Female Headship: Testing Theories of Linear Assimilation, Segmented Assimilation, and Familism among Mexican Origin Women Elizabeth Wildsmith Abstract This study examines how levels of female headship, non-marital fertility, and divorce among Mexican origin women aged 18-59 compare to levels among the non-hispanic white majority. Change in these family patterns are measured over time and across generations to test three theories of assimilation, linear assimilation, segmented assimilation, and familism. Whether Mexican Americans will follow the patterns of assimilation and integration experienced by European immigrants and their descendents is hotly debated. Some researchers argue that not enough time has passed or generational distance occurred in the Hispanic population for assimilation to be widespread (Alba, 1995). Others argue that the unique experience of particular ethnic groups once in the United States will prevent assimilation towards the mainstream population. Rather than experiencing socioeconomic improvement over time, these groups will remain distinct from the white majority both economically and culturally (Portes, 1995; Portes and Zhou, 1994). This analyses uses data from the IPUMS to analyze changes in Mexican American female headship from 1880-1990. Regression analyses of patterns of female headship, divorce, and non-marital fertility from 1960-1990, offer no support for either the linear assimilation of familism theories. Analyses of generational changes using the 1995 June CPS provide mixed support for the linear assimilation and segmented assimilation theories.
INTRODUCTION The United States has recently experienced a substantial growth in the Hispanic population, which by 1999 accounted for 12 % of the total population. The majority, 65 %, of the Hispanic population is comprised of people of Mexican origin, and this percentage has continuously increased over time. Mexicans (as well as most other Hispanic groups) are relatively economically disadvantaged, with poverty rates being three times higher among Hispanics than whites (Ramirez, 2000). While European origin immigrants that came to the US at the turn of the century also were economically disadvantaged and experienced discrimination, they slowly gained economic parity and cultural similarity with the majority population and now are considered fully assimilated. Whether this pattern of assimilation and integration will be experienced by other groups, such as Mexicans, who have clearly not achieved economic parity, is hotly debated. Some researchers argue that assimilation is a slow process; not enough time has passed or generational distance occurred in the Hispanic population for assimilation to be very widespread (Alba, 1995). Others argue that the unique experience of particular ethnic groups once in the United States will prevent assimilation towards the mainstream population. Rather than experiencing socioeconomic improvement over time, these groups remain marginalized and may even become more disadvantaged, remaining distinct from the white majority both economically and culturally (Portes, 1995; Portes and Zhou, 1994). This paper illuminates this debate through a comparative analysis of trends in Mexican American female headship and two of its components, divorce and non-marital fertility. The US has also recently experienced dramatic changes in family structure and family formation patterns such as higher levels of divorce, decreasing levels of fertility (decreasing marital, increasing non-marital), and later ages at marriage. Though different racial/ethnic 2
groups have been experiencing change at different rates, all have been experiencing change. We know that Hispanic origin women are less likely to be currently married and more likely to be female householders with no spouse present than are non-hispanic white women (Ramirez, 2000). However, we also know that family patterns are quite distinct for different ethnic groups. Research has shown that rates of marriage, divorce, fertility, and female headship vary considerably for Mexicans, Cubans, Puerto Ricans, and other Hispanics (Bean and Tienda, 1987). Given that Mexican Americans comprise the largest percentage of Hispanics, we know surprisingly little about certain aspects of family patterns among this group and how these patterns have changed with exposure to the majority culture. Family structure and behaviors are affected both by socioeconomic conditions (structural factors) and the norms and values of the community/communities to which a person belongs (cultural factors). Change in family behaviors will reflect changing socioeconomic situations and changing norms. Thus, once socioeconomic factors are controlled, family behavior becomes a crude indicator of cultural assimilation. As Keefe and Padilla (1987:129) point out, Family structure is an important aspect of ethnicity [it] is essential to understanding the three processes of culture change (acculturation, assimilation, and ethnic identification). While much research has focused on levels of fertility or divorce, very little has looked at levels of female headship, particularly among Mexican Americans. Examining female headship is especially interesting for two reasons. First, there is a strong association between poverty and female headship. Roughly 35% of all female-headed households live in poverty, and this percentage is much higher among black and Hispanic women (Garfinkel and McLanahan, 1986). Female headship most often occurs as a result of divorce (in which case a potential income is lost) or having children out of wedlock (in which 3
