Splitting up or getting married? Competing risk analysis of transitions among cohabiting couples in Sweden

(1)

Tilburg University

Splitting up or getting married? Competing risk analysis of transitions among

cohabiting couples in Sweden

Moors, G.B.D.; Bernhardt, E.

Published in: Acta Sociologica Publication date: 2009 Document Version

Publisher's PDF, also known as Version of record Link to publication in Tilburg University Research Portal

Citation for published version (APA):

Moors, G. B. D., & Bernhardt, E. (2009). Splitting up or getting married? Competing risk analysis of transitions among cohabiting couples in Sweden. Acta Sociologica, 52(3), 227-247.

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal

Take down policy

(2)

http://asj.sagepub.com

Acta Sociologica

DOI: 10.1177/0001699309339800

2009; 52; 227

Acta Sociologica

Guy Moors and Eva Bernhardt

Cohabiting Couples in Sweden

Splitting Up or Getting Married?: Competing Risk Analysis of Transitions Among

http://asj.sagepub.com/cgi/content/abstract/52/3/227

The online version of this article can be found at:

Published by:

http://www.sagepublications.com

On behalf of:

Nordic Sociological Association

can be found at: Acta Sociologica

Additional services and information for

(3)

Splitting Up or Getting Married?

Competing Risk Analysis of Transitions Among Cohabiting

Couples in Sweden

Guy Moors

Department of Methodology and Statistics, Tilburg University, Tilburg, The Netherlands

Eva Bernhardt

Department of Sociology, Stockholm University, Stockholm, Sweden

abstract: In this article, we investigate which ideational variables influence the propensity of cohabiting couples to transform their union into marriage, separa-tion or continued cohabitasepara-tion. The quessepara-tion is particularly relevant in the Swedish context of considerable social acceptance of unmarried cohabitation even among parents. A two-wave panel study including 705 never-married respondents co-habiting at the time of the first survey shows that ideational factors influence subse-quent behaviour, even when different sets of control variables are included in the model. Familistic attitudes, work-related values and reflections about the quality of the relationship prove to be predictors of the transition to either marriage or separa-tion even when intensepara-tions are taken into account.

keywords: attitudes ◆cohabitation ◆evaluations ◆intentions ◆marriage ◆ reasoned action ◆second demographic transition ◆separation

Introduction

Cohabitation seems to be a rising phenomenon throughout the Western world (Bernhardt, 2003). It is regarded as a crucial part of the transformation of Western family patterns, and has been called the second demographic transition (Lesthaeghe, 1995). The Scandinavian countries have the highest levels of cohabitation. In Sweden, it is rare for a marriage to start directly, without previous cohabitation – well over 90 per cent of first co-residential unions in Sweden begin before marriage (Statistics Sweden, 1995). Still, most people do get married eventually, many young people regarding marriage as a positive option in their lives (Bernhardt, 2002).

However, cohabiting unions are more fragile than marriages (for evidence for Sweden, see Hoem and Hoem, 1986; Gähler et al., 2009). While some cohabiting couples are ready to move on to a more committed relationship after some time of cohabitation, others find that they want to terminate their co-residential relationship and return to single life. In this article, we investigate – in the Swedish context of considerable social acceptance of cohabitation, even when the couple have become parents – which couples transform their cohabiting relation-ships into marriage and which couples separate. In particular, we are interested in whether, and in what way, attitudes, opinions about the quality of the relationship and intentions

(4)

contribute to explaining the transition from cohabitation to marriage or separation, as well as the extent to which the effects are the opposite for the two transitions, i.e. having a positive effect on the likelihood of marriage, but a negative effect on separation risks, or vice versa. A two-wave panel study among young adults in Sweden is used to estimate the effect of attitudes and socio-demographic characteristics on the likelihood of marriage versus separation.

Attitudes and living arrangements

There is a vast literature on the factors that influence the transition to marriage (see, e.g., Goldscheider and Waite, 1986; Cooney and Hogan, 1991; Axinn and Thornton, 1992; Cherlin, 1992; Goldstein and Kenney, 2001; Sweeney, 2002). Most of this literature, however, deals either specifically with the transition to marriage among unmarried men and women (disregarding cohabitation), or the choice between cohabitation and marriage when entering the first co-residential union (Liefbroer, 1991; Blom, 1994; Clarkberg et al., 1995). This article is about transitions that cohabiting couples experience after having established their relationship. Cohabiting couples obviously have the choice of separation, marriage or continued unmarried cohabitation. Literature focusing on this topic is much sparser, although rapidly growing in recent years (see, e.g., Smock and Manning, 1997; Duvander, 1999; Brown, 2000; Lichter et al., 2006). The present research contributes to that literature. Unlike some earlier studies, we are able to take into account both possible transitions, namely marriage and union disruption.

The relevance of including ideational factors in explaining union outcomes

for cohabitants

The question can be raised about what ‘benefits’ or ‘costs’ cohabiting couples perceive in trans-forming an intimate relationship into marriage. After all, when couples live together without being married, they have already completed the first step of the searching and matching process, and they have already obtained (or have the possibility to obtain) benefits such as income pooling and increasing returns to scale (Cherlin, 2000). There is little reason to believe that social stigma attached to living with a partner without being married would play an important role, except among conservative religious groups (Axinn and Thornton, 2000). This is even more the case in Sweden, where unmarried cohabitation became common as long ago as the 1970s, and where more children are born to cohabiting than to married parents (Bernhardt, 2004). However, the fact that Swedish cohabiting couples continue to get married demonstrates that for some reason marriage is still the preferred form of living together with a partner, even in a society where unmarried cohabitation is completely socially acceptable – except to a small minority of the population (Duvander, 1999). Apparently, cohabitation as a long-term arrangement is still seen by many young Swedes as inferior to marriage.1 _{It is}

therefore relevant to ask the question concerning what motivates cohabiting couples to marry, i.e. where attitudes, evaluations and intentions enter the picture.

(5)

perceptions or evaluations of their relationships. The analysis showed that conservative attitudes towards family issues had an indirect effect through relationship quality, but also a strong direct (negative) effect on disruption net of the other variables included in the analysis. The perceived relative benefits of a relationship compared to alternatives also had a very large effect: the rate of separation among those who thought their life would be better after separ-ation was three times that of those who felt it would be much worse. This indicates that sep-aration should be viewed as a process, and that panel data can be used to identify those who are in this process before the separation actually occurs.

Theoretical perspectives on attitudes, evaluations and intentions

From a theoretical point of view, perspectives on what links attitudes or values to demographic behaviour differ. In this section we focus on four issues: (1) the reciprocity of the attitude–beha-viour relationship; (2) the relevance of ideational theories compared to other types of explan-ation, (3) the attitudes that can be linked to union transitions, and (4) the role of intentions and evaluations.

First of all, it has to be emphasized that the attitude–behaviour relationship is reciprocal (Moors, 1997, 2000). Strong positive attitudes towards marriage, for instance, may increase the probability of marriage. In turn, the experience of marriage may affect the attitudes of couples by reinforcing them (becoming more positive) or altering them (adopting more negative views on marital life). Oppenheimer in particular (1994) was keen on arguing that the latter relationship was more often documented in research than the former argument about the effect of attitudes on family transitions. Measurements of attitudes before family transitions have occurred are necessary (Lesthaeghe and Moors, 2002). The data set used in this article meets this requirement. Attitudes as well as intentions and evaluations were measured in the first interview and changes in family situation are observed for more than four years afterwards.

