A Spatial Analysis of the Effects of International Migration on Infant Mortality in Mexico

Transcription

1 A Spatial Analysis of the Effects of International Migration on Infant Mortality in Mexico Miguel Flores T. S. Sunil Esteban Picazzo Palencia Dora Elia Cortez Hernandez Using spatial regression techniques, this article examined the role oj international migration on injant mortality in Mexican municipalities. The findings support the argument that through exposure to modern values and practices, migrant sending communities may gain health knowledge that may in turn lead to improvement in infant health outcomes. Keywords: GIS, infant mortality, international migration, Mexico, remittances, spatial econometrics Mexico is considered the second-largest recipient of financial remittances in the world (Newland & Patrick, 2Ü04). An early estimate of the amount of money remitted from the United States to Mexico in the early 1990s represents approximately two bijjion dollars (Massey, Goldring, & Durand, 1994). Over the last decade, remittances grew at an annual average rate of 15.6 percent, reaching $23.1 billion by 2006 (Cañas, Coronado, & Orrenius, 2007). The magnitude of total money remitted represents almost 80 percent of Mexico's oil exports and is roughly equivalent to 2.5 percent of the country's gross domestic product Miguel Flores. PhD. is research associate at the Institute for Demographic and Socioeconomic Research and adj unct assistant professor in the Department of Demography at the University of Texas at San Antonio. T. S. Sunil. PhD. is associate professor of Sociology and director of Institute for Health Disparities Research at the University of Texas at San Antonio. Estehan Picazzo Palencia. PhD. is research professor at the Institute of Social Research of Universidad Autónoma de Nuevo León, research associate at the Center for Research and Development on Health Sciences, and professor of Economics at the School of Puhlic Accounting and Administration at the Universidad Autónoma de Nuevo León. Dora Elia Cortez Hernandez. MD, MPH. is director of the Public Health Unit at the Center for Research and Development on Health Sciences and professor at the Medical School and School of Public Health at the Universidad Autónoma de Nuevo León. Direct correspondence to Miguel Flores. Institute for Demographic and Socioeconomic Research. University of Texas at San Antonio. 501 West Durango. San Antonio, TX 782Ü7-4415; miguel.flores utsa.edu inlernauonai ConsorUum for Social Development

2 2 Social Development Issues 34 (2) 2012 (International Monetary Fund, 2006).The income received by the families in Mexican communities from former migrants in the United States may then result in improvements and sustaining living standards and thus help to promote local economies (Cañas et al., 2007). The impact of financial remittances on health outcomes in Mexico has been widely studied by many scholars (Frank & Hummer, 2002; Hamilton, Villarreal, & Hummer, 2008; Kanaiaupuni & Donato, 1999; Sunil & Flores, in press). The general consensus among these studies is that migrant households tend to have better overall health outcomes compared to nonmigrant households. This advantage among migrant households was often attributed to the financial remittances individual households receive from the United States (or elsewhere). Another line of argument in this direction is the importance of social remittance on health outcomes. Hildebrandt and McKenzie (2005) posited the idea that households may gain health knowledge from the migration experience to the United States. This in turn may generate spillovers that can be transmitted from one migrant family to another, and eventually, this process may also benefit nonmigrant households. Several studies have shown that migrants may learn basic health knowledge about nutrition, diet, exercise, and family planning while living abroad and then pass this information to their family members (Frank, 2005; Kanaiaupuni & Donato, 1999; Levitt, 1998; Massey & Parrado, 1997; Menjivar, 2002). In addition, Lindstrom and Saucedo (2002) showed the effects of migration on women's fertility in Mexico. The basic assumption here was that the diffusion of innovative healthcare practices will slowly become the norm with significant impact on health outcomes and demographic behaviors. Both infant mortality and international migration are characterized by substantial variations across regions, states, municipalities, and rural-urban communities. More important, infant mortality rates are not randomly distributed across the geography and are strongly correlated to the level of socioeconomic development (Alvarez, Lara, & Harlow, 2009).The survival chances of infants are likely to show spatial variations and clustering within regions that may differ according to where they were born. Furthermore, as argued by Woodruff and Zenteno (2001) migrants have historically been selected disproportionately from certain regions of the country, particularly in certain central western states including Aguascalientes, Colima, Durango, Guanajuato, Jalisco, Michoacan, Nayarit, San Luis Potosi, and Zacatecas (Hildebrandt & McKenzie, 2005). This paper addresses the hypothesis that migrant sending communities tend to have better health outcomes and innovative health care practices due to diffusion of modern values. Hence, it examines the role of international migration on infant mortality in Mexico. In particular, spatial analysis methodology considering municipalities as unit of analysis aiming to model the extant spatial patterns of infant mortality rates as well as the variations in migration to the United States within the country was used. In doing this, this article attempts to address the spatial dependence of infant mortality in municipalities by avoiding