case there are more people to care for on the same income). The negative effects of single parenthood on children have been well documented. The children of these families are much more likely to be poor, to drop out of high school, to have a teenage birth, and suffer from developmental delays (McLanahan and Sandefur, 1994). Secondly, looking at changes in female headship will help us to better understand the factors that affect immigrants experiences as their population continues to grow. This paper investigates three theories of assimilation by comparing patterns of female headship, divorce, and non-marital fertility for the Mexican origin population to patterns for the white majority population. The analyses document how these behaviors have changed over time and across generations. THEORIES OF ASSIMILATION Assimilation refers to the social, economic, and political integration of a minority group into the mainstream, the majority group. For this to occur they must become acculturated as well as accepted by the dominant group (Keefe and Padilla, 1987). The theory implies that assimilation follows a linear path; there are a series of stages that an ethnic group will move through as they become more assimilated. While there are many aspects to becoming fully assimilated, three have been emphasized in earlier studies. These aspects are described most clearly by Gordon (1964), and while there have been minor revisions to the theory, the basic concepts have remained the same. The first three stages, in order, are cultural assimilation or acculturation, followed by structural assimilation, and lastly marital assimilation. Cultural assimilation refers to the adoption of norms, language, and values of the majority group by the minority. Widespread contact of the minority group with the majority group will result in structural assimilation, reflected by similarities in levels of education, income, occupation and 4
area of residence. Marital assimilation occurs when cultural and structural assimilation is widespread enough to weaken the boundaries between groups to overcome norms against intermarriage. People are more likely to marry people that are like themselves, thus intermarriage serves as a key indicator of the level of mutual acceptance between two different groups. Assimilation theory argues that this pattern will occur (and already has for many) the longer a particular ethnic group has been in the United States and the farther removed (generationally) individuals are from the country of origin. This type of change will arise over time as a result of contact between different cultural groups. Previous research has found that ethnic groups that have been in the United States longer tend to experience greater levels of assimilation than do groups that have been in the United States for less time (Alba and Golden, 1986). This greater likelihood of assimilation is the result of increasing levels of interaction between groups, resulting from increasing amounts of exposure of groups to each other. While this change may affect both groups, the assimilation model implies that the dominant group will exert a stronger influence. Change also occurs among the second and later generation because they are farther removed from the cultural influence of the homeland. As older, currently more ethnic generations are replaced by their children and grandchildren, who are less ethnic on average, the group as a whole becomes less ethnic (Alba, 1995:17). These children and grandchildren are going to have more exposure to the mainstream culture than to the ethnic culture, thus are more likely to be influenced by the mainstream. While at times linear assimilation theory has been criticized for supporting a melting pot view of society rather than a multicultural view, it has recently been reasserted as one of the best explanations for the changes experienced by ethnic minorities in the United States (Alba, 1995; Alba and Nee, 1997). 5
Linear assimilation theory was developed based largely on the successful acculturation and assimilation of the descendants of European Immigrants from the turn of the century (Pagnini & Morgan, 1990; Alba & Golden, 1986; Lieberson & Waters, 1988). Though Mexican immigration to the United States began prior to the immigration of many of the European groups that have already achieved socioeconomic equality with the majority population, we know that Mexican Americans are far from socioeconomic parity. The historical experience of Mexican immigrants and their descendents in the United States is distinct from that of the turn of the century European immigrants in at least two ways. First, many of the Mexican immigrants became involuntary US citizens when the boundaries of the United States expanded westward and absorbed Mexican lands (Martin and Midgley, 1994). Some have argued that this creates a psychological difference between European immigrants and Mexican immigrants, with many Mexican origin people feeling as if they were in their homeland rather than as aliens in a foreign place (Murguia, 1982). Another difference, probably the most discussed in relation to theories of assimilation, has been the continued high rates of immigration from Mexico throughout much of the 20th century. European immigration slowed to a trickle after 1920 and thus, over time, descendents of immigrants had less exposure to people who were born in the homeland leading to a decline and dilution of a specific cultural influence (Martin and Midgley, 1994). High levels of immigration have kept the proportion of the Mexican origin population who are first or second generation very high. According to the census, in 1970 roughly 23% of the Mexican origin population was foreign born increasing to 35% of the population in 1980 (Bean and Tienda, 1987). And, these figures are probably underestimates due to undocumented immigration. If the effect of cultural influence weakens the farther removed people become from the immigrant generation, one would expect current Mexican society to 6
