Are Remitters at Risk for Lower Food Security and Dietary Quality? An Exploratory Study of Mexican Immigrants in NYC

Abstract

Objective:

To examine whether remitting behavior among Mexican immigrants in the Bronx is associated with increased food insecurity and lower dietary quality, with a particular focus on potential gender differences in these associations.

Design:

Descriptive and bivariate statistics are shown, and binary logistic multivariate regression models are computed.

Setting:

Data come from a study exploring the social networks, dietary behaviors and outcomes of Mexican immigrants recruited from a Catholic Church in the Bronx between January 2019 and June 2019.

Participants:

81 Mexican immigrants 18 years or older living in the Bronx, New York City.

Results:

A statistically significant (p < .1) relationship was not found between sending remittances and food insecurity; however, we found that women remitters had higher odds than men remitters of having low dietary quality (p < .060). We also found that a higher Body Mass Index was associated with higher odds of experiencing low and very low food security (p < .037).

Conclusions:

Further research with nationally representative data is needed to investigate the full extent of the association between remittances and nutritional outcomes of remitters.

Introduction

Remittances, defined as money transfers by immigrants to their relatives or other people in their countries of origin, are an important aspect of immigrants’ lives.¹ Remittances confirm an immigrant’s continued social membership in their country of origin. Immigrants care deeply about those they left behind, and thus the emotions associated to remit are typically strong. Studies at macro and micro levels have tried to explain what factors may influence remittance behavior. Studies at the macro level have found that the means of transfers, the exchange rate and the interest rate on remittance shipments may shape the decision to remit. At the micro level, variables that may explain remittance behavior include age, gender, age at immigration, educational attainment, household income, migration status, acculturation, and marital status.^2,3 Increasingly, research also highlights the importance of the intersectionality between these roles, such that female immigrants are at higher risk for worse outcomes due to their double marginalization both as being women and as being a migrant.⁴

Diet and dietary quality are health-related outcomes that easily modified by many of the socioeconomic and sociocultural factors mentioned, including remittances. While there is extensive research on the nutritional impact of remittances on the people who receive them,⁵ the relationship between remittances and the nutrition status of the people who send them (“remitters”) is not well understood. Examining the role of remittances on the immigrant sender seems important given that they can lower an immigrant’s discretionary income, and food budgets are often the first thing that is manipulated when money becomes scarce,⁶ resulting in low food security or low dietary quality. Research among low-income populations suggests tighter budgets may result in greater consumption of calorie dense meals, or meals lacking essential vitamins and nutrients.⁷

However, most nationally representative survey that samples remitters lack data on diet-related outcomes. Most related studies have focused on the motives immigrants may have to send money to their home countries⁸ or on the impact remittances may have on the mental health of remitters.⁹ To the best of our knowledge, only two studies have examined the food security levels of remitter immigrants. One study from 2008 found that remitters were more likely to report hunger, but no meaningful associations were found for gender.¹⁰ The other study from 2013 found that gender roles were an important indicator for the differences in remittance behavior among undocumented Mexicans in NYC, where men (n = 301) remitted more than women (n = 130), 94% compared to 65%; however, there was no link between gender differences and food insecurity among remitters.¹¹

Research Question and Hypotheses

Our study addresses the gap in the literature by focusing on gender, remittances, and dietary outcomes of those who send them. This is done in several ways. First, for the measure of food security, we use the one defined by the US Department of Agriculture (“USDA”) since it has consistently been linked to multiple deleterious health conditions¹² The second domain of our analysis includes dietary quality measures, since food insecurity is associated with lower dietary quality, including low consumption of nutrient-dense foods such as vegetables and fruits.¹³ Third, our study focuses on Mexican immigrants living in the Bronx Borough of NYC (“the Bronx”), one of the largest immigration hubs in the US. Our paper aims to address to the following research question: Are remitters at risk for lower food security and dietary quality? The hypotheses we test in this paper relate to the subpopulation of Mexican Immigrants residing in the Bronx and include (a) remitters are more likely to suffer from food insecurity compared to those who do not remit; (b) remitters are more likely to have lower dietary quality compared to those who do not remit; and (c) the effect is more pronounced for women than men.

