14
C Demography, Volume 43-Number 1, February 2006: 127–140 127 WHY ARE COHABITING RELATIONSHIPS MORE VIOLENT THAN MARRIAGES?* CATHERINE T. KENNEY AND SARA S. MCLANAHAN In response to increases in cohabitation in the United States, researchers have recently focused on differences between cohabiting and marital unions. One consistent nding is a higher rate of do- mestic violence among cohabiting couples as compared with married couples. A prominent explana- tion for this nding is that cohabitation is governed by a different set of institutionalized controls than marriage. This article explores an alternative explanation, namely, that differences in selection out of cohabitation and marriage, including selection of the least-violent cohabiting couples into mar- riage and the most-violent married couples into divorce, lead to higher observed rates of violence among cohabiting couples in cross-sectional samples. Our results suggest that researchers should be cautious when making comparisons between married and cohabiting couples in which the dependent variable of interest is related to selection into and out of relationship status. ohabitation rates in the United States have increased dramatically during the past three decades, with cohabitation now replacing marriage as the modal form of rst coresidential union (Bumpass and Lu 2000). In response to this trend, researchers have begun to exam- ine the differences between these two types of unions. One of the more consistent (and po- tentially alarming) ndings in the emerging literature is the higher rate of intimate-partner violence and intimate-partner homicide among cohabiting couples than among married couples (Anderson 1997; Brownridge and Halli 2002; Stets 1991; Stets and Straus 1989; Wilson, Daly, and Wright 1993; Wilson, Johnson, and Daly 1995; Yllo and Straus 1981). Researchers have theorized that the higher level of violence among cohabiting couples is due to institutional differences between married and cohabiting relationships (Nock 1995). Ellis (1989), for example, argued that the presence of marital norms, a higher stake in conformity, and greater investment in the relationship contribute to lower levels of vio- lence in marriage. Similarly, Wilson, Daly, and their colleagues (Wilson and Daly 2001; Wilson et al. 1993; Wilson et al. 1995) argued that weaker norms favoring sexual exclu- sivity in cohabiting unions lead to heightened sexual jealousy, which increases the risk of violence by cohabiting men. Finally, researchers have claimed that because the social and nancial costs of ending a marriage are higher than the costs of ending a cohabiting relationship, married couples invest more time in developing strategies to deal with conict and to minimize violence (Gelles 1983; Stets and Straus 1989; Wilson and Daly 2001). In this article, we argue that although these interpretations are appealing, they are premature and that researchers need to consider how differential selection into and out of marriage and cohabitation affect the composition of couples in these unions at a point in time. If certain characteristics, such as violence or high conict, are positively (or nega- tively) associated with selection into (or out of) a particular status (here, cohabitation or marriage), then cross-sectional data will tend to overrepresent (or underrepresent) couples with this characteristic (Lundberg and Pollak 2001; Repetti 2001). If the selection process is different for married and cohabiting couples—which is likely, given that cohabiting *Catherine T. Kenney, Department of Sociology and Program in Gender and Women’s Studies, University of Illinois at Urbana-Champaign, 326 Lincoln Hall, 702 South Wright Street, Urbana, IL 61801; E-mail: ctkenney@ uiuc.edu. Sara S. McLanahan, Department of Sociology and Woodrow Wilson School of Public and International Affairs, Princeton University. An earlier version of this paper was presented at a MacArthur Family Network and NICHD Family Research Network conference, “Conict and Cooperation in Families,” March 2000. We are grateful to Joshua Goldstein, members of the University of Illinois Sociology Department’s junior faculty reading group, the Editors, and anonymous reviewers for their comments.

Why are cohabiting relationships more violent than marriages?

Embed Size (px)

Citation preview

Page 1: Why are cohabiting relationships more violent than marriages?

Violence in Cohabiting Relationships 127

C

Demography, Volume 43-Number 1, February 2006: 127–140 127

WHY ARE COHABITING RELATIONSHIPS MORE

VIOLENT THAN MARRIAGES?*

CATHERINE T. KENNEY AND SARA S. MCLANAHAN

In response to increases in cohabitation in the United States, researchers have recently focused on differences between cohabiting and marital unions. One consistent � nding is a higher rate of do-mestic violence among cohabiting couples as compared with married couples. A prominent explana-tion for this � nding is that cohabitation is governed by a different set of institutionalized controls than marriage. This article explores an alternative explanation, namely, that differences in selection out of cohabitation and marriage, including selection of the least-violent cohabiting couples into mar-riage and the most-violent married couples into divorce, lead to higher observed rates of violence among cohabiting couples in cross-sectional samples. Our results suggest that researchers should be cautious when making comparisons between married and cohabiting couples in which the dependent variable of interest is related to selection into and out of relationship status.

ohabitation rates in the United States have increased dramatically during the past three decades, with cohabitation now replacing marriage as the modal form of � rst coresidential union (Bumpass and Lu 2000). In response to this trend, researchers have begun to exam-ine the differences between these two types of unions. One of the more consistent (and po-tentially alarming) � ndings in the emerging literature is the higher rate of intimate- partner violence and intimate-partner homicide among cohabiting couples than among ma rried couples (Anderson 1997; Brownridge and Halli 2002; Stets 1991; Stets and Straus 1989; Wilson, Daly, and Wright 1993; Wilson, Johnson, and Daly 1995; Yllo and Straus 1981).

Researchers have theorized that the higher level of violence among cohabiting couples is due to institutional differences between married and cohabiting relationships (Nock 1995). Ellis (1989), for example, argued that the presence of marital norms, a higher stake in conformity, and greater investment in the relationship contribute to lower levels of vio-lence in marriage. Similarly, Wilson, Daly, and their colleagues (Wilson and Daly 2001; Wilson et al. 1993; Wilson et al. 1995) argued that weaker norms favoring sexual exclu-sivity in cohabiting unions lead to heightened sexual jealousy, which increases the risk of violence by cohabiting men. Finally, researchers have claimed that because the social and � nancial costs of ending a marriage are higher than the costs of ending a cohabiting relationship, married couples invest more time in developing strategies to deal with con� ict and to minimize violence (Gelles 1983; Stets and Straus 1989; Wilson and Daly 2001).

