International Migration and Gender in Latin America: A Comparative Analysis

Abstract

We review census data to assess the standing of five Latin American nations on a gender continuum ranging from patriarchal to matrifocal. We show that Mexico and Costa Rica lie close to one another with a highly patriarchal system of gender relations whereas Nicaragua and the Dominican Republic are similar in having a matrifocal system. Puerto Rico occupies a middle position, blending characteristics of both systems. These differences yield different patterns of female relative to male migration. Female householders in the two patriarchal settings displayed low rates of out-migration compared with males, whereas in the two matrifocal countries the ratio of female to male migration was much higher, in some case exceeding their male counterparts. Multivariate analyses showed that in patriarchal societies, a formal or informal union with a male dramatically lowers the odds of female out-migration, whereas in matrifocal societies marriage and cohabitation have no real effect. The most important determinants of female migration from patriarchal settings are the migrant status of the husband or partner, having relatives in the United States, and the possession of legal documents. In matrifocal settings, however, female migration is less related to the possession of documents, partner’s migrant status, or having relatives in the United States and more strongly related to the woman’s own migratory experience. Whereas the process of cumulative causation appears to be driven largely by men in patriarchal societies, it is women who dominate the process in matrifocal settings.

INTRODUCTION

Despite the fact that women comprise a large and growing fraction of immigrants worldwide, and are actually a numerical majority in many cross-border flows, research on international migration has focused disproportionately on males. In their comprehensive review of migration theory, for example, Massey et al. (1998) had little to say about the influence of gender on patterns and processes of migration. Likewise, the collection of empirical studies later assembled by Massey and Taylor (2004) failed to include a single paper on female migration. The short shrift given to gender in migration research was noted early on by feminist scholars (Simon and Brettell, 1986; Pessar, 1986; Tyree and Donato, 1986; Boyd, 1989; Pedraza, 1991; Tienda and Booth, 1991; Hondagneu-Sotelo and Cranford, 1999) and in response a growing number of studies have focused on female migration, allowing Pessar (1999) to conclude that “scholars have made great advances in moving beyond an earlier male bias in theory and research… [and that] …we are now moving toward a more fully engendered understanding of the migration process”.

Although the number of studies considering the migration of women may have multiplied, less attention has been focused on gender as a social construct. Most studies have simply specified and estimated comparable models of male and female migration and then compared results, an approach that Hondagneu-Sotelo (2003) calls “add women and stir”. Sometimes the only focus of the analysis is whether women are truly “independent” economic actors or simply moving for “family reasons” (Boyd, 1975, 1976, 1986; Findley and Williams, 1991; Tyree and Donato, 1985; Hugo, 1993), what Cerrutti and Massey (2001) have called the “auspices” of female migration. Focusing on such a narrow question is unlikely to shed much light on how the dynamics of gender play out in determining patterns and processes of male and female migration.

Another problem is that most studies to date have considered the migratory behaviour of males and females originating in a single culture. Much of the quantitative work, especially, has focused on Mexico, owing to the ready accessibility of data from the Mexican Migration Project. In her study, for example, Kanaiaupuni (2000) found that certain determinants of migration operated differently for Mexican men and women. Whereas higher education decreased the odds of male migration, it increased the odds of female migration; and whereas higher rates of female employment at the community level raised the probability of male migration it lowered the likelihood of out-migration by women. Likewise, Curran and Rivero-Fuentes (2003) found that male network ties were more important in raising the odds of emigration by men than by women, and that female network connections actually served to decrease the odds of male migration while very strongly raising the likelihood of female migration. Cerrutti and Massey (2001), meanwhile, showed that Mexican women generally followed male family members (either a husband or father) and that only a tiny minority initiated migration independently.

Although these studies help to clarify differences in the determinants and nature of migration for males and females in Mexico, they do not shed a much light on how gender itself influences international migration, for within any single country the prevailing gender system is relatively constant. Across countries, however, gender systems may be more or less patriarchal, and if Hondagneu-Sotelo’s (1992) leading hypothesis is true – that women use international migration as a means to overcome the restraints of patriarchal suppression within the family – then the process of female out-migration is apt to look very different across national settings.

Considerable understanding of the dynamic effects of gender may, of course, be gleaned by comparatively reviewing the results of qualitative studies done in different regions (as did Pessar, 1999); but quantitative research has been limited by the absence of a dataset capable of sustaining comparative statistical analyses across countries. Here, we make use of new data from the Latin American Migration Project and combine it with comparable information from the Mexican Migration Project to study patterns of male and female migration originating in five national settings whose gender systems differ in identifiable ways. This cross-national variation provides a basis for studying how gender affects the migratory behaviour of men and women rather than simply contrasting determinants across the sexes. We thus seek to redress Hondagneu-Sotelo’s (1994) criticism that “gender is typically considered in migration theory only when women are the focus” and agree with her that “gender is an analytical tool that is equally relevant to our understanding of men’s migration as it is to our understanding of women’s migration” (1994: 2–3).

SOURCES OF DATA

Since 1982, the Mexican Migration Project (MMP) has compiled detailed data on documented and undocumented migration from Mexico to the United States using a blend of ethnographic and survey methods to study specific origin communities and their US destinations (see Massey, 2004). To date the MMP has undertaken representative surveys of 16,840 households located in 93 binational communities and the validity and reliability of these data have been well documented (Zenteno and Massey, 1999; Massey and Zenteno, 2000; Massey and Capoferro, 2004). All MMP data are publicly available via the internet and have formed the basis for numerous empirical studies (for a recent selection see Durand and Massey, 2004).