3 Flores, Sunil, Picazzo Palencia, and Cortez Hernandez 3 some econometric problems, such as inefficient estimators, commonly found in estimating with spatially correlated cross-sectional data. Data Cross-sectional information for the 2,443 Mexican municipalities existing in the year 2000 was gathered. The dependent variable 2005 infant mortality rates estimates of the National Council of Population (Consejo Nacional de Población [CONAPO], 2007) was used. CONAPO estimates involved the use of indirect methods based on the proportion of children's deaths, history of pregnancies, and live births trends using different information sources such as National Population and Housing Census (1990,199 5,2000, and 200 5), and the National Survey of Demographic Dynamic (ENADID) To ameliorate issues of endogeneity data for the year 2000 was used for the rest of the independent variables. Different measures of international migration patterns were considered separately to examine the associated impact on infant mortality rates. The percentage of households receiving remittances is included in the regression models. As discussed above, remittances is the most important source of external funding with direct implications on households' absolute and relative income levels. Another variable included in the model is proportion of households with returning migrants from the United States. Finally the regressions also include the index of migration intensity (CONAPO, 2002) to the United States. This index based on a principal component technique, includes four variables: (1 ) the percentage of households that receive remittances, (2) the percentage of households with migrants in the United States in the last five years. (3) the percentage of households with the US circular migrants in the last five years, and (4) the percentage of households with returning migrants. The estimated index ranks each municipality into five levels of migration intensity: very high, high, medium, low, and very low. The regressions also include a set of control variables. For example, to control for the level of education in the municipality the average years of schooling among population aged fifteen years and older. Education has a direct positive social impact on health (Feinstein, Sabates, Anderson, Sorhaindo, & Hammond, 2006; Grossman & Kaestner, 1997). In particular, parental education levels have important implications for child survival. For example, father's education may infiuence child's survival through its direct implications on father's productivity that may result in higher wages and higher level of family consumption reaching higher socioeconomic status (Sastry 1996). The mother's higher level of education increases her involvement in child care, and she performs more effective health care practices (Streatfield & Diamond, 1990). A negative and significant correlation between average schooling in municipalities and infant mortality rates is expected. Community level variation in access to resources such as health care is often cited as a factor explaining why mortality in urban areas is generally lower than

