Incomplete Preliminary Draft! January 2006 A wage premium or penalty: Marriage migration and intermarriage effects among the children of immigrants? Aycan Çelikaksoy* JEL classification: J12, J61 Keywords: marriage premium / penalty, marriage migration, intermarriage, integration. Abstract: This article analyses the relationship between spouse type and economic assimilation for the children of guest worker immigrants in Denmark with regard to employment and wage, using panel register data for the years 1984-2001. For both males and females, we find a large marriage premium for having a native spouse on immigrant wages even after controlling for selection into employment and endogeneity of spouse type. The results also indicate that, for males, there is a marriage penalty for bringing their spouse from their country of origin as opposed to staying single. * CIM, SFI & Aarhus School of Business, Prismet, Silkeborgvej 2, DK-8000 Aarhus C, Denmark, Email: cel@asb.dk Acknowledgements: I would like to thank Nina Smith for helpful comments and suggestions. We appreciate financial support from the Graduate School for Integration, Production and Welfare and the Centre for Research in Social Integration and Marginalization (financed by the Social Science Research Council). The usual disclaimer applies.
1. Introduction The two separate topics of economic assimilation of immigrants 1 and wage differentials due to marital status 2 are of great interest to economists. However, studies at the junction of the two arrays of literature are very scarce. Most of the literature analyzing labour market integration of immigrants disregard the family s role in the assimilation process, whereas the literature on marital wage differentials mainly concentrate on white native males. In most countries Non-Western immigrants have lower employment and wage rates than the natives and their economic assimilation is a crucial issue. However, the economic fate of the children of immigrants in West European countries is yet to be decided. How these young immigrants sort themselves into household units through marriage may have very important implications for the integration of the current generation and their descendents. This paper analyzes the relationship between the spouse type of the children of guest worker immigrants in Denmark and their economic assimilation measured by their employment probabilities and wages. Rather than looking at marital status only, we investigate whether marrying a certain type of spouse have an effect on immigrant wages. Immigrants have the following spouse type options: 1.imported spouse from country of origin (marriage migration) 2.immigrant spouse who has already been living in the host country (Denmark) 3.native Dane. Since most societies face an increase of immigrants whose legal permit to enter the respective country is through marriage, the importance of the issue of marriage migration and its effect on the young immigrants in the destination countries is increasing. However, there are to our knowledge no studies on this issue. In the case of marriage to a native, most of the sociological literature 3 denotes interethnic marriage 4 as the final stage of immigrant assimilation. However, the direction of causality is an empirical question. So far, there have been very few attempts to answer this question empirically 5, though an agreement have not been reached regarding the direction of the causality. 1 See for example, Chiswick, 1978; Borjas, 1985, 1995, 1999. 2 See Ribar, 2004 for a review of the literature. 3 e.g. Gordon, 1964; Qian, 1999, 2001. 4 They define interethnic marriage, as marriage between immigrants and natives in the country of destination (host country). 5 See Meng and Gregory, 2005; Kantarevic, 2004; Duncan and Trejo, 2005.
This article investigates mainly the effect of spouse type on wages for the children of immigrants. The paper is organized as follows: Section 2 describes the theoretical background and explains our line of reasoning for the relationship of spouse type and economic outcomes. Section 3 describes the data set and variables used. Section 4 presents the empirical analyses to test our main hypotheses. Section 5 concludes. 2. Theoretical Background There are a large number of studies regarding the relationship between marital status and wages, analyzing wage differentials between married men and never married men. While most of these studies investigate the issue for white men, few of them study women or African American men. Although the relationship is not very clear in the case of female wages, all of the literature finds a positive relationship between male wages and marriage, and some of these studies establish a causal relationship. However, there are very few studies 6 trying to identify the reasons of such a causal relationship, such as intrahousehold specialization or other productivity increases for men due to marriage, as this type of analysis require a rich dataset such as time use data or employer employee records, etc. in addition to a panel register data. The existence of male marital wage premium has been primarily based on two theories: The first one is the selection effect 7, where it is argued that more productive men are also more likely to be married and that marriage itself does not have an effect on the productivity of the individual. The second is the productivity effect 8, where it is argued that marriage makes men more productive in the labour market due to various reasons; the most important one being the intrahousehold specialization 9. In the case of immigrants the wage differentials are mainly analyzed to compare immigrant and native wages or to compare the wages of different cohorts of immigrants to investigate the degree and the rate of economic assimilation for different groups of immigrants 10. The literature on economic assimilation mainly concentrates on the effects of number of years lived in the destination country and find a positive correlation between the two. One well known reason for this relationship is the language proficiency 6 See for instance, Rodgers and Stratton, 2005; Hersch and Stratton, 2000. 7 See for ex. Loh, 1995. 8 See Korenman and Neumark, 1991. 9 See Becker, 1973 for the theoretical underpinnings. 10 See Borjas, 1999 for the definitions of economic assimilation and a review of the literature.