A second issue regarding the role of attitudes and values in explaining family transitions is how it compares to other explanations. Most researchers seem to agree that ideational theories

complement explanations that refer to socio-economic and historical conditions as factors

influ-encing behaviour (Pollak and Watkins, 1993; Lesthaeghe, 1998; Lesthaeghe and Moors, 2002). The concept of the second demographic transition (Lesthaeghe, 1995), for instance, makes reference to the interdependence of socio-economic, demographic and cultural changes since the 1960s and links individual autonomy in ethical, religious and political domains to family formation in Western nations. The key argument is that culture in general, and values or attitudes in particular, at least partially guide the choices people make regarding their family formation. In this article we share this perspective and research whether attitudes and values influence the transition to marriage or separation independently of socio-demographic and socio-economic characteristics.

(6)

and independence that fosters new types of family life. The attitude scales used in this article refer to attitudes regarding family and work. In the next section we provide details about these scales and find that one ‘familistic attitude’ scale could be operationalized, one that stresses the importance of family life in the lives of people (a). Work-related attitudes unfolded into three scales measuring the importance of: (b) ‘socio-economic success’, (c) ‘socio-economic security’ and (d) ‘social engagement’. Given the two theoretical perspectives on attitudes in this section, four corresponding general hypotheses that guided our research can be formu-lated as follows:

(a) Positive/negative attitudes toward marriage and family life increase/decrease the likeli-hood of marriage among cohabiting couples. The theory of Reasoned Action and Planned Behaviour as well as the second demographic transition theory support this hypothesis. For the same reason we expect that a traditional family orientation would decrease the likelihood of separation because this is an undesirable option from a traditional family perspective.

(b) Work attitudes such as ‘being successful in work’, ‘possibilities to think and act independ-ently’ or ‘being proud of one’s work’ reflect how important ‘socio-economic success’ is to the respondents. It is hypothesized that these attitudes increase the likelihood of marriage as well as the likelihood of relatively short cohabitational unions. In Oppenheimer’s view (1994), cohabitation functions as an intermediate state (and not a preferred end-state) in the process of establishing a quality match in partnership. When a quality match is achieved, a marriage becomes more likely. From this perspective, socio-economic success is a prerequisite for marriage. If, however, a quality match is not established, a separation becomes more likely since stressing the importance of ‘socio-economic’ success reflects an individualistic perspective and individualism reduces union solidarity, i.e. separation is not an option that should be avoided by all means. Note that these arguments refer to

attitudes about socio-economic success and not about being economically successful. The

latter is a socio-economic covariate. Furthermore, these attitudes about socio-economic success should not be confused with materialistic or socio-economic security. Our analyses revealed these to be two distinct concepts.

(c) Materialism and socio-economic security issues have a central role in the second demo-graphic transition theory. If a person expresses non-materialistic concerns it is hypothe-sized that he or she will be less likely to opt for marriage. Vice versa, materialism increases the chance of a marriage. It is less clear what effect these materialistic attitudes have on the likelihood of separation. Materialism and social security attitudes in this research refer to issues such as the importance of ‘doing well economically’, ‘having a good salary’, ‘secure employment’ or ‘having good working hours’.

(d) The second demographic transition theory also links ‘new’ living arrangements to ‘new left’ attitudes that combine personal autonomy with social engagement. As such, we assume that work attitudes reflecting the need of social commitment, e.g. ‘helping people’, having a job that is ‘useful for society’, will increase the likelihood that cohabitation is a preferred end-state, and, hence, that the likelihood of marriage and separation will be lower.

Fourth, and finally, in addition to these attitudes, other ‘ideational’2_{factors, i.e. evaluations}

(7)

again, i.e. whether their general well-being and standard of living would improve after union dissolution. The general hypothesis applying to these ‘quality indicators’ is that:

(e) The likelihood of separation will increase, and marriage will decrease, when the quality of the current union is evaluated as negative, and being single again is evaluated positively.

Given that the quality of the relationship indicators is measured simultaneously with the attitudes and values, it remains unclear whether attitudes influence this evaluation or rather the other way round. For this reason, we have included the evaluations of the relationship next to the attitudes in our analyses.

Perceptions of the relationship also find their expression in intentions. We have included intentions in the final step of the analysis for two particular reasons. First, we hypothesize that an independent effect of attitudes on behaviour will be observed even after including inten-tions. A second reason to present models with and without intentions was that, from a socio-logical point of view, intentions can be defined as proxies of future anticipated behaviour, and hence as alternative dependent variables. Adherents of such a perspective would exclude intentions by arguing that the question is tautologous. Our analyses present models with and without intentions as covariates and, as such, avoid taking a stand on this epistemological debate on whether or not intentions should be included as mediators. Note also that we do not formally address the question of mediating effects in our analyses. Our key research question merely refers to the independent effect of attitudes on behaviour; hence, other explanatory variables only enter the model as control variables. Furthermore, there is a technical reason that prevents us from formally testing mediation effects. All established methods for mediation testing (Baron and Kenny, 1986) assume continuous measurements of the key variables, and especially the dependent variable. In this research the dependent variable is multinomial and indirect effects cannot be calculated (Hellevik, 2007). Even comparing estimates before and after controlling for other covariates should be done with care because adding covariates that are independent from the covariates already in the model does change the estimates of these latter covariates when multinomial logit regression models are estimated.

In this research, we included two intention variables, i.e. the ‘intention to marry’ and the ‘intention to have children’. The former is an obvious choice, whereas the latter is less directly related to the outcomes of our research. However, marriage is not necessarily a goal in itself, nor need it be an isolated event. Rather, marriage and parenthood are linked events. Hence, a desire to become a parent may induce a desire to marry. For this reason we hypothesize that:

(f) The intention to marry and the intention to have children will increase the likelihood of marriage and decrease the likelihood of separation. However, we also expect that attitudes remain significant predictors of the transition to marriage or separation.

The main reason why we expect that attitudes will have an independent effect on behaviour is, it has been argued (Rokeach, 1973), that general attitudes or values are far more stable characteristics of individuals than intentions, and hence that attitudes are more solid predic-tors of long-term behaviours. Empirical evidence seems to support this argument, since it has been demonstrated that intentions work well in explaining short-term transitions but less well in explaining long-term transitions (Liska, 1984; Liefbroer et al., 1994). In line with these arguments is Bumpass’s finding (2002) that conservative familistic attitudes had a direct negative effect net on other variables included in the analysis, such as reflections on the quality of the relationship.

(8)

effect for apparently ‘obvious’ explanatory variables such as intentions and evaluations of the relationship.