maintain some cultural influence due to the consistently high proportion of immigrants in the population. While linear assimilation theory still has its proponents, other theories attempt to account for continuing discrepancies between minority and majority groups. Familism theory focuses on differences in fertility, divorce, and family structure due to the persistence of strong cultural influences that shape values and behaviors (Vega 1990; Grebler, Moore and Guzman 1970). Since socioeconomic assimilation follows cultural assimilation, according to linear assimilation theory, once SES is controlled the effect of cultural factors should weaken over time and across generations. Familism on the other hand, argues that cultural values and norms specific to Mexicans will continue to exert an influence on behavior both over time and across generations. Because many Mexican Americans maintain considerable contact with others of Mexican origin, especially first generation members, certain cultural beliefs and values are more likely to persist. Some argue that this maintenance of culture occurs because it is psychologically rewarding to be with members of one s own group; ethnic loyalty and identity are promoted by one s own group members (Keefe & Padilla, 1987). Familism thus results in the lower levels of divorce, maintenance of stronger family ties, and higher levels of fertility associated with Mexican culture, despite socioeconomic characteristics associated with higher levels of family disruption. In reference to Mexican Americans, the popular view implies that they are very family oriented, more emotionally bound to extended family and children than are non-hispanic whites and that they are unusually warm, caring, and protective of their family (Del Castillo, 1984). Though there is no ideology that describes all Mexican Americans, most of what we know about Mexican origin families historically (prior to the 20th century) is based on the experiences of the middle and upper classes (Erickson, 1998; Del Castillo, 1984). While there did exist an 7
extraordinary romantic view of love, we know that one of the ideals was paternal authority, by which the father determined to whom and when his daughters would marry. This authority existed as a means to maintain racial and religious purity (Del Castillo, 1984). Additionally, other ideals stressed a strong orientation to marriage, motherhood, and fidelity and discouraged overt sexuality, either within or outside of marriage, among women (Erickson, 1998). Yet there were differences between the working class and the upper classes and between ideals and reality. While fidelity and marriage were held up as ideals, there existed a large number of union libres (common law marriages) among the working class, often because civil ceremonies were not feasible. In addition, even among the upper class and clergy there were many men who engaged in these free unions outside of their marriages. These free unions were not stigmatized, though the incidence of them did decline over time. Divorce also was not uncommon among the working class historically and was more common in this group than among the upper and middle classes. During the late 19th century up to 25% of Mexican American households were headed by women with no spouse present, though much of this was due to a husband s absence while they were looking for work (Del Castillo, 1984). It is most likely the religious prescriptions and ideals of patriarchy among the upper classes that has continued to fuel the belief in familism. An alternate view to familism and linear assimilation is segmented assimilation. This theory suggests that the experience, to various degrees, by most Mexican American groups, of racial prejudice, economic displacement, poverty, and marginalization when confronted with American conquest, has led to the emergence of a culture specific to Mexican Americans that is unique from both Mexican and white cultures (Del Castillo, 1984). Keefe and Padilla (1987:7), in their research on the Mexican American family, have found, 8
that the culture of Chicanos who have lived for generations in the US is distinctive, and moreover, a third way of life possessing many unique features, rather than simply an amalgamation of Mexican and American cultures. Some researchers argue that this alternative culture reflects differences in opportunities (economic, social networks) that can impact behaviors (Portes, 1995; Portes and Zhou, 1994). Some contend that certain familistic beliefs, such as strength of extended family, emerge from the experience of the Mexican American family rather than from the traditional Mexican family. Research has found that extended family networks are actually stronger among the second generation than the immigrant generation (Keefe & Padilla, 1987). Keefe and Padilla (1987:189) suggest that native-born Mexican Americans do acculturate but it is selective, and some ethnic traits especially maintenance of family ties are sustained and even strengthened from generation to generation. Though they are supporting a multidimensional model where assimilation occurs for some traits and not others, their description of a Mexican American community that is distinct from both the Anglo and recent immigrants leads to segmented assimilation theory. Segmented assimilation theory is an attempt to explain the more recent outcomes of second and third generation ethnic groups (in any number of outcomes, including marriage and divorce patterns as well as educational achievement) that are inconsistent with both the linear assimilation and familism theories. In certain cases, later generations (second and higher) become increasingly distant from the white majority. Over time and across generations certain groups tend to have increasingly higher levels of divorce than do whites, as well as lower overall levels of SES and completed education (Chapa, 1988). This theory suggests that two trajectories of assimilation occur depending on the ethnic group being examined, one toward the patterns of the white majority, the other toward patterns of groups marginalized by ethnic or racial prejudices (Portes, 1995; Portes and Zhou, 1994). Two explanations have been used to account 9
