIEWER

Figure 2.

Sex-specific ratios of spousal homicide victimization rates in cohabiting unions over registered marriages, United States, 1990-2005

Daly and Wilson (1988) noted that Canadian homicide victimization in married couples was maximal in the youngest age groups and declined steeply with age but that common law couples exhibited a very different pattern, with rates peaking in middle age. These distinct age patterns have also been found to prevail in Britain (Wilson & Daly, 2001), Australia (Wilson & Daly, 2001), and the United States (Shackelford, 2001a, 2001b). To assess whether (a) the dramatic decline in cohabiting couples’ homicide rates is peculiar to particular age groups, and (b) married and cohabiting couples have converged in the age pattern of homicide risk as well as in gross rates, we compared age-specific rates at the beginning and end of the analysis period. More specifically, we used population estimates for 1990 and 2005, and homicide rates that were collapsed over 1989-1991 and 2004-2006, to reduce noise caused by small numbers of homicides in some age groups. The results are presented in Figure 3, which shows (a) that homicide rates declined between the two periods in every age group, in both registered and cohabiting unions, (b) that the very different age patterns characteristic of registered versus cohabiting unions have nevertheless persisted, and (c) that a greater diminution of homicides in cohabiting than in married couples can be seen across a wide range of ages, although the data for cohabiting women aged 45 to 54 and wives above 64 are a little anomalous in this regard. Clearly, married and cohabiting couples have not converged with respect to the age pattern of risk: in both time periods, homicide victimization was maximal at the youngest ages in married couples and at middle age in cohabiting couples.

OPEN IN VIEWER

Figure 3.

Spousal homicide rates by victim–perpetrator relationship and age class of victim, comparing 1989-1991 and 2004-2006

Have Cohabiting and Marital Relationships Become Generally Similar?

The convergence of cohabiting and marital homicide rates might be the result of a general tendency for the two sorts of unions to have become less distinct in their attributes. Is this the case? Many differences between married and cohabiting couples have been reported. Cohabiting couples tend to be younger than married couples (Lichter, Turner, & Sassler, 2010), to have lower incomes (Hardie & Lucas, 2010; Prokos & Keene, 2010), to have lower education levels (Bumpass & Sweet, 1989), and to have higher rates of unemployment (Borooah, 2002; Rindfuss & Vandenheuvel, 1990). These are all plausible risk factors for spousal homicide (Wilson & Daly, 1998, 2001), so we assessed whether they, too, had converged from 1990 through 2005.

The age distributions of married and cohabiting couples in the United States have not converged: both age distributions have shifted somewhat toward older ages, but they remained as distinct in 2005 as they had been in 1990 (Figure 4). Distributions of income, unemployment, and education also fail to exhibit convergence. In 1990, the median income of married couples exceeded that of cohabiting couples by US$9,106; by 2005, this difference had grown to US$13,495, and the income advantage of married couples had increased in all age groups except those aged 55 and above. As regards employment, cohabiting and married persons indeed differ, but how they differ changed little between 1990 and 2005 (Figure 5). Finally, the trend for education is like that for income: the married advantage actually increased. In 1990, 21% of married persons held a bachelor’s degree or higher, compared to 15% of cohabitants; by 2005, this difference had grown to 30% versus 20%. Figure 6 shows that the married made greater gains in every age category except among seniors.

OPEN IN VIEWER

Figure 4.

Age distributions of married and cohabiting persons in the United States in 1990 and 2005

OPEN IN VIEWER

Figure 5.

Employment status, by union type, United States, 1990 and 2005

OPEN IN VIEWER

Figure 6.

Percentage of married and cohabiting populations holding a bachelor’s degree or higher, United States, 1990 and 2005.

In sum, none of the four most obvious demographic attributes that distinguish cohabiting and married persons converged between 1990 and 2005 in a way that might help explain why the dramatic difference in homicide rates disappeared.

Discussion

Figures 1 and 2 demonstrate a striking and fairly rapid change: the spousal homicide rates of cohabiting couples were severalfold higher than those of married couples in the early 1990s but then converged to virtual identity in little more than a decade.

