, Volume 48, Issue 2, pp 425–436

Simulating the Effects of Acculturation and Return Migration on the Maternal and Infant Health of Mexican Immigrants in the United States: A Research Note


    • Department of SociologyUniversity of Nebraska–Lincoln

DOI: 10.1007/s13524-011-0017-4

Cite this article as:
Ceballos, M. Demography (2011) 48: 425. doi:10.1007/s13524-011-0017-4


A significant body of research on minority health shows that although Latino immigrants experience unexpectedly favorable outcomes in maternal and infant health, this advantage deteriorates with increased time of residence in the United States. This study evaluates the underlying assumptions of two competing hypotheses that explain this paradox. The first hypothesis attributes this deterioration to possible negative effects of acculturation and behavioral adjustments made by immigrants while living in the United States, and the second hypothesis attributes this deterioration to the mechanism of selective return migration. Hypothetical probabilistic models are simulated for assessing the relationship between duration and birth outcomes based on the assumptions of these two hypotheses. The results are compared with the empirical research on the maternal and infant health of first-generation, Mexican-origin immigrant women in the United States. The analysis provides evidence that a curvilinear pattern of duration and birth outcomes can be explained by the joint effects of both acculturation and selective return migration in which the former affects health status over the longer durations, and the latter affects health status at shorter durations.


AcculturationSelectionReturn migrationImmigrationMaternal health


Minority health research shows that although Latinos experience unexpectedly healthy birth outcomes relative to other racial and ethnic groups, this health advantage appears to deteriorate with each subsequent generation of residence in the United States (Acevedo-Garcia et al. 2007; de la Rosa 2002; Ruiz et al. 2006). The principal explanation for this phenomenon is that the process of acculturation negatively affects behaviors that influence health outcomes for Latinos (Callister and Birkhead 2002; Collins and David 2004; Harley and Eskenazi 2006; Lara et al. 2005). Several studies have analyzed this relationship at an intragenerational level by examining the association between birth outcomes and length of residence among first-generation Mexican immigrants (Balcazar and Krull 1999; Ceballos and Palloni 2010; Guendelman and English 1995; Zambrana et al. 1997). As a consequence, a competing hypothesis has emerged: the apparent decline of immigrant health status is the result of the selective return migration of healthy migrants. This study assesses the viability of the acculturation and selective return migration hypotheses for explaining the intragenerational change in birth outcomes among Latino immigrants living in the United States.


According to the acculturation hypothesis, Latino immigrants enter the United States with a set of maternal characteristics, behaviors, and social relations—such as healthy diets, low levels of smoking and alcohol consumption, and strong familial ties—that facilitate favorable birth outcomes. Increased duration in the United States contributes to acculturative change, including less-healthy diets, increased smoking and alcohol consumption, and weaker familial ties that lead to increased negative maternal and infant health outcomes (Callister and Birkhead 2002; de la Rosa 2002; Lara et al. 2005). Most literature on birth outcomes shows a negative acculturation effect, but a few studies have found an opposite pattern (Balcazar and Krull 1999; Wolff and Portis 1996). The positive acculturation results are more consistent with traditional sociological theory on assimilation (Alba and Nee 1997; Rumbaut 1997), which would predict a positive association between immigrant health status and duration as a consequence of an improvement in socioeconomic status between, and possibly within, generations.

Although most research on Latino maternal and infant health examines intergenerational (versus intragenerational) decline, to my knowledge, only four studies have examined birth outcomes among first-generation Mexican immigrants by using the length of residence as a proxy for acculturation. Two studies of Mexican immigrants living in California (Guendelman and English 1995; Zambrana et al. 1997) found a negative effect of duration on birth outcomes, although only the Zambrana et al. (1997) study found it to be significant. Both studies make the case for including duration as a component of acculturation when studying immigrant birth outcomes. A third study found a positive and significant effect of length of residence on birth outcomes for Mexican immigrants living in Arizona (Balcazar and Krull 1999). The authors reasoned that duration and acculturation measure different aspects of sociocultural factors that in turn affect birth outcomes. These studies suggest either that duration may be a proxy for acculturation but with ambiguous effects on health status, or that it measures some other process.