In this article, we argue that although these interpretations are appealing, they are premature and that researchers need to consider how differential selection into and out of marriage and cohabitation affect the composition of couples in these unions at a point in time. If certain characteristics, such as violence or high con� ict, are positively (or nega-tively) associated with selection into (or out of) a particular status (here, cohabitation or marriage), then cross-sectional data will tend to overrepresent (or underrepresent) couples with this characteristic (Lundberg and Pollak 2001; Repetti 2001). If the selection process is different for married and cohabiting couples—which is likely, given that cohabiting

*Catherine T. Kenney, Department of Sociology and Program in Gender and Women’s Studies, University of Illinois at Urbana-Champaign, 326 Lincoln Hall, 702 South Wright Street, Urbana, IL 61801; E-mail: [email protected]. Sara S. McLanahan, Department of Sociology and Woodrow Wilson School of Public and International Affairs, Princeton University. An earlier version of this paper was presented at a MacArthur Family Network and NICHD Family Research Network conference, “Con� ict and Cooperation in Families,” March 2000. We are grateful to Joshua Goldstein, members of the University of Illinois Sociology Department’s junior faculty reading group, the Editors, and anonymous reviewers for their comments.

Page 2: Why are cohabiting relationships more violent than marriages?

128 Demography, Volume 43-Number 1, February 2006

couples have three choices (they can remain cohabiting, marry, or separate), whereas mar-ried couples have only two (they can stay together or separate)—the composition of these two populations will become increasingly different over time.

We use data from two waves of the National Survey of Families and Households (NSFH 1987–1988 and 1992–1994) to examine the extent to which differences in intimate-partner violence by marital status may be explained by differences in selection out of cohabitation and marriage. Using cross-sectional data for the NSFH1, we show that in relationships of short duration (up to one year), levels of violence are similar for cohabiting and married couples. For relationships of longer durations (up to � ve years), however, violence is more common among cohabiting couples than among married ones. Next, using longitudinal data, we show that violence is more positively associated with ending the relationship for married couples than for cohabiting couples. We also show that violence is negatively as-sociated with marriage for cohabiting couples. Taken together, these results suggest that much of the difference in violence between cohabiting and married couples observed in cross-sectional studies may be caused by selection rather than institutional factors.

BACKGROUNDAlthough previous research has not directly investigated the question of how selection processes affect cross-sectional differences in violence among married and cohabiting couples, we do have some indirect information on this topic. With respect to selection into marriage and cohabitation, many of the differentiating factors, which include age and education (Bumpass and Sweet 1989), income and employment stability (Rindfuss and Van-denHeuvel 1990), race (Raley 1996), family-structure history (Bumpass and Sweet 1989), and substance abuse (Yamaguchi and Kandel 1985), have also been shown to be associated with relationship violence (Anderson 1997; Cunradi, Caetano, and Schafer 2002; Ellison, Bartkowski, and Anderson 1999; Kantor and Jasinski 1998; Lupri, Grandin, and Brinkerhoff 1994; Rodriquez et al. 2001; Sorenson, Upchurch, and Shen 1996; Yllo and Straus 1981).

With respect to selection out of marriage (separation or divorce), several studies have documented a positive relationship between con� ict or intimate-partner violence and divorce (DeMaris 2000; Heaton 1991; Rogge and Bradbury 1999). This means that married couples with a lower risk of violence will tend to accumulate in the married population over time.

With respect to selection out of cohabitation, the picture is complicated by the fact that cohabiting couples can end their relationships either by marrying or by separating. Selection into marriage and selection into separation may have opposing effects on the composition of couples who remain in cohabitation. The “creaming off” into marriage of the “best” co-habiting relationships will tend to lower the average quality of cohabiting unions relative to marital unions in the population. Smock and Manning (1997) found that cohabiting couples were more likely to marry if the man had higher earnings, higher education, and full-time employment—all of which are negatively associated with intimate-partner violence. More directly, DeMaris (2001) found that violence by the female partner was associated with a reduced likelihood of marriage. These � ndings suggest that more-violent couples may tend to accumulate in the cohabiting population.

In contrast, if the worst cohabiting relationships tend to end quickly, this would raise the average quality of cohabiting unions. The evidence on this point is mixed. DeMaris (2001) reported that “intense” male violence increased the likelihood of separation, where-as Brines and Joyner (1999) found that higher-earning cohabitors (who presumably are less violent) were more likely than their poorer counterparts to separate. These con� icting forces may explain why DeMaris (2001) found that nonintense relationship violence did not predict separation among cohabiting couples.

The speed at which married and cohabiting couples end their relationships is also important in determining the composition of the two groups at any point in time. Prior research has consistently found that transitions from cohabitation to marriage take place

Page 3: Why are cohabiting relationships more violent than marriages?

Violence in Cohabiting Relationships 129

faster than those to separation (Bumpass and Sweet 1989; Sanchez, Manning, and Smock 1998). Thus, couples who are eventually going to separate, as well as the small propor-tion of couples who cohabit for relatively long periods, will tend to accumulate in the population of cohabiting couples and, relative to the entering population of cohabiting couples, will be overrepresented in cross-sectional samples, while those who marry will be underrepresented.

DATA AND METHODS

Data

Data for this study come from the two waves of the National Survey of Families and Households (NSFH 87–88 and NSFH 92–94), a national sample of 13,017 respondents who were initially interviewed between March 1987 and May 1988. In addition to hav-ing a nationally representative main sample of 9,643 respondents, the survey includes an oversample of certain minority groups and household types, including cohabiting couples (Sweet, Bumpass, and Vaughn 1988). For our purposes, the main advantage of the NSFH data, in addition to the oversample of cohabiting couples and the inclusion of questions on violence asked of both partners, is that these data include information on the starting dates of coresidence, which allows us to compare married and cohabiting couples in relationships of the same duration. Given the rapid changes in the prevalence of cohabitation, we expect that couples whose relationships began more than � ve years prior to NSFH1 may represent a population whose characteristics at entry into cohabitation were quite different from those of more recent cohorts. Thus, we limit our analyses to couples in coresidential relationships of up to � ve years’ duration.

Because the dependent variables for intimate-partner violence used in this study are created from both partners’ responses to NSFH 87–88, we include only cases for which complete information on violence was available from both partners, resulting in a sample of 1,232 married and 369 cohabiting couples for the analyses using NSFH 87–88. Previous research has shown that the characteristics of couples with completed NSFH spouse/ partner questionnaires differ slightly from those without completed questionnaires (McNally, Sassler, and Schoen 1997; Smock and Manning 1997); thus, our subsample may suffer from nonresponse bias. For the NSFH 87–88 sample, respondents with a completed spouse/ partner interview were less likely to be African American and female, but the correlations were modest (below .15). Moreover, having a spouse/partner interview had no correlation with whether the main respondent reported intimate violence. In addition, when we carried out our analyses using only information from the main respondent, our substantive results did not change (results available on request).