The Latin American Migration Project (LAMP) began in 1998 as a self-conscious attempt to replicate the design features of the Mexican Migration Project. To date, surveys have been carried out and made public on documented and undocumented migrants from Puerto Rico, the Dominican Republic, Costa Rica, Nicaragua, Peru, Haiti, and Paraguay. Additional surveys are planned or in process for Guatemala and Ecuador. Preliminary analyses of data from the LAMP suggest they are valid and accurate and that they yield a valid picture of patterns and process of international migration from the two Caribbean and two Central American nations (see Massey and Sana, 2004).

Table 1 assembles sampling information for the data used in the present analysis, which come from surveys undertaken in Mexico, Puerto Rico, the Dominican Republic, Costa Rica, and Nicaragua. Although Puerto Rico is not an independent nation, of course, and its inhabitants are US citizens by birth, it was included in the LAMP to represent the case of “international” migration in the absence of legal barriers to movement.

TABLE 1.

Sampling Information for Surveys Conducted in Selected Countries by the Latin American and Mexican Migration Projects

Sampling information	Mexico	Costa Rica	Puerto Rico	Nicaragua	Dominican Republic
Community samples
Number of      communities	93	7	5	9	7
Number of      households	15,171	1,391	585	1,598	904
Sampling fraction	31.3	22.0	17.3	19.0	13.1
Refusal rate	7.0	3.6	2.9	4.3	4.3
US samples
Number of      households	837	37	61	65	74
Number of      persons	3,522	168	319	303	370
Total sample
Number of      households	16,008	1,428	646	1,789	978
Number of      persons	80,621	7,414	2,878	11,168	5,913
Average survey      year	1995	2002	1998	2002	1999

As the table shows, a total of 15,171 Mexican households were surveyed across the 93 communities with an average sampling fraction of 31 per cent and an average refusal rate of 7 per cent. Naturally, none of the LAMP surveys can match the sample size of the MMP, which has been in the field for more than 20 years. The seven communities in Costa Rica were surveyed with an average sampling fraction of 22 per cent to yield 1,391 households and a refusal rate of around 4 per cent. Likewise, the five communities in Puerto Rico were surveyed at a rate of 17 per cent to yield 585 households and the same refusal rate. The 1,598 households from seven Nicaraguan field sties were obtained using a sampling fraction of 19 per cent and once again the refusal rate was around 4 per cent. Finally, in the Dominican Republic, the same percentage of households refused to participate in the survey but 904 households completed it across seven communities to produce a sampling fraction of 13 per cent.

The middle panel of the table shows the number of households and people surveyed by the MMP and LAMP in US destination communities. Because these samples are non-random, rates of refusal and sampling fractions were not computed. The number of people captured by the out-migrant surveys ranged from 168 for Costa Rica to 3,522 for Mexico. The bottom panel of the table shows the total sample compiled for each country. The Mexican sample is largest at 80,621 people and 16,008 households, followed by Nicaragua with 11,168 people and 1,789 households, Costa Rica with 7,414 people and 1,428 households, and the Dominican Republic with 5,913 and 978 households. The smallest sample was compiled for Puerto Rico with 646 households and 2,878 people.

Data were gathered using a semi-structured instrument known as the ethnosurvey, which in organization is midway between the highly structured instrument of the sample survey and the guided conversation of the ethnographer (see Massey et al., 1987). The ethnosurvey balances the goal of unobtrusive measurement with the need for standardization and quantification and yields an interview that does not use a standard question-answer format. It allows interviewers flexibility to collect the data in whatever way they believe works best, especially for sensitive information on foreign wages and documentation. But everyone collects the same information. Thus, a non-standard interview produces a standard set of data.

The interview schedule is arranged in a series of tables, with columns for different variables and rows referring variously to people, events, years, or other conceptual categories. While holding a natural conversation with the subject, the interviewer fills in the table by soliciting information in ways that the situation seems to demand, using his or her judgment as to the timing and wording of questions and probes. Each table is organized around a specific topic, giving coherence to the conversation. Specialized follow-up interviews are included from time to time to elaborate particular themes of interest.

Whereas the MMP employed the same ethnosurvey instrument at all field sites, total consistency was not possible in the LAMP. Geographic conditions; patterns of social and economic organization; and variables of interest, such as documentation, border crossing, and land tenure, differ from country to country. As a result, there is no a single “LAMP Questionnaire” in the same way that there is a uniform MMP questionnaire. Rather, LAMP investigators developed a set of core tabular forms to create a “Template Questionnaire”. This questionnaire was then adapted to each local situation to yield a standard body of data on international migration (questionnaires and documentation are available from the project website at http://lamp.opr.princeton.edu/).

The LAMP Template Questionnaire contains 16 tabular forms, lettered A through P, each covering a distinct topic. In this analysis, we rely mainly on data compiled using Forms A and D. Form A instructs interviewers to gather basic social and demographic information about the head of household; the spouse; all children, irrespective of whether they currently live in the household or have left; and other individuals living in the household. Variables include sex, relation to head, household membership, year of birth, place of birth, marital status, education, and occupation. Form D applies to each person listed in Form A who has ever been to the United States. It records, for the first and for the most recent US trips, the year of departure from country of origin, duration of stay, destination, occupation, and wage; it also ascertained the total number of US trips ever taken, and the migrant’s marital and legal status at the time of each trip.

Interviewing in Mexico typically occurred in the winter months because that country’s migration has historically been seasonal, and that is the time of year when circular or seasonal migrants are most likely to come home. This pattern contrasts with that of other countries. For example, virtually no Puerto Rican, Dominican, Nicaraguan, or Costa Rican migrants work in agriculture, the most seasonal of all industries. In the LAMP, therefore, no special efforts were made to concentrate interviewing at a particular time of year. Four of the five Puerto Rican community surveys were administered during the summer, and one during the autumn. Five of the Dominican communities were surveyed in the summer, one in the spring, and one in the winter. Two of the nine Nicaraguan community surveys took place in the spring, three in the summer, and four during the winter; and in Costa Rica, one survey was fielded in the spring, three in the summer, and three others in the winter. Further details about the LAMP surveys are available from Massey and Sana (2004).

