Katherine Baicker
Helen Levy
ABSTRACT
Employer health insurance mandates form the basis of many health care reform proposals. Proponents make the case that they will increase insurance, while opponents raise the concern that low-wage workers will see offsetting reductions in their wages and that in the presence of minimum wage laws some of the lowest wage workers will become unemployed. We construct an estimate of the number of workers whose wages are so close to the minimum wage that they cannot be lowered to absorb the cost of health insurance, using detailed data on wages, health insurance, and demographics from the Current Population Survey (CPS). We find that 33 percent of uninsured workers earn within $3 of the minimum wage, putting them at risk of unemployment if their employers were required to offer insurance. Assuming �in elasticity of employment with respect to minimum wage increase of -0.10, we estimate that 0.2 percent of all full-time workers and 1.4 percent of uninsured full-time workers would lose their jobs because of a health insurance mandate. Workers who would lose their jobs are disproportionately likely to be high school dropouts, minority, and female. This risk of unemployment should be a crucial component in the evaluation of both the effectiveness and distributional implications of these policies relative to alternatives such as tax credits, Medicaid expansions, and individual mandates, and their broader effects on the well-being of low-wage workers.
INTRODUCTION
Employer health insurance mandates form the basis of many health care reform proposals. Democratic presidential candidates Hillary Clinton, John Edwards, and Barack Obama have all proposed reforms that include pay or play mandates. Individual states are contemplating these mandates as well. In California, for example, the state legislature passed a law requiring employers above a certain size to provide a specified package of health benefits for their workers. California voters narrowly overturned the measure (Proposition 72) in 2004, but Governor Schwarzenegger recently unveiled a new plan to expand insurance coverage that requires employer contributions. Oregon and Washington enacted mandates that were later repealed. Hawaii implemented an employer mandate in 1974. The recent reform in Massachusetts, which combines an individual mandate, employer requirements, redirection of Medicaid funds, and the creation of a new insurance pooling mechanism, has garnered much attention and may spur similar reforms in other states.
The proponents of these measures make the case that they will increase insurance coverage while maintaining the role of the market in generating competition and efficiency in health insurance offerings. Opponents raise the concern that low-income workers will see offsetting reductions in their wages and that in the presence of minimum wage laws some of the lowest wage workers will become unemployed. Academics and the popular press alike cite increased health insurance costs as one of the causes of recent increases in unemployment (Porter, 2004). Estimates of the potential job loss from the mandates included in the failed Clinton health care proposal ranged from 600,000 to more than 2,000,000.
To determine how important the potential job loss from employer mandates is, we need to know how many workers are likely to be affected. Several factors affect the degree to which employer mandates will cause unemployment. First, what is the likely cost of the mandated health insurance? This clearly depends on the specifics of the mandated coverage.1 Second, how much of an increase in the cost of employing workers is borne by employees in the form of reduced wages? There is substantial evidence that the cost of health insurance mandates will be shifted to employees, resulting in lower wages.2 Third, how many workers not currently covered by employer-sponsored insurance are so close to the rninimum wage that their wages cannot be lowered enough to offset the cost of the new mandate? We focus on the last question. This article provides an estimate of how big the pool of workers at risk of unemployment is likely to be and what characteristics they are likely to have, taking into account mininium wage laws and patterns of employer health insurance offering and coverage.
We construct an estimate of the number of workers whose wages are so low that they cannot be lowered to absorb the cost of health insurance, using detailed data on wages, health insurance, and demographics from the Current Population Survey (CPS). We characterize the population of workers at risk in terms of their sociodemographic characteristics (age, race, gender, education, family structure), and industry of employment. We find that 33 percent of uninsured workers earn within $3 of the minimum wage, putting them at substantial risk of unemployment if their employers were required to offer insurance. These workers are disproportionately likely to be high school dropouts or racial minorities. Understanding which workers these laws are likely to affect should play an important role in the assessment of the effect of employer mandates on the level and distribution of employment and insurance coverage.
BACKGROUND
The estimated impact of an employer health insurance mandate on insurance coverage and employment depends on two sets of factors: (1) the specifics of the mandate and (2) what one assumes about the dynamics of wages, fringe benefits, and employment.
