Early Childhood Disadvantage for Sons of Mexican Immigrants: Body-Mass Index Across Ages 2–5

Abstract

Purpose

To distinguish the origins of higher weight status and determine when and why intra- and inter-racial/ethnic disparities emerge.

Design

Longitudinal analysis of the Early Childhood Longitudinal Study – Birth cohort (ECLS-B).

Setting

United States.

Subjects

Children of non-Hispanic White mothers and children of U.S.- and foreign-born mothers of Mexican origin from a nationally representative sample of children born in the year 2001 (N≈3700).

Measures

CDC growth charts determined sex- and age-specific weight status. Covariates were obtained from birth certificate records and parent interviews.

Analysis

Frequencies, growth curve trajectories, and ordinary least squares regression examined BMI and obesity across survey waves.

Results

Compared to their peers with non-Hispanic White mothers, children of Mexican heritage mothers have higher average BMI and greater rates of obesity. The BMI of boys with Mexican-born mothers is higher relative to Whites and children of U.S.-born Mexican mothers across early childhood, further increasing sharply at about age 4.5. This divergence is driven by increases in the BMI of boys, as girls do not show the same growth. A number of measures including descriptors of children’s nutritional intake, lifestyle factors, and acculturation, do not explain the increased obesity rates among sons of Mexican mothers.

Conclusion

Despite favorable perinatal health and weight, Mexican-American sons of foreign-born mothers show disadvantages in BMI that emerge close to the start of kindergarten.

Purpose

Obesity has become a major health concern as body mass levels have risen in the United States over the last several decades. The American Medical Association now recognizes obesity as a disease in hopes of bringing increased awareness and resources to the problem.¹ Childhood obesity is a particular concern because its rise and fall may determine the path to and away from obesity and chronic health problems at later ages: children who are overweight or obese are more likely to become obese or have health problems in adulthood.^2–3

Childhood obesity is also of public health import because it may help explain some of the sizable race/ethnic health disparities that persist throughout most of the life course. While nearly 17% of children and adolescents ages 2–19 in 2009–2010 were classified as obese, obesity rates were even higher among non-Hispanic Black (24.3%) and Hispanic (21.2%) children. Within the Hispanic pan-ethnic group, Mexican American children show a particularly high prevalence, with a 47–81% higher likelihood of being classified as obese relative to their non-Hispanic White counterparts.⁴

Researchers looking at childhood obesity are increasingly examining earlier ages to determine the origins of higher weight status.^5–7 Rates of obesity among preschool-aged children have doubled in recent decades, with nearly one in eight children ages 2–5 in 2009–2010 having a body-mass index (BMI) classified as obese.⁸ Data from 2008–2011 suggest that the increase has stabilized or reversed among children aged 2–5 in many U.S. states/territories, perhaps due to intervention efforts on the part of many agencies and organizations to improve nutrition and physical activity.^8–9 Despite this apparent progress, the prevalence is still nontrivial, and the obesity status of older cohorts of children still poses challenges to adult weight and health outcomes.

Early childhood also provides an opportunity to determine when and why inter- and intra- racial/ethnic disparities emerge. Although the infant health outcomes, including birthweight, of many Latinos are better than expected given their low socioeconomic status, which has been labeled a “Hispanic health paradox” (in perinatal outcomes)^10–13, children of Mexican immigrants have higher body-mass index (BMI) than White children of parents born in the United States as early as age 3.¹⁴ Higher BMI, and especially very high BMI levels, may be a meaningful divergence from favorable perinatal health that sheds light on other “crossover” processes in which immigrant health declines over time (for adult health, see Cunningham et al.¹⁵; for child health, see Van Hook and Balistreri¹⁶).

Research has yet to fully explain these health disparities and, particularly, the timing and mechanisms of high BMI levels among children of Mexican immigrants. The timing is particularly important since Hispanic children show increased obesity much earlier than Black children.¹⁷ However, our identification of this timing is constrained by the developmental context of early childhood. The Centers for Disease Control and Prevention (CDC) do not provide growth standards for overweight or obese until age 2¹⁸; for children younger than 2, researchers and clinicians use the term “high weight-for-length”, rather than “obese” or “overweight” (for example, Ogden et al.⁴). We therefore consider birthweight and weight-for-length in infancy, but do not examine these measures as outcomes.