for the divergent assimilation patterns for the different ethnic groups. The first focuses on the socioeconomic context and social network structure available to immigrants and their descendants in the United States. The second, stemming from the first, focuses on the emergence of an ethnic identity once in the US; this ethnic identity emerges as a response against the dominant majority as a result of discrimination. An example of these divergent trends is apparent in the experiences of Cubans and Haitians in Florida (Portes and Zhou, 1994). Cubans are considered the most successful Hispanic ethnic group; they have the highest levels of educational attainment, lowest rates of poverty, and lowest rates of fertility. These rates are very similar to the white majority. Haitians on the other hand, who are also a newer immigrant group, though not of Hispanic origin, have fared quite poorly, assimilating towards the black minority. This difference is determined to a large extent by the context the immigrants encounter upon reaching the US. The Cubans were welcomed by the government, did not meet great prejudice, and soon formed supportive communities (Portes and Zhou, 1994:22). Their communities offered opportunities for social mobility, supported by US government programs. Haitians, on the other hand, have experienced official and social hostility, as well as having virtually no resources in place to aide in upward social mobility (Portes and Zhou, 1994). Portes and Zhou (1994) argue that trends towards greater disadvantage can also be seen among the Mexican American community they studied in a small town in California. In this case they suggest that the emergence of an ethnic identity among later generation Mexican Americans in response to their socioeconomic context essentially prohibited upward mobility. They looked at the educational performance of first, second, and third generation Mexican American students. They found that later generation children generally formed two groups who, 10
exemplify a lost race between first-generation achievements and later generation expectations. Seeing their parents and grandparents confined to menial jobs, and increasingly aware of discrimination by the white mainstream, the U.S. born children of earlier Mexican immigrants readily join a reactive subculture as a means of protecting their sense of self-worth. Participation in this subculture erects serious barriers to upward mobility because school achievement is defined as antithetical to ethnic solidarity (Portes and Zhou, 1994:27-28). So, while the immigrant and some second-generation children generally do quite well in school, later generation children who make up the above subculture generally do poorly. Theoretical expectations and findings The assimilation of many different ethnic groups, including people of Mexican origin, has been examined using a number of different measures. Much research has focused on rates of intermarriage since it is considered the best indicator of assimilation as it is more likely to occur once cultural and structural assimilation have been achieved (Anderson and Saenz, 1994; Cazares, Murguia and Frisbie, 1984; Murguia, 1982; Bean and Bradshaw, 1970). While there is some indication that intermarriage of white men with Mexican women did occur late in the 19th and early in the 20th century, as a means to obtain land grants in many border areas, intermarriage among Mexican Americans and whites has historically been low (Murguia, 1982). Other family characteristics also can and have served as indicators measuring the extent of assimilation. Because familism makes claims about the strength of family values or family oriented behaviors, some research looks at fertility rates, divorce rates, and prevalence of extended families. The three theories of assimilation have different expectations for trends and differentials in these behaviors among the Mexican origin population compared to the Non- Hispanic white population. Linear assimilation asserts that socioeconomic differences will increasingly explain 11
differences in family behaviors between the Mexican origin population and whites over time and across generations, while cultural explanations will weaken. Thus, later generations should more closely resemble the majority population once controlling for socioeconomic differences. As a result, later generations should have the lower levels of fertility, higher levels of marital disruption, and lower rates of marriage consistent with the weaker family orientation of the white majority. Also, more recent cohorts of Mexican Americans should more closely resemble whites than did earlier cohorts, as a result of increased interaction between the two cultures over time. As Del Castillo (1984:122) argues, if linear assimilation was occurring we would expect the third generation to be more affluent, better educated, and more acculturated than the Mexican immigrant and second-generation population. Familism theory offers a cultural explanation for differences between Mexicans and whites, so we would we would expect to see the stronger family oriented behaviors persist once we control for socioeconomic and demographic differences that may affect these outcomes independently. Familism predicts that this stronger family oriented behavior (as reflected in higher rates of fertility, lower rates of divorce, and higher prevalence of extended families) would persist over time and across generations. If with each successive generation, controlling for socioeconomic and demographic factors, we see no convergence towards the majority norm we would then argue for familism, if we see a convergence we would then support linear assimilation theory. While linear assimilation theory implies that acculturation comes prior to assimilation and familism that acculturation does not necessarily occur completely, segmented assimilation suggests that certain types of cultural assimilation might be more likely to follow socioeconomic assimilation. Thus, ethnic groups with greater socioeconomic advantage will assimilate towards 12