Methods

Study Setting

Data come from a study that explored in-depth the social networks and dietary behaviors and outcomes of 81 Mexican immigrants recruited from a Catholic Church in the Bronx between January 2019 and June 2019. To recruit participants, the church representative identified prayer and activity groups that had a significant number of Mexican American members and provided telephone rosters. Three research assistants, fluent in both English and Spanish, reached out to individuals either in person or over the phone, explaining the study’s details. Following eligibility screening, individuals who expressed interest provided their consent to participate.

Participants

Inclusion criteria for the study included age (≥18 years of age) and ethnicity-nationality (self-identifying as Chicano, Mexican, or Mexican American). There were no exclusion criteria regarding participant’s selection beyond nationality (i.e., Mexican heritage) and age (18 and older). Individuals from the same family were eligible for recruitment. The social and health survey includes mostly closed-ended questions, and participants received incentives for each completed portion of the study in the form of a $25 gift card. Trained bilingual research assistants collected the data in person using social network data collection software, EgoWeb 2.0¹⁴ and participants’ dietary intake was assessed using a 24-h dietary recall (Automated Self-Administered 24-h Dietary Assessment Tool (“ASA24”) also available in Spanish.¹⁵ Health survey data were collected using Qualtrics platform.

Measures

Our conceptual framework (shown in Figure 1)focused on remittances, gender, and food security/dietary quality, and a decision tree analysis (shown in Figure 2, below) guided our decisions regarding variable selection and thresholds for constructing binary data.

Figure 1.

Double marginalization framework for the relationship between remittance behavior and dietary outcomes among remitters.

Figure 2.

Decision tree with food security as the dependent variable for the analysis (N = 81)^a. ^aDecision trees are robust to missing values since the counts in each node are based on the observed values only through that level in the tree. For this reason, N = 81 for our decision tree, but N = 77 in our regression analysis. ^bFood security levels (High or marginal/ Low/ Very low).

Dependent Variables

Food Security

The USDA defines food security as the ready availability of nutritionally adequate and safe foods, or the assured ability to acquire acceptable foods in socially acceptable ways.¹⁶ The ranges of food security include: “high food security” which refers to no reported indications of food access problems or limitations; “marginal food security” or one or two reported indications, such as anxiety over food sufficiency or a shortage of food at home, but little or no indication of changes in diet or food intake; “low food security” or reports of reduced quality, variety, or desirability of diet, but little or no indication of reduced food intake; and “very low food security” (formerly “food insecurity with hunger”) refers to multiple indications of disrupted eating patterns and reduced food intake.¹⁶ Participants were asked the same questions as in the US Household Food Security Survey developed by the USDA in their preferred language. The sum of affirmative responses to a specified set of items is referred to as the household’s raw score on the scale comprising those items. Raw scores of 0 and 1 are classified as “high or marginal food security”; raw scores from 2 to 4 are classified as “low food security”; and raw scores from 5 to 6 are classified as “very low food security”. For our study, the food insecurity data was binarized into 1 and 0, with 1 = denoting “low” and “very low food security” and 0 = denoting “high” or “marginal food security”.

Dietary Quality

The Healthy Eating Index (“HEI”) is a measurement tool of dietary quality that is scaled from 0 to 100, where 100 indicates completely alignment of a person’s diet with the 2015–2020 Dietary Guidelines of Americans.¹⁷ There are 13 components, of which 9 measure adequacy (consumption of total fruit, whole fruit, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, and fatty acids) and 4 measure foods to be consumed in moderation (refined grains, sodium, added sugars, and saturated fats). The score for adequacy components increases with increased consumption, whereas the score for moderation components increases with decreased consumption.¹⁷ Participants’ dietary intake was assessed using two 24-h dietary recalls with the ASA24 tool three to 4 days apart. Although each dietary assessment method has strengths and weaknesses, the 24-h recall provides highly-quality, detailed data on dietary intake for 1 day and is generally believed to be the least biased for dietary assessment of adults.¹⁸ The total food consumed by the participants was downloaded directly from the ASA24 algorithm. This algorithm calculates HEI scores by considering the 13 components previously mentioned. Continuous and binary data were used in this analysis. The binary data were obtained by assigning 1 = to HEI scores of less than 52, and 0 = to HEI scores greater than or equal to 52. This cut-off point was chosen based on the results of studies that included HEI and race/ethnicity in their analysis^19–21 and the decision tree results shown in Figure 2.

