Decisions on marriage? Couples’ decisions on union transition in Sweden

Marriage is an institution that has become optional for many. This study investigates how decisions are taken regarding marriage among Swedish cohabiting couples in the twenty-first century, specifically focusing on whose intentions to marry are most decisive. We use the Young Adult Panel Study conducted in 2009 with augmented register data for 2009–2014 in order to observe both partners’ intentions and to then follow up on which couples ultimately married. The study finds that women’s and men’s intentions to marry seem to be equally important, but that there are gendered differences by educational level: women’s intentions carry more weight among highly educated couples, while men’s intentions carry more weight among lower educated couples.


Introduction
Marriage is still a major life transition for many young couples, but it is today more optional than it has ever been. Pressure to marry is certainly lighter than before, and many of marriage's legal and formal differences from cohabitation are declining. In most Western countries, the normative expectation among young adults to marry is declining (Duncan, Barlow, & James, 2005;Hiekel & Keizer, 2015;Jamieson et al., 2002;Ohlsson-Wijk, Brandén, & Duvander, 2018), but this has not meant that marriage is disappearing as an important union form; in fact, marriage rates in the Nordic countries have increased in recent years (Ohlsson-Wijk, 2011). Decisions regarding whether or not to marry are still made; in this study, we investigate which partner is most influential in these decisions. We investigate marriage intentions and behaviour in Sweden, a country often noted as a forerunner in the second demographic transition away from tradition and towards more secular and individualistic norms and behaviour (Moors & Bernhardt, 2009;Lesthaeghe, 2010). Since the 1960s, marriage has been on the decline as a prerequisite to family building in Sweden, a trend initiated by women and men with less education and fewer economic means to protect (Andersson, 1998;Bernhardt & Hoem, 1985;Ohlsson-Wijk, 2011). Nevertheless, it has not disappeared as an important form of union and is still related to childbearing (Bracher & Santow, 1998;Kolk & Andersson, 2020).
Qualitative studies from various Western countries provide insights into the meaning of marriage today, indicating that cohabitation is the default situation, and that boundaries are blurred between cohabitation and marriage (Jamieson et al., 2002). Some consider this "do-it-yourself marriage", a concept that may be imbued by individuals with a variety of meanings (Duncan et al., 2005). Nevertheless, the security and stability of marriage are still emphasized, particularly with regard to children and the event of separation (Hiekel & Keizer, 2015). Some may also be resistant to marry due to being critical of the institution itself or because it may alter their union in negative ways (Hatch, 2017). A better understanding of how decisions to marry are made will increase knowledge regarding present perceptions of a traditional institution. In short, for whom is marriage important today?
One partner's intentions may weigh more heavily than those of the other depending on power relations in the union. Such power relations may be based on relative resources, but also on less overt forms of power rooted in gendered and cultural expectations. In this study, we investigate both partners' marriage intentions in 2009 and follow them for 5 years to see which couples ultimately married. We achieved this by connecting survey data to a follow-up in administrative marriage registers.

Theoretical framing of decisions to marry
In most cases, partners agree in their intentions to marry, but when they do not, the question is which partner's plans prove most important. We are interested in the decision to marry and whose plans or intentions are decisive for cohabiting couples. In line with theories of power stemming from relative resources in the household, one would expect that the individual with superior resources will have more to say about marriage decisions, just like they will also have more to say about how household tasks are divided (Blood & Wolfe, 1960;Evertsson & Nermo, 2004;Sullivan & Gershuny, 2016). However, the way in which relative resources are defined may be crucial: for instance, one study from the Czech Republic noted that the use of relative educational level and income as indicators may have yielded inconsistent results (Hamplová, Chaloupková, & Topinková, 2019). Relative income is more of a direct measure, while relative education can be seen as including prospects of resources, perhaps more important for a longterm decision such as marriage. It is therefore likely that the partner with higher education in a couple has a stronger say on whether to marry or not, and that relative education is a better measure than relative income for this kind of decision.