improvements of immigrants. However, we still know very little about the factors that effect the assimilation process of immigrants and surprisingly enough the role of the family has been mostly ignored in this literature. However, in recent studies attention has been directed to analyzing immigrant labour market outcomes in a family context with an approach which combines information about an individual s own labour market opportunities and the family context in which labour market decisions take place. Incorporating the family into the economics of integration, Duleep and Sanders (1993) argue that family investment strategy is an important determinant of immigrant women s labour force participation, in addition to language proficiency, years since migration and the presence of other relatives in the household. They argue that whether a husband invests in human capital specific to the US labour market, and also the extent of that investment, are factors affecting a woman s decision to work. The labour force participation of immigrant wives is inversely associated with their husbands years since migration, holding their own years since migration constant. Baker and Benjamin (1997) also argue that (for Canada) family investment model explains the gender assimilation differences as immigrant wives take on dead-end jobs to finance their husbands investments in human capital. They conclude that family composition is an important correlate of immigrants assimilation, and the family investment model can account for many of the patterns in the data. Blau et al. (2002) estimate annual hours and log wages according to family type: immigrant family and mixed family 11. In contrast to Baker and Benjamin, who found that immigrant wives in mixed families work less than immigrant wives in immigrant families but have a positive assimilation profile unlike wives in immigrant families, Blau et al. finds that wives in both types of families work significantly less than comparable natives upon arrival and both have significantly positively sloped hours assimilation profiles. However, hours shortfall at arrival is less for wives in mixed families than immigrant families and also, their hours assimilation profile is less steep than in immigrant families. This reflects a lower initial level of human capital and labour market knowledge among immigrant wives in immigrant families than in mixed families. The wage shortfall for wives relative to natives is also less for mixed than in immigrant families upon arrival. It may be that high human capital women are sorted into mixed families, or there is a positive effect of being in a mixed marriage for these immigrant wives; the issue of causality is not tackled in the paper. 11 Immigrant family refers to families, where both spouses are immigrants and mixed family refers to families, where one spouse is an immigrant and the other a native.
The sociological literature accepts mixed marriages (an immigrant marrying a native) as the final stage of assimilation (Klein, 2001; Gordon, 1964; Hutchinson, 1957; Qian et al., 2001; Kitano & Yeung, 1982). The research by Card et al. (1998) that deals with this issue as a small part of the paper also takes mixed marriages as an indicator of a high degree of assimilation. To our knowledge, there are only two studies that look at the relationship between marriage/spouse type and economic assimilation. Kantarevic (2004) finds, for the US case, that immigrants who are married to native-born spouses assimilate faster than comparable immigrants married to foreign born spouses. However, when she accounts for selection, this effect vanishes and she concludes that immigrants who are married to native born spouses are a favourably selected sub-sample from the population of all married immigrants, which mainly supports the sociological literature arguing that more assimilated immigrants marry natives 12 rather than the spouse type having an effect on the assimilation outcomes of the immigrants. Whereas Meng and Gregory (2005) argue that intermarriage is an important input into the economic and social assimilation of immigrants in the host country. They find that there is an interethnic marriage premium after accounting for human capital differences among individuals. This article combines the above mentioned two bodies of literature and extends the above studies by using a rich register based panel dataset and by specifically analysing a narrow group of immigrants, the children of guest workers, who have lived all or most of their lives in Denmark and followed the Danish educational system from the beginning. This way we can analyse the spouse choice of these young immigrants and the consequences of their choices as labour market outcomes. According to the discussion above, the relationship between spouse type and labour market outcomes may be boiled down to three main effects: First of all, following the productivity hypothesis for the marriage premium, we may expect different labour market outcomes with regard to spouse type due to differences in intrahousehold and labour market specialization among different family types. Immigrants and natives may have different preferences for home work, market work and leisure which in turn determine their gains from each type of work. 12 The definition of a mixed marriage to a native is not the same in the literature. For instance, Kantarevic (2004) defines it as marriage to native born individuals although they may be immigrants born in the destination country (the US) and Meng and Gregory (2005) defines it as marriage to English speaking individuals although they may be immigrants born inside or outside the destination country (Australia). Definitions in this paper are different from the above as we have more precise data. See the next section.
Secondly, marrying a native spouse may lead to the accumulation of the knowledge of the customs and language of the country at a higher level increasing one s productivity and marketability in the labour market whereas, marrying a marriage migrant spouse from the country of origin may lead to a movement in the opposite direction as country of origin language will be spoken at home etc. On the other hand, following the family investment model, for Danish immigrants we may expect individuals who marry marriage migrants to have higher employment rates and wages in the start, as they are financially responsible 13 from their newly arriving spouses for marriage purposes although, individuals who marry natives or immigrants who have already been living in Denmark do not necessarily have such a legal responsibility. As a result, it is reasonable to expect some mixed effects with regard to spouse type. Thirdly, marrying a native spouse may lead to the formation of a network (strong ties) for the individual which otherwise could not have been obtained leading to special contacts and connections due to the increase of the proportion of natives in one s social circle improving his/her job prospects. Whereas, marrying a marriage migrant again may have an effect in quite an opposite direction. On the other hand, the selection effect hypothesis postulates that spouse type does not have an effect on the labour market outcomes even though the two may be related. This relationship may be due to some unobservable characteristics that affects both the labour market outcomes and spouse choice outcomes of the individual. In addition, it may also be the case that, own labour market outcomes have a determining power in the spouse type choice. This paper intends to deal with the above listed problems and shed light on the relationship between the labour market outcomes of the children of immigrants and their spouse type. 13 Especially since 1983 individuals who intend to bring their spouses from another country are obliged to bare full financial responsibility of their spouses.