4 4 Social Development Issues 34 (2) 2012 Table 1 Descriptive Statistics of Infant Mortality Rates and Selected Socioeconomic Variables in Mexican Municipalities (N= 2,443) Variable M SD Minimum Maximum Infant mortality rate/per 1,000 % households recipient of remittances % households with returning migrants % households with circular migrants Very high migration intensity index High migration intensity index Medium migration intensity index Low migration intensity index Very low migration intensity index Average schooling % rural population % female head of family % population with social security % population without water connection % indigenous population rural areas (Baker, 1999). To account for such differentials in health eare access, the regressions include the percentage of population with health coverage and the percentage of population living in rural areas. The first is measured as the fraction of the population affiliated with some social security institution. The second corresponds to the proportion of population living in rural municipalities with less than 2,50(i inhabitants. A positive relationship between health coverage and infant mortality rates is expected. Furthermore, rural communities may experience higher mortality rates as a consequence of lack of access to medical facilities. Within the communities, the quality of housing can also have a strong influence on a chfld's survival. For example, access to water and lack of sanitation facilities expose children's health to environmental contaminants (Mosley & Chen, 1984). In converse, higher quality sanitary facilities and improved water supply are directly associated with lower mortality (Agha, 2000). To control for housing characteristics in each location, the analysis also considers the proportion of population without water connection. Finally, the fraction of indigenous population is added to control for segments of the population that have historically experienced lack of access to health eare and schooling facilities and in general lagged economic development. The descriptive statistics of all of the variables included in the analysis are displayed in table 1, Methodology The use of spatial analysis techniques is significant given that the dependent variable, infant mortality rates, is not randomly distributed across municipali-

5 Flores, Sunil, Picazzo Palencia, and Cortez Hernandez 5 ties. Hence, it is necessary to formally investigate whether such distribution of the variables is spatially autocorrelated creating significant spatial clusters within the Mexican geography If this is the case, such spatial autocorrelation must be formally addressed when modeling factors associated with our dependent variable. The first step of the analysis consists in determining whether the dependent variable is a stochastic phenomenon or contrary, whether infant mortality rates follow particular spatial patterns. The statistic Moran's I is widely employed for testing the presence of spatial dependence in observations. This provides a global statistic for assessing the degree of spatial autocorrelation between observations as a function of the distance separating them. In the present study, the geographic information system from the software ArcGIS 1 ( ) was used to calculate spatial autocorrelation, both globally and locally In performing this test, a Moran's I value of was obtained, with a positive 2 score and significant p value. This provides evidence to reject the null hypothesis that the data are randomly distributed. In other words, it indicates that the residuals are driven by positive and significant spatial autocorrelation related to infant mortality rates across municipalities in Mexico. Nonetheless, global statistics provide only a limited set of spatial association measurements by not considering the case of local variations in spatial autocorrelation. The local Moran's I or as known, local indicator of spatial association (LISA), allows the decomposition of the global indicator that in turn helps to explore the extent of significant clustering with values similar in magnitude around a particular observation (Lloyd, 2007). In search of local variations of spatial autocorrelation, or spatial clusters, the respective LISA statistic was performed. This facilitates the identification of municipalities with concentration of high values, concentration of low values, and spatial outliers. Figure 1 shows the distribution of significant spatial clusters of infant mortality rates and for the variables measuring international migration to the United States. The map shows the distribution of four different types of spatial clustering; (a) high-high (HH); a municipality with high value and its neighbors as well; (b) high-low (HL); a municipality with a high outlier and its neighbors low values; (c) low-high (LH); a municipality with a low outlier and its neighbors high values; (d) low-low (LL); a municipality with a low value and its neighbors as well. As shown in figure 1, spatial clusters of high values (HH) of infant mortality rates are exhibited mainly in the southern and southeast of Mexico, covering portions of the states of Guerrero, Chiapas, Oaxaca, and Veracruz; as well as the indigenous Tarahumara region of Chihuahua. In contrast, spatial clusters of low values (LL) are depicted particularly covering some portions of the northern border regions and coincidently central western Mexico, the region with high international migration tradition to the United States. The next step is to estimate spatial econometric models to account for spatial dependence on the relationship between infant mortality rates and

