The International Family Migration of Swedish-Speaking Finns: The Role of Spousal Education

Journal of Ethnic and Migration Studies 2012, 118, ifirst article The International Family Migration of Swedish-Speaking Finns: The Role of Spousal Education Jan Saarela and Fjalar Finnäs Using population register data from Finland, we study international family migration in the context of moves between two highly developed and gender-equal countries. The focus is on the Swedish-speaking population in Finland, a group with high international migration rates and good labour market prospects in the primary host country, Sweden. The influence of spousal education on emigration and return migration risks is of specific interest. We find that both the wife s and the husband s educational levels are decisive and independent determinants of migration. In families where the wife has a vocational education, the emigration risk is approximately 30 per cent lower than if she has a basic education only, and slightly greater if she is highly educated. The family s return migration is almost 40 per cent higher if the wife has a vocational education, and almost 50 per cent more if she has a higher education. These family-level estimates mirror results based on individual-level data. We argue that, unlike the case for many other countries, the gender blindness of the human-capital-based theory of family migration is not an impediment here. The results highlight that international migration, and particularly circulation as captured by return migration, has not only an individual, but also a family, dimension. Keywords: Family Emigration; Family Return Migration; Wife s Education; Husband s Education; Gender Equality; Swedish-Speaking Finns Introduction Family migration provides an essential tool for understanding the cooperative and conflicting behaviour of partnered individuals. While a single individual only has to consider his or her own interests, spouses must be concerned with the future prospects of the entire family. Jan Saarela and Fjalar Finnäs are respectively Adjunct Professor of Population Economics and Professor of Demography at Åbo Akademi University, Finland. Correspondence to: Dr J. Saarela, Åbo Akademi University, PO Box 311, FIN-65101 Vasa, Finland. E-mail: jan.saarela@abo.fi. ISSN 1369-183X print/issn 1469-9451 online/12/000001-18 # 2012 Taylor & Francis http://dx.doi.org/10.1080/1369183x.2013.733860

2 J. Saarela & F. Finnäs The most well-known theoretical framework for family migration decisions is the approach developed by Long (1974), Mincer (1978) and Sandell (1977), based on Sjastaad (1962), where migration is viewed as an investment in human capital. In the family context, migration is a joint welfare-maximisation decision of monetary and non-monetary gains, so that family, rather than personal, utility motivates the migration decision. The theory assumes that the potential gain of the husband and the wife are equally important in migration decisions, and that both partners are equally able to impose their own private interests on the family. This inherent symmetry with respect to spouses has led critics to accuse it of being gender-blind (see Nivalainen 2010). Many empirical findings contrast with the gender-neutral view of household bargaining. The husband s characteristics in terms of educational level and occupational prestige have generally been found to be more important for the family migration decision than the wife s characteristics (Boyle et al. 2001, 2003; Lichter 1980; Maxwell 1988). This shows that families are more likely to move in support of the man s career. Studies on whether the wife s education is of any relevance have produced contradictory results. Lichter (1982) and Swain and Garasky (2007), for instance, find a zero effect, whereas Shields and Shields (1993) and Pailhé and Solaz (2008) conclude that the wife s education is a significant determinant of family migration, albeit not as important as the husband s education. The gender disparities in labour market returns on migration have generally been attributed to women s lower levels of human capital and their secondary position when it comes to bargaining within the household (Shauman and Noonan 2007). Disentangling these two explanations is difficult, however, because human capital variables are closely related to a person s socio-economic position which, in turn, is probably the most decisive determinant of structural bargaining in relationships (Abraham et al. 2010). In societies characterised by gender equality and non-patriarchal families, the wife s bargaining power can be expected to be relatively high and gender differences in human capital to be modest. Smits et al. (2003, 2004) found that, in the Netherlands, the relative influence of the female partner s human capital characteristics on long-distance migration decisions has increased in importance since the 1970s. They attribute this to a societal shift towards a more-even power balance between the sexes and less-traditional gender roles. Results from the United States tell a different story, however. Tenn (2010) finds that the wife s characteristics are weak determinants of household migration, and did not increase in importance between 1960 and 2000. This low explanatory power of the wife s characteristics is argued to be the product of how married couples aggregate each spouse s migration preferences when arriving at the household migration decision. Few studies have explored international migrants on this issue, and particularly the role of each spouse s educational level on the family migration decision. One reason is that, in many societies, women are in a position where they are unable to initiate family

