INSIGHT INTO ETHNIC FLUX: MARRIAGE PATTERNS AMONG JEWS OF MIXED ANCESTRY IN ISRAEL*

Increases in rates of ethnic and racial intermarriages, coupled with larger flows of immigrants from diverse countries of origin, have led to growing proportions of persons of mixed ancestry and backgrounds in many societies. Findings from the 2000 U.S. census indicate that 2.4% of all the respondents and 4.0% of the respondents under age 18 identified themselves as belonging to more than one race (Jones and Smith 2001). Because ethnic and racial exogamy tends to be more common among second-generation than among first-generation immigrants, the growing size of the second generation in the United States will likely lead to larger proportions of multiracial and multiethnic individuals in the future (Gshur and Okun 2003; Portes 1994; Qian and Lichter 2001). Studies of intermarriage in the United Kingdom, France, Sweden, and Israel have suggested that future generations in these countries will also have significant proportions of multiracial and multiethnic persons (Coleman 1994; Cretser 1999; Okun 2001). Patterns of intermarriage are thus challenging concepts of race and ethnicity and are leading to complex views of ethnic and racial boundaries (Harris and Sim 2002; Hirschman 2003; Rockquemore and Brunsma 2002).

In this article, I tie together research on assimilation, ethnicity, and ethnic intermarriage by focusing for the first time on the marriage patterns of persons of mixed ethnic backgrounds, specifically the marriage patterns of Jews of mixed ancestry in Israel. There are several aspects of the Israeli case that make for interesting comparisons with other multiethnic societies. First, because of the fairly recent history of the first massive Jewish immigration to Israel, only a small minority of marriage-age Jews in Israel today can trace their roots in Israel past their grandparents' generation. Therefore, Israelis are at a much earlier stage in terms of generational effects than are European immigrant groups in the United States (Lieberson and Waters 1988). In contrast, Israelis are at a more advanced stage in terms of generational effects than are many of the recent immigrant groups to the United States, such as Koreans and Filipinos (Lopez and Espiritu 1990).

second, in Israel, large socioeconomic gaps remain between major Jewish ancestry groups, and residential segregation is pronounced in certain regions of the country. This situation contrasts with the case of the Southern and Eastern European immigrants in the United States, who benefitted from sustained economic expansion after 1940 and have been able to close socioeconomic gaps vis-r-vis earlier immigrants from Northern and Western Europe (Alba 1990). The gap in socioeconomic status between Jewish ancestry groups in Israel, however, is similar to that of some immigrant Hispanic and Asian groups in the United States today (Waters and Eschbach 1995).

Third, Israel may be unique in that its national ideology and governmental policy have been actively directed toward the realization of the "melting pot" ideology. In particular, Zionist ideology largely denies the importance of ethnicity that reflects Jewish experience in the Diaspora. Rather, Zionism has idealized a new national Israeli Jewish culture, and official Israeli policy has envisioned the complete assimilation of Jews from diverse countries of origin within three generations (Goldscheider 1996).

A brief review of the history of immigration to Israel highlights the period between 1948, when the State of Israel was founded, and 1951, during which about 700,000 immigrants arrived and doubled the previous Jewish population.1 Substantially lower, but still significant levels of immigration in the 1960s and 1970s continued to contribute to the diversity of Israeli society. Differences in culture, language, and demographic characteristics were dramatic between Jews from Eastern and Central Europe, Muslim North Africa, and the Muslim Middle East (Friedlander and Goldscheider 1979). Following a lull during the 1980s, the early 1990s brought another substantial wave of immigration-from the former Soviet Union.

Despite the immigration of Jews from diverse origins, Israeli society has always been dominated by Ashkenazim (immigrants from Europe and their descendants). Some of these Jews were the first to immigrate to preindependence Israel and were the founders of most political, economic, and cultural institutions. In general, Jewish immigrants from Muslim countries in North Africa and the Middle East were characterized by lower levels of socioeconomic status, as well as higher levels of fertility and mortality. For a variety of reasons, including the placement of newly arrived immigrants in different geographically and economically peripheral regions, as well as the differences in the timing of their arrival in Israel, Jews of North African and Middle Eastern ancestry have been disadvantaged socioeconomically (Cohen and Haberfeld 1998; Friedlander et al. 2002).

