Household Composition, Family Migration and Community Context. Migrant Remittances in Four Countries

Transcription

1 Household Composition, Family Migration and Community Context. Migrant Remittances in Four Countries Mariano Sana University of Pennsylvania Prepared for delivery at the 2003 meeting of the Latin American Studies Association, Dallas, Texas, March 27-29, 2003 Contact Information Mariano Sana, Project Manager, Latin American Migration Project (LAMP) Population Studies Center, University of Pennsylvania 3718 Locust Walk, room 239 Philadelphia, PA

2 Household Composition, Family Migration and Community Context. Migrant Remittances in Four Countries Mariano Sana University of Pennsylvania Worldwide migrant remittances exceeded $60 billion in 2000, with nearly 83% of that amount going to developing countries (International Monetary Fund 2001, Table B-19.) This figure is based on national balance of payments statistics, which cannot account for international transfers made through non-institutional channels. A recent survey of Latin American immigrants in the United States recorded that, while the vast majority of remittances are channeled through financial firms, 15% of them reach home through people traveling (Bendixen and Associates 2002). Hence, the global remittance flow is surely underestimated by billions of dollars. Not only is the sheer magnitude of migrant remittances impressive, but it has been continuously on the rise over the years. The IMF (2001) reports that worldwide remittances increased by about a third since The largest growth was recorded in Latin America and the Caribbean, where workers remittances rose from 9.7 billion dollars in 1994 to 17.1 billion in Among developing regions, Latin America and the Caribbean saw their share of remittances increase from 28.3% to 33.1%. In most Latin American countries migrant remittances exceed official development assistance inflows. They also amount to nearly one-third of the region s foreign direct investment, and account for at least 10% of GDP in five countries: Nicaragua (24%), Haiti (20%), El Salvador (14%), Jamaica (13%), and Ecuador (10%) (Multilateral Investment Fund 2002). Given these impressive figures, it is not surprising that remittances have attracted the attention of scholars from many disciplines. They have studied remittances at both the macro and micro level. Studies at the micro level focused on remitting behavior in two ways: from the point of view of the migrants who send money to their home countries, and from the point of view of the households that receive the money. This study should be classified among the latter. Using survey data collected between 1999 and 2001, I compare households in Mexico, the Dominican Republic, Nicaragua and Costa Rica. I examine how remittances vary by household composition, by whether there are different family members living abroad, and by various features of community context. Theory Only one comprehensive theory of international migration adjudicates a prominent role to remittances: the new economics of labor migration, developed under the lead of Oded Stark (Stark 1991). Stark and his colleagues initially looked at rural-to-urban migration in less developed countries. At the core of the new theory was the need for diversification of family labor, which arises in rural settings where local markets are too imperfect and incomplete. In Stark s words: [E]ntry to a specific labor market is often barred by constraints in capital, commodity, or financial markets. These characteristics tend to encourage migratory phenomena that would not have arisen if, for example, information were completely symmetric, if financial 2

