THE EFFECT OF OBESITY OR DISABILITY ON THE WAGES OF EMPLOYEES IN EMPLOYER- SPONSORED HEALTH INSURANCE PLANS

Transcription

1 THE EFFECT OF OBESITY OR DISABILITY ON THE WAGES OF EMPLOYEES IN EMPLOYER- SPONSORED HEALTH INSURANCE PLANS A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy By Megan Thomas, B.A. Washington, DC April 8, 2009

2 THE EFFECT OF OBESITY OR DISABILITY ON THE WAGES OF EMPLOYEES IN EMPLOYER- SPONSORED HEALTH INSURANCE PLANS Megan Thomas, B.A. Thesis Advisor: David Newman, Ph.D. ABSTRACT As labor can be compensated through either wages or non-wage (fringe) benefits, employers have incentives to reduce wages of employees in response to the costs of nonwage benefits (Woodbury, 1983). Health insurance is different than most fringe benefits as costs associated with it vary substantially across different individuals in an often uniform and observable way (Levy and Feldman, 2001). Evidence of whether wage penalties are made in response to the provision of employer-sponsored health insurance (ESI) is mixed, and evidence of whether potential penalties are levied more heavily on those with expected or actual high health care costs, like those who are obese or have physical disabilities, has not been well documented. This research explores this relationship using the University of Michigan Health and Retirement Study from This research tests the hypotheses that wages are lower for full-time workers with ESI obtained through their employer than those without; lower for obese full-time workers with ESI than obese full-time without; and lower for physically disabled fulltime workers with ESI than physically disabled full-time workers without. It also tests whether wages are similar between the obese and disabled, and if there are race or gender effects. The findings of this research do not support any of these hypotheses. Employersponsored insurance has no significant effect on wages, whether obese, disabled, or not. ii

3 I am extremely grateful for the guidance and continued support of Dr. Newman (despite juggling many other students), and thanks are also due to many of my coworkers for all of their advice and often last-minute help. iii

4 TABLE OF CONTENTS Introduction..1 Background..4 Literature Review.7 Conceptual Framework and Hypotheses...12 Data 14 Analysis Plan.15 Descriptive Results Regression Results Subpopulation Analyses.35 Discussion..48 References..52 iv

5 LIST OF TABLES AND FIGURES Table 1. Study Population: Full-Time Workers Ages 50-64, Table 2. Average Weekly Wage by Select Characteristics, Table 3. Percent of Full-Time Workers with ESI by Select Characteristics, Table 4. Average Weekly Wage by Disability and Coverage Status...26 Table 5. Average Weekly Wage by Insurance Status, BMI, Gender, and Race/Ethnicity, Table 6. Percent of Full-Time Workers with ESI by Job Industry, Table 7. Average Weekly Wage by Job Industry and Coverage Status, Table 8. Weekly Wage by BMI (continuous) and Worker Characteristics, Table 9. Weekly Wage by BMI (categorical) and Worker Characteristics, Table 10. Factors Significantly Associated with Wage Among Whites in the Multivariate Model, Table 11. Factors Significantly Associated with Wage Among Blacks in the Multivariate Model, Table 12. Factors Significantly Associated with Wage Among Hispanics in the Multivariate Model, Table 13. Weekly Wage Among Weight Groups by Worker Characteristics with Interaction Term for Race/Ethnicity and ESI, Table 14. Weekly Wage of Males by Worker Characteristics and Weight Class, Table 15. Weekly Wage of Females by Worker Characteristics and Weight Class, v

6 INTRODUCTION Employer-sponsored insurance (ESI) is the prevailing source of health insurance coverage in the United States, providing coverage for 158 million nonelderly individuals (Kaiser Family Foundation, September 2008). The vast majority of workers who are offered coverage take it, with 60% of workers covered by health benefits offered by the firm for which they work (Kaiser Family Foundation, September 2008). Firms offer health insurance primarily because workers want it, and it is often seen as a means of attracting more favorable candidates. Workers, on the other hand, prefer to obtain their health insurance through their employer for several reasons: the preferential tax treatment of employer-based coverage in the U.S. tax system, to avoid adverse selection, and it offers the possibility of maintaining coverage over several years (Glied, 2003). As health care costs have continued to rise in the U.S., fueled primarily by growth in new medical technology, but also due, in part, to the prevalence of chronic illnesses and the aging of the population, so too have the premiums for health insurance coverage (Newhouse, 1992). The average annual premium for single coverage in 2008 was $4,704, which was about 5% higher than the premium amount reported in 2007 ($4,479), and nearly 47% higher than the premium amount reported in 1999 ($2,196) (Kaiser Family Foundation, September 2008). Of the $4,704 premium for single coverage in 2008, covered workers, on average, paid only $721, or about 16%. Although premiums 1

