The Pre-engagement Cohabitation Effect: A Replication and Extension of Previous Findings

Abstract

Using a random telephone survey of men and women married within the past 10 years (N = 1050), the current study replicated previous findings regarding the timing of engagement and the premarital cohabitation effect (see Kline et al., 2004). Those who cohabited before engagement (43.1%) reported lower marital satisfaction, dedication, and confidence as well as more negative communication and greater proneness for divorce than those who cohabited only after engagement (16.4%) or not at all until marriage (40.5%). These differences were generally small, but could not be accounted for by length of marriage or by variables often associated with selection into cohabitation (i.e., age, income, education, and religiousness). Similar results were found in a subsample of individuals who cohabited only with the current spouse. There were no significant differences between those who cohabited after engagement and not at all before marriage, supporting a pre-engagement, but not a premarital cohabitation effect.

Living together before marriage has become the norm in the United States (Smock, 2000). Upwards of 70% of couples now do so (Bumpass & Lu, 2000; Stanley, Whitton, & Markman, 2004), yet there is evidence, even from couples who married in recent years, that premarital cohabitation is a risk factor for subsequent divorce (Kamp Dush, Cohan, & Amato, 2003), as well as lower marital quality (Cohan & Kleinbaum, 2002; Stanley et al., 2004). This phenomenon has been termed the “cohabitation effect.”

Although selection factors seem to account for a portion of the cohabitation effect (e.g., Woods & Emery, 2002), many have argued that there may also be something about the experience of cohabitation that could lead to higher marital instability and lower marital quality (e.g., Kamp Dush et al., 2003). Evidence from qualitative work suggests that many couples slide into cohabitation, without discussing what it means for the future of the relationship (Manning & Smock, 2005). Stanley, Rhoades, and Markman (2006) hypothesized that these couples who slide into cohabitation, and who do not make a mutual commitment to marriage first, would be most at risk for later marital difficulties. According to their inertia theory, cohabitation (compared to dating without cohabitation) increases the chances for marriage, partly because constraints to stay together (e.g., financial investments, pregnancy, social pressure) increase when partners begin cohabiting. Thus, the experience of cohabitation may lead some couples to marry, even though they would not have chosen to marry if they had not already been living together (Stanley et al., 2006). In essence, the idea of inertia does not so much suggest that cohabiting weakens relationships, but that cohabiting makes some already weaker relationships more likely to continue into marriage. In this way, inertia theory is compatible with selection theory; selection helps explain who cohabits and inertia helps explain why those relationships continue.

This inertia perspective has not been widely tested, but two studies support its most basic premise that the premarital cohabitation effect should be strongest for cohabitations that start before a mutual commitment to marriage is made. First, using data from 2000 individuals taking part in the 1987–88 wave of the National Survey of Families and Households, Brown and Booth (1996) found that currently cohabiting individuals without plans to marry had poorer relationship quality than married individuals, but that there was no significant difference in the relationship quality between cohabiting individuals with plans to marry and marrieds. Second, in a much smaller convenience sample of 136 couples, Kline et al. (2004) extended this research by showing that couples who had lived together before they were engaged had poorer marital quality than couples who lived together only after engagement or not at all before marriage. These findings are important because they demonstrate which cohabiting couples might be most at risk for later marital difficulties (should they marry), but the sample was limited in terms of generalizability.

The current study sought to replicate these findings on how engagement status at the time of cohabitation moderates the cohabitation effect in a recent, large, and random sample of married individuals. We hypothesized that married individuals who lived with their spouses before engagement would report having poorer quality marriages (in terms of general satisfaction, the negativity of their communication, confidence in the future of the relationship, commitment, level of friendship with each other, and satisfaction with their sexual/sensual relationship), as well as greater self-reported divorce potential than those who lived together after engagement or not at all before marriage. Because the concept of inertia predicts risk beyond what typical selection factors account for, we hypothesized that these findings would hold even controlling for length of marriage and characteristics that have been shown to be associated with selection into cohabitation, including religiousness, education, age, and income.

The current study also followed up on Teachman’s (2003) conclusion that women are not at greater risk for divorce because of premarital cohabitation if they cohabit only with their subsequent husbands (Teachman, 2003). Teachman speculated that people who cohabit only with their eventual spouse tend to have higher commitment to marriage in general and are therefore less likely to divorce. Consistent with the concept of inertia, however, we predicted that even those who cohabit with only a future spouse, if before engagement, will have lower marital quality. To test this prediction in the current sample, we limited some analyses to include only individuals who cohabited with their spouses and not with any other romantic partners.

