Abstract
Over half of young adults have cohabited, but relatively little is known about the role delinquency and substance use play in youths’ odds of cohabiting as well as the implications of cohabitation for early adult offending and substance use. This study focuses on the reciprocal relationship between cohabitation during late adolescence and young adulthood and self-reported offending and substance use. Using longitudinal data, we find that net of traditional predictors delinquency involvement is associated with increased odds of cohabitation and cohabiting at younger ages while substance use is not related to cohabiting during early adulthood. Further analysis indicates that cohabitation is associated with lower reports of substance use. However, cohabitation is not associated with self-reported offending. The results help to unravel the connection between cohabitation experience, offending and substance use, and early adult outcomes.
Keywords: offending, substance use, juvenile delinquency, cohabitation
Cohabitation has become a normative experience for many young adults; for example, two-fifths of young women have cohabited and nearly two-thirds of recent first marriages were preceded by cohabitation (Kennedy & Bumpass, 2008). As cohabitation has become more common, the median age at first marriage has increased to 25 for women and 27 for men (Goodwin et al., 2009). While such changes in union formation have been occurring, research on crime in adulthood continues to examine marriage as an important turning point in the life course, often referred to as the “good marriage effect” (Laub & Sampson, 2003). Few studies have specifically examined cohabitation and adult desistance from offending and substance use.
Drawing on longitudinal data from the Toledo Adolescent Relationships Study (TARS), we investigate the extent to which delinquency and substance use are related to cohabitation experience. We address two key questions: (1) does delinquency and substance use predict who is likely to cohabit; and (2) like marriage, is cohabitation tied to lower levels of offending and substance use in adulthood? This work contributes to both the family and delinquency literatures in at least two ways. First, prior studies have rarely considered the associations of early offending and substance use with subsequent cohabitation. Thus, a better understanding of who is likely to select into cohabitation may help to interpret the connection between cohabitation and well-being. Second, previous research on changes in offending and substance use in adulthood has focused almost exclusively on the prosocial effects of marriage when in fact cohabitation is the more common dyadic union among economically and socially disadvantaged men and women (Smock & Manning, 1997). Although a “good marriage effect” appears to exist, to date it seems unresolved whether cohabitation offers any of the same benefits as marriage in terms of desisting from crime or substance use in adulthood.
Background
Cherlin (2004) has argued that marriage has become “deinstitutionalized” in American society, citing the increasing prevalence of cohabiting unions as evidence. Early work by Bumpass (1998) documented a shift in what may be viewed as the conventional progression of courtship behaviors: dating, engagement, and marriage. Young adults are increasingly more likely to cohabit before getting married (Kennedy & Bumpass, 2008), with some of these unions being marriage-like and others appearing to be extensions of dating (Brown & Booth, 1996). These differences in the nature and meaning of cohabitation reflect the disparity in motives underlying cohabitation (Smock et al., 2006). Additionally, some young adults “slide” into cohabitation while others experience a more deliberate decision on the path to marriage (Smock, et al., 2005; Stanley et al., 2006). Consequently, cohabitation is not an indicator of strong, continuous commitment for all couples (Stanley et al., 2004). Our understanding of cohabitation will be expanded by further examining the characteristics of individuals who cohabit.
This paper draws on the union formation and adolescent risk behavior literatures to better understand young adults’ transition into cohabitation and the subsequent effects of cohabitation on well-being. Much literature has focused on how cohabitation influences the success of subsequent marriages, with premarital cohabitation often associated with low marital stability (Phillips & Sweeney, 2005; Smock, 2000; Stanley et al., 2006; Teachman, 2003). This is largely explained through selection arguments; that is, individuals who cohabit typically possess characteristics that are related to whether they cohabit, but these are the same characteristics that are linked to higher divorce rates. For example, Woods and Emery (2002) found that the negative effect of premarital cohabitation on marital stability is explained by several factors, including involvement in relatively severe forms of delinquency.
Few empirical studies have accounted for the potentially important influence of delinquency as a risk factor that may affect the experience and timing of cohabitation. As noted by Manning et al. (2007), delinquent youths were more likely to expect to cohabit than follow the traditional path toward marriage. Additionally, Yamaguchi and Kandel (1985) found that the use of marijuana and other illegal drugs was linked to a higher probability of premarital cohabitation. The majority of prior studies show the connection between economic disadvantage and cohabitation (Clarkberg, 1999; Smock et al., 2005), but to date no recent studies have examined whether delinquency and substance use influence cohabitation net of socioeconomic status.
At the same time, criminological studies treat romantic relationships (primarily marriage) during the transition to adulthood as an important factor related to adult desistance from crime. Laub and Sampson (2003) argued that a quality marriage increases informal social control as individuals enter adulthood and is associated with significant decreases in offending, that is, the “good marriage effect.” Although Laub and Sampson’s oft-cited results supported their hypothesis that marriage reduces crime, some have argued that the finding may be spurious insofar as there is differential selection into marriage and cohabitation (Gottfredson & Hirschi, 1990; see also Siennick & Osgood, 2008). Also, their earlier work focuses exclusively on marriage, and did not examine the association of cohabitation with later involvement in crime.
A significant limitation of Laub and Sampson’s analyses is that they focused on a cohort of men who matured into adulthood during the 1940s. Cohabitation was relatively quite rare during that era. For example, Sampson et al. (2006) indicated that cohabitation was reported in only 3% of the approximately 2,500 person-years of data collected retrospectively from 52 men (nearly 50 years of data per person) and found that cohabitation was related to decreased crime in adulthood. It is important, however, to continue this investigation with data obtained from a recent cohort, examining both men and women, and focusing on more contemporary union formation patterns.