7 have steadily increased, the employee contribution has remained relatively stable over the past few years (Kaiser Family Foundation, September 2008). The share of the premium paid by covered workers varies by firm characteristics and practices. For example, some firms vary premium contributions by workers wage level, and some firms do not require workers to pay co-payments or deductibles for medical care (Kaiser Family Foundation, September 2008). However, despite the fact that employees may have considerably different health care expenditures, the Health Insurance Portability and Accessibility Act of 1996 (HIPAA), prevents employers or insurers of group health insurance plans from varying premiums or rejecting employees on the basis of health status (Kaiser Family Foundation, April 2008). Federal discrimination laws further protect against employment discrimination on the basis of race/ethnicity, age, disability, and gender (Cave, 1992). Insurers may, however, vary rates from year to year, modify offered benefits or features, and decide not to renew the group. Due to either assumed lost productivity or discriminatory practices (or both), there is some evidence to suggest that some firms have attempted to work around this provision by penalizing workers wages to compensate for their actual or expected health care expenditures. Some studies have shown, for example, that because individuals who are obese tend to have higher health care costs on average than normal weight individuals, obese workers who obtain employer-sponsored health insurance through their own 2

8 employer earn a significantly lower hourly wage than their normal weight counterparts that cannot be attributed to the potential affect of wage on weight (Cawley, 2004; Quesenberry et al, 1998; Thompson, et al., 1999). Stated another way, there appears to be a significant negative association between body mass index and wages for individuals who obtain health insurance through the firm they work for. As the weight of an individual is a readily observable characteristic, wages are not often discussed openly in the workplace, and obesity rates in the United States have continued to increase over the years, documenting whether wage penalties exist, and if they exist, if they will continue, is important especially if health care and health insurance costs continue to rise as expected. It is also likely that there are a number of other physical characteristics or ailments which may also present opportunities for firms to penalize their workers wages in a similar manner. As such, do individuals with employer-sponsored health insurance obtained through their employer earn lower wages than individuals without employer-sponsored insurance obtained through their employer? Do obese full-time workers with health insurance obtained through their employer earn higher wages than obese workers without? Do individuals with an observable/physical disability (e.g. use a cane or wheelchair) who obtain their health insurance through their own employer make less income than individuals who also have an observable/physical disability but do not receive health insurance through their employer? Are obese 3

9 individuals and individuals with a physical disability who receive ESI treated similarly in terms of take-home pay? Are there race/ethnicity and/or gender effects? A number of cost containment strategies have been, and continue to be explored to both reduce the financial burden on employees who offer health insurance and also make individuals more accountable for their health care use. Investigating this relationship will document the importance of emphasizing such strategies, noting that such behavior may persist until viable strategies are found and then implemented. This research explores this relationship using the 2006 University of Michigan Health and Retirement Study. BACKGROUND Labor can be compensated through either wages or non-wage (fringe) benefits. Early research on the substitution of wage and non-wage benefits found that when fringe benefits were defined as health benefits and life insurance, they were relatively good substitutes with wage (Woodbury, 1983). Profit-maximizing employers have incentives to reduce wages of employees in response to the costs of non-wage benefits, and while economic labor market theory suggests that workers often bear the full cost of these job amenities, research is inconclusive (Levy and Feldman, 2001; Jensen and Morrisey, 2001). Unlike other fringe benefits, like tuition assistance or vacation, health insurance benefits are different because the costs associated with it vary substantially across 4

10 different individuals often in a uniform and observable way (Levy and Feldman, 2001). For example, older workers tend to have higher health care costs than younger workers, and those with family coverage rather than single coverage will have typically have higher health care costs (Levy and Feldman, 2001). Evidence of whether a wage adjustment is made in response to the provision of employer-sponsored health insurance is mixed, and further, despite federal laws which prohibit employee benefit plans providing health benefits from considering the health status of a member of the group in determining the member s eligibility for coverage, premium contribution, or cost-sharing requirements, evidence of whether a potential wage adjustment is individual or groupspecific levied more heavily on those with expected or actual high health care costs has not been well documented (Kaiser Family Foundation, April 2008). Obesity is one such observable condition for which the health care costs of these individuals are uniformly different from others, and would present employers with an opportunity to make a group-specific wage adjustment. As the risk for a number of illnesses, including hypertension, type 2 diabetes, and stroke, increases with increasing body mass index (BMI), obese individuals tend to have much higher health care costs than their normal weight peers, due, primarily to greater rates of use of, and expenditures on, hospital care (Andreyeva et al., 2004; Arterburn et al., 2005; Quesenberry et al., 1998; Thompson et al., 1999; Wee et al., 2005). As employer-sponsored health insurance is the prevailing source of coverage in the U.S. for those under age 65, and just over 60% 5

11 of adults in the U.S. were either overweight or obese in 2007, employers end up shouldering a substantial cost of overweight and obesity-related health care costs of employees (Kaiser Family Foundation, 2007). In fact, a study of the top ten most costly physical health conditions affecting six large U.S. employers in 1999 included four obesity-related conditions (Goetzel, et al., 2004). Further, a recent study by Finkelstein et al. (2003), estimated that overweight and obesity-related medical spending accounted for 8.2% of private health insurance expenditures between 1996 and 1998, or $19.8-$28.1 billion in 1998 dollars depending on the data source used (Finkelsetin et al., 2003). In , adults aged were more likely to be obese relative to younger and older adults (Ogden et al., 2007). A relatively small body of literature has documented insurance-related wage penalties, and an even smaller body has examined health insurance-related wage adjustments that are specific to certain individuals or groups (Bhattacharya and Bundorff, 2005; Cawley, 2004; Gruber, 1994; Jensen and Morrisey, 2001; Lehrer and Pereira 2007; Levy and Feldman, 2001; Miller, 2004; Morlock, 2000; Sheiner, 1999; Simon, 2001). However, it has yet to be examined how wage compares between older overweight and obese workers with and without health employer-sponsored health insurance, and if there is a difference, whether other groups with directly observable conditions or illnesses that are more likely to have higher health care costs are also likely to face such penalties. 6