Method

Participants

For this study, 1050 married men (n = 523) and women (n = 527) from different relationships completed a brief telephone survey regarding their cohabitation histories and several aspects of their marriages. Qualifying questions limited participation to those 18 – 34 years old (M = 30.67, SD = 3.00) and married 10 years or less (M = 6.20, SD = 2.91). (The sample therefore only includes individuals who married between 1996 and 2007.) The sample was 84.9% White, 5.2% Black or African American, 1.8% Asian, and 5.7% Hispanic or Latino; 2.4% identified themselves as some other race/ethnicity or did not answer the questions regarding race and ethnicity. On average, participants made $35,000–40,000 annually and had completed some post secondary-school education, but had not obtained a degree. The majority of participants (91.8%) had never been divorced. Regarding cohabitation history, 40.5% reported that they did not live with their spouse before marriage, 16.4% cohabited only after engagement, and 43.1% cohabited before engagement.¹

Procedure

Following a protocol approved by a university institutional review board, a professional calling center used a targeted-listed telephone sampling strategy to call households within the contiguous United States. This strategy employs purchased lists of telephone numbers acquired through local listings, survey data, public records, magazine subscriptions, and many other sources and is becoming increasingly useful as the number of working home (landline) telephone numbers declines in the United States. It was also the most cost effective way to obtain a large, random sample of married individuals. After a brief introduction to the study, respondents were asked qualifying questions about age and marital status. If they qualified, they provided verbal consent and completed a 7- to 10-minute telephone survey. Of those who were contacted, willing to answer the screening questions, and qualified for the study, 1050 (95%) completed the full interview; 51 (5%) either refused to complete the interview or became disconnected.

Measures

When possible, we chose measures that were used by Kline et al. (2004), in other phone surveys, or both. Specifically, dedication, confidence, and negative communication were measured by shorter versions of Kline et al.’s scales and the dedication, negative communication, relationship satisfaction, sexual satisfaction, and religiousness measures used in the current study have been used in other phone survey research (Stanley, Amato, Johnson, & Markman, 2006; Stanley et al., 2004). Details on all of these measures follow.

Continuous relationship variables

For relationship satisfaction, participants answered “All in all, how satisfied are you with your relationship?” on a 1 (completely satisfied) to 5 (not at all) scale (reverse coded for analyses). Dedication (or interpersonal commitment) was measured by four items from the Dedication Scale (also see Stanley & Markman, 1992). An example item is “My relationship with my spouse is more important to me than almost anything else in my life,” assessed on a 1 (strongly agree) to 5 (strongly disagree) scale (reverse coded for analyses). Cronbach’s alpha (α) for dedication was .72. Relationship confidence was measured by combining two items from the Confidence Scale (also see Whitton et al., 2007), “I believe we can handle whatever conflicts will arise in the future” and “I feel good about our prospects to make this relationship work for a lifetime,” assessed on a 1 (strongly agree) to 5 (strongly disagree) scale (reverse coded for analyses, r = .54 between the two items, α = .69). For sexual satisfaction, participants rated “We have a satisfying sensual or sexual relationship” on a 1 (strongly agree) to 5 (strongly disagree) scale (reverse coded for analyses). Friendship was measured by combining two items from Stanley, Markman, and Whitton (2002), “We regularly have great conversations where we just talk as good friends” and “We have a lot of fun together,” rated on a 1 (strongly agree) to 5 (strongly disagree) scale (reverse coded for analyses, α = .71). Negative communication was measured with four items rated on a 1 (never or almost never) to 3 (frequently) scale. An example is “When we argue, one of us withdraws, doesn’t want to talk about it anymore or leaves the scene;” α = .67.

There is precedence for treating these variables as separate constructs (e.g., Kline et al., 2004: Stanley et al., 2004), but the magnitudes of the correlations among these six continuous variables ranged from .37 to .61 (M = .48) so we ran additional tests. The items from the multi-item scales (dedication, confidence, friendship, and negative communication) were entered into confirmatory factor analyses. The results (obtainable from the first author) demonstrated that treating these constructs as four separate factors was a better fit for the data than using a single composite of all 12 items. Treating them separately also has conceptual advantages over using a marital quality composite that blends different constructs (Fincham & Bradbury, 1987).

Dichotomous divorce potential variable

We included an indicator of divorce potential that was based on work by Booth, Johnson, and Edwards (1983), “Have you or your spouse ever seriously suggested the idea of divorce?” Responses of “never” were coded as 0 and any response from “yes, within the last three months” to “yes, but not in the last three years” was coded as 1 for the current study.