In our second set of analyses, we use information from the NSFH 92–94 main re-spondent (but not the spouse or partner) regarding relationship transitions since NSFH 87–88. Because of some attrition between the two waves of the study, nonresponse bias is also possible in this sample. Among cohabiting couples, we found no correlation between physical violence in the relationship in the � rst wave and whether the main respondent was interviewed in the second wave, but African American respondents and respondents with less than a college degree were less likely to be reinterviewed. Among married couples, in addition to the same association with race and education, couples who reported violence were less likely to be reinterviewed. However, in an analysis (available on request) that made the strong assumption that all “lost” couples married (among cohabitors) or stayed married (among married couples), which would be the situation most likely to undermine our results, our substantive � ndings did not change. Given the characteristics of those who were lost to follow-up, the separation rate of these couples is likely higher than for the cou-ples who were reinterviewed. If that was the case, our � ndings here present a conservative estimate of the effect of the extent to which less-violent couples tend to accumulate in

Page 4: Why are cohabiting relationships more violent than marriages?

130 Demography, Volume 43-Number 1, February 2006

marriage over time. For the analysis using NSFH 92–94, our sample includes 1,032 married and 303 cohabiting couples.

Analytic StrategyIn this study, we assess the importance of selection into and out of relationship status for making comparisons of intimate-partner violence between married and cohabiting couples. First, using cross-sectional data from NSFH 87–88, we compare married and cohabiting couples in relationships of less than one year’s duration (an approximation of the “enter-ing” population1) and in relationships of less than � ve years’ duration. We are particularly interested in the entering populations of married and cohabiting couples because they have presumably not yet been subject to the effects of selection out of the relationship status or of institutional in� uences. We suggest that to the extent that married and cohabiting couples in entering populations differ, these differences are mainly indicative of differential selec-tion into marriage and cohabitation. If differences between married and cohabiting couples are greater for the sample of couples who coresided for up to � ve years than for the entering sample, this could imply that differential selection out of marriage and cohabitation causes couples with different kinds of characteristics to accumulate in the married and cohabiting populations, that couples’ behavior is changing as a result of spending time in these differ-ent institutions, or some combination of these two in� uences.

Using longitudinal data, we next examine the association between relationship violence and staying married versus separating or divorcing (for couples who were married at NSFH 87–88) and getting married, staying in cohabitation, or separating (for couples who were cohabiting at NSFH 87–88). Again, we focus on the relationship transitions of those in the entering populations of married and cohabiting couples at NSFH 87–88 because those couples have not already been subject to much selection out at the � rst observation. How-ever, because of the small sample size of couples in the entering populations, we also carry out this longitudinal analysis for couples in relationships of less than � ve years’ duration at NSFH 87–88. If the level of violence at Time 1 is lower among cohabiting couples who subsequently marry than among those who remain cohabiting, or higher among married couples who separate than among those who remain married, this would provide evidence for differential selection out of cohabitation and marriage.

Measures of ViolenceThe NSFH asked respondents and their partners several questions regarding relationship violence. Because previous research has found that it is more likely for one spouse to report violence than for both to agree it has occurred (Szinovacz 1983), and because it seems reasonable to assume that individuals are more likely to underreport violence than to overreport it, we make use of the couple data to construct indicators of violence that take a positive value if either member of a couple indicated that there was violence. The � rst indicator variable is created from a question that asks whether, in the past year, argu-ments with the spouse or partner have “become physical.” The second measure, constructed separately for male and female respondents, considers whether the male or female partner was victimized,2 based on questions asking how often the respondent hit or was hit by the spouse or partner in the past year.

1. Given the rapidity of exits from cohabitation, our approximation of the entering population is less satisfac-tory for cohabiting than for married couples. However, even using a one-year cutoff results in very small samples, which would be even smaller if we chose a shorter relationship duration for de� ning the entering population.

2. While previous survey research has found that women and men are about equally likely to report hitting by their partner, the consequences of relationship violence appear to differ by sex. Women are more likely to be hurt when there is hitting in the relationship (Brush 1993), and experiencing violence is more likely to undermine a woman’s sense of personal control (Umberson et al. 1998).

Page 5: Why are cohabiting relationships more violent than marriages?

Violence in Cohabiting Relationships 131

Explanatory Variables

Explanatory variables include measures of � xed characteristics, that is, characteristics of the individuals and their prior family experiences that they bring to the marriage or cohabi-tation and that are not likely to be changed by the marriage or cohabitation experience per se. Variable characteristics, on the other hand, are characteristics that might change based on the experience of being married or cohabiting and the differential social supports that surround each type of union. We distinguish between these two types of characteristics in order to isolate the effects of selection on differences in behavior. Our � xed characteristics are age, education (coded as less than high school, a high school degree or some college, or a college degree or more),3 race (coded 1 for African Americans),4 whether the respondent lived with both biological parents at age 16 (coded 1 if he or she did not), and whether either partner was married prior to the current relationship or brought children to the cur-rent relationship (both coded 1 if true). These characteristics have been identi� ed in prior research as being differentially associated with cohabitation and marriage and also as being associated with intimate-partner violence (see, e.g., Anderson 1997; Ellison et al. 1999; Sorenson et al. 1996).

The variable characteristics include individual- and couple-level factors that previous researchers hypothesized were associated with institutional differences between marriage and cohabitation, that is, factors that could be changed by the marriage or cohabitation experience and the social supports (or lack thereof) that surround different types of union (e.g., Ellis 1989; Stets 1991; Wilson and Daly 2001). These include the frequency (mea-sured as the approximate number of days per month) of socializing with relatives, social-izing with peers, socializing in a bar, or participating in religious groups; the number of times per month the couple argued about spending time together; and an indicator variable equal to 1 if either member of the couple believed that their chances of separating exceeded 50%. We also include in this category variables that indicate whether the male or female in the relationship had a problem with drug or alcohol use (Cunradi et al. 2002), as well as a variable measuring total couple income.

Our measure of relationship duration begins with the start of coresidence for both mar-ried and cohabiting couples. For couples who were married at NSFH1 but who cohabited before marriage, the duration measure begins with the start of cohabitation.

