ETHNIC INTERMARRIAGE IN TIMES OF SOCIAL CHANGE: THE CASE OF LATVIA*

In the social science literature, intermarriage is considered to be a major indicator of the quality of the relationships, or "social distance," among groups and the "cohesion" of societies. The existence of mixed marriages between members of different groups indicates that there may also be positive contacts of another nature, like friendship and work relationships, and that the members of the groups consider each other to be social equals (Gndz-Hosgr and Smits 2002; Kalmijn 1998). From a social cohesion point of view, mixed marriages are important because they link not only two individuals but also the larger groups to which these individuals belong. In a world full of ethnic, racial, and religious cleavages, intermarriage between members of different groups is one of the most important connecting elements. High levels of intermarriage are therefore expected to reduce the probability of violent conflicts among social groups (e.g., Blau and Schwartz 1984; Merton 1941). This double-sided character of intermarriage, as an indicator of the social distance among groups and a cohesion-producing agent, has made it one of the most important variables in the study of intergroup relationships.

In the past century, many empirical studies on intermarriage in the United States showed the existence of strong tendencies to marry within one's own ethnic, racial, or religious group (e.g., Alba and Golden 1986; Drachsler 1921; Gordon 1964; Johnson 1980; Kalmijn 1991, 1993; Kennedy 1944; Lieberson and Waters 1988; Qian 1997). There was also an increase in the number of studies on ethnic intermarriage in other countries, like Australia (Jones 1991), Canada (Jansen 1982), Czechoslovakia (Boguszak and Bozon 1989), England (Coleman 1985), France (Munoz-Perez and Tribalat 1984), Singapore (Lee 1988), Taiwan (Lu and Wong 1998), Yugoslavia (Botev 1994), and Turkey (Gndz-Hosgr and Smits 2002). However, almost all these studies were restricted to intermarriage in stable or gradually changing situations. Little is known about how intermarriage patterns are affected when the division of power in a society is drastically changed as a result of social upheaval or revolution.

The profound social transformations that took place in the late 1980s and early 1990s in Eastern Europe and the former Soviet Union offer important opportunities for such a study. In most of the Soviet republics that became independent at that time, the power relationship among ethnic groups was turned upside down, with the Russians, who had been the dominant group for half a century or more, losing their privileged position in favor of the native-born groups. A comparison of the ethnic intermarriage patterns in these republics before and after the transformation could yield new insights into how the marital decisions of individuals are shaped by (changes in) the power relationships in the larger society.

In this article, such a comparison is made for the former Soviet republic of Latvia. The ethnic composition of Latvia and the social upheaval caused by the transition to an independent democratic society after 50 years of Soviet occupation makes this country an interesting case for studying the dynamics of ethnic relationships in a rapidly changing environment. Using population registration data on the ethnic groups of brides and grooms and individual-level data from the Latvian Fertility and Family Survey (LFFS), we determined the patterns and trends in ethnic intermarriage between the Latvian majority and the Russian, Belarussian, and Ukrainian minorities over a period of more than three decades, from 1970 to 2003. In this way, we gained insight into the social distances among the groups in the period of Soviet occupation and the changes that took place after the country gained independence in 1991.

In the following sections, we present historical and demographic background information on Latvia and discuss the relations between the ethnic groups in the country before and after independence. Then we elaborate the hypotheses that we tested in our study. After we discuss the data and research methods, we report the intermarriage trends on the basis of the population registration data. Next, we present the results of a multivariate log-linear analysis in which we estimated trends in ethnic intermarriage while controlling for regional differences and the effects of education and educational assortative mating. We also report the results of a multivariate logistic regression analysis with the odds of marrying a person of the other group as the dependent variable. Although the earlier analyses remained at the aggregate level, the focus of this last analysis is on how the characteristics of individuals influence their marital decisions. The empirical part ends with some simulation analyses to gain deeper insight into the underlying causes of the trend that we found for the period after independence. Throughout the article, in addition to the term intermarriage, the term homogamy is used to denote the marriage of people with similar characteristics (i.e., ethnic homogamy or educational homogamy).

BACKGROUND

Demographic Developments

The first Latvian republic was created in 1918 after Latvia had been a part of the Russian empire for centuries. In 1940, the Soviets occupied Latvia, and in 1941 Latvia was forcefully incorporated into the Soviet Union. Interrupted by the German occupation (1941-1944), the Soviet rule lasted for almost 50 years. As can be seen in Table 1, after 1945, the ethnic composition of the population changed drastically. Russian party cadre, military personnel, and police were sent to ensure the incorporation of Latvia into the Soviet empire. Other immigrants, predominantly Russians, but also large groups of Belarussians and Ukrainians, soon followed. Many of these immigrants were factory workers. At the same time, the number of native Latvians was reduced by deaths and deportations during the war. Misiunas and Taagerpera (1993:354) estimated that the total population in Latvia decreased from 2 million to 1.4 million between October 1939 and the end of 1945. According to their educated guesses for the turbulent 1945-1955 period, 100,000 Latvians were deported, 100,000 Russian Latvians (who had moved to Russia after the October 1917 revolution or earlier in the century) returned, and 535,000 Russian and other Soviet immigrants arrived.

Zvidrins and Vanovska (1992) estimated that the net number of Russian-speaking immigrants who arrived in Latvia after 1960 was 400,000. Throughout the 1970s and 1980s, about 50,000 to 55,000 people settled in Latvia annually, whereas 43,000 to 46,000 moved back to other parts of the Soviet Union each year (Latvian Central Bureau of Statistics, LCBS, 2002a). As Table 1 shows, Latvians had almost become a minority, at least as a linguistic group, by the end of the 1970s. The vast majority of migrants were Russians, many of them army-related personnel and their families.