A fourth study of Mexican immigrant women living in the Midwest, together with a replication analysis with national data (Ceballos and Palloni 2010), found a curvilinear relationship between birth outcomes and duration: mothers living in the United States for 4 to 12 years were significantly more likely to have better birth outcomes than those with less than 4 or more than 12 years of residence. On the one hand, the unique outcomes of this study can be explained in part by the use of disaggregated duration data in a nonlinear analysis. On the other hand, the authors inferred the presence of a health-selective return migration mechanism: the long-run declines in birth outcomes may be a result of good health–selective return migration rather than from acculturative changes.

When studying immigrant health, it is important to consider the effect of selection bias because it has been shown to produce a significant effect on population health status (Jasso et al. 2004; Palloni and Ewbank 2004; Palloni and Morenoff 2001), although one study suggests otherwise (Akresh and Frank 2008). Previous research has not addressed the possible effect of selective return migration as an explanation for the decline in birth outcomes among Latino immigrants because the bulk of this work focuses on differences in health status intergenerationally. Arguably, for such research, the issue of return migration is of rather muted importance. However, when one studies health status changes within a single generation, the effect of return migration cannot be ignored, particularly because the streams of migrants returning to Mexico each year can be significant. The cumulative rate of return for Mexicans living in the United States has been estimated to be as high as 50% during the first two years and 70% after 10 years (Massey et al. 2002; Massey and Singer 1995; Reyes 2001). Thus, the effect of selective return migration on the distribution of immigrant health status may be nontrivial and therefore necessitates further examination.

Although most studies on health-selective return migration have examined its effect on adult Latino health status (Abraido-Lanza et al. 1999; Palloni and Arias 2004), two recent studies (Ceballos and Palloni 2010; Hummer et al. 2007) considered the effects on maternal and infant health; of these two studies, only Ceballos and Palloni focused on the relationship of birth outcomes and length of residence. Unlike previous studies, Ceballos and Palloni contended that young childbearing-age women may have different motivations for return migrating than elderly adults and thus hypothesized a positive health selection. They found that pregnant mothers of poor health may be less likely to return to Mexico prior to giving birth if they believe they have access to better healthcare in the United States.

The present study examines the effect of acculturation and selective return migration on the relationship between Latino immigrant birth outcomes and length of residence in the United States. Because of the methodological and empirical challenges to fully exploring a self-selection mechanism, I use a simulation analysis to assess the key assumptions of the acculturation and return migration hypotheses and to adjudicate the patterns resulting from these mechanisms and those found in empirical studies on intragenerational Latino immigrant birth outcomes. My overall goal is to determine whether the association between birth outcomes and duration is due to either acculturation or return migration, or to a combination of both.

The Models and Simulation Results

This analysis generates three simulations representing the individual and joint effects of acculturation and return migration. Each model will comprise the probability of favorable birth outcomes, assumed to be an endogenous, time-invariant average of population birth outcome probabilities; and the parameters representing the acculturation and return migration mechanisms that are assumed to be exogenous and time dependent. The parameters are based on the range of estimates available from empirical studies. These models will simulate the distribution of birth outcomes given the different assumptions: negative and positive acculturation, and good and poor health selection. The results will be evaluated on their replication of empirical analysis.

Because the expected similarity in outcomes for acculturation and selective return migration are based on cross-sectional data, the simulation model takes the same format. The simulation assumes a representative sample of Latinas who give birth after entering the United States such that each respondent enters the country at time zero and gives birth at time t some period after entry. It is also assumed that if the probability of a favorable birth outcome for any woman is represented by P and the probability of an unfavorable birth outcome is 1 – P at any time t, then the proportion of the sample with favorable outcomes will equal the constant P.1 This is the baseline case in which all relevant factors are controlled for in the model and there is no effect of acculturation or return migration.