12 LITERATURE REVIEW There is a vast literature attempting to explain how and why wages differ for employees. This literature suggests that differences stem from a mix of firm and employee characteristics or the interplay between the two (Groshen, 1990). Much of the research focused on the employee side of wages has examined how the attractiveness of employees may influence wages. In the same way, research has also focused on wage differences, or penalties, for employed persons with marked physical characteristics or particular mental or physical illnesses relative to those without such conditions. The majority of these studies focus earnings differentials for overweight, obese, and normalweight workers, examining how BMI affects rates of presenteeism (job productivity) and absenteeism, or how BMI affects the potential for occupational attainment. The literature on productivity and absenteeism consistently documents that those who are obese have a greater health-related loss of productivity than other employees, and higher rates of job absenteeism, which subsequently affect wage rates (Gates et al., 2008; Sullivan et al., 2008; Tsai et al., 2008). The literature examining occupational attainment also consistently documents wage penalties for employed overweight and obese women relative to normal-weight women, as they tend to be in lower-paying jobs and are largely excluded from managerial and technical positions (Averett and Korenman, 1995; Brunello and D Hombres, 2007; Haskins and Ransford, 1999; Mitra, 2001; Pagan and Davila 1997). However, studies that examine wage penalties for overweight and obese 7

13 men, or studies stratify by race and ethnicity present less consistent results (Averett and Korenman, 1995; Brunello and D Hombres, 2007; Loh, 1993; Mitra, 2001; Pagan and Davila 1997). At the same time, most of these studies do not focus on whether differences in earnings between obese and non-obese could be attributed to the provision of employer-sponsored health insurance. Only few studies examine whether compensation differentials resulting from the provision of health insurance benefits exist, and fewer studies examine whether wage penalties may be specific to particular groups or are individual specific. The literature regarding both is mixed. In an examination of wage differentials between workers with varying levels of employer-sponsored insurance, Morlock (2000) found that workers who received full health insurance had lower wages than workers without full health insurance. The difference in wages was substantial. Morlock estimated that those without health coverage that was paid fully by their employer had salaries that were 49% greater than those receiving full health coverage (Morlock, 2000). Miller 2004 also found evidence of a wage adjustment for the provision of employer-provided health insurance. In a study of working males between the ages of 25 and 55, Miller documented that workers who previously had, but then lost health insurance were compensated with a 10 to 11 percent increase in wage, which was roughly equivalent to the amount employers paid for health insurance coverage. At the same time, however, this model may not only capture the 8

14 effect of health insurance on wage; the author was unable to control for other fringe benefits, including paid vacation and sick leave (Miller, 2004). Evidence that employers may make wage adjustments in the provision of health insurance benefits specific to observable groups or individuals comes from a small body of literature on the pregnant, obese, and older workers. In 1994, Jonathan Gruber provided early evidence of how health costs are passed on to wages. The 1978 Pregnancy Discrimination Act stipulated that maternity benefits must be covered comprehensively in health insurance plans. Prior to this time, however, maternity coverage was often excluded from basic health benefits, or when included, often bore no relationship to actual delivery costs. Using the Current Population Survey (CPS), Gruber showed that 100% cost of the maternity benefit mandate was shifted to women of childbearing age, a directly observable group that was responsible for these costs (Gruber, 1994). Using data from the National Longitudinal Survey of Youth (NLSY), the Medical Expenditure Panel Survey (MEPS), and the National Health Interview Survey (NHIS), Bhattacharya and Bundorff 2005 found that obese workers with employer-sponsored health insurance pay for their higher expected medical expenditures through lower cash wages, with women suffering the greatest penalty. They found that obese workers earned, on average, $0.82 per hour less than normal or overweight workers, but did not document similar wage penalties for obese workers who 9

15 obtained health coverage through other means or through means in which BMI would not significantly impact the cost of health insurance coverage to the employer (Bhattacharya and Bundorff, 2005). Cawley 2004 provided further insight into this relationship by using multiple models to determine which of three primary explanations was responsible for the correlation between BMI and wages: 1) obesity lowers wages due to productivity or workplace discrimination; 2) low wages cause obesity; or 3) unobserved variables cause both obesity and low wages. Documenting wages for white, black, and Hispanic males and females using the National Longitudinal Study of Youth (NLSY) and the Third National Health and Nutrition Examination Survey, Cawley concluded that, for white females, a difference in weight of 65 pounds was associated with a difference in wages of 9%, which is equivalent to one and a half years of education or three years of work experience, suggesting that obesity lower wages due to productivity or workplace discrimination. He attributed other negative correlations between BMI and wage for other gender-ethnic groups to unobserved variables which cause both obesity and low wages. Overweight and obese black males, however, had a positive correlation between BMI and wage, and tended to earn more than their normal-weight and underweight counterparts (Cawley, 2004). Studying the cross-city variation of health costs for older workers using a number of data sources, Sheiner (1999) found that older workers, who tend to use more health 10