After Latvia gained independence from the Soviet Union in 1991, the migration pattern changed drastically. The average annual number of immigrants decreased sharply, from 50,000 in the 1980s, to 6,000 in 1991-1995, to 2,500 in 1996-2000, and to 1,400 in 2001-2003. In addition, the average annual emigration figures decreased greatly, from 42,000 in the 1980s, to 33,000 in 1991-1995, to 9,400 in 1996-2000, and to 4,000 in 2001-2003 (LCBS 2004). As a result, Latvia experienced a net emigration in the 1990s, whereas there had been a net inflow ever since World War II. The vast majority of emigrants were Russians, Belarassians, and Ukrainians. Among those who left during the emigration peak in 1992 and 1993, a large share was related to the Russian military, including personnel of the military-industrial complex and their spouses and dependent children (Codagnone 1998; Prikulis 1997; Zvidrins 1998). According to the Latvian-Russian agreement, the Russian military and all persons attached to it had to withdraw to the Russian Federation before August 1994. People who were related to personnel in the army of the Russian Federation and who were not ethnic Latvians were not allowed to stay. Exact figures on the share of army-related emigrants have not been published. Codagnone (1998) estimated that 9.2% of the ethnic Russians who were living in Latvia in 1989 had returned to the Russian Federation by 1996. The Russian emigration from other former Soviet republics was much larger in this period. For instance, in Uzbekistan, Azerbaijan, and Armenia, it was 22%, 35.5%, and 53% of the ethnic Russians, respectively (Codagnone 1998). Table 1 presents an overview of these population changes in Latvia during and after the Soviet rule.

Like most modern societies, Latvia has experienced a decrease in marriage rates, which was particularly substantial during the first half of the 1990s. This decrease stopped about 1997 and has turned into a slight increase in recent years. The number of people who married was about 8 per 1,000 for both Russians and Latvians in 2003 (LCBS 2004; our computations). During the 1990s, both the number of people who cohabited without being married (Usackis 1995) and the age at first marriage (LCBS 2004) increased, which suggests that the decrease in marriage was partly the result of delayed marriages.

Regarding the distribution of the ethnic groups throughout the country, it is important to note that, according to the 2000 census, the majority of Russians in Latvia (69%) lived in the cities (they made up 51% of the population of the seven largest cities), whereas the majority (65%) of the Latvians lived in the countryside (where they were the large majority of the population). The distribution of ethnic groups over the cities and countryside was largely stable throughout the second half of the twentieth century (Zvidrins and Vanovska 1992).

Ethnic Relations

Officially, ethnic relations throughout the Soviet Union were good and brotherly. Soviet ideology propagated friendship among nations and nationalities, and ethnicity officially played no role. Indeed, during Soviet rule, there were no violent uprisings, large-scale protests, or (public) ethnic conflicts in Latvia (Karklins 1986; Misiunas and Taagerpera 1993). With regard to marriage, the ideology was that people in a communist society would marry out of pure love. The homo sovieticus would be blind to a potential partner's wealth, occupation, intellect, or ethnicity.

However, despite this ideology, ethnicity was an important factor in many spheres of social life (as we describe later). Moreover, this ideology was not backed up with information about actual marriage behavior. During the Soviet period, hardly any statistics on intermarriage were available. Publications from the Soviet censuses (1959, 1970, 1979, and 1989) did not report rates of intermarriage. Official publications presented only the number of people who were living in mixed families-a crude indicator of the degree of intermarriage. No nationwide or unionwide research on intermarriage was published (Fisher 1980; Zvidrins and Vanovska 1992), although the political leadership claimed that intermarriage was continuously increasing (Karklins 1986:154-82). Susokolov (1987) provided a fragmentary overview of studies on intermarriage in the Soviet Union. On the basis of registration data from the cities of Tallinn (Estonia), Tbilisi (Georgia), and Kishinev (Moldavia), he concluded that intermarriage was more frequent after age 25 (however, the analyses are unclear). More recently, Botev (2002) provided descriptive analyses that were based on the 1989 census for the Russian Federation. For Latvia and the post-Soviet period in general, no empirical studies of ethnic intermarriage have been published.

One respect in which ethnicity was important in Soviet Latvia was with regard to language, especially because Latvian and Russian are very different languages. Soviet policy made the knowledge and use of Russian as the "language of interethnic communication" a priority (Karklins 1986:218). Russification through language instruction, starting at age 6 or 7, was successful in Latvia: most Latvians were bilingual, and Latvians had the highest proportion of Russian speakers among the non-Slavic people. However, most of the Russians who were living in Latvia spoke only Russian-a situation that aggravated the social and cultural segregation between the Russians-or, rather, Russian speakers-and Latvians.

Because of the relatedness of their languages and the Soviet Russification policy, virtually all Belarussian and Ukrainian immigrants in Latvia became Russian speakers (Misiunas and Taagerpera 1993). The 2000 Population Census showed that 73% of all Belarussians and 68% of all Ukrainians in Latvia regarded Russian as their mother tongue (LCBS 2002b). Of the Russians, Belarussians, and Ukrainians, only 4%, 7%, and 4%, respectively, reported that Latvian was their mother tongue.

During the Soviet period, Latvians always had their own primary and secondary schools, radio and television stations, and theaters. Nonetheless, until the second half of the 1980s, they were increasingly the object of Sovietization and Russification. Fluency in Russian was a prerequisite for a career in the Communist Party, army, government, or large industries. In all non-Russian republics, the local Russians were overrepresented in certain educational fields, especially engineering, law, and economics, and underrepresented in health care and agriculture. Although education at the republic level was available in the native languages, these languages were given no room in unionwide institutions, like the academic world, large enterprises, and organizations that operated throughout the Soviet Union.

Despite ethnic segregation, the awaking of national awareness in the second half of the 1980s did not lead to violent conflicts between Latvians and Russians. The first large-scale public protests against Soviet rule came at the end of the 1980s. An important reason to protest against new industrial plants and a subway system in Riga was the immigration of Russian-speaking workers that these projects would entail. Many Russians even actively supported the struggle for independence from the Soviet Union in 1990 and 1991 (Lieven 1993). On the other hand, the fiercest opponents of independence were also Russians.

In sum, during the Soviet period, the official ideology was pro-intermarriage, but at the same time, there were differences in political power and status among the ethnic groups; there was educational and cultural segregation along ethnic lines; and, in spite of all "Sovietization" attempts, ethnic consciousness does not seem to have disappeared. We therefore think that ethnicity was also an important characteristic with regard to preferences for a marriage partner before independence and that it thus remains to be seen whether intermarriage was really on the rise, as the Soviet authorities claimed.

After Independence

As in other former-Soviet states, ethnic consciousness (or national awareness) increased in Latvia during the period of perestroika and was especially strong during the first years of independence (Grigorievs 1996; Lieven 1993; Misiunas and Taagerpera 1993). Political power became the domain of Latvians, and Latvian became the single state language. The political climate changed drastically, also with regard to ethnic and language issues. After independence, only descendents of citizens of the first Latvian republic (1918-1940) were automatically granted citizenship in the new republic. Thus, although citizenship did not involve nationality de jure, de facto 99% of the Latvians could change their Soviet passports for Latvian passports, whereas only a small minority of the Russian-speaking groups could do so. In the first years, a complex system of quotas and age limits was proposed that would allow only a limited number of non-Latvians to acquire citizenship and naturalization from 1996 onward. Language laws were introduced, specifying Latvian language requirements for certain professions. Russian language lessons in Latvian school were reduced in favor of English, and Latvian language instruction was intensified in Russian schools.