Across childhood, genetics and epigenetics, intrauterine exposures, birthweight, diet, energy expenditure, TV viewing, and other factors have been identified as determinants of child obesity levels.¹⁹ In early childhood, birthweight, infant feeding patterns, taking a bottle to bed, mother’s weight status, consumption of sugar sweetened beverages, and sleep patterns have all been shown to have associations with children’s weight status.^20–25 While Hispanic children have greater rates of most of these risk factors,²⁶ these mechanisms do not explain greater BMI among Hispanic or Mexican-American children compared to their White or Black counterparts in early or mid-childhood.^{14, 20, 21}

More research is needed to understand how another possible set of mechanisms—measures of immigrant “acculturation” to societal values, health behaviors, and adaptation to U.S. life more broadly—that may lead to the adoption of behaviors and risk factors associated with obesity. There is limited information on the role of family acculturation in shaping children’s BMI, especially among ages 2–5, and the findings are not conclusive. Mother’s age at migration does not appear to be associated with disparities between non-Hispanic White children of U.S.-born parents and Hispanic children of foreign-born parents, although living in an English-speaking household is associated with increased risk of obesity.^27–28

Much of the literature examining childhood obesity controls for race/ethnicity, nativity, class, and gender but does not explore their intersectionality, even though prior findings suggest that certain groups experience disproportionate risk factors and rates of obesity.

Our study aims to advance knowledge on obesity in early childhood, ethnic disparities in child obesity, and the crossover from advantaged perinatal health to disadvantaged health among children of Mexican immigrants. To our knowledge, this study is the first to chart BMI trajectories across early childhood (ages 2–5) for a nationally-representative cohort. To accomplish these goals, this study examines BMI across ages 2–5 among a nationally representative group of White children of U.S.-born parents, children of Mexican heritage with U.S.-born parents, and second-generation Mexican American children.

Methods

Sample

We use data from the Early Childhood Longitudinal Study – Birth Cohort (ECLS-B). This dataset follows a nationally representative sample of U.S. born children from birth in 2001 until the start of kindergarten, up to 2007. We use information from Waves 1, 2, 3, and 4, when children are approximately 9 months, 2 years, 4½ years, and 5½ years old, respectively. Of the ~7000 children that participated in Wave 4, we analyze the ~3800 that are children of U.S.-born non-Hispanic White mothers, U.S.-born Mexican heritage mothers, and foreign-born Mexican mothers. About 100 children (<3%) are missing on BMI at Wave 4, leaving a sample of about 3700. All samples are rounded to the nearest 50 per security requirements by the ECLS-B. The descriptive statistics for this sample are displayed in Table 1. Attrition rates across the ethnic groups were not significantly different. We retain as much information as possible for each analysis, so samples other than the 3700 were used for descriptive information from Waves 2 and 3 and growth curve analyses. Wave 2 has fewer children, because children under age 2 do not have a defined obesity status and Wave 3 has more children because Wave 4 was conducted on an 85% subsample due to budget constraints. Growth curve analyses include any children with two or more points of information. Restricting descriptive statistics and growth curve analyses does not bring about any substantive differences.

Table 1.

Descriptive statistics

	Non-Hispanic white, U.S.-born mother		Mexican heritage, U.S.- born mother		Mexican heritage, foreign- born mother
Population	79.22	%	7.53	%	13.24	%
Wave 4 Overweight (≥85th percentile)	30.66	%	40.16	%	47.19	%
Wave 4 Obese (≥95th percentile)	14.10	%	22.85	%	31.15	%
Wave 4 Body-mass index	16.58		17.00		17.50
Age in months (Wave 4)	64.68		64.71		64.85
Male	0.50		0.51		0.55
SES Quintile (Wave 4)
1 (lowest)	0.09		0.23		0.65
2	0.16		0.29		0.22
3	0.22		0.25		0.08
4	0.25		0.16		0.05
5 (highest)	0.27		0.07		0.01
Region (Wave 4)
Northeast	0.16		0.00		0.02
Midwest	0.28		0.13		0.13
South	0.38		0.32		0.30
West	0.18		0.55		0.56
Urbanicity (Wave 4)
Urbanized area	0.61		0.78		0.88
Urbanized cluster	0.14		0.19		0.09
Rural	0.25		0.03		0.02
Birth order (1=firstborn)	1.95		2.15		2.38
Birthweight
Normal	0.82		0.85		0.85
Low	0.05		0.05		0.05
Very low	0.01		0.01		0.01
High	0.12		0.09		0.08
Weight-for-length in infancy (Wave 1)
<5%	0.07		0.06		0.07
5–95%	0.79		0.75		0.72
>95%	0.14		0.19		0.22