Journal of Ethnic and Migration Studies 3 migration. Another is that feminist concern has focused on the inequalities that women face, rather than on their bargaining potential within a couple (Raghuram 2004). Existing evidence suggests that female characteristics are of less importance in families international than internal migration decisions (e.g. Boyle 2002; Man 1995; Willis and Yeoh 2000). Yet, these studies have generally been concerned with moves between countries whose economic and social development differs greatly, where patriarchal household structures limit the re-negotiation of household tasks, or in which migration and union dissolution are strongly interrelated (Frank and Wildsmith 2005; Landale and Ogena 1995; Ortiz 1996). Third, and perhaps the most important, is the lack of reliable data on international migration flows. Few countries have a population registration system that allows researchers to analyse the international migration of families according to their characteristics and to account for the possibility of return migration. This paper contributes the first register-based study on international family migration between two highly developed and gender-equal societies. Using Finnish population register data on partnered individuals, we investigate the role of each spouse s educational level on the family s likelihood of emigration and return migration. Context In the 1970s and 1980s, there was a substantial migration flow from Finland to Sweden. It was fuelled by a shortage of jobs in Finland when the large postwar birth cohorts entered the labour market (Saarela and Finnäs 2009a). The need for labour in Sweden, together with the agreement of a common Nordic labour market, meant that 170,000 Finns migrated to Sweden*three-quarters of all migration from Finland during this period. However, over the past two decades, Sweden has lost its dominant position as the primary destination country for Finnish emigrants (see Figure 1). Our focus lies with the Swedish-speaking population of Finland. Swedish-speakers have notably different prerequisites for emigration than Finnish-speakers. Having Swedish as their mother tongue, their language skills are fully applicable in Sweden and, as a result, they have been found to integrate much better into Swedish society (Saarela and Finnäs 2007). Their employment rates and income levels have parity with native Swedes, whereas the situation of observably similar Finnish-speaking immigrants is inferior (Rooth and Saarela 2007a). Swedish-speakers in Finland live geographically concentrated on the coast, number barely 6 per cent of the total population and have a low internal migration rate. Their international migration rate, on the other hand, is three times that of Finnish-speakers and is dominated by emigration to Sweden. As illustrated by Figure 2, which is based on our dataset, emigration has a very temporary nature. Many of the emigrating families move back to Finland after a few years. The likelihood of return migration is highest during the first four years after emigration, after which time it gradually reduces. After four years, 34 per cent of

4 J. Saarela & F. Finnäs Figure 1. Annual number of migrants from and to Finland, 19712009 Sources: Institute of Migration (2006) and Statistics Finland (2011) emigrating families had returned to Finland, and an additional 12 per cent after 10 years. These numbers roughly correspond to those where the observation unit is the individual (Finnäs 2003). Reverse migration, and particularly the influence of each spouse s educational level, is here used to gain further insights into the family dimension of migration decisions. Hazard 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.00 0 1 2 3 4 5 6 7 8 9 10 11 12 13 Years abroad Figure 2. One-year risk of return migration for Swedish-speaking Finnish families

Journal of Ethnic and Migration Studies 5 Swedish-speaking society, like Finland as a whole, is in many respects very genderequal (Saarela and Finnäs 2003a, 2003b, 2006). In the 1980s, Finland was regarded as one of the best places in which a woman could live, just slightly behind top-ranked Sweden (Population Crisis Committee 1988). A recent assessment says that Finland now ranks third on gender equality, after Iceland and Norway, but in front of Sweden (Hausmann et al. 2010). Women s health, education, economic position and legal conditions have been at the top of the agenda for a long time. In 1906, the country was the first in the world to give women universal and equal rights to vote and stand for election. This started a consistent development towards equal gender opportunities. The welfare system had begun to provide women with substantial assistance in child-bearing and child-rearing, and many notable legislative gains have brought them close to full equality with men. Three-quarters of adult Finnish women work outside the home, making up almost half of the work force. In the cohorts studied here, women and men are almost equally educated; however, among younger generations, women are even better educated than men. We do not claim that full equality has been achieved. Women s wages are still only 80 per cent of those of men. They more often work part-time, with temporary job contracts, and have greater responsibility for household chores. In an international perspective, however, there is an equal power balance between the spouses in the families we studied. This implies that a wife s personal characteristics are likely to be highly influential on the likelihood of international migration. The overall aim of this paper is to investigate this issue by focusing on the effect of each spouse s educational level on the likelihood of emigration and return migration. Education and Other Migration Determinants Education is strongly related to Swedish-speakers emigration from Finland (Finnäs 2003). Emigrants tend to be selected from both the upper and the lower echelons of educational distribution, conforming to a J-shaped pattern between educational level and emigration risk (Saarela and Finnäs 2009b). The high relative emigration rate of the lowest-educated can be explained by the work opportunities provided abroad. In the 1970s and 1980s, many emigrants from Finland worked in low-skilled manufacturing industries in Sweden (Ekberg 1991; Leiniö 1984). The high emigration rate of higher-skilled individuals is presumably the result of low internal migration rates for the Swedish-speaking population. A reluctance to move within the country means fewer available career opportunities in Finland, specifically if a person wishes to work in the Swedish language. The labour market in Sweden, on the other hand, provides extensive opportunities in this respect. Many Swedish-speaking Finns in Sweden are to be found in higher-ranked occupations (Rooth and Saarela 2007a). Analyses of individual-level data say that return migrants have higher educational levels than those who remain abroad (Finnäs 2003). One possible explanation is that higher-educated emigrants have more to gain from return migration than lessereducated ones, as the return on education is higher in Finland than in Sweden (Rooth