Over time, ethnic flux in Israel has led to the evolution of new ethnic groups (Goldscheider 1996; Matras 1973; Shavit and Stier 1997). For example, a broadening of the basis of ethnicity has resulted in the creation of a panethnic identity-Israelis of Middle Eastern or North African descent-known in Hebrew as Mizrahim (persons of Eastern origin). Consistent with a framework outlined by Lopez and Espiritu (1990), it is likely that the creation of the panethnic identity came about, despite cultural and socioeconomic diversity among Israelis of Middle Eastern and North African descent, because of broad differences between these groups and the Ashkenazi group in socioeconomic status, residential location, cultural and religious practices, and the generally darker skin phenotype of the Mizrahim. The split between Mizrahim and Ashkenazim has become dominant in Israeli Jewish society, especially because it is understood largely in terms of class inequality (Ben-Rafael 1982; Smooha 1993).

Ben-Rafael and Sharot (1991) discussed why the North African and Middle Eastern groups have a much greater ethnic consciousness than have the Ashkenazim. In coming to Israel, Ashkenazim made a conscious and desired break from the cultural and social features of the Jewish communities from which they originated. In contrast, Jews from North Africa and the Middle East conceived of immigration to Israel as a means of continuing their cultural practices and traditions.

THEORETICAL CONSIDERATIONS

Ethnicity and race are increasingly understood among social scientists as fluid social constructs, rather than as fixed or biologically determined characteristics. Identification with particular ethnic or racial groups is often viewed as stemming from economic or political interests in group or community membership or from advantages gained through social networks and other forms of social capital (Portes and Zhou 1993). Alternatively, identification can be seen in terms of "symbolic" ethnicity, which is increasingly "optional" for some multiethnic or multiracial groups. In cases of symbolic ethnicity, ethnic identity does not affect social behavior or life chances, but is understood as an individual choice that may be more salient or less salient at different stages of the life cycle (Gans 1979).

Alba (1990) suggested that among U.S. non-Hispanic whites, ethnicity is largely symbolic, and that ethnic intermarriage among U.S. non-Hispanic whites has contributed to the formation of a new European American ethnic group. This European American group, which is largely homogeneous across component groups with respect to a wide range of socioeconomic characteristics, is seen as partially replacing other more specific forms of ethnicity, such as German American or Polish American. Moreover, there has been evidence of growth in the proportion of non-Hispanic U.S. whites who have reported their ethnicity simply as "American" and speculation about the formation of a new American ethnic group (Lieberson and Waters 1988).

The evolution of symbolic ethnicity among non-Hispanic whites in the United States is largely consistent with classic assimilation theories, which suggest that persons of mixed ancestry are only weakly identified with their particular ethnic backgrounds (Gordon 1964). Among non-European Americans in the United States-such as some Hispanic or Asian groups-ethnicity is also seen as increasingly symbolic, particularly for those of mixed ancestries (Eschbach and Gomez 1998).

When ethnicity or race is symbolic, there may be individual- or family-level variation in ethnic or racial identity that is associated with socioeconomic status. The relationships between socioeconomic status and ethnic identity appear to depend on the aggregate socioeconomic status of relevant groups, as well as on the extent to which ethnic identity is seen as a cultural resource (Eschbach, Supple, and Snipp 1998; Qian 2003; Waters 1994). For example, among Hispanics in the United States, higher family status is associated with less-consistent positive responses to questions about Hispanic ethnicity (Eschbach and Gomez 1998; Qian 2003). In contrast, parents' greater educational attainment is associated with Asian racial identification among third-generation biracial Americans of Asian descent (Xie and Goyette 1997). The associations between educational attainment and ethnic identity are important for understanding patterns of intermarriage because education is an important direct factor in the selection of mates via well-known patterns of educational homogamy (Kalmijn 1991; Mare 1991), as well as a potentially important indirect factor via its associations with ethnicity and ethnic identity and therefore with ethnic endogamy (Qian, Blair, and Ruf 2001).