6 Social Development Issues 34 (2) 2012 Figure 1 Spatial Clusters of infant Mortality, Mexican Municipalities, LL = low-low (a municipality with a iow value and its neighbors as weii); LiH = low-high (a municipality with a low outiier and its neighbors high values); HL = high-low (a municipality with a high outiier and its neighbors iow values); HH = high-high (a municipality with a high value and its neighbors as well). international migration. A general regression equation modeling spatial dependence can he represented as follows: e = XWe + ll ~ N(0, o^ where Y represents the vector of dependent variables, X is a matrix of independent variables, e is a vector of random residuals, and W is the spatial weight matrix. These spatial weight matrices represent the "degree of potential interaction" between neighboring locations (Anselin, 1988). The parameters p and X are scalar spatial parameters measuring the degree and type of spatial dependence. When p ^ 0 and A, = 0, the result is what is called a spatial lag model or spatial autoregressive model: (2)

7 Plores, Sunil, Ficazzo Falencia, and Cortez Hernandez 7 where p reflects the spatial dependence or the average influence that the neighboring observations have on one specific observation. In this case, estimating equation 1 by ordinary least squares (OLS) will lead to biased and inconsistent estimators of the endogenous dependence variable. The other type of spatial dependence arises when p = 0 and X ^ 0. In this case the regression equation takes the form of a spatial error autocorrelation model; y=x^ + t, (3) e = XW where X is the scalar parameter measuring the degree of spatial dependency in the residuals. In this case, OLS estimation of spatial error models leads to inefficient estimated coefficients. One theoretical difference between equations 2 and 3 is that in the autoregressive model it is typical that all error terms are only correlated with their immediate neighbors, as specified in weigh matrix W (Fotheringham, Brunsdon, & Charlton, 2000, p. 170). A crucial aspect of this analysis is to discuss the differences when modeling and interpreting the results obtained from these two types of spatial regression. On the one hand, the spatial lag model assesses the possibility of "spiflovers" effects among municipalities that directly influence a particular outcome of their neighbors. In terms of the variable of interest, this model distinguishes whether mortality rates in a particular municipality are indeed influenced by the level of its neighbors. On the other hand, the spatial error model suggests the possible existence of omitted variables in the right hand side of the regressions (Voss, Long, Hammer, & Friedman, 2006). This means that infant mortality rates may be related to additional factors other than those included in the model and that these factors are related through the residuals among municipalities. As discussed above, given the presence of autocorrelation in the error term, OLS would lead to inefficient estimators. However, maximum likelihood estimation (MLE) gives consistent and efficient parameters (Anselin, 198 8). All the variables in the model were transformed by applying logarithms, except to the index of migration intensity. The estimated coefficients therefore indicate changes in infant mortality rates related to percentage changes in the dependent variables. Results Tables 2 to 4 show the results from the estimation of OLS and MLE regressions. Although OLS estimates are biased, we provided both OLS and MLE estimates for the purpose of understanding the effects of autocorrelation. Each table contains the estimates of the scalar parameter for the spatial lag model (p) and the spatial error model (k). These tables present different statistical tests such as spatial autocorrelation in the residuals (Moran's I), heteroskedasticity

