The Demographic and Socioeconomic Consequences of Restricting Access to Marriage for Young Immigrant Women in Denmark

Abstract

Introduction

Migration increasingly shapes European populations. In 2023, more than 1 in 5 European residents were first- or second-generation immigrants, up from 1 in 7 in 2008 (Eurostat 2024). The speed at which immigrant populations adapt to the host country's marriage and fertility schedule and patterns depends on differences between the sending and receiving countries, potentially extending over generations (e.g., Fernández and Fogli 2009; Hannemann et al. 2020; Hannemann et al. 2018; Kulu and Hannemann 2019; Milewski 2011). The degree of intermarriage might also reflect immigrant group size, access to culturally similar marriage partners, and cultural distance between the host and sender societies (e.g., Dribe and Lundh 2011; Qian and Lichter 2018). Alignment to the host country's fertility and marriage schedules can ultimately be seen as a marker of assimilation or adaption (Adserà and Ferrer 2015). Further, marriage and fertility decisions are linked to educational attainment and labor market activity such that changes in one dimension spill over into and affect the others (e.g., Alderotti et al. 2021; Arendt et al. 2021; Kleven et al. 2019; Nitsche et al. 2018).

Reforms enacted in July 2002 limited marriage opportunities for all Danish and non–European Union (EU) citizens living in Denmark who wished (or whose parents wished for them) to marry someone from outside the European Union and the Nordic countries. A declared aim of the reforms was to reduce arranged marriages between an immigrant in Denmark and a spouse from their origin country (Jørgensen 2014). First, one reform prohibited family reunification between a person residing in Denmark and a spouse from abroad if either partner was younger than 24. Given that most individuals younger than 24 who found a spouse abroad were first- or second-generation immigrants to Denmark, this reform predominantly affected this population. Second, another reform required couples to prove a stronger affiliation to Denmark than to other countries (as opposed to the previous requirement of an equally strong affiliation) for a spouse to gain residency in Denmark. Beginning in 2004, the requirement was waived for those who had been Danish citizens or lived legally in Denmark for more than 28 years. Both reforms substantially limited opportunities for younger Danish residents to marry nonresidents (Schultz-Nielsen and Tranæs 2010).

This study thus focuses on women of non-EU background in Denmark—a population group generally prone to marrying early—to examine how delaying and abstaining from marriage affect subsequent fertility behavior, human capital accumulation, and labor market activities. The decision to marry and to do so at a young age has direct implications for subsequent educational attainment (e.g., Alexander and Reilly 1981; Bozick and DeLuca 2005; Buchmann et al. 2008; Goldin et al. 2006; Jacobs and King 2002; Marini 1978; Teachman and Polonko 1988). Increased educational attainment might also affect the choice of a spouse and family formation (Blossfeld and Huinink 1991; Lawrence and Breen 2016; Marini 1978; Oppenheimer 1988; Raymo 2003; Sabbah-Karkaby and Stier 2017). Because educational activity decreases with age, populations prone to marrying young will likely be more affected by delayed marriage than those prone to marrying at older ages. In the Danish educational system, the typical age at graduation from five-year tertiary education has remained at roughly 24–27 since at least 1998 (OECD 1999, 2002, 2012, 2022; see Table X1.1 in all four publications). Thus, limiting the available marriage market for individuals younger than 24 might have substantially impacted educational completion.

Marrying young has been seen as a partial consequence of culture (e.g., Sabbah-Karkaby and Stier 2017) and the absence of good alternatives (e.g., Waite and Spine 1981). This study examines how family formation, mate choice, and human capital accumulation are affected by the curbed possibility of intermarriage among a minority population. We consider the consequences of the policy for fertility, educational attainment, and labor market attachment to reveal insights relevant to migration policy and theories on associations between marriage markets, marriage decisions, and subsequent life course trajectories. Our findings show that the reforms delayed marriage and increased nonmarital cohabitation. After accounting for ongoing trends, we find that those affected by the reform also showed decreased childbearing rates, with some indications of an increase in out-of-wedlock births. Although we find plausible evidence of a short-term decrease in labor force participation that coincided with increased educational attainment, we find no long-term consequences for labor force participation.

Family Formation, Education, and Immigrant Background

Migrants bring cultural scripts and traditions that shape how they integrate into the host society's social and economic fabric (Fernández 2011; Fernández and Fogli 2009; Foner 1997; Read and Oselin 2008). Although such practices represent identity and history, they might also impede integration into the host society. The maintenance of distinct practices of intramarriage has often been tied to the size of immigrant groups (Blau 1977) and the strength of group identification and social sanctions for leaving the group's borders (Kalmijn 1998). Practices and barriers to intermarriage typically erode as time in the host country increases (Chiswick and Houseworth 2011; Fernández and Fogli 2009; Holland and de Valk 2013). Additionally, the incorporation of the host country's family-formation practices (including intermarriage) has been taken as a clear sign of integration and assimilation (or adaption) (Adserà and Ferrer 2015).