The Acculturation Case

For the case of negative acculturation, the proportion of favorable birth outcomes declines over time as the average probability of favorable outcomes deteriorates due to immigrant women engaging in behaviors detrimental to their health. This decline is represented by the function γi, where 0 ≤ γi ≤ 1 for time t = i. For positive acculturation, the function is inversed, representing the opposite effect, where γi ≥ 1. Therefore, the proportions of favorable and unfavorable birth outcomes at time t = i equals γiP and 1 – γiP, respectively.2

I use the rate of Latino low birth weight as the measure of negative birth outcomes and its change over the first and second generations as a proxy for the acculturation parameter (γ). The rate of low birth weight is taken from empirical studies on Mexican-origin women, which show that the average rate ranges from 3.7% among first-generation immigrants to 7.1% among second- and later-generation immigrants (Acevedo-Garcia et al. 2007; English et al. 1997; Martin et al. 2006; Scribner and Dwyer 1989; Ventura and Taffel 1985; Williams et al. 1986). For the negative acculturation simulation, γiP is assigned a probability of favorable outcomes ranging from .96 to .94 based on the aforementioned rates from the empirical studies, assuming that the highest rate of low birth weight for the first generation approaches that of the second generation. For the positive acculturation simulation, the inverse is assumed. The parameter is assumed to take the shape of a logistic function as acculturation is expected to increase or decrease monotonically.3 The purpose of this shape is to suggest that the effect of acculturation is gradual during the short term, greatest over the medium term, and slight over the long term as it approaches the level of acculturation similar to the second generation.

Figure 1 displays the simulated cases representing both assumptions of positive and negative acculturation. The figure illustrates an expected positive and negative relationship based on the respective assumptions about the effect of acculturation on birth outcomes. It also shows that the resulting pattern of this relationship is entirely described by the shape of γ.
Fig. 1

Simulation of the positive and negative acculturation hypotheses

The Selective Return Migration Case

In the return migration case, any change in the proportion of favorable birth outcomes will be due to changes in the health status distribution of the immigrant population rather than to changes at the individual level. The simulation model requires two time-dependent parameters: the proportion of return migrants with favorable outcomes, λt, and the rate of return migration, μt, for time t = i. For this case, the proportion of favorable birth outcomes at time t = i equals the proportion of favorable birth outcomes among all immigrants from the previous period less the product of the proportion of favorable outcomes among return migrants (λi) and the rate of return migration (μi).4 Likewise, the proportion of cases with unfavorable outcomes equals the proportion of unfavorable outcomes from time t = i – 1 less the product μi(1 – λi).

The parameter μ is assigned values derived from two empirical studies of Mexican immigration that estimate the rate of return migration: one uses a survey of Mexican immigrants to document the number of trips made to the United States between 1985 and 2000 (Reyes 2001), and the other uses the Current Population Survey (CPS) to match the resident status of immigrants living in the United States for two years between 1996 and 2003 (Van Hook et al. 2006). These studies improve upon the estimation of the return migration for recently arrived and undocumented immigrant population (Hogan 1993; Robinson et al. 1993; Van Hook et al. 2006) as compared with previous approaches using indirect methods based on data from the Immigration and Naturalization Service and the U.S. census (Ahmed 1994; Warren and Peck 1980).

The estimates of return migration for these two studies vary widely, from a high of 33% (Reyes 2001) to a low of 7% (Van Hook et al. 2006). The higher rates from the Reyes (2001) study owe partly to a more accurate reporting of return migration behavior by migrants with an uncertain or tenuous legal status obtained from survey data collected in Mexico that contain life histories documenting the number of return trips from the United States.

The simulation of the return migration rate over the migrant’s length of residence, based on the Reyes’ study, approximates a negative exponential function, with the highest rate of return occurring during the first year of residence, rapidly declining with each subsequent year, and leveling off by the 10th year. Three distributions of these rates with different initial levels are simulated for this analysis, including the following rates: .33, .28, and .07. The two highest rates are based on Reyes’ (2001) estimates of documented and undocumented migrants, respectively, and the lowest rate is based on the Van Hook et al. (2006) study.5

In this analysis, the proportion of favorable outcomes of the return migrants (λ) is assigned four constant values6 representing levels of health that are higher and lower among return migrants than among those who stay behind, based on empirical studies of Mexican immigrant birth outcomes (Acevedo-Garcia et al. 2007; English et al. 1997; Martin et al. 2006; Scribner and Dwyer 1989; Ventura and Taffel 1985; Williams et al. 1986). For the purpose of comparison, the value of P is assigned the same initial values as the acculturation functions: .96 for the negative acculturation simulations and .94 for the positive acculturation simulations. Figure 2 displays the results of these simulations for the different levels of μ and λ.
Fig. 2