Other studies utilizing more recently collected data have examined the implications of cohabitation on crime and substance use. While marriage had prosocial effects, cohabitation actually was related to increases in crime (Horney et al., 1995; Piquero et al., 2002). However, these samples were quite select and limited to incarcerated offenders (Horney et al., 1995) or parolees (Piquero et al., 2002). For substance use, research has shown that declines in alcohol, marijuana, and other illicit drugs followed (and in some cases preceded) the formation of cohabiting relationships (Duncan et al., 2006; Yamaguchi & Kandel, 1985). These studies are informative, but there are limitations. For example, some used data collected at a time (e.g., 1970s) when cohabitation was less normative (Yamaguchi & Kandel, 1985), and others relied on a time window of “change” that is quite long (11 years) (Duncan et al., 2006).
Current family trends, such as the rising age at first marriage and the high prevalence of premarital and nonmarital cohabitation (Kennedy & Bumpass, 2008), highlight the importance of investigating predictors of cohabitation and the timing of cohabitation using recently collected data. We move beyond prior work by addressing whether problem behaviors in adolescence are factors associated with entering into cohabiting unions while taking into account traditional covariates of union formation. Family background is included because children whose parents have cohabitated or divorced are more likely to cohabit (Axinn & Thornton, 1996; Sassler et al., 2009). Socioeconomic status is also recognized as an important factor in a person’s decision to cohabit; economic disadvantage has been linked with a greater likelihood of cohabiting (Clarkberg, 1999; Smock et al., 2005). We also include a measure of religiosity due to the finding that more religious people are less likely to cohabit than those who hold more “liberal” values (Thornton et al., 1992). Premarital pregnancy is another factor that has been linked to the decision to cohabit (Manning, 1993; Raley, 2001), and this is included in our models.
Furthermore, we assess whether cohabitation is associated with better (or worse) outcomes for young adults. Owing to the mixed findings in the literature, we approach analyzing the role played by cohabitation separately for offending and substance use. A contribution of this study is that it incorporates several important control variables previously shown to be related to crime and substance use. Measures of social learning such as friends’ delinquency involvement and substance use are among the strongest correlates identified in the criminological literature, and adolescents with delinquent peers are more likely to report delinquency involvement themselves (Haynie, 2002; Warr, 2002). Low self-control is also identified as an important factor linked with increased rates of offending (Pratt & Cullen, 2000; Vazsonyi & Klanjsek, 2008). Socioeconomic status is a significant predictor as well, and most theories of crime posit that disadvantaged youth are at an elevated risk for drug use and delinquency while other formulations argue the opposite (Agnew et al., 2008; Wright et al., 1999).
Methods
Data
The sample for this study is drawn from the Toledo Adolescent Relationships Study (TARS) that was derived from the enrollment records of students registered for the 7th, 9th, and 11th grades in Lucas County, Ohio, which is mostly composed of the city of Toledo. Devised by the National Opinion Research Center, a stratified random sample, that includes over-samples of African-American and Hispanic youths, was drawn from the records of 62 schools over seven school districts. Interviews were conducted at home with the help of laptop computers programmed with the survey questionnaire. The TARS data currently consist of four waves, following 1,321 adolescents since 2001. Follow-up interviews occurred at one, three, and five years in which 83% of the original sample were retained by the final wave.1 At the fourth wave, age, measured in years, ranges from 17 to 24 with a mean of 20. U.S. Census data indicate that our initial sample mirrors the characteristics of the Toledo, Ohio, MSA and the socio-demographic characteristics of the Toledo area closely parallel those of the nation in terms of education (80% in the Toledo MSA vs. 84% in the U.S. are high school graduates), median family income ($50,046 vs. $50,287), marital status (73.5% vs. 75.9% married two-parent households), and race (13% vs. 12% African-American).
The TARS data are well-suited for the current investigation and may be preferable over other data sources. For example, the National Longitudinal Study of Adolescent Health (Add Health) asked respondents at the third wave, when respondents ranged in age from 18 to 28, to retrospectively report if they have “ever lived with someone in a marriage-like relationship for one month or more” (emphasis added). Work by Manning and Smock (2005) suggests that this wording may not capture the full range of meanings attributed to living with another of the opposite sex, thereby underestimating cohabitation experience. The TARS data refer to cohabiting relationships as “[living] with a boy/girlfriend (not as a roommate) without being married” and most likely captures a more complete group of cohabitors. Additionally, the TARS sample at the fourth wave has a maximum age of 24 years, compared to 28 in Add Health, possibly reducing reporting error, particularly with the date of the start of one’s first cohabiting union. Last, school attendance was not a requirement for inclusion in the TARS sample because the original sample was drawn from school district enrollment records which include truants as well as those who may have dropped out. This is important when considering delinquency and substance use because samples administered during school hours may miss youths with the most serious and frequent involvement in deviant behaviors (Cernkovich et al., 1985).
The analytic sample draws on 1,028 respondents who were re-interviewed at each of the interview waves. The sample further excludes adolescent and young adults who are or have been married (n = 60) and who began cohabiting prior to the second interview (n = 34). We also omit 20 additional adolescents who were in the “other” race/ethnicity category. These omissions result in a final sample of 914 respondents.
Measures
Dependent variables
All dependent variables are measured at the final wave, and descriptive statistics (including scale reliability measured with Cronbach’s alpha) are reported in Table 1. Cohabitation experience is constructed from answers to the question: “Have you ever lived with a boy/girlfriend (not as a roommate) without being married?” Responses are coded 1 for yes and 0 for no. Age at first cohabitation is measured in years and was constructed by subtracting the century month of the respondent’s birth from the century month corresponding to the start of their first cohabiting relationship; this value was then divided by 12.
Table 1.