Specific mandate proposals vary widely from state to state.3 Most include exemptions for smaller firms (e.g., those with fewer than 20 employees in California) and for employees with few hours (e.g., fewer than 20 hours per week in Hawaii, or 100 hours per month in California). Most include minhnum employer contributions (such as 80 percent of premiums in California, or 75 percent for employees in Oregon) and minimum coverage requirements (benchmarked to other plans offered in the state in Hawaii; including prescription drugs and preventive care in California). Three of these features are likely to be particularly important for the analysis of any particular mandate. First, which employers and employees are affected? Any exemptions, such as those for small firms or part-time workers, will dilute both the positive and negative effects of a mandate. second, what is the marginal cost of the newly mandated benefits, both in terms of specific benefits and in terms of lost flexibility for employers? A mandate can specify a generous benefits package that all employers must provide (thus increasing costs for some employers already providing insurance), or it can require minimal coverage that affects only employers who do not already provide insurance. Third, what fraction of these costs must nominally be borne by the employer? When nominal wage rigidities prevent accommodation of increased costs through reduced wages, the statutory incidence may have a substantial effect. Policies that require firms to offer insurance but not pay for it would likely have little effect on rates of coverage because uninsured workers do not appear to be very responsive to the availability of benefits unless they are very heavily subsidized (Chernew et al., 1997; for a review of the recent literature on price elasticities of demand for health insurance among uninsured workers, see Gruber and Washington, 2003).4
The second set of issues-what assumptions one maintains about the dynamics of wages, fringe benefits, and employment-comes into play when a significant share of the cost of the newly mandated health benefits falls on employers. There is a consensus among most economists that these costs, like the cost of any fringe benefit that workers value, will be passed on to workers in the form of reduced wages whenever possible (see Gruber and Krueger, 1991; Gruber, 1994; Fishback and Kantor, 1995; Olson, 2002). The implication of this is that when an insurance mandate accomplishes its stated goal of extending coverage to a previously uninsured worker, that worker will also experience a reduction in her wage or the growth of her wage relative to what would have happened otherwise. In the best-case scenario, the worker's wage will be sufficiently high to absorb the entire cost of the benefit, and the mandate will have changed the composition of compensation (less wages, more benefits) but not the total value of compensation.
The problem arises when the worker's wage is not high enough to absorb this cost without bumping into the minimum wage. When this is the case, the insurance mandate has the same effect on employment as an increase in the minimum wage. Suppose, for example, that an uninsured worker earning the minimum wage becomes subject to an insurance mandate that requires the employer to provide benefits that cost $1 per hour worked. Since there is no scope to reduce wages, the hourly cost of employing the worker is now the minimum wage plus $1. Economists have long believed that this is likely to result in lower employment, as employers substitute machines for workers when workers become more expensive. The size of this 'elasticity' of employment with respect to the minimum wage has been the subject of considerable recent controversy: there is little consensus on the magnitude of the unemployment effect associated with an increase in the minimum wage (see Brown, 1999, for a review). Regardless of one's beliefs about the employment effect of minimum wage increases, however, the employment effect of an employer health insurance mandate that increases employer costs ought to be the same as the effect of a change in the minimum wage. In the analysis that follows, we present estimates of the population at risk of being affected by the imposition of employer mandates, to which different estimates of the elasticity of employment with respect to changes in the minimum wage can be applied. Our analysis shows how many uninsured workers are within different ranges of the minimum wage (such as within $3), so that readers can consider mandates that impose different levels of cost on employers and a range of estimates of the effect of changes in the minimum wage on employment.
DATA AND METHODS
The primary data for analysis come from the CPS, conducted annually by the Bureau of the Census. The CPS collects information from about 50,000 households each month about household composition, sociodemographic characteristics, earnings, and employment in eight different monthly surveys over the course of 16 months. Each month's survey provides detailed demographic data such as age, race, education, marital status, and family composition of respondents. The March survey in each year also collects information about health insurance coverage. We combine these variables with information provided by respondents about their labor force status, whether or not they are paid hourly, usual hours worked, and wages in an exit ('outgoing rotation') interview.5 We use the standard use data from 2000 to 2006, the most recent CPS survey available. We restrict our sample to respondents' age from 22 to 65.
To these data we add information on the minimum wage, which varies by state and over time (see Nelson 2000, 2001, 2002, 2003; Nelson and Fiztpatrick 2004; Fitzpatrick 2005,2006,2007; Fiscal Policy Institute, 2004, for details of state law changes, also shown in Appendix Table Al). While the federal minimum wage was $5.15 throughout our study period, several states enacted minimum wages that were higher than the federal minimum, so workers and employers in these states faced a higher minimum wage. We then compare workers' wages to the minimum wage in effect in January in their state and year (which corresponds best with the period from which respondents in the CPS report their wages).