Many Russians in Latvia thought they were being treated as second-class citizens. However, the quota system for citizenship was abolished in 1994, and by 1998, the age limits were also removed (Zaagman 1999). Recent studies have found no evidence of structural ethnic discrimination in the labor market (Aasland and Flotten 2001). Still, compared to other postcommunist states, ethnic polarization seems to be strong in Latvia. Latvians are not eager to support more rights for Russians, whereas many Russians are demanding more rights (Evans and Need 2002).

The rise of nationalism among Latvians and the loss of privileges enjoyed by Russians led to tensions both within Latvia and between Latvia and Russia (Mezs, Bunkse, and Rasa 1994). These tensions reached their highest levels in 1995 and 1996, but no open violent conflicts occurred. Since 1998, some signs of decreasing segregation can be observed. For instance, more bilingual schools have opened, and the government is paying greater attention to the "integration" of non-Latvians into Latvian society. Since 2001, the Latvian government has funded the National Program for Latvian Language Training, aimed at increasing the proficiency of Latvian among Russians. Despite these positive developments, however, segregation in schools, cultural life, and the media is still widespread. Since the introduction of freedom of speech and a market economy, the segregation in the (printed) media has even become stronger (e.g., Mediju Tilts 2001).

In sum, ethnic consciousness increased after independence, and the power relations between Latvians and Russians were reversed. How this situation might have affected intermarriage compared to the Soviet period is discussed in the next section.

HYPOTHESES

Increased Consciousness or Stable Preferences?

In the literature on intermarriage, a distinction is generally made between the preferences of the partners and their families for a marriage candidate from their own or another social group and the opportunities and restrictions present in the marriage market in which the partners compete (Gunduz-Hosgor and Smits 2002; Kalmijn 1998). Our first two hypotheses focus on how the changing circumstances in Latvia as a result of independence may have influenced the preferences of the persons who married and their families. The first hypothesis predicts a decrease in Latvian-Russian intermarriage after independence as a result of the ethnic tensions, national movements, restrictive citizenship laws, and political conflicts surrounding and following independence. These tensions were broadly discussed in the media, which also paid much attention to the position and historical experiences of the in-group (Latvians) and expressed increasing distrust of the members of the out-group (Russians) who became "colonists" instead of "Soviet brothers." In such a polarized and (seemingly) unstable situation, with all the media attention on the differences and tensions between the groups, it seems likely that ethnicity, not only in society at large, but also in personal relations, would become an issue and that people who were looking for marriage partners might increasingly prefer partners from the "right" (that is, their own) ethnic group. Consequently, we would expect a reduction of Russian-Latvian intermarriage in the years surrounding and after independence. This is the increased ethnic consciousness hypothesis.

However, there are also arguments for a competing hypothesis that states that the preferences of both Latvians and Russians did not change substantially in the period of transformation, but that the political turnaround made it possible for these preferences to be publicly expressed. As we discussed in the section on ethnic relations, it is likely that during the period of Soviet occupation, the relationships between Latvians and Russians were less "brotherly" than the Russian propaganda suggested and that a large proportion of the Latvians considered the Russians to be occupiers (Karklins 1986; Misiunas and Taagerpera 1993). During the 1970s and 1980s, individuals, Latvians and Russians alike, were conscious of their ethnic identity and may have had a high preference for marrying partners of their own ethnic groups. Since there were no restrictions on marriage choice under the old system, we may argue that people already married according to these preferences and hence that there may have been little change after independence. This is the stable preferences hypothesis.

Migration Dynamics

A third hypothesis predicts changes in intermarriage as a result of changes in the migration pattern after independence. As we noted in the section on demographic developments, the inflow of new immigrants, which was high before independence (50,000 annually), almost completely dried up after independence (to only 2,500 annually in the second half of the 1990s). At the same time, emigration was relatively high in the first years after independence, but then also decreased to a relatively low level (to fewer than 10,000) in the second half of the 1990s. These changes in the migration pattern may have affected the composition of the group of Russians in Latvia in two main ways. First, the Russians who emigrated from Latvia after independence were probably those who were the least integrated into Latvian society (for example, they did not speak Latvian or were part of the Russian military). We expect that they had a higher preference for marrying partners from their own ethnic group than did those who stayed behind. As a result, their departure can be expected to have increased the average preference for intermarriage among the remaining Russian population.

Second, the strong reduction in the inflow of Russians after 1991 means that during the 1990s, the share of Russians who were recent immigrants decreased. Because a substantial portion of immigrants had no intention of staying in Latvia permanently, they tended to be focused on the Russian community and to have little contact with native Latvians. This tendency suggests that the reduction of the inflow after 1991 led to a higher average level of integration of Russians in Latvia.

As a result of these two processes, the Russian population in Latvia may increasingly consist of people who are more integrated into the Latvian society and hence would be more likely to marry Latvian partners. This prediction of a positive effect of the changed migration patterns on intermarriage preferences among Russians is called the changed migration hypothesis. It is important to note that this hypothesis implies a change in preferences only among the Russians, and not among the Latvians.

Social Exchange

The foregoing hypotheses are rather naive from the perspective of social exchange theory in the area of partner selection (e.g., Kalmijn 1998; Merton 1941; Schoen and Wooldredge 1989) because they take into account only one characteristic that affects the selection of partners. In reality, people select their partners on more than one characteristic. Social exchange theory focuses on the fact that different characteristics of individuals may differ in their value in the marriage market and that persons may compensate unfavorable characteristics in one respect with more favorable characteristics in other respects. With regard to ethnic intermarriage, it means that persons whose social status in society is low because they belong to a subordinate ethnic group have a better chance of marrying members of the higher-prestige dominant group if they can offer the partners something in return, like high socioeconomic status or physical attractiveness. In the same way, persons with a low socioeconomic status or less physical attractiveness who belong to the dominant group may marry higher-status or more attractive partners from the subordinate group. Both partners may profit from this kind of exchange.

