Determinants of private health insurance coverage among Mexican American men 2010–2013

Abstract

Background: Private health insurance (PHI) represents the largest source of insurance for Americans. Hispanic Americans have one of the lowest rates of PHI coverage. The largest group in the US Hispanic population are Mexican Americans; they account for about two in every three Hispanics. One in every three Mexican Americans aged 64 years and under did not have health insurance coverage. Mexican Americans have the most unfavorable health insurance coverage of any population group in the nation.

Objectives: The objective is to determine the factors associated with the gap in PHI coverage between Mexican American and non-Hispanic American men.

Methods: This study used the National Health Interview Surveys (2010–2013) as the sample. A non-linear Oaxaca-Blinder decomposition was run, estimating the explained and unexplained gap in PHI coverage between the groups. Several robustness tests of the model were also included.

Results: This study estimates that 44.4% of employed Mexican American men are covered by PHI compared to 79.5% of non-Hispanic American men. Nearly 60% of employed Mexican American men were found to be foreign born, 35% have an educational attainment less than a high school degree, and 40% are likely to have language barriers. Decomposition results show that income, low educational attainment, being foreign-born, and language barriers diminished the probability of private health insurance coverage for Mexican Americans, and that 10% of the gap is unexplained.

Conclusions: Most of the difference in the PHI rate between Mexican American men and non-Hispanic men is explained by observable differences in group characteristics: education, language, and immigration status. About 10% of the difference can be attributed to discrimination under the traditional interpretation of an Oaxaca-Blinder decomposition. The PHI rate gap is large and persistent for Mexican American men.

Keywords:

• Mexican American men
• private health insurance
• Oaxaca decomposition
• health disparities

JEL Classification Codes:

• I13
• I14
• J15

This article is related to:

The whole is greater than the sum of its parts: the importance of integrated, universal, and equitable healthcare coverage

Introduction

This paper provides an insight into the major determinants of the gap in the rate of private health insurance coverage (insurance obtained through the workplace or individual purchased insurance) for employed Mexican Americans compared to other non-Hispanic employed men prior to the implementation of key provisions of the Affordable Care Act (ACA) in 2014. The results provide an outlook into the health insurance coverage prospects for Mexican American men in the workplace under the historic new legislation.

In the US, the lack of health insurance coverage among Hispanics is striking; ∼ 27% of Hispanics aged 19–64 years were not covered in 2014 by either private or public insurance, as compared to 17% of blacks and 11% of whites¹. Hispanics are a heterogeneous population; Mexican Americans, Puerto Ricans, Cubans, and other Spanish-origin groups comprise the 55.6 million Hispanics currently residing in the US. The largest group in the Hispanic population are Mexican Americans and account for about two in every three Hispanics residing in the US². Mexican Americans have the most unfavorable health insurance coverage of any population group in the nation; one in every three Mexican Americans aged 64 years and under did not have health insurance coverage³. Nearly one quarter of the nation’s 46.7 million uninsured population in 2013 were Mexican Americans². Most studies on the lack of health insurance among Hispanics do not often disaggregate by Hispanic groups (Mexican Americans, Puerto Ricans, Cubans and other Hispanics) or separately by gender of a specific Hispanic group^4,5. Studies which control for health insurance coverage and the heterogeneity of Hispanics, specifically the Mexican American population, often aggregate men and women in the analysis and include the Medicare age 65 years and older eligible population in their analysis of insurance coverage⁶. The study of the Hispanic population as an homogenous population without respect to a specific origin group or gender is understandable given the relatively small sample size of Hispanics available in public data sources (e.g. National Health Interview Surveys, the Current Population Surveys, and the Community Tracking Surveys), and the overall aim to generalize about the population group as a whole.