Other explanatory variables

The main focus of this article is the impact of attitudes on marriage versus separation among cohabiting couples in Sweden. However, there are also other explanations that might change the relationship between attitudes and family transitions. Hence, controlling for other covariates is necessary. Very few people would doubt that the couple’s socio-economic characteristics have an impact. Socio-economic security (employment status) and education are key explana-tory variables in economic theories of demographic transitions (Easterlin, 1980; Becker, 1981). A second important set of indicators is defined by the socio-demographic characteristics or cycle experiences of the couple. The effect of attitudes may depend on the period in the life-cycle in which they are expressed or measured. Hence, life-life-cycle characteristics that should be included are the age of the respondent, the length of the current relationship, age at leaving home and parental status. Research on childlessness has repeatedly demonstrated that ‘post-ponement for a definite time’ may in the long run turn into ‘post‘post-ponement for an indefinite time’ (Veevers, 1980). The same rationale may apply to marriage; we might therefore expect that the likelihood to marry increases with age and length of the union, but that there will be a turning point in the life-cycle in which this relationship is the reverse.

It is unclear what to expect regarding the presence of children in the household. For child-less couples, for instance, one might expect that the desire to become parents may foster the idea of marriage. If there are already children in the household, however, there may be less incentive to marry for some respondents, whereas others may be more focused on expressing their commitment to one another by marriage. Whether there were children in the household at the time of the first wave of the survey was included in the preliminary analysis, but later excluded as this was not found to have a significant effect. However, when we distinguished childless couples in their intention to become parents, we observed that the majority expressed such an intention while a minority of about 18 per cent either said they wanted to remain childless or expressed hesitation about parenthood.3 _{Childless couples who expressed their}

intention to become parents behaved in a similar way to couples with children in the house-hold, i.e. they were more likely to marry and less likely to separate than childless couples who did not have any such plans. This indicated that ‘absence of birth plans’ is the key issue in estimating the effect of ‘children’ on the likelihood of marriage or separation after cohabita-tion. As mentioned before, we classify this indicator in the category of ‘intentions’.

Religiosity – or rather secularization – is regarded as an important cultural explanatory factor of recent demographic changes in the second demographic transition theory (Lesthaeghe, 1995). Of course, the second demographic transition refers to aggregate changes, but if the argument regarding the effect of secularization holds at the individual level, religiosity of the parents should be included together with religiosity of the respondent. Other parental background or early socialization characteristics should be included as well. The most comprehensive body of evidence regarding the importance of including these characteristics comes from the research conducted by Thornton and colleagues (e.g. Thornton, 1991; Barber and Axinn 1998; Barber et al., 2002). Their data set is – to the best of our knowledge – unique, since it includes attitudes from the parents (mothers) measured when they were young adults. Our data are less advanced, but we have information regarding socio-economic and socio-demographic characteristics of the parents, i.e. educational level, whether the mother was working outside the home when the respondent was a child, the economic situation of the parents when growing up, whether the respondent has an intact family background and the number of siblings.

(9)

Data and methodology

Data

If estimating the effect of attitudes on the transition to marriage or separation, attitudes should be measured before this transition is observed. The panel data analysed in this article fit this requirement. The first wave of the survey ‘Family and Working Life in the 21st Century’ was undertaken in the spring of 1999 and the second in the spring of 2003. Data collection was by Statistics Sweden. A nationally representative sample of young adults with two Swedish-born parents was drawn, and 3,408 individuals were asked to fill in a mail questionnaire with questions about their plans, expectations and attitudes regarding family and working life. Factual information about their current situation and background characteristics was included. In addition, information such as the respondent’s education was taken from registers. The response rate was 66.7 per cent; thus, 2,273 respondents returned their questionnaires. Details about the project are available at www.suda.su.se/yaps.

For the purpose of this article, we selected a subsample of 914 never-married respondents who were cohabiting at the time of the first survey. For 899 of these respondents, we had full information about the timing of their relationship at the first interview and 705 (78.4 per cent) of them participated in the second wave of the survey. We checked the selectivity of the panel attrition by comparing compositional and attitudinal differences between participants and drop-outs.4_{In most cases, there were no significant differences. However, participation in the}

second survey was significantly higher among women (exp(b) = 2.44) than among men; and also increased with ‘age at leaving the parental home’ for the first time (exp(b) = 1.18). As far as attitudes, evaluations and intentions were concerned, we found no significant differences between participants and drop-outs. Only ‘familistic attitudes’ were slightly higher among participants (exp(b) = 2.65; p-value = 0.09) than among drop-outs. The minor differences between participants and drop-outs suggest a low potential for bias due to panel attrition.

Dependent variable and method

The dependent variable distinguishes between 113 separations, 168 marriages and 414 respon-dents who were still cohabiting with the same partner 4 years after the first interview. We emphasize this finding, since it highlights the particularity of the Swedish context. In the United States, for instance, one would expect well over half of cohabiting couples to have separated because cohabitations tend to be short-lived. In our sample, nearly 60 per cent indicated that they were still living with their partner without being married, 24 per cent had married their partner and only about 16 per cent had separated in this 4-year period.

The duration (time) variable is expressed in months since the first round of the survey in 1999. All respondents thus enter the risk set at the time of the first survey in 1999. There are several reasons why this date of entry is chosen. First of all, we selected never-married respon-dents who were cohabiting at that particular time. Second, and more importantly, the key question of this article is whether attitudes influence the two ways of leaving a cohabiting rela-tionship, i.e. marriage or separation. These attitudes are measured at the first round of the survey and, consequently, this date is the most appropriate. Of course a second time axis, i.e. the duration of the cohabitation, is important as well; hence, we followed Yamaguchi’s advice (1991) to include this variable as a time-varying covariate. Note that this variable refers to the ageing of the relationship rather than a change in status. The data are organized in a person-month file (Allison, 1984). The first person-month or date of entry into the risk set is the date of the first interview, and each respondent contributes as many months of observation as he or she is at risk of experiencing the event, i.e. marriage, separation or end of study. All models are estimated by means of a discrete-time multinomial logistic regression analysis treating marriage and separation as competing risks.5

(10)

Attitudes

The procedure to construct attitudinal scales may be summarized as follows: the scales are the outcome of exploratory principal components analysis, which is a statistical tool that reduces the number of attitudinal items to a meaningful limited number of attitudinal dimensions or scales. Attitudinal scales were computed as composite scales of the items that identified partic-ular dimensions and scale values were re-scaled to a range from 0 to 1. Details about the scales are presented in the Appendix. We present these attitude scales, the quality of the relationship indicators and the intention variables in Table 1.