In the period of Soviet occupation, during the "Russification" of the country, Russia's dominance in the Soviet Union and the fact that mastery of the Russian language was a prerequisite for reaching the highest social positions in the country suggest that the Russian minority was the dominant ethnic group. Social exchange theory would thus predict an overrepresentation of marriages between Latvians of a higher socioeconomic status and Russians of a lower socioeconomic status.

After 1991, however, the situation was reversed with regard to ethnic dominance. Suddenly, the Latvians became the dominant ethnic group, and the Russian-speaking minority lost their privileged position. Social exchange theory, therefore, would predict that after 1991, the Russians would have had to be of a higher socioeconomic status to marry persons from the dominant Latvian group. Hence, we would expect to find an overrepresentation of marriages between higher-status Russians and lower-status Latvians after independence.

Control Factors

An association between the ethnic groups of spouses does not necessarily imply that ethnicity plays a significant role in the choice of a partner. Such an association may also be a by-product of selection on the basis of other characteristics or a result of the specific structure of the marriage market. To gain more insight into the strength of the preferences for (or aversion to) a partner from one's own or the other group, we controlled for a number of potentially confounding factors in our analysis.

In this section, we briefly describe these factors. First, as we discussed earlier, people select their partners generally on the basis not of one single characteristic, but of many characteristics. Because these characteristics may be related to each other, a resemblance between spouses with regard to one characteristic may be the result of partner selection on the basis of another characteristic (cf. Ultee, Dessens, and Jansen 1988). Earlier research found, for example, that ethnic homogamy may be strengthened by educational homogamy (Gndz-Hosgr and Smits 2002). In addition to such an effect of educational homogamy, a direct effect of education on intermarriage can also be expected because schools are homogeneous marriage markets (Kalmijn 1998; Mare 1991). If schooling is extended, the chances increase that individuals will marry persons they meet at school. This implies that in the United States, with its ethnically mixed school system, people with higher educational levels can be expected (and have been found) more often to marry partners with the same educational levels (Mare 1991) and less often to marry partners from the same ethnic group (Lieberson and Waters 1988; Qian 1997). In Latvia, however, the school system is ethnically segregated to a large extent. This was especially the case before independence. As a result, individuals who complete more schooling have a higher chance of marrying partners not only with the same educational level but also from their own ethnic group (and hence there may be a positive association between ethnic and educational homogamy). Given the possible confounding effects of both educational level and educational homogamy, we controlled for these factors in our models.

In addition to being influenced by the preferences of the partners and their families, the outcomes of marital decisions are also influenced by structural characteristics of the marriage market (Kalmijn 1998). An important factor in this respect is group size. All other things being equal, the probability of coincidentally meeting somebody from a large group is higher than the probability of meeting somebody from a small group (e.g., Blau and Schwartz 1984). Furthermore, the size of the smallest groups is a restrictive factor. The number of mixed marriages cannot be higher than the number of persons in the smallest group. In the same way, the sex ratio of the group may play a role. Because more male than female Russians immigrated to Latvia, even with completely random mating, more marriages between Russian men and Latvian women can be expected. To gain more insight into the role of the preferences of the marriage partners, intermarriage measures like odds ratios or log-linear parameters that are insensitive to differences in group size have to be used (Kalmijn, 1998; Qian 1997).

A final factor that can be expected to play a role is spatial segregation. Because the proportion of Russians is substantially lower in the countryside than in the cities (LCBS 2004), it seems likely that the Russians would intermingle more with Latvians in the countryside (that is, live more often in the same neighborhoods and go to the same schools). Hence, a higher number of mixed marriages among Russians can be expected in the countryside than in the cities. For Latvians, the reverse is true. Because of the relatively small number of Russians living in the countryside, the chance of Latvians meeting potential Russian partners is relatively low there. To take this factor into account, we also controlled for place of residence in our analysis.

DATA AND METHODS

Data

We used annual marriage tables obtained from the LCBS and individual-level data from the 1995 Latvian Fertility and Family Survey (Zvidrins, Ezera, and Greitans 1998). The marriage tables contain information on ethnicity of brides and grooms for all marriages registered from 1970 to 2003. For 1981, LCBS was not able to provide the complete table and could give information only for Russians. To study intermarriage patterns while controlling for the influence of other factors, we also used individual-level data from the 1995 Latvian Fertility and Family Survey (LFFS), which is based on a nationally representative sample of 1,501 men and 2,699 women aged 18-49 who were interviewed in September 1995. Women were oversampled because of the survey's focus on fertility. The nonresponse owing to refusals to be interviewed was about 5%. All the respondents provided information on their current and all previous marriages, including the dates of marriage and cohabitation and sociodemographic information for all partners. We used the information on all legal marriages presented in the survey.

Variables

In the LFSS, ethnicity is self-defined. In the marriage tables, information from the respondents' passports is used. Soviet and Latvian passports state both citizenship and "nationality." In many Eastern European countries, the word nationality actually means ethnicity. Therefore, it is possible to have Latvian citizenship and at the same time have a different nationality. In our analyses, we distinguished among Latvians, Russians, Belarussians, Ukrainians, Poles, Lithuanians, and Jews (who were defined as a "nationality" by the Soviet administration).

The differences among Russians, Belarussians, and Ukrainians are small. Most of the Belarussians and Ukrainians reported that Russian is their native language (73% and 68%, respectively). The vast majority of Belarussian and Ukrainian immigrants arrived in Latvia during the same period as the Russians, came for similar work, lived in the same neighborhoods and flat blocks of apartments, and sent their children to Russian schools because there were no Belarussian or Ukrainian schools before 1991. Because of the strong resemblance among these groups, we combined them in most analyses into one category called "Russians." Age at the time of marriage was measured in years. For educational level, we used the categories (1) primary education or less, (2) secondary education, and (3) tertiary education. The educational difference between the partners was measured with a variable indicating whether the partners had the same level (0) or differed by either one (1) or two levels (2). We defined four marriage cohorts: before 1972, 1972-1980, 1981-1990, and 1991-1995. Urbanization was dichotomized: (1) Riga (the capital) and the six other largest cities (Daugavpils, Ventspils, liepaja, Jurmala, Jelgava, and Rezekne), where 51% of the inhabitants are "Russian," and (2) towns and the countryside, where only 21% of the inhabitants are "Russian" (figures from the 2000 census). Mixed family background was measured by a dummy variable indicating whether (1) or not (0) the respondent's parents were of different ethnic groups. The distribution of the variables is presented in Table 2. Latvians and "Russians" differed substantially only with regard to the percentage of mixed family background and place of residence.