How is ethnicity among persons of mixed ancestry reflected in marriage behavior? It may be that over time, a new "Israeli" ethnicity is being created, and that there is a concomitant decline in the salience of a distinct "Mizrahi" ethnicity. If so, I hypothesize that there is a tendency for persons of mixed ancestry to marry among themselves, a tendency that I term "ethnic endogamy" among persons of mixed ancestry. There are at least two reasons why these changes may be expected. First, according to Zionist ideology, ethnic intermarriage is a means of creating a single, unified culture among Jews in Israel (Rosen 1982). second, there has been little immigration to Israel from countries of North Africa and the Middle East since the mid-1960s, thereby cutting off, in Massey's (1995:642) words, "the supply of raw materials for the grist mill of ethnicity" and enabling generational effects to weaken ethnic identity.

In contrast, it may be that persons of ethnically mixed unions do not form a new "Israeli" identity, but rather identify and are identified primarily with one of the two major origin groups. If so, I hypothesize that persons of mixed ethnic ancestry exhibit one-sided marriage behavior in the sense that, among those who marry Mizrahim or Ashkenazim, marriage with one origin group is more common than marriage with the other. Consequently, the proportion of people who identify themselves with the "favored" group may actually increase more over time than if there were no ethnic intermarriage. This has been shown to be true in the case of Irish Americans (Hout and Goldstein 1994). The salience of Mizrahi identity could increase or decrease, depending on which is the favored group.

Another possibility is that persons of mixed ancestry marry with ethnic groups according to their educational attainment. Assimilation theories suggest that education tends to weaken ethnic attachments and to increase exposure to people of different ethnic origins, thereby leading to a loss of ethnic identity among upwardly mobile members of disadvantaged ethnic groups (Alba 1990; Qian et al. 2001). In the Israeli context, the assimilation hypothesis suggests that the greater contact of educated Mizrahim with Ashkenazim has led to the erosion of their Mizrahi ethnic consciousness, more frequent intermarriage, and the consequent concentration of Mizrahi ethnicity in the lower classes (Ben-Rafael 1982). If assimilation theory is applied to the case of persons of mixed ethnic ancestry, the more educated among them should be expected to have a less-well-defined Mizrahi ethnic consciousness and therefore to be less likely to marry Mizrahim, whereas those who are less educated should be expected to be more likely to marry Mizrahim.

In contrast to the assimilation hypothesis, the awareness hypothesis states that education may increase social sensitivity to ethnicity and ethnic distinctiveness, so that when ethnicity is optional, educated ethnics are more likely to have a stronger ethnic identity and are thus less likely to marry out of their ethnic groups (Qian 2003; Xie and Goyette 1997). In the case of Israel, the awareness hypothesis suggests that among Mizrahim and persons of mixed ancestry, the educated are more likely to have a strong Mizrahi ethnic identity and hence are more likely to marry Mizrahim.

DATAANDVARIABLES

The source of data for this study was a file that contains linked records from the 1995 20% sample of the Israeli census and the 1983 100% sample of the Israeli census. As is explained later, the linked records were necessary to define the ethnicity of third-generation persons in 1995 (Friedlander et al. 2002). The data provide information on current marital status, number of times married, and year of first marriage as of 1995.1 defined the sample of 15,997 married couples as husbands and wives who were living in the same household in 1995 (1) who married in Israel, (2) in which the wives married for the first and only time between 1986 and 1995, and (3) in which the wives were aged 18-40 in 1995. By restricting the study to marriages that occurred within a relatively short period prior to the census, I reduced potential biases that occur if marital dissolution of exogamous couples is more common than that of endogamous couples. According to Eisenbach (1992), such biases are not large, partially because overall rates of marital dissolution in Israel are still lower than those of other countries, such as the United States.2

Census data from 1995 contain information on the respondents' countries of birth, as well as the countries of birth of the respondents' parents. However, these data do not contain information on the grandparents' places of birth, and this omission poses a problem for defining ethnicity among third-generation Israelis (those with at least one native-born parent).3 Therefore, I turned to the linked data file. In the 1995 census, persons were linked to the 1983 census records of their parents. In the 1983 records, parents responded to questions on their fathers' countries of birth (that is, the countries of birth of the grandfathers of the marriage-age persons in 1995). Because the data I analyzed provide objective information on the countries of birth of the respondents, their parents, and grandfathers, my research does not suffer from the distortion in reported ethnicity and race that is common in subjective questionnaires, such as that included in the U.S. census (Lieberson and Waters 1988).