16 care, pay for their more expensive benefits through lower wages (Sheiner, 1999). Similarly, using the 1994 and 1998 Health and Retirement Survey (HRS), Jensen and Morrisey (2001) found a fragile wage-health insurance trade-off for older workers, with annual wages for older workers roughly $6,300 lower as a result of the provision of health insurance. They also noted that this wage offset was consistent with employer survey data which showed that annual family premiums in 1998 were between $5,800 and $6,900 (Jensen and Morrisey, 2001). Some studies failed to find any evidence of compensating differentials for the provision of health insurance. In 2001 Levy and Feldman set out to determine whether the incidence of health insurance costs was individual-specific, but failed to even find evidence that a tradeoff between wage and fringe benefits exists (Levy and Feldman, 2001). Simon (2001) examined the wages of displaced workers who either lost or gained health insurance through new jobs, and her results were the opposite of what was expected. Workers that lost health insurance through a job change also lost wages, but workers who gained health insurance also had wage gains (Simon, 2001). Lastly, in a recent study, Lehrer and Pereira 2007 similarly found no evidence of a wage adjustment for the provision of health insurance, but did find that the effect of health insurance on wage of workers in firms providing health insurance has increased by 50% between decades (Lehrer and Pereira 2007). 11

17 These studies provide at least some empirical evidence that a tradeoff between health insurance and wages may exist, and when it does, it often depends on at least two observable characteristics, age and obesity. However, there is no evidence to date that wage penalties are specifically levied against insured workers with physical disabilities or illnesses who are expected to have higher health care costs. Further, there is no evidence that within an older age group, wage penalties are specifically levied against insured workers with higher BMIs. CONCEPTUAL FRAMEWORK AND HYPOTHESES Using weekly wage as the dependent variable, this research will examine variation in weekly wage for obese and physically disabled full-time workers who obtain employer-sponsored health insurance through their own employer relative to obese and physically disabled individuals who do not receive health insurance through their employer in one primary model. The model will control for several wage explanatory variables, including age, race/ethnicity, gender, census region, education, health status, job industry, and tenure. The dependent variables of interest are self-reported body mass index (BMI), physical disability status, and health insurance status. Interaction terms for BMI and insurance coverage and disability status and insurance coverage are included to investigate whether wage is affected differentially by the particular combination of an 12

18 individual s insurance status and weight, or insurance and disability statuses. The primary model is outlined below: Model 1: Weekly Wage= β 0 + β 1 Wage Explanatory Variables + β 2 Insurance Status + β 3 BMI + β 4 Disability Status + β 5 Insurance Status x BMI + β 6 Insurance Status x Disability Status Subanalyses will also be conducted to determine whether workers wages are differentially affected by the interaction of insurance and weight and/or insurance and disability within a racial/ethnic group (Model 2), whether the interaction of insurance and race differentially affect wages within a weight class (Model 3), and lastly, whether sharper differences are observed when models are stratified by gender (Model 4). Model 2 (stratified by race/ethnicity): Weekly Wage= β 0 + β 1 Wage Explanatory Variables + β 2 Insurance Status + β 3 BMI + β 4 Disability Status + β 5 Insurance Status x BMI + β 6 Insurance Coverage x Disability Status Model 3 (stratified by weight/disability category): Weekly Wage= β 0 + β 1 Wage Explanatory Variables + β 2 Insurance Status + β 3 BMI + β 4 Disability Status + β 5 Insurance Status x Race/Ethnicity + β 6 Insurance Coverage x Race/Ethnicity 13

19 Model 4 (stratified by gender): Weekly Wage= β 0 + β 1 Wage Explanatory Variables + β 2 Insurance Status + β 3 BMI + β 4 Disability Status + β 5 Insurance Status x Race/Ethnicity + β 6 Insurance Coverage x Race/Ethnicity Together, these main and subanalyses will test the hypotheses that wages are lower for full-time workers with insurance obtained through their employer have than those without who do not obtain ESI through their employer; lower for obese full-time workers with ESI than obese full-time workers without ESI; and lower for physically disabled full-time workers with ESI than physically disabled full-time workers without ESI. This research will also test the hypotheses that obese and physically disabled workers with ESI earn similar wages, and that there are race/ethnicity and gender effects. DATA This research uses data from the University of Michigan Health and Retirement Study (HRS) from Supported by the National Institute on Aging (NIA), with supplemental support from the Social Security Administration (SSA), this longitudinal, face-to-face and telephone-based study surveys older Americans every two years since 1992 on indicators of health, insurance, economic well-being, and labor market status. The 2006 HRS, conducted between March 2006 and February 2007, sampled 18,469 individuals from 12,605 households (response rate not currently available), and is 14