We also use information on health insurance premiums by state, year, and policy type (family or single) collected by the Medical Expenditure Panel Survey from 1999 to 2005. We merge these data with the individual observations from the CPS for those years (using the previous year's survey to most closely match the timing of the CPS questionnaire) to impute a health insurance premium for each observation, attributing family policy premiums to those with a spouse or children and single policy premiums to those without.6 We deflate all dollar amounts to year 2006 dollars using the Consumer Price Index (CPI).
Together, these data allow us to estimate both the likely effect of different employer mandates on wages and employment, as well as the distributional implications for workers with different characteristics. In the analysis that follows, we aggregate data from the CPS across years, and report workers' insurance status, wages relative to the minimum wage, and various demographic characteristics such as age, race, marital status, and education. We use the weights provided in the CPS so that the numbers and proportions we report are representative of the full-time private sector workforce as a whole. see the Appendix Tables that follow for more detail.
RESULTS
We use these data to estimate which workers would be at risk of unemployment with the imposition of employer mandates. We present data on the health insurance and wage distribution of all workers, as well as different demographic subgroups, focusing in particular on workers with wages close to the minimum wage since it is these workers whose wages may have the least flexibility to be lowered in response to mandates that make employing them more costly, and thus may be most likely to face adverse employment consequences.7 We focus our analysis on workers employed more than 20 hours per week, as those with fewer hours are likely to be exempt from employer mandates.8 Much more detailed data are shown in the Appendix tables that follow.
Workers at Risk
More than 15 percent of private sector workers employed more than 20 hours a week (whom we call 'full-time') are currently uninsured (Table 1). Note that our estimate of uninsured workers includes those who decline insurance offered to them by their employers, but does not include workers who get insurance from a source other than their own employer.9
Who are these uninsured workers? They are more than three times as likely to be high school drop-outs as insured workers, and twice as likely to be from a minority racial or ethnic group. They are 50 percent more likely to be under the age 35 years and to be unmarried. They are almost twice as likely to be single parents (Table 2).
Uninsured workers are thus demographically quite different from insured workers. Several of these characteristics make them economically vulnerable-and also make them the target population for policies intended to expand health insurance coverage.
Many of the employer mandates being considered by different states exempt small firms. More than 55 percent of all uninsured workers are employed in firms with more than 25 employees (compared to more than 80 percent of insured workers)-which means that they would be covered by many proposed mandates (Table 3).10 Of course, mandates that cover only firms above this size would have commensurately smaller effects both on insurance coverage and on the risk of decreasing employment than proposals without such limitations.
Benchmark Insurance Costs
How likely these uninsured workers are to face unemployment depends on whether the minimum wage is binding-that is, if the hourly cost per worker of newly mandated health insurance is greater than the gap between the worker's wage and the minimum wage. While a more detailed calculation requires knowledge of (or assumptions about) workers' family structure, health status, the elasticity of labor supply and demand, workers' valuation of health insurance benefits, long-run labor market dynamics (such as substitution toward part-time employees), and the like, we calculate several informative back-of-the-envelope benchmarks using aggregate insurance costs (Table 4). The average annual premium for employer-sponsored health insurance in our data was approximately $9,046 for family coverage and $3,429 for single coverage (for the period 2000-2006, expressed in 2006 dollars), for an average hourly premium of $3.66 for a full-time worker.11 If employers were required to pay 80 percent of premiums, the average hourly wage for this group of workers would thus have to decrease by about $3 to absorb fully the cost of providing the average health insurance package. Here, clearly, the costs would be different if the mandated insurance coverage were more or less generous than the typical plan already provided to most workers or if workers were required to pay more of the premium directly.
We also calculate a more sophisticated benchmark based on the insurance cost facing individual workers, rather than a broad average. We impute the insurance cost for each worker based on state of residence, year, and family structure, divide that number by 2,000 to generate an average hourly cost of insurance, and compare the difference between hourly wages and the minimum wage to that hourly insurance cost.