A few studies have concentrated on the lack of health insurance for men among the different Hispanic groups. Employed Mexican American men aged 25–64 years were more likely than Cubans and Puerto Ricans to be without private health insurance and differences in human capital attributes among the groups accounted for much of the insurance disparity⁷. Results from a study of employer-based health insurance among Mexican American employer based insurance showed that the coverage disparity cannot be exclusively accounted for by over-representation of Mexican American men in occupations and industries with low insurance coverage⁸. These studies on the lack of insurance coverage for Mexican Americans and Hispanics in general were done prior to 2010, and well before the passage of the Affordable Care Act in 2013. The objective of this paper is to describe the determinants of private health insurance among employed Mexican American men in the workplace using the National Health Interview Surveys (NHIS) from 2010–2013.

Methods

Description of sample

This study uses the Integrated Health Interview Survey (IHIS) version of the NHIS⁹. Table 1 presents characteristics of employed men aged 18–64 for 2010–2013 for the full sample as well as the following sub-groups: non-Hispanic men and Mexican American men. Men who self-identified as Mexican or Mexican-American in the IHIS are classified as Mexican Americans, irrespective of race for the purpose of this study. Nearly all employed men, including Mexican Americans, were working more than 30 hours or more per week. The IHIS provides occupational and industry data for each worker; the sample does not provide information or identify workers who work for the same employer. The proportion of married men is uniform (60%) for Mexican Americans and other men; the average family size is, however, larger for Mexican American men. Although Mexican American men are working the same number of hours per week as other workers, they are also the most disadvantaged in private health insurance coverage, i.e. insurance obtained through the workplace or individual-purchased insurance. Before the full implementation of the ACA in 2014, eligibility for Medicaid and other public insurance was generally not favorable toward working-aged men. In our sample, ∼ 2% of non-Hispanic men and ∼3% of Mexican American men report Medicaid insurance coverage. Eighty per cent of non-Hispanic men were covered by private insurance as compared to ∼45% of Mexican Americans. Mexican Americans in comparison to non-Hispanics are less likely (61% vs 73%) to report being in very good or excellent health.

Table 1. Characteristics of employed men aged 18–64 years in the National Health Interview Survey, 2010–2013.

Percentage (except where noted)	Full sample	Non-Hispanic	Mexican Americans
Private Health Insurance	0.76	0.80	0.44
Mexican American	0.11	0.00	1.00
Mean age (years, SD)	40.66 (12.47)	41.19 (12.51)	36.6 (11.36)
White	0.83	0.82	0.94
Foreign born	0.16	0.11	0.60
English Interview Language	0.95	0.99	0.60
Federal Poverty Level (FPL) or less	0.07	0.06	0.21
Between 1–2 FPL	0.13	0.11	0.31
2 plus FPL	0.79	0.83	0.48
Less than high school (HS)	0.08	0.05	0.37
HS or equivalent	0.29	0.28	0.33
More than HS	0.63	0.68	0.30
Married	0.61	0.61	0.60
Family size (count, SD)	3.06 (1.58)	2.93 (1.46)	4.04 (2.00)
Very good or excellent health	0.71	0.73	0.61
30 hours or more per week	0.89	0.89	0.89
Northeast	0.17	0.19	0.04
North Central/Midwest	0.24	0.26	0.11
South	0.35	0.35	0.36
West	0.24	0.20	0.50
Observations	70,899	59,372	11,527

Other important contributors to the disproportionate private insurance coverage among employed Mexican American men are poverty, age, education, nativity, language, and regional concentration. The ability to purchase group insurance through the workplace or individual insurance in the market place represents a financial barrier for employed Mexican American men. Using the family poverty income levels as proxy for the ability to purchase insurance, employed Mexican Americans are disadvantaged by income; nearly a third (31%) are living below the federal poverty level (FPL) as compared to 11% of other men. Mexican American workers are generally 4 years younger on average than other workers, and half as likely (30% vs 63%) to have an education attainment more than a high school degree. Nearly two of every five employed Mexican American men have not completed a high school degree. The majority (60%) of employed Mexican Americans are foreign born; the legal residency status question is not asked in the IHIS. It is, therefore, not surprising that only 60% of interviews of Mexican Americans were conducted in English, while nearly all interviews conducted of non-Hispanics were in English. In comparison to other men, the vast majority of Mexican Americans are geographically concentrated, with nearly 90% residing in the South or West of the country.