Control variables: socio-demographic characteristics

Covariates that (might) explain the transition to marriage or separation, and at the same time (might) influence the relationship between attitudes and either one of these two transitions, can be classified into different sets. We distinguish among socio-economic and socio-demographic characteristics of the respondents and the partner, parental background information, early childhood indicators and religiosity of the couple and the respondent’s parents. A number of these theoretically relevant control variables are not included in the models that we discuss in the next section. Although it is tempting to include all possible covariates in a model from a theoretical point of view, we should be prudent to do so in this research. The main reason is that including control variables not related to the dependent variable may bias the estimates of other variables, which is an artefact of the logistic regression technique (Hellevik, 2007). Furthermore, our hypotheses refer to the effect of ideational factors on marriage or separation. We take little substantive interest in the effects of these control variables as such. To prevent inclusion of control variables that are unrelated to the dependent variables, we used a backward elimination procedure dropping the variable with the lowest significance (Wald statistic) and continue to re-estimate the model until only significant control variables are included in the model. We used a tolerant criterion of a p-value < 0.10; hence, even control variables with minor effects are still included. Note that ‘religiosity’ of the respondent is not included in the set attitudes. An analysis revealed that the religiosity of the respondent’s parents, which is in the model, was more strongly related to ‘marriage versus separation’ than religiosity of the respondent. Of course, this parental religiosity indicator should be regarded as a ‘cultural’ characteristic that influences family transitions. However, it is much more a background char-acteristic than it is a personally expressed attitude of the respondent. For this reason it was included in the set of background characteristics. Preliminary analyses also led to the following choices: (a) Combining the employment status of the partners was the stronger predictor of the different employment variables we have tested. (b) Educational level of the respondent and/or the partner proved not to be significant; educational level of the male partner was the only variable that was relevant. (c) Only two of the parental background characteristics appeared to be relevant in this study: religiosity of the parents and whether the respondent comes from an intact family. Socio-economic characteristics of the parents, such as information on education and occupation, had little effect. The descriptive statistics of the selected control variables and ideational factors are presented in Table 1. The latter variables are included in the model without applying a backward selection mechanism, since these constitute the core of our research questions.

Procedure

(11)

235

Table 1 Descriptives. Percentages or means of key variables

Total (N = 705)

Dependent (transition) Percentage

no transition (cohabiting) 60.1 Separation 16.0 Marriage 23.8 Socio-demographic characteristics Cohort 68 32.8* 72 37.7 76 29.5 Gender men 39.3* women 60.7

Age difference between partners (in years) 2.23 Age at leaving the parental home (ALPH) 19.24 Employment status of the partners at first interview

both partners unemployed 7.5*

‘man employed, woman unemployed’ 9.8

‘man unemployed, woman employed’ 24.5

part-time + full-time employment 18.4

both partners full-time employed 39.7

Intact family (are parents of respondent separated or divorced?)

intact family 78.7*

non-intact family 21.3

Number of siblings (mean) 1.22

Religiosity of the parents (member of a religious congregation + level of affiliation)

father and mother not members 21.3*

father or mother member 4.5

‘father and mother members, but low affiliation’ 63.0 both members and at least one parent with high affiliation 11.2 Educational level of the male partner

low 39.9*

middle 41.1

high 19.0

Time since first interview (in months) 2.4 Length of the co-residential union (in years/12) 3.85 Attitudes (scale values range between 0 and 1) Mean

Familistic attitudes 0.62

Work attitudes: socio-economic success 0.60

Work attitudes: social/engagement 0.43

Work attitudes: security/materialism 0.67

Evaluations of current living arrangement (range: 0–1)

Living alone again: improvement personal well-being 0.54 Satisfaction with current relationship 0.91 Intentions

Intention to marry Percentage

no intention to marry 28.4*

don’t know 15.4

‘intends to marry, but later on’ 37.1

intends to marry within 2 years 19.1

Absence of birth plans (childless without intention to become parent)

NO (children in the household or intention to become parent) 81.7* YES (childless without intention to become parent) 18.3

(12)

Following Baron and Kenny’s advice (1986), background ‘control’ variables are included before the key ‘independent’ attitude variables. Intentions are included in the final step of the analysis for reasons we explained in the previous section, i.e. we hypothesize that direct effects of attitudes and evaluations on marriage or separation remain to be observed even after including intentions. Furthermore, we present both models with and without intentions, since it can be theoretically questioned whether intentions can be viewed as predicting behaviour or as proxies of anticipated behaviour.

Results

The main focus of this article is the relationship between, on the one hand, attitudes, evalua-tions and intenevalua-tions and, on the other, the transition to marriage or separation among cohab-iting couples. The other background covariates that were included function as control variables. Hence, in discussing the findings from our analyses we first focus on the comparison of the results obtained in the second and third steps of the analyses. The results of the stepwise analysis are summarized in Table 2. Effect parameters (betas and odds) indicate the likelihood of marriage or separation relative to no transition (= remaining cohabiting), which is the reference category. The reference category of the categorical covariate is listed next to the label of the variable. Attitudes and evaluations are defined as metric scales ranging from 0 to 1. As such, we are estimating the maximum effect of any attitude scale, i.e. moving from ‘completely disagree’ to ‘completely agree’.6_{The length of time since the first survey in 1999 and the length}

of the co-residential union are time-varying covariates. The age difference between the partners and the age at first leaving the parental home are two other metric covariates. Other variables are treated as categorical. Preliminary analyses have revealed a non-linear effect (= quadratic term) of age at first leaving the parental home and of the length of the co-residential relation-ship, and hence a quadratic term was included to model this non-linearity. The Wald statistic indicates the overall significance of each variable in the analysis, and the z-value (= β/s.e.) the relative magnitude of a particular effect since z-values are unaffected by the chosen metric of a variable. Comparing z-values indicates the magnitude of effects but not the strength of effects, since z-values have no upper limit.

Starting with the attitudes (set 2) in step 2, we find that ‘familistic attitudes’ and ‘socio-economic success’ are significantly related to the transition to marriage and separation as hypothesized (hypotheses (a) and (b)). Familistic attitudes increase the likelihood of marriage substantially (exp(b) = 10.211) and decrease the likelihood of separation (exp(b) = 0.231). Stressing the importance of ‘socio-economic success’ increases the likelihood of both types of transition, i.e. on the one hand, it makes marriage more likely (exp(b) = 2.016), but it also increases the likelihood of separation (exp(b) = 2.641). The other two work-related attitude scales are not significantly related to either transition. Furthermore, the direction of the effect of the ‘materialism/security’ dimension was the opposite of what we expected (cf. hypothe-sis (d)). Even when no control variables were included (results not shown) we did not observe a significant effect. It is therefore safe to conclude that no effects of the ‘materialism/security’ and ‘social engagement’ dimension of work values were found.

(13)

237 T able 2 Stepwise discr ete-time multinomial r egr

ession analysis of the transition to marriage or separation among cohabiting couples in

Sweden

dependent

Step 1: Backgr

ound characteristics

Step 2 : Step 1 + attitudes + evaluations

Step 3 : Step 2 + A + E + intentions (r ef. = no transition) beta z-value exp(b) W ald p-value beta z-value exp(b) W ald p-value beta z-value exp(b) W ald p-value

Independents Set 1: Backgr

ound characteristics Cohort (r efer ence category = 68) 20.137 0.000 21.437 0.000 22.934 0.000 72 Separation 0.224 0.825 1.251 0.226 0.824 1.253 0.226 0.817 1.253 Marriage 0.324 1.740 1.383 0.296 1.586 1.344 0.595 3.152 1.812 76 Separation 0.530 1.787 1.699 0.630 2.050 1.877 0.508 1.581 1.662 Marriage –0.639 –2.324 0.528 –0.684 –2.455 0.505 –0.377 –1.320 0.686 Gender (r ef. = men) 10.808 0.005 11.031 0.004 11.288 0.004 women Separation 0.779 3.153 2.179 0.838 3.218 2.312 0.865 3.277 2.375 Marriage –0.161 –0.916 0.851 –0.149 –0.812 0.862 –0.136 –0.727 0.872 Age dif fer

ence between partners (in years)