Traditionally, women were more protected in marriage than today because they were economically dependent on their partner. However, in a society in which both partners are or will be active in the labor market and in which there is no alimony after a separation, there are virtually no obvious gendered reasons for women to be more eager to realize marriage plans. Nevertheless, it has repeatedly been found that the gender dimension is important for couples' decision-making, often phrased as "doing gender", that is, acting to sustain gender roles even when relative resources point in another direction (West & Zimmerman, 1987). One example is women earning more are doing a greater amount of household work than their male partner. Such behavior may vary by context, for example, being visible in the USA but not in Sweden (Evertsson & Nermo, 2004). Another example close at hand is the choice of name after marriage in Finland, where subtle forms of patriarchy continue to shape the decision-making process (Castrén, 2019). In addition, higher educational level is often used to indicate genderequal attitudes and leads to more equal sharing, for example, through parental leave in Germany (Geisler & Kreyenfeld, 2011). We may therefore expect that couples with lower education are the ones who tend to adhere to "gendered decision making." Similarly, one may expect that among couples with high education, the intentions to marry (or not) of both partners are equally decisive. However, returning to one general explanation for the decline in marriage being a lack of "marriageable men", that is, men with earning potential (Koball, 2004), we also have to consider the strength of intentions, or what is at stake in a marriage. One American study has indicated that when men have weaker economic prospects, the results include a potential decline in demand for marriage among women as well as among these men (Koball, 2004). Therefore, one may expect that women seek to marry highly educated men (with economic prospects), and that these men are more open to marriage than their less educated counterparts. Earlier Swedish studies indicated that the higher the education of both the man and the woman, the greater the likelihood of marriage (e.g., Duvander, 1999).
Previous research has also pointed to gendered differences in how men and women approach marriage. Research in the USA has indicated that women's marriage intentions are declining over time and by cohort (Vespa, 2014). Wiik, Bernhardt, and Noack (2010) have found that in Scandinavia, commitment to the union is the most important factor determining women's marriage intentions, whereas for men, commitment in combination with their own and their partner's socioeconomic characteristics (education and income) are salient. Thus, the socioeconomic resources of each partner may be decisive, perhaps particularly for men. Cho, Cui, and Clardige (2018) used US data from the Fragile Family and Child Wellbeing Study to test whether mothers' or fathers' intentions to marry mattered more, the answer being the former. However, the literature is not consistent on whose intentions are more important. It is also not clear that women have a stronger say than men in carrying out marriage intentions. For example, one Swedish study has suggested that intentions certainly influence actual marriage, but that there is no significant difference between couples where only the man or only the woman intends to marry (Duvander, 2001). Inconsistency regarding who is more decisive regarding marriage is unsurprising given the changing and different meanings of marriage over time and between contexts.
In sum, expectations about who decides regarding marriage are not as clear as they may have been in a traditional setting. It is likely that gender is less important today but that relative resources based on education (indicating prospective resources) continue to matter. Nevertheless, it is plausible that gender matters differently depending on one's educational level owing to differences in gains by being married, possibly in combination with varied attitudes to traditional gender roles by educational level.

Development of marriage in Sweden
Marriage rates started to decline in the 1960s in Sweden and were quickly replaced by cohabitation as the first union. For several decades, cohabitation has constituted the start to almost all marriages (Duvander, 1999) and, over time, has become longer in duration. Marriage today does not always precede children: more than half of all children are born out of wedlock, although in most cases, to a stable union (Thomson & Eriksson, 2013). Similar trends are observed worldwide, although they start somewhat later (see, for example, Kuperberg, 2018;Manning, Longmore, & Giordano, 2007;Perelli-Harris & Lyons-Amons, 2015).
As is true in most countries, in Sweden, marriage is still less likely to break up than cohabitation, and fewer couples divorce than separate, even in unions with children (Hoem & Hoem, 1992;Thomson, Winkler-Dworak, & Beaujouan, 2019). There are few formal reasons to marry, but some regulations regarding inheritance and the division of resources after the end of a union differ between marriage and cohabitation. In essence, cohabitants do not inherit from each other and do not share the resources they brought with them to the union in the case of separation (Ohlsson-Wijk et al., 2018; see also Perelli-Harris & Sanchez-Gassen, 2012). Since the end of the 1990s, a marriage rate that was declining has turned into an increase in Sweden and the rest of the Nordic countries, which is still somewhat puzzling. In part, the trend can be explained by compositional changes-there being more young women and men at an age when people are likely to marry-but this do not fully explain the development (Ohlsson-Wijk, 2011).

Why would a couple marry in Sweden?