I defined the ethnicity of persons in the married-couples sample as (1) Middle Eastern or North African (Mizrahi); (2) European4 (Ashkenazi); or (3) mixed, with one Mizrahi parent and one Ashkenazi parent. For first-generation Israelis (immigrants to Israel), 1 defined ethnicity as Mizrahi or Ashkenazi on the basis of information on the respondents' places of birth.5 For persons in the married-couples sample who were born in Israel, I defined ethnicity as a function of both parents' ethnicity. In the case of persons who were second-generation Israelis (their parents were immigrants to Israel), I defined the parents' ethnicities as a function of their countries of birth. In the case of third-generation Israelis, the parents were themselves born in Israel, so that the parents' ethnicities had to be defined as a function of their fathers' places of birth (i.e., the places of birth of the grandfathers of those in the married-couples sample). My reliance on grandfathers' places of birth to define ethnicity among third-generation Israelis is potentially problematic, since it is possible that individuals' grandmothers may have had different places of birth. However, given the overwhelming prevalence of ethnically endogamous unions in the past, I do not believe that this problem is of practical significance (Okun 2001).

By this definition, persons of ethnically mixed ancestries were a minority in the sample. Among brides and grooms, 11% and 9% were of mixed ancestry, respectively. The proportions of brides and grooms who were of Mizrahi ethnicity were 58% and 60%, with the remaining 32% of the brides and grooms of Ashkenazi origin. The small proportions of ethnically mixed persons among recently married Israeli Jews reflect low levels of Ashkenazi-Mizrahi intermarriage in the more distant past. In the future, I expect to see large increases in these proportions as a consequence of significant increases in Ashkenazi-Mizrahi intermarriage over the past 40 years (Okun 2001).

Explicit consideration of educational attainment is essential in an analysis of ethnic marriage patterns in Israel because of the strong association between ethnicity and educational attainment and because educational homogamy is a central feature of marriage patterns (Friedlander et al. 2002; Okun 2001). I categorized persons in the sample into five ascending groups of educational attainment, according to whether the highest degree or diploma they ever received was (1) a primary or intermediate school diploma, (2) a high school diploma without matriculation certification, (3) a matriculation diploma, (4) a post-secondary school diploma (such as a teaching certificate), or (5) an academic degree (BA or higher). Attaining a matriculation diploma is a critical step in the Israeli educational system because this diploma is a prerequisite for most types of postsecondary education. The completion of high school does not guarantee attainment of the matriculation diploma because the latter requires passing a series of standardized national examinations. As can be seen in Table 1, Ashkenazim are the most educated group and Mizrahim are the least educated, with persons of mixed ancestry in an intermediate position.6

To provide a descriptive assessment of marriage patterns, I present, in Table 2, the distribution of persons in the married-couples sample by sex, ethnicity, educational attainment, and spouses' ethnicity. First, note that overall and within almost every sex-education group, the probability of marrying a spouse of Ashkenazi origin is the greatest among persons of Ashkenazi ancestry, the lowest among persons of Mizrahi ancestry, and intermediate among persons of mixed ancestry. Similarly, the probability of marrying a spouse of Mizrahi origin is the greatest among persons of Mizrahi ancestry and the lowest among persons of Ashkenazi ancestry. In contrast, the probability of marrying a spouse of mixed ancestry varies little by ethnic origin. Thus, there is a strong suggestion of ethnic endogamy among Ashkenazim and Mizrahim, but no prima facie evidence of endogamy among persons of mixed ancestry.