20 representative of the community-based population in the U.S. (HRS 2008). The RAND Center for the Study of Aging, with funding and support from the NIA and SSA, created RAND HRS data files, which contain cleaned and processed variables with consistent and simplified naming conventions, and model-based imputations (RAND HRS, February 2008). The RAND HRS data file is used in this research. For the purposes of this analysis, the sample is restricted to individuals between the ages of 50 and 64 who are currently working full-time for pay, are not self-employed, and who either have employer-sponsored insurance (ESI) that is obtained through the respondent s employer and covers only the individual, or those who do not have ESI that is obtained through their employer. These individuals may be uninsured or may receive health insurance through a source other than Medicaid, a health insurance for the poor that has strict income eligibility limits. There are 1,402 observations in this subpopulation, representing 8,836,204 community-based individuals. ANALYSIS PLAN This research focuses on how obesity and disability affect the wages of working individuals who have employer-sponsored health insurance plans, as a way of offsetting real or potential high health care costs of these individuals. As such, the subpopulation of interest is paid workers who currently obtain health insurance through their employer. More specifically, this research focuses on adults between the ages of 50 and 64 who are 15

21 currently working full-time for pay, and belong to one of two insurance categories: 1) Have single coverage, employer-sponsored health insurance as their primary source of health insurance coverage and receive it through their own employer, or 2) do not receive health insurance through their employer. Those who belong to the second group may either be uninsured or have a primary source of health insurance obtained through another source. Individuals with Medicaid coverage, a health insurance program for the poor that has strict income eligibility requirements, have been eliminated. Further, the age of the population limits the number of individuals with Medicare as their primary source of coverage to only the permanently disabled, who are eligible for coverage before the age of All other observations, including individuals who are self-employed and those who do not receive pay for work, have been excluded from the dataset. Additionally, as family coverage is more expensive than single coverage and could impact an employer s wage decisions, individuals whose employer health plan covers both the respondent and dependents have also been excluded. Lastly, because the wages of part-time workers are substantially less than those of full-time workers, and the likelihood of being offered health insurance is greater for full rather than part-time workers, this research has been restricted to full-time employees. The Health and Retirement Survey, which sampled 18,469 individuals in 2006, contains a sufficient sample size to support this analysis. 1 Dual eligibles, low-income elderly and disabled persons with both Medicare and Medicaid, are eliminated from this analysis through the exclusion of Medicaid beneficiaries from the sample. 16

22 This research uses ordinary least squares regression to examine the relationship between weight and wages and disability and wages among those with employersponsored health insurance. The relationship between these variables is expected to be linear. The dependent variable is weekly wage rate in nominal dollars, and was calculated for respondents who were working at the time of the HRS interview using the usual hours worked per week, usual weeks worked per year, and pay rate, and are reported as gross wages. If the respondent was working but the wage rate was left missing, the respondent was excluded from the dataset. Three key variables of interest are examined in this research: body mass index, physical disability status, and health insurance status. Self-reported body mass index (BMI) is operationalized as both a continuous and categorical variable (separately), with weight categories based on the Centers for Disease Control and Prevention s (CDC) calculations and interpretations. BMI is calculated as weight in kilograms divided by height in meters squared, and is interpreted similarly for adults of both sexes. Individuals with a BMI below 18.5 are classified as underweight, , normal, , overweight, and 30.0 and above as obese (CDC 2009). It is important to note that perceptions of weight and weight categories may differ. Human capital variables such as sex and race are said to explain about 6% of the variation in wages (Groshen, 1988). As such, additional human capital variables included in the analysis are gender, age of the respondent at the time of the interview at the beginning of the month, and region of the 17

23 country of residence (Northwest, Midwest, South, West, and other). The classification of race/ethnicity in this research involved a set of mutually exclusive categories of non- Hispanic blacks, Hispanics, and non-hispanic whites. The racial/ethnic categories were derived from two sets of questions in the HRS. To establish race, respondents were asked to self-identify as one or more of three racial groups: white/caucasian, black/african American, and other. The race that the respondent considers to be his or her primary race is used in this research. To establish the ethnicity of respondents, the HRS asks individuals to choose between two ethnic groups Hispanic or Not Hispanic. Persons who reported Hispanic as their ethnicity regardless of their race are classified as Hispanic. Additional wage explanatory variables included in the model were education, which is derived from an HRS question that recodes years of education into categories ranging from less than high school, GED, high school graduate, some college, to college and above, and the respondent s self-reported general health status (excellent, very good, good, fair, and poor), which can affect rates of work absenteeism or productivity, thereby affecting earnings. To analyze the effects of a physical/observable disability on wages, disability status, in this research, is derived from a question in the HRS that asks whether the respondent uses equipment to walk across a room (i.e. cane, wheelchair). All individuals responding in the affirmative were classified as disabled for the purposes of this analysis. 18

24 Individuals with cognitive or mental disabilities were not incorporated into this definition as these types of disabilities may not be readily observable. Firm characteristics are equally important determinants of wage, and several were included in this analysis. Previous literature suggests that much of the unexplained variation in wages among employees can be linked to employer characteristics (Groshen, 1990). As previous studies have documented that employed overweight and obese women tend to sort into lower-paying jobs and are largely excluded from managerial and technical positions relative to normal-weight women, job industry is an important control variable in this model (Averett and Korenman, 1995; Brunello and D Hombres, 2007; Haskins and Ransford, 1999; Mitra, 2001; Pagan and Davila 1997). Job industry has been estimated to account for 29% of the variation in wages (Groshen, 1988). Tenure, which is measured in this research as job tenure in years at the current job, has been documented as one of the strongest determinants of wage, and is also included in all models (Groshen, 1991). The final key variable of interest is the health insurance coverage of the respondent. This variable was derived from two questions in the original HSR dataset (non-rand version) that asked respondents whether their primary source of health insurance was from a current employer, and, separately, whether anyone else was covered under the plan. Respondents who answered yes to the question that asked whether the respondent s primary source of health insurance was from a current 19