Methods

We first present the percentages of mixed marriages and odds ratios for all yearly marriages in the Latvian marriage registers. The percentages of mixed marriages are important from a social cohesion perspective: they show to what extent there are cohesionincreasing marital ties between the groups. However, because percentages are sensitive to differences in size and sex ratio between the ethnic groups, they say less about the preferences for (or the aversion to) a marriage with somebody of another group. Therefore, we also present odds ratios, which show the strength of the tendency to marry within one's own ethnic group independent of the differences in sizes of the groups (Kalmijn 1998). The odds ratio is defined as the odds that a Latvian man will marry a Latvian woman (rather than a Russian woman), divided by the odds that a Russian man will marry a Latvian woman (rather than a Russian woman), that is, (LL/LR)/(RL/RR) (cf. Kalmijn 1998:405). The odds ratio for women is equivalent. An odds ratio of 1 means that the likelihood of marrying within one's ethnic group is no greater than the likelihood of marrying outside one's ethnic group. An odds ratio greater than 1 indicates a tendency to marry within one's own ethnic group relative to marrying outside one's ethnic group.

Another way to control for differences in group size is by using log-linear analysis (Agresti 2002). This method has the advantage that it offers the possibility of studying patterns of ethnic intermarriage in a multivariate way and hence controlling for the effects of potentially confounding factors (Kalmijn 1991; Qian 1997). Our second analysis is a multivariate log-linear analysis in which the trend in ethnic intermarriage is studied while controlling for differences in the size of the ethnic groups and for differences between city and countryside, for differences between educational groups, and for the influence of educational homogamy. For this analysis, we constructed a 2 2 3 3 4 2 table (ethnicity of husband ethnicity of wife education of husband education of wife cohort urbanization) on the basis of the LFFS data.

The registration data and the multivariate log-linear models show the degree to which intermarriage varies over time and subgroup differences at the aggregate level. However, these patterns of intermarriage observed at the aggregate level are the outcomes of decisions made by individuals, who, on the basis of their own and their (potential) partners' characteristics, decide whether or not to marry partners from the other group. If we want to understand what happens at the societal level, it is also important to understand what happens at the individual level. In our third analysis, we used the LFFS data to analyze the association between the characteristics of an individual and the individual's decision to marry somebody from the other group. For this purpose, we performed a logistic regression analysis with a dummy variable indicating whether (1) or not (0) the respondent was married to someone from the other ethnic group as the dependent variable and the respondent's age, gender, educational level, ethnicity, marriage cohort, urbanization, and whether or not the respondent has a mixed family background as the explanatory variables. We conclude with a simulation analysis to explore the extent to which the trends in ethnic intermarriage after independence may be due to selective emigration.

RESULTS

Descriptive Analyses

Figure 1 reports the percentage of people who married partners outside their own ethnic group. These percentages represent all annually registered marriages for the largest ethnic groups in Latvia from 1970 until 2003 (with the exception of 1981). During this period, more than 1 in 3 marriages was between partners of different ethnicities. The highest rates of intermarriage were among the smaller ethnic groups, with rates of 80% to 90%. This finding is no surprise, since for individuals who belong to smaller groups, the probability of a coincidental meeting with a potential partner from the same ethnic group is smaller than for members of a larger group. The larger ethnic groups, Latvians and Russians, had much lower rates-about 20% among the Latvians and 35% to 40% among the Russians. Large proportions of Belarussians (87%) and of Ukrainians (85%) married outside their own groups. An additional analysis (not presented) showed that these two latter groups mostly married Russians. For example, during the 1980s, 55% of all Ukrainians and 51% of all Belarussians married Russians.

With regard to trends, three observations can be made from Figure 1. First, for the five smaller groups, the rate of intermarriage seems to have increased between 1970 and 2003. The increase was especially strong among Jews. One reason for this increase may have been the continuing exodus of Jews to Israel. As the group became smaller, the chance of finding partners within their own ethnic group declined. Moreover, the most orthodox Jews were probably the first to emigrate, and they probably also had the strongest preferences for marrying within their own group. Second, the trend for Latvians showed little change: between 1970 and 2003, the proportion of Latvians who married outside their own group fluctuated slightly around 20%. Third, there was an increase in the proportion of mixed marriages among the Russians in the 1990s.

Given the high number of mixed marriages between Russians and Belarussians and between Russians and Ukrainians and the similarities, discussed earlier, among these three groups, we combined these groups in the remainder of this article (denoted as "Russians"). Figure 2 presents the intermarriage rates between Russians and Latvians by gender. Two interesting patterns appear. First, intermarriage was rather stable between 1970 and 1989-1990. Then the percentage of Latvians who married Russians dropped slightly to an all-time low in 1996, after which it rose again. At the same time, the proportion of Russians who married Latvians increased substantially, from about 16% in 1990 to 25% in 2003. Second, although there were slightly more mixed marriages between Russian men and Latvian women during most of the 1970s and 1980s, the pattern reversed in the early 1990s, after which there was a clear overrepresentation of marriages between Russian women and Latvian men.

The trends in Figure 2 do not support the increased ethnic consciousness hypothesis. Instead of a decrease in intermarriage after independence, we found stability among the Latvians and an increase among the Russians. The fact that intermarriage among the Latvians remained at about the same level after independence seems at first sight to be consistent with the stable preferences hypothesis. However, because of the reduction of the Russian population owing to emigration after independence, stable preferences would actually imply decreasing intermarriage rates among Latvians. To draw more definite conclusions about this hypothesis, we therefore have to look at odds ratios. The increasing percentages of intermarriages among the Russians after independence seems to lend support to the changed migration hypothesis, which predicts such an effect as a result of the changed migration patterns after independence.

To find out whether the changing intermarriage percentages after independence reflect changes in the social distances among the groups or only changes in the relative number of Russians and Latvians who married, we calculated odds ratios, which are insensitive to such differences in group size (see the right y-axis of Figure 2). The trend in the odds ratios makes clear that changes in group size are not the only explanation for the higher intermarriage rates among Russians in the 1990s. Whereas the value of the odds ratio fluctuated around 24 during the 1970s and 1980s, it clearly decreased (indicating increasing intermarriage) after the independence year 1991 and-with the exception of a slight increase in the mid-1990s (the period of highest ethnic tension)-it continued to decrease to about 14 in 2003. This finding suggests that contrary to the predictions of the increased ethnic consciousness hypothesis and the stable preferences hypothesis, the social distance among the groups became smaller in the 1990s.