6 J. Saarela & F. Finnäs and Saarela 2007b). The positive selection based on educational attainment in return migration flows also corresponds to the circulation of high-skilled individuals between developed countries observed elsewhere (Ouaked 2002). Recent empirical evidence indicates that temporary migration is widespread among highly skilled migrants (Mayr and Peri 2008). The worldwide movement of highly skilled workers in transnational companies, for instance, has long been known (Peixoto 2001). We have no information about the employer or the workplace of the migrants, but it is plausible that part of the emigration and return migration observed here is a result of work assignments in transnational companies. Other common determinants of family migration, which we use primarily as controls, relate to household and family structure, labour-force participation and place of residence. Migration, and international migration in particular, generally occurs at young adulthood (e.g. Saarela and Finnäs 2009b). A primary reason is that the benefits of migration are more likely to exceed the costs and inconvenience for younger people than for older ones (Sjastaad 1962). Children generally reduce the chances of opportunities for couples to make long-distance moves (Kulu and Milewski 2007), as the economic and psychological costs of moving are relatively high. The fear of disrupting school-aged children s education, and considerations about arranging day-care for younger children, may inhibit migration (Davies and Flowerdew 1992; Smits et al. 2004). A need for support from or contact with relatives can also influence the family migration decision (Michielin et al. 2008). Studies of internal migration suggest that many couples move soon after marriage, implying that union duration and migration are negatively correlated (e.g. Grundy and Fox 1985). Since both union duration and children reflect stages in the family s life-cycle, the net effect is not, however, obvious. Life-course and migration events may interact in affecting family well-being and other outcomes (De Jong and Roempke Graefe 2008). Labour force participation tends to deter long-distance migration, since ties to the current locality are stronger, particularly in families where both spouses work (Greenwood 1985; Lichter 1980). In support of this argument, unemployed individuals are found to be more prone to emigrate and to return migrate (Finnäs 2003; Saarela and Finnäs 2008). Place of residence is generally used to reflect labour market opportunities in the potential migrant s place of origin, so that migration risks increase with the area-level unemployment rate and, hence, with poorer career opportunities (Nivalainen 2004). It has been argued that, in the context of Swedish-speakers, regional variation in emigration rates may also reflect sociohistorical linkage and cognitive distance (Hedberg 2007). In the next section, we describe the variables used and give an overall outline of the data and methods applied. Data and Methods The data are based on the Finnish longitudinal census register, which includes individual information from all the quinquennial censuses between 1970 and 2000.

Journal of Ethnic and Migration Studies 7 From all persons who were present in at least one census, we have a 50 per cent random sample of Swedish-speakers, who can be identified since the register contains information about a person s mother tongue. The individual census data in the register were completed with information about the exact year of all emigrations, immigrations and deaths, the start and end of every union, and information about the partner. The data on unions, migrations and deaths were available until 2003. Before 1987, they contained information on marriages only; however, since then, consensual unions have also been included. Thus, we have information about all existing marriages in 1970, including the year of marriage, and new marriages contracted after that. We restrict our analyses to emigrations during the period 197190. One main reason is that we have no information about the country of destination, but we do know that, during this period, over 80 per cent of all Swedish-speakers who emigrated went to Sweden, and most of the others to another Nordic country (Finnäs 1995; Hedberg and Kepsu 2003). However, in the 1990s, Sweden lost much of its dominance as a destination country. The set-up also implies that all emigrants can be observed for as long as 13 years with respect to potential return migration. At present, almost all couples start their family-building career with a period of informal cohabitation and many remain unmarried but, until the 1990s, consensual unions were rare as a permanent family type (Finnäs 1993). Thus, the restriction to formal marriages for most of the study period cannot have any decisive effect on the results. The unit of observation is the family and the primary event of interest is migration. To calculate the risk or hazard of emigration, we have determined the corresponding families under risk each year. Since we want to focus on families where both spouses were of active working age (also at the time of potential return migration), we have restricted the analyses to families with men aged 2247 years and women aged 2045 years at the year of potential emigration. Over 85 per cent of all family emigrations occurred within these ages. The total number of families included in the analyses is 73,068, of whom 1,637 emigrated. The lion s share of all emigration is undertaken by young singles, but the spouses in these families account for over one-fifth of all emigration in the same age groups. The families who emigrated form the risk population for return migration. To study the return migration risk, we used Cox regressions with the length of stay abroad as the duration parameter. After 13 years almost half, or 762 families, had returned to Finland. Families who apparently separated while abroad*since only one of the spouses returned*were treated as right-censored in the analyses, although these separations were rare. Of the families where at least one spouse returned, singles amounted to only 6 per cent. We have calculated one-year migration risks, and accounted for the characteristics of both the spouses and of union-specific variables. Since the primary focus is on the