8 8 Social Development Issues 34 (2) 2012 in the residuals (Breusch-Pagan), and multicolinearity in the regressors (multicolinearity number). The first task in interpreting the results is to identify what type of spatial model, either spatial lag or spatial error, is more appropriate in defining the relationship between infant mortality and remittances. Following Anselin's (200 5, p. 198) decision rule, significant and higher values of robust Lagrange multiplier (LM) test, favor the spatial error model over the spatial lag model in all estimated models. Hence, the interpretation of the results is based on the results of spatial error models. Migration Variables Table 2 shows the estimated coefficients from regressing infant mortality rates against the percentage of household recipient of remittances including as well the set of control variables previously described. To control for a possible nonlinear relationship between international migration and infant mortality as discussed in Kanaiaupuni and Donato (1999). a variable with the square of remittances is added into the model. As expected, higher levels of household recipient of remittances is significantly and negatively associated with infant mortality rates. The estimated coefficient indicates that for a 10% increase in percentage of households receiving remittances there is a reduction in approximately 1.4 percent in infant mortality rates. It is interesting to note the effects of the estimates associated to remittances once accounted for spatial dependence. In particular, the estimated coefficient of remittances is greater in magnitude in the spatial error model than in the OLS model. This means that the effects of remittances on infant mortality rates are higher once it is controlled for spatial autocorrelation. Furthermore, there is a significant but modest effect related to nonlinear effects on infant mortality, as measured by the remittances square term. Table 3 shows the effects of returning Mexican migrants from the United States on infant mortality in the presence of addition to all the control variables. The results support the expectations; higher percentage of returning migrants in Mexican municipalities is negatively associated with infant mortality. The results suggest that by any 10 percent increase in the percentage of households with returning migrants this can be associated with a reduction on mortality rates of 5.5 percent. The results related to the level of international migration intensity of Mexican municipalities are presented in table 4. Four dummy variables were included denoting the level of international migration, leaving medium migration levels as the reference group. Although very low and low migration intensity show statistical significance levels in the OLS regression, only low levels of international migration show significance and negative effects on infant mortality rates in the spatial error model.

9 F/ores, Sunil, Picazzo Palencia, and Cortez Hernandez Table 2 Ordinary Least Squares (OLS) and Spatial Regressions of Infant Mortality Rates and Remittances Variable Migration % households receiving remittances Square % households receiving remittances Control % rural population P X % female headed households % population with health coverage % population without water connection Average schooling % indigenous population Constant R' Multicolinearity number Jarque-Bera Breusch-Pagan Moran's 1 Robust Lagrange multiplier (lag model) Robust Lagrange multipiier (error model) Likelihood ratio test OLS *** (469) 03** (015) 88*** (038) *** (273) -167* (098) 0.111*** (067) *** (0.1097) -01 (048) 36.67*** (1.9659) * 2,08*** 321.7*** 19.81*** 21.10*** 284.9*** Spatial lag -66* (386) 02 (012) 61*** (033) *** (225) 09 (080) 76*** (058) *** (039) -01 (039) *** (1.7063) 0.571*** (1774) *** 784.4*** Spatial error *** (454) 03*** (013) 74*** (036) *** (258) -09* (077) 0.105*** (069) *** (877) 07 (038) *** (3.7570) 0.704*** (181) *** 936.6*** Note. Standard errors are reported in parentheses. *p<.1. **p<.05. ***p<.01. Control Variables In general, the results related to each of the control variables show the expected effects across models. For example, higher percentage of rural population is positively associated with infant mortality rates, whereas "female headed households" is negatively associated with mortality rates. In examining the effects of the percentage of population with health coverage, there was a negative but limited and marginally significant (at 10%) effect on mortality rates. The

10 1 o Social Development Issues 34(2)2012 Table 3 Ordinary Least Squares (OLS) and Spatial Regressions of Infant Mortality Rates and Households with Returning Migrants from the United States Variable OLS Spatial lag Spatial error Migration % households with returning migrants Square % households with returning migrants Control % rural population % female headed households % population with health coverage % population without water connection Average schooling % indigenous population Constant p X R^ Multicolinearity number Jarque-Bera Breusch-Pagan Moran's 1 Robust Lagrange multiplier (lag model) Robust Lagrange multiplier (error model) Likelihood ratio test (0.1579) 13 (178) 879*** (879) *** (267) -15* (098) 0.111*** (068) *** (0.1097) -01 (048) 36,061** (1.9530) *** 21.4*** 286.2*** (0.1301) 09 (146) 62*** (032) *** (219) 09 (080) 76*** (058) *** (905) -01 (039) *** (1.6976) 0.571*** (177) *** 788.6*** *** (0.1561) 38*** (157) 73*** (036) *** (253) -09* (077) 0.104*** (069) *** (876) 01 (038) *** (3.7665) 0.706*** *** 944.9*** Note Standard errors are reported in parentheses. *p<.1. **p<.05. * * * p <.01. percentage of population without water connection, a proxy for sanitary conditions, is found to be positive and highly significant across models. Finally, higher education levels in Mexican municipalities are signfficant and negatively related to infant mortality This is indicated as the coefficient associated with average schooling is the highest estimated coefficient among the control variables. Two points need to be emphasized on the results presented above. First, in the spatial error model the relationship between remittances and returning migration is influenced by the "unobservable" predictors that were not included in the