Empirical model

Our model is a non-linear Blinder-Oaxaca decomposition, as originally used in Fairlie¹⁰ and as described in Fairlie^11,12. The dependent variable, private health insurance (PHI), is equal to one for men with private health insurance and zero otherwise. Our model includes all the variables discussed above (see Description of sample section and Table 1). The omitted categories for education and income are less than high school diploma or equivalent and the FPL or less, respectively. All models include dummies for year of survey.¹

The decomposition was performed using a pooled sample weighting—as opposed to choosing one group or the other. For examples of pooled weighting estimates see Neumark¹³ and Oaxaca and Ransom¹⁴. The decomposition method also requires one-to-one sample sizes with the smaller sample, Mexican Americans in our case, as the sample size defining group. The decompositions were performed 100 times and averaged. The average of 100 replications of the decompositions gave essentially the same results over several trials, so random number seeding was not performed. To test the importance of the ordering of the variables included in the decomposition, we performed the decompositions with the variables ordered, as shown in Table 2 and also with random ordering of the variables. The results were essentially the same; the random order results are reported in this paper. Additionally, our decomposition results take into account the complex sample design of the IHIS in the calculation of the standard errors.

Table 2. Detailed decomposition—percentage point contribution (SE), percentage contribution.

Sample	Total sample	West only	Small sample	Small w/industry	Small w/occupation
Age	0.0028	0.0023	0.0013	0.0005	0.0009
	(0.0009)	(0.0016)	(0.0015)	(0.0015)	(0.0016)
	0.89%	0.73%	0.41%	0.16%	0.29%
White	–0.0023	0.0021	–0.0032	–0.0042	–0.0039
	(0.0008)	(0.0016)	(0.0013)	(0.0012)	(0.0013)
	–0.73%	0.67%	–1.02%	–1.34%	–1.24%
Foreign born	0.0451	0.0366	0.0408	0.0354	0.0426
	(0.0038)	(0.0055)	(0.0065)	(0.0062)	(0.0067)
	14.37%	11.66%	13.00%	11.28%	13.57%
English	0.0535	0.0522	0.0599	0.0500	0.051
	(0.0046)	(0.0065)	(0.0092)	(0.0083)	(0.009)
	17.05%	16.63%	19.09%	15.93%	16.25%
Between 1–2 FPL	–0.0184	–0.0215	–0.0186	–0.0185	–0.0195
	(0.002)	(0.0039)	(0.0031)	(0.0032)	(0.0035)
	–5.86%	–6.85%	–5.93%	–5.89%	–6.21%
2 plus FPL	0.1393	0.1388	0.1427	0.1271	0.1359
	(0.0036)	(0.0067)	(0.0057)	(0.0073)	(0.0059)
	44.38%	44.22%	45.47%	40.50%	43.30%
HS or equivalent	–0.001	–0.0076	–0.0021	–0.0017	–0.0017
	(0.0007)	(0.0019)	(0.0012)	(0.0009)	(0.0011)
	–0.32%	–2.42%	–0.67%	–0.54%	–0.54%
More than HS	0.0759	0.0896	0.0697	0.0491	0.0482
	(0.0036)	(0.0079)	(0.0063)	(0.0061)	(0.0065)
	24.18%	28.55%	22.21%	15.64%	15.36%
Married	–0.0026	–0.0043	–0.0055	–0.0039	–0.0052
	(0.0005)	(0.0009)	(0.0009)	(0.0008)	(0.0009)
	–0.83%	–1.37%	–1.75%	–1.24%	–1.66%
Family size	0.006	0.0102	0.0065	0.0037	0.0034
	(0.002)	(0.0036)	(0.0038)	(0.0036)	(0.0039)
	1.91%	3.25%	2.07%	1.18%	1.08%
Very good/excellent	0.0049	0.0049	0.0047	0.0046	0.0047
health	(0.0005)	(0.0013)	(0.0009)	(0.0009)	(0.0009)
	1.56%	1.56%	1.50%	1.47%	1.50%
30 hours or more/week	0.0003	0.0004	0.0000	–0.0002	–0.0001
	(0.0001)	(0.0003)	(0.0003)	(0.0003)	(0.0003)
	0.10%	0.13%	0.00%	–0.06%	–0.03%