8 J. Saarela & F. Finnäs influence of each spouse s educational level on the migration risk, the other variables are used primarily as controls. The data make it possible to separate people with no further schooling above the basic level, those with vocational qualifications and those with higher education. The first category refers to people with nine years of compulsory education. The second category includes people with 23 years of additional education directed towards some specific lower-level occupation such as mechanic or orderly. The third category refers to all people with an education above and beyond the vocational level. This includes those with university entrance requirements, college educations and university degrees. The age of each spouse and the year of emigration were categorised into five-year intervals. Labour-market status refers to a situation where either both partners were employed, only the husband was unemployed, only the wife was unemployed, or both were unemployed. The region of residence consists of the four major areas in which 95 per cent of all Swedish-speakers in Finland live, plus an additional category to cover the rest of the country. Union duration separates recently established partnerships (53 years old) from those which are 49 years old, and those over nine years. The child variable captures the stages of the family s life-cycle, as it reflects the age of the oldest child. Compulsory education starts in the year when a child becomes seven years old. Here we separate families without children from those with the oldest child aged under seven, and from those with the oldest child aged at least seven. This taxonomy turned out to fit the data better than focusing on the age of the youngest child, the total number of children, or the number of children in different age groups. All variables are measured at the year of potential emigration, except for labourmarket status and region of residence, which are measured at the year of the most recent quinquennial population census prior to potential emigration. In the study of return migration, we used the same variables. Since there is no information about persons living abroad, the variables had to be measured at the time of, or before, emigration. Therefore, the estimates should be interpreted with caution. This is especially true for the child variable and labour-market status. Complementary analyses which we undertook (not shown) revealed, however, that, even if all the control variables were to be excluded from the models, the conclusions about the influence of education on the return migration risk would be the same as those discussed here. Table 1 gives the variable distributions for all families at risk of emigration, for families who emigrated, and for families who return migrated. For the variable of central interest*education*we can see that, in 40 per cent of the families, the husband has a basic education, compared with 38 per cent of the wives. About 24 per cent of the husbands have vocational qualifications and the remaining 36 per cent higher education. The corresponding figures for the wives are 28 and 34 per cent respectively. Table 2 displays the within-family educational distribution. There is clear evidence of educational homogamy: in almost 55 per cent of all the families, the husband and the wife have the same level of education. The distribution around the main diagonal is almost symmetric, however, which indicates that the marriage market is gender-neutral with respect to education.

Table 1. Variable distributions (%) Journal of Ethnic and Migration Studies 9 Families at risk of emigration Families who emigrate Families who return migrate Husband s education (basic) 40.4 30.5 21.9 Vocational 23.6 17.3 15.6 Higher 36.0 52.2 62.5 Wife s education (basic) 38.1 30.9 22.0 Vocational 27.6 19.6 19.0 Higher 34.3 49.4 59.0 Husband s age (2226 years) 10.1 21.9 24.6 2731 years 21.1 30.7 32.8 3236 years 24.4 22.0 22.4 3741 years 23.4 16.4 14.2 4247 years 20.9 9.0 6.0 Wife s age (2024 years) 9.5 20.2 20.4 2529 years 21.7 32.1 36.8 3034 years 24.9 24.1 22.0 3539 years 23.6 14.3 13.7 4045 years 20.2 9.2 7.0 Union duration (5 3 years) 19.5 36.3 39.9 49 years 28.9 31.1 36.1 9 years 51.6 32.6 24.0 Children (no) 18.4 37.3 33.3 Oldest B7 years 30.0 33.6 43.5 Oldest ] 7 years 51.6 29.1 23.1 Employment (both employed) 82.5 66.1 59.0 Only husband unemployed 4.4 9.0 9.9 Only wife unemployed 6.9 10.2 12.5 Both unemployed 6.3 14.7 18.6 Region of residence (Nyland) 39.6 42.8 44.0 Åland and Egentliga Finland 14.8 14.1 14.1 Österbotten and Mellersta 30.3 30.0 28.1 Österbotten Östra Nyland 10.3 5.2 4.6 Other areas 5.1 7.9 9.3 Observation year (197175) 25.2 19.7 16.9 197680 24.9 43.5 40.2 198185 24.9 13.3 16.1 198690 25.0 23.5 26.7 # families 73,068 1,637 762 Results Emigration The estimates of the hazard models for emigration are presented as risk ratios in Table 3. We begin by presenting models that include the educational level of one spouse only. Model 1 is with the husband s education and Model 2 with the wife s.

10 J. Saarela & F. Finnäs Table 2. Distribution of families at risk of emigration by educational level of both spouses (%) Wife s education Husband s education Basic Vocational Higher Basic 23.1 11.2 6.1 Vocational 8.7 9.0 5.9 Higher 6.7 7.4 22.3 The estimates are similar across these two models, which is a reflection of educational homogamy. As with research on individual-level data, we see a J-shaped relationship between educational level and emigration. Families in which the husband has vocational qualifications have almost 20 per cent lower emigration risks than those in which he has a basic education, whereas those in which the husband has higher education have over 50 per cent higher risks. The figures including the wife s education are 30 per cent lower and 30 per cent higher respectively. Hence, the husband s education tends to have a stronger relative effect than the wife s, and a stronger statistical power, but each variable significantly improves the model fit. Model 3 further shows that there is also a notable improvement of the log likelihood function when we add the other spouse s education level to either model. Estimates for the effect of the husband s education are practically unaffected, whereas those for the wife s education conform to a more U-shaped pattern. The influence of the wife s education on emigration might still depend on the level of the husband s education. To be assured that this is not the case, we proceed by using a model that includes an interaction between the husband s education and the wife s education. Table 4 summarises the results in terms of hazard ratios across the different potential combinations. A family in which both spouses have a basic education serves as the reference category. The emigration risk of a family in which both spouses have higher education is 1.60 of the reference family, and that of a family in which both spouses have vocational education, 0.54. The interaction improves the statistical fit of the model, but the previous conclusions remain practically the same. Roughly speaking, there is a J-shaped relationship between family emigration and the husband s education and a U-shaped relationship between family emigration and the wife s education. Hence, the wife appears to have an independent role on the family s emigration decision. Take, for instance, families in which the husband has a basic education. If the wife has vocational qualifications, the emigration risk is 0.64 that of families in the reference category, and 1.07 if she has higher education. In families where the wife has a basic education, a husband with vocational qualifications is associated with an emigration risk that is 0.94 that of families in the reference category, and 1.22 if he is higher-educated. The estimated effects of the control variables are stable across different model specifications, generally as expected, and will not be discussed at length (see Table 3). The risk of family emigration falls, notably, with both husband s age