RESULTSTable 1 presents descriptive statistics for cohabiting and married couples by relationship duration. The results in the � rst two columns of Table 1, comparing married and cohabiting couples in the “entering” sample, imply relatively little difference in the characteristics of couples selecting into marriage and cohabitation on many of our measures. Most important, physical violence is not substantially higher among cohabiting couples than among married couples. In addition, married and cohabiting couples in the entering population are similar in terms of age, race, living with both biological parents growing up, having been previ-ously married, and having had children by previous partners. A slightly higher proportion of married respondents have a college education or more (25% vs. 17%). By contrast, the entering population of cohabiting respondents spends signi� cantly less time participating in religious groups (0.49 days per month compared with 1.41 for married respondents) and signi� cantly more time socializing in bars (1.7 days per month compared with 0.87 for

3. Age and education are included as a � xed characteristics because they cannot decrease with relationship duration.

4. Because of the small sample size of Hispanics, and because (in this sample) Hispanics were similar to whites in rates of cohabitation and physical violence, Hispanics and whites were combined.

Page 6: Why are cohabiting relationships more violent than marriages?

132 Demography, Volume 43-Number 1, February 2006

Table 1. Sample Frequencies and Means, by Marital Status, Couples Coresiding Up to One Year and Up to Five Years

Relationship Duration Relationship Duration Less Th an 1 Year Less Th an 5 Years _____________________ _____________________Variable Married Cohabiting Married Cohabiting

Violence LevelsCouple’s arguments got physical (%) 19 21 19 **31**Female respondent hit (%)a 14 17 11 **23**Male respondent hit (%)b 14 17 15 **25**Mean age of respondent 29.43 28.66 30.43 29.49 (11.75) (9.59) (10.13) (9.11)

Fixed CharacteristicsAfrican American (%) 12 10 8 **14**

Less than high school education (%) 13 17 12 **20**

High school diploma or some college (%) 63 66 64 64

College degree or more (%) 25 *17* 24 **15**

Respondent not with both parents at age 16 (%) 29 31 25 **38**

Either partner previously married (%) 43 47 44 **57**

Child from previous relationship (%) 31 37 31 **48**

Variable CharacteristicsMean days per month socializing with relatives 3.86 3.05 3.18 2.96 (4.79) (4.48) (4.27) (4.57)

Mean days per month socializing with peers 5.53 6.73 5.41 **6.85** (7.16) (7.52) (6.75) (8.18)

Mean days per month socializing at bar 0.87 **1.70** 0.91 **1.64** (1.69) (3.18) (2.13) (3.39)

Mean days per month in religious group 1.41 **0.49** 1.44 **0.36** (3.21) (1.54) (3.59) (1.37)

Mean days per month arguing about 2.28 1.72 1.96 *2.80*time together (6.17) (4.96) (5.27) (6.62)

Male has a drug/alcohol problem (%) 3 3 4 6

Female has a drug/alcohol problem (%) 1 1 1 2

Either thinks likelihood of separation is > 50% (%) 16 **33** 18 **42**

Total couple income 33,198 31,446 38,804 *33,644* (22,364) (26,952) (41,155) (40,025)

Mean Duration of Relationship (years) 0.59 **0.45** 2.71 **1.66** (0.29) (0.28) (1.41) (1.39)

N 187 164 1,232 369

Notes: Numbers in parentheses are standard deviations. Signifi cance level indicates the results of tests of diff erence of means or proportions.

aFor female respondent hit: < 5 years duration sample, (N = 622 married, N = 169 cohabiting); < 1 year duration sample, (N = 88 married, N = 76 cohabiting).

bFor male respondent hit, < 5 years duration sample, (N = 579 married, N = 183 cohabiting); < 1 year duration sample, (N = 94 married, N = 84 cohabiting).

*p < .05; **p < .01

Page 7: Why are cohabiting relationships more violent than marriages?

Violence in Cohabiting Relationships 133

married respondents). In addition, about twice as many cohabiting couples (33% vs. 16%) include a partner who thinks the couple’s chances of separation exceed 50%.

The second two columns of Table 1 compare couples who have been together for up to � ve years. These results suggest that different kinds of individuals may accumulate in cohabitation and marriage. This explanation is particularly relevant for the comparisons of � xed characteristics, such as age, race, and family history, which cannot be explained by changes in behavior that are due to differences in the institutional supports or social norms surrounding the two relationship types.

When the sample includes longer-term relationships, signi� cantly higher proportions of cohabiting than married respondents had less than a high school education (20% vs. 12% of married respondents), did not live with both parents at age 16 (38% vs. 25% of married respondents), were previously married (57% vs. 44% of married respondents), or had a child from a previous relationship (48% vs. 31% of married respondents). The differ-ences between the two sets of results suggest that differential selection out of cohabitation contributes substantially to the signi� cant differences in the characteristics of married and cohabiting couples in cross-sectional samples.

Differences between married and cohabiting couples in the longer-duration sample are also greater on all of the violence measures and a number of other variable character-istics. For example, 31% of cohabitors reported that their arguments “got physical” in the past year, compared with 19% of married couples; cohabitors spent more time socializing with peers (6.85 days per month compared with 5.41 days for married respondents); and cohabitors reported arguing about spending time together an average of 2.80 days per month, compared with 1.96 days per month for married couples. None of these measures showed signi� cant differences between married and cohabiting couples in the “entering” sample. The greater differences in violence could arise from differences in selection out of relationship status, from changes in behavior associated with time spent in marriage or cohabitation, or from both.

To explore further whether the changes in violence and the other variables over time are due, at least in part, to “selection out” effects, we use longitudinal data to examine transitions out of marriage and cohabitation. Table 2 presents couples’ relationship status at NSFH2 (92–94) separately for NSFH1 cohabiting and married couples. The results provide important con� rmation of the patterns of selection out of cohabitation and marriage implied by the previous cross-sectional results.

The � rst three columns of Table 2 compare the characteristics of cohabitors in the entering sample who separated from their NSFH1 partners to the characteristics of those who continued to cohabit and those who married their original partners. The results show that for violence, couples who were still cohabiting at NSFH2 appear to be quite different from couples who married, although the small sample size of those still cohabiting (n = 15) suggests caution in interpreting these results. Although the difference does not reach statistical signi� cance, almost 27% of the couples still cohabiting at NSFH2 experienced physically violent arguments at NSFH1, but only 14% of couples who went on to marry their partners experienced violence at NSFH1. The difference is larger in magnitude, and signi� cant, for violence against the male partner: while over 27% of the still-cohabiting couples reported violence against the male at Time 1, only about 6% of those who later married reported such violence.