Log-Linear Analysis

Table 3 presents a number of log-linear models and their measures of fit. To select the best model to describe the data, we relied largely on the Bayesian information criterion (BIC; Raftery 1995). The BIC is a model-selection tool that is applicable when one uses maximum likelihood estimation. It can be used to compare nonnested models, and it prefers parsimonious models to complex ones. The BIC is calculated as likelihood ratio - (df) log(N). In general, the model with the lowest BIC value is the one that should be selected (Raftery 1986). In case we found almost-similar BIC values, we also conducted a likelihood ratio test (Agresti 2002), in which the difference in the likelihood ratio (G^sup 2^) among the models is compared with the difference in degrees of freedom (df).

Model 1 is the model of independence, which supposes that the frequency of occurrence of each marriage combination is determined only by the number of men and women in the margins of the multivariate marriage table. Table 3 shows that this model controls for the differences in frequencies among all the combinations of categories of factors that are included in the analysis. After controlling for these differences, the model supposes marriage to be completely at random. All the other models in Table 3 suppose a tendency toward homogamy. Model 2 tests for the presence of ethnic homogamy. The large decrease of both the G^sup 2^ (for 1 df) and the BIC value point to the presence of strong ethnic homogamy in Latvia. Models 3 to 5 test for the presence of educational homogamy. Model 3 describes this form of homogamy with a uniform association parameter, which indicates the extent to which the educational levels of the spouses resemble each other. Model 4 adds two diagonal parameters to the model, and Model 5 includes parameters for all (four) independent combinations of the educational variables. Model 3 fits the best of these three models (according to BIC) and fits better than Model 2. We therefore conclude that in addition to ethnic homogamy, there is educational homogamy in Latvia and that the uniform association parameter gives a reasonable description of the degree of educational homogamy.

Models 6 to 8 test for changes in ethnic homogamy over time. Model 6 fits a separate ethnic homogamy parameter for each cohort, Model 7 tests for the presence of a linear trend, and Model 8 tests whether the degree of ethnic homogamy differed before and after independence. None of these models shows an improvement in fit compared to Model 3. We also tested the same three trend models for educational homogamy and found none of them to be an improvement (not presented in Table 3). We therefore conclude that the level of both ethnic homogamy and educational homogamy remained stable over time. This finding seems to be in contrast with the data in Figure 2, which show a clear decrease in ethnic homogamy after independence. However, it is probably due to the fact that the LFFS data capture only the first years (1991-September 1995) after independence. In the beginning of this period, ethnic intermarriage was at an all-time low level, and it was not before 1993 that the odds ratio dropped below the average of the preindependence years for the first time (see Figure 2). The low level of intermarriage is also reflected in the average odds ratio over the registered marriages in the period 1991-1995, which, with a value of 22, is not much lower than the value of 24 for the period 1970-1990.

To test the social exchange hypothesis with regard to ethnicity and education, Model 9 assumes an overrepresentation of mixed couples in which the Latvian partner had a higher educational level before 1991 and in which the Russian partner had a higher educational level after 1991. Since this model offers no improvement over Model 3, we conclude that the data lend no support to this hypothesis. Model 10 tests whether ethnic homogamy differs between the countryside and the city. This model fits better than Model 3 (it has the same BIC value, but is clearly better on the likelihood ratio test), so we conclude that a small countryside-city difference indeed exists. Model 11 tests whether ethnic homogamy is related to the educational levels of the partners, and Model 12 tests whether it is related to educational homogamy. Model 11 is no improvement, but Model 12 clearly is. The difference in the BIC values between Model 12 and the second-best models (3 and 10) is more than 6, which, according to the criteria presented by Raftery (1995), is strong evidence that Model 12 fits the data better than do those models. Therefore we chose Model 12 for the description of our data.

The relevant log-linear parameter estimates and the corresponding odds ratios of Model 12 are presented in Table 4. The odds ratio for ethnic homogamy, averaged over city and countryside, has a value of 8.5, which means that for both Latvians and Russians, the odds of marrying a partner from one's own ethnic group is 8.5 times as high as the odds of marrying a partner from the other group. This odds ratio is substantially lower than the bivariate odds ratio over the period 1970-1995 (which has a value of 29 in the LFFS data and of 24 for the registered marriages over this period). This finding may be due, in part, to the fact that the log-linear model controls for differences among city and countryside, educational homogamy, and the marginal distribution of all the included variables.

Table 4 also shows that there is substantial educational homogamy. The uniform association parameter has a multiplicative value of 5.4. This parameter can be interpreted as the average of odds ratios in the 3 3 educational homogamy table. It indicates that an individual is 5.4 times as likely to be married to a partner with the same level of education (e.g., both with secondary education) as to a partner with a one-step difference in educational level (e.g., an individual with a secondary education married to a partner with either a primary education or a tertiary education).

The interaction between urbanization and ethnic homogamy indicates that the chances of intermarriage are higher in the countryside than in the cities. Finally, the positive coefficient for the interaction between educational and ethnic homogamy shows that in Latvia, ethnic homogamy goes together with educational homogamy to a certain extent.

Logistic Regression Analysis

To gain insight into the individual-level factors that are associated with being married to a partner of the other group, Table 5 presents the results of the logistic regression analysis with the odds of ethnic intermarriage as the dependent variable. To make the interpretation of the interaction effects more straightforward, we used deviation-from-mean coding for the dummy variables in the model. Only significant interactions are included in the model. Table 5 shows that the odds of intermarriage differ significantly between the ethnic groups: Latvians are less likely to intermarry than are Russians-a finding that is also suggested in Figure 2. Gender and age are not significantly associated with the odds of marrying someone from the other ethnic group. The parameters of the cohort variable show the changes in ethnic homogamy over the period studied, with the 10 years preceding independence as the reference category. None of the other cohorts deviates significantly from this category. Also the interaction effects of cohort with ethnicity, gender, and age are not significant. As we discussed earlier, the finding that intermarriage did not increase after independence (although the coefficient is in the expected direction) may be due to the fact that the LFFS data document only the first years of the upward trend, which started at an exceptionally low level in 1991.