Most immigrants from outside Europe originate from countries where people generally marry young. Research spanning multiple decades shows that marrying early—especially for women—is detrimental to educational attainment (e.g., Alexander and Reilly 1981; Bozick and DeLuca 2005; Buchmann et al. 2008; Goldin et al. 2006; Jacobs and King 2002; Marini 1978; Teachman and Polonko 1988). This likely reflects cultural and normative roles but might also be a function of childbearing following marriage. Concurrently, educational attainment delays marriage, increases intermarriage, and delays fertility (Blossfeld and Huinink 1991; Chiswick and Houseworth 2011; Lawrence and Breen 2016; Marini 1978; Oppenheimer 1988; Raymo 2003; Sabbah-Karkaby and Stier 2017). Marriage and fertility delays result from the incapacitation effect of being enrolled in an educational institution. In contrast, increased intermarriage following education likely occurs via exposure to a new marriage market (Kirkebøen et al. 2021) and acquiring host country–specific capital that makes immigrants more likely to intermarry with out-group partners (Chiswick and Houseworth 2011; González-Ferrer 2006). A clear causal sequence thus connects educational decisions to subsequent family formation.

Accordingly, although increased educational attainment delays marriage, it is uncertain whether constraining immigrants’ traditional marriage markets increases intermarriage, educational activity, and subsequent returns to human capital. Because individuals often make the decision to pursue higher education before marital decisions, this literature gap has received inadequate attention. Yet, abstaining from early marriage and thereby early childbearing might facilitate increased educational attainment among population groups that tend to marry young. Cross-national research has shown a positive association between open family reunification policies and marrying a spouse from outside the host country (Carol et al. 2014). Nielsen et al. (2009) found that abstaining from marriage to an immigrant partner while young decreased the educational dropout of young men with an immigrant background but had no detectable effect on women. Schultz-Nielsen and Tranæs (2010) found a positive effect of the reform on educational enrollment among young immigrant men and women, but their results were sensitive to time trends. Moreover, studies of how shocks to (intra-)marriage markets affect family formation and education have mainly considered them in more traditional settings—for example, by examining changes to dowry and mahr payments (e.g., Chowdhury et al. 2020; Corno et al. 2020), historical cases (e.g., Eriksson et al. 2022), or constraints imposed by events such as mass incarceration (Charles and Luoh 2010) and deindustrialization (Autor et al. 2019). In this study, we consider the longer term family-formation and human capital consequences of age-constraining access on the previous main channel of acquiring spouses (marriage migration) among a minority population prone to marrying young.

Recent History of Migration to Denmark and Danish Migration Policy

From 1900 to 1960, Denmark generally saw more emigration than immigration, with immigrants predominantly arriving from the rest of Scandinavia, Germany, the United Kingdom, and North America (Matthiessen 2009). However, from 1960 onward, most years saw a net positive inflow of immigrants, increasingly from outside the European countries and North America. In 1974, 89,855 foreign citizens resided in Denmark, with 56% originating from European countries and North America. In 2000, the number of foreign citizens residing in Denmark had increased to 290,490, with 30% from European countries and North America. From 1974 to 2000, the Danish population grew by 6%, but the immigrant population grew by 323%, and Denmark received more immigrants per capita than, for example, Sweden and Germany. In 2002, the largest immigrant and descendant groups outside the EU countries originated from Muslim-dominated countries: Turkey, Iraq, Lebanon, former Yugoslavia, Pakistan, Somalia, and Iran.¹ Notably, many more 1.5-generation (immigrants arriving before age 15) and second-generation immigrants to Denmark from Muslim-dominated countries intramarry ethnically than those from non-Muslim-dominated countries (Qvist and Qvist forthcoming). Before 2002, 80% of male and 70% of female immigrants and descendants from non-European countries—who were aged 18–25, had lived at least 10 years in Denmark, and had married—found spouses from their parents’ origin country. Thus, a substantial amount of immigration to Denmark occurred through intramarriage or chain migration, making it (and asylum seekers and refugees) the most important driver of permanent migration to Denmark from 1973 until the early 2000s (Bauer et al. 2004).