Descriptive Statistics (n=914)
| Mean | SD | Range | |
|---|---|---|---|
| Cohabitation Experience (W4) | .33 | — | 0–1 |
| Current Cohabitation (W4) | .18 | — | 0–1 |
| Age at First Cohabitation (W4) (n = 301) | 19.08 | 1.68 | 14–23 |
| Delinquency (W2) (α = .77) | .14 | .42 | 0–5.43 |
| Offending (W4) (α = .71) | .15 | .45 | 0–4.57 |
| Substance Use (W2) (α = .79) | 0.93 | 1.41 | 0–7.00 |
| Substance Use (W4) (α = .69) | 1.86 | 1.68 | 0–7.33 |
| Age (W4) | 20.17 | 1.74 | 17–24 |
| Female (W1) | .52 | — | 0–1 |
| Race/Ethnicity (W1) | |||
| White | .66 | — | 0–1 |
| African-American | .24 | — | 0–1 |
| Hispanic | .10 | — | 0–1 |
| Parents' Income (in thousands) (W1) | 61.16 | 35.70 | 0–160 |
| Mother's Education (W1) | |||
| Less than 12 years | .10 | — | 0–1 |
| 12 years | .35 | — | 0–1 |
| More than 12 years | .54 | — | 0–1 |
| Family Type (W1) | |||
| Married, biological parents | .54 | — | 0–1 |
| Cohabiting parent | .04 | — | 0–1 |
| Step parent | .12 | — | 0–1 |
| Single parent | .23 | — | 0–1 |
| Other | .08 | — | 0–1 |
| Currently in School (W3) | .79 | — | 0–1 |
| Employment Status (W3) | |||
| Unemployed | .50 | — | 0–1 |
| Part-time | .34 | — | 0–1 |
| Full-time | .17 | — | 0–1 |
| Adolescent's Income (W2) | 69.99 | 149.82 | 0–2500 |
| Religiosity (W2) (α = .72) | .00 | .88 | −1.50–1.37 |
| Been Pregnant/Gotten Someone Pregnant (W3) | .12 | — | 0–1 |
| Low Self-Control (W3) (α = .70) | 1.88 | .62 | 0–4 |
| Friends' Delinquency (W3) (α = .85) | .50 | .96 | 0–7.14 |
| Friends' Substance Use (W3) (α = .77) | 2.34 | 1.96 | 0–8 |
Offending is measured using a 7-item scale composed of the mean of reported frequencies of the following behaviors: “stolen (or tried to steal) things worth $5 or less”; “carried a hidden weapon other than a plain pocket knife”; “damaged or destroyed property on purpose”; “stolen (or tried to steal) things worth more than $50”; “attacked someone with the idea of seriously hurting him/her”; “sold drugs”; and “broken into a building or vehicle (or tried to break in) to steal something or just to look around.” Substance use is a scale of the mean frequency of alcohol use, public drunkenness, and drug use. Possible responses for each offending and substance use variable range from never (coded 0) to more than once a day (8). The behaviors included in the two measures are selected from the longer self-report scale by Elliott and Ageton (1980).
Independent variables
Delinquency involvement and substance use, measured at the second wave (mean age = 16) are constructed from scales identical to those comprising the offending and substance use measures discussed above. We also create a variable measuring current cohabitation indicating whether the respondent was cohabiting at the time of the wave 4 interview.2 The respondents who are not currently cohabiting (the reference category) are either single or dating and exclude those who have ever been married.
Controls
The following variables excepting age are constructed from data gathered at the initial interview. Age at the final wave is measured in years, gender is coded 1 for female and 0 for male, and race/ethnicity is composed of three dummy variables for White (reference category), African-American, and Hispanic. Parental income is in thousands of dollars, calculated by taking the sum of a parent’s response to their own as well as their partner’s income. We also include mother’s education with three dummy variables representing less than 12 years of education, 12 years (reference category), and more than 12 years. Family type is a series of dummy variables indicating the following types: married, biological parents (reference category), single parent, cohabiting parent, step-parent, and other.
The next several control variables were measured at the second wave. Adolescent’s income is the sum of responses to the following question: “In a typical week, how much money do you make from working?” and “How much money do you get from your parents?” Religiosity is a scale constructed from the mean of two standardized items: “How important is religion in your life?” (0 for not at all important to 4 for very important) and “In the past 12 months, how often did you attend religious services?” (0 for never to 3 for once a week or more).
Data from the third wave are used to construct the remaining control variables. Currently in school is coded 1 if respondents are currently attending an educational institution (or if interviewed during the summer, attended during the previous academic year) and 0 otherwise. Employment status consists of three dummy variables: unemployed, part-time (reference category), and full-time. A pregnancy measure is constructed from the question: “How many times have you gotten pregnant?” (male respondents were asked, “How many times have you gotten someone pregnant?”). All responses greater than or equal to one (responses ranged from 1 to 6 pregnancies) are coded as 1 and otherwise coded 0. Low self-control is measured using the mean response (0 for strongly disagree to 4 for strongly agree) to the following 6 statements concerning impulsivity, risk-taking, and self-centeredness (see Gottfredson & Hirschi, 1990): “When making a decision, I go with my ‘gut feeling’ and don't think much about the consequences of each alternative”; “I like to take risks”; “I live my life without much thought for the future”; “When nothing is happening I usually start looking for something exciting”; “I like it when people can do whatever they want, without strict rules and regulations”; and “Sometimes I think I am too self-centered.” Friends’ delinquency and friends’ substance use are obtained by asking the respondents about their friendship group’s delinquency involvement and substance use behaviors (e.g., “In the last 12 months, how often have your friends drunk alcohol?”). The scales include mean responses for the same behaviors as in self-reported offending, delinquency, and substance use.