Demographic variables

Single items measured religiousness (“All things considered, how religious would you say that you are?”), premarital cohabitation history (before vs. after engagement or not until marriage for the current relationship, total number of cohabitation partners), length of marriage, education, and income.

Missing data

Due to our error, the items about friendship and sexual satisfaction were not asked during the first 88 interviews. These data are assumed to be missing completely at random and we therefore treated them with listwise deletion. Missing data were negligible on all other key variables, with no more than three observations missing from any item.

Results

Because the sample is large, the alpha for all hypothesis tests was set at p < .01. To test the hypothesis that married individuals who lived with their spouses before engagement would report having poorer quality marriages in terms of satisfaction, negative communication, confidence, dedication, friendship, and sexual satisfaction than those who lived together after engagement or not at all before marriage, six one-way analyses of variances (ANOVAs) were conducted (see Table 1). There were significant main effects of premarital cohabitation history for four of these six dependent variables: satisfaction (F(2, 1038) = 4.90, p < .01), negative communication (F(2, 1040) = 12.12, p < .001), confidence (F(2, 1039) = 6.13, p < .01), and dedication (F(2, 1040) = 7.12, p < .01). For these four variables, t-tests were conducted to compare the three types of cohabitation history. For the four variables with significant differences, those who cohabited before engagement reported significantly poorer quality marriages than those who did not live together premaritally. For satisfaction, dedication, and confidence, those who lived together before engagement reported significantly lower scores than those who cohabited only after engagement. In addition to these ANOVAs, a chi-square showed that there were significant differences in self-reported divorce potential based on cohabitation history, χ²(2, N = 1040) = 13.39, p < .01, with those who cohabited before engagement being significantly more likely to have ever suggested divorce (18.7%) than those who did not live together premaritally (10.2%, χ²(1, N = 869) = 12.48, p < .001). In contrast to our hypothesis, those who cohabited before engagement were not significantly more likely to have ever suggested divorce than those who cohabited after engagement (12.3%, χ²(1, N = 620) = 3.64, p = .06). There were no significant differences between those who cohabited after engagement and not at all before marriage on any of these variables.

Table 1.

Means, Standard Deviations, and Effect Sizes for Pre-engagement Cohabitation History Comparisons

Construct	Before engagement M (SD)	After engagement M (SD)	Not until marriage M (SD)	Before vs. after engagement ES (with cov.)	Before vs. not until marriage ES (with cov.)	After vs. not until marriage ES (with cov.)
Relationship satisfaction	4.46 (.74)	4.64 (.54)	4.55 (.60)	−.27* (−.23*)	−.13 (−.10)	.15 (.14)
Relationship dedication	4.55 (.56)	4.66 (.39)	4.66 (.46)	−.25* (−.13)	−.23* (−.12)	.00 (.00)
Relationship confidence	4.59 (.77)	4.77 (.42)	4.72 (.61)	−.30* (−.27*)	−.18* (−.12)	.10 (.16)
Sexual satisfaction	4.45 (.80)	4.48 (.73)	4.48 (.76)	−.05 (.00)	−.05 (.02)	.00 (.03)
Friendship	4.45 (.72)	4.53 (.55)	4.51 (.60)	−.12 (−.08)	−.09 (−.02)	.03 (.06)
Negative communication	1.46 (.45)	1.39 (.34)	1.33 (.36)	.18 (.13)	.33* (.25*)	.18 (.14)

Notes. Effect sizes are Cohen’s d values; effect sizes reported in parentheses, “(with cov.)” are the effect sizes controlling for marriage length, religiousness, and education.

^*Indicates significant differences between two groups, based on t-tests, p < .05.

We also hypothesized that these findings would hold even when controlling for selection factors. To know what selection factors to include as covariates in ANCOVAs, we first tested for possible differences across the groups based on cohabitation history. We found significant differences for length of marriage, religiousness, and education level (ps < .001), but no significant differences on income or age, ps > .30. Thus, we entered marriage length, religiousness, and education level as covariates in ANCOVAs with satisfaction, negative communication, confidence, dedication as dependent variables. For dedication only, the overall main effect of cohabitation history fell to non-significance when accounting for these covariates, F(2, 948) = 1.90, p > .05.² The effect sizes were weaker when these covariates were included (see Table 1). A logistic regression (not reported for sake of space) showed that the difference in self-reported divorce potential between those who lived together before engagement and those who did not cohabit at all before marriage remained significant (p < . 01) when controlling for education, length of marriage, and religiousness.