GENDER AND FAMILY IN FIVE SETTINGS

Across the five sets of samples it is not difficult to locate the two extremes of the distribution from least to most patriarchal. Despite recent changes (see Salles and Tuirán, 1998), the Mexican family system remains remarkably patriarchal in structure and organization, with formal authority invested in a male household head who exercises power over wives and daughters (Oliveira, 1998). Patriarchal relations are especially prevalent in provincial communities and rural villages, where most Mexican migrants originate (Durand et al., 2001). In Mexico, unwed childbearing, informal unions, divorce, and separation are still quite rare (Sana, 2004); and single person households are unusual first because most young people do not move out of parental households until they are married, and second because widowed parents generally return to reside with an adult child. The Mexican family pattern has been labelled “traditional and patriarchal” by Stromquist (1998).

In the Dominican Republic, in contrast, informal unions are prevalent, unwed childbearing is common, and union disruption is frequent, yielding a wide variety of household types (Safa, 1995). Households are differentiated not just by life cycle stage, but also by the presence or absence of an adult male, the nature of a woman’s relationship to that male (formal or informal), and the order of their union (first, second, third, etc.). Although children almost always live with their mother, there is a great deal of variation in whether they live with their fathers. The Caribbean family system is often called matrifocal because mother and children comprise the basic family unit, into and out of which adult males come and go (Smith, 1956; Clarke, 1957; Barrow, 1996).

In Table 2 we cull census data from the various countries to confirm that Mexico and the Dominican Republic anchor the ends of a continuum of family and gender relations. In Mexico a relatively large share of those aged 15 and older are married, few are in consensual unions, and there is little divorce or separation. As the last column in the table shows, among Mexicans aged 15+ in 1990 there were 3.9 married persons for every person living in a consensual union, divorced, or separated, yielding the highest ratio in the table. Although rates of marital disruption and cohabitation had both increased somewhat by 2000, the ratio of 2.9 was still high compared to other countries.

TABLE 2.

Population Aged 15+ by Marital Status, Country, and Year: Census Data From Selected Latin American Settings

Country and year	Married	Consensual union	Divorced or separated	Ratio of married to consensual and disrupted
Mexico
1990	41.0	8.7	1.9	3.9
2000	46.1	12.4	3.4	2.9
Costa Rica*
1984	39.0	8.2	2.9	3.5
2000	38.2	13.6	5.4	2.0
Puerto Rico
1990	48.5	5.5	10.2	3.1
2000	52.0	2.3	13.9	3.2
Nicaragua
1971	35.2	21.1	2.0	1.5
1995	28.6	29.0	8.3	0.8
Dominican Republic
1981	20.2	28.8	5.7	0.6
1993	20.6	30.7	5.8	0.6

Note:

^*Population aged 12+.

Within the Dominican Republic, in contrast, consensual unions outnumber legal marriages and divorce and separation are far more prevalent than in Mexico, yielding a ratio of married to cohabiting, divorced or separated persons of just 0.6 in both census years (1981 and 1993), the lowest ratios in the table. We performed the same basic calculations in Costa Rica, Puerto Rico, and Nicaragua and the results are arrayed in separate panels between Mexico and the Dominican Republic in what we judge to be descending order of patriarchy. Costa Rica, like Mexico, has a relatively patriarchal system of family and gender relations characterized by relatively high rates of marriage, though with somewhat higher rates of cohabitation and marital disruption than in Mexico, yielding a marriage ratio of 3.5 in 1984 and 2.0 in the year 2000. In Puerto Rico, legal marriage is generally the norm and consensual unions have become less common over time, though rates of marital disruption are quite high so that marriage continues to be relatively unstable as a social institution. Owing to a change in the way the US census classified households in 2000, the relative frequency of consensual unions is probably underestimated, thus helping to account for Puerto Rico’s slight increase in the marriage ratio from 3.1 to 3.2 between 1990 and 2000. We consider Puerto Rico to be roughly midpoint on the continuum between Mexico and the Dominican Republic.

Although gender relations in Puerto Rico may mix elements of both the Mexican and Dominican systems to create a somewhat confusing picture, Nicaragua very clearly lies closer to the Dominican end of the spectrum, reflecting a history of rural proletarianization and landlessness that denied males access to real property on which a patriarchal system of inheritance and wealth could be constructed (cf Baud, 1995; Hall, 2000; and Bugajski, 1990). The relative standing and independence of women may also have been enhanced by social reforms undertaken during the Sandinista period (Gilbert, 1988). As can be seen from the table, the prevalence of consensual unions in Nicaragua is second only to the Dominican Republic, and divorce and separation are even more prevalent, yielding a ratio of married persons to those divorced, separated, or cohabiting of 0.8 according to the latest data (compared with 0.6 among adult Dominicans).

Census data has its limitations, of course, and rigid census definitions are often at odds with fluid and changeable social constructs such as “family” and “household”. In order to further explore the continuum of gender relations just described, and to reveal how these relations are expressed among respondents to the MMP and LAMP, Table 3 shows the current marital status of male and female householders in the five national contexts under investigation. For our purposes, a householder is an adult man or women living independently with or in partnership with a member of the opposite sex along with other family members, usually children.

TABLE 3.