Notes: The table displays detailed results of non-linear Blinder-Oaxaca decomposition between other non-Hispanic men and Mexican American men. Year of survey dummies explain <0.10% of the gap.

Results

Overall, our decomposition explains ∼90% of the gap in PHI (see Table 3).² The unexplained difference, the discrimination part of the decomposition¹⁵, is only ∼10%. This is consistent across all decompositions.

Table 3. Decomposition.

Sample	Full sample	West only	Small sample	Small sample	Small sample
	(1)	(2)	(3)	(4)	(5)
Non-Hispanic PHI rate	0.7954	0.7963	0.7987	0.7987	0.7987
Mexican PHI rate	0.4440	0.4589	0.4565	0.4565	0.4565
PI gap	0.3514	0.3374	0.3422	0.3422	0.3422
Explained difference – Rate	0.3139	0.3039	0.3051	0.3086	0.3123
Explained difference – %	89.3%	90.1%	89.2%	90.2%	91.3%
Includes Region dummies	X		X	X	X
Includes Industry dummies				X
Includes Occupation dummies					X
Observations	70,899	20,186	29,545	29,545	29,545

Notes: The table displays the main results of non-linear Blinder-Oaxaca decomposition between other non-Hispanic men and Mexican American men. All regressions include year of survey dummies. All results have been obtained from the same sample; column 3 is a reproduction of column 1 but using the sub-sample which includes occupation and industry codes. Columns 3 and 4 include the industry and occupation codes respectively.

The explained difference is only marginally improved by including industry or occupation, with occupation results having the highest percentage explained, 91.3%.³ This represents a 1.9% difference or a 2.4% increase in the explained difference over the decomposition without occupation dummies (see Table 3, columns 4 and 5). Table 3 also highlights the PHI rate gap. It is consistently ∼34% (see Table 3, Rows 1 and 2; 79% vs 45%) different or, said another way, Mexican American men are 43% less likely ($\frac{(45-79)}{79}\times 100=-43\%$) to have PHI than other non-Hispanic working men.

The results of the detailed decomposition are shown in Table 2. Each variable’s coefficient is presented as a percentage point contribution (with its standard error) and the percentage contribution to the explained difference below. Income relative to the FPL is the highest contributing explanatory variable in each model. If Mexican Americans had the same fraction of employed men as non-Hispanics in the group, with two or more times the FPL, Mexican Americans would be 13–14% more likely to have PHI. After income relative to the FPL, the largest contributors to the explained difference are (in order): educational level, English as the interview language, and being foreign born. The low educational level and the English language limitations that are more prevalent among the foreign born are likely strong contributors to low family income, which accounts for most of the PHI disparity among employed Mexican American men.

Discussion

Although most Americans obtain health insurance through the workplace, employed Mexican American men are the least likely to have private health insurance coverage. Despite the similarities in hours worked per week, the lack of evidence that occupation or industry explain the gap in PHI, and no indication of changes in the years immediately preceding the implementation of the ACA, Mexican American men have systematic and persistent lower health insurance coverage through PHI.

The empirical results show that nearly all of the health insurance coverage gap between Mexican American men and non-Hispanic men are attributable to differences in human capital characteristics such as education, language, and immigration status. The unfavorable family income status as measured by the FPL of Mexican American men relative to non-Hispanics is the single biggest contributor to their health insurance gap. These findings suggest that the outlook for ameliorating the health insurance coverage of employed Mexican American men through the provisions of the ACA is discouraging.