11 Flores, Sunil, Picazzo Palencia, and Cortez Hernandez 11 models. Second, focusing on the spatial error models, the variahle of returning migration to the United States shows the highest estimated effect on infant mortality rates. Some researchers attribute those "unohservables" to certain norms and attitudes that cannot he measured but may play an important role in the variable of interest (Chamarhagwala, 2009). Table 4 Ordinary Least Squares (OLS) and Spatial Regressions of infant Mortality Rates and Remittances the Index of Migration Intensity Variable Migration Very high migration intensity High migration intensity Low migration intensity Very low migration intensity Control % rural population % female headed households % population with health coverage % population without water connection OLS -0,559 (0.5812) -03 (0.4528) *** (0.3852) -72* (0.3684) 859*** (038) *** (265) -155 (097) 0.113*** (067) Spatial lag -23 (0.4785) (209) *** (0.3178) (0.3032) 60*** (032) *** (219) 08 (080) 78*** (058) Spatial error 387 (181) (0.3703) *** (0.3114) (0.3268) 71*** (036) *** (252) * (077) 0.106*** (069) P X Average schooling % indigenous population Constant Multicolinearity number Jarque-Bera Breusch-Pagan Moran's 1 Lagrange multiplier (lag model) Lagrange multiplier (error model) Likelihood ratio test *** (0.1095) 01 (048) *** (1.9191) ,257.3*** 377.4*** 87.6*** 34.1*** 240.5*** *** (906) 05 (03) *** (1.669) 0.567*** (177) *** 776.1*** *** (878) 01 (038) *** (3.7431) 0.705*** (181) *** 303.6*** 941.1*** Note. Standard errors are reported In parentheses. *p<.1. **p<.05. ***p<.01.

12 12 Social Development Issues 34(2)2012 This latter flnding may support to some extent the hypothesis that through migration to the United States, migrants may gain health knowledge and behaviors (or attitudes) that in turn positively impact infant health outcomes at the community level. However, this statement should be made with caution. Spatial error models assume that the spatial effect comes from the "errors." However, it should be noted that these are just variables that were either omitted from the regressions or those that could not be measured (Beck & Gleditsch, 2006). At this point a more conclusive analysis relating to those factors that influence such spatial effect on the relationship between infant mortality and particularly with returning migration is not available. Further research is needed and the next section discusses some venues. Discussion Infant mortality is often considered as a global welfare indicator because it is sensitive to environmental changes in a broad sense and it also reflects the living conditions not only for the newborn, but also for a much wider segment of the population (Ferrarini & Norstrom, 2010). During the past three decades, research on the impact of migration on health outcomes of Mexicans, in particular on infant health outcomes, has gained the attention of scholars in different fields of social sciences (Amuedo-Dorantes & Pozo, 2006; Durand, Massey, & Zenteno, 2001; Durand, Parrado, & Massey, 1996; Hamilton et al., 2008; Hildebrandt & McKenzie, 2005; Kanaiaupuni & Donato, 1999). This paper developed a spatial analysis methodology considering municipalities as unit of analysis to capture the extent of spatial patterns of the migration to the United States on infant mortality in Mexico. The findings support the argument that remittances may decrease the level of infant mortality. Nevertheless, this positive effect might be diminished by the lag of socioeconomic development that already exists in communities where these tend to be grouped in particular regions across the Mexican territory. In terms of the analysis, where the level of aggregation is municipalities, it is argued that accounting for those spatial patterns within the Mexican geography is crucial to provide reliable estimates when assessing the impact of international migration on the level of local and regional socioeconomic development. The findings also provide additional insights on how spatial matters and the possible mechanism through which international migration might have an effect on reducing infant mortality rates. It is also clear from the analysis that municipalities with a higher percentage of returning migrants tend to be negatively correlated with infant mortality rates. The findings may also support to some extent the hypothesis that through exposure to US practices, municipalities may gain health knowledge that may deliver better child health outcomes. This study suffers from several limitations. Due to data availability, medical care infrastructure in Mexican municipalities was not included. This is an important control variable contributing significantly to community health improvements.