Distribution of Male and Female Householders by Current Marital Status

	Legally married	Consensual union	Divorced or separated	Ratio of married to consensual and disrupted
Male householders
Mexico	88.9	5.8	1.0	13.1
Costa Rica	79.7	9.2	3.6	6.2
Puerto Rico	57.0	13.5	15.2	2.0
Nicaragua	61.7	26.9	4.7	2.0
Dominican Republic	55.8	31.6	7.0	1.4
Female householders
Mexico	81.1	5.4	3.6	9.0
Costa Rica	69.6	7.9	11.6	3.6
Puerto Rico	44.0	11.5	19.7	1.4
Nicaragua	48.5	21.4	17.0	1.3
Dominican Republic	45.1	26.3	14.5	1.1

The distinctive and highly stable nature of the Mexican family system is at once apparent. At the time of the survey, 89 per cent of male householders were legally married, 6 per cent were in a consensual union, and just 1 per cent were separated or divorced (another 2% were widowed and 2% were never married but these data are not shown). Female householders display a similar profile, with 81 per cent married, 5 per cent in consensual unions, and 4 per cent separated or divorced (the incidence of widowhood is higher than men’s at 8%). Within our Mexican samples, in other words, roughly 90 per cent of both male and female householders were either currently married or only left marriage because of the death of a spouse. Marriage among respondents to the MMP is clearly a very stable and enduring social institution. The ratio of married households to those that are separated, divorced, or cohabiting is 13 to 1 among males and 9 to 1 among females.

Next on the continuum is Costa Rica. Legal marriage is still very common at 80 per cent among men and 70 per cent of women, but cohabitation and marital disruption are more common than in Mexico. Around 9 per cent of male householders and 8 per cent of females are living in a consensual union and whereas the frequency of separation and divorce is only 4 per cent among male Costa Ricans, it reaches nearly 12 per cent among females. The ratio of married to disrupted or cohabiting householders is thus 6.2 for male Costa Ricans and 3.6 for females, considerably lower than the respective figures for Mexicans, but still relatively high compared to the other cases.

As with the census data, Puerto Ricans appear to occupy the middle point on the patriarchy continuum constructed with LAMP data. Only a minority of female householders (44%) were married at the time of the survey, compared with 11 per cent cohabiting and 20 per cent separated or divorced, yielding a marriage ratio of just 1.4. The pattern is similar for male householders from Puerto Rico, except that marriage is slightly more prevalent compared with cohabitation and marital disruption, yielding a ratio of 2.0.

As before, Nicaragua and the Dominican Republic lie close to one another and at the opposite end of the continuum from Mexico. Among female householders, only 48 per cent of Nicaraguans and 45 per cent of Dominicans are legally married, but 21 per cent of the former and 26 per cent of the latter were in consensual unions along with respective figures of 17 per cent and 15 per cent being separated or divorced, yielding marital ratios of only 1.3 and 1.1. Thus for every married female household in these countries an almost equal number are cohabiting, divorced, or separated. For male householders the picture is similar, though legal marriages outnumber other states by 2.0 to 1 in Nicaragua and 1.4 to 1 in the Dominican Republic.

MARRIAGE, GENDER, AND MIGRATION

In sum, whereas most women in Mexico and Costa Rica live in a husband-wife household, most women in the Dominican Republic, Nicaragua, and Puerto Rico do not. The pattern of female family structure is very similar for Nicara-guans and Dominicans, and the main thing separating Puerto Rican women from the other two origins is the greater prevalence of marital disruption compared to cohabitation. Thus, whereas just 13 per cent of households were headed by women in Mexico and only 22 per cent in Costa Rica, the corresponding figure for Nicaragua and the Dominican Republic was around 30 per cent and in Puerto Rico it reached 42 per cent. These patterns are correspondingly associated with relatively low rates of female labour force participation in Mexico (29% according to MMP data) and Costa Rica (37% according to the LAMP) compared with Puerto Rico (46%), Nicaragua (48%), and the Dominican Republic (42%).

As suggested earlier, in terms of female autonomy and independence, Mexico and Costa Rica lie at one extreme and Nicaragua and the Dominican Republic lie at the other, with Puerto Rico having a somewhat ambiguous status in-between. Table 4 begins to consider how these differences in gender relations are reflected in rates and patterns of international migration by men and women. It shows rates of lifetime migration to the United States (the percentage of people who had ever been to the United States by the time of the survey) for male and female householders by place of origin and marital status. Overall rates of lifetime migration from each place are shown at the bottom of each panel. As can be seen, overall, male migration is greatest in Mexico and Puerto Rico (at 40%–41%), lowest in Nicaragua (at 10%), and in-between in Costa Rica (at 15%) and the Dominican Republic (at 18%).

TABLE 4.

Percent of Male and Female Householders who have Ever been to the United States by Country and Current Marital Status

	Mexico	Costa Rica	Puerto Rico	Nicaragua	Dominican Republic
Male householders
Never married	39.4	6.7	25.0	10.9	7.1
Legally married	41.7	16.2	41.4	12.4	22.4
Consensual union	29.4	10.5	32.7	3.2	11.9
Separated or divorced	51.5	19.5	45.8	13.3	18.0
Total	40.9	15.3	41.0	9.7	17.8
Female householders
Never married	12.4	5.5	36.7	3.1	10.2
Legally married	11.8	5.6	30.7	9.1	14.7
Consensual union	12.6	5.9	31.7	1.7	4.7
Separated or divorced	16.7	9.3	47.6	7.6	22.5
Total	12.1	5.9	35.3	6.8	12.5
Ratio of female to male migrants within category
Never married	0.31	0.82	1.47	0.28	1.44
Legally married	0.28	0.35	0.74	0.73	0.66
Consensual union	0.43	0.56	0.97	0.53	0.39
Separated or divorced	0.32	0.48	1.04	0.57	1.25
Total	0.30	0.39	0.86	0.70	0.70

In Mexico, the prevalence of lifetime migration was greatest among men who were separated or divorced (51%) and lowest among those in consensual unions (29%), with the legally married and never married lying close to the average. Likewise, among Puerto Rican male householders the highest migration prevalence occurred among the separated and divorced (46%) and the lowest among the never married (25%). Puerto Rican males in consensual unions also displayed a relatively low rate of lifetime migration (33%). Among Dominican males, however, migration was greatest among the legally married (22%) and those separated or divorced (18%), whereas among Nicaraguan males the top two categories were separated or divorced (13%) and legally married (12%). All categories of Costa Rican male householders displayed rates of US migration that were slightly above or below the overall rate, yielding a narrow range from 10 per cent to 20 per cent, except for the never married, who evinced a rate of just 7 per cent.