3. Data The data used is based on a large dataset originating from administrative registers of Statistics Denmark. It is a panel data set, where we have yearly observations for the period 1984-2001. Our focus is on the children of guest workers from Turkey, Pakistan and Ex-Yugoslavia. Denmark imported its guest workers mainly from these three countries in the 1960s and early 1970s. In the Danish registers, it is not possible to distinguish the reason of immigration, such as labour migration, refugee migration or family reunification. The immigrant flows from Ex-Yugoslavia in the 1990s have mainly consisted of refugees. As refugees are quite a different group than that of guest workers and it is not possible to identify for Ex-Yugoslavian immigrants who is a refugee or not from the register data, we have excluded this group from the analysis. In the selected dataset we have 137,053 observations when we restrict the sample to descendents and middle generation immigrants aged 15-55 from Turkey and Pakistan. The dataset contains information on a large number of demographic and labour market characteristics of the individuals. In this dataset families are identified and spouses, mothers and fathers are merged to the main individual. The dataset enables us to identify individual transitions between different states in the marriage market and the labour market on a yearly basis. We look at the individuals who are legally married to their first spouses rather than taking into consideration second or third marriages. Legal marriages rather than cohabitation are taken into consideration for two reasons. First of all it is not possible to identify partners from the cohabiters, as this information only reveals the living arrangement rather than the relationship at stake. Secondly, cohabitation is still unaccepted by the immigrant groups studied and most partnerships are legal marriages. In addition, there has to be a legal marriage for marriage migration to take place. In this paper the focus is on the children of the guest worker immigrants. Thus we analyse both second generation immigrants 14 and middle generation immigrants. The children of guest workers who are born in Denmark are called second generation immigrants or descendents. We define middle generation immigrants as the children of immigrants, who are not born in Denmark and who have immigrated to Denmark before the age of eight. We start following the individuals when they are 15 years old or older and we follow them through the years 1984-2001. 14 Second generation immigrants are immigrants who are born in Denmark (the country of destination) to immigrant parents.
The immigrants have the following spouse type options: 1.imported spouse from country of origin (marriage migration) 15 2. native Dane 3. immigrant spouse who has already been living in the host country (Denmark). The reason for such a distinction of spouse type is to see whether marrying a native has a different effect than marrying a spouse with any immigrant origin but who has been living in Denmark as apposed to a spouse who have migrated for marriage purposes. In addition, integrating the third category into either spouse type option one or two would not be entirely correct. Table1 shows sample means with regard to sex and spouse type for the children of guest workers who are between the ages 15-55. It can be seen that mean log hourly wage rate for females who are married to native Danes is 4.80 whereas; it is 4.70 for females who are married to marriage migrants. Although the difference is not as large as expected, we can see a greater difference in the case of employment. It can be seen that 72 percent of females who are married to native Danes are employed on average as opposed to 44 percent of females who are married to marriage migrants, which may indicate the importance of taking selection into employment 16 into account. Both education and experience levels are much higher for males and females who are married to natives relative to other types of spouses and singles. 15 A person is defined as a marriage migrant (imported spouse) if he or she has migrated before or after two years from his/her marriage date. 16 In a country like Denmark with a compressed wage structure and high unemployment benefits selection into employment rather than into the labour force is taken into account.
Table 1. Sample means for the children of guest workers aged 15-55 by sex and spouse type spouse type for males marriage migrant native immigrant single for females marriage migrant native immigrant variable single country of origin (Turkey) 0.60 0.61 0.44 0.64 0.63 0.76 0.72 0.83 middle generation 0.38 0.88 0.96 0.98 0.33 0.69 0.74 0.73 log hourly wage 4.34 4.84 4.89 4.78 4.31 4.70 4.80 4.66 employment 0.43 0.49 0.63 0.41 0.40 0.44 0.72 0.41 experience 1.41 4.77 7.11 5.37 1.15 2.76 5.49 2.90 experience square 5.17 37.40 74.02 43.58 3.08 12.65 54.53 14.89 education 9.60 10.25 10.81 8.86 9.56 10.16 12.28 9.68 nb. of child. (under 6) 0.12 0.88 0.37 0.37 0.19 1.04 0.47 0.83 age 19.26 34.62 44.29 45.83 18.23 24.58 33.37 28.80 age square 394.24 1326.87 2022.03 2164.31 341.18 620.93 1203.16 925.10 sex ratio 0.87 0.86 0.80 0.82 1.15 1.12 1.18 1.14 relative group size 0.34 0.33 0.23 0.28 0.40 0.46 0.41 0.45 Number of obs. 42835 14328 1866 20394 34419 9413 229 3202 3.1 Spouse type among the children of guest worker immigrants The initial guest workers who migrated to West European countries in the 1960s and 1970s were mainly men; as a result it is not surprising that the first generation brought their existing wives or fiancées to the destination countries. Thus we are interested mainly in the marriage 17 behaviour of middle generation immigrants and descendents, which is rather a very different group from the initial guest workers. These groups of immigrants have lived all or almost all their lives in Denmark, have followed Danish educational system from the beginning and are expected to be closer to the Danish culture than their parents. As a result, it could be reasonable to expect more and more children of immigrants to marry Danes and/or each other and assimilate to the Danish norms. However, it is not quite the case, where most of the marriages that take place in most 17 However, it should be noted that in 2001 for ex. 69 percent of males and 67 percent of females are single.