25 employer, and indicated that the plan covered only the respondent were identified as having single coverage employer-sponsored health insurance through their employer. Those who indicated that they received insurance through their employer but that the plan covered individuals other than the respondent (received family coverage) were eliminated from the dataset, as family coverage tends to be more expensive than single coverage and may be a factor in wage determination. Individuals who indicated that they did not receive health insurance through their current employer whether they were unemployed or received health insurance through a source other than Medicaid, a health insurance program for the poor were identified as not having employer-sponsored health insurance through their employer. 2 Restricting this population to age 64 years and under limits the number of respondents who receive health insurance through the Medicare program, a federal health insurance program for those aged 65 and older, which serves as the primary source of health insurance for most elderly. Individuals in the analysis with Medicare coverage are only those with permanent disabilities. DESCRIPTIVE RESULTS The study population on which this analysis is based represents an estimated 8.8 million older U.S. adults who were working full-time for pay in 2006 and were either insured primarily through an employer health plan that covered only the individual, or 2 In this instance, whether or not the insurance is family coverage is not important, as the insurance was not obtained through the respondent s employer, and respondent s wages would not be at risk for penalty. 20

26 were insured through another (non-medicare, non-medicaid) source or were uninsured. Nearly 60% of adults aged in this analysis received health insurance through their employers, which equals the percentage reported in Employer Health Benefits Survey for 2007 (Table 1). Notably, the mean age of the population was 57 years, 79% were white, and women comprised 56% of the sample. Additionally, more respondents resided in the South relative to any other one region (42%), and three-quarters of the population was overweight or obese (76%). Despite this, most considered themselves to be in good or very good health. Literature examining wage rates suggests that wages differ greatly depending on the characteristics of both the employee and the employer, including gender, education, and firm type (Groshen, 1990). The descriptive results of this analysis are supported by this literature. For example, while the average weekly wage for the population was $919, women earned an average weekly wage of $803, which was about $260 less than the wage of males who earned $1066 per week (Table 2). Wage was also split along racial/ethnic lines, with whites earning the highest weekly wage ($973), followed by those of the other racial/ethnic group, which is likely comprised mostly of Asians, Native Hawaiians and Pacific Islanders and American Indian/Alaska Natives ($867), blacks ($776), and Hispanics who earned an average of just less than $600 per week. Individuals who were overweight had the highest weekly wage of all weight categories ($992), followed by those of normal weight ($970), the obese ($813), and those who 21

27 were underweight ($743). Individuals who received health insurance through their employer earned only $20 more per week than those who did not. Another large difference in wages was found between the disabled and nondisabled ($556 and $743, respectively). This may be largely due to the differences in occupation or job industry between the disabled and non-disabled, however, these effects will be examined through multivariate analysis. As exhibited in Table 3, the majority of individuals in the study population were insured through the firm they worked for, with the lowest rates of ESI coverage seen for those of the other racial/ethnic group (44%) and Hispanics (49%). Individuals who were underweight, female, disabled, or black had the highest rates of ESI coverage (71%, 66%, 64%, and 63%, respectively). The percentages of males and females with ESI may be distorted as this variable excludes individuals whose employer plan also covers their family (family coverage), a large proportion of whom are likely to be men. Even more pronounced were the differences in weekly wage are when wage and insurance status are analyzed with one or more variables. Though Table 4 shows that the non-disabled with ESI earned consistently higher wages than the disabled with ESI, wages become more equivalent when both groups do not have ESI coverage through their employers. Table 5 shows that weekly wage differs widely by BMI, gender, and race/ethnicity. In most cases, individuals with ESI obtained through the employer earned higher weekly wages than their counterparts without ESI with a few exceptions: white 22

28 men of all weight categories without ESI obtained through their employer earned higher weekly wages than obese white men with ESI (except for the underweight for which there were no observations); obese and overweight black men without ESI earned higher wages than their counterparts with ESI; obese males and females of the other racial group without ESI earned higher weekly wages than their counterparts with ESI; and normal weight females from the other racial group without ESI earned higher weekly wages than those with ESI ($1,354 vs. $498). When looking across racial/ethnic groups within a BMI category, however, those with ESI obtained through their employer tended to earn higher wages than those without it, with the exception of overweight individuals without ESI who had higher weekly wages than those with ESI. Looking at the particular characteristics of the industries examined in this analysis, Table 6 shows that the majority of workers in eleven of thirteen job industries represented in this research were insured by single coverage employer-sponsored insurance through the firm they worked for. Percentages of ESI coverage ranged from 29% in the agriculture industry to 81% in non-durable manufacturing. Although, interestingly, the two industries with the lowest percentages of ESI agriculture and mining and construction had the highest average weekly wages, workers without ESI did not consistently have the highest (or lowest) wages across these two industries (Table 7). In the agriculture industry, workers with no ESI coverage had an average weekly wage that was over $3000 more than workers with ESI ($953 vs.$4,150) while in the 23