Analytic Strategy
The first set of analyses examine the odds of cohabiting using binary logistic regression models for the full analytic sample (n = 914). Models are estimated separately for self-reported delinquency involvement and substance use as the focal predictors. Coefficients greater than one indicate greater odds of cohabiting and those less than one indicate lower odds. The odds ratio (OR) is computed by taking the exponential value of the coefficient. Ordinary least squares (OLS) regression is used to examine the age at first cohabitation among adolescents with a history of cohabitation experience (n = 301). The second set of analyses focus on if cohabitation is a significant correlate of offending and substance use. Because our goal is to determine whether cohabitation is related to lower involvement in deviant behaviors, these analyses are limited to respondents with a high level of early involvement. Thus, those who reported high rates (above average) of adolescent delinquency (n = 146) or substance use (n = 297) (second wave reports) were included in the analysis. Given the distributions of the dependent variables, tobit regression is employed for analyses of offending while OLS regression (coefficients are standardized) is used for estimates of substance use. Also, the extent to which these associations vary for male and female respondents is assessed by adding to each full model a multiplicative interaction term.
Results
Table 1 shows that 33% (n = 301) of the 914 respondents have ever cohabited and 18% (n = 165) reported cohabiting at the time of the final interview. Current cohabiting relationships make up 55% of those who have ever cohabited. These cohabitation levels are on par with national estimates for similar age groups. The measures of delinquency and offending are positively skewed with means of 0.14 and 0.15, respectively, indicating that many respondents report rarely engaging in delinquent acts (far less than one or two times per year). Substance use is slightly more frequent at both time points. The means are 0.93 (about once or twice a year) at the second wave and 1.86 (about once every two to three months) at the fourth. These low rates of involvement are not surprising since the TARS is school-based and not a high-risk sample.
Table 2 presents the association between self-reported delinquency and cohabiting using binary logistic regression models. At the zero-order the odds ratio for delinquency is 2.35 (p < 0.001), indicating that with each unit increase in delinquency respondents have 135% higher odds of cohabiting. After adding controls for demographic characteristics, socioeconomic indicators, religiosity, and pregnancy, delinquency remains significantly related to cohabitation experience (OR = 2.61; p < 0.001). The odds of cohabiting increase 161% with each additional unit of delinquency involvement.
Table 2.
Logistic Regression of Cohabitation Experience and OLS Regression of Age at First Cohabitation on Delinquency and Controls
| Cohabitation Experience (n=914) |
Age at First Cohabitation (n=301) |
|||
|---|---|---|---|---|
| Zero Orders | Full Model | Zero Orders | Full Model | |
| OR | OR | B | B | |
| Delinquencya | 2.35*** | 2.61*** | −.15** | −.12* |
| Agea | 1.44*** | 1.33*** | — | — |
| Female (reference=Male) | 1.87*** | 2.45*** | −.02 | −.05 |
| Race/Ethnicity (reference=White) | ||||
| African-American | 1.75*** | 1.06 | .11 | .15* |
| Hispanic | 2.04*** | 1.09 | −.03 | −.00 |
| Parents' Incomea | .99*** | .99* | .10 | .12 |
| Mother's Education (reference=12 years) | ||||
| Less than 12 years | 1.77* | 1.16 | .09 | .15* |
| More than 12 years | .57** | .76 | .13* | .10 |
| Family Type (reference=Married, biological parents) | ||||
| Single parent | 2.65*** | 1.59* | −.02 | .02 |
| Cohabiting parent | 6.06*** | 4.03** | −.09 | −.06 |
| Step parent | 2.39*** | 2.09** | −.09 | −.15* |
| Other parent | 4.47*** | 1.97* | −.08 | −.09 |
| Currently in School (reference=Not currently) | 4.57*** | 1.85** | .04 | .08 |
| Employment Status (reference=Part-time) | ||||
| Unemployed | .97 | .94 | −.29*** | −.34*** |
| Full-time | 2.38*** | 1.60 | −.09 | −.11 |
| Adolescents' Incomea | 1.00*** | 1.00 | .09 | .07 |
| Religiositya | .76*** | .83 | .07 | .02 |
| Been Pregnant/Gotten Someone Pregnant (reference=Never) | 7.15*** | 2.90*** | .02 | .01 |
p < .05;
p < .01;
p < .001
Variable is centered around its mean
Consistent with prior studies, the full model also indicates that older adolescents, female adolescents, and those respondents with parents reporting lower incomes are more likely to cohabit. Race/ethnicity in the full model is not significant and the zero order effect is fully explained by family structure. After adding to the model measures for parent’s income, family structure, and school status, mother’s education is longer significant. Family structure is associated with cohabitation with respondents raised outside of two biological parent families having higher odds of cohabiting in early adulthood. It is notable that having lived with a parent who cohabited is highly associated with experiencing cohabitation in early adulthood. Respondents’ socioeconomic circumstances matter as well. Those who are in school or employed full-time (compared to part-time) have higher odds of cohabiting. However, in the full model full-time employment is not significantly tied to cohabitation. Becoming pregnant or getting someone pregnant is significantly and positively related to the odds of cohabiting at the zero order and in the full model. The interaction of gender and delinquency is not statistically significant indicating that delinquency has a similar effect on the odds of cohabitation for male and female respondents (results not shown).
The next set of models show that delinquency involvement is significantly associated with younger ages at first cohabitation. As delinquency level increases, age at first cohabitation decreases by 0.15 (p < 0.01) in the zero-order model and by 0.12 (p < 0.05) in the full model with control variables. In addition, a gender and offending interaction term is not statistically significant, suggesting that the delinquency-age of cohabitation relationship is similar for male and female respondents (results not shown). The covariates significantly associated with age at cohabitation include race and ethnicity, mother’s education, family income, and employment status. The suppression observed from the zero order to the full model for African-American respondents seems to be explained by multiple factors, all reflecting socioeconomic status (i.e., mother’s education, parents’ and adolescent’s income, employment status, and currently in school).