Independent Variable

Sends Remittances

Participants were asked whether they sent money to family in Mexico, and the variable was coded 1 = yes and 0 = no. The original question came from the National Latino Asian American Survey (“NLAAS”), which is the only national survey in the US to specifically ask about remittances in a representative sample of US adults.²² We modified the questions to be specific to family in Mexico.

Effect Modifier

Gender

Participants were asked whether they identify themselves as female (coded as 1) or male (coded as 0). Gender identity is a multidimensional construct that could include non-binary terms²³; however, such classifications are beyond the scope of our analysis. We note that the proportions of males (83%) in our sample who remitted is much higher than that for females (65%), and the difference of these proportions is statistically significant (p-value = .059) at 10% level of significance. For this reason, we have tested gender as an effect modifier for remittance sending, rather than studying its mean effects only.

Control Variables

For control variables, we identified the socioeconomic and demographic factors that have been reported in the literature to be associated with food security or dietary quality^24–26 as well as the decision tree.²⁷

For the decision tree analysis, we used the R statistical software for an exploratory analysis of the data. Decision trees are a useful tool for identifying the most informative covariates in modeling the dependent variable in an intuitive way. Decision trees are constructed with recursive binary partitions of the feature space.²⁸

Figure 2 shows a decision tree with food security as the dependent variable. The variables that best explain food security include HEI, gender, age, age at immigration, BMI and lack of US citizenship. For our dataset, the decision tree helps inform the split point for HEI (<52).

We briefly explain the output of the decision tree in Figure 2. For more detailed treatment of the methodology, refer to any standard text on machine learning, see for example,²⁹. Each box with a set of numbers is an end point. These end points describe how many participants following the tree’s conditions fall into each of the three categories of food security (High or marginal/Low/Very low). For example, the left most box that contains (16/0/0) means that for those who (1) lack US citizenship, (2) have a BMI of less than 36, (3) immigrated when they were younger than 28, and (4) are females, 16 of them experienced high or marginal food security, while none experienced low or very low food security.

Considering the literature and the decision tree analysis, certain variables were excluded (e.g., family living in Mexico) while the following variables were included:

Annual Household Income

We used a question that has been validated among a representative sample of Latinos in the NLAAS.²² Participants were asked to provide the best estimate of their income for the entire household before taxes based only on wages or stipends from employment (i.e., participants were specifically reminded to not include pensions, investments, interest, dividends, rent, social security, alimony or child support, unemployment insurance and armed forces or veteran’s allotment). Annual household income was classified into four categories as follows: 1 = Less than $10,000, 2 = $10,000 to $29,999, 3 = $30,000 to $49,999,4=$50,000 or higher. For descriptive statistics, pairwise correlations and logistic regression models, we used the four-level version of the variable.

Educational Attainment

The participants were asked about their highest level of education completed. If a participant attended school outside the US, they were asked to mark the US equivalent. Responses were classified into three categories: 1 = 11^th grade or less, 2 = High school graduate or GED, 3 = Some college or higher. For descriptive statistics, pairwise correlations and logistic regression models, we used the three-level version of the variable.

Age

Participants were asked for the year in which they were born, and their age was calculated based on the date the interview took place.

Age at Immigration

Participants were asked how old they were when they first came to the US. This question also came from NLAAS.²² Evidence suggests that more time spent in a foreign country may increase the risks of unfavorable dietary changes, especially when it comes to in the US and Canada.⁴ Furthermore, in the US, immigrants have been found to be in better health than their US-born counterparts, but their health advantages consistently decrease as the duration of their US residence increases.³⁰

Lack of US Citizenship

Modified from NLAAS, this question asked whether a participant was born as a US citizen or became a citizen through naturalization. There is evidence that suggests that households composed entirely of noncitizens are more likely to be food insecure compared to households composed of natural born citizens.^31,32 Benefits of obtaining US Citizenship include bringing family members to the US (allowing for reunification), becoming eligible for federal jobs (which may lead to a higher income), and accessing the Supplemental Nutrition Assistance Program benefits.^31,33 Lack of US citizenship in our study is classified as 1 = “participants does not have US citizenship”, and 0 = “participants does have US Citizenship”.