At the turn of the millennium, non-Western immigrants experienced a low employment rate (Schultz-Nielsen 2001). Then, in 2000, the Danish government established the Think Tank on Challenges for Integration Efforts in Denmark. Several of its reports discussed integration problems and argued that the low employment rate among new citizens was financially problematic for a small country with a comprehensive and income tax–financed welfare state, presupposing high labor force participation and solidarity across population groups (Danish Ministry of Refugees, Immigrants and Integration 2004a, 2004b; Danish Ministry of the Interior 2001; Haagen Pedersen 2002). After the national election in November 2001, a right-wing government replaced a left-wing one and tightened the rules for family reunification and limited refugees’ access to Denmark while easing access for students and high-skilled immigrant workers (Hvidtfeldt and Schultz-Nielsen 2018).² These measures limited migration from countries with traditionally low labor market integration but did not affect EU citizens (except Danish citizens), whose right to family life is extensively protected by EU law.³

Stemming from human rights conventions, the right to family reunification encompasses family unity and the related right to family life (Christensen et al. 2006:127). Denmark's family reunification rules have been regulated through Section 9 of the Aliens Act for more than three decades. For spouses to access family reunification, the Danish resident must be a citizen of Denmark or a Nordic country, a refugee, or a foreigner who has held a permanent residence permit for Denmark for a specified number of years (which has varied over time). Before 2000, Danish family reunification policy had been gradually tightened. In the early 1980s, it was considered one of the most humanitarian refugee policies worldwide, but Denmark has subsequently introduced more requirements for family reunification (Bauer et al. 2004). Adamo (2022) argued that the Danish legal framework for migration has moved from “a humanitarian past” to “a restrictive present” over the last three decades.

The early 2000s then represented a watershed moment in which reforms curtailed general possibilities for migration, especially for marriage migration. Table 1 lists the main changes in the rules regarding family reunification for spouses between 2000 and 2012.⁴ The “24-year rule” states that both spouses must be aged 24 before family reunification can occur with a partner from outside Nordic or European countries. Considered the most important of these reforms, the “24-year rule” applies only to family reunification; spouses on other visas (e.g., students and green card holders) are unaffected. For an overview of other changes in the Aliens Act, see Hvidtfeldt and Schultz-Nielsen (2018) and Adamo (2022).

These four legal changes have affected cohorts at different ages, as the lexis surface in Figure 1 illustrates. The horizontal and vertical axes indicate calendar time and age, respectively; the diagonal shows the events over time for a given cohort. The white area represents the marriage regime before July 2000, when neither the “attachment rule” nor the “24-year rule” restricted family reunification for anybody. The attachment rule has affected everyone since July 2000 (highlighted in green), regardless of age. The 24-year rule was implemented in July 2002 (highlighted in orange), affecting all individuals younger than 24; the attachment rule was strengthened in the same year. The exception to the attachment rule (the 28-year rule) has been in force since January 2004, and only those younger than 28 (highlighted in yellow) are subject to this rule.

Different birth cohorts were subject to different lengths of time under the various policy regimes. Whereas the policy changes regarding spouses’ attachment relied on a civil servant evaluation about whether a potential couple is more attached to Denmark than elsewhere, the 24-year rule sets a clearly defined cutoff for determining when individuals are affected by the reform. Qualitative evidence suggests the 24-year rule is well-known among young immigrants in Denmark, but the attachment rule is less known (Schmidt et al. 2009). Previous research on the 2002 reform has generally shown that the rule likely provided the key shock to the marriage market for immigrants and descendants in Denmark (Andersen et al. 2022; Nielsen et al. 2009) and that the surrounding policies had a much smaller impact (see also Figure 2). We extend this evidence by demonstrating that the marriage rate changed for different age groups among non-European immigrants and descendants in Denmark and that this substantial change coincided with the reform.

The Reform Impact on Marriage Rates

Figure 2 shows the age-specific marriage rates (measured as the number of newly married women during a year per 1,000 unmarried women at the start of the year) for first- and second-generation non-European immigrant women compared with ethnic Danish women and the relative change in marriage rate compared with women born in 1978, the last cohort unaffected by the 24-year rule. The top panel's right graph shows that age at marriage increased across birth cohorts and marriage rates generally declined for non-European first- and second-generation immigrant women (see also Figure A1; all tables and figures designated with an “A” appear in the online appendix). However, the top panel's left graph shows that while native-born women saw a decline in marriage rates, it occurred later in life and to a lesser degree. The bottom panel shows the change in age-specific marriage rates relative to 1978 cohorts. There was a general decline for pre-1978 cohorts; for those born after 1978, the gray shading shows the part of the life course (until 24) affected by the reform for their given cohorts. The decline in marriage rates for first- and second-generation non-EU immigrant women coincided with the life course portion covered by the reform (lower bottom panel). Conversely, native-born women did not see a decline in their marriage rate before much later—a decline matching the onset and aftermath of the Great Recession in 2008–2009, when marriage rates in Denmark declined by 30% within two years.⁵ The earlier reform in 2000 or the later reform in 2004 did not substantially impact marriage rates for first- and second-generation non-EU female immigrants.