West-European countries are to spouses who arrive for marriage purposes 18, which have made this type of marriages an important issue in the media and the political debate in these countries, making the various consequences of such marriages an important topic to analyse. It can be seen in Figure1 that starting from 1988 for females and 1997 for males, a majority have been married to marriage migrants rather than any other spouse type and by year 2001, 78 percent of females and 69 percent of males were married to marriage migrants, where only 1 percent of females and 3 percent of males were married to native Danes 19. The high proportion of marriages to marriage migrants and the very low proportion of marriages to native Danes are quite striking especially when taking into consideration that these are the children of immigrants who are expected to be relatively more assimilated when compared to the initial guest workers. There are various possible reasons for the existence and increase of marriages to marriage migrants (Çelikaksoy, 2005), where one of the most important reasons is the sex ratio imbalances in the marriage market. One of the most important structural factors of a marriage market is the sex ratio. It is generally defined as the number of men divided by the number of women and it is one of the most important determinants of how people sort themselves into marriages in that marriage market, as have been argued 20 by most sociologists and demographers. It becomes a crucial factor especially when studying ethnic marriages. To be more precise, sex ratios in this paper are defined according to sex and country of origin, year and within a specific age group: 15-55, such that sex ratio for males from Turkey is: f m nt f SRt =, where n m t is the number of females from Turkey aged 15-55 at year t, and nt n is the number of males from Turkey aged 15-55 at year t, and vice versa for females. m t 18 See Çelikaksoy, 2005. 19 In this study native Danes are defined as people born in Denmark to native parents, that is, the definitions are made according to country of birth and parental origin, rather than citizenship or country of birth alone as a result we can distinguish between descendents and natives. In most studies these two categories can not be distinguished from each other which in our opinion is an important conceptual problem. 20 See Groves and Ogburn (1928) and Cox (1940) for the earliest empirical studies.
Figure 1. Proportion of different spouse types among the married 21 children of guest workers, age 15-55. 90% 80% 70% 60% 50% 40% 30% marriage migrants native Danes immigrants 20% 10% 0% 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Males Females In the literature of interethnic marriages it is argued that there are two main factors that determine the rate of intermarriages; group size and sex ratio. The rationale behind this is quite simple; as the group size gets smaller, that is, the size of an ethnic group in the destination country gets smaller, the higher the probability that they will marry natives. In this paper the probability of within marriage is assumed to depend on both relative group size and sex ratio. Group size is defined with regard to country of origin, sex, age and year, such that the relative group size variable for males from Turkey is: f m nt Pit ( M) =, where N f t f t f nt is the number of females from Turkey, aged 15-55 at year t, and N is the total number of females in Denmark, aged 15-55 at year t. The rationale behind sex ratio imbalances is similar as well. It is expected that as the sex ratio for males gets low, that is, the number of females marriageable to males within a 21 See the Appendix for a figure including the singles as well.
group gets low and as a result, the more likely the males to marry outside that ethnic group. However, on contrary to the studies on intermarriages, in this paper, we also model explicitly immigrants who marry marriage migrants in addition to other types of spouses. Thus immigrants in this case have more options rather than only marrying Danes, when faced with small relative group size or sex ratio imbalances, such as bringing their spouse from their country of origin. Table 2 shows the sex ratios for the children of guest workers as described above. It can be seen that it is quite unbalanced 22, especially in the first half of the period, for ex. there are 66 females per 100 males in 1984. Table 2. Sex ratios by gender for the children of guest worker Turkey Pakistan for males for females for males for females 1984 0.69 1.45 0.62 1.62 1985 0.72 1.39 0.65 1.55 1986 0.74 1.35 0.66 1.51 1987 0.77 1.30 0.67 1.48 1988 0.81 1.24 0.71 1.42 1989 0.84 1.19 0.74 1.36 1990 0.86 1.16 0.76 1.31 1991 0.88 1.13 0.79 1.27 1992 0.89 1.12 0.80 1.25 1993 0.90 1.12 0.82 1.22 1994 0.90 1.11 0.84 1.19 1995 0.91 1.10 0.85 1.18 1996 0.91 1.10 0.86 1.17 1997 0.92 1.09 0.87 1.16 1998 0.92 1.08 0.89 1.14 1999 0.92 1.09 0.90 1.11 2000 0.91 1.09 0.92 1.09 2001 0.91 1.09 0.93 1.08 22 This can be a consequence of the fact that guest workers were mainly males.