Source: ECLS-B

Notes: Accounts for complex sampling design. N ≈ 3700

Measures

Outcome

Our outcome of interest is BMI, a measure based on child height and weight collected by ECLS-B personnel. We also examine obese status, defined as at or above the 95th percentile of BMI for children of the same age and sex using CDC growth charts.²⁹ As mentioned above, age 2 is the youngest we examine because most studies do not interpret higher weight-for-length for younger ages as indicative of obesity or overweight (for example, see Baird et al.³⁰; see Moss and Yeaton³¹ for a counterexample).

Race/ethnicity and nativity

As indicated above, we categorize children based on the ethnicity and nativity of their mothers. We include children with non-Hispanic White U.S.-born mothers, Mexican heritage U.S.-born mothers, and Mexican-born mothers. To be concise, we refer to children of non-Hispanic White U.S.-born mothers as “White.” We refer to children of Mexican heritage as Mexican American, either 2nd generation (Mexican-born mother) or 3rd generation+ (U.S.-born mother).

Controls

Time-invariant control variables include child sex (1= male), birth order (1=firstborn), birthweight (very low <1500 g, low ≥1500 g and <2500 g, normal ≥2500 g and < 4000, and high ≥ 4000 g), and weight-for-length percentile in infancy (collected at Wave 1 and adjusted for age in months; categorized as between 5th and 95th, below 5th, and above 95th percentiles based on World Health Organization [WHO] growth charts). We use the WHO percentiles for infancy, as the CDC does not provide figures for children under the age of two. We also use the following covariates collected at each wave, with the referent given as the first in each list of categories: child age in months, region (Northeast, South, West, and Midwest), urbanicity (urbanized area [population ≥ 50,000], urbanized cluster [population ≥ 2500 and <50,000], or rural [population<2500]), and household socioeconomic status quintiles (5=highest, 4, 3, 2, and 1=lowest). These quintiles were based on an ECLS-B constructed, continuous variable that averages standardized scores on mother’s occupation and education, father’s occupation and education, and household income.

Behavioral risk factors (across the study period)

As a possible explanation for ethnic disparities, we examine behavioral risk factors including nutrition, practices around food, television watching, sleep, and mother’s BMI. A measure of regular consumption of sugar-sweetened beverages is defined as drinking sodas, sports drinks, or fruit drinks that are not 100% fruit juice every day for the last seven days (Waves 3 and 4) or usually drinking these beverages with meals or snacks (Wave 2). Frequent fast food consumption is defined as eating a meal or snack from a fast food restaurant four or more times per week at Waves 3 and 4. At Wave 1, parents reported on breastfeeding and solid food consumption, which were used to create dichotomous measures of whether the child was breastfed for at least six months and if the child ate solid foods before four months. Data limitations prevent us from distinguishing exclusive breastfeeding. At Waves 1 and 2, parents reported whether the child took a bottle to bed, a factor associated with increased likelihood of obesity.³² Sleep curtailment has also been associated with higher weight among children, and indicators identified whether the child usually slept less than ten hours or went to bed past 10 pm at Waves 3 and 4.^22,24 A dichotomous variable of watching more than the recommended guideline of two hours of television per day was taken from Waves 2, 3, and 4. Finally, mother’s BMI was captured at each wave. This measure was created using the interviewer’s weight measurements at each wave and mother’s self-reported height in Wave 1.

Additional risk factors in preschool (Wave 3 only)

To further explore possible influences on BMI, analysis examined additional nutritional intake factors and childcare variables at Wave 3 (preschool). Additional variables represented parent-reported nutritional intake. Parents reported the frequency that their children ate sweet snacks, salty snacks, fresh fruits, vegetables (other than french fries or fried potatoes), and 100% fruit juices in the last seven days. Wave 3 child care measures included a dichotomous measure indicating if the child was in center care and a continuous measure representing the total number of hours per week spent in nonparental care.