Figure A1 reports the annual marriage rate for first- and second-generation immigrants in Denmark. The annual marriage rate was affected differentially by the reforms across ages. During the 2002 reform year, the marriage rate for both men and women declined drastically: from 2001 to 2003, the marriage rate for men and women younger than 24 declined by 60% to 65%, and the marriage rate for 24- to 27-year-olds declined by 50%. No discrete change emerged for women aged 28–42, but men older than 28 saw a small discrete change, likely because of the tendency for men to marry women younger than them. Hence, the focus of this study is on women.

Potential Further Consequences of the Reform

As discussed earlier, this study focuses on an immigrant population comprising many immigrants from Muslim-dominated countries. More than any other religious group, Muslims are unlikely to report engaging in premarital sexual relationships (Adamczyk and Hayes 2012) and view premarital sex as morally unacceptable (Barber 2018; Finke and Adamczyk 2008). Using data from the Pew Research Center (2014), Figure A2 maps the relationship between the share of the population who are Muslim and the share who report viewing premarital sex as unacceptable across 40 countries. Respondents in non-European countries, particularly Muslim-majority ones, are more likely to view premarital sex as unacceptable. This finding holds for several of the most-represented countries in our sample (see Tables A1 and A2). Thus, marriage for this population likely represents an initiation (and/or intensification) of sexual relationships and the possibility of childbearing, potentially further constraining the likelihood of maintaining educational activities. Given that educational attainment most often is concluded in the early to mid-20s (OECD 1999, 2002, 2012, 2022), delaying marriage in early adulthood will likely have the most impact on educational attainment. Overall, the reform may have had a substantial effect on individuals’ life courses beyond delaying marriage.

Data and Methods

Data

We utilized full-population, deidentified administrative data from Statistics Denmark. All Danish residents are assigned a unique personal identification number at birth or on the day of immigration, enabling us to follow individuals over time and link information on demographics, education, and labor market affiliation. Data include information on immigrant and descendant status, year of migration, and own and parental origin country. Because those born before July 1, 1978, were not subject to the 24-year rule and those born after July 1, 1984, were subject to the rule from their 18th to their 24th birthday, we included cohorts born from 1972 to 1990.

We refer to children born in Denmark to parents not born in Denmark or to Danish citizens as “descendants” until at least one parent born in Denmark became a Danish citizen, those born outside of Denmark to parents who were not Danish citizens or born in Denmark as “immigrants,” and those born to at least one parent born in Denmark with Danish citizenship as “native-born.” To avoid including potential marriage migrants in the sample and ensure sufficient educational information from Danish sources for all sample members, we excluded immigrants and descendants from inside the EU and Nordic countries from the main sample (because the 24-year rule did not affect spousal migration from these countries) and first-generation immigrants arriving after their 15th birthday. Accordingly, we obtained the age at migration for first-generation immigrants and the origin country of first- and second-generation immigrants. Following the practice of Statistics Denmark, we assigned the mother's origin country (if known) to first- and second-generation immigrants. Table 2 shows the distribution of origin countries across birth cohorts. The population of women with a migration background increased across cohorts, and descendants increasingly account for more of those with a migration background. Table A1 shows the distribution across countries of origin.

Outcome Variables

We considered four outcome variables. The first is the probability of marriage at each year of age between 25 and 30 using information from the Danish Population Register, which contains day-level data on marriage dates. Because the reform mechanically limited the marriage market for those younger than 24, we began examining marriage at age 25 to determine whether the reform caused only short-term postponement. To examine whether the composition of spouses also changed or whether people forwent marriage, we included three marriage outcomes: (1) married a Danish native; (2) married an individual who was born as a non-EU second-generation immigrant in Denmark, migrated to Denmark before turning 15, or is a Danish native; and (3) married an individual born in the EU (or migrated before turning 15), including Denmark. We also considered an alternative to marriage: cohabiting in a nonmarried union. We captured this status with dwelling-level data from Statistics Denmark using their definition of cohabitation: two opposite-sex individuals who are living together, unmarried, and either (1) have a child together or (2) are not related by blood, are the only two adults in the dwelling, and have no more than a 15-year age gap.

Second, using data from the Danish population database, we measured the transition to parenthood at each age between 25 and 30 years using a binary indicator for whether the woman had had at least one child. We considered the probability of having a child while not married as an additional outcome.