4. Empirical Strategy The empirical section is composed of two parts where, in the first subsection we analyse two marriage types as opposed to being single, as this is the way immigrant marriages are classified in most studies as discussed in the previous sections. In this section, two marriage types, namely, exogamous 23 and endogamous marriages will be analysed in relation to employment and wages. However, in the second subsection marriage type will be decomposed further into total four categories, namely, singles, marriages to marriage migrants, native Danes and immigrants who have already been living in Denmark. Different methods of analysis will be used in the different subsections. 4.1 Endogamous versus exogamous marriages The analysis is based on the estimation of a traditional human capital wage function (1) ln W it = β t + Xitβ + M it β M + ε it where X it is a vector of explanatory variables: experience, experience squared, years of education; and Mit is a vector of marital types (indicators for being single, being endogamously or exogamously married). The subscripts i and t index the individual and 2 time, respectively, and εit is an error component, Nid(0, σ ). Following Meng and Gregory (2005), we begin by estimating a pooled OLS, where we assume that marital type is exogenous. It can be seen in Table 3 that experience and education has a significant positive effect on wages as expected. An extra year of schooling brings a wage increase of 1.6% for males and 2.9% for females. When we look at the coefficients of the different marriage types, we can see that endogamous marriages are associated with a wage premium of 15% for both males and females whereas; exogamous marriages are associated with a marriage premium of 19% for males and females relative to singles. Our results are similar to the results of Meng and Gregory at this stage, in the sense that they also find a higher premium for exogamous marriages. They find an endogamous marriage premium of 9.9% and an exogamous marriage premium of 14.7% indicating a 4.8% intermarriage premium when compared with those who are endogamously married. Our results indicate an intermarriage premium of about ε 23 Exogamous marriages are defined as marriages to natives, that is, native Danes in this case, whereas endogamous marriages are defined as marriages among immigrants in Denmark, that is, non-natives.
4% when compared with those who are endogamously 24 married. In addition, the Wald F-tests indicate that the differences between the marriage types are statistically significant, although that is not the case for women. However, there are various potential problems with the current analysis. First of all, following the relevant literature on marriage premium, as discussed in Section 2, we have to take account of the potential endogeneity of marriage type. In addition, the formal tests indicate that spouse type is endogenous in our current wage equation. However, we also need to take account of other potential problems such as unobserved heterogeneity and selection into employment, which will be discussed in the next subsection. To deal with the above mentioned problems we follow the methodology of Wooldridge (2002). The first equation 25 below is a fixed effects model for log wages, where α i is the unobserved 2 heterogeneity term, assumed to be individual-specific, time invariant and Nid(0, σ α ). This equation is the main equation of interest. Equations 2 and 3 are linear projections for the endogenous marriage type variable and the forth equation is the selection equation, where we observe wages only when the individual is employed 26, that is, when S it = 1. (1 ) ln W it = β t + Xitβ + M it β M + λ itβ λ+ αi+ εit (2) M 2it = π 2t + Zitπ 21+ αi2+ u2it (3) 3it 3t it 31 i3 3it M = π + Z π + α + u (4) Sit = 1( Zitδ + Vit > 0) where we use standard probit for S it = 1 such that, PS ( it = 1\ Zit) =Φ ( Zδ it ) to compute λ i = φ( δ' Zit)/ Φ ( δ' Zit). To estimate equations 2 and 3 we use two instrumental variables, namely sex ratio and relative group size, which will be introduced in the next section below. The identifying assumptions are untestable when there is only one instrument for each endogenous variable. The instruments are valid because it is self evident that sex-ratio and relative group size that they only influence wages through their effect on marriage behaviour. In addition, we need some variables that affect employment but not wage offer. For this purpose we use number of children under age 7, age and age 24 However, it should be noted again that our definitions of marriage types are different, please see the explanations in the previous sections. 25 We don t go into detail regarding any of these equations in this section; we just deal with them very superficially. This will be dealt with in the next version. 26 In a country like Denmark with a compressed wage structure and high unemployment benefits selection into employment rather than into the labour force is taken into account.
squared. All variables except marriage type and wage offer are treated as exogenous. Following Wooldridge (2002), all exogenous variables, which are denoted as Zit in the above equations, appear in the selection equation, and all are listed as instruments in estimating marriage type in equation 1 by 2SLS. The results can be seen in Table 3. Table 3. Estimation of hourly wage function for the children of guest workers by 2SLS. Selected coefficients. (1) (2) Pooled OLS FE Exogenous spouse type Endogenous spouse type Selection into employment Males Females Males Females Variables Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Experience 0.099 0.002 0.095 0.004 0.103 0.023 0.143 0.033 Experience, squared/1000-0.004 0.000-0.004 0.000-0.003 0.001-0.007 0.001 Education 0.016 0.001 0.029 0.002 0.062 0.008 0.062 0.007 Marriage Type: Endogamous 0.149 0.008 0.153 0.009-0.212 0.094-0.369 0.162 Exogamous 0.187 0.016 0.192 0.039 11.362 1.535 7.899 2.726 Lambda 0.309 0.108 0.354 0.127 Constant 3.925 0.017 3.630 0.053 3.070 0.073 3.178 0.103 Wald F -test for Endog=Exog p 7.47 1.03 0.01 0.31 53.62 8.96 0.00 0.00 R-squared 0.25 0.19 N 29384 15369 29384 15369 figures in bold indicate significance at the one percent level; figures in italic indicate significance at the 5 percent level. all standard errors are in italic Note 1: All models include year dummies. Note 2: standard errors in the second estimation method have not been corrected It can be seen in Table 3, in the second estimation method that marrying endogamously is associated with a wage penalty of 21% for males and 37% for females and a very large wage premium for marrying exogamously. Although these coefficients are too large,