From these data, it appears difficult to generalize about the effect of gender on patterns of male migration. Migratory prevalence is high in both Mexico, which has the most patriarchal system, and Puerto Rico, which is in the middle of the continuum, and there is no clear association between a country’s gender system and the prevalence of migration in any particular marital status group. The relation between rates and patterns of US migration and gender are clearer in the case of female householders, however.

The middle panel of the table shows the absolute percentage of female householders who had ever been to the United States by the time of the survey, and the bottom shows ratios compared with the corresponding figures for male householders, which we take as the best overall indicator of females’ relative propensity to migrate. Female householders in Mexico and Costa Rica – the two most patriarchal and least matrifocal of the countries surveyed – are notable for their relatively low rates of lifetime migration compared with males, with ratios of just .30 and .39. Among Mexican women, only those in a consensual union displayed a notably higher migration ratio of 0.43, whereas among Costa Ricans it was widowed women who stood out with a ratio of 0.46 (the computations for widows are not shown). Thus, in these two relatively patriarchal settings, the two categories that display higher-than-normal levels of lifetime migration involve women who are unmarried.

In the remaining countries, which tend to be less patriarchal, relative rates of female migration are much higher. The ratio of female to male migration is 0.86 overall in Puerto Rico and 0.70 in both the Dominican Republic and in Nicaragua – much higher than the 0.30–0.39 observed in Mexico and Costa Rica. Indeed, within several marital categories, female householders are actually more likely than their male counterparts to migrate to the United States. Among Puerto Rican females, for example, the relative level of lifetime migration is 1.47 among the never married and 1.04 among those who are separated or divorced. The lowest ratios are observed among those with marital experience – 0.74 for those currently in legal marriages and 0.64 for widows (not shown). Similar patterns prevail among Dominican female householders (see Safa, 1995). The migration ratio is 1.44 among the never married and 1.25 among those separated or divorced, compared to just 0.66 among legally married women and 0.39 among those in consensual unions. In Nicaragua, none of the female categories exceed 1.0, perhaps reflecting the recent and rather political origins of its migration (see Lundquist and Massey, 2004).

GENDER AND THE PROCESS OF OUT-MIGRATION

In general, the foregoing data suggest that the relative propensity for females to migrate internationally is higher in societies that are matrifocal and lower in those that are patriarchal. Among matrifocal societies the propensity for unattached women to migrate is markedly greater than in patriarchal societies, at times even greater than males in the same marital category. Apart from gender and marital status, however, many other factors – such as human capital, physical capital, and social capital – affect the likelihood of international migration, and we now turn to an analysis of these effects in the context of contrasting gender systems.

In their analysis of male and female migration from Mexico, Cerrutti and Massey (2001) used bivariate probit models to estimate simultaneous functions predicting the migration of husbands and wives and sons and daughters, while taking account of the degree to which their decisions were interconnected. Because marriage is so unstable in many of the settings under investigation, an analysis based only on currently married couples and intact families would exclude many – and in some cases most – men and women. Here we instead estimate separate logit regression models to predict the migration of male and female householders to the United States while controlling for the presence of a formal or informal spouse and relevant spousal characteristics on the right hand side of the prediction equation.

Our descriptive analysis of gender and migration clearly suggested that Mexico and Costa Rica were united in a common adherence to a patriarchal gender system, whereas Nicaragua and the Dominican Republic were similar in their exhibition of matrifocal gender patterns, and that Puerto Rico lay in-between, mixing features of a patriarchal gender system (a preference for legal marriage over cohabitation) with elements of a matrifocal family organization (a high prevalence of marital dissolution through divorce and separation). We, therefore, sought to estimate three models of male and female migration and compare the results to learn more about the interplay between gender and international migration: one for Mexico and Costa Rica combined (the patriarchal case), one for Nicaragua and the Dominican Republic combined (the matrifocal case), and a third one for Puerto Rico by itself (the middle case).

Unfortunately, the small size of the Puerto Rican sample and the relatively small number of men and women who had migrated within the three years prior to the survey did not yield sufficient degrees of freedom to produce stable equation estimates. As Massey and Sana (2004) demonstrated empirically, migration from Puerto Rico to the mainland peaked in the late 1960s and early 1970s, so that the migratory experience of Puerto Rican men and women was rather dated by the time the surveys were fielded in the late 1990s and few new US trips were being taken. As a result, none of the logit models we specified converged to a stable solution. We tried pooling the Puerto Rican data with that of Mexico and Costa Rica, on the one hand, and Nicaragua and the Dominican Republic, on the other; but in both cases, we found the addition of Puerto Rico yielded a significant reduction in goodness of fit.