Third, we captured permanent investment in human capital as the length (in months) of completed educational degrees. Using data from the Education Registry, we transformed women's highest obtained clearinghouse-approved educational degrees into the number of months of full-time study it would take to attain them. For example, a Danish high school degree would require 144 (12 × 12) months of education; an undergraduate degree would require 180 (12 × 15) months.

Finally, studies of nonlabor immigrants’ outcomes in host countries often use employment indicators to measure productive participation in society (e.g., Cangiano 2014; Dustmann et al. 2024; Hainmueller et al. 2016; Krieger 2020; Lai et al. 2022; Schultz-Nielsen 2016) and as a proxy for having better health (Lai et al. 2022). Unlike wages (for example), where employment effects on the extensive margins are difficult to distinguish from productivity on the intensive margin, an employment indicator captures the total effect on the extensive margin. Therefore, we measured the utilization of human capital using employment. With information from the Register-Based Labor Force Statistics, we obtained annual labor force status at the end of November in the years women turned ages 25–30. In Denmark, with its high de facto minimum wages and labor force participation among women, employment can be viewed as successful economic and societal integration. We generated a binary indicator equal to 1 if the main activity at the end of November was in paid employment or self-employment. Table A3 reports all outcome statistics by age, birth cohort, and immigrant status.

Analytic Strategy

where $yiA$ represent the outcomes above for each woman i at age A, BY is a vector capturing birth year, Immigrant indicates being a first- or second-generation immigrant, Generation is a vector distinguishing between first- and second-generation immigrants, Age at entry is a vector equal to 0 for everyone born in Denmark and to 0–14 for all first-generation immigrants, and Country of origin is a vector of non-EU countries. If $Immigranti×BYi$ is conditionally independent of the error term, the parameter vector $γ$ describes the reform's impact on the outcome and should be 0 only for cohorts born before July 1, 1978. We did not distinguish reform effects on first- versus second-generation immigrants. The results expressed by $γ$ were captured relative to the baseline behavior of the nonmigrant control group. The fundamental assumption for identifying a causal effect of the policy is that of parallel trends: absent policy dosage, rule-affected cohorts would have had parallel outcome trends with the Danish-born. We next discuss the threats to this assumption and how we aim to address them.

Threats to Identification

First, the share of immigrants in the full sample increases across birth cohorts, the composition of the first and second generations changes (see Table 2), and the origin-country composition of the immigrant group changes (Schultz-Nielsen and Tranæs 2010). If unaddressed, any change driven by composition changes might be misconstrued as resulting from the reform. To account for the changing composition of immigrants, we controlled for first-generation immigrant and second-generation (descendant) status, age at arrival for first-generation immigrants, and origin country.

Second, individuals might leave Denmark to marry a spouse from outside the EU, inducing collider bias (Elwert and Winship 2014): those prone to marrying may be more prone to leaving the sample, driving down the marriage rate by selecting out of the data. Bratu et al. (2020) documented an increasing rate of out-migration from Denmark to the neighboring country of Sweden following the 24-year rule. Annual out-migration increased from roughly 3% to 4% among Danish citizens with an immigrant background, completely driven by increased moves to Sweden. Within eight years, more than half of those who out-migrated had returned. The out-migration increase was concentrated among immigrants living near Copenhagen (which was connected by a bridge to Sweden in 2000). To ensure that our results would not be driven by selection out of the sample, we reestimated the main regressions while excluding people living on Zealand (where Copenhagen is located); the results were robust to the exclusion of this part of the country.

Third, differing trends between women with an immigrant background and women born to native Danish parents before the reform challenge identification. Given that Denmark has seen large-scale immigration from outside the EU since the 1970s, the processes of integration, social acclimatization, and adaption are likely ongoing, affecting intermarriage (Schwartz 2013), fertility patterns (Adserà and Ferrer 2015), and human capital accumulation (Adserà and Ferrer 2015). These secular trends could violate the parallel trends assumption underlying our identification strategy. The top panels of Figure 3 show the marriage and first-birth rates for women aged 18–32 in Denmark across migration backgrounds. They show no indication of different trends before the 24-year reform, including after the attachment rule introduction in 2000. In 2002, when the reform was introduced, the marriage rate declined drastically and then settled at a new, lower level in 2003; there is no indication of a dynamic effect in the years following the reform. However, a different issue is apparent: the Great Recession lowered marriage and birth rates (for a discussion of the latter, see Comolli et al. 2021) but only for native-born Danes. Thus, estimates of the reform's effect are likely biased toward 0 for marriage and fertility outcomes measured at older ages.

where $τ$ is a linear time trend for immigrants, and $γy$ represents birth year dummy variables for all cohorts affected by the 24-year rule. Table A4 reports the estimates for all linear trends. In Table A5, we allow for a curvilinear trend, but a linear trend suffices across 29 out of 30 regressions.