Acculturation

Acculturation scores were derived using the Marin’s Short Scale.³⁴ The scale primarily captures changes in language use such as the language one speaks at home, the language one speaks with friends, the language one reads in, and the language one thinks in. In total, twelve questions were asked. Scores of at most three indicate low acculturation, and greater than three indicates high acculturation. For our analysis acculturation was represented as a continuous variable with 2 decimal points. Acculturation in the US is shown to be connected with worse dietary choices such as low consumption of fruit and vegetables, higher consumption of fats and sugar, higher body max index (“BMI”), and low physical activity.^35,36

BMI

Trained interviewers determined each participant’s ability to stand and found that all participants were able to stand on both feet. After this assessment, participants were asked to remove anything from their pockets before being asked to step on the scale, which was a SECA Robusta 813 digital scale calibrated before each measure. The interviewers then used a portable height measuring board (SECA) to measure height. Height was recorded in meters and weight was recorded in pounds. Using the standard formula of the Centers for Disease Control and Prevention,³⁷ a person’s weight in kilograms was divided by the squared high in meters. A BMI less than 25 means the person’s weight falls within the healthy weight range; a BMI between 25 and 29.9 means it falls within the overweight range; and a BMI of 30 or higher means the weight falls within the category of obesity.³⁷ Current literature states that BMI is associated with both food insecurity and the quality of the diet. Additionally, higher readings of immigrants’ BMI is associated with higher levels of acculturation, lower levels of income, lower levels of education, and correlated with age and gender.^38–40 It is for these reasons why BMI is included as a control variable in our analysis. Finally, continuous BMI data was utilized, up to the first decimal point.

Marital Status

Participants were asked, What is your current marital status? where 1 = Single, never married, 2 = Married, 3 = Living with partner but not married, 4 = Divorced, 5 = Separated, and 6 = Widowed. For our analysis we converted Marital status to binary where 1 = Married and 0 = not Married. Literature suggests that marital status influences remittance behavior and health outcomes.⁴¹

Analysis

Any participant who did not know (n = 1) or refused (n = 3) to answer any given question was excluded from the data set, resulting in a final analytical sample of N = 77. All descriptive and bivariate statistics using pairwise correlations and decision tree analysis, as well as binary logistic multivariate regression models, were computed using IBM SPSS Statistics Version 27.⁴² Pairwise Pearson sample correlations between the response variable and the possible predictors are computed to perform an initial analysis by determining the subset of predictors that are individually correlated with the response variable, and hence have predictive ability. The decision tree approach is a well-known strategy for identifying important predictors in the context of regression analysis and model selection with many possible predictors.²⁸ We used these to build two different models to test our hypotheses. The first model tests whether Mexican immigrant remitters in the Bronx are more likely to experience food insecurity than those who do not remit after accounting for control covariates. This hypothesis is tested using a binary logistic multivariate regression (see Model 1). The model also tests the hypothesis that sending remittances has a greater effect on females than on males by adding the interaction term Remittances × Female to examine whether gender acts as an effect modifier for the independent variable.

$$\text{Logit}{\left(P\left(\mathit{\text{Food}}\mathit{\text{Insecurity}}\right)\right)}_i=\mathit{\text{Constant}}+{\beta }_1*{\mathit{\text{Remittances}}}_{\mathit{\text{i}}}+{\beta }_2*{\mathit{\text{Remittances}}}_{\mathit{\text{i}}}*{\mathit{\text{Female}}}_{\mathit{\text{i}}}+{\beta }_3*{\mathit{\text{ControlCovariates}}}_{\mathit{\text{i}}}+{\epsilon }_i$$ (1)

Logit refers to the natural log of the odds of a participant experiencing food insecurity. The adjusted odds ratio (“AOR”) is a measure of association between exposure (e.g., remittances) and outcome (e.g., food insecurity),⁴³ controlling for HEI, annual household income, education, gender, age, age at immigration, BMI, lack of US citizenship, and acculturation.