The migration of male householders

We, therefore, chose to focus on Mexico-Costa Rica and Nicaragua-Dominican Republic as contrasting cases, indicating migration processes that prevail at opposite poles of a societal continuum running from patriarchy to matrifocality. Equation estimates predicting the out-migration of male householders are shown in Table 5. As already explained, these models are cross-sectional, with the dependent variable being whether or not the subject migrated during the three years preceding the survey date, which is predicted from indicators of general human capital (age, education greater than six years, the ratio of workers to members in the household), migration-specific human capital (whether or not the householder held legal documents enabling migration and employment in the United States and number of prior US trips), physical capital (whether or not the respondent owned real property or a business in the country of origin), social capital (whether or not a parent, sibling, or child of the respondent had ever been to the United States), marital status (whether or not the subject was currently in a legal marriage or informal union), and the migration status of the spouse or partner (whether he or she ever migrated to the United States, possessed legal residence documents, duration of the first US trip, and total number of trips). We also indicated country origins within each set of pooled two-country regressions using a dummy variable. These controls do not exhaust the list of factors potentially affecting migration, of course, especially those operating at the community or state level, but they do cover a range of salient individual characteristics that earlier work has found to be important in predicting international migration (Massey and Espinosa, 1997) and which are lacking in most data sets (such as legal status and cumulative migration experience).

TABLE 5.

Logit Models Predicting the Likelihood of Male Migration to the United States within the Three Years Prior to the Survey

	Mexico and Costa Rica		Nicaragua and Dominican Republic
Independent variables	B	SE	B	SE
General human capital
Age	−0.083***	0.003	−0.024	0.016
Education >6 years	−0.635***	0.068	0.101	0.419
Workers/HH members	0.049	0.149	−0.718	0.831
Migration-specific capital
Documented	1.502***	0.082	0.991	0.991
Number of prior trips	0.180***	0.009	0.600**	0.237
Physical capital
Owns property at home	0.193***	0.067	−0.270	0.425
Social capital
Has relatives in the US	0.795***	0.103	0.422	0.434
Marital status
Married or in union	0.122	0.167	0.296	0.775
Spouse’s migration status
Spouse a migrant	1.031***	0.122	−0.404	1.066
Documented	−0.219	0.154	0.441	0.872
Duration of first US trip	−0.010***	0.001	−0.006	0.005
Number of US trips	0.013	0.035	0.692*	0.356
Country
Costa Rica	−0.038	0.153	-	-
Dominican Republic	-	-	−0.780+	0.472
Intercept	−0.028	0.225	−3.020***	1.202
Number of male householders	14,677		1,810
Likelihood ratio	3,228.905		49.522
Somers’ D	0.706		0.582

Notes:

⁺p<.10;

^*p<.05;

^**p<.01;

^***p<.001.

The degree of a coefficient’s statistical significance is indicated by serial asterisks, with more asterisks indicating higher p-values. We also conducted tests to determine whether Mexican and Caribbean coefficients were statistically different from one another (simple t tests), and those rows where the difference was indeed significant are highlighted in bold (P<.05). The intercepts indicate the propensity for US migration among householders from each region, controlling for variables included in the model. As can be seen, the odds of male migration are much higher in the two patriarchal countries than in the pair of matrifocal nations. Moreover, we can detect no real differences between Mexico and Costa Rica among the patriarchal nations or between Nicaragua and the Dominican Republic among matriarchal countries, controlling for other variables in the equation.

There are clear contrasts in the process of male out-migration, however, as indicated by significant differences in coefficients between the two models. Male migration from the patriarchal countries is strongly and significantly selected on indicators of general human capital, but the selection is negative. People who are older (and thus have more work experience) and have achieved higher educations are significantly less likely to migrate. In contrast, male out-migration from the two matrifocal settings is unrelated to any indicator of general human capital. It thus seems that labour markets in patriarchal societies function to reward human capital and thus deter males who possess it from migrating abroad, whereas matrifocal societies do not, at least to the same extent.

Male migration from Mexico and Costa Rica seems to be more connected to capital endowments generally than that from Nicaragua and the Dominican Republic. Among the former countries, the likelihood of male out-migration is significantly and positively related to the ownership of physical capital and access to social capital, but there is little evidence of these effects among males from the matrifocal nations (though coefficients from the two models are not significantly different from one another).

Male emigration from Mexico and Costa Rica is also strongly related to migration-specific human capital – personal resources accumulated in the course of international migration itself. The probability of taking a US trip is thus positively related to the number of trips already taken and, not surprisingly, is strongly boosted by the possession of legal documents. Only for migration-specific human capital are similar results found for the matrifocal countries of Nicaragua and the Dominican Republic. The effect of legal status is rather large and positive, and though it is not significantly different from zero (likely because of smaller degrees of freedom), though neither is it different from the large coefficient obtained in the patriarchal nations. It is not much of an inferential leap to conclude that the odds of male migration are similarly raised by the possession of legal documents in both settings.

However, although the likelihood of taking a US trip is significantly and positively related to the number already taken in both equations, the effect is significantly stronger among males from Nicaragua and the Dominican Republic – more than three times greater. Thus, the forms of capital that typically figure in the decision to initiate migration (education, experience, and property ownership) appear relevant more in patriarchal than matrifocal settings, whereas those that typically sustain migration once it has begun are generally as influential – or in the case of prior experience, more influential – in both contexts.

The other great contrast has to do with the role of wives in promoting male migration across the two settings. In both sets of countries, being married or cohabiting has no bearing on the decision to migrate, what matters is that if a man is formally or informally linked romantically to a woman, whether she herself has been to the United States and the number of trips she has taken. In Mexico and Costa Rica, a male is much more likely to migrate if his wife (and partners in these countries are overwhelmingly wives rather than cohabiting partners) has prior migratory experience, whereas in Nicaragua and the Dominican Republic, where unions are typically unstable and of limited duration, the migratory experience of the wife or partner (mostly partners in this case) is irrelevant to male decision-making. The odds of male migration from matrifocal settings is more dependent on the number of prior trips a partner has taken rather than the simple fact of her having ever migrated.