3 (insurance, credit) institutions functioned smoothly, or if returns to exchange among agents exhibited linear regularities. (Stark 1991:4). The new economics of labor migration does not apply to settings that feature markets with smooth, well-functioning financial institutions. It is not readily applicable, for example, to the study of internal migration in developed countries. Incomplete and imperfect markets are a necessary condition for the migration of labor. Stark and his colleagues define migration as a tool that households use to manage the various risks that result from underdeveloped markets. By sending a family member to work away from home, a household makes an investment that is expected to pay off when the migrant's remittances arrive. These remittances are used to cope with local market failures, such as the lack or malfunctioning of capital markets, crop insurance, futures markets, unemployment insurance or social security. There is a second necessary condition for the occurrence of migration: an implicit contractual arrangement between the family at home and the migrant (Stark and Lucas 1988). In contrast with previous economic models, migration does not arise as a result of a decision made by an isolated individual. It is, instead, part of a family strategy. The family can be conceived as a coalition, a group of players committed by choice to act as one unit vis-à-vis the rest of the world (Stark 1991:5). Just like an individual, the family seeks to maximize utility. A skillful diversification of its allocation of labor is crucial to achieve this goal. Although the migrant typically the oldest son relocates away from home, he remains bound to his family of origin. The monies that he sends home are part of an implicit contractual arrangement between two parties: the household at home and the migrant. Without it, there would be no guarantee of reciprocity and remittances might not happen. An obvious question is what would enforce compliance with the terms of the contract, especially after the migrant has obtained good employment at his destination and is generating income. The answer is altruism, which enforces the contractual arrangement and eliminates the need for costly legal safeguards (Stark and Lucas 1988:469). Stark and Lucas naturally understand that altruism is a more powerful force among members of the same family than among, for instance, randomly assembled people. Of course there are exceptions, but in the aggregate altruism within families should prevail over altruism among non relatives. Stark and Lucas reckon, however, that altruism is likely to wane over time. It follows that the reliability of the contractual arrangement between the migrant and the family left at home will depend on the strength of altruism. The degree of altruism is likely to vary considerably across families, being strongest among cohesive, traditional families, and weaker among nontraditional families with unstable bonds. A fairly strong degree of altruism is needed to justify the family investment on the migrant s trip and to expect remittances from him. In other words, the new economics of labor migration implicitly assumes a cohesive, traditional family, the members of which share common goals, and are likely to trust, and remain loyal to, each other. The theory does not have an explicit hypothesis or assumption regarding trip duration, but it certainly makes more sense when the migrant expects to return home. A common theme in the literature, for example, is the expectation to inherit land or other forms of wealth as a factor that at least partially motivates remittances (Stark and Lucas 1988, Hoddinott 1992, de la Brière et al. 2002). A migrant who expects to return home will have ever more incentives to comply with the contractual arrangement between him and his family. This is acknowledged in the concept of tempered altruism or enlightened self-interest, which considers both altruism and individual 3

4 gain as motivations to remit when the migrant s return is likely (Lucas and Stark 1985). Thus, the theory focuses on a particular type of migrant: the short-term, target-oriented worker, who identifies with his family of origin at home, and who expects to return, as opposed to settling in the host country. This migrant is most likely male (hence my use of he and his above), and he could be either a son of the household head or the household head himself. Provided that local markets are incomplete and imperfect, and that the family is cohesive enough to guarantee a self-enforcing contractual arrangement, migration and remittances may happen, as long as the family has the means to finance a trip. The household member who moves will probably conform to the profile of the temporary and goal-oriented labor migrant outlined above. When migrants are not sojourners who pursue an earnings target, however, this remittances narrative is less appealing. In particular, even though it takes the household as the unit of analysis, the theory overlooks gender and is especially rigid with respect to household dynamics. The complete absence of gender in the formulation of the theory suggests that male and female migrants should behave the same. Yet, researchers have found that investment motivations typically split across gender lines, with sons remittances being more responsive to parental inheritable assets than daughters remittances (Hoddinott 1992, De La Cruz 1995, de la Brière et al ) Changing households pose further problems to the theory. The migrant could marry and establish a new household in the host country. Alternatively, the original household that sent a migrant to work abroad may eventually relocate to the destination. The first case most typically involves migrant children and imposes a loyalty conflict upon the migrant-family contractual arrangement. In the second scenario the (typically male) migrant is usually joined by his family and the contract becomes obsolete. The empirical record, however, shows that large numbers of migrants keep sending remittances to their home countries event after these events. Taking into account these limitations, however, the new economics of labor migration does provide a theoretical rationale useful for the study of migrant remittances. In this paper, I focus on household composition, family migration and community context because these factors are at the core of the theory. First, we should expect remittances to be associated with cohesive families, which presumably should be better enforcers of the contractual arrangement between the migrant and the family left at home. Second, the migration of immediate family members should be more strongly associated with remittances than the migration of more distant kin. Finally, since remittances are a tool for risk diversification, remittances should be more likely to flow to communities were markets are underdeveloped. If, in addition to risk diversification, remittances are meant to be used for local investment, underdeveloped markets should nevertheless be accompanied by certain local infrastructure that can support investment. Data I use data gathered by the Mexican Migration Project (MMP) and the Latin American Migration Project (LAMP), twin collaborative research projects based at the University of Pennsylvania and the University of Guadalajara. Since 1982 the MMP carries out surveys in Mexican communities, most of which are located in the western part of the country, where migration to the United States is most prevalent. The LAMP began activities in 1998 with a set of surveys in Puerto Rico, and expanded later to cover communities in the Dominican Republic, Nicaragua, Costa Rica, Haiti and Peru. In each survey site the MMP and the LAMP draw a random sample of between 100 and 4