The Role of the Minimum Wage
A large fraction of uninsured workers earn little more than the minirnum wage. Insurance costs potentially represent an enormous increase in the minimum compensation for this group of workers. The federal minimum wage is $5.15, and the average minimum wage in our sample (taking into account state minimums that are sometimes higher) is $5.98-so the benchmark cost of $3 represents 50 percent of the effective minimum wage.12 There is clearly a great deal of disagreement about the effect of ininimum wages on employment, but even under relatively conservative elasticity estimates this could result in significant effects on rninimum wage workers.
Uninsured workers earning within $3 of the rninimum wage represent 5 percent of the workforce and a third of all uninsured workers (Table 5). (Using the more sophisticated benchmark based on individual insurance costs yields answers very similar to the $3 benchmark, both of which are reported in the Appendix tables.)
Figure 1 shows a more detailed distribution of the hourly wages of uninsured workers relative to the minimum wage.
Thus, while the overall fraction of private sector workers who are 'at risk' is moderate, since only 5 percent of all workers are uninsured workers earning within $3 of the minimum wage, a potentially very large fraction of the group supposedly targeted for help by employer mandates might in fact be hurt, since 33 percent of uninsured workers earn within $3 of the rnininium wage. So, of the roughly 114 million United States private sector workers, 105 million of whom work more than 20 hours per week, 16 million are uninsured, and more than 5 million of those earn within $3 of the minirnum wage.13
As Table 2 suggested, low-skilled workers are more likely to be uninsured. Figure 2 shows this wage distribution for workers with different levels of education. Workers with less than a high school degree are significantly more likely to have earnings close to the minimum wage.
Thus, among the uninsured, those with the least education face the highest risk of losing their jobs under employer mandates. The same is true for nonwhites, those under age 35, single parents, and women (as seen in Appendix Table A4).
Potential Job Loss
How many of those workers are likely to lose their jobs? We calculate an approximate answer to this question in the following way. First, we compare the individual-specific hourly insurance costs described above to the cushion between an uninsured worker's wage and the nunimum wage. If a worker's wage is sufficiently high that it can adjust downward by the full cost of insurance without hitting the minimum wage, we assume this worker is not at risk of losing her job. If, however, the minimum wage constraint binds, we calculate the percentage increase in total compensation implied by the health insurance mandate. For example, if a worker earning $6 per hour is mandated to have health insurance costing the firm $2 per hour, we assume that her wage will adjust downward by 85 cents to the minimum wage of $5.15. However, the remaining $1.15 of the cost of the mandate cannot be absorbed by reducing wages and increases her total compensation to $7.15-an increase in compensation of almost 20 percent ($1.15/$6.00 = 0.19). Assuming an employment elasticity with respect to the minimum wage of -0.1, meaning that a 10 percent increase in the minimum wage would lead to a 1 percent reduction in employment, this worker has a 2 percent chance of losing her job.14 Performing a similar calculation for all the workers in our sample suggests that about 224,000 workers would lose jobs as a result of a mandate with these costs (Table 6). More than 60 percent of these workers would be racial or ethnic minorities and about one-third would have less than a high school education. The burden of the mandate would thus fall disproportionately on these groups since, for example, racial and ethnic minorities are only 30 percent of the workforce in this sample.
To the extent that mandates impose additional costs on firms (such as reduced flexibility or more generous coverage than they were already offering), these figures represent a lower bound on the increase in unemployment likely to result from such mandates. As noted above, if mandates apply only to some workers this will dilute both the positive and negative effects of a mandate. For example, establishments with fewer than 25 workers employ 44 percent of uninsured workers (Table 3); if these small employers were exempted from a mandate, our estimate of job loss would drop to about 45 percent of the number above.
Regional Variation
These results are not confined to any particular area of the country. As Appendix Table 7 shows, the Northeast, Midwest, South, and West have very similar fractions of workers at risk for unemployment. Looking at individual states shows that there is local variation in this at-risk pool, however (although sample size limits our ability to compare individual states).
Individual states should be more concerned with employment effects of their own minimum wage laws and health insurance mandates than the federal government, since firms and jobs may move across state lines if nearby states place fewer constraints on employers.
DISCUSSION
Understanding the labor market consequences of employer mandates is a key component in evaluating their effectiveness relative to other policies such as tax credits, Medicaid expansions, and individual mandates. Several studies have analyzed the effect of different versions of employer mandates on insurance premiums and on workers' wages. This study contributes an important missing piece to the analysis: how large is the potential risk of unemployment? Our analysis suggests that one-third of the targeted population of uninsured workers have hourly wages close enough to the minimum wage that employers will not be able to lower their wages enough to accommodate fully the increase in compensation costs that employer mandates would impose. These workers, who tend to be disproportionately low education, minority, and female, thus face a risk of unemployment. This risk of unemployment should be a crucial component in the evaluation of both the effectiveness of these policies in reducing the number of uninsured and their broader effects on the well-being of low-wage workers.