Simulation of the selective return migration hypothesis with a constant λ

These results show that the level of μ determines the concave shape of the relationship between health status and duration. The differences between the levels of μ for a given level of λ changes the level of the curve, but it has little effect on the change in overall health status over the long term. On the other hand, λ determines both the direction of the relationship and the level of health status. As expected, when λ is greater than P, there is a negative association due to a good health–selective return migration; the association is positive when λ is less than P. The concavity of the relationship, which represents the effect of return migration on birth outcomes, also becomes more pronounced with larger differences between λ and P, even when these differences are relatively slight.

Finally, the concave relationship between birth outcomes and duration exists during the short term and disappears following the third period as the slope of the curve approaches zero. This implies that any long-term change in population birth outcomes found in the empirical studies is unlikely to be explained by selective return migration.

These results demonstrate the dilemma in distinguishing between the two mechanisms when conducting a linear analysis. A nonlinear analysis would be a better alternative for differentiating between these two effects. Furthermore, Goldman (2001) reminds us that these two mechanisms are not necessarily mutually exclusive and that if it exists, the selection mechanism must operate “alongside a complex set of causal pathways” (p. 112). Therefore, I now consider the combined effect of acculturation and return migration.

The Joint Acculturation and Selective Return Migration Case

When acculturation and return migration operate simultaneously, γ changes with the rate of return migration.7 Figure 3 displays the results of a simulation of the individual and joint effects of negative acculturation and poor health–selective return migration. This simulation model shows that the joint effect initially follows the pattern of return migration and then converges with the acculturation curve in a negative direction by the sixth year. The result is an inverted-U pattern, with health status increasing in the very short term and declining in the medium and long terms. The prominence of the inverted-U shape effect is determined by the level of μ: the higher the μ, the more prominent the inverted-U shape. When assuming positive acculturation and poor health–selection return migration, the results show a curve that follows a similar but smaller inverted-U shape8 over the short term, which converges with the acculturation mechanism in an upward direction over the long term.
Fig. 3

Simulation of individual and combined hypotheses of negative acculturation and poor health–selective return migration

This simulation demonstrates that when combined, the effects of selective return migration and acculturation vary by duration: the former in the short term, and the latter in the long term. Therefore, if return migration and acculturation mechanisms are operating together, any change in health status over the short term (less than five time periods) must be due to the effects of return migration, and any long-term changes must be due to acculturation.


This study assesses the underlying assumptions of the acculturation and selective return migration hypotheses as they apply to first-generation Latino immigrants. The simulations of each hypothesis confirm the expected direction of the relationship between population health status and duration: a decline in population health status may result from the negative effects of acculturation or selective return migration on good health; conversely, an improvement in health status may result from the positive effects of acculturation or selective return migration on poor health. The difference between the effects of these two mechanisms on health status is their speed and duration. Selective return migration affects health status over the short term and is short lived, whereas acculturation affects health status over medium and long terms and thus persists. While these results are consistent with empirical studies on intragenerational Mexican immigrant birth outcomes (Balcazar and Krull 1999; Guendelman and English 1995; Zambrana et al. 1997), distinguishing between the two hypotheses is, at best, problematic when linear analyses are used.

In the case of the joint acculturation and selective return migration model, the simulations indicate a curvilinear relationship between health status and duration. In general, there is an inverted-U shaped relationship that is affected by selective return migration over the short term and a linear relationship affected by acculturation over medium and long terms. This result is similar to that of the empirical study by Ceballos and Palloni (2010) that shows a concave downward relationship between birth outcomes and length of residence in the United States: evidence of short-term selective return migration on poor health and a long-term negative acculturation effect.

The results also indicate that the level of the return migration rate during the first few years of the immigrant’s length of stay can have a significant effect on the relationship between birth outcomes and duration. A low return migration rate results is a slight curvilinear effect, and a high rate results in a stronger effect. The estimates used from two different studies on return migration represent the upper (Reyes 2001) and lower bounds (Van Hook et al. 2006) of the return migration rate; the former is due to the more accurate accounting of undocumented migration available in the Mexico-based Reyes (2001) study.