Second, the proportion of Mizrahim who marry other Mizrahim declines with educational attainment, while the proportion of Ashkenazim who marry other Ashkenazim tends to increase with educational attainment. Thus, it appears that the extent of ethnic endogamy may vary by educational attainment. The proportion of persons of mixed ancestry who marry others of mixed ancestry does not seem to vary systematically by education.

Third, there appears to be some one-sided behavior in the marriage patterns of persons of mixed ethnic ancestries. A larger proportion of these individuals marry spouses of Mizrahi ancestry than of Ashkenazi ancestry, but this pattern may be expected, given the larger proportion of Mizrahim than of Ashkenazim in the overall sample.

Fourth, it appears that one-sided behavior by persons of mixed ancestry is a function of education; as education increases, the proportion of persons of mixed ancestry who are married to Mizrahim declines relative to the proportion who are married to Ashkenazim.

Finally, marriage patterns are fairly similar for men and women. Men appear to be slightly more likely than women to marry spouses of mixed ancestry, overall and in most ethnicity-educational groups.

My interpretations of Table 2 may not be reliable because of large ethnic differences in the sizes of groups and in sex-educational composition. I therefore used various log-linear models that control for the marginal distributions, by ethnicity and education, of married men and women. Log-linear models enabled me to distinguish marriage patterns that reflect the associations between spouses' traits from patterns that result from the marginal distributions of husbands' and wives' traits (Mare 1991).7 In addition, multivariate log-linear models allowed me to describe the patterns of ethnic and educational assortative mating net of each other.

LOG-LINEAR MODELS

The log-linear analyses describe mating patterns in a marriage table of 225 cells (225 = (5 3)2, that is, five categories of educational attainment and three categories of ethnicity for each of the sexes). Table 3 presents the fit statistics for eight models (the model equations are available from me on request). Baseline Model 1 involves random mating after marginal distributions of married persons by ethnicity and education are controlled. Model 1 also controls for the joint distribution of ethnicity and education among husbands, as well as among wives, but does not allow for cross-spouse interactions of any kind.

Model 2, a quasi-symmetry model of educational mating, adds to Model 1 ten sex-symmetric, off-diagonal parameters, each representing intermarriage patterns between pairs of different educational groups (Haberman 1979). This sex-symmetric approach assumes that fitted frequencies do not depend on whether the husband or wife is the more educated in each educationally heterogamous combination.8 I selected this approach after I tested a variety of models, including models that allow for asymmetry, as well as crossing models (Mare 1991). The large reductions in the G^sup 2^ and BIC statistics9 between Model 1. which does not allow for assortative mating of any kind, and Model 2, which allows for educational assortative mating, indicate that educational homogamy is an important element in Israeli marriage patterns. The values of the G^sup 2^ and BIC statistics remain large in Model 2, however, indicating that marriage patterns are not well described by a model that considers only educational assortative mating.

All further models build on Model 2. Therefore, patterns of ethnic assortative mating in Models 3-8 are described net of patterns of educational assortative mating, thus avoiding confounding between the two. Fits are substantially better in Models 3-8 than in Model 2. These improvements in fit indicate that ethnic assortative mating makes significant independent contributions to explaining marriage patterns, beyond the patterns described by educational assortative mating. Models 3-8 fit the data reasonably well according to the BIC statistics.

Model 3 adds to Model 2 three ethnic variable diagonal parameters to examine the varying extent of ethnic endogamy among the three different ethnic groups.10 Model 4 differs from Model 3 in that it constrains the extent of ethnic endogamy among Ashkenazim to equal that among Mizrahim and allows only the extent of ethnic endogamy among persons of mixed ancestry to differ from the two. The parameterization in Model 4 thus saves one degree of freedom compared with Model 3.

In Models 3 and 4, no distinction is made between different types of ethnically exogamous unions. In contrast, Model 5 builds on Model 4 by adding an ethnic distance parameter, which allows the fitted frequency of ethnic intermarriages between Ashkenazim and Mizrahim to differ from the fitted frequency of ethnic intermarriages involving spouses of mixed ancestry. Model 5 allows for a cleaner test of the relative frequency of endogamous versus exogamous unions by persons of mixed ethnic ancestry.