which is suspicious and will be looked into thoroughly for possible multicollinearity problems etc., they are in the same direction as the existing studies such as Meng and Gregory (2005), who find a 15% penalty for females marrying endogamously and a 31% premium for exogamous marriages. 4.2 Marriage migration effects analysed separately In this section we decompose spouse type categories further into marriage migrant spouses and other immigrant spouses. First we estimate a pooled OLS, where spouse type is assumed to be exogenous. It can be seen in Table 4 that the results are similar to the results from the previous section except that the premium for being married to a marriage migrant for males is 14.6% and 15.4% for being married to another type of immigrant, where the Wald F-tests indicate that the differences between the spouse types are statistically significant for males. The results for females are 15.1% and 16% respectively. We can also see that marrying a native is associated with a large positive significant wage premium of 19% for both males and females relative to singles. However, there are various potential problems with the current analysis. These problems will be discussed and tried to be dealt with below. Omitted variables: This problem is especially important in marriage premium studies, as it is the case that some unobservable individual characteristics are likely to appeal to both employers and potential spouses (such as cultural assimilation or good looks), and maybe especially to certain types of potential spouses. To deal with this problem we use a fixed effects 27 method controlling for unobserved individual specific effects. It can be seen in Table B1, under method (2), that the marriage premium is decreased to 7% for males and to 6.1% for females who are married to marriage migrants. Whereas, for males, marrying a native has a premium of 6.6% and 3.9% for marrying an immigrant, but with a significance at the 10% level this time instead of a significance at the 1% level. The other spouse types are not significant for females. Thus, these results show the importance of the problem and may indicate that individuals marrying especially natives and immigrants have favourable characteristics that are valued by both employers and potential spouses relative to marriage migrants. Endogeneity of spouse type: As also discussed in the previous sections, the direction of causality between earnings/wages and marital status/marital type is debated in the literature. Marital type is assumed to be exogenous in the previous methods however, it is likely that high wage individuals are more attractive potential spouses and as a result are 27 It can be noted that the RE method is rejected by the Hausman tests.
more likely to be married and maybe individuals with relatively higher wages are more likely to be married especially to certain types of spouses. The problem of the endogeneity 28 of marital type in our wage equation still remains to be tackled with. Thus, the previous results are biased and inconsistent. To deal with this problem we use instrumental variables method. First, we estimate marital type with pooled multinomial logit 29 using two instruments 30, namely, sex ratio and relative group size. Instruments: 1. Country of origin, age, year and gender specific sex ratios: SR m it n = for males and n f t m t SR f it n = for females n m t f t 2. Country of origin, age, year and gender specific relative group size: f m m nt f nt Pit ( M) = for males and Pit ( M) = for females N N f t m t The instruments are strong, as they have a large significant effect in the multinomial logit model, and from the appendix we can see that they do have a very strong effect. However, the identification problems we have in this section will be discussed in the next version of the paper. Regarding these instruments, it is hard to imagine that an individual s wages would depend on the number of individuals of the opposite sex in the same age-ethnic group in that year as a ratio of the total stock of the opposite sex of marriageable age. The same rationale 31 goes for the sex ratios as well. The first stage estimation of the pooled multinomial logit for spouse type can be seen in Appendix Table B2. 28 It can be noted that formal tests indicate that spouse type is endogenous. 29 J j β Mj β h h= 1 Py ( M X) exp( X ) /[1 exp( X )] = = +, where j M indicates marital type, which takes four values, singles (the base category), marriages to marriage migrants, native Danes, and other immigrants. 30 For details of the instrumental variables see Section 3.1. 31 However, it should be noted that the IVs are not region specific, which could have been problematic, as it could reflect internal migration differences due to better job opportunities.
For females education has a negative significant effect on marrying a marriage migrant and other immigrants, whereas it has a positive effect on marrying a Dane. Regarding the instruments, relative group size has a negative effect on marrying a Dane for males and a positive effect on marrying immigrants in Denmark for females. The estimated predicted probabilities are used instead of the actual spouse types in the second stage wage equation, where the results can be seen again in Table B1. The descriptives for the actual and the predicted probabilities of spouse types can be seen below in Table 4. Table 4. Descriptives for the actual and estimated spouse types. Males Females Males Females Males Females Spouse type: actual marriage migrant 0.18 0.20 0.39 0.00 1.00 0.00 1.00 native Dane 0.02 0.00 0.15 0.00 1.00 0.00 1.00 other immigrants 0.26 0.07 0.44 0.00 1.00 0.00 1.00 Spouse type: predicted marriage migrant 0.25 0.35 0.12 0.18 0.05 0.75 0.00 0.71 native Dane 0.03 0.01 0.04 0.03 0.00 0.34 0.00 0.82 other immigrants 0.37 0.11 0.26 0.08 0.01 0.85 0.00 0.77 N (actual) 79423 47263 N (predicted) 79423 47263 Mean Std. Dev. Min - Max values Selection bias: In our wage equation only the sample of wage earners are used, which causes a selection bias especially in the case of women because of the lower rate of employment among immigrant women. To deal with this problem we will try to control for selection into employment 33. First we estimate a pooled probit 34 model for being 33 In a country like Denmark, where there is a compressed wage structure and high unemployment benefits; it makes more sense to take account of selection into employment rather that into labour force. 34 The results can be seen in Appendix Table B4.