29 mining and construction industry, workers with ESI made an average of $1500 more than workers without ESI per week ($2,334 vs. $794). Table 1. Study Population: Distribution of Full-Time Workers Age 50-64, 2006 Age (n=1402) 57.0 years BMI (n=1402) Gender (n=1402) Obese 37% Male 44% Overweight 39% Female 56% Normal Weight 23% Race/Ethnicity(n=1402) Underweight 1% White 79% Disability Status (n=457) Black 10% Disabled 3% Hispanic 7% Non-Disabled 97% Other 3% Insurance Coverage (n=1402) Education(n=1402) ESI 59% Less than HS 9% No ESI 41% GED 4% Job Industry (n=1097) HS 28% Agriculture/Farming 1% Some College 31% Mining and Construction 4% College and Above 28% Manufacturing (non-durable) 6% Region (n=1380) Manufacturing (durable) 11% Northeast 14% Transportation 6% Midwest 25% Wholesale 5% South 42% Retail 9% West 19% Financial/Insurance/Real Est. 8% Other 0% Business Services 5% Health Status (n=1401) Personal Services 2% Poor 2% Entertainment/Recreation 1% Fair 12% Professional Services 37% Good 32% Public Administration 7% Very Good 38% Tenure (n=1396) 12 years Excellent 16% Wage (n=1402) $919/wk Notes: Weighted data. Body Mass Index (BMI) and Health Status are self-reported. Racial groups are non-hispanic. Employer-sponsored insurance (ESI) refers whether or not respondent has health insurance through their own employer. Percentages may not sum to 100% due to rounding effects. Observations: 8,836,204. Data: Health and Retirement Study,

30 Table 2. Average Weekly Wage by Select Characteristics, 2006 Weekly Wage Gender Male $1066 Female $803 Race/Ethnicity White $973 Black $776 Hispanic $568 Other $867 BMI Obese $813 Overweight $992 Normal Weight $970 Underweight $743 Disability Status Disabled $743 Non-Disabled $556 Insurance Coverage ESI $927 No ESI $908 Notes: Weighted data. Body Mass Index (BMI) and Health Status are self-reported. Racial groups are non-hispanic. Employer-sponsored insurance (ESI) refers whether or not respondent has health insurance through their own employer. Observations: 8,836,204. Data: Health and Retirement Study,

31 Table 3. Percent of Full-Time Workers with ESI by Select Characteristics, 2006 % ESI Coverage Gender Male 51% Female 66% Race/Ethnicity White 61% Black 63% Hispanic 49% Other 44% BMI Obese 60% Overweight 60% Normal Weight 60% Underweight 71% Disability Status Disabled 64% Non-Disabled 58% Notes: Weighted data. Body Mass Index (BMI) and Health Status are self-reported. Racial groups are non-hispanic. Employer-sponsored insurance (ESI) refers whether or not respondent has health insurance through their own employer. Observations: 8,836,204. Data: Health and Retirement Study, Table 4. Average Weekly Wage by Disability and Coverage Status, 2006 ESI No ESI Disabled (n=17) $557 $554 Non-Disabled (n=439) $808 $625 Notes: Weighted data. Employer-sponsored insurance (ESI) refers whether or not respondent has health insurance through their own employer. Observations: 2,761,259. Data: Health and Retirement Study,

32 Table 5. Average Weekly Wage by Insurance Status, BMI, Gender, and Race/Ethnicity, 2006 Obese Overweight Normal Weight Underweight Total White Total ESI no ESI Total ESI no ESI Total ESI no ESI Total ESI no ESI Male $926 $923 $929 $1,314 $1,222 $1,420 $1,210 $1,133 $1,317 NSD NSD NSD $1,159 Female $752 $790 $665 $831 $878 $726 $915 $950 $860 $943 $1,529 $285 $828 Total $826 $836 $812 $1,090 $1,041 $1,166 $1,006 $1,004 $1,008 $943 $1,529 $285 $973 Black Male $918 $809 $1,012 $645 $627 $671 $863 $1,074 $737 NSD NSD NSD $758 Female $910 $1,117 $459 $658 $680 $587 $683 $842 $476 $415 $415 NSD $788 Total $912 $1,056 $673 $650 $652 $646 $736 $891 $573 $415 $415 NSD $776 Hispanic Male $552 $662 $428 $585 $586 $585 $802 $1,058 $601 NSD NSD NSD $594 Female $566 $685 $426 $507 $620 $372 $553 $637 $441 NSD NSD NSD $539 Total $558 $734 $405 $549 $605 $504 $646 $770 $511 NSD NSD NSD $568 Other Male $1,080 $935 $1,111 $866 $980 $801 $1,108 $1,111 $1,107 NSD NSD NSD $970 Female $445 $320 $529 $787 $913 $457 $872 $498 $1,354 NSD NSD NSD $707 Total $749 $497 $853 $836 $944 $728 $1,036 $870 $1,166 NSD NSD NSD $867 Total $813 $849 $760 $992 $968 $1,027 $970 $984 $950 $743 $932 $285 $919 Notes: Weighted data. Body Mass Index (BMI) and Health Status are self-reported. Racial groups are non-hispanic. Employer-sponsored insurance (ESI) refers whether or not respondent has health insurance through their own employer. NSD: Not Sufficient Data. Observations: 8,836,204 Data: Health and Retirement Study,