Model 6 builds on Model 5 by allowing the fitted frequency of intermarriages between persons of mixed ancestry and Mizrahim to differ from the fitted frequency of intermarriages between persons of mixed ancestry and Ashkenazim. This model is useful for testing the one-sided hypothesis. Model 7 also builds on Model 5 and allows for gender asymmetry in one-sided behavior among persons of mixed ethnic ancestry.

To test the assimilation hypothesis, I constructed a more complex interaction model, which allows patterns of ethnic mating to vary by educational attainment. Model 8 allowed me to test, for example, whether the extent of ethnic endogamy varies by educational attainment, and whether one-sided behavior by persons of mixed ethnic ancestry varies by educational attainment. In this model, parameters of ethnic mating were estimated for couples in which both spouses had matriculation diplomas or higher degrees, while distinct ethnic mating parameters are estimated for all other couples. As Table 3 shows, Model 8 offers the best fit, indicating that patterns of ethnic mating vary by the educational levels of spouses.

RESULTS

I turn to a test of my first hypothesis, which states that a new "Israeli" ethnic identity is manifest in the tendency of persons of mixed ancestry to marry among themselves, rather than among Ashkenazim and Mizrahim. I present here the relative fitted frequencies, net of marginal distributions, of ethnically endogamous unions among Ashkenazim, Mizrahim, and persons of mixed ancestry, based on Model 3.11 The reference category is all exogamous marriages. I computed the relative fitted frequencies of endogamous unions separately for each ethnic group and found the following: while ethnically endogamous unions among Mizrahim and among Ashkenazim have much higher fitted frequencies (2.94 and 2.35, respectively) than all ethnically exogamous unions, ethnically endogamous unions among persons of mixed ancestry actually have lower fitted frequencies (0.72) than all ethnically exogamous unions.12 These relative fitted frequencies differ from 1.00 at the .001 level of significance. Thus, I did not find support for the first hypothesis and suggest instead that persons of mixed ancestry do not tend to marry among themselves.

According to Model 3, the extent of ethnic endogamy among Ashkenazim appears to be similar to that of Mizrahim. In addition, a comparison of the BIC statistics for Models 3 and 4 in Table 3 suggests that it is possible to save a degree of freedom without suffering a significant loss of fit by constraining the extent of ethnic endogamy among the two main ethnic groups to be equal. The exponentiated parameter estimate representing the relative fitted frequency of ethnically endogamous unions among Mizrahim and Ashkenazim is 2.64, while that among persons of mixed ancestry is 0.73. Again, reported fitted frequencies are relative to the fitted frequency of all ethnically exogamous unions and differ significantly from 1.00 at the .001 level. Thus, from Model 4, there is still no evidence that ethnically endogamous unions among persons of mixed ancestry are relatively more frequent than are all exogamous unions.

For a more direct comparison of ethnic endogamy versus ethnic exogamy among persons of mixed ancestry, I turn to the results of Model 5. The relative fitted frequency of ethnically endogamous unions versus ethnically exogamous unions, among couples involving at least one spouse of mixed ancestry, is 0.72 (not significantly different than 1 at the 10% level). Thus, among couples in which at least one spouse is of mixed ancestry, ethnically endogamous couples are no more common than are ethnically exogamous couples. Again the results are not supportive of the new Israeli-ethnicity hypothesis.

I now turn to Model 6 to test the one-sided hypothesis. Recall that this hypothesis states that among persons of mixed ethnic ancestry who marry either Mizrahim or Ashkenazim, marriage with one group is more common than marriage with the other group. The results from Model 6 are not consistent with the hypothesis. The fitted frequency of unions between persons of mixed ancestry and Mizrahim is 0.94 and is not significantly different from the fitted frequency of unions between persons of mixed ancestry and Ashkenazim (the reference category) at the 10% level. Likewise, Table 3 indicates that there is no improvement in fit in Model 6 over Model 5, so that distinguishing between unions involving persons of mixed ancestry and Ashkenazim and unions involving persons of mixed ancestry and Mizrahim does not significantly improve the understanding of marriage patterns. Thus, the findings suggest that persons of mixed ethnic ancestry are not overwhelmingly identified as and do not overwhelmingly identify with either of the two major ethnic groups to the exclusion of the other.