6.389 0.041 3.624 0.160 3.161 0.210 Separation 0.158 2.291 1.171 0.124 1.799 1.132 0.122 1.759 1.130 Marriage –0.059 –1.057 0.943 –0.035 –0.616 0.966 –0.015 –0.250 0.986

Age at leaving the par

ental home (=ALPH)

(14)

238 T able 2 Continued dependent Step 1: Backgr ound characteristics

Employment status of the couple at first interview

12.158 0.140 12.859 0.120 15.979 0.043 (r ef. = both ar e unemployed) ‘man employed, Separation 0.301 0.616 1.351 0.591 1.182 1.805 0.701 1.385 2.015 woman Marriage 0.191 0.422 1.210 0.173 0.380 1.189 0.131 0.274 1.139 unemployed’ ‘man unemployed, Separation 0.481 1.123 1.618 0.841 1.91 1 2.318 0.903 2.027 2.468 woman Marriage 0.323 0.805 1.381 0.249 0.607 1.283 0.550 1.260 1.733 employed’ full- + part-time Separation 0.788 1.816 2.199 1.126 2.515 3.082 1.217 2.683 3.377 Marriage 0.807 2.038 2.241 0.693 1.732 1.999 0.883 2.105 2.418 both full-time Separation 0.533 1.270 1.704 0.869 2.012 2.385 0.999 2.264 2.716 Marriage 0.588 1.532 1.800 0.490 1.240 1.633 0.683 1.623 1.980 Par

ental ‘intact’ family (r

ef. = intact family)

5.870 0.053 6.698 0.035 8.288 0.016 non-intact Separation 0.402 1.637 1.495 0.291 1.147 1.337 0.325 1.283 1.383 Marriage 0.387 1.796 1.473 0.509 2.326 1.663 0.581 2.584 1.787 Number of siblings (r ef. = ‘0’) 8.740 0.068 9.443 0.051 16.512 0.002 1 sibling Separation 0.001 0.004 1.001 0.099 0.324 1.105 0.131 0.428 1.140 Marriage 0.768 2.576 2.155 0.787 2.589 2.197 1.004 3.148 2.728 2 or mor e Separation 0.197 0.659 1.218 0.272 0.872 1.313 0.299 0.962 1.349 siblings Marriage 0.490 1.549 1.632 0.463 1.436 1.589 0.446 1.320 1.562

Religiosity of the par

ents (r

ef. father and mother not members)

11.656 0.070 14.062 0.029 8.255 0.220 father or mother Separation –0.882 –1.187 0.414 –1.406 –1.847 0.245 –1.372 –1.813 0.253 is member Marriage 0.022 0.047 1.022 0.153 0.327 1.166 0.177 0.370 1.193

father and mother

Separation 0.189 0.762 1.208 0.215 0.842 1.239 0.173 0.675 1.189 members with Marriage 0.569 2.479 1.766 0.606 2.625 1.833 0.376 1.599 1.456 low af filiation both members Separation 0.407 1.229 1.502 0.397 1.170 1.488 0.376 1.106 1.457

and at least one

(15)

Educational level of male partner (r

ef. = low) 15.189 0.019 16.728 0.010 17.21 1 0.009 middle Separation 0.339 1.504 1.404 0.310 1.341 1.363 0.252 1.071 1.286 Marriage –0.31 1 –1.558 0.733 –0.361 –1.790 0.697 –0.485 –2.271 0.616 high Separation 0.391 1.325 1.478 0.173 0.575 1.189 0.058 0.190 1.060 Marriage 0.416 1.979 1.516 0.459 2.092 1.582 0.153 0.651 1.165 else or not Separation 0.256 0.571 1.292 0.346 0.771 1.413 0.378 0.832 1.459 stated Marriage 0.389 1.085 1.476 0.366 1.006 1.443 0.903 2.461 2.468 T

ime since first interview (in months)

14.235 0.001 12.792 0.002 2.480 0.290 Separation –0.016 –2.131 0.984 –0.014 –1.750 0.986 –0.012 –1.565 0.988 Marriage –0.019 –3.125 0.981 –0.019 –3.127 0.981 –0.001 –0.181 0.999

Duration of the co-r

esidential union (in years/12) (= DURA

TION) 6.139 0.046 7.961 0.019 7.425 0.024 Separation 0.074 0.594 1.077 0.075 0.583 1.078 0.086 0.661 1.090 Marriage 0.262 2.408 1.300 0.310 2.763 1.364 0.302 2.646 1.353 Duration 2 6.204 0.045 7.686 0.021 8.994 0.01 1 Separation –0.007 –0.708 0.993 –0.008 –0.764 0.993 –0.008 –0.840 0.992 Marriage –0.019 –2.391 0.981 –0.022 –2.667 0.978 –0.024 –2.882 0.976 Set 2: Attitudes Familistic attitudes 24.874 0.000 6.463 0.040 Separation –1.464 –2.465 0.231 –1.171 –1.853 0.310 Marriage 2.324 4.327 10.21 1 0.962 1.730 2.616 W

ork attitudes: socio-economic success

9.891 0.007 8.893 0.012 Separation 0.971 2.401 2.641 0.993 2.448 2.699 Marriage 0.701 2.041 2.016 0.586 1.714 1.797 W

ork attitudes: social/engagement

1.481 0.480 2.751 0.250 Separation 0.143 0.417 1.154 0.139 0.402 1.149 Marriage –0.335 –1.141 0.715 –0.485 –1.607 0.615 W

ork attitudes: security/materialism

(16)

Set 3: Evaluations of curr

ent living arrangement

Living alone again: impr

ovement personal well-being (mean rating)

4.437 0.1 10 1.242 0.540 Separation 0.236 0.566 1.267 0.319 0.750 1.375 Marriage –0.667 –2.026 0.513 –0.288 –0.820 0.750

Satisfaction with curr

ent r

elationship (mean rating)

29.564 0.000 25.243 0.000 Separation –1.814 –4.915 0.163 –1.964 –4.994 0.140 Marriage 1.212 2.302 3.361 0.295 0.528 1.343

Set 4: Intentions Intention to marry (r

ef. = no intention to marry)

156.604 0.000 don’t know Separation –0.1 19 –0.367 0.888 Marriage 0.829 1.855 2.291 ‘Y es, but Separation 0.231 0.913 1.260 later on’ Marriage 1.403 3.61 1 4.066 ‘Y es, within Separation 0.597 1.786 1.817 2 years’ Marriage 3.291 8.496 26.875

Absence of birth plans (r

ef. = childr

en in the household or intention to become par

(17)

afterwards and less likely to separate (exp(b) = 0.163), suggesting that a positive evaluation of the current relationship increases the likelihood of consolidating the current relationship or reaffirming it by marriage.