Similarly, in the next two columns of Table 2, we compare the NSFH1 level of violence of couples who were married in the entering sample but who had divorced or separated from their NSFH1 spouses to those who were still married to their NSFH1 spouses. One-third of the couples who had divorced or separated by NSFH2 experienced physically violent arguments at NSFH1, compared with about 13% of couples who were still married to their Time 1 spouses. In addition, almost 22% of couples who separated or divorced experienced violence against the female partner, and 40% experienced violence

Page 8: Why are cohabiting relationships more violent than marriages?

134 Demography, Volume 43-Number 1, February 2006

Tabl

e 2.

N

SFH

1 V

iole

nce

and

Rel

ated

Cha

ract

eris

tics

of M

arri

ed a

nd C

ohab

itin

g C

oupl

es, b

y R

elat

ions

hip

Stat

us a

t NSF

H2

Sa

mpl

e: R

elat

ions

hip

Dur

atio

n <

1 Y

ear a

t NSF

H1

Sam

ple:

Rel

atio

nshi

p D

urat

ion

< 5

Year

s at N

SFH

1

___

____

____

____

____

____

____

____

____

____

____

____

___

__

____

____

____

____

____

____

____

____

____

____

____

____

Re

latio

nshi

p St

atus

Re

latio

nshi

p St

atus

Re

latio

nshi

p St

atus

Re

latio

nshi

p St

atus

at

Tim

e 2

at T

ime

2 at

Tim

e 2

at T

ime

2

(Coh

abiti

ng a

t Tim

e 1)

(M

arrie

d at

Tim

e 1)

(C

ohab

iting

at T

ime

1)

(Mar

ried

at T

ime

1)

___

____

____

____

____

____

____

__

____

____

____

____

___

__

____

____

____

____

____

____

___

___

____

____

____

____

Still

Sepa

rate

d/

Still

Sepa

rate

d/Va

riabl

e Se

para

ted

Coh

abiti

ng

Mar

ried

Div

orce

d M

arrie

d Se

para

ted

Coh

abiti

ng

Mar

ried

Div

orce

d M

arrie

d

Vio

lenc

e Le

vels

Argu

men

ts “g

ot p

hysic

al”

(%)

c 31.9

c 26

.7

14.3

33

.3g

12.6

c 42

.6c

b 35.3

b 21

.9

g 31.1

g 15

.5

Fem

ale

resp

onde

nt h

it (%

) 26

.1

25.0

14

.3

21.7

h 07

.8

c 34.0

c 28

.6

14.7

g 16

.8g

09.2

(n

= 2

3)

(n =

4)

(n =

35)

(n

= 2

3)

(n =

51)

(n

= 5

0)

(n =

21)

(n

= 6

8)

(n =

101

) (n

= 4

24)

Mal

e re

spon

dent

hit

(%)

d 19.1

d a 27

.3a

0,5.

9 40

.0g

09.1

d 29

.4d

b 30.8

b 14

.9

g 28.3

g 10

.8

(n =

21)

(n

= 1

1)

(n =

34)

(n

= 1

0)

(n =

66)

(n

= 5

1)

(n =

26)

(n

= 6

7)

(n =

92)

(n

= 3

88)

Fixe

d C

hara

cter

istic

sM

ean

age

of re

spon

dent

d,

f 25.9

6d,f

29.2

7 28

.83

g 25.1

7g 30

.03

e 27.9

7e a 32

.02a

29.0

1 27

.98g

30.6

0

0(7.

48)

0(7.

70)

0(9.

95)

0(9.

42)

(11.

00)

0(7.

50)

0(9.

46)

(8.6

9)

(8.4

6)

0(9.

20)

Afric

an A

mer

ican

(%)

c 14.9

c 33

.3a

01.4

11

.1

08.4

16

.5c

a 25.5

a 05

.1

g 12.1

g 05

.8

Less

than

hig

h sc

hool

edu

catio

n (%

) 17

.0

20.0

15

.7

16.7

10

.9

21.7

25

.5

17.5

13

.2

09.3

Hig

h sc

hool

dip

lom

a or

som

e co

llege

(%)

63.

8 73

.3

61.4

h 75

.0h

57.1

60

.0

62.8

65

.0

h 69.3

h 62

.3

Col

lege

deg

ree

or m

ore

(%)

19.2

b0

6.7b

22.9

g 0

8.3g

31.9

18

.3

11.8

17

.5

g 17.6

g 28

.4

Not

with

bot

h pa

rent

s at a

ge 1

6 (%

) 31

.9

b 46.7

b 24

.3

33.3

24

.4

39.1

b 47

.1b

33.6

g 36

.9g

20.9

Eith

er p

artn

er p

revi

ously

mar

ried

(%)

29.8

c 46

.7

50.0

38

.9

46.2

47

.8d

60.8

59

.9

45.2

43

.2

Chi

ld fr

om p

revi

ous r

elat

ions

hip

(%)

36.0

33

.0

36.0

31

.0

33.0

49

.6

54.9

48

.2

35.0

30

.8

Page 9: Why are cohabiting relationships more violent than marriages?

Violence in Cohabiting Relationships 135Va

riabl

e C

hara

cter

istic

sM

ean

days

per

mon

th so

cial

izing

3.

40

4.58

3.

07

4.13

3.

96

2.74

3.

17

2.96

g 3.

81g

3.03

with

rela

tives

(5

.10)

(5

.61)

(4

.33)

(5

.20)

(4

.82)

(4

.74)

(4

.64)

(4

.24)

(4

.76)

(4

.03)

Mea

n pe

r mon

th d

ays s

ocia

lizin

g 7.

80

7.31

5.

50

7.03

4.

87

7.33

7.

46

5.74

g 6.

95g

4.81

with

pee

rs

(8.2

3)

(7.2

2)

(6.6

4)

(8.4

0)

(6.8

8)

(9.0

0)

(7.9

1)

(6.8

0)

(7.9

0)

(6.0

6)

Mea

n da

ys p

er m

onth

soci

alizi

ng a

t bar

e 2.

41e

0.87

1.

73

0.66

0.

98

1.99

1.

41

1.41

1.

01

0.86

(4

.15)

(1

.45)

(3

.20)

(1

.00)

(1

.92)

(3

.85)

(3

.12)

(2

.94)

(2

.10)

(2

.03)

Mea

n da

ys p

er m

onth

in re

ligio

us g

roup

s 0

.27

e 0.20

b 0.

65

e 0.39

g 1.

70

0.21

0.

23

0.38

h 1.

12h

1.58

(0

.68)

(0

.32)

(2

.06)

(1

.00)

(3

.64)

(0

.62)

(0

.65)

(1

.54)

(3

.31)

(3

.74)

Mea

n da

ys p

er m

onth

arg

uing

abo

ut

2.07

4.