employed then use the computed Inverse Mills ratio in our wage equation. Table B3 in the Appendix shows that there is a huge significant positive effect of having a Danish spouse on the likelihood of being employed for both males and females. Using standard probit for S it = 1 such that PS ( it = 1\ Wi) =Φ ( Wδ i ), compute λ i = φ( δ' Wi) / Φ ( δ' Wi) and estimate ln W = β + X β + M β + λ β + α + ε (4) it t it it M it λ i it where α i is the unobserved heterogeneity term, assumed to be individual-specific, time 2 invariant and Nid(0, σ α ). The results of the above fixed effects model can be seen below in Table 5, after taking account of individual specific unobserved heterogeneity, endogeneity of spouse type and selection into employment. The identifying assumption is that the number of children below age 7 only influence wages through their effect on employment. Mit is a vector of marital types (indicators for being single, having a marriage migrant spouse, Danish spouse, and immigrant spouse). It can be seen, regarding spouse type, that for males marriage to a marriage migrant is associated with a wage penalty of 18.4%, whereas marriage to a native Dane is associated with a very large wage premium. In the case of females as well we find a very large premium for marrying Danes, where the premium for marrying marriage migrants are much lower, indicating a large intermarriage premium when compared with those who are married to marriage migrants. However, again we have too large coefficients, which will be investigated thoroughly.
Table 5. Estimation of hourly wage function for the children of guest workers. Selected coefficients. (1) (2) Pooled OLS FE Exogenous spouse type Endogenous spouse type Selection into employment Males Females Males Females Variables Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Experience 0.099 0.002 0.095 0.004-0.023 0.008-0.020 0.021 Experience, squared -0.004 0.000-0.005 0.000 0.001 0.000 0.000 0.001 Education 0.016 0.001 0.029 0.002 0.124 0.004 0.074 0.006 Spouse Type: Marriage Migrant 0.146 0.008 0.151 0.100-0.184 0.065 0.564 0.118 Dane 0.189 0.016 0.193 0.039 3.139 0.324 1.602 0.369 Immigrant 0.154 0.009 0.160 0.015 2.150 0.111 0.969 0.229 Lambda 0.500 0.039 0.093 0.029 Constant 3.922 0.017 3.627 0.054 1.746 0.081 2.839 0.110 Wald F -test for M2=M3=M4 p 4.15 0.67 194.59 5.01 0.02 0.51 0.00 0.01 R-squared N 0.25 0.19 29384 15369 0.28 0.22 29384 15369 figures in bold indicate significance at the one percent level, figures in italic indicate significance at the 10 percent level. all standard errors are in italic Note 1: All models include year dummies. Note 2: the standard errors for method (2) will be corrected in the next version 5. Conclusion Although our empirical analysis is far from being complete, our results indicate that spouse type has an important effect on immigrant wages and that marrying a native Dane is associated with a very large wage premium when compared to singles and other spouse types, whereas marrying a marriage migrant is associated with a penalty for males.
References Baker, M. and Benjamin, D. (1997) The Role of the Family in Immigrants Labor-Market Activity: An Evaluation of Alternative Explanation, American Economic Review, 87 (4): 705-727. Borjas, George J. (1995) Assimilation and changes in cohort quality revisited: What happened to immigrants earnings in the 1980s?, Journal of Labor Economics 13(2): 201-45. Borjas, George J. (1999) Economic analysis of immigration, in Orley Ashenfelter and David Card (eds.) Handbook of labor economics 3A:1697-1760, Elsevier Science, Amsterdam. Çelikakcoy, A. (2005) Marriage Migration: The case of Denmark Draft Çelikakcoy, A., H. S. Nielsen, M. Verner (forthcoming) Marriage Migration: Just another case of positive assortative matching?, Review of Economics of the Household. Chiswick, Barry R. (1978) The effect of Americanization on the earnings of foreign-born men, Journal of Political Economy 86(5): 897-921. Duleep, H. O. and Sanders, S. (1993) The decision to work by married immigrant women, Industrial and Labor Relations Review, 46 (4): 677-690. Duncan, Brian and Trejo, Stephen J. (2005) Ethnic identification, intermarriage, and unmeasured progress by Mexican Americans, IZA discussion paper series No. 1629. Gordon, M. M. (1964) Assimilation in American Life, Oxford University Press, New York. Hersch, J. and Stratton, L. S. (2000) Household specialization and the male marriage wage premium, Industrial and Labor Relations Review, 54(1): 78-94. Kantarevic, J. (2004) Interethnic marriages and economic assimilation of immigrants, IZA discussion paper series, No. 1142. Klein, T. (2001) Intermarriage between Germans and Foreigners in Germany, Journal of Comparative Family Studies, Calgary, Summer, 32 (3) pp: 325-347. Meng, Xin and Gregory, Robert G. (2005) Intermarriage and the Economic Assimilation of Immigrants, Journal of Labor Economics 23(1). Qian, Z. (1999) Who Intermarries? Education, Nativity, Region, and Interracial Marriage, 1980 and 1990. Journal of Comparative Family Studies 30(4): 579-597. Qian, Z.; Blair, S. L. and Ruf, S. D. (2001) Asian American interracial and interethnic marriages: Differences by education and nativity, The International Migration Review, 35(2): 557-586.