Journal of Ethnic and Migration Studies 11 Table 3. Hazard ratios of emigration, main effects models Model 1 Model 2 Model 3 Husband s education (basic) 1 % 1 % Vocational 0.814* 0.825* Higher 1.570** 1.476** Wife s education (basic) 1 % 1 % Vocational 0.716** 0.683** Higher 1.311** 1.079 Husband s age (2226 years) 1 % 1 % 1 % 2731 years 0.870* 0.883* 0.861** 3236 years 0.733** 0.746** 0.724** 3741 years 0.754** 0.758** 0.743** 4247 years 0.536** 0.531** 0.525** Wife s age (2024 years) 1 % 1 $ 1 % 2529 years 0.863 0.877 0.839 3034 years 0.701* 0.724 0.671** 3539 years 0.504** 0.532** 0.479** 4045 years 0.455** 0.482** 0.428** Union duration (5 3 years) 1 % 1 % 1 % 49 years 0.931 0.923 0.930 9 years 1.557** 1.520** 1.553** Children (no) 1 % 1 % 1 % Oldest B7 years 0.689** 0.697** 0.706** Oldest ] 7 years 0.447** 0.449** 0.462** Employment (both employed) 1 % 1 % 1 % Only husband unemployed 1.344** 1.509** 1.351** Only wife unemployed 1.192** 1.134* 1.145* Both unemployed 1.438** 1.496** 1.362** Region of residence (Nyland) 1 % 1 % 1 % Åland and Egentliga Finland 0.915 0.929 0.946 Österbotten and Mellersta Österbotten 1.001 0.970 1.038 Östra Nyland 0.536** 0.524** 0.549** Other areas 1.317** 1.354** 1.342** Observation year (197175) 1 % 1 % 1 % 197680 2.384** 2.375** 2.413** 198185 0.770 0.765 0.783 198690 1.349** 1.331** 1.376** 2 Log Likelihood 42,826 42,870 42,789 Model Chi-Square/DF 1,354/24 1,307/24 1,396/26 # families 73,068 # migrating families 1,637 # family years (total risk time) 692,084 Notes: *p50.05; **p 50.01 as compared to the reference category; $ p 50.05; % p 50.01 for the entire variable. and wife s age. Children, and particularly older children, deter emigration, whereas unemployment increases it. Families with an unemployed husband, and those in which both spouses were unemployed, have an approximately 35 per cent higher

12 J. Saarela & F. Finnäs Table 4. Hazard ratios of emigration, results of model with interaction between husband s and wife s education Wife s education Husband s education Basic Vocational Higher Basic 1 % 0.639** 1.074 Vocational 0.936 0.539** 0.752 Higher 1.218* 1.056 1.601** 2 Log Likelihood 42,778 Model Chi-Square/DF 1,416/30 Notes: The model includes controls for husband s age, wife s age, union duration, children, employment, region of residence and observation year. *p 50.05; **p 50.01 as compared to the reference category; $ p 50.05; % p 50.01 for the entire variable. emigration risk than those in which both spouses were employed. The effect is smaller if only the wife was unemployed, which could be the result of a greater importance of the man s career on the migration decision, but also because the women are of child-bearing age. Emigration risks are the same across the three major source regions, whereas the most distant region from Sweden (Östra Nyland) is associated with the lowest emigration risk. Persons originating outside the Swedish-speakers main settlement have relatively high emigration rates, and also tend to be mobile within the country (Finnäs 1986). Variation across periods coincides with a peak in overall levels of emigration in the latter part of the 1970s, but is also indicative of a distinctive pattern for families, as the emigration risk is relatively high during the latter part of the 1980s. Couples with long union durations appear to have high emigration risks. This is an artefact of a strong correlation with the child variable. In a model without the child variable, there is no effect of long union duration on the emigration risk (not shown). Since the estimated effects of all the variables, including the child variable, are insensitive to whether or not union duration is excluded, we have chosen to present the results of the comprehensive models. Return Migration As with research based on individual-level data, education is highly related to return migration. The family s return migration risk*as shown in Table 5* increases notably with both the husband s and the wife s educational levels. If the husband s education is included together with the control variables (Model 1), the risk is 27 per cent higher for the vocationally educated and over 90 per cent greater for the higher-educated, compared to those with a basic education. Model 2 reveals that the wife s education has a somewhat lower explanatory power, but the overall pattern across levels of education is roughly similar to that for the husband s education. Families in which the wife has a vocational education have a 50 per cent