The second hypothesis tests the associations among Mexican immigrant remitters residing in the Bronx, and their dietary quality, compared to those who do not remit. This hypothesis is tested using the binary logistic multivariate regression (see Model 2). In addition, the interaction term Remittances × Female was included to assess the significance of gender as an effect modifier among remitters.

$$\text{Logit}{\left(P\left(HEI\right)\right)}_i=\mathit{\text{Constant}}+{\beta }_1*{\mathit{\text{Remittances}}}_{\mathit{\text{i}}}+{\beta }_2*{\mathit{\text{Remittances}}}_{\mathit{\text{i}}}*{\mathit{\text{Female}}}_{\mathit{\text{i}}}+{\beta }_3*{\mathit{\text{ControlCovariates}}}_{\mathit{\text{i}}}+{\epsilon }_i$$ (2)

Logit (P(HEI)) refers to the natural log of the odds of a participant having a low HEI score (less than 52). The control covariates included food insecurity, annual household income, education, gender, age, age at immigration, BMI, lack of US citizenship, and acculturation.

We set the statistical significance level at 0.1 since this is a pilot study based on a relatively small sample with a robust set of control variables. Our goal is to evaluate the feasibility of conducting a preliminary study in this population and lay the groundwork for the design of a larger-scale study. The sample size of 77 was sufficient to allow detection of standardized effect sizes of 0.28 in regression models with 12 predictors at a significance level of 0.1 and a power of 0.8. The significance level of 0.1 is consistent with various other studies published in public health with small community samples.^44–46 Our study is, therefore, exploratory, but given the lack of research on the relationship between remittances and the nutritional status of the people who send them, it can inform future work with larger sample sizes.

Results

Table 1 represents the descriptive statistics of our data set. It illustrates that in our sample, 37.7% of participants had low and very low food security (35.2% of women and 43.5% of men). In addition, HEI scores were roughly the same between men and women, with an average of 57.5. Regarding remittances, 70.1% of participants sent money transfers to Mexico, and a higher proportion of men did so (82.6%) than women (64.8%). Women participants who sent remittances represented 45% of the participants, as denoted by the interaction variable ‘Female × Remittances’. Most of our sample (49.3%) had an annual household income between $10,000 and $29,999, and education level of 11^th grade or less (58.4%). Our sample includes individuals aged between 20 and 68, with an average age of 43.5, and most of our sample (46.7%) is in between ages 46–68. The participants moved to the US when they were between 9 and 65 years old, and the mean age at immigration was 23.4. The average BMI among participants was 29.9, where women on an average were obese with a BMI of 30.2. The average acculturation score in our sample is 1.6 (reflecting low acculturation). Most of our sample (62.3%) is not married. Finally, only five participants in our sample had US Citizenship compared to 72 who did not.

Table 1.

Descriptive Characteristics of Participants in the Bronx, NYC.