Acculturation

Citizenship, household primary language, interview language, the mother’s years in the United States, and a scale of English language abilities serve as proxies for acculturation. For mothers not born in the U.S., citizenship is self-reported. A dichotomous variable captures this report, with U.S.-born individuals coded as citizens. Household primary language and interview language are both dichotomous variables (0=English, 1=non-English). Mother’s years in the United States represented the difference between the age of the mother and the age she reported coming to the United States. We average the respondent’s reports of how good their reading, writing, understanding, and speaking skills are, with higher scores indicating more difficulties with English. Those reporting that English was the main household language were not asked this question, so they were given the lowest scores for all measures to represent no problems with English.

Preschool household and neighborhood resources

Additional measures include ZIP code measures and household resources at Wave 3. ZIP code measures represent the characteristics of the child’s residential area, as provided by the 2000 Census. Measures included the percentage distributions of race/ethnicity (White, Black, and Hispanic), educational attainment (less than high school, high school diploma, some college, college degree or more), income levels (low income, below poverty), female household head, unemployment, vacancy, renters, and working class. Townsend and Carstairs indices captured broader measures of deprivation, with the former using unemployment, non-car ownership and non-home ownership, and overcrowding³³ to compute a score of material deprivation and the latter incorporating low social class, non-car ownership, overcrowding, and male unemployment³⁴. Resources at Wave 3 include the receipt of government benefits (SNAP or food stamps, Medicaid, TANF/welfare, and WIC), health insurance, and assets (car ownership, investments, bank account, income-to-needs, and mother’s education), and social resources (household size, receipt of parenting help or advice, and no emergency contact).

Analysis

We first examine wave-specific frequencies of BMI and obesity by gender in each of the ethnic groups (with no controls). We then estimate multivariate growth curve models by gender, examining trajectories of BMI over time after controlling for SES and health behaviors. This approach nests time points within children, with age and age squared as the level 1 units and individual as the level 2 unit (see Van Hook and Balistreri¹⁶). To investigate how ethnic and nativity groups diverge over time, models include terms that interact Mexican U.S.-born mothers and Mexican foreign-born mothers with age and age squared. Prior research has assessed associations with weight status for children over time³⁵, but has not charted these trajectories. These growth curve models use time-varying covariates available at Waves 2, 3 and 4 (age, region, urbanicity, SES quintile, mother’s BMI), time-invariant covariates that do not change over time (child sex, birth order, birthweight, weight-for-length at Wave 1, breast fed for six or more months, received solid food before four months), and time-invariant counts of risk factors that summed the number of waves at which the risk factor was present (sugar-sweetened beverage, fast food, too much television, bottle to bed, sleeping less than 10 hours, bedtime past 10 pm). While we include possible influences on BMI in these growth models, we do not examine preschool nutrition and childcare, because they are specific to the middle of the observation period. Multicollinearity prevents us from including acculturation alongside race/ethnicity and nativity categories and including the many possible SES household and neighborhood measures in the growth curve models. To analyze these factors, we perform descriptive statistics among boys of Mexican-origin mothers and regression analysis on sons of Mexican, Mexican-American, and White mothers.

We then present descriptive statistics comparing obese and average weight boys with Mexican immigrant mothers on a range of factors, including the SES and health behaviors presented above as well as the preschool measures described above. We also present results from ordinary least squares (OLS) regressions predicting Wave 4 BMI controlling for Wave 3 BMI, focusing on the ages when sons of Mexican immigrant mothers demonstrate the largest increase. Multicollinearity again prevents inclusion of all the different health behavior, acculturation, and SES measures, but models with different specifications did not produce different results. All analyses account for complex sampling design using probability weights and jackknife replication weights provided by NCES, except growth curve models that account for probability weights and clustering in primary sampling units. While some families live in the same ZIP codes, there is not enough clustering to allow for multilevel models nesting individuals in ZIP codes. Adjusting for complex sampling design accounts for clustering in counties, which were used as sampling units³⁶.

Results

Obesity statistics for each wave are presented in Figure 1. The percentage of obese children ranges from 14.3% (White males, Wave 2) to 36.4% (2nd generation Mexican-American males, Wave 4). Non-Hispanic White sons and daughters of U.S.-born mothers show relatively similar prevalence rates across the three waves. Sons and daughters of U.S.-born mothers of Mexican heritage display higher rates of obesity, but these rates do not appear to change much over time. In contrast, sons of foreign-born Mexican mothers have a much higher obesity rate in Wave 4; daughters do not show this increase.