Journal of Ethnic and Migration Studies 13 Table 5. Hazard ratios of return migration, main effects models Model 1 Model 2 Model 3 Husband s education (basic) 1 % 1 % Vocational 1.270* 1.216 Higher 1.916** 1.678** Wife s education (basic) 1 % 1 % Vocational 1.502** 1.372** Higher 1.811** 1.457** Husband s age (2226 years) 1 1 1 2731 years 0.840 0.874 0.846 3236 years 0.781 0.838 0.793 3741 years 0.683 0.681 0.680 4247 years 0.500* 0.497* 0.498* Wife s age (2024 years) 1 1 1 2529 years 1.088 1.050 1.031 3034 years 0.829 0.830 0.794 3539 years 1.022 1.184 1.033 4045 years 0.958 1.169 0.993 Union duration (5 3 years) 1 % 1 % 1 % 49 years 0.933 0.905 0.919 9 years 0.604** 0.569** 0.586** Children (no) 1 % 1 % 1 % Oldest B7 years 2.016** 2.027** 2.070** Oldest ] 7 years 1.764** 1.765** 1.830** Employment (both employed) 1 1 $ 1 Only husband unemployed 1.019 1.101 1.017 Only wife unemployed 1.131 1.162 1.128 Both unemployed 1.267* 1.327** 1.243* Region of residence (Nyland) 1 1 1 Åland and Egentliga Finland 0.956 0.922 0.942 Österbotten and Mellersta Österbotten 0.806 0.801 0.824 Östra Nyland 1.089 1.116 1.162 Other areas 1.229 1.204 1.195 Observation year (197175) 1 1 1 197680 0.956 0.963 0.927 198185 1.267 1.225 1.193 198690 1.291 1.139 1.171 2 Log Likelihood 10,702 10,715 10,680 Model Chi-Square/DF 211/24 199/24 228/26 # families 1,637 # migrating families 762 # family years (total risk time) 14,590 Notes: *p50.05; **p 50.01 as compared to the reference category; $ p 50.05; % p 50.01 for the entire variable.

14 J. Saarela & F. Finnäs higher return migration risk than those where she has a basic education, whereas the risk is 81 per cent greater if she has a higher education. Including both spouses educational level (Model 3) clearly improves the model fit compared to the inclusion of either the husband s or the wife s education. The difference in return migration risks across educational levels is somewhat reduced, but the overall pattern for each spouse remains the same. Studying family-level education by allowing for interaction between the husband s and the wife s education also suggests gender neutrality with respect to the family s return migration decision (Table 6). Families in which the husband is higher educated and the wife is at least vocationally educated, or vice versa, have approximately twice the return migration risk of other families. The matrix is almost symmetric. Hence, the effect of the wife s education within a specific level of the husband s education is practically the same as the effect of the opposite scenario. Take, for instance, families where the wife has vocational qualifications. The relative effect of the husband s education here is 1.13, 1.17 and 2.41. If we instead study families in which it is the husband who has the vocational education, there is a corresponding pattern for the wife s education*1.00, 1.17 and 2.06. Thus, the effect of education on the family s return migration risk is more or less independent of which spouse is used as the index person. In spite of the relatively small number of observations, the interaction effect is close to statistically significant at the 10 per cent level. The control variables generally provide a poor fit to the models (Table 5). This was expected, as they are measured prior to emigration and the number of observations is modest. Age, region of residence and observation year display no statistically significant effects. Families without children (prior to emigration) tend to have notably lower return migration risks than those with children, whereas people who had formed a union closer to the date of emigration appear to have relatively high return migration risks. Families in which both spouses were unemployed before emigration are more prone to return migrate than those where Table 6. Hazard ratios of return migration, results of model with interaction between husband s and wife s education Wife s education Husband s education Basic Vocational Higher Basic 1 % 1.134* 1.118 Vocational 0.996 1.168* 2.056** Higher 1.308* 2.407** 2.128** 2 Log Likelihood 10,673 Model Chi-Square/DF 238/30 Notes: The model includes controls for husband s age, wife s age, union duration, children, employment, region of residence and observation year. *p 50.05; **p 50.01 as compared to the reference category; $ p 50.05; % p 50.01 for the entire variable.

Journal of Ethnic and Migration Studies 15 both were employed. Conclusions about the influence of education, as discussed above, would remain the same, even if all the control variables were to be excluded. Conclusions Numerous studies (Boyle et al. 2001, 2003; Lichter 1980; Maxwell 1988) have found that the husband s characteristics are substantially more important for the family migration decision than the wife s. This is certainly true in societies where women are less educated than men, and are considered to have a secondary position when it comes to household decision-making. We have argued that the prerequisites are notably different when it comes to studying the international migration of Swedishspeaking Finns. These Swedish-speakers in our study migrated mainly to Sweden, where their language skills were fully applicable, and where they had labour-market outcomes in parity with those of native Swedes. More important still is that they moved between two countries that are top-ranked on gender equality and in which women s equal opportunities have been at the forefront for a very long time. Under such circumstances, one would expect the wife s educational level to be a decisive and independent determinant of the family s migration decision. Our empirical results also support this hypothesis. We find that, in families where the wife has a vocational education, the emigration risk is approximately 30 per cent lower than if she has a basic education only, whereas it is slightly greater if she is highly educated. This pattern mirrors results based on individual-level data. The wife s education has an effect that goes beyond that of the husband s on the emigration risk, even though the latter provides more statistical power. There is a U-shaped relationship between the wife s education and family emigration, independent of the husband s education. For return migration risks, the effect of the wife s education is also obvious. Like the results based on individual-level data, we find the return migration propensity to increase notably with family-level education. Controlling for the husband s education, families with vocationally educated wives have an almost 40 per cent higher return-migration risk than families with wives who have completed only basic education, and those with higher-educated wives almost 50 per cent greater. Further, we see that the influence of the wife s education within each level of the husband s education is similar to the influence of the husband s education within each level of the wife s. This indicates that the educational level of both spouses is important for the return migration decision. Hence migration, and particularly circulation as captured by return migration, has not only an individual but also a family dimension. The present results should be considered as supporting previous research from other countries, as we have studied easily integrated persons who move between advanced and gender-equal countries where women are, at least in an international perspective, not in an inferior position to men. In our case, the gender neutrality of the human-capital-based theory of family migration is a minor impediment.