1 Yelowitz ('The Cost of California's Health Insurance Act of 2003,' EPI, 2003), for example, shows that costs and benefits of California's law depend crucially on the subsidy for low-income workers, the generosity of the plan required to fulfill the play or pay requirements, etc. see also Zedlewski et al., 'Play-or-Pay Employer Mandates: Potential Effects,' Health Affairs, Spring 2002; and Krueger and Reinhardt, 'The Economics of Employer Versus Individual Mandates,' Health Affairs, Spring 1994; California Health Care Foundation (2004); Chollet (1987).
2 See, for example, Jonathan Gruber and Alan Krueger, 'The Incidence of Mandated EmployerProvided Insurance: Lessons From Workers' Compensation Insurance,' Tax Polin/and the Economy, 1991; Norman Thurston, 'Labor Market Effects of Hawaii's Mandatory EmployerProvided Health Insurance,' Industrial and Labor Relations Review, October, 1997; Currie and Madrian (2000).
3 Yelowitz (2004) illustrates the importance of understanding the specifics of California's proposed mandate in order to estimate the proposal's cost.
4 If workers are required to take up the insurance, the degree to which workers value the benefits and the elasticity of labor supply and demand would determine the ultimate effect on wages (and the 'incidence' of the mandate)-as discussed below. See Summers (1989) for a discussion of how worker valuation affects the incidence of mandated benefits.
5 We are able to match just over 70 percent of March respondents to their corresponding exit ('outgoing rotation') interview. Reasons for failing to find a respondent across months include household mobility, nonresponses, and noise in the identifiers. There is also a known decline in match quality following the expansion of the CPS sample size in 2002 (driven in part by the way that household identifiers were assigned to the new sample). We household and person identifiers to match across months, and then screen for match quality using respondent demographics (such as age and gender), based on methodology outlined in Madrian and Lefgren (2000). About 5.1 percent of observed 'matches' appear to be false, and these observations are dropped.
6 The health insurance questions in the March CPS refer to coverage in the previous calendar year. Swartz (1986) presents evidence that people actually respond to these questions as if they were reporting their coverage at the time of the survey.
7 While hourly workers may be more susceptible to binding minimum wages than salaried workers, minimum wage laws apply to almost all salaried workers as well. We impute an hourly wage for those workers on salary using the usual hours worked per week and weekly wages from the CPS. Workers paid hourly are much more likely to be close to the minimum wage than those paid on salary, but we include both in our analysis.
8 Many proposed mandates only apply to full-time workers. Employers might thus have the incentive to substitute away from full-time employees toward part-time employees. We ignore these dynamics. We are also implicitly assuming here that wages adjust independently of whether workers would have taken up insurance or not-insofar as there is no mechanism for employers to know ahead of time (when offering a wage and insurance package) whether a worker is going to take up that coverage or not.
9 Implicitly, we are assuming that the wages of workers who turned down have not already adjusted downward by the cost of the insurance that they declined. Analysis of the February Contingent Work Supplements to the Current Population Survey in 1995, 1997, 1999, 2001, and 2005 shows that about one-quarter of uninsured workers were offered insurance. We also assume that workers with coverage from another source, which is typically a spouse's employersponsored policy, would not be affected by mandates.
10 It is not clear how accurate employees' reports of their establishment size are.
11 These data are consistent with other survey results on insurance costs, suggesting that our algorithm for assigning premiums in our sample is representative. For example, the Kaiser/HRET survey reported average employer premiums in 2006 of $11,500 for family policies and $4,200 for single policies, while in our CPS/MEPS sample the average premium in 2006 was $10,700 for family policies and $4,000 for single policies.
12 On average, wages represent about 70 percent of compensation in the private sector, with health insurance costs accounting for an additional 7 percent, other voluntary fringe benefits accounting for 14 percent, and legally required benefits (such as social security) accounting for the remaining 9 percent (Department of Labor 2007). In theory, then, employers might respond to insurance mandates by reducing other fringe benefits. Low-wage workers are less likely than the typical worker to have these other benefits, however (Schwabisch 2004), so it is unclear in practice how much of a buffer other benefits provide.