Table 3 presents the results of models focusing on the association between substance use on cohabitation experience and age at first cohabitation. At the zero-order, substance use is significantly related to higher odds of cohabiting (OR = 1.15; p < 0.01). Each unit of substance use is associated with a 15% increase in the odds of cohabiting. However, the coefficient for substance use (OR = 0.97) is no longer significant with the inclusion of several controls, specifically gender and age. Older teens are more likely to cohabit and are also more frequently involved in substance use. Also, there is a gendered pattern of substance use with male respondents reporting higher use, but we find that substance use has a similar relationship with cohabitation for male and female respondents (results not shown). The covariates are similar in this model as Table 2 with the exception that religiosity is negatively associated and full-time employment (compared to part-time) is positively associated with cohabitation in the full model.
Table 3.
Logistic Regression of Cohabitation Experience and OLS Regression of Age at First Cohabitation on Substance Use and Controls
| Cohabitation Experience (n=914) |
Age at First Cohabitation (n=301) |
|||
|---|---|---|---|---|
| Zero Orders | Full Model | Zero Orders | Full Model | |
| OR | OR | B | B | |
| Substance Usea | 1.15** | .97 | .08 | .08 |
| Agea | 1.44*** | 1.33*** | — | — |
| Female (reference=Male) | 1.87*** | 2.18*** | −.02 | −.03 |
| Race/Ethnicity (reference=White) | ||||
| African-American | 1.75*** | 1.06 | .11 | .16* |
| Hispanic | 2.04*** | 1.17 | −.03 | −.01 |
| Parents' Incomea | .99*** | .99* | .10 | .12 |
| Mother’s Education (reference=12 years) | ||||
| Less than 12 years | 1.77* | 1.12 | .09 | .17* |
| More than 12 years | .57** | .77 | .13* | .11 |
| Family Type (reference=Married, biological parents) | ||||
| Single parent | 2.65*** | 1.63* | −.02 | .02 |
| Cohabiting parent | 6.06*** | 3.74** | −.09 | −.05 |
| Step parent | 2.39*** | 2.01** | −.09 | −.13* |
| Other parent | 4.47*** | 2.07* | −.08 | −.09 |
| Currently in School (reference=Not currently) | 4.57*** | 1.86** | .04 | .08 |
| Employment Status (reference=Part-time) | ||||
| Unemployed | .97 | .97 | −.29*** | −.34*** |
| Full-time | 2.38*** | 1.64* | −.09 | −.10 |
| Adolescents' Incomea | 1.00*** | 1.00 | .09 | .09 |
| Religiositya | .76*** | .80* | .07 | .05 |
| Been Pregnant/Gotten Someone Pregnant (reference=Never) | 7.15*** | 2.94*** | .02 | .01 |
p < .05;
p < .01;
p < .001
Variable is centered around its mean
The next two models indicate that substance use is not significantly related to age at cohabitation in the zero-order (B = 0.08) or in the full model (B = 0.08). A gender and substance use interaction term is not significant indicating that substance use has a similar relationship among male and female respondents (results not shown). The control variables’ coefficients are similar to those reported in Table 2.
Next we determine if cohabitation is a factor related to young adult offending among those adolescents who reported above average delinquency at the second wave. Among the high-rate offenders, one-quarter (24%) were cohabiting at the time of the last interview, and mean offending at the final wave is 0.45 with a standard deviation of 0.74.3 Table 4 presents the tobit regression estimates for self-reported offending. Both at the zero order (b = −0.10) and with controls added (b = −0.06), the coefficients for current cohabitation are nonsignificant, suggesting that respondents who are cohabiting share similar levels of offending as those who are not currently cohabiting. A nonsignificant interaction of gender and cohabitation indicates an invariant relationship for male and female respondents (results not shown). The factors associated with offending in the full model include friends’ delinquency and gender.
Table 4.
Tobit Regression of Offending and OLS Regression of Substance Use on Current Cohabitation and Controls
| Offending (n=146) |
Substance Use (n=297) |
|||
|---|---|---|---|---|
| Zero Orders | Full Model | Zero Orders | Full Model | |
| b | b | B | B | |
| Current Cohabitation (reference=Not current) | −.10 | −.06 | −.18** | −.13* |
| Agea | .07 | .01 | .15* | .11 |
| Female (reference=Male) | −.73** | −.48* | −.23*** | −.15** |
| Race/Ethnicity (reference=White) | ||||
| African-American | .31 | .30 | −.09 | −.07 |
| Hispanic | .22 | .40 | .02 | .06 |
| Mother's Education (reference=12 years) | ||||
| Less than 12 years | −.16 | −.27 | −.05 | −.02 |
| More than 12 years | .02 | .11 | .07 | −.00 |
| Currently in School (reference=Not currently) | .62* | .36 | −.02 | −.08 |
| Employment Status (reference=Part-time) | ||||
| Unemployed | .27 | −.12 | −.05 | −.03 |
| Full-time | .56 | .41 | .03 | −.01 |
| Low Self-Controla | .52** | .31 | .13* | .10 |
| Friends’ Delinquencya | .38*** | .36*** | — | — |
| Friends’ Substance Usea | — | — | .37*** | .31*** |
p < .05;
p < .01;
p < .001
Variable is centered around its mean
Table 4 also shows the regression of cohabitation on substance use among respondents who reported above average use at the second interview wave. Among respondents who reported higher rates of prior substance use, 25% are currently in a cohabiting relationship, and mean substance use by the fourth wave is 2.88 with a standard deviation of 1.65. At the zero-order, respondents who are cohabiting report significantly lower substance use (B = −0.18; p < 0.01) than those who are not cohabiting. The cohabitation coefficient remains statistically significant in the full model (B = −0.13; p < 0.05).4 This finding is notable given that the model includes key covariates such as self-control and friends’ substance use. In the full-model gender and friends’ substance use are significantly associated with substance use. Further analyses indicate that cohabitation has a similar negative association with substance use for male and female respondents (results not shown). Although for many adolescents drug and alcohol use is increasing during this time period, young adults in cohabiting relationships have lower substance use than their non-cohabiting counterparts, but this is not the case for offending behavior.