Education is significantly associated with ethnic intermarriage: individuals with higher educational levels have lower likelihoods of being married to partners from outside their group. The completion of only secondary education, as opposed to tertiary education, is associated with a 36% increase in the odds of intermarriage. For individuals with only a primary education, the increase is as large as 88%. Because the interaction effects of education with ethnicity and gender were not significant, we conclude that the effects of education are about equal for men and women and for Latvians and Russians. We further conclude that the existence of an educational difference between the partners significantly increases the odds of intermarriage.

Having an ethnically mixed family background has a strong positive effect on ethnic intermarriage. With regard to urbanization, respondents in the countryside have higher odds of marrying outside their ethnic group than do respondents in the cities. The fact that the interaction between urbanization and ethnicity is significant gave us the opportunity to gain more insight into this city-countryside difference. On the basis of the coefficients presented in Table 5, we calculated the extent to which the odds ratios for the different combinations of ethnicity and urbanization differ from the average odds ratio. This computation clarified that Latvians in the city are more likely to marry Russians than are Latvians in the countryside (odds ratios of 0.97 and 0.69 times the average, respectively), whereas Russians in the countryside are more likely to marry Latvians than are their counterparts in the cities (odds ratios of 1.96 and 0.76 times the average, respectively). No other interaction of ethnicity, gender, and cohort with the other variables was significant.

Emigration or Integration?

Our analysis of marriage-registration data for the 1970-2003 period suggests an increase in intermarriage between Russians and Latvians in the 1990s. This change was not evident in our analyses of the LFFS data, but, as we noted, these data, which were collected in 1995, may not capture a long-enough period within which intermarriage was changing for the Russian population in Latvia. The suggestion of an increase in Russian-Latvian intermarriage in the late 1990s is important from a social cohesion perspective. It indicates that after periods of social upheaval and conflict, processes of renewed social integration may emerge that seem to be able to overcome the tension and mistrust between the antagonistic groups to a certain extent. Of course, the nature of the underlying processes would determine whether we can denote such a development as merely positive. Our changed migration hypothesis suggests two possible reasons for the increasing tendency toward intermarriage among the Russians: selective emigration of the least-integrated Russians or a reduction of the immigration of the nonintegrated Russians. If the increasing social integration is mainly the result of the departure of individuals who did not feel safe and at home in Latvia after independence, we have to denote the process as more negative than if it is the effect of a reduced inflow of nonintegrated newcomers and the increasing integration of earlier immigrants.

Given the high level of emigration of Russians in the first years following independence, it seems likely that the proportion of Russians who married Latvians increased to some extent in those years because the Russians who emigrated were predominantly those who would have married within their own group. The push factors for mixed families (or for individuals who had the possibility of marrying Latvian partners) were much smaller because the Latvian partners and their children could obtain Latvian citizenship. In addition, the pull factors for migration to Russia were probably greater for all-Russian families and for Russians with a strong preference for homogamous marriages.

The fact that selective migration did take place does not imply that it was the major factor behind the Russians' increasing intermarriage rates. It all depends on the size of this effect relative to the alternative ones. To increase our understanding of the impact of selective migration, we performed a simple simulation analysis. Starting with the inter-marriage table from 1991, which has an odds ratio of 26.6, we estimated what reduction of Russian marriages would be needed to obtain the odds ratio of 14.2 that was observed for 2003. If we assume that only the number of homogamous Russian marriages decreases and that the number of mixed marriages and homogamous Latvian marriages remains the same (the situation in which migration is most selective), the proportion of homogamous Russian marriages would have to be reduced by 46.5% to get the 2003 odds ratio. Because we assume that the number of intermarrying Russians was stable, the reduction of 46.5% of the homogamous marrying group implies a decrease of 38.3% of the total marriageable Russian population (based on 1991 figures).

If we compare this figure of 38.3% with the actual reduction of the Russian population in the period 1991-2003 through emigration, which was about 19% (our computations on the basis of figures obtained from LCBS), it becomes clear that, at most, half the reduction in the odds ratios can be explained by selective emigration. Cross-tabulations of emigrants by age, gender, marital status, and ethnicity are not available. However, the distributions for marital status, age, and ethnicity suggest that people who were "at risk of marrying" (i.e., were single) and people aged 20^9 were not substantially overrepresented among emigrants after independence. The overall emigration rates are thus expected to reflect the emigration of marriageable Russians. Note that our estimate for the contribution of selective migration to increased intermarriage rates presents an upper bound because it assumes that all marriageable Russians who left the country would have married within their own group if they had stayed in Latvia. If we drop this assumption, the part of the increase in intermarriage that is explained by selective emigration becomes smaller.

We also conducted this simulation for the period of highest emigration, 1991-1994. A reduction of homogamous Russians marriages by 28.5% (which means 23.5% fewer marriageable Russians) would be needed to produce the observed reduction in the odds ratio to the value of 19.0 in 1994. The actual emigration in this period was 11% at most (Codagnone 1998; LCBS 2004).

Thus, although the increasing tendency toward intermarriage among the Russians after independence may have been due, in part, to selective emigration, our simulation results indicate that half or more of it must have been due to processes that were taking place within Latvia and to the slowdown of immigration. That intermarriage continued to increase after the mid-1990s, when emigration was much lower, and even after 2000, when emigration had almost completely stopped, also suggests that integrative processes in Latvia are responsible, to a substantial extent, for the decreasing odds ratio for within-group marriage.

CONCLUSION

In our analysis of the patterns and trends in ethnically mixed marriages between Latvians and Russians (including Belarussians and Ukrainians) in Latvia between 1970 and 2003, we found that the tendency toward ethnic intermarriage was strong during the entire period. About 1 in 3 marriages was mixed when we considered all ethnic groups separately, whereas about 17% of all marriages were mixed when we considered intermarriage between Latvians and Russians (including Belarussians and Ukrainians) only. The average odds ratio for Latvian-Russian intermarriage from 1970 to 2003 was 22.0, which means that the odds of marrying within one's own ethnic group were 22 times larger than the odds of marrying outside one's own ethnic group (irrespective of group size). This finding may be seen as a clear preference for marrying within one's own ethnic group. However, the odds ratio is relatively weak compared to, for example, the odds ratios of 104 for Hispanic-white intermarriage in the United States, of 361 for Czech-Slovak intermarriage in the former Czechoslovakia, of 793 for Kurdish-Turkish intermarriage in Turkey, and of 11,291 for black-white intermarriage in the United States (Boguszak and Bozon 1989; Gndz-Hosgr and Smits 2002; Qian 1997:275, calculated from table Al for 20- to 29 year olds in 1990).