16 J. Saarela & F. Finnäs Acknowledgements The comments from the anonymous JEMS referees and the seminar participants at the UCL migration conference in April 2011 are gratefully acknowledged. References Abraham, M., Auspurg, K. and Hinz, T. (2010) Migration decisions within dual-earner partnerships: a test of bargaining theory, Journal of Marriage and Family, 72(4): 87692. Boyle, P.J. (2002) Population geography: transnational women on the move, Progress in Human Geography, 26(4): 53143. Boyle, P.J., Cooke, T.J., Halfacree, K.H. and Smith, D. (2001) A cross-national comparison of the impact of family migration on women s employment status, Demography, 38(2): 20113. Boyle, P.J., Cooke, T.J., Halfacree, K.H. and Smith, D. (2003) The effect of long distance family migration and motherhood on partnered women s labor market actitivites in Great Britain and the US, Environment and Planning A, 35(12): 2097114. Davies, R.B. and Flowerdew, R. (1992) Modelling migration careers, using data from a British survey, Geographical Analysis, 24(1): 3557. De Jong, G.F. and Roempke Graefe, D. (2008) Family life course transitions and the economic consequences of internal migration, Population, Space and Place, 14(4): 26782. Ekberg, J. (1991) Vad Hände Sedan? En Studie av Utrikes Födda på Arbetsmarknaden. Acta Wexionensia. Växjö: Högskolan i Växjö, Series 2 Economy and Politics, No. 21. Finnäs, F. (1986) Den Finlandssvenska Befolkningsutvecklingen 19501980. En Analys av en Språkgrupps Demografiska Utveckling och Effekten av Blandäktenskap. Helsingfors: Svenska Litteratursällskapet i Finland, unpublished doctoral thesis. Finnäs, F. (1993) Formation of Unions and Families in Finnish Cohorts Born 193867. Helsinki: Central Statistical Office of Finland. Finnäs, F. (1995) Finlandssvenskarna 1993: En Statistisk Översikt. Helsingfors: Svenska Finlands Folkting, Finlandssvensk Rapport No. 29. Finnäs, F. (2003) Migration and return migration among Swedish-speaking Finns, in Höglund, R., Jäntti, M. and Rosenqvist, G. (eds) Statistics, Econometrics and Society: Essays in Honour of Leif Nordberg. Helsinki: Statistics Finland, Research Report No. 238, 4154. Frank, R. and Wildsmith, E. (2005) The grass widows of Mexico: migration and union dissolution in a binational context, Social Forces, 83(3): 91948. Greenwood, M.J. (1985) Human migration: theory, models, and empirical studies, Journal of Regional Science, 25(4): 52144. Grundy, E.M.D. and Fox, A.J. (1985) Migration during early married life, European Journal of Population, 1(2/3): 23763. Hausmann, R., Tyson, L.D. and Zahidi, S. (2010) The Global Gender Gap Report. Geneva: World Economic Forum. Hedberg, C. (2007) Direction Sweden: migration fields and cognitive distances of Finland Swedes, Population, Space and Place, 13(6): 45570. Hedberg, C. and Kepsu, K. (2003) Migration as a cultural expression? The case of the Finland- Swedish minority s migration to Sweden, Geografiska Annaler, 85B(2): 6784. Institute of Migration (2006) Tilastot. http://www.migrationinstitute.fi/db/stat/fin/. Kulu, H. and Milewski, N. (2007) Family change and migration in the life course: an introduction, Demographic Research, 17: 56790. Landale, N.S. and Ogena, N.B. (1995) Migration and union dissolution among Puerto Rican women, International Migration Review, 29(3): 67192.