33 Table 6. Percent of Full-Time Workers with ESI by Job Industry, 2006 % ESI Coverage Agriculture/Farming 29% Mining and Construction 48% Manufacturing (non-durable) 81% Manufacturing (durable) 73% Transportation 63% Wholesale 59% Retail 65% Financial/Insurance/Real Est. 62% Business Services 52% Personal Services 54% Entertainment/Recreation 70% Professional Services 65% Public Administration 74% Notes: Weighted data. Employer-sponsored insurance (ESI) refers whether or not respondent has health insurance through their own employer. Observations: 6,936,384. Data: Health and Retirement Study,

34 Table 7. Average Weekly Wage by Job Industry and Coverage Status, 2006 Weekly Wage ESI No ESI Total Agriculture/Farming $953 $4,150 $3233 Mining and Construction $2,334 $794 $1531 Manufacturing (non-durable) $986 $1,342 $1053 Manufacturing (durable) $942 $759 $892 Transportation $875 $1,197 $993 Wholesale $962 $1,179 $1052 Retail $587 $678 $619 Financial/Insurance/Real Est. $1,040 $962 $1010 Business Services $1,050 $888 $971 Personal Services $745 $831 $785 Entertainment/Recreation $803 $482 $707 Professional Services $903 $962 $923 Public Administration $992 $1,182 $1041 Total $946 $1,015 $970 Notes: Weighted data. Employer-sponsored insurance (ESI) refers whether or not respondent has health insurance through their own employer. Observations: 6,936,384. Data: Health and Retirement Study, REGRESSION RESULTS This research seeks to examine variation in weekly wage for obese and physically disabled full-time workers who obtain employer-sponsored health insurance through their own employer relative to obese and physically disabled individuals who do not. The multivariate regression results in Table 8 show the effect of worker characteristics on the weekly wages of 1,072 full-time workers aged in the United States, representing an 29

35 estimated 6.8 million Americans. Controlling for several employee and employer-based determinants of wage, including gender, age, race/ethnicity, job industry, and tenure, this model is able to explain about 40 percent of the variation in weekly wages for these individuals. Unless otherwise noted, only differences that are significant at the p<.05 level or higher are discussed in the text. Race/ethnicity had a large, significant impact on wages. Holding all else constant, blacks earned 16.2% less than whites, and Hispanics earned 30.1% less than whites in No statistical difference was found between the wages of those of the other racial group and whites. Gender similarly had a significant effect on weekly wage, with female employees earning 14.8% less than males. The region of residence was also found to significantly impact wages in Relative to the Northeast, residents of the Midwest earned 19.2% less per week, and Southerners earned 12.9% less. The wages of workers from the West and those from the other region did not significantly differ from the wages of Northeasterners. As expected, education had a strong impact on wage, with higher levels of education yielding higher wages. While the wages of individuals with a GED did not differ statistically from the wages of those with less than a high school degree, individuals who graduated high school earned 15.4% more than those with less than a high school degree, workers with some college education earned 38.2% more, and those with a college degree or more education earned nearly 75% more per week. 30

36 Self-reported health status also affected wage, though better health status did not consistently result in higher wages. Controlling for other covariates, employees who reported their health status as very good earned 15.7% less than those who reported excellent health, and those reporting good health earned 19.0% less. Interestingly, adults reporting poor health made 24.5% less than those who reported excellent health, but those who reported fair health status made 26.0% less. Although the wages between these two categories differed slightly, as this is a self-reported measure, there may be little actual difference in the health status of individuals reporting in each group. Job industry was found to be a negative, but non-significant factor in determining wages across all industries included in this analysis. Tenure, however, was a highly significant, yet moderately strong factor in affecting wages. For every additional year of tenure at a workers current job, wages increased by 1.0%. Though not shown in the tables, 17 occupation categories were also included in an iteration of this regression model. While occupation explained an additional 7.6% of the variation in wages, and most of the 17 occupation categories included in the model were significant, the inclusion of the variable did not drastically impact results, and was ultimately excluded from the final analysis due to concerns of over-specification. Of the key variables of interest in this research, neither BMI nor insurance significantly affected wage. Additionally, the sample size for the disabled in this analysis (n=17) precluded the inclusion of the variable in multivariate regression models, and thus 31

37 the effect of disability on the wages of those with employer-sponsored insurance could not be studied. BMI, run as both a continuous and categorical variable, was not significantly associated with wage in the multivariate model (Table 9). Though the signs of the coefficients were negative in the model with BMI run as a continuous variable (-0.57%) and in the model that included a dummy variable for obesity (-7.8%), the nonsignificant coefficients indicate that, among this population in 2006, weight did not have a significant impact on weekly wage. Similarly, the coefficients for overweight and normal weight were positive but also not significant (1.2% and 7.9%, respectively), suggesting that the wages of overweight and normal weight individuals did not differ from the wages of employees of other weight classes. Due to small sample size, the dummy variable for underweight was dropped from the model. Health insurance status, the other key variable of interest, was positive yet nonsignificant in the regression model (Table 8 and Table 9). This suggests that when controlling for other factors, weekly wage was similar throughout the study population whether or not employees obtained health insurance through the firm they worked for. The interaction terms for weight and insurance included in these models were also not significant, indicating that wage was not differentially affected by these particular combinations of insurance and weight. 32