I turn now to the results of Model 7, which allowed me to examine sex asymmetry in ethnic out-marriage among persons of mixed ancestry. There is some evidence that among persons of mixed ancestry who out-marry, men are more likely than women to marry an Ashkenazi, rather than a Mizrahi. In particular, the fitted frequency of unions between grooms of mixed ancestry and Ashkenazi brides relative to the fitted frequency of unions between grooms of mixed ancestry and Mizrahi brides is 26% greater than the analogous fitted frequency of unions between brides of mixed ancestry and Ashkenazi grooms relative to the fitted frequency of unions between brides of mixed ancestry and Mizrahi grooms. This ratio of relative frequencies is significantly different from 1.00 at the .01 level of significance.

Thus, it appears that men of mixed ancestry are closer in marriage behavior to Ashkenazim than are women of mixed ancestry. This finding may reflect broader gender issues, in that it suggests that men of mixed ancestry have greater opportunity to marry up than do women of mixed ancestry. This may be particularly true of educated men of mixed ancestry who are able to exchange their high educational status and presumably greater earnings potential for marriage with Ashkenazi women (Davis 1941; Fu 2001; Merton 1941). Although Model 7 does not provide direct evidence for this exchange theory, a closer look at Table 2 supports this notion. Among persons of mixed ancestry who had academic degrees, 67.0% of the men were married to Ashkenazi women, and only 57.4% of the women were married to Ashkenazi men.

I now turn to a test of the assimilation hypothesis, basing my findings on Model 8. I focus on comparing several estimated parameters of ethnic mating among couples in which both spouses had matriculation diplomas or higher to estimated parameters of ethnic mating among couples involving one or more spouses who did not have at least matriculation diplomas.13 The results are presented in Table 4.

As I discussed earlier, the assimilation hypothesis suggests that more-educated persons of mixed ancestry are more likely to marry Ashkenazim than Mizrahim, while less-educated persons of mixed ancestry are more likely to marry Mizrahim than Ashkenazim. To test this hypothesis, I compared one-sided behavior among persons of mixed ancestry in more-educated couples with one-sided behavior among persons of mixed ancestry in less-educated couples. The fitted frequency of marriages between persons of mixed ancestry and Mizrahim relative to the fitted frequency of marriages between persons of mixed ancestry and Ashkenazim is 1.28 among less-educated couples (see the left-hand panel of Table 4). The analogous relative fitted frequency is 0.57 among more-educated couples (see the right-hand panel of Table 4). The difference in the relative fitted frequencies is statistically significant at the .001 level and indicates that among less-educated couples, persons of mixed ethnic ancestry are more likely to be married to Mizrahim than to Ashkenazim, while among more-educated couples, persons of mixed ethnic ancestry are more likely to be married to Ashkenazim than to Mizrahim.

The log-linear models I presented control for the marginal and joint marginal distributions of the sample, as well as for patterns of educational homogamy. Therefore, the finding that one-sided behavior among persons of mixed ancestry differs as a function of educational attainment does not stem from the facts that Ashkenazim tend to be more educated in general and that persons tend to marry spouses with similar levels of education.

The assimilation hypothesis also suggests that, in the case of Israel, more-educated Mizrahim exhibit less ethnic endogamy than their less-educated counterparts. This indeed appears to be the case in that the relative fitted frequency of ethnically endogamous marriages among less-educated couples involving at least one Mizrahi spouse (4.14) is significantly greater at the .001 level than is the relative fitted frequency of ethnically endogamous marriages among more-educated couples involving at least one Mizrahi spouse (1.32). The difference in the extent of ethnic endogamy among Mizrahim in the two types of couples is striking and is substantively and statistically larger than the differences in the extent of ethnic endogamy by educational attainment among the other ethnic groups.14 Overall, my empirical results are more consistent with the assimilation hypothesis than with the awareness hypothesis.