13 Bureau of Labor Statistics series CES0500000001 (total private employment) is 113,753,000 in March 2006 and about 114 million in other months of 2006 also.
14 This is a relatively conservative estimate of the sensitivity of employment to minimum wage laws. See Brown (1999) for a review of the wider range of estimates of this elasticity.
REFERENCES
Agency for Healthcare Research and Quality, Medial Expenditure Panel Survey-Insurance Component, US Department of Health and Human Services, 1999-2005.
Baicker, K., and A. Chandra, 2006, The Labor Market Effects of Rising Health Insurance Premiums, Journal of Labor Economics, 24(3).
Brown, C, 1999, Minimum Wages, Employment, and the Distribution of Income in: O. Ashenfelter and D. Card eds., Handbook of Labor Economics, 3B (Holland: Elsevier Science).
Bureau of Labor Statistics, 2006, Current Employment Statistics, U. S. Department of Labor.
California Health Care Foundation, 2004, State Employer Health Insurance Mandates: A Brief History, http://www.chcf.org/documents//EmployerInsuranceMandates.pdf (accessed October 31,2004).
Chernew, M., K. Frick, and C. McLaughlin, 1997, The Demand for Health Insurance Coverage by Low-Income Workers: Can Reduced Premiums Achieve Full Coverage? Health Services Research.
Chollet, D., 1987, Public Policy Options to Expand Health Insurance Coverage Among the Nonelderly Population, In Government Mandating of Employee Benefits, Employee Benefits Research Institute, (accessed October 31, 2004).
Currie, J., and B. C. Madrian, 2000, Health, Health Insurance and the Labor Market, in: O. Ashenfelter and D. Card eds., The Handbook of Labor Economics, 3 C (Holland: Elsevier Science).
Fiscal Policy Institute, 2004, State Minimum Wages and Employment in Small Businesses, (accessed October 31, 2004).
Fishback, P., and S. E. Kantor, 1995, Did Workers Gain From the Passage of Workers' Compensation Laws? Quarterly Journal of Economics, UO: 713-742.
Fitzpatrick, J. J., 2005, State Labor Legislation Enacted, Monthly Labor Review, January.
Fitzpatrick, J. J., 2006, State Labor Legislation Enacted, Monthly Labor Review, January..
Fitzpatrick, J.J., 2007, State Labor Legislation Enacted, Monthly Labor Review, January.
Gruber, J., 1994, The Incidence of Mandated Maternity Benefits, American Economic Review, 622-641.
Gruber, J., and A. Krueger, 1991, The Incidence of Mandated Employer-Provided Insurance: Lessons from Workers' Compensation Insurance, in: D. Bradford ed., Tax Policy and the Economy, vol. 5 (Cambridge, MA: MIT Press), pp. 111-144.
Gruber, J., and E. Washington, 2003, Subsidies to Employee Health Insurance Premiums and the Health Insurance Market, National Bureau of Economic Research Working Paper 9567.
Kaiser/HRET, Survey of Employer-Sponsored Health Benefits, Various years.
Krueger and Reinhardt, 1994, The Economics of Employer Versus Individual Mandates, Health Affairs, Spring.
Madrian, B. C, and L. J. Lefgren, 2000, An Approach to Longitudinally Matching Population Survey (CPS) Respondents, Journal of Economic and Social Measurement, 26(1): 31-62.
Nelson, R., 2000, State Labor Legislation Enacted, Monthly Labor Review, January.
Nelson, R., 2001, State Labor Legislation Enacted, Monthly Labor Review, January.
Nelson, R., 2002, State Labor Legislation Enacted, Monthly Labor Review, January.
Nelson, R., 2003, State Labor Legislation Enacted, Monthly Labor Review, January.
Nelson, R., and J. J. Fitzpatrick, 2004 State Labor Legislation Enacted, Monthly Labor Review, January.
Olson, C, 2002, Do Workers Accept Lower Wages in Exchange for Health Benefits? Journal of Labor Economics, 20(2): S91-S114.
Porter, E., 2004, Rising Cost of Health Benefits Cited as Factor in Slump of Jobs, New York Times, August 19.
Schwabisch, J. A., 2004, Accounting for Wages and Benefits Using the ECI, Monthly Labor Review, September: 26-41.
Summers, L. H., May 1989, Some Simple Economics of Mandated Benefits, American Economic Review, 177-183.