13 Flores, Sunil, Picazzo Palencia, and Cortez Hernandez 13 Furthermore, the presence of heteroskedasticity in the residuals was neither eliminated nor reduced in the spatial models. An extension of this study should consider econometric techniques to deal with this issue and also consider the use of instrumental variables regression models in a spatial context as a way to control for endogeneity among the regressors. References Agha, S. (2000). The determinants of infant mortality in Pakistan. Social Science & Medicine, 51, Alvarez, G., Lara, F, & Harlow, S. (2009). Infant mortality and urban marginalization: A spatial analysis of their relationship in a medium size city in northwest Mexico. Pan American Joitrnal of Public Health, Amuedo-Dorantes, C, & Pozo, S. (2006). Remittances as insurance: Evidence from Mexican immigrants. Journal of Poptdation Economics, 19, Anselin, L. ( ). Spatial econometrics: Methods and models. Dordrecht, The Netherlands: Kluwer Academic. Anselin, L. (2005). Exploring spatial data with GeoDa: A workbook. Urbana; University of Illinois. Urbana-Champaign. Retrieved from Baker, R. (1999). Differential in child mortafity in Malawi. Social Networks Project Working Papers, 3,1-29. Beck, N., & Gleditsch, K. (2006). Space is more than geography: Using spatial econometrics in the study of political economy International Studies Quarterly, 50, Cañas, J., Coronado, R., & Orrenius, P (2007). Explaining the increase in remittances to Mexico. The Southwest Economy: Eederal Reserve Bank of Dallas, 4, 3-7. doi: RePEc:fip:feddse:y:2007:i:jul;p:3-7;n:4 Chamarbagwala, R. (2009). Social interactions, spatial dependence, and children's activities; Evidence from India. The Journal of Developing Areas, 42, Consejo Nacional de Población. (2002). Index of migration intensity Mexico- United States, México City Mexico: Author. Consejo Nacional de Población. (2007). Infant mortality estimation states and municipalities in Mexico, México City México: Author. Retrieved from metodo.pdf Durand, J.. Massey D.. & Zenteno, R. (2001). Mexican immigration to the United States: Continuities and change. Latin American Research Review, 36, Durand, ],, Parrado, E., & Massey D. (1996). Migradollars and development: A reconsideration of the Mexican case. International Migration Review, 30,