The interplay between a wife or partner’s migrant status and her number of trips in affecting male propensities to migrate is shown in Figure 1, which generates predicted probabilities of migration for a married or cohabiting male depending on migrant status and number of trips taken by his partner or wife in patriarchal versus matrifocal societies. We used the separate equations for patriarchal and matrifocal settings and inserted into each one the mean values observed for Mexico-Costa Rica (thus holding background experience constant) and then varying the wife or partner’s migrant status and experience. As can be seen, the odds of a man’s migration from a patriarchal setting receives a significant boost if his wife (or partner) has prior experience in the United States, going from 0.06 for no trips to 0.15 with one trip; but thereafter there is no additional payoff for additional trips. In contrast, the odds that a male will emigrate from a matrifocal society is hardly affected by whether the partner (or wife) has migrated (compare the predicted probabilities at zero trips and one trip), but thereafter the likelihood of male migration rises with the number of trips taken by the partner, crossing over the patriarchal curve at three trips.

FIGURE 1.

Predicted Probability that a Male Household Migrates To the United States from a Patriarchal Versus Matrifocal Setting by Wife’S Migratory Experience

The migration of female householders

Table 6 replicates the analysis of the determinants of out-migration for female householders in the two contrasting settings. As can be seen, the intercept estimated for females from Mexico-Costa Rica is significantly above that estimated for women from Nicaragua-Dominican Republic, suggesting that the women from patriarchal settings have a higher propensity to migrate. This contrast is deceiving, however, because one needs to take into account marital status. In the two patriarchal nations, being married or cohabiting hugely and significantly reduces the odds of female migration whereas in the two matriarchal countries, it does not. In Nicaragua and the Dominican Republic being married, if anything, raises the odds of out-migration, though the effect is not significant.

TABLE 6.

Logit Models Predicting the Likelihood of Female Migration to the United States Within the three Years Prior to the Survey

	Mexico and Costa Rica		Nicaragua and Dominican Republic
Independent variables	B	SE	B	SE
General human capital
Age	−0.031***	0.005	0.006	0.020
Education >6 years	0.110	0.123	0.186	0.521
Workers/HH members	1.324***	0.211	0.967	0.808
Migration-specific capital
Documented	2.667***	0.133	1.540**	0.619
Number of prior trips	0.189***	0.028	1.348***	0.336
Physical capital
Owns property at home	−0.499***	0.114	−0.367	0.528
Social capital
Has relatives in the US	1.662***	0.322	1.090+	0.585
Marital status
Married or in union	−2.216***	0.218	0.585	0.619
Spouse’s migration status
Spouse a migrant	2.244***	0.198	1.491 +	0.914
Documented	0.008	0.140	−2.876***	0.995
Duration of first US trip	−0.005***	0.001	−0.008	0.007
Number of US trips	0.023**	0.011	0.347	0.436
Country
Costa Rica	0.220	0.309
Dominican Republic	-	-	−0.572	0.517
Intercept	−3.916***	0.402	−6.706***	1.346
Number of male householders	16,178		2,329
Likelihood ratio	1,691.940		84.002
Somers’ D	0.832		0.751

Notes:

⁺p<.10;

^*p<.05;

^**p<.01;

^***p<.001.

As a result, when the coefficient for marriage is added to the intercept in both settings, the resulting sum is virtually identical. In a patriarchal society, the typical woman is married and this union subjects her to male control, which reduces her likelihood of international out-migration, whereas in a matrifocal setting the typical woman is unmarried, but whether she is or is not in a union has a very minor influence on her odds of leaving for the United States, and what influence there is appears to be positive. For most women, therefore, the underlying propensity to migrate (net of other variables in the equation) is roughly equal across matrifocal and patriarchal cultural contexts.

In addition to the contrasting effects of marriage or cohabitation itself, the migrant status of the husband or partner also differs markedly across settings. Within a patriarchal context, a woman’s migration behaviour is very closely related to that of her husband, as Cerutti and Massey (2003) have shown. Thus, in Mexico and Costa Rica, though married women are less likely to migrate generally, they are far more likely to do so if their husbands are US migrants, and the propensity rises the more trips he has taken (though it is reduced if the husband had a long duration of stay on the first trip).

In matrifocal settings, however, female migratory behaviour is less clearly related to the migration characteristics of the husband or partner. Female out-migration is only weakly related to the migrant status of the male spouse or partner (p<.10) and not significantly related to the number of prior trips or the duration of his first US trip. Moreover, unlike females from Mexico and Costa Rica, those from Nicaragua and the Dominican Republic are markedly less likely to migrate if their husbands have legal documents.

We also observe contrasts across settings with respect to the influence of various forms of capital. Whereas women from Mexico and Costa Rica are negatively selected with respect to age and positively associated with respect to the number of workers per household, Nicaraguan and Dominican women are not significantly selected at all (though all the coefficients are positive). Moreover, whereas women in both contexts are more likely to migrate if they are themselves documented, the effect is significantly stronger for those from Mexico and Costa Rica. We also observe a significant negative association with respect to property ownership in patriarchal but not matrifocal countries (though admittedly the coefficients do not differ from each other).

Both groups of women respond positively to social capital, however, being significantly more likely to migrate if they have relatives living in the United States than if they do not, though the effect is only marginally significant and somewhat smaller among women from Nicaragua and the Dominican Republic. Perhaps the greatest contrast in capital-related effects is the key indicator of migration-specific human capital, here measured by the number of prior US trips. Prior work has found this to be a key nexus for the cumulative causation of migration, at least among men (Massey and Espinosa, 1997; Massey and Zenteno, 1999). The more trips a man has taken, the more likely he is to take an additional trip because his experience migrating, crossing the border, finding a job, and working in the United States simultaneously reduce the costs of additional migration and increase the potential value of the next trip (Massey, 1990). A strong positive coefficient thus indicates that the person in question is an autonomous actor operating to maximize the returns to migration across specific trips.