Alternative Identification of Reform Effect

The reform could be regarded as age-, period-, and cohort-level (APC) treatment for immigrants’ and descendants’ marriage likelihood, given that it happened at a specific time, had different effects on specific cohorts, and involved an age limit. Fosse and Winship (2019a, 2019b) showed that the APC problem is a linear-in-means problem and that nonlinear effects, such as those caused by policy shocks to specific cohorts at specific ages and at specific time points, are readily identifiable in observational data. To account for the compositional change over time in the sample of immigrants and descendants, we residualized the probability of being married by regressing it on the immigrant/descendant indicator, origin country, and age at migration. We present the results of the APC model in the online appendix.

Results

Marriage Behavior

Figure 4 depicts the identified marriage and relationship outcomes at each age from 25 to 30 relative to the year women turned 24 (x-axis), with the dashed vertical lines indicating the first cohort affected by the introduction of the 24-year rule (the farther to the right from the dashed vertical line, the more fully a cohort is treated from ages 18 to 24). Under the assumption of parallel trends (absent the introduction of the 24-year rule), all estimates should be interpreted as difference-in-differences estimates of changes relative to the baseline cohort (those turning 24 in 2001) and changes in the ethnic Danish population from the same cohorts. The top six panels examine the probabilities of marriage and nonmarried cohabitation. The probability of being married declined following the introduction of the reform. At age 25, women who spent their first six years of adulthood under the 24-year rule (turning 24 in 2010) saw a decline of nearly 20 percentage points in the probability of being married relative to those who turned 24 before the reform; at age 30, they still had a 6-percentage-point-lower probability of being married. We find no evidence of trend differences from the pretreatment cohorts relative to the control group before the 24-year rule or a pre-reform trend for nonmarital cohabitation. We observe an increase in the probability of cohabitation at younger ages following the reform, which decreased as these women grew older and married, ending postponement.

The bottom panels of Figure 4 examine the probability of being married to (1) a Danish native; (2) an individual who was born in Denmark, an EU, or a non-EU European country unaffected by the 24-year rule; and (3) an individual who was born in or migrated to Denmark before turning 15. The composition of spouses among those married apparently changed following the reform. The probability of marrying a spouse born in Denmark to Danish parents increased, with marriage to spouses born to native-born Danish parents accounting for up to one third of the increase and marriage to a spouse born in Denmark to immigrant parents (or who arrived before age 15) accounting for the remaining two thirds. Effects on the probability of marrying a spouse from any EU country were practically identical to the results for marrying one born in or already residing in Denmark, indicating that individuals shifted to a national rather than a European marriage market.

Pre-reform trends could affect both national marriage outcomes. Figure A3 examines the trend-adjusted outcomes, and Table A4 reports the parameters for the trend adjustment. The trend adjustment is significant for the probability of marrying a partner who was born in Denmark or who arrived before turning 15 at all ages except 29, but it is significant only at ages 25 and 26 in regard to marrying a Danish-born person. The trend adjustment drastically increases standard errors because of variance inflation, and some parameter estimates become negative. The finding that the 24-year rule led to a change in spousal composition should be interpreted cautiously, given that pre-reform developments might have trended toward an increase in finding a minority-member spouse in Denmark.

Robustness of Marriage Results

This study's findings hinge on the reform affecting marriage rates; thus, we further tested the robustness of these findings via an APC approach where we identified the nonlinear effects across age, period, and cohort. Here, we exploit the findings from Fosse and Winship (2019a, 2019b) that age, period, and cohort effects can be separated into a linear term for each and a set of nonlinearities. The reform should affect the nonlinear effects discretely (as a jump). Figure A4 displays estimates for nonlinear effects for age, period, and cohort for immigrant and descendant women, indicating that the reform provided an immediate period shock in 2002 (with no indications of a shock in 2000, when the previous reform was introduced) and a shock to the cohort effect beginning with the 1979 birth cohort (aged 23 when the reform was introduced). Accordingly, the APC results align with the findings that women postponed and abstained from marriage in their 20s (Figure 4, panel a); marriage postponement is an immediate and then declining nonlinear period effect, whereas abstaining is an increasing cohort effect (which is ultimately larger than the period effect). We tested excluding people who resided on Zealand, an island that houses Denmark's capital region and saw the largest out-migration to Sweden following the 24-year rule's implementation. This exclusion did not impact the findings on the marriage probability, as shown in Figure A5, which reproduces the findings from Figure 4 (panel a) but with larger standard errors because of the decreased sample size.