Our focus has been on the desistance process, and, because a certain level of prior offending is necessary to observe any possible downward change, the investigation was directed primarily at small groups of respondents with high-rates of delinquency and substance users. It is important, however, to know the implications of cohabiting for young adults with lower reported levels of involvement at the second wave. Thus, as a supplemental analysis we estimated the models in Table 4 with tobit regressions for 768 low-rate and non-offenders and 618 low-rate and non-substance users, separately (results not shown). Wave 4 current cohabitation is negatively related to offending (b = −0.42; p < 0.05) in adulthood while it is not related to substance use (b = −0.18; p > 0.05). Findings are similar when employing a measure of cohabitation at the third wave. The results indicate that for these low-rate/non-offenders, cohabitation is related to lower involvement at the final follow-up. This group of respondents started out with low mean delinquency, but their cohabitation experience is accompanied by a slight drop in offending across waves; non-cohabitors’ rates remain about the same as they transition to adulthood.
Discussion
Higher levels of delinquency, in contrast to substance use, are related to greater odds of cohabiting and doing so at younger ages. The results suggest that delinquent behaviors explain variation in cohabitation experience beyond that of traditional demographic and socioeconomic factors. Delinquent youth may establish cohabiting unions at earlier ages. Moreover, such unions may be linked with poorer economic well-being. Some may argue that delinquency is an attempt to be adult-like while at the same time being denied access to an adult status. In other words, cohabitation may represent an attempt to develop autonomy from one’s parents and live independently, but without some of the resources necessary to ensure stable relationships. Certainly, a subgroup of cohabiting young adults is not the most prosocial, and recognizing this may be important in considering the effects of cohabitation on adult well-being.
Among teens who initially reported high rates of substance use, those who cohabit report lower use in young adulthood compared to non-cohabitors, even after adding controls for low self-control and friends’ substance use, important social and personal factors. These findings are similar to those reported by Duncan et al. (2006) although the time window in our analyses is 4 years in contrast to 11 years in their work. Supplemental analyses of low-rate or non-substance users indicate that cohabitation does not accelerate substance use. Substance use is an activity that may be taking place with the cohabiting partner in the home, offering opportunities for actual regulation of partners’ behaviors such as drinking and smoking. Partners may also influence behaviors outside the home by controlling the amount of time spent hanging out with friends or going to bars and clubs.
The same pattern of results is not found for self-reported offending among the high-rate delinquents, however. The null finding regarding offending behaviors may be an artifact of being enmeshed in deviant social networks (Haynie, 2002; Lonardo et al., 2009). Highly delinquent youth may require more prosocial influence than merely moving in with a boyfriend or girlfriend. Our supplemental analysis involving the low-rate/non-offenders offers a prosocial view of cohabitation. Among low-rate/non-offenders, significantly lower involvement is reported by cohabitors than non-cohabitors. Therefore, cohabitation appears to have a positive impact on adult outcomes, depending on an individual’s history of delinquency and substance use.
Although the data used for this investigation are from a longitudinal study of adolescents and young adults, we cannot entirely rule out issues related to causality. It is indeed possible that changes in substance use had begun or had already occurred preceding the formation of the cohabiting unions; thus, our results must be interpreted with caution. However, supplemental analyses indicate that, like our contemporaneous measure of cohabitation experience, a measure of cohabitation prior to the final wave of data produces similar results for young adult offending and substance use.
Unlike prior studies of parolees or adult incarcerated offenders that find cohabitation is tied to increases in offending, our work on a more general population of young adults suggests that cohabitation does not have a negative influence on offending. It is possible that the cohabiting partner’s offending and substance use behaviors matter as well; that is, cohabiting with someone who is not involved in offending would influence desistance. A quality cohabiting union (high attachment and/or commitment) may also be of greater importance than cohabitation status alone. Studies investigating these two hypotheses are warranted.
It is important to note that our sample is limited to respondents in one metropolitan area. While its characteristics are similar to national estimates, further research should use data collected from other communities. The TARS sample is limited to respondents who were 17–24 so it reflects the experiences of a relatively young sample. The median age of first cohabitation for similarly aged women according to national data (National Survey of Family Growth) is 19 years. It is possible that the relationship between delinquency and cohabitation may dampen with an older sample when cohabitation is more normative and perhaps more akin to marriage. Additionally, our sample excluded married individuals, and with further data collection an examination of the joint role of cohabitation and marriage on crime in adulthood is possible. Another issue is that our analysis focuses on a single cohabitation experience. More disadvantaged adults experience serial cohabitation (multiple stints of cohabiting relationships) (Cohen & Manning, 2009), and this should be incorporated in analyses with larger samples. Future research could delve into the implications of cohabitation among delinquents in terms of quality and duration of the relationship.
The findings we have presented are an attempt to examine the complex relationships involving union formation, delinquency, offending, and substance use. Traditionally, family scholars focusing on cohabitation distinguish types of cohabitors based on marriage plans (Brown, 2000; Guzzo, 2009; Manning & Smock, 2005), but our work and that of others (Woods & Emery, 2002) indicate it may be important to distinguish cohabitors based on risk factors such as offending and substance use. One route into cohabitation is potentially less prosocial with adolescent delinquents choosing early cohabitation and perhaps prematurely assuming adult roles. The implications of cohabitation for children and adults most likely depend on the level of disadvantage and characteristics of individuals forming cohabiting unions. Finally, although we cannot make definitive conclusions about causal order given the nature of this investigation, our findings appear consistent with the notion of a “good cohabitation effect,” and future research should incorporate a measure of cohabitation experience and consider the importance of these romantic relationships in the desistance process.