the socioeconomically advantaged white population, adopting those norms and values, while those at a disadvantage do not assimilate towards the majority and may even become increasingly disadvantaged. This is found both at the aggregate and individual level. For example, Cubans, who as a group have higher levels of education than all other Hispanic groups tend to have lower fertility rates overall (Ramirez, 2000; Ventura et al., 2000). On an individual level, Mexican origin women with high levels of education tend to have fewer children than those with lower levels of education (Frisbie, 1986). With segmented assimilation then, we might tend to see an initial thrust towards assimilation from the immigrant to the second generation as a result of expectations from the immigrant community for the second generation to do well. Yet, ethnic groups that have disadvantaged opportunity networks (i.e. less access to education and employment), as Mexicans do, will be discouraged. This theory suggests that family behaviors among later generation members will be shaped more by the disadvantaged context in the United States, and thus one should see a dramatic movement away from the majority norms towards norms of groups that have been marginalized. This would be reflected in higher levels of divorce and fertility among the third and later generations compared to the immigrant and second generation. Depending on which outcome is measured, support for these theories varies. Research looking at assimilation patterns through measurement of family oriented behaviors has focused primarily on rates of fertility, divorce, and the presence of extended families. Historically, Mexican American fertility has been higher than that of the Anglo population, largely due to the high fertility of the immigrant population. Though fertility rates in Mexico declined during the 1970 s, in the 1980 s the total fertility rate was still more than double that in the US (Stephen, 1989). In general, marital fertility has been declining in the US since the late 1980 s for all 13
populations, though non-marital fertility has been increasing (Ventura et al., 2000). Most of the support for linear assimilation has been found looking at rates of fertility. Stephen (1989) and Stephen & Bean (1992) both find a convergence of fertility rates between Mexican origin women and the white population when looked at over generations. Thus rates for second and later generation women begin to approach the rates for white women. Yet, not all research looking at fertility has found unequivocal support for linear assimilation. Bean, Swicegood, and Berg (2000) go beyond the basic native-born/foreign-born distinction and distinguish 2nd from 3rd generation members. They find that while 2nd generation fertility rates do approach those of whites, 3rd generation far exceed them, lending more support to a segmented assimilation type argument. This same pattern was found looking at non-marital fertility though the results were not reported in the article. Little research has looked at non-marital fertility, especially among the Mexican American population. It is known that rates are higher for Hispanics and blacks than they are for white women, though much of this difference is accounted for by differences in education, income, access to health care, sexual activity, and contraceptive use (Ventura, 1995:1). Birth rates for unmarried Hispanic women are more than twice as high as whites and about 10% higher than black women (Ventura, 1995). The higher overall rate among Hispanic women is due to high birthrates for unmarried women in all age groups, and not just among teenagers. Unfortunately this research does not present birth rates by specific Hispanic ethnicity. Most of the research looking at non-marital fertility specifically among Mexican origin and Mexican American women has focused on the experience of teens. Erickson (1998), focuses on the impact of acculturation on the patterning of teen pregnancy. Mexican origin teens have a lower age at first birth than do whites, though this is largely influenced by socioeconomic status. 14
And though, out of all Hispanic groups, Mexican Americans have the highest teen birth rate, they continue to have the lowest rate of non-marital births. This low rate of non-marital births reflects a greater prevalence of marriage among Latina teens. Among this group of teens, premarital sex is common (and often unprotected) and many of the teens expect that marriage will follow pregnancy, even though often it does not. Erickson (1998) suggests that perhaps there exists an alternative normative order for Latina teens, one that encourages a strong adherence to traditional female roles and one that allows a timing and sequencing of events that is distinct from both the middle class white culture and the middle and upper class Latin American culture. She proposes that, While the Latin American middle and upper class ideal is premarital chastity prior to marriage in the Catholic church, the reality among the working classes is often premarital pregnancy followed by informal union, which is sometimes legitimated, sooner or later, by formal marriage (Erickson, 1998:20). Here, the ideal of familism in the upper classes is not reflected in behavior of working classes. Erickson (1998) explores this alternative normative order as possibly being a rational choice made in the face of bleak life options. Yet, teenager s account for only 30% of the births to unmarried women and the non-marital birth rate among teens has actually been declining significantly in the 1990 s (Ventura et al., 2000). There is virtually no research on patterns of non-marital fertility among adult Mexican American women and how they differ from patterns for Non-Hispanic White or African American women. Another measure of family behavior researchers have investigated is marital disruption. Earlier studies of marital disruption by Frisbie and colleagues (1985; 1986) found some support for familism, reflected in lower rates of divorce for Mexican origin people. This research argued that cultural factors possibly explained some of the stability edge since this pattern held even when controlling for educational differences. Yet, this research did not distinguish between 15