A general finding when adding the two ‘intention’ variables in set 4, i.e. the ‘intention to marry’ and ‘absence of birth plans’ (among childless couples), is that they decrease the effect of the attitudes and evaluations on the likelihood of marriage, whereas the effects of attitudes and evaluations on separation remain virtually the same. This is pronounced in the case of familistic attitudes, for which the odds of marriage have decreased to 2.616 compared to 10.211 when intentions are not included. This illustrates that family-related ideational factors, whether they are attitudes or intentions, tend to cluster together (or are highly correlated) in such a way that they reflect a general family orientation steering the decision to marry. What is more important though is that an independent effect of attitudes and evaluations net of intentions is observed. This indicates that there is more about attitudes and evaluations than mere reflections of intentional behaviour. As expected from the perspective of the second demographic transition theory, ‘being childless without birth plans’ decreases the likelihood of marriage (exp(b) = 0.392) and slightly increases the likelihood of separation (exp(b) = 1.615). The fact that the absence of birth plans is the key issue here was concluded from a more detailed analysis revealing that the likelihood of marriage was similar for couples with children in their household compared to childless couples who intended to have children in the future. Complementary to these findings is that the likelihood to marry increases the more firmly this intention was expressed. When a respondent indicated planning to marry within two years, the odds of marriage peak at a value higher than 26. This finding does not come as a surprise, since short-term planning reflects anticipated behaviour. What is remarkable is that even while controlling for such a strong intention we still find evidence that attitudes are related to behaviour. Effects of intention to marry on separation were small and insignificant.

In our conceptual framework, background characteristics were mainly included as control variables, since we hypothesized that ideational factors would have significant effects on union transitions net of these characteristics. Some of the effects of the background character-istics deserve comment. First of all, it is worth noting that in comparing the effects of socio-demographic variables in the three steps of the analysis, few important changes in effect size of the control variables are observed. Furthermore, if a change is observed, it was rarely a decrease in effect size. Indeed, background variables such as ‘employment status’, ‘coming from an intact family’ and ‘number of siblings’, even gained significance when ideational vari-ables were included. This finding, of course, provides evidence for the theoretical argument that ideational factors complement other types of explanation.

(18)

(exp(b) = 0.984). The reason for separation being highest among early nest-leavers and younger cohorts is probably due to the fact that these categories are still in the process of finding the ‘right’ partner (Oppenheimer, 1994). Marriage, on the other hand, is more prevalent among the middle cohort (exp(b) = 1.383) and less among the youngest cohort (exp(b) = 0.528) of our sample compared to the oldest cohort. The likelihood of marriage decreases with time since first interview (exp(b) = 0.981), but increases with the length of the co-residential union (exp(b) = 1.300). At older ages of the co-residential union this increase is slowed down (exp(b) of duration2 _{= 0.981). These findings on the timing of marriage suggest that – paraphrasing}

Veevers (1980) – if marriage is postponed for a long time, it might be postponed for an in-definite time.

Summary and conclusions

Even with a relatively small sample and a short (4-year) period of observation, we were able to demonstrate important effects of attitudes and evaluations on the likelihood of transitions out of a cohabiting relationship in Sweden, a country where informal cohabitation is highly socially accepted, but where marriage seems to remain the preferred form of living arrange-ment for the overwhelming majority. Thus, we contribute to the rapidly growing field of family research, investigating what makes people transform their cohabiting relationship into marriage, split up or remain in long-term cohabitation.

The main focus of this article is the relationship between ideational factors and the transi-tion to marriage or separatransi-tion among cohabiting young adults in Sweden. A key finding is that ideational factors which directly refer to the anticipated behaviour – as indicated by the theory of Planned Behaviour – do influence the choices of cohabiting couples regarding their relationship. Familistic attitudes, satisfaction with the relationship and intention to marry influence decisions about marriage, separation or continued cohabitation. Agreement with familistic attitudes, being satisfied with the relationship increases the likelihood of marriage and decreases the likelihood of separation. Not surprisingly, a strong intention to marry is strongly related to the transition to marry.

We also found clear evidence for arguments put forward by the theory of the second demo-graphic transition. First, significant effects of attitudes on transitions were observed after controlling for a series of socio-economic background characteristics. Even adding intentions did not completely weaken the effect of attitudes on transitions. Second, work-related values of success proved to be important, since these increased the likelihood of both transitions, indi-cating that ‘career-orientation’ leads to a more stable form of living arrangement, i.e. marriage, or, if the relationship is not satisfactory, to a break-up. Other work-related attitudes, however, proved not to be significantly related to the transitions young cohabiting couples experience. Our analyses are not decisive on the question whether a general values climate inhibits or fosters transitions – as suggested by the second demographic theory. For that we needed addi-tional information on attitudes such as politics and morality. However, we did find evidence that attitudes which do not reflect dispositions to respond favourably to the event of interest are important in explaining transitions, hence extending the framework defined by the theory of planned behaviour.

(19)

Appendix

Attitudinal scales

Procedure

An attractive feature of the survey is that a fairly large set of attitudinal questions was included in the questionnaire. For this research we selected attitudes that refer to the significance and meaning of family and work. The first step of our analysis consisted of an exploratory principal component analysis for each set of questions, i.e. familistic attitudes and work attitudes. The former identified a single dimension, whereas issues regarding work unfolded into three separate scales. Items that did not consistently correlate with one particular dimension (or principal component) or that reduced the reliability of the scale (Cronbach’s alpha) were omitted from the construction of composite scores. The familistic attitude scale included 10 items; 12 items were used to operationalize the work attitude scales. In the second step, composite scales were constructed by summing the scores on items that identify a particular dimension. These composite scales are rescaled to have a range from 0 to 1. Since the magnitude of the regres-sion coefficients varies with the range of the observed values of the covariates, a uniform rescaling of the composite scales facilitates comparison of their effects. Similarly, we have rescaled the two evaluation variables, i.e. satisfaction with the relationship and evaluation of being single again. Comparing z-values, which are the effect parameters divided by their standard deviations, also indicates the relative importance of variables, since their values are not influenced by scale range.

Because not all respondents answered all questions, we imputed missing data by means of multiple regression estimates of missing values and added a random component to the regression estimates for each missing value. Imputing estimated values to replace missing data in this way has the effect that the mean and standard deviations of the items hardly change and that the intercorrelations between items are unaffected. This method is to be preferred since it only assumes a missing at random mechanism (MAR, see Rubin, 1976; Little, 1982; Little and Rubin, 1987), which is less restrictive than the missing completely at random (MCAR) mechanism that is assumed with list-wise deletion of missing cases or mean substitution of missing data. Unlike MCAR, MAR assumes that the pattern of non-response may depend on the observed variables in the non-response pattern, which is – of course – a much more likely assumption.

Scales

In what follows, we present the selected questions as administered in the survey (translated from Swedish), indicate the scales that could be constructed – including the reliability (Cronbach’s alpha) of the scale – and label the scales that are used in the article.

Set 1. Familistic attitudes

a. What is your view of the following statements?

Note: Answers rate from 1 ‘don’t agree at all’ to 5 ‘agree completely’. 1) To have children is part of what gives life meaning.

2) Something is missing if a couple never have children. 3) It is too easy to get divorced in today’s Sweden.

b. People have different opinions about what is important in life. Can you tell us how important it is to you to achieve the following?

Note: Answers rate from 1 ‘unimportant’ to 5 ‘important’.

(20)

4) To live in a good cohabiting or married relationship. 5) To have children.

c. Here are some statements about children and family. What do you think? Note: Answers rate from 1 ‘don’t agree at all’ to 5 ‘agree completely’.

6) I enjoy children.