SUMMARY AND DISCUSSION

My analyses lead to related findings regarding the marriage patterns of persons of mixed ancestry in Israel. There is no suggestion of ethnic endogamy among persons of mixed ancestry. That is, the as-yet small, but growing, group of marriage-age Israelis who are of mixed ethnic origins do not marry among themselves to any great extent. Rather, the findings suggest that ethnic mating among persons of mixed ancestry in Israel is largely a function of educational attainment. In particular, among persons of mixed ancestry, those who are more educated are more likely to marry persons of Ashkenazi origin, while those who are less educated are more likely to marry persons of Mizrahi origin.

That the structural integration of Mizrahim into the dominant society is imperfect is evidenced by the continuing existence of group differences in education, occupational status, income, and place of residence-despite a convergence in some types of social and demographic behavior over time (Friedlander et al. 2002). The continuing association between ethnicity and life chances is well known both in academic and popular circles. Indeed, the recent success of some Mizrahi politicians has often been a result of their appeal to disadvantaged Mizrahi ethnic groups, and much of Israeli politics has been ethnically based.

My findings suggest that if ethnicity is more than symbolic in nature-as is the case in Israel-then ethnic intermarriage may not have the effect of reducing the salience of ethnicity. This counterintuitive conclusion stems from two streams of research. The first, based on a well-established literature on Israel and other societies, shows that there is selective out-marriage of the more educated members of disadvantaged ethnic groups (Eisenbach 1992; Fu 2001; Goldscheider 1996; Qian 1997). The second, presented here, takes the literature one step further by examining the marriage behavior of the offspring of these interethnic marriages. I showed that the marriage behavior of persons of mixed ancestry concentrates more-educated persons of mixed ancestry in families with Ashkenazi spouses and less-educated persons in families with Mizrahi spouses. This pattern perpetuates the historically important association between Mizrahi ethnicity and low socioeconomic status, which is the basis of the popular understanding of current forms of Mizrahi ethnicity. Thus, I suggest that ethnic intermarriage itself should not necessarily be viewed as a vehicle for reducing group differences over time or for reducing the salience of ethnicity among disadvantaged groups.

To the extent that ethnic identification is reflected in marriage patterns, ethnic identity in Israel is conditioned on socioeconomic status, with more-educated persons of mixed ancestry indicating less identification with Mizrahim compared with their less-educated counterparts. My findings stand in contrast to the limited extant evidence on the effects of individual-level socioeconomic status on ethnic identity in cases in which ethnicity is symbolic or optional. In these instances, ethnic awareness among more-educated individuals may be stronger than assimilation effects that are associated with education (Eschbach 1995).

I now return to my negative finding-that I did not find evidence in marriage patterns of the formation of a new national Israeli ethnicity. My interpretation of this finding is that persons of mixed ancestry do not form a cohesive group and do not share a common identity. They may not do so partly because Jews of mixed Mizrahi-Ashkenazi origin are a group that knows no ethnic boundary among Jewish Israelis; there is no "other" group that defines them as distinct, since all other Jews fall into one of their component ancestries. In addition, I suggest that a necessary condition for the development of a new national ethnic identity in an immigrant country is the elimination of important structural differences between the various component groups. I suggest that gaps in socioeconomic status between Mizrahim and Ashkenazim have impeded the assimilation of Mizrahim along other dimensions, thus working to maintain the ethnic salience of Mizrahim and hindering the development of an assimilated Israeli ethnicity.

In contrast to the case of the Mizrahim in Israel, structural differences have been eliminated among U.S. non-Hispanic whites, and there are the makings of a new European American ethnicity (Alba 1990) and perhaps a new American ethnicity (Lieberson and Waters 1988). However, there are examples of recent immigrant groups in the United States whose future socioeconomic assimilation is still in doubt (Portes and Zhou 1993). In this sense, the relative socioeconomic status of these groups is similar to that of Mizrahim in Israel. Thus, the marriage behavior of some persons of mixed ancestry in the United States may parallel that of persons of mixed ancestry in Israel. Future research could investigate the marriage behavior of persons of mixed backgrounds in the United States and other countries to shed light on changes in the meaning of race and ethnicity in increasingly multiracial and multiethnic societies.