Fertility

Figure 5 reports the probability of transitioning to motherhood and having a child out of wedlock, with and without adjustment for the pre-reform trend. The probability of childbearing by the mid-20s decreased substantially, with women fully covered by the reform roughly 10 percentage points less likely to become mothers at age 25 than those not affected. No significant pretrend emerges for having entered motherhood (see Table A4), and point estimates remain similar after trend adjustment (Figure A6). This decrease lessens across age but somewhat remains at age 30, with parameters likely biased toward 0 because of the Great Recession's impact on native-born women's fertility (cf.Figure 3, panel b). Moreover, the results seemingly show an increase in the probability of becoming a mother while unmarried, but significant pretrends are evident for younger ages (25 and 27; cf.Table A4); after we adjust for pretrends, estimates move downward to 0 (cf.Figure A6). Entry into motherhood thus declined following the 24-year rule, with some indications of an increase in out-of-wedlock fertility.

Education and Employment

Figure 6 reports the results for the length (in months) of accredited education (top panel) and the probability of educational enrollment (bottom panel). We show results with and without detrending. Assuming similar trends, we see consistently higher levels of completed education among the birth cohorts who spent more of their early adulthood under the reform; once detrended, the standard errors increase dramatically because of variance inflation from high multicollinearity, and estimates approach 0. We find evidence of significant differences in trends only at ages 25 and 26 (Table A4). Further, the bottom panel of Figure 6 (using an indicator of educational enrollment) shows that constraining access to spouses from outside the EU increased educational attainment. Although the women in our study were measured at ages 25 or older, the trend remains clear: the likelihood of remaining in education is higher at ages 25–27, albeit not always significantly.

Moreover, Figure A7 and Table A4 (the latter of which reports significant pretrends for four of the six ages: 25, 26, 28, and 29) display substantial differences in employment trends before the first treated cohort. Adjusting for the differences in pretrends, we see substantial negative development in labor market attachment for the treated cohorts at age 25; the results are insignificant and closer to 0 for those who are older. Two likely explanations exist for the unexpected finding of decreased employment at age 25: (1) ongoing educational activity might delay labor force entry; and (2) the last three studied cohorts turned 25 after 2008, when the financial crisis and Great Recession might have stalled labor force entry (see also Figure 3, panel d). We thus find no evidence of long-term changes to women's labor force participation resulting from the reform.

Discussion

The literature on women's marriage postponement has predominantly focused on how increased educational enrollment and shocks to the quality of local marriage market candidates (e.g., Autor et al. 2019; Chowdhury et al. 2020; Corno et al. 2020) affect marriage age and probability. We took a different approach by studying how the marriage probability, fertility behavior, and human capital were affected when the marriage market for women who often marry young became constrained. Accordingly, we extended the literature on the effect of the legal framework on the opportunities for union formation and family reunification from outside host countries by going beyond the associational relationship and considering long-run consequences for union formation and spillovers into other family and social life spheres.

Unsurprisingly, when the marriage market became constrained, women's marriage probability declined, their marriage age increased, and their relationship behavior changed. Nonmarital cohabitation increased, women seemingly shifted their focus to the national marriage market, and fertility transitions were delayed. Lastly, educational length increased, with no indication of that increased education translating into a stronger labor market attachment.

Marriage postponement potentially increased education in the group of non-European immigrant women considered in this study and allowed women access to the new marriage market via educational attainment. However, for most affected women, the shift was not from bringing in a spouse from outside Denmark to intermarrying with a Danish spouse. Instead, although the likelihood of intermarriage increased, the main shift occurred to marrying a second- or 1.5-generation spouse already residing in Denmark. Once a substantial aspect of adaption to host-country relationship behavior occurred, nonmarital cohabitation rates increased substantially.

Research on marriage behavior changes among immigrants and other minority groups has often been concerned with availability in the local marriage market (Huschek et al. 2012; Lievens 1998; Muttarak and Heath 2010), cultural adaptation due to the length of stay in the host country (Chiswick and Houseworth 2011; Fernández and Fogli 2009; Hannemann et al. 2018; Holland and de Valk 2013; Huschek et al. 2012), and socioeconomic resources among minority individuals seeking to marry (Chiswick and Houseworth 2011; González-Ferrer 2006)—that is, how changes in the host country's opportunity structure affect marriage behavior among minority groups. We instead demonstrated what happens when the opportunity in the host country remains the same while the outside option of bringing in a spouse from abroad is constrained. We identified an increase in intermarriage to a Danish spouse, but the largest increase was to a spouse who was a minority group member already living in the host country. Because pretrends indicated somewhat that this was an ongoing process before the reform, the policy seems to have greatly decreased transnational marriage while having a lesser effect on decreasing intramarriage beyond broadly deterring women from marriage.