Footnotes
This research was supported by grants from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (HD036223 and HD044206), the Department of Health and Human Services (5APRPA006009), and by the Center for Family and Demographic Research, Bowling Green State University, which has core funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R24HD050959-01). An earlier version of this paper was presented as “Delinquency, Substance Use, and Cohabitation Experience in Young Adulthood: Unraveling Selection and Influence Processes” at the 2009 annual meeting of the Population Association of America.
Logistic regression results (not shown) indicate that survey attrition (non-inclusion at the final wave) is not significantly associated with initial levels of delinquency involvement and substance use. The logit model included the following controls measured at the first wave: gender, race/ethnicity, age, family structure, and mother’s education.
The choice of constructing a measure of current cohabitation from the fourth wave of data was guided primarily by substantive and theoretical reasons. We argue that a contemporaneous measure of current cohabitation is useful because of the greater incidence of cohabitors at older ages. As with most studies in the area of romantic relationships, longitudinal associations are difficult to interpret due to the fact that couples often break up in the two year period between waves of data collection. Also, it may be more meaningful to use a measure of cohabitation that is as temporally close to the outcome as possible because any influence that may be occurring will happen during the relationship and not after the partners have separated. However, we estimate additional models with wave 3 cohabitation.
About 51% of the 146 delinquent adolescents reported no offending at the last wave. A logistic regression model predicting having no involvement compared to any involvement was estimated. Results (not shown) accord with the findings presented from the tobit model estimation.
To assess whether the results we obtain for the contemporaneous measure of current cohabitation are robust with respect to the timing of the cohabitation experience, we investigate the relationship between wave 3 cohabitation and offending and substance use at wave 4. Although the numbers of currently cohabiting respondents at the third wave for the sample of high-rate offenders (n = 15) and high-rate substance users (n = 27) are small, the results are similar to the ones obtained from the contemporaneous measure. Wave 3 cohabitation, net of controls, is significantly associated with substance use (OLS; B = −0.17; p < 0.01) and not significantly related to offending (tobit; b = −0.47; p > 0.05). These findings also do not differ for boys and girls. The longitudinal results add to our confidence that cohabitation is meaningful for substance use behaviors in young adulthood.
References
- Agnew Robert, Matthews Shelley Keith, Bucher Jacob, Welcher Adria N, Keyes Corey. Socioeconomic Status, Economic Problems, and Delinquency. Youth and Society. 2008;40(2):159–181. [Google Scholar]
- Axinn William G, Thornton Arland. The Influence of Parents’ Marital Dissolutions on Children's Attitudes toward Family Formation. Demography. 1996;33(1):66–81. [PubMed] [Google Scholar]
- Brown Susan L. Union Transitions among Cohabitors: The Significance of Relationship Assessments and Expectations. Journal of Marriage and the Family. 2000;62(3):833–846. [Google Scholar]
- Brown Susan L, Booth Alan. Cohabitation versus Marriage: A Comparison of Relationship Quality. Journal of Marriage and the Family. 1996;58(3):668–678. [Google Scholar]
- Bumpass Larry L. The Changing Significance of Marriage in the United States. In: Mason Karen Oppenheim, Nilsuya Noriko, Choe Minja., editors. The Changing Family in Comparative Perspective: Asia and the United States. Honolulu, HI: East-West Center; 1998. pp. 63–79. [Google Scholar]
- Cernkovich Stephen A, Giordano Peggy C, Pugh Meredith D. Chronic Offenders: The Missing Cases in Self-Report Delinquency Research. Journal of Criminal Law and Criminology. 1985;76(3):705–732. [Google Scholar]
- Cherlin Andrew J. The Deinstitutionalization of American Marriage. Journal of Marriage and Family. 2004;66(4):848–861. [Google Scholar]
- Clarkberg Marin. The Price of Partnering: The Role of Economic Well-Being in Young Adults’ First Union Experiences. Social Forces. 1999;77(3):945–968. [Google Scholar]
- Cohen Jessica A, Manning Wendy D. Working Paper No. 2009-04, Center for Family and Demographic Research. Bowling Green, OH: Bowling Green State University; 2009. Estimates and Correlates of Serial Cohabitation. [Google Scholar]
- Duncan Greg J, Wilkerson Bessie, England Paula. Cleaning Up Their Act: The Effects of Marriage and Cohabitation on Licit and Illicit Drug Use. Demography. 2006;43(4):691–710. doi: 10.1353/dem.2006.0032. [DOI] [PubMed] [Google Scholar]
- Elliott Delbert S, Ageton Suzanna S. Reconciling Race and Class Differences in Self-Reported and Official Estimates of Delinquency. American Sociological Review. 1980;45(1):95–110. [Google Scholar]
- Goodwin Paula, McGill Brittany, Chandra Anjani. NCHS Data Brief, No. 19. Hyattsville, MD: National Center for Health Statistics; 2009. Who Marries and When? Age at First Marriage in the United States, 2002. [PubMed] [Google Scholar]
- Gottfredson Michael R, Hirschi Travis. A General Theory of Crime. Stanford, CA: Stanford University; 1990. [Google Scholar]
- Guzzo Karen Benjamin. Marital Intentions and the Stability of First Cohabitations. Journal of Family Issues. 2009;30(2):179–205. [Google Scholar]