Figure 1. Percent obese by ethnicity and sex, Waves 2, 3, and 4.

Source: ECLS-B.

Notes: Statistics account for complex sampling design. Wave 2 N≈3450; Wave 3 N≈4850; Wave 4 N≈3700.

Because of the sharp increase among sons of foreign-born Mexican mothers, we display multivariate, longitudinal analyses of males. We also analyzed girls, but there were no significant differences in the rate of change in obesity over time for the ethnic groups with daughters of Mexican immigrants exhibiting a higher BMI at the start of the study period. The growth models in Figure 2 illustrate the interactions between ethnic/nativity group and age. Full tables are available upon request. Panel A shows trajectories over time for the ethnic groups, without any controls. These patterns should be interpreted within the developmental context of the normal J-shape in body mass gain across childhood²⁹. Looking at the 50th percentile, CDC growth charts exhibit a decline in BMI from ages 2 to about 5 (girls) or 6 (boys), when it begins to rise again. Although most groups evinced this general pattern, the apparent inflexion point where BMI starts rising again began to occur somewhat earlier for sons of Mexican mothers at around age 4.5, after this group also had a higher baseline BMI than White boys. Mexican-American girls had higher BMI than White girls but did not exhibit this additional jump around pre-school age. Panel B illustrates trajectories after the inclusion of risk factors and controls that are time invariant or time varying. The inclusion of these covariates equalized BMI over the time period for Mexican-American sons with U.S.-born mothers, but the accelerated increase among sons of Mexican-born mothers remained.

Figure 2. Male BMI trajectories by ethnicity, across Waves 2–4.

Source: ECLS-B.

Notes: Statistics account for probability weights and clustering in primary sampling units. Time-invariant covariates include birth order, birthweight, Wave 1 weight for length, count of regular consumption of sugar sweetened beverages, count of fast food consumption 4 or more times per week, count of watching TV more than the recommended guideline, count of taking a bottle to bed, breastfeeding for 6 or more months, eating solid foods before 4 months of age, count of sleeping less than 10 hours, and count of bedtime later than 10 pm and time-varying variables include region, urbanicity, SES quintile, and mother’s BMI. N≈2300 individuals or 5500 person-waves.

Given the BMI risks identified among sons of Mexican-origin mothers, we investigated their demographic, health behavior, and acculturation profiles. Table 2 displays the results of F-tests comparing obese and non-obese members of this disadvantaged group. Obese sons were significantly more likely to be in the lowest SES quintile, have high birthweight, and have a mother with higher BMI. Contrary to expectations that obesity is related to increased acculturation, they also were more likely to have a mother with English difficulties and significantly less likely to have a mother that was a U.S. citizen (or married). Mothers also reported that obese sons drank fewer sugar-sweetened beverages than boys of normal weight. Despite these differences, the other indicators did not reveal significant differences, though the small sample size may preclude significant findings for small effect sizes. While power analyses for Adjusted Wald tests adjusting for complex sampling design are not straightforward, a t-test with the sample sizes for these children and a power of .8 or higher should reveal effect sizes of .4 or greater³⁷. In other words, to detect a small effect such as the difference in taking a bottle to bed in Wave 2 (OR=1.6; d=.26), our sample only provides a power of .5. Thus, there may be other small differences between obese and normal weight sons of Mexican immigrants.

Table 2.