Although the 24-year rule likely increased education, postponing individuals’ marriage and fertility transitions through marriage market–limiting policies might have drawbacks beyond immediate discriminatory and paternalistic aspects. Per Oppenheimer's (1988) theory of marriage timing, the age at which people transition to adult economic roles and marry affects whether people forgo marriage or settle for a more poorly matched spouse. However, removing the option of bringing in an immigrant marriage partner could hasten integration into the host society by providing an exogenous excuse for finding a partner outside the group wherein search behavior is traditionally enforced (Adserà and Ferrer 2015). This proposition holds only insofar as a partner from the host country represents someone more aligned with the host country's culture than a partner from abroad. Given that time in the host country generally lowers cultural differences (Chiswick and Houseworth 2011; Fernández and Fogli 2009; Holland and de Valk 2013), the identified shift in marriage partners would likely have caused an increase in social integration.

Some limitations of this study should be noted. First, our study covered the period during and after the Great Recession, which substantially impacted marriage behavior (Figure 3, panel a) and fertility behavior (Figure 3, panel b; see also Comolli et al. 2021) among ethnic Danes. Therefore, the negative effect of the 24-year rule on marriage and fertility was likely underestimated at higher ages. The recession also greatly impacted the Danish labor market—even more for groups with weaker labor market attachment, which have historically included immigrants. Thus, the effect on labor market attachment especially at older ages might be negatively biased. For several outcomes, the assumed parallel trends underlying the identification strategy were rejected. Although we provided evidence from detrended regressions, the detrending relies on the functional form of the trend line being correctly specified. Given the limited cohorts, relying on linear trends will suffice but cannot completely rule out the presence of nonlinear trends. Finally, cohort-dependent out-migration might have compromised our sample because of the policy. If those most likely to marry migrated at high rates, we would arrive at similar findings. Relatedly, Bratu et al. (2020) demonstrated that although out-migration increased temporarily when the reform was enacted, it was geographically confined to Zealand in Denmark (particularly the Greater Copenhagen area), and most people returned within the period we studied.⁶ The results remained similar when only parts of Denmark were unaffected by out-migration, so future work should consider changing migration patterns when shocks to marriage markets occur.

Present-day Perspective

We considered a population of first- and second-generation immigrants from outside the EU and Nordic countries (non-EU+) who were no older than 15 upon arrival in Denmark for the first generation. In 2001, 7.5% of all 15-year-olds in Denmark had a non-EU+-migrant background (Table A2); for those who were 15 at the start of 2024, the equivalent share was 10.9%. An increasing population share thus remains affected by the 24-year rule that remains in place. Beyond the former Yugoslavia (broken up in 2006 following the civil war), most countries that contributed the largest number of 15-year-olds to the Danish population in 2001 were those that contributed the most 15-year-olds in 2024 (Table A2). The only exceptions were Syria and Ukraine, from which recent waves of refugees have arrived following the Syrian civil war (beginning in 2012) and the Russian invasion of Ukraine (beginning in 2022). Because the composition of the population with an immigrant background has remained quite stable, we expect findings from the 2001 reform to describe mechanisms that remain relevant in Denmark and likely across most of Europe.

Conclusion

Migration policies affect not only the inflow of immigrants into society but also the family-formation behavior of immigrants already in the country. Further, immigrants often arrive with life scripts that differ from those of the host society. Limiting access to out-of-country marriage markets by tightening rules around marriage to immigrants delays union formation and forces marriage-aged individuals to search within the local market, fostering increased adaption in union formation patterns—at the cost of constraining choices for some in society based on ethnic or national origin.

Acknowledgments

This work was funded by the ROCKWOOL Foundation (#1207). Peter Fallesen acknowledges additional funding from the Swedish Research Council for Health, Working Life and Welfare (#2016-07099). All work was conducted independently from the funders. Andreas Eklundh Sørensen provided highly competent research assistance. We thank Tim Bruckner, Alicia Adsera, Thomas Crosley, Fabrizia Mealli, Deadric Williams, Youngmin Yi, Belén Rodríguez Moro, and seminar participants at the ROCKWOOL Foundation and European University Institute for helpful comments. Previous versions were presented at annual meetings of the Population Association of America and the RC28. Authorship is shared equally across authors, and author order was randomized using the AEA author randomization tool (confirmation code X1ub2DRZuv6c).

Notes

References