21

1.01

e 4.

89g

1.18

2.

49

b 4.84

b 2.

33

g 2.78

g 1.

67tim

e to

geth

er

(4.1

7)

(10.

08)

(3.4

3)

(9.4

0)

(3.9

1)

(5.0

3)

(9.9

3)

(6.0

2)

(6.4

5)

(4.6

3)

Mal

e ha

s a d

rug/

alco

hol p

robl

em (%

) 6.

0 0.

0 3.

0 g 8.

0g 1.

0 9.

6d 7.

8 4.

4 g 6.

8g 3.

2

Fem

ale

has a

dru

g/al

coho

l pro

blem

(%)

2.0

0.0

0.0

0.0

1.0

4.3d

3.9

0.7

0.9

0.6

Eith

er th

inks

like

lihoo

d of

se

para

tion

is >

50%

(%)

e 49c

e 47a

20.0

28

.0g

12.0

c,

e 60.0

c,e

39.2

28

.9

g 33.0

g 12

.6

Tota

l Cou

ple

Inco

me

($)

e 25,1

54c

30,3

29

36,4

32

e 26,0

73g

36,6

62

33,2

75

33,7

13

36,4

68

g 34,8

22g

40,8

99

(16,

772)

(1

8,70

7)

(36,

759)

(1

4,03

3)

(24,

910)

(5

7,38

0)

(32,

972)

(3

1,79

3)

(27,

940)

(4

5,70

2)

N

47

15

70

36

119

115

51

137

206

826

Not

e: Ex

cept

whe

re o

ther

wise

indi

cate

d, n

umbe

rs in

par

enth

eses

are

stan

dard

dev

iatio

ns.

a Valu

e fo

r tho

se w

ho a

re st

ill c

ohab

iting

is d

iff er

ent f

rom

that

of t

hose

who

mar

ried

(p <

.05)

.b Va

lue

for t

hose

who

are

still

coh

abiti

ng is

diff

eren

t fro

m th

at o

f tho

se w

ho m

arrie

d (p

< .1

0).

c Valu

e fo

r tho

se w

ho h

ave

sepa

rate

d is

diff e

rent

from

that

of t

hose

who

mar

ried

(p <

.05)

.d Va

lue

for t

hose

who

hav

e se

para

ted

is di

ff ere

nt fr

om th

at o

f tho

se w

ho m

arrie

d (p

< .1

0).

e Valu

e fo

r tho

se w

ho h

ave

sepa

rate

d is

diff e

rent

from

that

of t

hose

still

coh

abiti

ng (p

< .0

5).

f Valu

e fo

r tho

se w

ho h

ave

sepa

rate

d is

diff e

rent

from

that

of t

hose

still

coh

abiti

ng (p

< .1

0).

g Valu

e fo

r tho

se w

ho h

ave

sepa

rate

d or

div

orce

d is

diff e

rent

from

that

of t

hose

still

mar

ried

(p <

.05)

.h Va

lue

for t

hose

who

hav

e se

para

ted

or d

ivor

ced

is di

ff ere

nt fr

om th

at o

f tho

se st

ill m

arrie

d (p

< .1

0).

Page 10: Why are cohabiting relationships more violent than marriages?

136 Demography, Volume 43-Number 1, February 2006

against the male partner, compared with about 8% and 9%, respectively, among couples who remained married.

Because the entering sample of NSFH1 couples was small, in the � nal � ve columns of Table 2, we repeat the longitudinal analysis on the larger sample of couples in relation-ships of less than � ve years’ duration at NSFH1. As we noted earlier, the longer-duration relationships in this sample were already exposed to “selection out” effects, any in� uences of marital-status differences in social norms and supports on behavior, or both, by NSFH1. However, it is reasonable to assume that processes of selection out will continue to occur over time, even to couples who had been together for a longer time at Time 1. The results for the larger sample con� rm our � ndings for the entering sample: violence is higher among NSFH1 cohabiting couples who were still cohabiting at NSFH2 than among those who married their partners by NSFH2. About 35% of the couples still cohabiting at NSFH2 ex-perienced physically violent arguments at NSFH1, compared with only 22% of couples who went on to marry their partners. About 30% of couples who were still cohabiting at NSFH2 reported violence against the male or female partner at NSFH1 (28.6% against the woman and 30.8% against the man), while only about 15% of those who later married reported such violence. The last two columns of Table 2 reinforce the effects of selection out for the larger sample of married couples. About 31% of the couples who had divorced or separated by NSFH2 experienced physically violent arguments at NSFH1, compared with 15.5% of couples who were still married to their Time 1 spouses. Almost 17% of couples who sepa-rated or divorced experienced violence against the female partner, and 28.3% experienced violence against the male partner, compared with about 9% and 11%, respectively, among couples who remained married.

These results suggest that differential selection out of cohabitation and marriage be-cause of violence is an important part of the explanation for the differences in violence seen in cross-sectional comparisons. Because cohabiting couples with lower levels of violence tend to select out of cohabitation into marriage, couples with higher levels of violence ac-cumulate in the cohabiting population, even though a substantial proportion of cohabiting couples in violent relationships separate. Indeed, those remaining in cohabitation are more similar to those who separate (who had the highest levels of violence) than they are to those who marry. Similarly, because violent married couples tend to end their relationships, less-violent couples accumulate in the married population.

Table 2 also provides evidence that a number of marital-status differences in other characteristics result (at least in part) from differences in selection out of cohabitation and marriage. For example, African American couples, as well as couples who did not live with both parents at age 16, appear to accumulate in the cohabiting population relative to the en-tering samples, whereas couples who are not African American, who have a college degree or more, and who lived with both parents at age 16 appear to accumulate in the married population relative to the entering samples.

Selection explains some differences in our variable (non� xed) characteristics as well. For example, among cohabiting couples, those who were still cohabiting at Time 2 spent more time at NSFH1 arguing about spending time together than either those who separated or those who married their partners. Meanwhile, married couples who separated or divorced spent signi� cantly more time at NSFH1 arguing about spending time together than those who stayed married. Similarly, differential selection out of cohabitation and marriage is as-sociated with participation in religious groups. For the entering sample, cohabiting couples who remained cohabiting participated less often in religious activities than cohabitors who eventually married. At the same time, married couples who subsequently separated or divorced spent signi� cantly less time participating in religious groups than those who remained married. Thus, couples who argue more frequently, as well as less-religious couples, tend to accumulate in cohabitation, while couples who argue less frequently, as well as more-religious couples, tend to accumulate in marriage.