Descriptive statistics of sons of Mexican immigrants, by Wave 4 obesity status

	Not obese	Obese	Sig diff?
Population	64%	36%	n/a
Behavioral risk factors
Regular sugar-sweetened beverages	0.47	0.33	*
Fast food 4× per week+	0.20	0.20
# veggies per week	8.84	9.23
# fruit per week	12.59	12.35
Juice per week	12.23	12.23
Sweet snacks per week	6.57	4.78
Salty snacks per week	4.50	3.98
Regular family dinner	4.69	4.48
Bottle to bed (Wave 2)	0.28	0.39
Breast feed for 6 mo. + (Wave 1)	0.43	0.49
Solid food before 4 mo. (Wave 1)	0.14	0.20
Watches too much TV	0.34	0.34
Sleep 10+ hours	0.81	0.77
Bedtime past 10pm	0.05	0.03
Mom BMI	28.25	31.28	**
Controls
SES Quintile
Quintile 1 (lowest)	0.56	0.74	**
Quintile 2	0.23	0.18
Quintile 3	0.15	0.03	***
Quintile 4	0.05	0.05
Quintile 5 (highest)	0.01	-
Region
Northeast	-	0.06
Midwest	0.10	0.20	+
South	0.39	0.25	+
West	0.52	0.49
Urbanicity
Urbanized area	0.84	0.91
Urbanized cluster	0.14	0.07
Rural	0.03	0.03
Mom married	0.68	0.50	*
Birth order (Wave 1)	2.38	2.27
Birthweight (Wave 1)
Average	0.87	0.74	*
Low	0.06	0.04
Very low	0.01	0.00	+
High	0.06	0.21	**
Childcare
Center care	0.52	0.52
# childcare hours	16.62	17.73
Acculturation
Respondent is citizen	0.18	0.09	+
HH primary lang not English	0.92	0.97
Interview not in English	0.81	0.87
Years in US (Wave 1)
More than 10	0.46	0.46
5-less than 10	0.20	0.22
less than 5	0.35	0.31
Std. scale of English difficulties (Wave 1)	2.00	2.42	+
Resources
SNAP/food stamps	0.18	0.22
Medicaid	0.46	0.51
TANF/welfare	0.02	0.03
WIC	0.72	0.73
No health insurance for child	0.11	0.12
Household owns car	0.87	0.82
Household has investments	0.04	0.03
Household has bank account	0.38	0.28
Income-to-needs	1.21	1.05
Mom educational attainment	10.75	10.08	+
Household size	5.35	5.36
Received parenting help/advice	0.07	0.08
No emergency contact	0.02	0.02
Neighborhood characteristics (by ZIP code)
% White	0.43	0.41
% Black	0.11	0.13
% Hispanic	0.39	0.39
% Less than high school	0.31	0.33
% High school	0.25	0.25
% Some college	0.26	0.25
% College degree or more	0.18	0.17
% Low income	0.23	0.24
% Below poverty	0.16	0.17
% Female household head	0.61	0.60
% Unemployed	0.28	0.28
% Vacant	0.06	0.05
% Renting	0.42	0.44
% Working class	0.71	0.73
Townsend Index	3.42	3.75
Carstairs Index	2.95	3.22

Source: ECLS-B

Notes: Accounts for complex sampling design. Variables are from Wave 3 except where noted. Dashes indicate cell sizes too small to estimate. N ≈ 250

⁺p<.10

^*p<.05

^**p<.01

^***p<.001

Additional analyses (available upon request) focused on ethnic differences in the period of steep BMI growth. These models predicted boys’ BMI at Wave 4 (about age 5.5) after controlling for BMI at Wave 3 using ordinary least squares regression. Boys with Mexican-born mothers had significantly higher BMI compared to Whites, which was reduced slightly after controlling for SES and birthweight, but persisted beyond the inclusion of behavioral risk factors, type and number of childcare hours, and ZIP code characteristics. Despite the rich set of covariates available in these data, the mechanisms for the differences in BMI did not emerge.

Discussion

This study makes several contributions to our understanding of child obesity. First, it highlights young Mexican-heritage sons of immigrant mothers as particularly susceptible to higher weight status in early childhood. This finding corresponds to the conclusions of Van Hook and Baker³⁸, who investigated differences in the effect of acculturation on boys and girls ages 5–11. They found that boys with parents raised outside the United States weighed more and gained weight faster, but the same pattern did not emerge for girls. They were not able to test the mechanisms underlying the increased risk among Mexican heritage boys in middle childhood, but the authors^38:210 posited that the additional risk may be due to “gendered beliefs and parenting practices that are more likely to indulge boys.” Gender is important to understanding family functioning among immigrant families,³⁹ and boys generally take on less responsibility than girls in Mexican immigrant households.⁴⁰ Mexican parents may view heavier weight status as more favorable or healthier, but a prior study showed that Mexican immigrants’ beliefs, attitudes, and knowledge of weight status were similar, regardless of whether their preschooler was below or above the 95^th percentile for BMI.⁴¹ Future research should further examine how race/ethnicity, nativity, and gender combine to produce different dimensions of risk.