14 14 Social Development Issues 34 (2) 2012 Feinstein, L., Sabates, R,, Anderson, T., Sorhaindo, A,, & Hammond, C. (2006). What are the effects of education on health. In R. Desjardins & T, SehuUer (Eds.), Measuring the ejjects oj education on health and civic engagements: Proceedings oj the Copenhagen symposium. Paris: Organization for Economic Cooperation and Development. Ferrarini, T,, & Norstrom, T. (2010). Family policy, economic development and infant mortality: A longitudinal comparative analysis. International JournaloJSocialWelJare, 19, Fotheringham, S,, Brunsdon, C, & Charlton, M. (2000). Quantitative geography perspectives on spatial data analysis. London: Sage. Frank, R. (2005). International migration and infant health in Mexico. Journal oj Immigrant Health, 7, Frank, R., & Hummer, R. (2002). The other side of the paradox: The risk of low birth weight among infants of migrants and nonmigrant households within Mexico, International Migration RevieviJ, 36, , Grossman, M,, & Kaestner, R, (1997), Effects of education on health. In J. R, Berhman & N, Stacey (Eds.), The social benefits oj education (pp, ). Ann Arbor: University of Michigan Press. Hamilton, E., Villarreal, A,, & Hummer, R. (2008), Mother's, household, and community: US, migration experience and infant mortality in rural and urban Mexico, Population Research and Policy Review, 28, Hildebrandt, N., & McKenzie, D, (200 5). The ejjects oj nngration on child health in Mexico (World Bank Policy Research Working Paper No. 3573). Washington, DC: World Bank. International Monetary Fund. (2006). Mexico: Selected issues (Country Rep, No. 06/351). Washington, DC: Author. Kanaiaupuni, S., & Donato, K, (1999). MigradoUars and mortality: The effects of migration and infant survival in Mexico. Demography, 36, , Levitt, P, (1988). Social remittances: Migration driven, local-level forms of cultural diffusion. International Migration Review, 32, , Lindstrom, D, P, & Saucedo, S, (2002), The short and long term effects of U,S, migration experience on Mexican women's fertility. Social Forces, 80, Lloyd, C. (2007), Local models Jor spatial analysis. Boea Raton, FL: CRC Press, Massey, D., Goldring, L,, & Durand, J, (1994). Continuities in transnational migration: An analysis of nineteen Mexican communities. American Journal oj Sociology, 99, Massey, D., & Parrado, E, (1997). International migration and business formation in Mexico. Social Science Quarterly, 79,1-20. Menjivar, C, (2002). The ties that heal: Guatemalan immigrant women's network and medical treatment. International Migration Review, 36, , Mosley, W, & Chen, L, (1984), Child survival: Strategies Jor research. New York: Cambridge University Press,

15 Flores, Sunil, Picazzo Palencia, and Cortez Hernandez 15 National Population and Housing Census. (1990). Aguascalientes, Mexico; Instituto Nacional de Estadística y Geografía. Retrieved from http;//www. inegi. org.mx /est / contenidos / proyectos / ccpv / cp v 19 9 O / default, aspx National Population and Housing Census. (1995). Aguascalientes, Mexico; Instituto Nacional de Estadistica y Geografia. Retrieved from http;// National Population andhousing Census. (2000). Aguascalientes, Mexico; Instituto Nacional de Estadistica y Geografia. Retrieved from http;// National Population andhousing Census. (2005). Aguascalientes, Mexico; Instituto Nacional de Estadistica y Geografia. Retrieved from http;//www inegi.org.mx/est/contenidos/proyectos/ccpv/cpv2005/default.aspx National Survey of Demographic Dynamic (ENADID). (2006). México City México; Consejo Nacional de Población. Retrieved from http;// Newland, K., & Patrick. E. (2004, July). Beyond remittances: The role of diaspora in poverty reduction in their countries of origin (Migration Policy Institute report). Washington, DC; Department for International Development. Sastry, N. (1996). Community characteristics, individual and household attributes, and child survival in Brazil. Demography, 33, Streatfield, S., & Diamond, I. ( 1990). Maternal education and child immunization. Demography, 27, Sunil, T., & Flores, M. (in press). New evidence on the effects of international migration on the risk of low birthweight in Mexico. Maternal & Child Nutrition. Voss, P, Long, D., Hammer, R., & Friedman, S. (2006) County child poverty rates in the U.S.; A spatial regression approach. Population Research and Policy Review, 25, Woodruff, C, & Zenteno, R. (2001). Remittances and microenterprises in Mexico (Working paper). University of California at San Diego, Graduate School of International Relations and Pacific Studies.

16 Copyright of Social Development Issues is the property of Lyceum Books, Inc. and its content may not be copied or ed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or articles for individual use.