	Female (n = 54)	Male (n = 23)	Total (n = 77)
Continuous variables	Mean (25%, 75%)a	Mean (25%, 75%)	Mean (25%, 75%)
Healthy eating index	57.4 (51.0,65.4)	57.5 (50.6, 67.6)	57.5 (50.9,65.7)
Age	44.4 (38.2, 51.0)	41.4 (34.0, 54.0)	43.5 (35.0, 51.0)
Age at immigration	23.1 (17.0, 26.0)	24.0 (17.5, 27.5)	23.4 (17.0, 27.0)
Body Mass index	30.2 (26.0, 33.8)	29.1 (26.4, 32.5)	29.9 (26.2, 33.4)
Acculturation	1.6 (1.2, 1.8)	1.7 (1.3, 1.87)	1.6 (1.3, 1.8)
Female × Remittances	0.45 (0.0, 1.0)	-	0.45 (0.0, 1.0)
Categorical variables	Count (%)	Count (%)	Count (%)
Healthy eating index (HEI)
Low HEI (<52)	15 (27.8)	8 (34.8)	23 (29.9)
High HEI (≥52)	39 (72.2)	15 (65.2)	54 (77.1)
Food security
High/Marginal	35 (64.8)	13 (56.5)	48 (62.3)
Low/Very low	19 (35.2)	10 (43.5)	29 (37.7)
Sends Remittances
Yes	35 (64.8)	19 (82.6)	54 (70.1)
No	19 (35.2)	4 (17.4)	23 (29.9)
Annual household income
Less than 10K	17 (31.5)	1 (4.3)	18 (23.4)
10K to 29.9 K	26 (48.1)	12 (52.2)	38 (49.3)
30K to 49.9 K	6 (11.1)	5 (21.7)	11 (14.3)
50K or higher	5 (9.3)	5 (21.7)	10 (13)
Education attainment
11th grade or less	30 (55.6)	15 (65.2)	45 (58.4)
High school graduate or GED	14 (25.9)	4 (17.4)	18 (23.4)
Some college or higher	10 (18.5)	4 (17.4)	14 (18.2)
Lacks US citizenship
Yes	51 (94.4)	21 (91.3)	72 (93.5)
No	3 (5.6)	2 (8.7)	5 (6.5)
Marital status
Married	19 (35.2)	10 (43.5)	29 (37.7)
Not married	35 (64.8)	13 (56.5)	48 (62.3)
Age
20–30	7 (13)	5 (21.7)	12 (15.6)
31–45	20 (37)	9 (39.1)	29 (37.7)
46–68	27 (50)	9 (39.1)	36 (46.7)

^a25% denotes the 25th percentile and 75% denotes the 75th percentile.

Next, pairwise correlations were assessed to identify highly correlated covariates in the data set, as illustrated in Table 2. Gender-female and income have a low negative correlation of −0.308. Similarly, lack of US citizenship and age had a low negative correlation (−0.369), body mass index and age had a low positive correlation (0.348), and marital status and age had a low positive correlation (0.339). Only two moderate positive correlations were observed. The first one was found for acculturation and education scores (0.504). The second was found with age at immigration and age (0.519).

Table 2.

Pairwise Correlations^a.

Food Security

1

Healthy eating index

−0.006

1

Sends Remittances

0.002

0.062

1

Annual household income

−0.191

−0.045

−0.034

1

Education attainment

−0.151

−0.084

−0.119

0.112

1

Gender-female

−0.082

−0.002

−0.178

−0.308b

0.064

1

Age

0.138

−0.047

−0.141

0.038

−0.082

0.121

1

Age at Immigration

0.152

−0.261

−0.17

−0.118

0.236

−0.044

0.519c

1

Body Mass index

0.230

−0.056

−0.109

−0.072

−0.128

0.104

0.348b

0.103

1

Acculturation

−0.244

0.137

−0.229

0.289

0.504c

−0.102

−0.057

0.03

−0.136

1

Lacks US citizenship

0.109

−0.045

0.289

−0.122

−0.069

0.058

−0.369b

−0.113

0.008

−0.224

1

Marital status

−0.196

0.080

0.097

0.233

0.058

−0.078

0.339b

0.013

0.259

0.051

−0.230

11

Food security

Healthy eating index

Sends Remittances

Annual household income

Education

Female

Age

Age at immigration

Body Mass index

Acculturation

Lacks US citizenship

Marital Status

^aFood Security categories were used, and Healthy Eating Index was continuous.

^bLow positive (negative) correlation: ± 0.30 to ± 0.50.

^cModerate positive (negative) correlation: ± 0.50 to ± 0.70.

Table 3 shows the results for the binary logistic multivariate regression that tested the association between food insecurity and remittance senders as well as the gender interaction (see Model 1). This model estimated an AOR for BMI of 1.1147 (p < .037), meaning that higher BMI was associated with greater odds of having low or very low food security. In addition, Table 4 shows the analogous results for dietary quality (as measured by HEI) as the dependent variable (see Model 2). The results show an improvement in the p-value for remittances, but the result is not statistically significant at the 0.1 level. Women were less likely overall to have low HEI scores than men (AOR 0.010, p < .062). However, the interaction term between remittances and gender is the highlight of our results. It shows that women who remitted had higher odds of having a low HEI score (OR = 114.479, p < .060) compared to men who remitted. Greater acculturation was significant and negatively associated with low HEI scores (OR = 0.018, p < .001), meaning more acculturated individuals were less likely to have low HEI scores.