5 200 households. Both projects complement each survey with a small snowball sample no more than 20 households of migrants from the same community who have settled in the United States. The MMP and LAMP methodology relies on the ethnosurvey questionnaire (see Massey et al. 1987: chapter 2), a semi-structured interview schedule that gathers a rich array of demographic, socioeconomic and migration-specific information. In 1999, both the MMP and the LAMP introduced a new question into their survey schedules. MMP and LAMP fieldworkers asked whether each household interviewed in the home community regularly received remittances sent from the United States. When the answer was positive, the interviewers inquired whether the amount received, when compared to the total of household income from all sources, was small, intermediate or substantial. The information revealed by those two questions are at the core of this paper. 1 In addition to remittances received from the United States, the LAMP-Nicaraguan surveys collected data on remittances received from Costa Rica, the primary country of destination for Nicaraguan international migrants. Furthermore, in one Dominican community with an established network of migrants in Madrid, LAMP fieldworkers applied the same questions to remittances received from Spain. In this paper I combine the answers to all these questions to study whether households received remittances irrespective of their source, and what the size of those remittances (compared to total household income) is. In any case did the MMP and the LAMP collect data on who the sender of these remittances was, but the projects did collect information on family members and other relatives living abroad. Table 1 provides some information on the communities included in the analysis: ten located in Mexico, seven in the Dominican Republic, five in Nicaragua and four in Costa Rica. All of them were interviewed between 1999 and They range from small rural villages with about 1,000 inhabitants, to portions of large neighborhoods in densely populated cities. Naturally, survey sites contained only small proportions of the total population of large cities, while representing much larger shares in small towns. Within survey sites, sampling fractions (the proportion of randomly selected and interviewed households) show variation, oscillating between.04 and.40, plus a few small communities were all households within the survey site were interviewed. Remittances among the interviewed households Taking official IMF figures and adding an estimate of unreported remittances, the Multilateral Investment Fund of the Interamerican Development Bank estimated that a total of 1.8 billion dollars in migrant remittances entered the Dominican Republic in The figure for Mexico was 9.3 billion, while the estimate for Nicaragua was 610 million (Multilateral Investment Fund 1 Since their inceptions, both the MMP and the LAMP provide detailed data on remittances sent by migrant household heads during their last trip to the United States. That data has been used extensively in both micro studies of remitting behavior and macro studies of remittances and development (Durand, Parrado and Massey 1996; Durand et al. 1996; Massey and Parrado 1994 and 1998; Sana 2002 and 2003). In this study, I use the new data on remittances received by households, rather than the alternative data on remittances sent by household heads in the United States. 5

6 2002). After dividing each figure for the population of each country, we obtain $217 per capita in the Dominican Republic, $124 per capita in Nicaragua, and $95 per capita in Mexico. Table 2 shows statistics on remittances received by the households interviewed by the MMP and the LAMP. In the Dominican Republic, 29% of the households received remittances, while the figure decreases to 15% among Nicaraguan households and 13% among Mexican household, an order that strongly parallels the ordering of these countries by remittances per capita. Not surprisingly since the incidence of remittances is much lower in Costa Rica, only 5% of Costa Rican households interviewed by the LAMP receive money from migrants abroad. Table 2 also shows the relative size of remittances with respect to total household income. Mexican households appear to be the most dependent on remittances, with 40% of them reporting that remittances constitute a substantial part of their income. For the other three countries, this figure ranges from 24 to 29 percent. This statistic is telling, since Mexico features the type of family that most approximates the cohesive type that the theorists of the new economics of labor migration had in mind, which is necessary to enforce the contractual arrangement between the migrant and the family left at home. A comparison of census statistics between Mexico and the Dominican Republic, for example, shows much lower incidence of divorce and of consensual unions in the former (Sana 2002). In addition, MMP and LAMP data reveal that a history of repeated unions is much more common among Dominican household heads than among Mexican household heads (data not shown.) Remittances and household composition Table 3 provides more information on the different types of families that we find in these countries. Households are classified based on the classical typology of families developed by Hammel and Laslett (1974), opening the simple category following De Vos (1995:40). Extended refers to a nuclear household plus other relatives. Multiple (or multiple nuclear ) households contain two couples (each of them a nucleus.) The single nuclear household, which consists of a couple and their children, is the norm among the households interviewed by the MMP: almost two thirds of Mexican households in the sample belong of this type. The Dominican Republic and Nicaragua show a lower, and similar, incidence of nuclear households (40%). These two countries also show the same proportion of extended households (about 26-27%) and single-parent households (about 10-11%). Multiple households are most commonly found in Nicaragua. The table also shows that 2-generation households are the most prevalent in all four countries, involving between 64 and 78 percent of all households. The maximum is found in Mexico and the minimum in the Dominican Republic and Nicaragua. In the Dominican Republic, the proportion of three-generation households doubles that of Mexico s, while in Nicaragua, it triples it. In part, this may be the result of a matrifocal family structure (see Barrow 1996), in which households cluster around women of successive generations in the absence of stable male breadwinners, or as a survival strategy to maximize income and minimize housing expenses per capita. Table 4 provides remittances statistics for the different types of households. Single nuclear households appear to be the least likely to receive remittances in nearly all four countries. The table raises questions of endogeneity: are remittances especially directed to certain types of households because of these households composition, or is it household composition what 6