Swartz, K., 1986, Interpreting the Estimates From Four National Surveys of the Number of People Without Health Insurance, Journal of Economic and Social Measurement, 14: 233-242.
Thurston, N., 1997, Labor Market Effects of Hawaii's Mandatory Employer-Provided Health Insurance, Industrial and Labor Relations Review, October.
United States Bureau of the Census, Current Population Survey, various years.
United States Department of Labor, 2007, Employer Costs for Employee Compensation, Publication, September: 07-1434.
Yelowitz, A., 2003, The Cost of California's Health Insurance Act of 2003, Employment Policies Institute.
Yelowitz, A., 2004, The Economic Impact of Proposition 72 on California Employers, Employment Policies Institute.
Zedlewski, S., G Acs, and C. Winterbottom, 2002, Play-or-Pay Employer Mandates: Potential Effects, Health Affairs, Spring.
Katherine Baicker is professor of Health Economics, Department of Health Policy and Management, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115; e-mail: kbaicker@hsph.harvard.edu . Helen Levy works with the University of Michigan, Economic Research Initiative on the Uninsured, 555 S. Forest Street, Ann Arbor, MI48104; phone: (734) 6159587; e-mail: hlevy@uminch.edu . We are grateful to participants at the conference 'Healthcare Reform: The Economics of Pay-or-Play Mandates' (Washington, DC, September 14,2007), especially Jared Bernstein and Elise Gould for their very helpful comments and suggestions. Financial support of an earlier version of this work from the Employment Policies Institute is gratefully acknowledged.
COMMENT ON 'EMPLOYER HEALTH INSURANCE MANDATES AND THE RISK OF UNEMPLOYMENT'
Jared Bernstein
Elise Gould
Katherine Baicker and Helen Levy (hereafter, BL) provide readers with an accessible and useful analysis of the potential employment effects of an employer mandate for health insurance. While we offer numerous critiques of the study, its clarity is a virtue, and the simulation model the authors present in this study may well prove to be of use to others interested in future extensions of the work.
BL set out to predict the impact on the employment of low-wage workers of an employerside health care mandate. Why low-wage workers? Because in the presence of minimum wages, employers will not be able to fully trade off wages for health care benefits. The logic here is simple: employers can painlessly pay the full cost of the mandate by simply trading off health costs for wages, dollar for dollar. At least, they can do so until they run into the minimum wage. At that point, the employer cannot complete the trade-off and something else has got to give. BL assume that something is employment, although as we argue below, there are other absorption mechanisms, which they ignore.
Using a negative elasticity from the minimum wage literature-the percentage of employment reduction with respect to the increase in the minimum-BL apply this wage margin (the dollar amount that the employer was prevented from trading off) to the elasticity and derive a count of jobs lost. They also look at the characteristics of laid-off workers.
FINDINGS AND ASSUMPTIONS
Our first point is that if BL are right, and we offer some reasons to question their fundamental assumptions, the number of displaced workers looks to us to be very small, both in terms of actual magnitude and in terms of the number who benefit from the mandate. They estimate that 190,000 workers will lose their jobs, a tiny fraction of even the low-wage sector of the labor market. The latter implies that millions more will continue to work but will now benefit from health coverage paid for at least in part by their employer (as assumed by the authors-i.e., since the employer can not trade off the full cost of the mandate, she has to absorb it herself).
According to BL's Table 5, there are about 100 million workers who work at least 20 hours per week (one of BL's sample-selection requirements, since they assume those who work less will be exempted from the mandate), so there are 20 million in the bottom fifth of the wage scale, one definition of the low-wage work force. Thus, the mandate leads to job loss of just under 1 percent of the low-wage workforce, by this definition. Note also that in 2006, net job gains per month average about 190,000. Obviously, one does not want to see any low-wage worker hurt by a policy intended to help them, but no social policy can avoid unintended consequences, and this strikes us as a very mild cost given the benefits of the mandate. So, our first point is if BL are right, policy makers might want to grab this deal.
But while we agree with the simple logic, we question whether simple logic is correct in this case. There are two strains of literature that we believe raise tough questions about their two main assumptions: the wage trade-off and the negative wage elasticity. We also raise margins other than employment through which higher wage mandates have been absorbed.
WAGE/BENEFIT TRADE-OFF?