To test whether the main results would hold for individuals who cohabited only with their current spouses, we excluded individuals who had cohabited with someone else (remaining n = 718) and ran ANOVAs with satisfaction, negative communication, confidence, and dedication as dependent variables. The results of the overall F-tests indicated significant main effects of cohabitation history on all four variables (ps < .05) in the expected directions. Additionally, chi-square tests showed that among this subsample there was a significant difference (p < .01) in the expected direction between those who cohabited before engagement and not at all before marriage on self-reported divorce potential. Differences across the groups remained significant when controlling for marriage length, religiousness, and education for divorce potential and negative communication, though not for dedication, confidence, or satisfaction.

Discussion

With a random sample of individuals who were married within the last decade, the current study sought to replicate and extend previous findings that cohabiting before engagement is associated with marital dissatisfaction and instability (e.g., Kline et al., 2004). The basic findings are in line with previous research, as there were no significant differences between individuals who cohabited after engagement versus not until marriage, but those who cohabited before engagement reported significantly lower quality marriages and greater proneness for divorce than those who cohabited only after engagement or not at all until marriage. Most of these effects remained significant when controlling for length of marriage, religiousness, and education. We did not find significant differences based on cohabitation history for the level of friendship between partners or for satisfaction with the sexual/sensual relationship. Previous research has found that cohabiting individuals have lower sexual satisfaction than married individuals (e.g., Stanley et al., 2004), but to our knowledge, this study is the first to examine either of these constructs in relation to type of premarital cohabitation (before or after engagement).

Overall, the effect sizes in this more representative sample were smaller than what Kline et al. (2004) found with a convenience sample. There are several differences between the two studies that could account for the smaller effect sizes in the current study. First, Kline et al. used longer versions of several measures than were used in the current study. It could be that the full versions of these measures are more reliable and valid, leading to stronger results in Kline et al.’s study. Second, Kline et al. used a convenience sample of couples who were marrying within religious organizations in a single metropolitan area; the current study’s sample is much more generalizable, suggesting that in the broader populations, the pre-engagement cohabitation effect may be less pronounced. Third, the timeframes for data collection differed substantially. Kline et al. used data collected premaritally and 10 months into marriage whereas the current study used data collected six years into marriage, on average. Therefore, the current study excludes some couples who divorced early in their marriages whereas Kline et al. included all such couples. Nevertheless, this study highlights the importance of considering the timing of the decision to marry in understanding the cohabitation effect, especially because cohabitation is increasingly common. These effects may be small, but they are applicable to a substantial number of people.

Future research should distinguish pre-engagement cohabitation from general premarital cohabitation, as those who live together only after making a mutual, public commitment to marry may not be at greater risk for marital dissatisfaction or instability. To be clear, it is not that data show benefits of cohabiting after engagement relative to waiting until marriage, but the accumulating evidence shows some added risks for cohabiting before a mutual commitment to marriage. These findings are consistent with the theory that some cohabiting couples may go on to marry partly because of constraints associated with living together (e.g., tangible investments, social pressure; see Stanley et al., 2006), since settling on marriage during cohabitation is a risk factor for having more problems in marriage.

The results of the current study run counter to a previous conclusion that the cohabitation effect may be explained by cohabitations with multiple partners (Teachman, 2003). We found evidence that cohabiting before engagement, even only with one’s future spouse, is associated with more negative communication and higher divorce potential. This new finding is consistent with the perspective that the experience of cohabitation may add to the explanation of the cohabitation effect beyond the typical understanding of selection. If future research continues to support inertia theory, it would be valuable for practitioners and relationship education programs to review the potential risks associated with pre-engagement cohabitation and to help individuals evaluate possible costs and benefits of cohabiting before deciding to marry.

Although the current study provides important new information about premarital cohabitation in an up-to-date, random sample, it is not without limitations. As is true in all studies on the cohabitation effect, steadfast conclusions about causality cannot be drawn from these cross-sectional findings. We ruled out several possible confounding variables, but there may be other variables that account for the associations between pre-engagement cohabitation and marriage quality. Additionally, the sample is not a perfect representation of the United States population, as it was dependent on individuals having telephones and being willing to complete a telephone survey. Further, the effects here may be underestimated because some couples would have divorced already; prospective longitudinal studies are needed to better estimate the magnitude of the cohabitation effect across different stages of marital development. Lastly, given the nature of the data collection, the measures used in the study were necessarily brief and in some cases were single items. Estimates of their reliability are therefore limited. With these limitations in mind, the study shows that cohabiting before engagement is a risk factor for lower marital quality and divorce, suggesting that the premarital cohabitation effect could be more fundamentally cast as the pre-engagement cohabitation effect.