7 determines migration (and subsequent remittances) in the first place? Selecting households that receive remittances, we can see the percentage for which these remittances represent intermediate or substantial amounts. It is interesting to note that, for nuclear households, this statistic is highest in Mexico (65%), while, among all other types of households, the percentage among both Dominicans and Nicaraguans is always higher than that recorded for Mexico. While the incidence of remittances is undoubtedly higher among Dominicans and Nicaraguan households than among Mexican households, these figures show that Mexican nuclear households are actually more dependent on remittances than Dominican and Nicaraguan households. 2 Remittances and family living abroad Undoubtedly, there will be a close association between family migration experience and remittances. Table 5 shows statistics on family migration and on relatives living abroad. By far, the most likely migrant in this sample is a Mexican male householder. This result is not surprising, since Mexican migration to the United States is a massive phenomenon, but it is also of mostly temporary and seasonal nature (Massey, Durand and Malone 2002). Given that these surveys were conducted in the home countries of these migrants, households that moved entirely to the United States would be censored out of the sample, an occurrence that is much more common among Dominican than among Mexican households. Female householders from the Dominican Republic and Nicaragua are most likely to migrate than those from Mexico. (Figures on sons and daughters pending). How do these household migration profiles translate into family members living abroad at the time of the survey? While Dominicans and Mexicans show the highest percentage of male householders abroad at the time of the survey (3.8% and 3.5% respectively), Dominicans alone are on top when it comes to female householders abroad (1.2%). Mexicans report almost one son abroad every four households, while among Dominicans it is about one every five households. Figures for daughters show a slightly larger number for Dominicans than for Mexicans. A look at other relatives suggests that, in this cross-sectional picture, Dominicans have the largest kin networks in the United States, followed by Mexicans. Table 6 shows the relationship between remittances and family and other kin living abroad. For all categories of immediate family members currently abroad, both Dominican and Nicaraguan households are more likely to receive remittances than Mexican households. I attribute this result to the different types of migration prevalent in these countries. Mexican migrant trips to the United States are shorter than Dominican or Nicaraguan trips, and are also recurrent. MMP and LAMP data show an average of 2.18 trips for Mexican migrants, and 1.25 for both Dominicans and Nicaraguans. In turn, the last trip among these migrants was shortest for Mexicans: 66 months on average (with the 25 th percentile at 8 months), compared to 138 for Dominicans (25 th percentile at 54 months) and 102 for Nicaraguans (25 th percentile at 24 months. Frequently, Mexican migrants do not need to send remittances home in order to repatriate savings earned in the United States. They may just carry the savings back home with them at the end of their trips. The same is less likely to happen among Dominican and Nicaraguan migrants. For them the costs 2 The difference between Mexico and the Dominican Republic, however, is not significant. 7