The question of why couples marry is certainly a valid one in contemporary Sweden. The normative pressure to marry is negligible for most couples and, if we relate Sweden to the idea of the second demographic transition (Lesthaeghe, 2010), values of individuality, secularization, and a challenging of authority more or less dominate contemporary discourse and are rarely challenged. Other than some resistance to same-sex marriage from the Christian Democratic Party (Committee on Civil Affairs, 2008/ 2009), marriage is mainly a non-existent political question, unlike in the USA, for instance (see, for example, Lichter & Qian, 2008). Furthermore, religious motives are downplayed in the secular Swedish context, where the dominant Swedish Lutheran Church does not condemn divorce and has even appointed an investigator to formulate a ritual for facilitating it (Swedish Church, 2017). Nonetheless, there may still be reasons to marry. In addition to the obvious reasons of love and commitment, the main factors leading to marriage may be a specific life course stage, the potential economic benefits of marriage, and positive attitudes toward this type of union. Even if marriages are almost always a consequence of successful cohabitation, they are more likely to occur at certain ages, after a period of cohabitation and in relation to childbearing (Bracher & Santow, 1998;Duvander, 1999;Manning & Smock, 1995;Vergauwen, Neels, & Wood, 2017). This is probably related to the still valid fact that marriages are more stable than cohabiting unions, and that stability is sought in the case of childbearing. Given that the present study focuses on how couples decide on marriage, these factors are used as control variables.
In addition, there are various benefits of co-residential living compared to single living, including the sharing of collective goods, economic gains from specialization, extending credit, coordinating investments, and risk pooling (Weiss, 1997). Given the long-term and stable nature of marriage, most of these benefits may be more efficient in marriage than in cohabitation. Formally, resources are more protected in marriage (Duvander, 1999;Perelli-Harris & Sanchez-Gassen, 2012). It is often claimed that people with higher education have more to gain in marriage because they have more economic resources to protect, now and in the future. Furthermore, higher educated women and men are more likely to marry or have intentions to marry (Duvander, 2001;Wiik et al., 2010). The same educational gradient is not found in the process of entering cohabitation (Thomson & Bernhardt, 2010), which indicates that decisions regarding different types of unions are made on different grounds.
Since 2009, Sweden has had a completely gender-neutral marriage law (Andersson & Noack, 2010), potentially signaling a new meaning of marriage. This possibility is reinforced by the fact that couples with gender-equal attitudes tend to see no problem with marriage (Ohlsson-Wijk et al., 2018). In fact, it seems that it is the gender-equal couples that most often end up married. This phenomenon may be interpreted to mean that individual values and commitment to a union do not stand in conflict once gender equality within partnerships is achieved (Goldscheider, Bernhardt, & Lappegard, 2015). In gender-equal societies, the meaning of marriage may change to indicate a commitment that does not hinder an individual's life project (Ohlsson-Wijk et al., 2018). Perhaps, marriage will change meaning worldwide to include a variety of life plans (see, for example, Deutsch, Kokot, & Binder, 2007).
Attitudes and intentions are both determinants of marriage, but it is debatable how great influence attitudes have in addition to their indirect influence through intentions. Zilincikova and Hiekel (2018) have found that attitudes have a direct effect on the outcome of marriage in a large number of European countries, but to differing degrees. They conclude that it is necessary to scrutinize attitudes towards marriage to better understand cohabitants' marital trajectories. Finally, Moors & Bernhardt (2009) have claimed that for Sweden at the turn of the century, familial values still predict marriage intensity.

Data and methods
This study used the Young Adult Panel Study (www.suda.su.se/yaps), which includes a nationally representative sample of men and women born in 1968in , 1972in , 1976in , and 1980in . There were panels in 1999in , 2002in , and 2009 in which respondents received postal questionnaires. This study used the 2009 wave and additionally included information on the co-residential partner of the main respondent. The main respondent was thus between 29 and 40 years old. The response rate was 56%. Seventy-one percent of the co-resident partners of the main respondents responded to a questionnaire with very similar content. The questionnaire included major demographic events, such as children's births, work, and partner histories as well as retrospective questions on childhood circumstances and parental characteristics. The questionnaire also contained a large set of questions on values, attitudes, and intentions, such as regarding marriage and divorce/separation. In total, 1079 cohabiting or marital couples participated in the 2009 wave, making it possible to execute couple-level analyses on a wide range of topics.