Second, our results establish early childhood as a critical period for the exacerbation of health disadvantage for Mexican-heritage sons of foreign-born mothers. While these children have displayed higher BMI across childhood ages, this study is the first to identify the timing of the further widening of the obesity disparity: about age 4.5. For many children this timing coincides with the prekindergarten year, an age when most children enroll in center-based preschools and may experience corresponding changes in diet, exercise, and other health behaviors. Children from Mexican immigrant families are less likely to receive center-based care,⁴² which could potentially spur divergent experiences at this age. At around the same age, children are aware of and begin to enact differences related to race, class, and gender,^43–45 which could shape the emergence of gender disparities in health behaviors.

Third, this study points to a complex web of factors that influence child health, which may require more detailed data collection methods. Other research has been unable to identify the mechanisms linking disadvantaged children to higher rates of obesity, and this study produced similar results, despite the focus on early childhood and the inclusion of a wealth of measures describing the children’s nutritional intake and other lifestyle factors. However, our data are limited, and we do not have information on patterns of behavior that lead to increased weight gain, such as portion size or physical activity levels. An alternative explanation is inaccurate reporting. It is unlikely that reporting differences are due to language difficulties, since ECLS-B provided bilingual interviewers and both Spanish and English programs for computer assisted personal interviewing, and only ~0 parents were unable to conduct an interview due to language difficulties³⁶. For example, mothers may underreport the eating habits of their sons because of social desirability or because they do not realize what they are eating.

Nonetheless, the socioeconomic and language disadvantages associated with obesity among these young Mexican boys suggest that these families may experience hardships, such as discrimination or stigma, as opposed to experiencing simpler forms of negative “acculturation” in which being “culturally” closer to the mainstream is related to worse BMI outcomes.⁴⁶ The role of language may be particularly salient, as low-proficient English speaking Mexican mothers have disproportionately more obese sons, a surprising finding that parallels that of Van Hook and Baker.³⁸ There is not a source to directly compare Mexican children of the same age in the same year to children in this study (that uses the same obesity thresholds), but similar estimates for Mexican children appear to be lower.⁴⁷ Further, unlike their adult counterparts, prior research identifies U.S. exposure as important for children of Mexican immigrants, as children in Mexico with a high propensity for immigration have lower BMI compared to both their Mexican counterparts with a lower propensity and Mexican-born children living in the United States.⁴⁸

The mechanisms for the influence of parent acculturation –including the disadvantages that produce said acculturation⁴⁹– on weight status among boys of Mexican heritage will be an important research target for the future. The more frequent occurrence of high birthweight (i.e., ≥ 4000 g) among Hispanics and the association between high birthweight and childhood body mass.^50,51 suggests that these mechanisms are in place prior to birth, but continue to operate into early childhood since birthweight did not explain racial/ethnic differences in obesity in regression analyses. An additional consideration is gestational diabetes, unmeasured in the ECLS-B, since this condition is associated with increased risk of high birthweight and is more prevalent among Hispanic women.⁵² Further, the higher BMI of the Mexican mothers of obese sons points to family-level influences.

In conclusion, while this study does not solve the origin of the crossover between favorable perinatal health and weight and unfavorable body mass levels in early childhood, it does push its origin to earlier in childhood. Moreover, this study also contributes to mounting evidence of the existence of an additional obesity bump for second-generation Mexican American boys around pre-school age, suggesting that some aspects of Mexican American health may be worse for the (male) children of immigrants than for third and subsequent generations. Future research should confirm the existence and improve understandings of this generational gap.

SO WHAT? Implications for Health Promotion Practitioners and Researchers.

What is already known on this topic?

Hispanic children display greater rates of obesity than their White counterparts. However, the causes and timing of the emergence of this disparity are yet unknown.

What does this article add?

This study aims to advance knowledge on obesity in early childhood, ethnic disparities in child obesity, and the crossover from advantaged perinatal health to disadvantaged health among children of Mexican immigrants. To our knowledge, this is the first study to chart growth trajectories of BMI across early childhood. We uncover a sharp increase in BMI and obesity rates among sons of Mexican immigrants starting at about age 4.5.

What are the implications for health promotion practice or research?

Our results suggest that young sons of Mexican immigrants are particularly at risk for unhealthy weight status. Interventions should target this population prior to their increased risk at age 4.5.