8 of the move are higher and they feel stronger incentives to stay in the U.S. for a longer time. The households at home, however, may not be able to wait that long. For the most part, remittances are also positively associated 3 with the existence of other kin living abroad: siblings, uncles/aunts and cousins of the household head. The figures in the bottom row summarize remittances among households that have any of these relatives living abroad. While the percentage of households that receive remittances remains at more or less moderate levels, the incidence of remittances in the total income of these households is remarkable, with more than 50% of households in Mexico, the Dominican Republic and Costa Rica reporting that remittances are at an intermediate or substantial level. Remittances and community context As mentioned in the data section above, the survey sites range from small rural towns to neighborhoods in large cities. Table 7 shows that the Mexican survey sites cover the whole range of community population sizes. The surveyed Dominican communities concentrate in the 5,000-25,000 range and large cities. Nicaraguan data comes from relatively small towns. The four Costa Rican communities are neither too small nor too large. These places of various sizes recognize different degrees of development, although nearly all of them feature paved streets, electricity, potable water, public lighting, telephone service, and at least one post office, school and bank. All these institutions and services, however, were introduced at different points in time. Table 7 shows, for each country, the mean number of years elapsed since the introduction of each of these institutions and services. The simple mean of all of these, labeled mean development age, is also provided in Table 7. We see that, in the aggregate, these institutions and services have a longer history in the Dominican communities (48 years on average), and are newest in Nicaragua (30 years). The MMP and LAMP collect some information on local businesses in the survey site, but these data are not always comparable across countries. Table 7 shows the four variables that can be compared: the existence of a daily market, the number of video stores, the number of bank offices, and the existence of taxi service. We see considerable variability across countries. According to the new economics of labor migration, households send migrants away in order to accumulate capital to cope with imperfect local markets. A straightforward implication is that remittances should arrive in places where there are no banks or there are too few of them to guarantee competitive interest rates. In addition, if migrant households intend to invest in businesses, they would need business opportunities and a minimum infrastructure. Business opportunities are, of course, impossible to measure. As an approximation, we can speculate that, where there is infrastructure but there are not enough businesses, then business opportunities would be available. In general, we would expect to find less attractive prospects for new small businesses in large cities than in small towns, since competition in large cities is fiercer. In addition, large cities generally offer more job opportunities for household members, in a way that facilitates the diversification of risk that migration and remittances are supposed to respond to. Table 8 shows, for all the variables presented in Table 7, Pearson coefficients of correlation with a dichotomous variable representing whether the household receives remittances, 3 Compare the figures in the bottom panel with the totals at the top of the table. 8

9 and with a discrete continuous variable on the relative size of those remittances compared to total household income, where 1=small, 2=intermediate, and 3=substantial. There is no strong correlation between community population and the reception of remittances. However, consistent with the reasoning above, as a share of household income remittances lose importance as population size increases. Correlations between remittances and the number of years that each service or institution has been available are overwhelmingly positive and significant. The summary measure, mean development age, shows a strong and positive correlation with both the existence of remittances and the relative size of remittances. These statistics indicate that remittances are more likely to flow toward communities with a certain history of infrastructure and services. The bottom panel of Table 8 shows negative correlations between all the business variables and whether households receive remittances. These correlations are significant (video stores and taxi service) or marginally significant (p<.15 for daily market and bank offices). These results are consistent with the reasoning above, according to which a smaller number of businesses might be an indicator of business opportunities. However, the same reading does not apply to the correlations between the business variables and the relative size of these remittances Multivariate models After this review of different factors that presumably have an effect on remittances, it is worth considering how they all interact together. Table 9 shows estimates from four specifications of a logistic regression that predicts a household s probability of receiving remittances. All four specifications share dummy variables for country. The first model adds immediate family and other kin living abroad to the country dummies. The second model considers the country dummies together with household composition. The next model combines the country dummies with measures of community context. Model 4 is the full model that includes all the variables. 4 The models control for country, assuming that family living abroad, household composition, and community context have the same effect for households in different countries, a relatively strong assumption as we will see soon. In general, coefficients do not change much across specifications. Most of those for immediate family and other relatives living abroad are nearly identical in model 1 and model 4. The odds of receiving remittances for households with their male householder abroad are an impressive 29.3 (model 1) and 39.6 (model 4) times the odds for households without a male householder abroad. Surprisingly, a female householder living abroad does not appear to be a predictor of remittances (more on this below). The odds ratio for households with at least one son abroad versus those with no sons abroad are slightly above 7, declining to 5.6 for daughters (model 4). For households with other kin living abroad, the effect is still positive but moderate: the odds ratio falls near 1.5. The estimates for the household composition variables confirm that single nuclear households (the omitted category) are the least likely to receive remittances. Interestingly, this effect remains even after controlling for immediate family living abroad. The effect of the number of generations does vary considerably, from (odds ratio=1.21) in model 2, to (odds 4 Ordinal logit models for the relative size of remittances will be introduced in a revised version of this paper. 9