While this effect is not absent among women from Mexico and Costa Rica, it is much smaller than that observed among those from Nicaragua and the Dominican Republic. The coefficient on prior trips of 0.189 for females from patriarchal settings is about the same as that for their male counterparts (0.180). Among women from matrifocal settings, the coefficient on number of prior trips is not only more than twice that of their male counterparts (1.35 versus 0.60), it is more than seven times that of women in patriarchal countries. In matrifocal societies, in other words, women are more likely to migrate autonomously as active participants in international labour markets and to contribute to the process of cumulative causation, both in comparison to women in patriarchal societies and in comparison to men in their own setting.

CONCLUSION

In this paper we followed theorizing laid down by Hondagneu-Sotelo (1994) and hypothesized that the social construction of gender in different societies affects the migration of men and women in different and distinct ways. We tested this idea by undertaking a comparative analysis of male and female out-migration in five societies with very different gender-family systems: Mexico, Costa Rica, Puerto Rico, Nicaragua, and the Dominican Republic (Safa, 1995). Comparable data sets from the MMP and the LAMP were combined and used to document these differences, in concert with data compiled from national censuses.

A preliminary reconnaissance of data from these countries suggested that they could be objectively placed along a continuum of gender relations ranging from patriarchal to matrifocal. Both census data and information from the LAMP/ MMP were consistent in placing Mexico and Costa Rica close together at the patriarchal end of the spectrum, and in locating Nicaragua and the Dominican Republic close together at the matrifocal end. Puerto Rico occupied a middle position, blending characteristics of both gender systems – a preference for marriage over cohabitation as in patriarchal countries and a high rate of marital dissolution as in matrifocal settings.

These differences in gender-family systems were reflected in different patterns and propensities of female relative to male migration. Female householders in the two most patriarchal settings displayed relatively low rates of out-migration compared with males whereas in the two matrifocal countries the ratio of female to male migration was much higher. In fact, within some marital categories – generally those in which women were unencumbered by men, such as never married, separated, divorced, or widowed – female householders were more likely than their male counterparts to migrate to the United States.

Multivariate analyses of male and female migration revealed differences in the process of international out-migration that were interpretable in light of contrasting gender systems. In patriarchal societies such as Mexico and Costa Rica, male migration is strongly related to capital endowments, being negatively selected with respect to education and labour market experience (which are rewarded in patriarchal societies) and positively selected with respect to physical capital (which, controlled by men, can be used to finance trips and serve as targets for investment of their earnings). Migration is also promoted by migration-relevant human and social capital, increasing steadily as the number of prior US trips rises and as the number of social connections to the United States increases. Male migration is unrelated to marital status, but if a wife is a migrant, men are much more likely to migrate themselves.

In matrifocal societies such as Nicaragua and the Dominican Republic, in contrast, the odds of out-migration are much less dependent on capital endowments. The only form of capital that consistently predicted out-migration was the migration-specific human capital indexed by the number of prior US trips. In general, the forms of capital that typically figure in the decision to initiate male migration (education, experience, and property ownership) appear relevant more in patriarchal than matrifocal settings, whereas those that typically sustain male migration once it has begun are generally as influential – or in the case of prior experience, more influential – in both contexts. In contrast to the situation in patriarchal societies, the propensity of men to emigrate from matrifocal settings is unrelated to the migrant status of the wife or partner, though it does increase as the number of spousal trips rises.

Contrasts in the process of international out-migration are greater for women than for men across patriarchal versus matrifocal settings. In the context of patriarchy, a formal or informal union with a man dramatically lowers the odds of female out-migration, other things equal, whereas in matrifocal societies marriage and cohabitation have no real effect. The most important determinants of female migration from patriarchal settings are the migrant status of the husband or partner, the presence of other relatives in the United States, and the possession of legal documents. Female migration also falls sharply with age, and although the odds of out-migration increase with the number of prior US trips, this effect is not dominant in the estimates derived from the two patriarchal settings. In general, female householders in patriarchal settings appear to behave much more as “tied” movers accompanying other family members compared with those from matrifocal contexts.

In matrifocal settings such as Nicaragua and the Dominican Republic, female migration is less related to the possession of documents, less related to a husband or partner’s migrant status, and less connected to the presence of other relatives in the United States, and it is more strongly related to the woman’s own prior migratory experience. Thus, whereas the process of cumulative causation appears to be driven largely by men in patriarchal societies, it is women who dominate the process in matrifocal settings. Far more than their male counterparts in either patriarchal or matrifocal settings, women from matrifocal societies such as Nicaragua and the Dominican Republic are far more likely to migrate again once they have begun, and the odds of taking an additional trip rise very steeply with each trip taken. All in all, women emigrating from matrifocal settings appear more to be independent actors than tied movers, a conclusion that is underscored by the fact that the odds of female migration in such settings is negatively related to the documentation status of husbands or partners.

The foregoing analysis thus offers quantitative evidence for the argument that gender does matter in international migration. Not only do the determinants of male and female migration differ in different contexts, but the pattern of differences is conditioned in interpretable ways by the prevailing system of gender relations – whether patriarchal or matrifocal. In societies where women are more autonomous, independent, and less tied to men as partners, they are more likely to migrate as independent agents. Gender not only influences which determinants of migration matter and how much, but also determine what the characteristics of the immigrant population ultimately are. To the extent that there are gendered differences in patterns and processes of assimilation, therefore, immigrant populations with different gender compositions can be expected to assimilate in different ways and at different rates.