Given that this study focused on opposite-sex couples' marital behavior, we derived a subsample consisting of 521 cohabiting couples from the 2009 wave. We excluded all cohabiting same-sex couples (n 14) as our focus was on gendered decisions; hence, the analytical subsample included 507 non-married cohabiting opposite-sex couples. The data were supplemented with register data on the vital demographic events of the main respondent, i.e., dates of marriages, divorces, and children born up to 31 December 2014. This approach enabled analyses of marriage propensities after survey participation during the follow-up period. We followed respondents in cohabiting couples over time to analyze their marriage intentions using register data records of the date of marriage over the 2009-2014 period by employing logistic regression analysis in which the outcome was marriage (yes/no). The regression models estimated the probability of getting married by couples' marital intentions, controlling for relative educational level, women's age, parental status, and duration of union over the 2009-2014 period. In separate models, we also controlled for childbearing plans and marriage attitudes. In total, 201 respondents in cohabiting couples married over this period.
The variable measuring couples' marital intentions was based on two identical questions, one asked to the main respondent and the other to the partner: Do you and your partner plan to get married? The answer alternatives were (1) yes, within the next 2 years; (2) yes, but later; (3) I would like to, but my partner would not; (4) my partner would like to, but I would not; (5) no, we do not have intentions to get married; and (6) do not know. The created variable comprised four categories: (1) Both partners want to get married; (2) Woman wants to get married but man does not; (3) Man wants to get married but woman does not; and (4) Both partners do not want to get married (including "Do not know"/"Missing"). Table 1 displays the cross-tabulation of women's and men's marital intentions and indicates that the vast majority of the couples had harmonized answers. A total of 45% of the couples agreed to marry within 2 years or later (228 couples), and 15% agreed that they did not intend to get married. Only 5% agreed on only one partner wanting to marry. Hence, for most couples, coding the combination variable was not problematic because the partners' answers were not conflicting. To categorize the 136 couples who reported conflicting answers, we had to make some additional considerations. As an overruling principle, the ego's answer ruled out the partner's answer, meaning that what a person stated about him-or herself was considered truer than the partner's statement. Table 5 in the Appendix shows the coding schema for couple's marital intentions. Couples' education level was a combined variable in which both partners' highest achieved education level was taken into consideration. The variable was categorized into the following: (1) Both have tertiary education, (2) the woman has less than tertiary education, and the man has tertiary education, (3) the woman has tertiary education, and the man has less than tertiary education, and (1) both partners have less than tertiary education.
Furthermore, as our focus was on the importance of gender and educational level, we made a combined variable to identify the marriage intention of (1) both, (2) only the woman, (3) only the man, or (4) neither, tabulated by (1) both tertiary, (2) man tertiary, (3) woman tertiary, or (4) noth less than tertiary education. The 16 categories are displayed in a model where "both have tertiary education and both want to get married" was the reference category (see Table 7 in the Appendix). The odds ratios for all with "both have tertiary education" are shown in Fig. 2a, while in Fig. 2b-d, the only difference is that the reference category is recalculated, and the idea being to show how intentions varied within the educational group.
To indicate each couple's life course stage, we combined the question Do you plan to have (more) children in the future? Asked to both partners with information on the age of (any) children. We labeled the new variable couple's childbearing intentions and categorized it as follows: (1) Both partners plan to have children; (2) Either the man or the woman plans to have children; (3) no partner plans to have children; (4) have children aged 0-3; and (4) have children older than 3. Couples with children and who planned to have more were coded as having children, and we disregarded their plans to have additional children (if any). We assumed that cohabiting couples who already had children were in another life phase and that if these couples had not already married, they may have been less likely to do so. Nevertheless, the childbearing intentions of childless couples were compared to couples who already had children, regardless of their intentions of continued childbearing. Furthermore, we tested parental status in other ways. In our main model, we simplified the variable to the couple having children or not in order to avoid overloading it. We also considered the number of children in the household rather than the age of the children, but found no difference from the main conclusion (not shown). In addition, we made use of the fact that we could follow childbearing after the time of the survey for the couples. In a complementary log-log regression, we used childbearing as a time-varying variable, thus considering children after the time of interview, but before potential marriage. In total, 232 main respondents had a child in the follow-up period: 95 had their first child, and 137 had one or several additional children. These results are shown in Table 8 in the Appendix. In essence, the model yielded results leading to the same conclusion regarding marriage intentions, but we also found that having three or more children increased couples' propensity to marry.