10 ratio=1.36, and statistically significant) in the full model, where family living abroad and community context are accounted for. Finally, measures of community context appear to be only weak predictors of the likelihood of receiving remittances, except for the log of community population, which shows the negative association between population size and remittances. The last results are disappointing when it comes to evaluate the rationale of the new economics of labor migration. However, these results are an effect of pooling together data from different countries, with very different family structures and degrees of market development, in addition to different migration dynamics. In previous work (Sana 2002) I showed that its traditional patriarchal family structure, combined with its underdeveloped markets, makes Mexico, and especially Western Mexico, and ideal setting for the rationale proposed by the new economics of labor migration. Mexican migrants fit the profile of the temporary and goaloriented male labor migrant, committed to his family of origin, that the theorists of the new economics had in mind. In contrast, Dominican migrants do not fit this profile. Dominican family structure resembles a Caribbean model, where marital disruption and unstable commitment are more prevalent than in Mexico. As a consequence, investment-oriented remittances for family businesses are less likely to happen, since family bonds are weaker. While there is an important business-oriented remittance flow in Mexico (Massey and Parrado 1998), most Dominican remittances seem to be destined to consumption and family subsistence (Grasmuck and Pessar 1991, Itzigsohn 1995). I explored this idea further by separately running the full model from Table 9 on Mexican and Dominican households. Table 10 shows the results, and we quickly see that there are important differences between the two countries. First, a female householder living abroad is a strong negative predictor of remittances for the Dominican Republic. This is consistent with research findings that show that, once Dominican couples relocate abroad, females tend to strongly favor settlement over return, embarking in strategies of investment in the destination country and opposing the detour of resources toward the home country (Grassmuck and Pessar 1991, Pessar 1990). Among children of the household head, we see that the presence of sons living abroad is a stronger predictor of remittances among Mexican households, while daughters living abroad are more strongly associated with remittances among Dominican households. These results seem to reflect the two different types of family structure in both countries: among Mexicans, sons are expected to be loyal remitters. In the Dominican society, men are perceived as less reliable, and remittances from daughters take prevalence. The coefficients for household composition are revealing as well. We see that, among Mexican households, single nuclear households are no longer in a disadvantaged position when it comes to the likelihood of receiving remittances. Since nuclear households are the most typical representation of the traditional family, this result is telling, even though endogeneity problems with these variables persist. The estimated coefficients for the variables on community context also show different results between the two countries. Signs for the Mexican coefficients are consistent with the reasoning outlined above and derived from the new economics of labor migration: population size has a negative effect on the likelihood of remitting, the effect of development age is positive, and all the coefficients for the business variables show negative signs. The combined reading of these coefficients suggests that Mexican remittances are more likely to flow toward communities 10

11 of moderate size, with certain development history and infrastructure, but limited number of businesses. All combined, these factors suggest the existence of business opportunities. These conclusions must be taken cautiously, since this rationale involves assumptions, and some of the coefficients themselves show high standard errors. The estimated equation for Dominican households, in turn, shows mixed results. Some coefficients suggest a similar interpretation to the one just outlined for Mexican households (see the coefficients for community population and daily market in community). However, other coefficients point in the opposite direction (bank offices, taxis), indicating the operation of even more complex forces. Conclusions In this paper I study remittances among households surveyed by the Mexican Migration Project (MMP) and the Latin American Migration Project (LAMP) in Mexico, the Dominican Republic, Nicaragua and Costa Rica, between 1999 and These remittances are sent mostly from the United States, but also from Costa Rica (to Nicaragua) and Spain (to the Dominican Republic). The first basic statistics confirmed that the percentages of households that receive remittances in the countries under study are consistent with what is known from official balance of payments statistics: almost 30% of Dominican households in the sample receive remittances, about 15% of Nicaraguan households, 13% of Mexican households, and about 5% of those sampled in Costa Rica. The analysis focused on three sets of factors: immediate family members and other kin living abroad, household composition, and community context. I used the rationale of the new economics of labor migration to derive expectations. The analysis of immediate family members and other kin living abroad showed a not surprising positive association between family and kin expatriates and remittances. However, the data revealed an interesting gender twist: a female householder living abroad is not a strong predictor of the likelihood of receiving remittances. When the analysis was carried out separately for Mexico and the Dominican Republic, the effect of a female householder living abroad actually became negative among Dominican households, consistent with research findings that show that Dominican women in the United States tend to strongly favor settlement over investment in the Dominican Republic. The same separate analysis revealed that the negative association between remittances and single nuclear households, verified at the aggregate level, does not hold among Mexicans once all factors considered here are controlled for. Together, these findings suggest that family structure and gender roles play an important role as determinants of remittances. While in Mexico remittances seem to be associated to the patriarchal traditional family, in the Dominican Republic the opposite seems to take place. The new economics of labor migration postulates that households adopt international migration as a strategy for risk diversification in the absence or scarcity of capital and insurance. If this is true, it follows that migrants will send remittances to communities with few businesses (underdeveloped markets). In addition, if families send migrants abroad to repatriate capital for investment, certain local infrastructure provides an environment conducive to such strategy. Businesses are more likely to succeed in places were there is infrastructure but at the same time there is no fierce competition. In view of this, we should expect remittances to be negatively associated with population size, positively associated to degree of development (infrastructure), and negatively associated with the number of businesses in the local community. When these 11