7) I think I can be satisfied with my life if I am a good parent. 8) Spending time with the family is more rewarding than work. 9) Children need siblings.

10) To have children is confirmation of a good partner relationship.

Scale: FAMATT ‘traditional familistic attitude’ (reliability = 0.80)

Set 2. Work attitudes

b. People have different opinions about what is important in life. Can you tell us how important it is to you to achieve the following?

11) To do well economically. SECURITY 12) To be successful in my work. SUCCESS d. What does a good job mean to you?

13) That I can think and act independently. SUCCESS 14) That it offers good possibilities to advance. SUCCESS 15) That I can be proud of my work. SUCCESS 16) That my work is useful for society. SOCIAL 17) That I meet many people. SOCIAL 18) That I get a high salary and/or other benefits. SECURITY 19) That I can help other people. SOCIAL 20) That I have many good work mates. SECURITY 21) That I have secure employment with a regular income. SECURITY 22) That I can take parental leave and/or work part-time SECURITY

without facing difficulties at work.

Scales

SECURITY ‘materialism and social security’ (reliability = 0.65) SUCCESS ‘socio-economic success’ (reliability = 0.69)

SOCIAL ‘engagement and social meaning of work’ (reliability = 0.68)

Appendix Notes

(a) Item 18 ‘high salary and/or other benefits’ correlated about equally with the ‘success’ and ‘security’ issues, indicating that a ‘high salary’ refers to professional success as well as to materialistic concerns. Rather than dropping the item in the analyses, we have decided to include it in the ‘security’ scale to indicate materialistic concerns. Since effects of partic-ular scales on transitions are controlled for other scales, the consequence of this decision

(21)

is that the ‘success’ dimension refers to the importance of social status per se irrespective of materialistic orientation.

(b) One could argue that for some of these scales the reliability is modest (i.e. < 0.70). However, these values pertain to composite scales that include only a few items. As Carmines and Zeller (1979) argue, if we were able to increase the number of items with equivalent items that correlate at approximately the same level, the reliability would increase significantly. Hence, the reliability may be interpreted as fair. Nevertheless, we should keep in mind that these attitudinal scales are only rough indicators of the constructs we would like to introduce in the event-history regression models.

Notes

1. For a cross-sectional analysis of the relationship between cohabitation and commitment, see Bernhardt et al. (2007).

2. Note that we use the term ‘ideational’ in its broadest sense, referring to the different expressions of respondents. These might include attitudes and values as well as intentions, evaluations or sentiments. 3. The question asked was: ‘Do you think you will have children in the future?’ and the response

alter-natives were ‘yes’, ‘perhaps’ and ‘no’.

4. Significance is tested with a logistic regression analysis. ‘Participation’ was the dependent variable; attitudes, evaluations, intentions and background variables mentioned in the article were the covariates. 5. A competing risk model makes the assumption of independence of irrelevant alternatives, which states that the odds of an event do not depend on other alternatives given the variables included in the model. This is especially so when not all alternatives in the dependent variable are present; this assumption is an issue of concern. In our analyses, the three possible alternatives for cohabiting couples are present, i.e. marriage, separation and continued cohabitation. Note that tests for assessing violation of this assumption yield conflicting results and, for this reason, are regarded as not very useful (Long and Freese, 2006).

6. Note that estimating maximum effect sizes is also used when the effect of metric variables is compared with the effects of dummy variables in OLS (Ordinary Least Squares) regression. Since original scale points for all attitudes range from 1 to 5, the effect of moving from one scale point to the next equals the estimated beta divided by five.

References

Ajzen, I. and Fishbein, M. (1980) Understanding Attitudes and Predicting Social Behaviour. Englewood Cliffs, NJ: Prentice-Hall.

Allison, P. D. (1984) Event History Analysis: Regression for Longitudinal Event Data. Beverly Hills, CA: Sage. Axinn, W. G. and Thornton, A. (1992) ‘The Influence of Parental Resources on the Timing of the Transition

to Marriage’, Social Science Research 21: 261–85.

Axinn, W. G. and Thornton, A. (2000) ‘The Transformation in the Meaning of Marriage’, in L. Waite (ed.) Ties that Bind: Perspectives on Marriage and Cohabitation. New York: Aldine de Gruyter.

Baanders, A. N. (1998) Leavers, Planners and Dwellers. The Decision to Leave the Parental Home. Wageningen: Thesis Wageningen.

Barber, J. S. and Axinn, W. G. (1998) ‘The Impact of Parental Pressure for Grandchildren on Young People’s Entry into Cohabitation and Marriage’, Population Studies 52: 129–44.

Barber, J. S., Axinn, W. G. and Thornton, A. (2002) ‘The Influence of Attitudes on Family Formation Process’, in R. Lesthaeghe (ed.) Meaning and Choice: Value Orientations and Life Course Decisions, pp. 45–93. The Hague and Brussels: NIDI CBGS Publication.

Baron, M. B. and Kenny, D. A. (1986) ‘The Moderator-Mediator Variable Distinction in Social Psycho-logical Research: Conceptual, Strategic, and Statistical Considerations’, Journal of Personality and Social Psychology 51: 1173–82.

Becker, G. (1981) A Treatise on the Family. Cambridge, MA: Harvard University Press.

(22)

Bernhardt, E. (2002) ‘Cohabitation and Marriage among Young Adults in Sweden: Attitudes, Expecta-tions and Plans’, in Nordic Demography: Trends and Differentials, Scandinavian Population Studies, pp. 157–70. Oslo: Unipub.

Bernhardt, E. (2003) ‘Cohabitation’, in Encyclopaedia of Population. New York: Macmillan Reference. Bernhardt, E. (2004) ‘Cohabitation or Marriage? Preferred Living Arrangements in Sweden’. Available

at: http://www.oif.ac.at/sdf/sdf04–04-bernhardt.pdf.

Bernhardt, E., Noack, T. and Wiik, K. A. (2007) ‘Cohabitation and Commitment: Is Cohabitation Really Indistinguishable from Marriage in Norway and Sweden?’. Paper presented at the Annual Meetings of the Population Association of America, New York, April 2007.

Blom, S. (1994) ‘Marriage and Cohabitation in a Changing Society: Experiences of Norwegian Men and Women Born in 1945 and 1960’, European Journal of Population 10: 143–74.

Brown, S. (2000) ‘Union Transitions among Cohabitors: The Significance of Relationship Assessments and Expectations’, Journal of Marriage and the Family 62: 833–46.

Bumpass, L. (2002) ‘Family-Related Attitudes, Couple Relationships and Union Stability’, in R. Lesthaeghe (ed.) Meaning and Choice: Value Orientations and Life Course Transitions, pp. 161–84. The Hague and Brussels: NIDI CBGS Publication.

Carmines, E. G. and Zeller, R. A. (1979) Reliability and Validity Assessment. Sage University Papers, Series on Quantitative Applications in the Social Sciences, no. 017. Beverly Hills, CA and London: Sage. Cherlin, A. (1992) Marriage, Divorce, Remarriage. Cambridge, MA: Harvard University Press.

Cherlin, A. (2000) ‘Toward a New Home Socioeconomics of Union Formation’, in L. Waite, (ed.) Ties that Bind: Perspectives on Marriage and Cohabitation, pp. 126–44. New York: Aldine de Gruyter.