These findings suggest that recent immigrants may be more likely to return to Mexico when in need of healthcare, perhaps because of stronger and more dependable sources of social support and healthcare. On the other hand, immigrants with longer residency may have developed the necessary social networks, social ties, and healthcare access that increase the likelihood of staying in the United States. This is consistent with studies showing that immigrants with longer residence in the United States have higher healthcare utilization than those with shorter lengths of residency (Durden 2007; LeClere et al. 1994). The results also provide evidence of negative acculturation that takes effect after many years of living in the United States.

Therefore, this simulation demonstrates that by examining the nonlinearities of birth outcomes and length of residency, an important step is taken toward differentiating between the sociobehavioral and data-artifactual effects on a population distribution for the case of Latino immigrant health. It also demonstrates the limitations of linear analysis when studying the effects of health-selective return migration.

A limitation of this analysis is the use of duration as a proxy for acculturation, which fails to incorporate the complexity and multidimensionality of that concept (Lara et al. 2005; Marin 1992; Marin and Gamba 1996). To the extent that duration does capture the effects of acculturation on health, this analysis allows for detecting the selective return migration mechanism.

This study provides evidence that the acculturation hypothesis alone is insufficient in explaining the relationship of health status and length of residence among first-generation Latinos in the United States. A full understanding of this relationship requires further examination of the role of selective return migration. Both mechanisms must be evaluated, individually and jointly. In sum, this study points to the necessity for further research of the effect of acculturation and selective return migration on the health status of first-generation Latino immigrants.


Proportion of favorable outcomes equals


Proportion of favorable outcomes in the case of acculturation effects equals


The effect of acculturation is simulated using a logistic function: for the positive acculturation effects beginning at a low bound of P (0.937):, and its inverse for the negative acculturation effects beginning at an upper bound of P (0.963): A linearized γ with similar parameters has little overall effect on the final results of this analysis (results not shown).


The ratio of the proportion of favorable outcomes for the selective return migration case equals, where Fi and Ui are equal to the number of cases with favorable outcomes and unfavorable outcomes, respectively, at time t = i.


The distribution of the return migration rates are based on the first, second, and fifth years of return migration in the Reyes (2001) study and simulate the following inverse exponential functions: (a) high, μi = 0.33et(i); (b) medium, μi = 0.28et(i); and (c) low, μi = 0.07e–t(i). The Van Hook et al. (2006) study distinguishes between the “emigration rate” (those who return to Mexico and stay) and the “return migration rate” (those who make multiple return trips). The return migration estimates used for this simulation is a combination of both rates.


Constant values for λ are assigned as follows: high λ = .99, medium-high λ = .98, medium-low λ = .94, and low λ = .92. Although it may be the case that λ changes over time, the results show only a slight difference from a constant λ for all cases except where λ declines from .99 to .92. (See Fig. 4 in Online Resource 1.) However, these differences have little effect on the final results (results available upon request.


The distribution of the joint effects of acculturation and selective return migration equals:, where Fi – 1i) and Ui–1i) equal Fi–1 and Ui–1, respectively, with a probability of favorable outcomes of γiP at time t = i.


See Figs. 5–7 in Online Resource 1 for results of joint simulations showing similar patterns, given various assumptions.



This research was supported by the National Science Foundation (Grant No. SES-0082704), the National Institute of Child Health and Human Development (Grants No. 3-F31-HD08740-02, L60 MD000933-01), the Robert Wood Johnson Foundation (Reference ID 030613), and funding from the International Migration Program of the Social Science Research Council with funds provided by the Andrew W. Mellon Foundation. The author acknowledges appreciation for comments on earlier drafts from Rodrigo Cantarero, Gustavo Carlo, and Julia McQuillan, and to Mike Spittel and Alberto Palloni for contributions at its early conceptual development. Any opinions, findings, and conclusions or recommendations expressed in this article are those of the author and do not necessarily reflect the views of these individuals, agencies, or foundations.

Supplementary material

13524_2011_17_MOESM1_ESM.pdf (93 kb)
ESM 1(PDF 93.3 kb)

Copyright information

© Population Association of America 2011