The variable couple's marital attitudes was based on a gender-specific index that included responses to eight statements to which respondents could answer on a 5-level scale, from "Agree completely" to "Do not agree." The statements were as follows: 1. People ought to get married for the sake of the children 2. People ought to get married for economic reasons. 3. Married persons are under greater pressure to conform. 4. It is more difficult to break up when married. 5. The wedding ceremonies are too expensive. 6. It is tradition to get married. 7. It is romantic to get married. 8. Marriage shows you are serious about the relationship.
To assess the reliability of the battery of marital attitudes, Cronbach's alpha was tested, a commonly applied tool to measure the degree of relation of items or indicators in a group (index). Our test yielded an alpha of 0.75, which can be considered relatively high internal consistency (the alpha coefficient varied from 0 to 1) (Tavakol & Dennick, 2011). Statements 3, 4, and 5 were rescaled because the original scale indicated that the higher the number, the less positive the respondent was towards marriage. Individuals who answered that they did not know or had missing answers on one or a few statements were recoded as the middle category "3". Sensitivity analyses without such imputations did not change the results. The distributions of the gender-specific indexes are displayed in Fig. 1. Subsequently, the indexes were dichotomized based on the means (mean women 23.25; mean men 21.99) into the categories "Not positive towards marriage" and "Positive towards marriage". Couple's marital attitudes had four categories: (1) Both partners positive towards marriage, (2) woman positive towards marriage but man not, (3) man positive towards marriage but woman not, and (4) neither partner positive towards marriage.
Furthermore, given that individuals' intentions and attitudes may be highly correlated and potentially be proxies for each other, we performed a Pearson's chi-square test for independence between marital intentions and attitudes ( Table 2). The test showed that intentions and attitudes were not completely consistent and correlated. Hence, we could conclude that it was appropriate to include both variables in the analyses.
In the models, we also controlled for women's age and duration of union in years. We chose to dichotomize women's age at 32, as the mean age of the women in the Fig. 1 Distribution of the scores of the indexes on attitudes towards marriage by gender. Striped bars represent statistically non-significant estimates survey was 33. We further adjusted the model for union duration, which was included as a continuous variable. Descriptive statistics of all variables are presented in Table 3.
There were two major shortcomings with the data at hand. First, we could not be certain that the main respondents married their cohabiting partner from 2009 as we only had information on marriage from the main respondent. We therefore conducted sensitivity analyses in which we only included marriages up to 2 years after survey participation. 1 Some of the associations found were somewhat stronger, but the overall patterns were similar to the results presented below, giving credence to the study's main findings (see Table 9 in the Appendix). Given the Swedish context, in which marriage in the absolute majority of cases happens after a long period of cohabitation, it is very rare that someone will have had time to end one cohabiting union, start a new one, and turn it into marriage within the time frame of 2 years.
The second major shortcoming with the data was that we were not able to censor for separation for all couples. This meant that in a substantial number of cases, we were probably observing couples' marriage risk in cases they have instead ended the union. To minimize this flaw in the analysis, our main strategy was to use a number of indicators of a union being at high risk of separation at the time of the survey. In separate models, we in turn excluded couples where at least one partner claimed to have considered breaking up (n 170), does not think the relationship is serious (n 29), does not think they are in a good relationship (n 38), and is not satisfied with the relationship (n 112). When excluding these cases, the differences between both intending to marry and only the woman or the man intending to marry diminished somewhat, but the major conclusion did not change (see Table 9 in the Appendix). Pearson chi 2 (9) = 46.7 P value = 0.000 The second strategy was a sensitivity analysis based on the fact that for couples who were already parents (337 cases), we had information about whether and when the partners moved apart, this being the same as separation. In an analysis of only parental couples, we excluded those who moved apart during the observation period (sixth model in Table 9 in the Appendix). We found that this analysis also did not change our main conclusion. Again, given the Swedish context where marriage is neither normative nor very important for practical reasons, we did not consider marriage and separation as alternatives, although surely the couples that were close to separation were not less likely to marry. Table 3 displays the descriptive statistics of the cohabiting couples in the year of the survey in 2009. We found that in half of all couples, both partners intended to get married. It proved somewhat more common for only the woman to intend to get married

Couple's childbearing intentions
Both partners plan children 23 28 Either the man or the woman plans children 5 5 Neither of the partners plans children 5 3 Have children aged 0-3 3 8 4 0 Have children older than 3 30 23 Couple's marital attitudes than the opposite. In more than a quarter of the couples, neither of the partners intended to get married. Regarding education level, approximately two fifths of the couples had tertiary education. In one fifth of the couples, only the woman had tertiary education and in less than half of the couples, only the man had tertiary education. In 43% of the couples, the woman was aged over 32. The vast majority had children or planned to have children, but in 5%of the couples, only one of the partners planned to have children. In approximately two fifths of the couples, both were positive towards marriage and in one fifth of the couples, the man was positive, and the woman was not.