Clarkberg, M., Stolzenberg, R. M. and Waite, L. J. (1995) ‘Attitudes, Values, and Entrance into Cohabita-tional Versus Marital Unions’, Social Forces 74: 609–32.

Cooney, T. M. and Hogan, D. P. (1991) ‘Marriage in an Institutionalized Life Course: First Marriage among American Men in the Twentieth Century’, Journal of Marriage and the Family 53: 178–90. Duvander, A.-Z. (1999) ‘The Transition from Cohabitation to Marriage: A Longitudinal Study of the

Propensity to Marry in Sweden in the Early 1990s’, Journal of Family Issues 20: 698–717.

Easterlin, R. (1980) Birth and Fortune. The Impact of Numbers on Personal Welfare. New York: Basic Books. Fishbein, M. and Ajzen, I. (1975) Belief, Attitude, Intention and Behavior: An Introduction to Theory and

Research. Reading, MA: Addison-Wesley.

Gähler, M., Hong, Y. and Bernhardt, E. (2009) ‘Parental Divorce and Union Disruption among Young Adults in Sweden’, Journal of Family Issues. Online first. doi: 10.1177/0192513X08331028.

Goldscheider, F. and Waite, L. (1986) ‘Sex Differences in the Entry to Marriage’, American Journal of Sociology 92: 91–109.

Goldstein, J. and Kenney, C. (2001) ‘Marriage Delayed or Marriage Foregone? New Cohort Forecasts of First Marriage for U.S. Women’, American Sociological Review 66: 506–19.

Hellevik, O. (2007) ‘Linear Versus Logistic Regression When the Dependent Variable is a Dichotomy’, Quality & Quantity. Online first. doi: 10.1007/s1113500790773.

Hoem, J. and Hoem, B. (1986) ‘The Disruption of Marital and Non-Marital Unions in Contemporary Sweden’, in J. Trussell (ed.) Demographic Applications of Event History Analysis, pp. 61–93. Oxford: Clarendon Press.

Kaufman, G. (2000) ‘Do Gender Role Attitudes Matter? Family Formation and Dissolution among Tradi-tional and Egalitarian Men and Women’, Journal of Family Issues 21: 128–44.

Lesthaeghe, R. (1995) ‘The Second Demographic Transition in Western Countries: An Interpretation’, in K. O. Mason and A.-M. Jensen (eds) Gender and Family Change in Industrialized Countries. Oxford: Clarendon Press.

Lesthaeghe, R. (1998) ‘On Theory Development: Applications to the Study of Family Formation’, Population and Development Review 24: 1–14.

Lesthaeghe R. and Moors, G. (1994) ‘Living Arrangements, Socio-Economic Position and Values among Young Adults: A Pattern Description for France, Germany, Belgium and the Netherlands 1990’, in J. C. van den Brekel and F. Deven (eds) Population and Family in the Low Countries. Amsterdam: Kluwer. Lesthaeghe, R. and Moors, G. (2002) ‘Life Course Transitions and Value Orientations: Selection and Adaptation’, in R. Lesthaeghe (ed.) Meaning and Choice: Value Orientations and Life Course Decisions, pp. 1–44. The Hague and Brussels: NIDI CBGS Publication.

(23)

Lesthaeghe, R. and Surkyn, J. (1988) ‘Cultural Dynamics and Economic Theories of Fertility Change’, Population and Development Review 14: 1–45.

Lichter, D. L., Qian, Z. and Mellot, L. M. (2006) ‘Marriage or Dissolution? Union Transitions among Poor Cohabiting Women’, Demography 43: 223–40.

Liefbroer, A. C. (1991) ‘The Choice Between a Married or Unmarried First Union by Young Adults: A Competing Risk Analysis’, European Journal of Population 7: 273–98.

Liefbroer, A. C., Gerritsen, L. and de Jong Gierveld, J. (1994) ‘The Influence of Intentions and Life Course Factors on Union Formation Behaviour of Young Adults’, Journal of Marriage and the Family 56: 193–203. Liska, A. E. (1984) ‘A Critical Examination of the Causal Structure of the Fishbein/Ajzen Attitude–

Behavior Model’, Social Psychology Quarterly 47: 61–74.

Little, R. J. A. (1982) ‘Models for Nonresponse in Sample Surveys’, Journal of the American Statistical Asso-ciation 77, 237–50.

Little, R. J. A. and Rubin, D. B. (1987) Statistical Analysis with Missing Data. New York: Wiley.

Long, J. S. and Freese, J. (2006) Regression Models for Categorical Dependent Variables Using Stata. College Station, TX: Stata Press.

Moors, G. (1997) The Dynamics of Values-Based Selection and Values Adaptation. With an Application to the Process of Family Formation. Vrije Universiteit Brussel, IPD-Working Paper 1997–4.

Moors, G. (2000) ‘Values and Living Arrangements: A Recursive Relationship’, in L. Waite (ed.) Ties that Bind: Perspectives on Marriage and Cohabitation, pp. 212–26. New York: Aldine de Gruyter.

Oppenheimer, V. K. (1994) ‘Women’s Rising Employment and the Future of the Family in Industrial Societies’, Population and Development Review 20: 293–342.

Pollak, R. A. and Watkins, S. C. (1993) ‘Cultural and Economic Approaches to Fertility: Proper Marriage or Mésalliance?’, Population and Development Review 19: 467–95.

Rokeach, M. (1973) The Nature of Human Values. New York: Free Press. Rubin, D. B. (1976) ‘Inference and Missing Data’, Biometrica 581–92.

Smock, P. J. and Manning, W. D. (1997) ‘Cohabiting Partners’ Economic Circumstances and Marriage’, Demography 34: 331–41.

Statistics Sweden (1995) Kvinnors och Mäns liv. Del 2. Parbildning och Separationer (The life of women and men. Part 2. Couple formation and separations). Stockholm: Statistics Sweden.

Sweeney, M. M. (2002) ‘Two Decades of Family Change: The Shifting Economic Foundations of Marriage’, American Sociological Review 67: 132–47.

Thornton, A. (1991) ‘Influence of the Marital History of Parents on the Marital and Cohabitational Experi-ences of Children’, American Journal of Sociology 96: 868–94.

Veevers, J. E. (1980) Childless by Choice. Toronto: Butterworth & Co. Yamaguchi, K. (1991) Event History Analysis. Newbury Park, CA: Sage.

Biographical Note: Guy Moors is Assistant Professor at the Department of Methodology and Statistics, Faculty of the Social and Behavioural Sciences, Tilburg University. His research interests are in the field of social demography, values research, survey research methodology and latent class analysis. Address: Tilburg University, FSW-MTO, Room S110, P.O. Box 90153, 5000 LE Tilburg, The Netherlands.

[email: guy.moors@uvt.nl]

Biographical Note: Eva Bernhardt is Professor Emerita of Demography at the Sociology Department, Stockholm University. Her research interests are in the field of family demography, with special emphasis on gender roles.

Address: Sociology Department, Stockholm University, SE-106 91 Stockholm, Sweden. [email: eva. bernhardt@sociology.su.se]