Table 3.

Logistic Regression With Food Insecurity as the Dependent Variable.^a

		95%CI
Independent variables	Odds ratio	Lower	Upper	p-value
Healthy eating index	1.018	0.973	1.065	.441
Sends Remittances	0.188	0.005	7.437	.373
Annual household income	0.700	0.340	1.444	.335
Education attainment	0.779	0.323	1.877	.578
Gender- Female	0.078	0.002	3.206	.179
Age	1.029	0.959	1.105	.425
Age at immigration	1.018	0.935	1.109	.677
Body Mass index	1.1147	1.008	1.305	.037**
Acculturation	0.363	0.082	1.609	.182
Lacks US citizenship	8.654	0.221	338.735	.249
Marital status	0.398	0.111	1.430	.158
Female × Remittances	5.160	0.107	248.320	.406
Constant	0.016	-	-	.224

^a1 = Low and very low food security and 0 = High and marginal food security.

^***p < .01,

^**p < .05,

^*p < .1.

Table 4.

Logistic Regression With Low Healthy Eating Index as a Dependent Variable.^a

		95%CI
Independent variables	Odds ratio	Lower	Upper	p-value
Food security categories	0.572	0.204	1.608	.289
Sends Remittances	0.028	0.000	3.836	.140
Annual household income	1.714	0.766	4.904	.190
Education attainment	1.647	0.553	4.261	.370
Gender- Female	0.010	0.000	1.260	.062*
Age	0.971	0.891	1.059	.512
Age at immigration	1.081	0.984	1.187	.105
Body Mass index	1.080	0.943	1.236	.267
Acculturation	0.018	0.002	0.204	.001***
Lack US citizenship	2.035	0.018	232.790	.769
Marital status	0.538	0.117	2.486	.427
Female × Remittances	114.479	0.825	15890.513	.060*
Constant	76.394	-	-	.277

^a1 = low Healthy Eating Index (<52) and 0 = high Healthy Eating Index (≥52).

^***p < .01,

^**p < .05,

^*p < .1.

Discussion

Our cross-sectional analysis is a first step in elucidating the potential impact of remittances on nutritional and health outcomes among immigrants. Our analysis suggests that women remitters within the sample were more likely to have a lower quality diet (HEI < 52) than men remitters. However, no significant relationships were found between food insecurity, gender, and remittances, as originally hypothesized. In contrast, we found that a higher BMI was associated with low and very low food security.

Remittances are highly influenced within the context of one’s family; they can represent contractual patterns (a sense of obligation) among family members, or they can be ingrained from social norms rather than conscious decisions made by an immigrant.⁴⁷ They can also exert negative pressure on the individual, to the point where senders may skip or cut down their meals. Though our data did not provide evidence that remittances are associated with food insecurity, it does suggest that dietary quality could be negatively affected, especially for women.

Altruistic motives may explain this result. There is evidence to suggest that, in the case of women, they are, and are expected to be, more altruistic than men and that the more a woman describes herself as independent or dominant compared to warm or tender, the less is her altruism.⁴⁸ In other words, the more a woman describes herself using male characteristics, the lower her altruistic scores can become. A qualitative exploration of altruism, specifically intergenerational altruism among Mexican immigrants, especially women, may elucidate how this particular construct shapes remittance and health behaviors in our community of interest. This would represent an important initial step in public health research on this topic.