Ribar, David C. (2004) What do social scientists know about the benefits of marriage? A review of quantitative methodologies, IZA Discussion Pape Seriesr No. 998, Bonn. Rodgers, William M. and Stratton, Leslie S. (2005) The male marital wage differential: Race, training, and fixed effects, IZA discussion paper series, No. 1745.
Appendix A. Figure A1. Proportion of different marital types among the children of guest workers, age 15-55. 90% 80% 70% 60% 50% 40% 30% singles marriage migrants native Danes immigrants 20% 10% 0% 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Males Females
Appendix B. Table B1. Estimation of hourly wages for the children of guest workers, age 15-55. Selected coefficients. (1) (2) (3) (4) Pooled OLS Exogenous spouse type FE Exogenous spouse type FE Endogenous spouse type FE Endogenous spouse type Selection into employment Males Females Males Females Males Females Males Females Variables Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Experience 0.099 0.002 0.095 0.004 0.065 0.004 0.074 0.008-0.036 0.008-0.022 0.021-0.023 0.008-0.020 0.021 Experience, squared/1000-0.004 0.000-0.005 0.000-0.003 0.000-0.004 0.000 0.001 0.000 0.000 0.001 0.001 0.000 0.000 0.001 Education 0.016 0.001 0.029 0.002 0.089 0.003 0.058 0.004 0.107 0.004 0.070 0.005 0.124 0.004 0.074 0.006 Spouse Type: Marriage Migrant 0.146 0.008 0.151 0.100 0.070 0.013 0.061 0.016 0.079 0.062 0.527 0.118-0.184 0.065 0.564 0.118 Dane 0.189 0.016 0.193 0.039 0.066 0.040 0.086 0.073 1.861 0.309 1.478 0.367 3.139 0.324 1.602 0.369 Immigrant 0.154 0.009 0.160 0.015 0.039 0.021 0.044 0.028 1.328 0.091 0.899 0.229 2.150 0.111 0.969 0.229 Lambda 0.500 0.039 0.093 0.029 Constant 3.922 0.017 3.627 0.054 3.234 0.033 3.236 0.068 2.450 0.060 2.876 0.110 1.746 0.081 2.839 0.110 Wald F -test for M2=M3=M4 p 4.15 0.02 R-squared N 0.25 0.19 0.27 0.22 29384 15369 29384 15369 0.27 0.22 29384 15369 figures in bold indicate significance at the one percent level, figures in italic indicate significance at the 10 percent level. all standar errors are in italic Note 1: All models include year dummies. Note 2: standard errors are to be corrected 0.28 0.22 29384 15369
Table B2. Estimation of a multinomial logit model for marital type for the children of guest workers, age 15-55. Spouse type marriage migrant native Dane other immigrant Males Females Males Females Males Females Variables Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Coeff. S.E. Experience 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Experience, squared/1000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Education 0.000 0.006-0.129 0.010 0.034 0.010 0.213 0.032-0.153 0.006-0.194 0.014 sex ratio -4.923 0.587-2.890 0.587-0.527 1.164-0.566 2.004-1.007 0.610-4.054 0.846 relative group size 1.211 1.127 0.583 0.124-1.400 0.359 0.454 0.604-0.207 0.153 0.953 0.210 Constant 2.497 0.402 2.823 0.929-2.026 0.766-4.555 3.091 2.321 0.413 6.291 1.287 Log Likelihood Pseudo R-squared N -46484-20253 -46484-20253 -46484-20253 0.26 0.15 0.26 0.15 0.26 0.15 53074 24366 53074 24366 53074 24366 figures in bold indicate significance at the one percent level, figures in italic indicate significance at the 10 percent level. Note 1: All models include year dummies.
Table B3. Descriptives for the actual and estimated spouse types. Males Females Males Females Males Females Spouse type: actual m arriage m igrant 0.18 0.20 0.39 0.00 1.00 0.00 1.00 native Dane 0.02 0.00 0.15 0.00 1.00 0.00 1.00 other immigrants 0.26 0.07 0.44 0.00 1.00 0.00 1.00 Spouse type: predicted m arriage m igrant 0.25 0.35 0.12 0.18 0.05 0.75 0.00 0.71 native Dane 0.03 0.01 0.04 0.03 0.00 0.34 0.00 0.82 other immigrants 0.37 0.11 0.26 0.08 0.01 0.85 0.00 0.77 N (actual) 79423 47263 N (predicted) 79423 47263 Mean Std. Dev. Min - Max values Table B4. Estimation of a probit model for being employed for the children of guest workers, age 15-55. Males Females Variables Coeff. S.E. Coeff. S.E. Education 0.083 0.005 0.104 0.007 Spouse type: marriage migrant -0.797 0.065 1.084 0.089 native Dane 11.991 0.300 7.723 0.719 other immigrants 3.766 0.071 3.740 0.254 Nr. of child. under age 7-0.080 0.009-0.448 0.013 age -0.199 0.005-0.307 0.010 age, squared 0.002 0.000 0.004 0.000 constant 1.645 0.087 3.175 0.202 Log Likelihood Pseudo R-squared N -32941-14718 0.20 0.16 59831 26512 figures in bold indicate significance at the one percent level, Note 1: All models include year dummies.