Page 11: Why are cohabiting relationships more violent than marriages?

Violence in Cohabiting Relationships 137

On a number of other characteristics, we found evidence for the accumulation of “ better” couples (i.e., those who were less disposed to violence) couples in marriage, even where the effects of selection on who accumulates in longer-term cohabitations are less clear. NSFH1 married couples who stayed married had signi� cantly higher average in-comes than those who separated ($40,899 vs. $34,822 in the larger sample). Selection out of marriage by couples in which the male had a drug or alcohol problem also appears to result in couples without this problem accumulating in the married group over time.

CONCLUSIONSAs Casper and Sayer (2000) and Manning and Smock (2002) emphasized, an entering co-hort of cohabitors is a heterogeneous group.5 Some of them have marriage plans and tend to be better educated, to have grown up in nonpoor, two-parent households, and to be less physically aggressive in arguments. The results presented here suggest that as this subgroup moves out of cohabitation and into marriage, those who remain in cohabiting unions are, on average, less educated, more likely not to have lived with both parents while growing up, less committed to the relationship, and more likely to have used violence against their partners. Given the brief duration of the majority of cohabiting relationships, further research using larger entering samples of cohabiting couples, as well as more-frequent follow-up measures of relationship transitions, are needed to con� rm these results. However, couples who re-main in cohabiting relationships seem to be more similar on many measures to couples who separate than they are to couples who marry. The result of this selection out of cohabita-tion into marriage, as Lundberg and Pollak (2001) predicted, is that the characteristics and behaviors of those who ever cohabit differ in important ways from the characteristics of the population of cohabiting couples at a point in time. Cross-sectional samples are likely to overrepresent cohabiting couples who are violent and who have a variety of other char-acteristics associated with violence.

Similarly, entering populations of married couples are relatively heterogeneous, and they are also affected by selection processes operating over time. Indeed, the results pre-sented here are stronger regarding the effects of selection out of marriage than out of co-habitation. Unlike cohabiting couples who can separate or marry, however, married couples have only one exit option: separation. Our results show that married couples with violent relationships, as well as those with lower levels of education and lower income, those who are younger, and those who are less religious, are more likely to separate. Again, as Repetti (2001) speculated, the result is that less-violent couples (and those with higher socioeco-nomic status) tend to accumulate in the married population.

Although differential selection out of marriage and cohabitation appears to be an important part of the explanation for observed cross-sectional differences in violence between married and cohabiting couples, we cannot rule out the possibility that some of the difference may also be attributable to causal effects of spending time in relationships with different social supports and institutional characteristics. To fully explore the relative contributions of differential selection and institutional causes, we would need longitudinal data that includes a substantially larger sample of entering couples, as well as frequent and consistent measures of violence and other non� xed characteristics over time. With such data, we could determine whether, in addition to the effects of differential selection, violence decreases with duration among couples who remain married and increases with duration among couples who remain cohabiting. Because of missing NSFH2 spouse/ partner

5. Cohabitation is also changing rapidly and spreading to segments of the population in which it was previ-ously rare, such as those with college degrees (Bumpass and Lu 2000; Seltzer 2000; Smock 2000). Such increases in cohabitation will likely make current entering populations of cohabitors even more heterogeneous than they were at the time of NSFH1. At the same time, evidence that the average duration of cohabitations is getting shorter (Bumpass and Lu 2000) suggests that selective exits may be happening even more rapidly than in the past. Thus, understanding how such exits affect the stock of cohabiting couples at a point in time continues to be important.

Page 12: Why are cohabiting relationships more violent than marriages?

138 Demography, Volume 43-Number 1, February 2006

interviews and missing values on NSFH2 violence and other measures, we could not carry out such an analysis on the already-small sample of NSFH1 entering couples.

Nonetheless, the results presented here suggest that some researchers may have been premature in attributing higher levels of violence among cohabiting couples to the failure of institutional controls. Indeed, taken together, the differences in selection out of cohabitation and marriage suggest that cross-sectional comparisons of intimate- partner violence by marital status are likely to be biased. Our results also have implications for other comparisons between married and cohabiting couples in which the dependent variable of interest—including other aspects of relationship quality, job stability, or household divi-sion of labor—is likely to be related to selection out of relationship status.

In his review of research on marital quality during the 1980s, Glenn (1990:820) ar-gued that “any correlation of the dependent variable with a tendency to move out of the population poses serious problems for causal inference.” Glenn continued, “the population of married persons is not approximately closed, and the tendency to move out of it is rather highly correlated with marital quality.” Because cohabiting relationships tend to be shorter than marriages (Bumpass and Lu 2000), the population of cohabiting couples is even less “ approximately closed” than that of married couples. Accordingly, accounting for the effects of selection into and out of these populations on the characteristics of their members is a critical � rst step that must be taken before attributing differences between the groups to institutional in� uences on behavior.

REFERENCESAnderson, K.L. 1997. “Gender, Status, and Domestic Violence: An Integration of Feminist and Fam-

ily Violence Approaches.” Journal of Marriage and Family 59:655–69. Brines, J. and K. Joyner. 1999. “The Ties That Bind: Principles of Cohesion in Cohabitation and

Marriage.” American Sociological Review 64:333–55. Brownridge, D.A. and S.S. Halli. 2002. “Understanding Male Partner Violence Against Cohabiting

and Married Women: An Empirical Investigation With a Synthesized Model.” Journal of Family Violence 17:341–61.

Brush, L. 1993. “Violent Acts and Injurious Outcomes in Married Couples: Methodological Issues in the NSFH.” Gender and Society 4(1):56–67.

Bumpass, L. and H. Lu. 2000. “Trends in Cohabitation and Implications for Children’s Family Con-texts in the United States.” Population Studies 54(1):29–41.

Bumpass, L.L. and J.A. Sweet. 1989. “National Estimates of Cohabitation.” Demography 26:615–25.

Casper, L. and L. Sayer. 2000. “Cohabitation Transitions: Different Attitudes and Purposes, Different Paths.” Paper presented at the annual meeting of the Population Association of America, Los Angeles, March 23–25.

Cunradi, C., R. Caetano, and J. Schafer. 2002. “Socioeconomic Predictors of Intimate Partner Violence Among White, Black, and Hispanic Couples in the United States.” Journal of Family Violence 17:377–89.

DeMaris, A. 2000. “Till Discord Do Us Part: The Role of Physical and Verbal Con� ict in Union Disruption.” Journal of Marriage and the Family 62:683–92.