Other characteristics unique to women may help explain our results. Immigrant women may be doubly disadvantaged, which may arise from the discrimination they face as women and as immigrants⁴⁹. To elaborate further, women not only find themselves in more vulnerable and low-paying employment sectors but also experience the reinforcement of gender roles through their placement in traditionally feminine positions. These positions, such as nursing, domestic work, and caregiving, are often shunned by many locals in the US, further exacerbating the challenges faced by women in the labor market.³ Essentially, these disadvantages manifest themselves is via income inequalities. The pay gap between non-immigrant men and immigrant women in high income countries is estimated at 20.9%, which is greater than the aggregate gender pay gap of 16.2%.⁵⁰ Even though the literature suggests that men send more remittances than women in absolute terms,⁵¹ women tend to put in a greater effort while sending remittances since the proportion of such remittances is bigger compared to their income, and they send with greater frequency.³ In our analysis, we control for annual household income; however, due to the categorical nature of our data on annual household income, it could be that gender income disparities related to the double disadvantages women immigrants face explain our results.

Furthermore, our analysis suggests that immigrants in our sample who have a higher BMI are more likely to experience low and very low food security. This result is supported by the existing literature. Food insecurity has been shown to be associated with poor diets and malnutrition. High rates of obesity are more likely in low-income communities that have poor food environments or that exist in food desserts.⁵² Stigma among obese or overweight individuals has also shown to increase the length unemployment spells, compared to non-obese or non-overweight populations. For example, women who are not obese are more inclined to secure higher-paying positions in comparison to their obese counterparts. These findings indicate that discriminatory preferences may be influencing the occupational choices of obese women, contributing to the wage penalty they face.⁵³

Our study has several limitations that should be considered in the context of our results. First, the primary focus of the original study was not on the impact of remittances on health. As a result, our analysis was limited by the coarse granularity of our data on remittances such as the absence of precise data on the amount or proportion of annual household income sent by immigrants to their country of origin and the frequency with which remittances are sent or the motives behind such money transfers (e.g., such as altruism). In our study, we have focused on cash remittances. Nevertheless, the incorporation of other forms of support, such as physical goods and cash-equivalent forms of payment like gift cards, might yield different conclusions. In addition, the exploratory nature of the study allowed for data collection from one community in the Bronx using non-probabilistic methods, which precludes our ability to generalize to the larger Mexican American population in the US. Our sample has fewer citizens, is older, and has less income and educational attainment compared to national statistics of Mexicans in the US.⁵⁴ In addition, there were no exclusion criteria for participants in our study. As a result, family members in the sample can potentially affect the independence assumption of linear regression, and hence the standard errors of the point estimates. In a future study with a larger sample and with more families and a higher number of subjects represented, one could incorporate cluster effects to estimate more robust standard error estimates. The small sample also contributed to lower precision for point estimates for our model parameters than what was needed for the significance level of 0.05. We were limited to the small sample size due to the original study itself. Finally, the observational nature of our study does not allow us to make any causal claims about the relationship between remittances and food security or dietary quality.

Conclusion & New Contribution to the Literature

Most existing literature on remittances examines to their impacts on the people that receive them. Research on the effect of remittances on food security and diet quality of senders is scarce. Our study fills this gap by exploring this topic with a community-based sample of Mexican immigrants in New York City and finds that women remitters experienced worse dietary quality than men remitters and that those with higher BMI experienced greater food insecurity. Research is needed with nationally representative data and/or longitudinal data that can uncover precise causal mechanisms between remittances and nutritional outcomes of the senders.

Biographies

Daniela Cruz-Salazar has a master’s in Public Health Nutrition CUNY SPH, and she researches immigrant health, food insecurity, and chronic diseases.

Neil Hwang is an Associate Professor at Bronx Community College, where he teaches and investigates computational and machine learning methodologies and their applications to social and public heath domains.

Shirshendu Chatterjee is an Associate Professor at CCNY and the Graduate Center, research topics include analysis of probabilistic models that arise from questions in Biosciences, Social Sciences, Physics and Computer Science; and Statistical Inference problems that arise from questions in Biosciences and other network data analysis.

Kathryn P. Derose has a dual appointment at the RAND corporation and University of Massachusetts Amherst, focuses on understanding and addressing health inequalities among Latino and immigrant populations.

Karen R. Flórez is an Associate Professor at CUNY SPH, researching sociocultural and neighborhood-level determinants of diet and diet-related diseases among high-burden populations.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author, [KRF], upon reasonable request.