Our central goal in this article was to describe how Latvian-Russian intermarriage was affected by Latvia's independence. We tested three hypotheses about the expected direction of this effect. According to the increased ethnic consciousness hypothesis, ethnic tensions, nationalistic movements, and political conflicts after independence would lead to heightened ethnic awareness and greater social distances among the groups and hence to less Latvian-Russian intermarriage. The stable preferences hypothesis, on the other hand, argues that these tensions should be interpreted as the public expression of deeper emotions that existed earlier but could not be expressed during the Soviet period. Since marriage choice was free in the Soviet period, these sentiments were probably already reflected in the intermarriage rates. Therefore, this hypothesis predicts intermarriage rates to be hardly influenced by independence. The third hypothesis focuses on the effects on intermarriage of the changed migration patterns after independence. This changed migration hypothesis predicts an increase in Latvian-Russian intermarriage after independence because of the selective out-migration of Russians and the drying up of the inflow of new immigrants.

After having been more or less stable until the end of the 1980s, Latvian-Russian intermarriage (as measured by the odds ratio) started to increase immediately after the independence year 1991 and continued to increase until 2003, the last year for which data are available. Only around 1996, the period of highest ethnic tension, was there a decrease, which was completely compensated for in the next year by an extra increase (suggesting perhaps that planned marriages were not canceled but only delayed). Because of this unexpected increase in intermarriage after independence, we have to reject both the increased ethnic consciousness hypothesis and the stable preferences hypothesis. Apparently, the political tensions at the macro level did not make the distances between the ethnic groups larger. An interesting finding was that the increase in Latvian-Russian intermarriage after independence was due largely to changes by Russians (especially Russian women), for whom intermarriage increased from about 16% to almost 25%. For Latvians, the changes in the percentages of intermarriage were much smaller. There was a slight decrease around the time of independence and during the crisis year 1996, but afterward, they increased again to almost the preindependence level. This relative stability among Latvians, in combination with the reduction in the Russian population because of emigration indicates that the preference for intermarriage, in fact, increased among the Latvians. Given the decrease in the Russian population, stable preferences among the Latvians would have implied decreasing percentages of intermarriage for the Latvians.

The increasing intermarriage percentages of the Russians are consistent with the changed migration hypothesis. It therefore seems likely that at least part of the increase in Latvian-Russian intermarriage after independence was caused by changes in the migration patterns. Simulation analyses suggested that selective emigration can explain at most half the increase in Latvian-Russian intermarriage (measured by the odds ratio) in the period 1991-2003. Moreover, after 2000, when emigration had largely come to a halt, the increase in intermarriage continued. This finding implies that other integrative processes played an important role as well.

In addition to hypotheses about the effect of independence on the level of intermarriage, we tested a hypothesis on the exchange of favorable and unfavorable characteristics in the (inter)marriage market. Our results show that this hypothesis has to be rejected for educational attainment. We did not find an overrepresentation of marriages between less-educated members of the dominant group and more highly educated members of the subordinate group.

What kind of people tend to intermarry? The results of our logistic regression analysis make clear that less-educated Latvians and Russians more often choose partners from the other group than do those who are more highly educated. A likely explanation for this finding is the ethnically segregated Latvian school system (as it existed during the Soviet period). Staying longer in this system means staying longer in both an ethnically and an educationally homogeneous environment. We also found that ethnic homogamy interacts with educational homogamy: partners with the same educational level have higher odds of being of the same ethnicity. As may be expected, people from mixed family backgrounds have a higher probability of having mixed marriages themselves. With regard to place of residence, we found that Russians in the countryside intermarried more than Russians in the cities and that Latvians in the countryside intermarried less than Latvians in the cities.

The fact that the rise in intermarriage of Russians after independence can only partly be explained by selective emigration makes one wonder which other factors may be responsible. The drying up of the large stream of nonintegrated Russian immigrants after independence is one candidate. However, there are also complementary, more substantial explanations. First, Russians (especially younger Russians) who were born in Latvia have more incentives to take an active part in Latvian society now than in the past. For instance, the Latvian language has been taught much more actively in Russian schools since independence than it was during the Soviet period. Moreover, Latvian language skills are needed in the labor market, and young Russians often go on to Latvian language tertiary education (since Russian programs are no longer available for all subjects). The change in the overall attitude toward integration is also reflected in the percentage of children from mixed marriages who choose Latvian (as opposed to Russian) nationality when they receive their passports at age 16. In 1989, 78% of children from Latvian-Russian marriage opted for Latvian nationality, whereas in 1998, 93% did so (Ezera 1999). Whether the decrease in marriage rates and the increase in cohabitation in the early 1990s have played a significant role remains to be seen. The fact that marriage rates were similar for Latvians and Russians and that the decrease in marriage rates was also similar does not point in this direction. To check for possible cohabitation effects, we repeated our analyses on the LFFS data for all relationships, including cohabitation. We obtained the same results as were presented here for marriages.

Finally, we would like to draw some more general conclusions on the use of intermarriage for the study of ethnic relations. Our results suggest that intermarriage figures are sensitive to changes in the relationships among groups. We observed clear changes in both intermarriage percentages and odds ratios at the time of independence. Also, the period of highest ethnic tension in the mid-1990s seems to be reflected in the intermarriage rates. In this respect, our results support the claims in the social science literature that intermarriage is a major indicator of the quality of the relationships among groups. With regard to the cohesion-producing (or conflict-reducing) aspect of intermarriage, our findings also seem in line with the claims in the literature. Before independence, a substantial proportion of the marriageable population (about 16%) tended to marry partners from the other ethnic group. This rate is relatively high compared with ethnic intermarriage rates in other societies, where more serious ethnic conflicts have occurred. For example, in Turkey in the 1980s, about 8% of Kurds married Turkish partners (Gunduz-Hosgr and Smits 2002), and in the former Yugoslavia the percentages of Croatians, Muslims, and Albanians who married Serbian partners were only 4%, 1%, and 0.4%, respectively, before the outbreak of the civil wars (Smits 2002). Of course, many factors play a role in ethnic conflicts, and the available evidence is much too sparse to draw any firm conclusions. Still, we think that with 25% mixed marriages among the Russians in 2003, Latvia is heading toward a safe haven with regard to the risk of violent conflict. In the coming years, a substantial proportion of the children born in Latvia will have mixed-family backgrounds, and our results make clear that children born in mixed families have 50% higher odds of choosing mixed marriages themselves.