The complexities of the ACA are especially challenging for Americans with limited English, low educational attainment, and poorly assimilated or acculturated immigrants. Employment opportunities or requirements to obtain private health insurance coverage and the coverage provisions of the ACA may disproportionately impact Mexican Americans given their low family income, the percentage of this population with limited English, lower levels of education, and immigration status to obtain insurance coverage. Although government subsidies are available to low income persons to purchase health insurance, the family income status of employed Mexican American men is more unfavorable relative to other workers. The exclusion of undocumented persons to obtain any health insurance under ACA coverage is especially dismal for a population where the majority of its workers are foreign born.

The study does have some limitations. This paper only includes employed men aged 18–64, which inherently creates some selection bias—although the unemployment rates for the two groups do not differ greatly and the same decomposition results were obtained when the unemployed were included in the sample. Thirty hours or more/week was chosen because that is an important part of the ACA, but so is firm size, specifically 50 or more FTE employees. Also, data on firm size or whether a group of workers are employed by the same employer is unavailable in the IHIS. Other data sets may have firm size, but the IHIS has a language of interview variable which is an advantage over other data sets. This trade-off is warranted because language is a major contributor to the explained difference and the impact of language on PHI is understudied and merits further research.

Conclusions

This research clearly shows that most of the difference in the PHI rate between Mexican American men and other non-Hispanic men is explained by observable differences in group characteristics. The low educational attainment of employed Mexican American men (nearly 40% do not have a high school education) interweaves with employment opportunities, type of employer, income, and contributes to the disparities in private health insurance coverage. Only ∼10% of the differences in PHI can at best be attributed to discrimination under the traditional interpretation of a Blinder-Oaxaca decomposition¹⁶. Other factors not included in the model such as quality of schooling, neighborhoods, or workplace settings could account for or chip away at the unexplained variation. This does not mean though that there is an absence of discrimination based on observable characteristics—birthplace is not a human capital investment like education or English language proficiency. The PHI rate gap is large and persistent for Mexican American men.

The next steps in this research are to explore how the gap has changed as newer data, post-ACA implementation, become available. Plans include further exploration of the PHI gap and shifts to public insurance programs, specifically Medicaid expansion states, and examining the differences in both of those changes between the groups of interest. Less favorable employment opportunities for foreign born workers including requiring legal residency for ACA coverage will likely have a disparate impact on the health insurance coverage for a group of workers whose nativity is outside the US. Further research should also explore private or public data that control for health insurance disparity among groups of workers who are employed by the same firm. Finally, language as a major determinant of the Mexican American PHI gap and access to the ACA raises important issues for the debates over official state-level constitutional languages, and will require further research.

Transparency

Declaration of financial/other relationships

There are no financial or other conflicts of interests to report for this submission. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Previous presentations

Earlier versions of this paper were presented at the American Society of Health Economics 2016, Philadelphia, PA, and the Allied Social Science Associations Meeting, American Society of Hispanic Economists Panel Session, 2016, San Francisco, CA.

Data availability statement

The data that support the findings of this study are openly available through IPUMS Health Surveys at https://nhis.ipums.org/nhis/. IPUMS Health Surveys harmonizes data from the National Health Interview Survey (NHIS) and allows users to create custom NHIS data extracts for analysis.

Acknowledgements

We thank the UNM Department of Economics, the Combined BA/MD program at UNM, and the Robert Wood Johnson Foundation Center for Health Policy at UNM.

Notes

Additional information

Funding

There is no funding to report for this submission.

Notes

1 The results are fundamentally the same with and without year dummies.

2 The smaller sample (see Tables 2 and 3) has the same inclusion/exclusion criteria discussed in Section 1, and additionally the individual needed to be selected for the supplemental survey which included questions on industry and occupation. The smaller sample characteristics are not reported, but are very similar to the large sample, and are available upon request.

3 There are 21 industry categories and 23 occupation categories among the employed men; the data is not able to identify workers who are employed by the same firm.