the native-born and foreign-born population. Bean, Berg, and Van Hook (1996) find that lower disruption is actually a characteristic of immigrants rather than all Mexican Americans, and that stability is due to the structural and economic situations of the immigrants. So, socioeconomic factors rather than particular cultural strengths explained most of the differences in divorce rates. The strongest support for familism comes from the prevalence of extended families and from the higher rates of interaction that the Mexican origin population have with their primary and secondary family members compared to whites. Tienda and Angel (1982) finding a higher prevalence of extended family structure among Hispanic families compared to whites concluded that while much of the difference was due to economic hardship (as a means of coping), at least some was the result of cultural differences. In their work on the Mexican American family, Keefe and Padilla (1987), find not only that extended families are more common among Mexican Americans than whites, but also that Mexican Americans have higher levels of contact with primary and secondary family members. This increased contact is actually more common in later generations than in the immigrant generation. But, they also admit that non-acculturation is only characteristic of certain traits. They conclude that, Native born Mexican Americans acculturate but it is selective, and some ethnic traits especially maintenance of family ties, are sustained and even strengthened from generation to generation at other times they are Chicanos, practicing new and emergent cultural patterns and sustaining an ethnic community set apart from both Anglo and recent immigrant Mexicans (Keefe and Padilla, 1987:190). This statement closely echoes the description of segmented assimilation provided by Portes and Zhou (1994). While not specifically testing theories of assimilation, work done by Oropesa, Lichter, and Anderson (1994) and Oropesa (1996) looks at the effects of cultural beliefs and socioeconomic factors on attitudes towards marriage and actual marriage rates. They find that 16
controlling for socioeconomic factors, cultural measures play a role in predicting stronger adherence to pro-nuptial beliefs, yet when looking at actual behavior they find that cultural factors are insignificant in explaining the higher rates of marriage among Mexican origin people. ANALYTICAL APPROACH There is no consensus as to how and in what way exposure to the white majority has affected the Mexican American family, nor on what changes one might expect in the future. Given the higher prevalence of extended families among Mexican Americans, not all of which is explained by socioeconomic discrepancies, familism theory receives some empirical support. Yet, the biggest debate centers around whether linear assimilation theory or segmented assimilation theory provides the better framework for understanding the interaction between minority and majority populations and the effects on the family. I look at tends and differentials in female headship to test these theories. A woman can become a female head in two ways, either by being a married mother who divorces (or widows) or by being unmarried and having a child. Among women who were once married, rates of female headship will be affected by changes in the proportion of women who marry, who have children, and who become formerly married (either though divorce, separation or widowhood). Rates of female headship among never married women will be influenced by changes in the proportion that never marry and in the proportion who have an extra marital birth. Thus, rates of female headship are affected by changes in rates of divorce, marital and non-marital fertility, and marriage, rates that sometimes change in different directions. Female headship is a composite of family behavior that incorporates all aspects of family change. The rate of female headship has increased rather dramatically, especially in the latter part 17
of the twentieth century, and is of concern due to its association with economic disadvantage. The rate of single parenthood for all women remained relatively constant from 8.5 % at the turn of the century to a rate of 9.1 % in 1960, after which the rate began to increase substantially (Gordon and McLanahan, 1991). While the main component of this rate at the turn of the century was widowhood, by 1960 it was divorce. As of 1990, the overall female headship rate (measured as the percentage of households with children under the age of 18, headed by a single female) was 18 %. This varies dramatically by race/ethnicity with whites having a rate of 14 %, blacks 48 %, and Hispanics 22 % (Lichter et al., 1997). While the factor contributing most to the rate of female headship among white women is divorce, the growth in out of wedlock births has been the most important factor since 1970 for black women (Wojtkiewicz et al., 1990). Little is known about female headship among Mexican Americans and how it has changed over time. To understand the assimilation process, this research looks at trends in female headship among Mexican origin women and compares them to trends for white women white women. The analyses also compare Mexican born and U.S. born Mexican origin women. To more fully examine the process of assimilation, this research looks at trends across generations. Previous research on European immigrants shows that assimilation did not occur by the second or third generation, Italians and other ethnic minorities experienced high levels of discrimination over many years and many generations (Alba and Nee, 1997). Additionally, because assimilation is more likely to occur the longer the ethnic group has been in the United States, it is necessary to look at change over a substantial period of time to truly understand the experiences of Mexican Americans in the United States. 18
Data The analyses in this paper are based on data from two sources. The first part of the analysis uses IPUMS census data samples prepared and distributed by the University of Minnesota (Ruggles and Sobek, 1997). These data files contain samples for all the decennial census years from 1850 through 1990, with the exception of 1890 (where the censuses have been lost), and 1930 (where confidentiality rules have not yet allowed the sample to be constructed). Because of the small number of Mexicans prior to 1880, the analysis reported here begins with the 1880 sample. Because rates of female headship do not really begin to change until 1960, the decomposition and multivariate analysis focus on 1960 through 1990. A great strength of these data, especially post 1950, is the availability of a substantial number of relatively uniform variables for all years. Another strength is the large sample provided by Census data. Though the IPUMS do not specifically identify people of Hispanic origin until 1980 the work done by Gratton and Gutmann (2000) allow for the identification of Hispanics, by specific ethnicity, from 1880 through 1970. The identification of Hispanics is done by linking parent/child records and is based on place of birth, parents place of birth, grandparents place of birth (when possible), and possession of a Spanish surname. In some years these identification factors are supplemented with information on mother tongue, parent s mother tongue, language spoken, and a Spanish race write-in on the census. Each person identified as Hispanic by this method is then assigned an ethnicity by familial place of birth or geography. This data set gives researchers the opportunity to identify Mexicans separate from Hispanics, using nationally representative census data going back for over 100 years. Using these data I assembled a series of smaller data files (one for each census year) that consisted of all women aged 18-59 who were non-hispanic white, non-hispanic black, and Mexican origin. Because it is impossible to 19