———. 2001. “The In� uence of Intimate Violence on Transitions Out of Cohabitation.” Journal of Marriage and Family 63:235–46.

Ellis, D. 1989. “Male Abuse of a Married or Cohabiting Female Partner: The Application of Socio-logical Theory to Research Findings.” Violence and Victims 4:235–55.

Ellison, C.G., J.P. Bartkowski, and K.L. Anderson. 1999. “Are There Religious Variations in Domes-tic Violence?” Journal of Family Issues 20(1):87–113.

Gelles, R. 1983. “Toward a Theory of Intra-Familial Violence: An Exchange/Social Control Theory.” Pp. 151–65 in The Dark Side of Families: Current Family Violence Research, edited by D. Finkelhor, R. Gelles, G. Hotaling, and M.A. Straus. Beverly Hills, CA: Sage.

Page 13: Why are cohabiting relationships more violent than marriages?

Violence in Cohabiting Relationships 139

Glenn, N. 1990. “Quantitative Research on Marital Quality in the 1980s: A Critical Review.” Journal of Marriage and the Family 52:818–31.

Heaton, T.B. 1991. “Time-Related Determinants of Marital Dissolution.” Journal of Marriage and the Family 53:285–95.

Kantor, G. and J. Jasinski. 1998. “Dynamics and Risk Factors in Partner Violence.” Pp. 1–43 in Partner Violence: A Comprehensive Review of 20 Years of Research, edited by J.L. Jasinski and L.M. Williams. Thousand Oaks, CA: Sage.

Lundberg, S. and R. Pollak. 2001. “Bargaining and Distribution in Families.” Pp. 314–40 in The Well-being of Children and Families: Research and Data Needs, edited by A. Thornton. Ann Arbor, MI: University of Michigan Press.

Lupri, E., E. Grandin, and M. Brinkerhoff. 1994. “Socioeconomic Status and Male Violence in the Canadian Home: A Reexamination.” Canadian Journal of Sociology 19:47–73.

Manning, W. and P. Smock. 2002. “First Comes Cohabitation and Then Comes Marriage?” Journal of Family Issues 23:1065–87.

McNally, J., S. Sassler, and R. Schoen. 1997. “Misplaced Affection: The Use of Multiple Imputation to Reconstruct Missing Cohabiting Partner Information in the NSFH.” PSTC Working Paper #97-09. Population Studies and Training Center, Brown University, Providence, RI.

Nock, S. 1995. “A Comparison of Marriages and Cohabiting Relationships.” Journal of Family Is-sues 16(1):53–76.

Raley, R.K. 1996. “A Shortage of Marriageable Men? A Note on the Role of Cohabitation in Black-White Differences in Marriage Rates.” American Sociological Review 61:973–83.

Repetti, R. 2001. “Searching for the Roots of Marital Con� ict in Uxoricides and Uxorious Husbands.” Pp. 47–55 in Couples in Con� ict, edited by A. Booth, A.C. Crouter, and M. Clements. Mahwah, NJ: Lawrence Erlbaum Associates.

Rindfuss, R. and A. VandenHeuvel. 1990. “Cohabitation: A Precursor to Marriage or an Alternative to Being Single?” Population and Development Review 16:703–26.

Rodriquez, E., K. Lasch, P. Chandra, and J. Lee. 2001. “Family Violence, Employment Status, Welfare Bene� ts, and Alcohol Drinking in the United States: What Is the Relation?” Journal of Epidemiology and Community Health 55:172–78.

Rogge, R. and T. Bradbury. 1999. “Till Violence Does Us Part: The Differing Roles of Communication and Aggression in Predicting Adverse Marital Outcomes.” Journal of Consulting and Clinical Psychology 67:340–51.

Sanchez, L., W. Manning, and P. Smock. 1998. “Sex-Specialized or Collaborative Mate Selection? Union Transitions Among Cohabitors.” Social Science Research 27:280–304.

Seltzer, J. 2000. “Families Formed Outside of Marriage.” Journal of Marriage and the Family 62:1247–68.

Smock, P. 2000. “Cohabitation in the United States: An Appraisal of Research Themes, Findings, and Implications.” Annual Review of Sociology 26:1–20.

Smock, P. and W. Manning. 1997. “Cohabiting Partners’ Economic Circumstances and Marriage.” Demography 34:331–41.

Sorenson, S., D. Upchurch, and H. Shen. 1996. “Violence and Injury in Marital Arguments: Risk Patterns and Gender Differences.” American Journal of Public Health 86:35–40.

Stets, J. 1991. “Cohabiting and Marital Aggression: The Role of Social Isolation.” Journal of Mar-riage and the Family 53:669–80.

Stets, J. and M. Straus. 1989. “The Marriage License as a Hitting License: A Comparison of Assaults in Dating, Cohabiting, and Married Couples.” Journal of Family Violence 4:161–80.

Sweet, J., L. Bumpass, and C. Vaughn. 1988. “The Design and Content of the National Survey of Families and Households.” Working Paper #1. Center for Demography and Ecology, University of Wisconsin–Madison.

Szinovacz, M. 1983. “Using Couple Data as a Methodological Tool: The Case of Marital Violence.” Journal of Marriage and the Family 45:633–44.

Page 14: Why are cohabiting relationships more violent than marriages?

140 Demography, Volume 43-Number 1, February 2006

Umberson, D., K. Anderson, J. Glick, and A. Shapiro. 1998. “Domestic Violence, Personal Control, and Gender.” Journal of Marriage and the Family 60:442–52.

Wilson, M. and M. Daly. 2001. “The Evolutionary Psychology of Couple Con� ict in Registered Ver-sus De Facto Marital Unions.” Pp. 3–26 in Couples in Con� ict, edited by A. Booth, A.C. Crouter, and M. Clements. Mahwah, NJ: Lawrence Erlbaum Associates.

Wilson, M., M. Daly, and C. Wright. 1993. “Uxoricide in Canada: Demographic Risk Patterns.” Canadian Journal of Criminology 35:263–91.

Wilson, M., H. Johnson, and M. Daly. 1995. “Lethal and Nonlethal Violence Against Wives.” Cana-dian Journal of Criminology 37:331–61.

Yamaguchi, K. and D. Kandel. 1985. “Dynamic Relationships Between Premarital Cohabitation and Illicit Drug Use: An Event-History Analysis of Role Selection and Role Socialization.” American Sociological Review 50:530–46.

Yllo, K. and M. Straus. 1981. “Interpersonal Violence Among Married and Cohabiting Couples.” Family Relations 30:339–47.