Estimation of the numbers of cohabiting couples in the U.S. population at large is potentially problematic. Puzone et al. (2000) and Shackelford (2001a, 2001b) derived their estimates from the Current Population Survey, but Baughman, Dickert-Conlin, and Houser (2002) note that this source provides substantially smaller estimates of cohabitant numbers than either the Decennial Census or the American Community Survey and suggest that the latter sources are probably more accurate. If so, then the excess risk associated with cohabitation will have been overestimated in prior research; indeed, our estimates of the cohabitation/marriage ratio of homicide rates in the early 1990s (Figure 2), while substantial, are nevertheless much lower than Shackelford’s (2001a, 2001b) estimates of 8.9 for women and 10.4 for men in 1976-1994. Of course, our lower estimates may also be due in part to a downward trend in these ratios that began before 1990. But in any case, these complications do not gainsay the reality of our central finding since both the homicide counts and the population estimates used to generate the data portrayed here were derived from the same sources. There can be no doubt that cohabiting unions in the United States entailed substantially greater risks of intimate partner homicide than registered marriages as recently as the 1990s and that this was no longer the case by 2005.

The fact that registered and common law marriages have become much more similar in homicide risk between 1990 and 2005 suggests that they must somehow have become more similar in other relevant attributes. Nevertheless, the obvious demographic distinctions have not disappeared, nor even diminished: differences between the two types of unions with respect to age distributions and unemployment rates were essentially unchanged over the relevant time period, and differences in education and income actually grew larger. However, some subtler differences between the two sorts of unions have been documented in prior research, and remain to be explored with respect to their possible relevance to the trends portrayed in Figures 1 and 2.

Wilson and Daly (2001) proposed three major reasons why conflict and violence might be more extreme among cohabiting couples than in married couples: (a) differential commitment, “exit costs,” and stability, (b) differences in sexual infidelity, and (c) differences in the numbers of coresiding children from previous unions. Whether cohabiting and married couples have become more similar in these domains is unknown.

Wilson and Daly (2001) reviewed a diversity of evidence that cohabiting unions are more likely to dissolve than registered marriages, even when correlated factors such as income, previous duration, and the presence of children are controlled. If men are inspired to use coercive violence against partners who are perceived, for whatever reason, as inclined to leave, then the lesser stability of cohabiting unions could explain their enhanced violence. Wilson and Daly (1992a, 1993, 1996) argued forcefully that “male sexual proprietariness” is a motivating factor in a large majority of spousal homicides, with the triggering events being either the woman’s threatening to leave the relationship or the man’s suspicions of infidelity, and it is noteworthy that surveys generally find that sexual infidelity is much more prevalent among cohabiting women than among registered marriage wives (e.g., Forste & Tanfer, 1996; Treas & Giesen, 2000). We have not found evidence bearing on the possibility that these contrasts between cohabiting and marital unions may have diminished between 1990 and 2005, but this is a possibility worth exploring.

Finally, the differential presence of stepchildren has been a strong candidate explanation for the differential homicide rates in marriage versus cohabitation. First, there is abundant evidence that stepchildren, unlike children of the current union, exacerbate couple conflict (Daly & Wilson, 1996), and their presence is associated with a substantial elevation of homicide risk (Brewer & Paulsen, 1999; Campbell et al., 2003; Daly, Wiseman, & Wilson, 1997). Second, this is one of the largest differences between marriage and cohabitation: in both the United States and Canada, almost half of all cohabiting unions with children have been found to include stepchildren, compared to approximately 6% of married unions (Brown, 2003; Statistics Canada, 2004; Stewart, 2001). Again, we have found no evidence bearing on possible trends in the magnitude of this difference, but this, too, warrants further exploration.

The elevated homicide risk associated with cohabitation has been so large and so consistent in previous research that the contrast is often stated in the present tense, as a sort of timeless truth. Clearly, that is not the case, and further research on temporal trends is needed. While it appears that there has been negligible change in the union profile of cohabiting couples, limitations of the survey data that we have examined mean that our picture was painted with a very restricted palette. More research is needed to determine whether factors such as financial arrangements, relationship satisfaction, fidelity, commitment, and the prevalence of stepchildren have converged before one can confidently judge whether cohabiting unions in the United States are coming to resemble registered marriages.

In homicide generally, and in spousal homicide in particular, the United States is sometimes an outlier among developed nations (e.g., Wilson & Daly, 1992b). Thus it is important to ask whether the phenomena that we report here are peculiar to the United States or more widespread. Preliminary analyses indicate that excess risk in cohabiting unions, as compared to registered marriages, has been shrinking in Canada, too (James, 2011). Whether our findings will be echoed in other countries warrants investigation.