Journal of Ethnic and Migration Studies 17 Leiniö, T.-L. (1984) Inte Lika men Jämlika? Om Finländska Invandrares Levnadsförhållanden enligt Levnadsnivåundersökningarna 1968, 1974 och 1981. Stockholm: University of Stockholm, Swedish Institute for Social Research, 1981 års Levnadsnivåundersökning Rapport No. 6. Lichter, D.T. (1980) Household migration and the labor market position of married women, Social Science Research, 9(1): 8397. Lichter, D.T. (1982) The migration of dual-worker families: does the wife s job matter?, Social Science Quarterly, 63(1): 4857. Long, L. (1974) Women s labour force participation and the residential mobility of families, Social Forces, 52(3): 3428. Man, G. (1995) The experience of women in Chinese immigrant families: an inquiry into institutional and organizational processes, Asian and Pacific Migration Journal, 4(2/3): 30326. Maxwell, N.L. (1988) Economic returns to migration: marital status and gender differences, Social Science Quarterly, 69(1): 10821. Mayr, K. and Peri, G. (2008) Return Migration as a Channel of Brain Gain. Cambridge, MA: National Bureau of Economic Research Working Paper No. 14039. Michielin, F., Mulder, C.H. and Zorlu, A. (2008) Distance to parents and geographical mobility, Population, Space and Place, 14(4): 32745. Mincer, J. (1978) Family migration decisions, Journal of Political Economy, 86(5): 74973. Nivalainen, S. (2004) Determinants of family migration: short moves vs long moves, Journal of Population Economics, 17(1): 15775. Nivalainen, S. (2010) Essays on Family Migration and Geographical Mobility in Finland. Helsinki: Pellervo Economic Research, PTT Publications, No. 21. Ortiz, V. (1996) Migration and marriage among Puerto Rican women, International Migration Review, 30(2): 46084. Ouaked, S. (2002) Transatlantic roundtable on high-skilled migration and sending countries issues, International Migration, 40(4): 15366. Pailhé, A. and Solaz, A. (2008) Professional outcomes of internal migration by couples: evidence from France, Population, Space and Place, 14(4): 34763. Peixoto, J. (2001) The international mobility of highly skilled workers in transnational corporations: the macro and micro factors of the organizational migration of cadres, International Migration Review, 35(4): 103053. Population Crisis Committee (1988) Country Rankings of the Status of Women: Poor, Powerless and Pregnant. Washington DC: Population Crisis Committee, Population Briefing Paper No. 20. Raghuram, P. (2004) The difference that skills make: gender, family migration strategies and regulated labour markets, Journal of Ethnic and Migration Studies, 30(2): 30321. Rooth, D.-O. and Saarela, J. (2007a) Native language and immigrant labour market outcomes: an alternative approach to measuring the returns to language skills, Journal of International Migration and Integration, 8(2): 20721. Rooth, D.-O. and Saarela, J. (2007b) Selection in migration and return migration: evidence from micro data, Economics Letters, 94(1): 905. Saarela, J. and Finnäs, F. (2003a) Social background and education of Swedish and Finnish speakers in Finland, European Journal of Education, 38(4): 44556. Saarela, J. and Finnäs, F. (2003b) Unemployment and native language: the Finnish case, Journal of Socio-Economics, 32(1): 5980. Saarela, J. and Finnäs, F. (2006) Can the low unemployment rate of Swedish speakers in Finland be attributed to structural factors?, Journal of Socio-Economics, 35(3): 498513. Saarela, J. and Finnäs, F. (2007) Adjustment failures in an immigrant population: Finns in Sweden, Social Indicators Research, 82(4): 54563. Saarela, J. and Finnäs, F. (2008) Cross-country employment propensity of Finnish migrants: evidence from linked register data, Migration Letters, 5(1): 6377.

18 J. Saarela & F. Finnäs Saarela, J. and Finnäs, F. (2009a) Return migrant status and employment in Finland, International Journal of Manpower, 30(5): 489506. Saarela, J. and Finnäs, F. (2009b) The influence of education and socio-economic background on age-specific migration from Finland, Open Sociology Journal, 2: 428. Sandell, S. (1977) Women and the economics of family migration, Review of Economics and Statistics, 59(4): 40614. Shauman, K.A. and Noonan, M.C. (2007) Family migration and labor force outcomes: sex differences in occupational context, Social Forces, 85(4): 173564. Shields, M.P. and Shields, G.M. (1993) A theoretical and empirical analysis of family migration and household production: US 19801985, Southern Economic Journal, 59(4): 76882. Sjastaad, L.A. (1962) The costs and returns of human migration, Journal of Political Economy, 70(5): 8093. Smits, J., Mulder, C.H. and Hooimeijer, P. (2003) Changing gender roles, shifting power balance, and long-distance migration of couples, Urban Studies, 40(3): 60313. Smits, J., Mulder, C.H. and Hooimeijer, P. (2004) Migration of couples with non-employed and employed wives in the Netherlands: the changing effects of the partners characteristics, Journal of Ethnic and Migration Studies, 30(2): 283301. Statistics Finland (2011) Maahan- ja Maastamuuttaneet Lähtö: Ja Määrämaan iän ja Sukupuolen Mukaan 19872009. Http://pxweb2.stat.fi/database/StatFin/vrm/muutl/muutl_fi.asp. Swain, L.L. and Garasky, S. (2007) Migration decisions of dual-earner families: an application of multilevel modeling, Journal of Family and Economic Issues, 28(1): 15170. Tenn, S. (2010) The relative importance of the husband s and wife s characteristics in family migration, 19602000, Journal of Population Economics, 23(4): 131937. Willis, K. and Yeoh, B. (2000) Gender and transnational household strategies: Singaporean migration to China, Regional Studies, 34(3): 25364.