In somewhat more couples, the woman was positive towards marriage, and the man was not. The average length of union was 7.6 years. We started with a model only including the couples' marriage intentions and then added our control variables in steps. The results of both the binary and the controlled models showed that couples in which both partners had intentions to marry within a few years were more likely to do so than couples in which only one of the partners had intentions to marry. We did not find any statistically significant gender differences; that is, it did not matter whether the woman or the man was the partner intending to get married. We further display these results by changing the reference categories in Table  6 in the Appendix to facilitate the differentiation of the men's and the women's intentions. Unsurprisingly, the couples in which both partners did not intend to marry were the least likely to get married, but a slight tendency was apparent for marriage to occur more often if only the man intended to marry than if no one intended to do so (significant at 10%).

Logistic regression analyses
In the second model, we controlled for relative education, woman's age, and parental status. Couples where (at least) the woman was below age 33 and childless couples were found to have higher marriage intensities, while relative education had no significant effect. In the third model, duration of union was added, but had no significant impact. In the fourth model, child intentions and children's age were added, but had no significant effect. Given that marriage and child intentions may be linked at the same life course stage, we expected marriage to take place when the couple planned children. In the final model, we excluded child intentions and children's age and included marriage attitudes. As expected, marriage was more likely when both partners were positive towards marriage compared to no one being positive, but we also found that the couples where only the man was positive did not have statistically significantly different marriage intensities compared to both being positive. Marriage intensity was lower when only the woman was positive than when both partners were positive.
To further test whose intentions of marriage proved more decisive by educational group, we analyzed a combination variable based on the couples' marital intentions and relative education using the full sample and controlled for mother's age, duration, marital attitudes, and parental status. The results are presented in bar charts in Fig. 2a, d and in Table 7 in the Appendix. The reference category was both partners having intentions to marry within each educational group (both tertiary, only man tertiary, only woman tertiary, neither tertiary). For example, among couples in which both partners had tertiary education, the reference category was "Both have tertiary education, both want (to get married)." The striped bars indicated statistically non-significant estimates. The results showed that among couples in which both partners had tertiary education, there was no statistically significant difference between couples in which both partners intended to get married and those in which only the woman intended to get married. This may be interpreted as the woman having a stronger say in these couples. If the man is the only partner who intends to get married or neither intends to get married, the propensity to marry is lower. Moreover, we found that in couples in which only the man had tertiary education, both the woman's and the man's sole intention to marry indicated a lower marriage propensity than if both intended to marry. Among couples in which only the woman had tertiary education, there seemed to be no statistically significant differences by couples' marital intentions. Here, we also had one category missing because it was very uncommon for the man to intend to marry in this category (woman tertiary/man less than tertiary, man wants to get married). We should thus be very cautious in interpreting the findings. However, the results indicated that the intentions of both, one, or neither partner were not a strong determinant in cases where only the woman had tertiary education. Among low-educated couples, we found that if the man was the only partner intending to get married, the marriage propensity was as high as if both intended to marry. If only the woman or neither partner intended to marry, the likelihood was significantly lower. The findings of high-and low-educated couples were thus mirror images: in highly educated couples, the woman has a stronger say, while in couples with lower education, the man has a stronger say.

Additional sensitivity analyses
To check the robustness of our findings, we performed several additional analyses. First, we tried another operationalization of marital intentions and separated those who stated that they did not intend to marry from those who stated "Do not know". The logistic regression analyses produced similar results as those presented here, except that individuals who answered "Do not know" manifested a very low likelihood of marrying. Second, we included the couples' income levels, which did not produce any statistically significant results. Third, couples' employment status did not influence the main results. See also the sensitivity analyses to deal with flaws in the data in Tables 8 and 9 in the Appendix and described above.