- Haynie Dana L. Friendship Networks and Delinquency: The Relative Nature of Peer Delinquency. Journal of Quantitative Criminology. 2002;18(2):99–134. [Google Scholar]
- Horney Julie, Osgood D Wayne, Marshall Ineke Haen. Criminal Careers in the Short-Term: Intra-Individual Variability in Crime and Its Relation to Local Life Circumstances. American Sociological Review. 1995;60(5):655–673. [Google Scholar]
- Kennedy Sheela, Bumpass Larry L. Cohabitation and Children's Living Arrangements: New Estimates from the United States. Demographic Research. 2008;19:1663–1692. doi: 10.4054/demres.2008.19.47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Laub John H, Sampson Robert J. Shared Beginnings, Divergent Lives: Delinquent Boys to Age 70. Cambridge, MA: Harvard University; 2003. [Google Scholar]
- Lonardo Robert A, Giordano Peggy C, Longmore Monica A, Manning Wendy D. Parents, Friends, and Romantic Partners: Enmeshment in Deviant Networks and Adolescent Delinquency Involvement. Journal of Youth and Adolescence. 2009;38(3):367–383. doi: 10.1007/s10964-008-9333-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Manning Wendy D. Marriage and Cohabitation Following Premarital Conception. Journal of Marriage and the Family. 1993;55(4):839–850. [Google Scholar]
- Manning Wendy D, Longmore Monica A, Giordano Peggy C. The Changing Institution of Marriage: Adolescents Expectations to Cohabit and Marry. Journal of Marriage and Family. 2007;69(3):559–575. [Google Scholar]
- Manning Wendy D, Smock Pamela J. Measuring and Modeling Cohabitation: New Perspectives from Qualitative Data. Journal of Marriage and Family. 2005;67(4):989–1002. [Google Scholar]
- Phillips Julie A, Sweeney Megan M. Premarital Cohabitation and Marital Disruption among White, Black, and Mexican American Women. Journal of Marriage and Family. 2005;67(2):295–313. [Google Scholar]
- Piquero Alex R, MacDonald John M, Parker Karen F. Race, Local Life Circumstances, and Criminal Activity. Social Science Quarterly. 2002;83(3):654–670. [Google Scholar]
- Pratt Travis C, Cullen Francis T. The Empirical Status of Gottfredson and Hirschi’s General Theory of Crime: A Meta-Analysis. Criminology. 2000;38(3):931–964. [Google Scholar]
- Raley R Kelly. Increasing Fertility in Cohabiting Unions: Evidence for the Second Demographic Transition in the United States. Demography. 2001;38(1):59–66. doi: 10.1353/dem.2001.0008. [DOI] [PubMed] [Google Scholar]
- Sampson Robert J, Laub John H, Wimer Christopher. Does Marriage Reduce Crime? A Counterfactual Approach to Within-Individual Causal Effects. Criminology. 2006;44(3):465–508. [Google Scholar]
- Sassler Sharon, Cunningham Anna, Lichter Daniel T. Intergenerational Patterns of Union Formation and Relationship Quality. Journal of Family Issues. 2009;30(6):757–786. [Google Scholar]
- Siennick Sonja E, Osgood D Wayne. A Review of Research on the Impact on Crime of Transitions to Adult Roles. In: Liberman Akiva M., editor. The Long View of Crime: A Synthesis of Longitudinal Research. New York: Springer; 2008. pp. 161–187. [Google Scholar]
- Smock Pamela J. Cohabitation in the United States: An Appraisal of Research Themes, Findings, and Implications. Annual Review of Sociology. 2000;26:1–20. [Google Scholar]
- Smock Pamela J, Huang Penelope, Manning Wendy D, Bergstrom Cara A. Working Paper No. 2006–10, Center for Family and Demographic Research. Bowling Green, OH: Bowling Green State University; 2006. Heterosexual Cohabitation in the United States: Motives for Living Together among Young Men and Women. [Google Scholar]
- Smock Pamela J, Manning Wendy D. Cohabiting Partners’ Economic Circumstances and Marriage. Demography. 1997;34(3):331–341. [PubMed] [Google Scholar]
- Smock Pamela J, Manning Wendy D, Porter Meredith J. Everything’s There except Money: How Economic Factors Shape the Decision to Marry among Cohabiting Couples. Journal of Marriage and Family. 2005;67(3):680–696. [Google Scholar]
- Stanley Scott M, Rhoades Galena Kline, Markman Howard J. Sliding versus Deciding: Inertia and the Premarital Cohabitation Effect. Family Relations. 2006;55(4):499–509. [Google Scholar]
- Stanley Scott M, Whitton Sarah W, Markman Howard J. Maybe I Do: Interpersonal Commitment and Premarital or Nonmarital Cohabitation. Journal of Family Issues. 2004;25(4):496–519. [Google Scholar]
- Teachman Jay. Premarital Sex, Premarital Cohabitation, and the Rise of Subsequent Marital Dissolution among Women. Journal of Marriage and Family. 2003;65(2):444–455. [Google Scholar]
- Thornton Arland, Axinn William G, Hill Daniel H. Reciprocal Effects of Religiosity, Cohabitation, and Marriage. American Journal of Sociology. 1992;98(3):628–651. [Google Scholar]
- Vazsonyi Alexander T, Klanjsek Rudi. Test of Self-Control Theory across Different Socioeconomic Strata. Justice Quarterly. 2008;25(1):101–131. [Google Scholar]
- Warr Mark. Companions in Crime: The Social Aspects of Criminal Conduct. New York: Cambridge University; 2002. [Google Scholar]
- Woods LaKeesha N, Emery Robert E. The Cohabitation Effect on Divorce: Causation or Selection? Journal of Divorce and Remarriage. 2002;37(3/4):101–122. [Google Scholar]
- Wright Bradley R Entner, Caspi Avshalom, Moffitt Terrie E, Miech Richard A, Silva Phil A. Reconsidering the Relationship between SES and Delinquency: Causation but Not Correlation. Criminology. 1999;37(1):175–194. [Google Scholar]
- Yamaguchi Kazuo, Kandel Denise B. Dynamic Relationships between Premarital Cohabitation and Illicit Drug Use: An Event-History Analysis of Role Selection and Role Socialization. American Sociological Review. 1985;50(4):530–546. [Google Scholar]