separate 2nd generation from later generation members in 1980 and 1990, the analysis using the IPUMS only distinguishes between U.S. and foreign-born Mexican Americans, determined by responses to questions about place of birth. Using the IPUMS I look at long-term trends in female headship among non-hispanic white, non-hispanic black, U.S. born Mexican, and Mexican born women between 1880 and 1990. I specifically look at changes over time and changes that occur between the immigrant and U.S. born Mexicans and how they compare to changes for white women. To measure this, a dichotomous variable is created for each woman in each of the samples, female head of household or not. To qualify as a female head of household, one must be a woman between the ages of 18-59, who has a child or children under the age of 18 living at home, who is not currently married, and who is heading her own household. I do not include women older than 59 because beyond that age very few women have children under 18 living in the household. Women who are never married, married spouse absent, divorced, widowed, or separated are considered formerly married. To determine whether differences between groups can be attributed to cultural differences, I examine trends and differentials in female headship controlling for SES and other variables that may be independently associated with the dependent variable. To examine the association between ethnicity and nativity net of the controls, I estimate logistic regression models using the PROC LOGISTIC function in SAS. Since it has been generally accepted as the best measure of socioeconomic status, level of education is used to control for differences in socioeconomic status (Stephen and Bean, 1992; Bean et al., 1996; Frisbie, 1986). In the regression equation I include five dummy variables corresponding to specific levels of education achieved, less than high school, some high school (reference), high school graduation, some 20
college, and four or more years of college. Dummy variables are also included in the model to statistically control for age, region, and metropolitan status (when available). This paper also conducts analysis using the June 1995 Current Population Survey. This data source includes a nationally representative sample of all civilian non-institutionalized persons living in households and includes a supplement on the marital and fertility histories for all women aged 15-65 (U.S. Dept. of Commerce, 1998). This survey includes questions on mother and father s place of birth as well as the individual s place of birth, thus allowing one to distinguish 3rd generation from first and 2nd generation members. The analysis with the CPS is restricted to non-hispanic white, non-hispanic black, and Mexican origin women (women who identified themselves as Mexican American, Chicano, or Mexican in response to the Hispanic ethnicity question) aged 18-59, resulting in a working sample of 37,304 women. The measure for female headship is created in virtually the same way in the CPS as in the IPUMS, though since there is no specific question on number of children under the age of 18, I use an alternative approach to identify the presence of one s own children under the age of 18 in the household. The marital and fertility supplement includes questions on the month and year of birth for each child up to five children (first four and the last), as well as information on where they currently reside. This information was used to identify children who were under 18 at the time of the interview who resided in their parent s household. The strongest advantage of the CPS is that it allows one to conduct analysis distinguishing the second generation from third and higher generations. These data are also used to extend the IPUMS analysis to 1995. 21
RESULTS Trends in female headship 1880-1990 Figure 1 presents levels of female headship between 1880 and 1990. For all populations the level of female headship stayed relatively constant until 1960. In general, the level for white women was lowest within each year while the level for black women was generally the highest. The level of female headship for foreign-born Mexican women was higher than the level for U.S. born Mexican women until 1970 when this pattern reversed. We also see that until 1950, levels for foreign-born Mexican women and black women were more similar than different. [FIGURE 1 ABOUT HERE] The level of female headship began to increase for all women starting in 1960. From 1960 to 1990 there is an increase from 3.1 to 5.9 among the white population, of 6.7 to 8.1 among the Mexican born population, and of 5.8 to 11.1 among the native born Mexican population. Aside from the black population, the most rapid increase is occurring among the native born Mexican population while the slowest increase is for the Mexican born, whites fall in the middle. Over this period of time there is a decrease in the difference between the white and the Mexican born population. While Mexican born women begin with higher levels than the white population, as the rate among whites begins to increase more rapidly, the gap between these groups begins to close with levels among white women beginning to approach that of the immigrant population. For U.S. born Mexicans on the other hand, the rate of increase in female headship is greater than among the white population. So, while levels were already higher for them in 1960 than for the white population, by 1990 this difference has increased with the level for U.S. born Mexicans more than double the level for white women. Much of this difference may be related to differences in socioeconomic status. 22