Summary and conclusion
This study has found that marriage takes place when both the woman and the man intend to marry, and that there are no general gender differences in the decisiveness of intentions in Sweden at the beginning of the new century. Both partners may act out a veto, and one partner does not seem more decisive than the other. However, we have found evidence that in highly educated couples, the woman has a stronger say, while men have a stronger say in low-educated couples. Couples where either the woman or the man has a higher education are harder to interpret, probably because of the small sample size and so we refrain from interpreting and discussing the results here. Women's stronger influence on decisions in highly educated couples may be related to the notion of the men in these couples being "more marriageable" in a traditional sense of possessing economic resources, strengthening women's marriage intentions and increasing men's openness to this institution. Men's stronger say in low-educated couples can instead be interpreted as couples with low education being more likely to retain traditional male dominance in decision-making. Such differences in interpretation by educational level refer back to earlier studies, but it is of course also possible to stop with the interpretation that there may be a form of female dominance regarding marriage decisions among highly educated couples and a form of male dominance regarding marriage decisions among lower educated couples. It may also be that marriage has different meanings for different groups. These two groups are often distinguished as forerunners (highly educated) and laggards (low educated), and it may be that they attach different meanings to marriage which play out in a gendered way.
The study has not found support for the idea that a couple's education level on its own influences marriage transition. The measure may be too crude, the sample may be a b c d Fig. 2 a-d Odds of marriage by combination variable between marital intentions and education (the figures display separate models adjusted for woman's age, couples' education level, parental status, union duration, and marital attitudes). The striped bars indicate statistically non-significant results too small, or the finding may actually be a sign that the meaning of marriage is changing, in that it is no longer an institution for the protection of resources. In addition, the study has found that couples with positive marital attitudes are ultimately more likely to marry, but also that the man's positive attitudes tend to be more important than the woman's. Our measure of positive attitudes towards marriage may thus indicate a stronger desire to marry among men and slightly greater scepticism among women.
It is necessary here to mention the limitations of this study and to point out that the reader should be somewhat cautious when drawing conclusions from it. First, the response rate was around half of the drawn sample. Even though such a low response rate is common today, it is important to recognize that individuals who are younger, of ethnic backgrounds and with lower education, are often underrepresented. Therefore, we may also suppose that the responses were dominated by respondents who were interested in answering about their family and work behavior and attitudes. They may have been individuals with more conscious views on marriage and marriage intentions, potentially leading to a stronger association between intentions and behavior than is true of the population at large. Second, we were not able to fully control for separation during the observation period, even though we tried to deal with this in different ways. Nevertheless, separation may vary by education, and we had no means to control for this, which should be considered when interpreting the results. Third, we could only observe marriage for the main respondents and had to assume that the person married was the respondent's partner at the time of the survey. Given the Swedish context where marriage is optional and in the majority of cases happens after years of cohabitation, we do not consider this the kind of major problem it might be in other contexts where marriage is more common, quick, and expected. Lastly, we were restricted by our relatively small sample, and the cell sizes were sometimes small. This is why only dichotomizing educational level and using a larger survey would surely be able to nuance the conclusions drawn here.
The results of this study indicate that the meaning of marriage is indeed changing. Even if we cannot yet determine the ways and the new meaning, "old indicators" of life course stages or the protection of resources seem to matter less today, at least in Sweden. It may be, as Duncan et al. (2005) conclude, that marriage is currently a variation of cohabitation. The blurred boundaries between marriage and cohabitation are further emphasized by the non-correlation with education. Additionally, the attitudinal index that this study used included a variety of questions regarding marriage, and less than 40% of the couples were characterized by both partners being positive towards marriage. However, the questions included in the index may not correspond to contemporary reasons and motivations to marry. Thus, future research should use different, potentially open-ended questions to determine why people marry today. Qualitative studies may also shed light on this issue.
We want to conclude by noting that the vast majority of partners are aware of the intentions of their partners. Even if they do not always agree, they have a good idea of their partner's opinion on the matter. This may be interpreted to mean that marriage is still a relevant option or topic for discussion and not something that has been forgotten or an irrelevant part of relationships.  One couple excluded due to small number of cells in the category "Woman tertiary/man less than tertiary, man wants to get married"