12 expectations were tested using pooled data from the four countries under study, they could not be confirmed. However, when the same analysis was carried out separately for Mexico and the Dominican Republic, remittances among Mexican households did conform to expectations. This result suggests, consistently with previous research, that Mexico provides an ideal setting for the new economics of labor migration, while the new economics rationale may not be readily applicable to other countries such as the Dominican Republic. The ultimate reason for this mismatch lies in the non-traditional Dominican family structure, which potentially undermines the risk-diversification and investment logic of the new economics of labor migration. References Bendixen and Associates "Survey of Remittance Senders: U.S. to Latin America." Presented at the Second Conference on Remittances as a Development Tool. Washington, DC, February 26. On line at de la Briere, Bénédicte, Elisabeth Sadoulet, Alain de Janvry, and Sylvie Lambert "The Roles of Destination, Gender, and Household Composition in Explaining Remittances: An Analysis for the Dominican Sierra." Journal of Development Studies 68(2): De La Cruz, Blanca E "The Socioeconomic Dimensions of Remittances: A Case Study of Five." Berkeley McNair Journal 3(Summer):1-7. De Vos, Susan M Household Composition in Latin America. New York and London: Plenum Press. Durand, Jorge, William Kandel, Emilio A. Parrado, and Douglas S. Massey "International Migration and Development in Mexican Communities." Demography 33(2): Durand, Jorge, Emilio A. Parrado, and Douglas S. Massey "Migradollars and Development: A Reconsideration of the Mexican Case." International Migration Review 30(Summer): Grasmuck, Sherri and Patricia R. Pessar Between Two Islands. Dominican International Migration. Berkeley, Los Angeles, London: University of California Press. Hammel, Eugene A. L. P "Comparing Household Structure Over Time and Between Cultures." Comparative Studies in Society and History 16: Hoddinott, John " Modeling Remittance Flows in Kenya." Journal of African Economies 1(2): International Monetary Fund (IMF) Balance of Payment Statistics Yearbook. Washington, D.C.: International Monetary Fund (IMF). 12

13 Itzigsohn, Jose "Migrant Remittances, Labor Markets, and Household Strategies: A Comparative Analysis of Low-Income Household Strategies in the Caribbean Basin." Social Forces 74(2): Massey, Douglas S., Jorge Durand, and Nolan J. Malone Beyond Smoke and Mirrors. Mexican Immigration in an Era of Economic Integration. New York: Russell Sage Foundation. Massey, Douglas S. and Emilio Parrado "International Migration and Business Formation in Mexico." Social Sciences Quarterly 79(1): "Migradollars: The Remittances and Savings of Mexican Migrants to the United States." Population Research and Policy Review 13:3-30. Multilateral Investment Fund "Remittances to Latin America and the Caribbean." Washington, D.C.: Interamerican Development Bank. Pessar, Patricia R "Dominican International Migration: The Role of Households and Social Networks." Pp in In Search of a Better Life. Perspectives on Migration From the Caribbean, editor Ransford W. Palmer. New York, Westport (Connecticut), London: Praeger. Sana, Mariano "Family Structure, Market Development, and Remittances." Paper presented at the Annual Meeting of the Population Association of America (PAA), Atlanta, GA, May "International Monetary Transfers: Three Essays on Migrant Decision-Making." Doctoral Dissertation. Graduate Group in Demography, University of Pennsylvania. Stark, Oded The Migration of Labor. Cambridge, Massachusetts: Basil Blackwell. Stark, Oded and Robert E. B. Lucas "Migration, Remittances, and the Family." Economic Development and Cultural Change 36(3):

14 14

15 15

16 16

17 17

18 18

19 19

20 20

21 21