Table 1 presents the results from the regression analysis, which controls for SES and other factors known to be associated with female headship. Figure 1 showed that both Mexican born and U.S. born Mexican women had consistently higher levels of female headship than white women in every year from 1960 1995. Once controls are added, the difference in levels of female headship relative to white women drops significantly for both groups. The level of headship for foreign-born Mexican women is virtually the same as for white women from 1970 onwards, occasionally falling slightly below the level for white women. While the difference in level of female headship between U.S. born Mexican and white women narrows within each year once controls for SES are added, the overall difference between them in each year remains relatively constant over time. U.S. born Mexican women have headship levels that are between 1.4 and 1.7 times as high as levels for white women. Controlling for education allows one to begin to understand the association between SES and female headship, and the findings are not surprising. Women with less than a complete high school education (less than 9 years and some high school) are significantly more likely to be a female head compared to women with a high school degree, while those with higher levels of education are less likely to. The advantage of having a college degree remains consistent, and by 1995 women with a college degree are almost 50% less likely to be a female head than a women who finished high school. Though differences in educational distributions explain much of the difference between Mexican origin women and white women, female headship remains higher among U.S. born Mexican women than among white women. [TABLE 1 ABOUT HERE] One finding of note is that levels of female headship are higher among both U.S born Mexican and Mexican born women compared to white women in 1960 and 1970, even after 23
controlling for SES. We do not see substantially lower levels of female headship for Mexican born women compared to white women or among U.S. born Mexican women. These findings call into question the popular assumption of familism, which suggests a stronger adherence to traditional family structure. It is clear though, that consistent with the findings in Figure 1, the rate of growth in levels of female headship among Mexican born women is much slower than that of either U.S. born Mexican or white women. The difference in headship rates between the U.S. born Mexican and white women stays relatively constant over time. While these results do not offer much support for familism, neither is there a clear pattern of convergence between the native born Mexican origin women and white women as would be suggested by linear assimilation theory. These findings raise the question, what behaviors contribute to these ethnic differences in female headship? The next part of the analysis looks specifically at this question. Decomposition Analysis of female headship Decomposition allows researchers to express the difference between two rates in terms of differences in a specific set of factors that go into the composition of that rate. Holding other factors constant, the relative importance of one factor in relation to the others in explaining the difference in overall rates between two populations can be determined (Das Gupta, 1993). For this analysis, I look at the level of female headship from 1960 through 1990. Comparisons are made between the white population and Mexican born population, between the white population and the U.S. born Mexican population, and between the Mexican born and U.S. born Mexican women within each year. The proportion of women who head families (with no spouse present) can be defined as the sum of the proportion of women who head households who are formerly married with 24
children and the proportion of women who head households who have never been married and have children (Wojtkiewicz et al., 1990). The decomposition analysis focuses on the components that go into being a formerly married head of household and a never married head of household. Four factors contribute to female headship among ever married women, the proportion of all women who have ever been married (factor A), the proportion of ever married women who have children under the age of 18 living with them (factor B), the proportion of mothers who are formerly married (factor C), and the proportion of formerly married mothers who head their own household rather than live with a family member or in some other situation (factor D). Three factors contribute to female headship among never married women, the proportion of all women who have never married (which is the same as 1-factor A), the proportion of the never married who have children under the age of 18 in the household (factor E), and the proportion of never married mothers who head their own household (factor F). Each of these factors tells us something about a particular aspect of family behavior. The proportion married (or never married) tell us about marriage rates. Among those ever married, the proportion with children tells us about levels of marital fertility. The proportion formerly married informs us primarily about levels of divorce, though it is also affected by separation and widowhood, therefore inn the rest of the paper I refer to factor C as divorce. The proportion heading their own household is an indicator of the propensity to live in an extended family. Among those never married, we get rough estimates rates of levels of non-marital fertility and the propensities to head households among that population. Table 2 displays the results of the decomposition analysis. Presented are the levels of female headship for non-hispanic white, U.S. born Mexican, and Mexican born women and the differences between each of the pairings possible between these groups. These are the same as 25