While it is conventional in the labor economics literature to assume a dollar-for-dollar trade-off of benefits for wages, there is actually a fair bit of evidence to the contrary. In fact, in a recent article by Schwabish (2004) that reviews the trade-off literature, he notes that more articles seem to contradict than prove the conventional wisdom: 'These papers are unique in the literature in that they are generally successful in finding tradeoffs between these mandated benefits and wages.' Schwabish himself finds scant evidence of a full trade-off.
Buchmueller and Lettau (1997) have a particularly strong article in that they have longitudinal, firm-level data, with high-quality measures of employers' costs for both wages and benefits They use the microdata from the Bureau of Labor Statistics Employer Costs Survey. The challenge in this literature is to control for individual productivity factors that influence both the wage and the receipt of health coverage. By differencing out fixed effects, Buchmueller and Lettau arguably control for such effects. They write: 'We ... find no evidence that higher health care costs growth translates into lower wage growth. This 'non-finding' is robust to a variety of specifications and sample definitions.'
Simon (2001), a contributor to this volume, uses a different data set and comes up with a similar result: 'I find that even after controlling for an extensive set of productivity factors I obtain results indicating a wrong-signed tradeoff... a strong correlation between changes in wages and changes in benefits.'
More recently, Tyson and Reber (2004) engage in a related analysis using a very different approach. They hypothesize that health care costs were one factor behind the jobless recovery of the early 2000s. Of course, if employers could trade off wages for benefits, this hypothesis would not make sense. But the authors find strong, industry-level correlations between health care costs and weak job growth. A further comparison with Canadian industries over the same time period supports the view that higher health costs were depressing job growth, an insight incompatible with a simple wage/benefit trade-off.
Finally, as we show in Figure 1, also using recent data, workers whose real wages grew the least between 2000 and 2005 had the largest loss in health coverage. This too challenges a simple trade-off model and suggests that perhaps worker bargaining power and compensation inequality is playing a role in these dynamics. One interpretation of this literature is that as labor demand and other wage determinants (less union power, more global competition, lower minimum wages) have evolved over the past few decades to favor high relative to low-wage workers, the former have been able to claim a larger share of the economy's growth in both their wages and their benefits. In work such as Schwabish (2004), who runs quantile regressions on compensation, this shows up quite clearly in the absence of a trade-off for workers earnings both higher wages and benefits.
So, we are skeptical of the wage/benefit trade-off that is a critical assumption of the BL model, and would urge them to do a lot more to convince the reader that this is a valid assumption. Perhaps the literature cited above is mostly picking up shorter-term disequilibria, and the trade-off is a longer-term phenomenon. But the BL model assumes instant trade-off, so simply citing lags is not sufficient. They may also want to argue that a tradeoff is more likely for low-wage workers, an argument that would fit well within our bargaining power framework (though Schwabish's, 2004, findings do not support this conclusion). But to do so would call for a more nuanced model, one that controlled other factors that play an important role in bargaining power, such as the unemployment rate (note that both wages and benefits grew for low-wage workers in the full-employment period of the latter 1990s, even as the minimum wage was also increased).
MINIMUM WAGE ELASTICITY?
As noted, BL need to borrow an employment elasticity from the minimum wage literature to implement their model. <Their> assumption that employment falls by 1 percent for a 10 percent increase in the minimum wage is by no means indefensible. But neither is an assumption of a zero disemployment effect.
Of course, the negative assumption reflects conventional wisdom, but the spate of interesting research in the 1990s, often associated with David Card and Alan Krueger, has had a significant impact on economists' thinking on the matter of the disemployment effects of minimum wage increases. A good example of how thinking has evolved comes from comparing the statement in two editions of Alan Blinder's undergraduate text book.
In the 1979 edition of Blunder's textbook, he told students: 'The minimum wage effectively bans the employment of workers whose marginal product is less than [the minimum wage]. The primary consequence of the minimum wage law is not an increase in the incomes of the least skilled workers but a restriction of their employment opportunities' (p. 519).
His 10th edition (2006), however, framed the issue this way: 'Elementary economic reasoning . . . suggests that setting a rninimum wage . . . above the free-market wage . . . must cause unemployment . . . Indeed, earlier editions of this book, for example, confidently told students that a higher minimum wage must lead to higher unemployment. But some surprising economic research published in the 1990s cast serious doubt on this conventional wisdom' (p. 493).
Recently, an interesting article by Dube et al. (2007) take the closest look at the employment effects of the hundreds of different minimum wages in play around the country. Using high-quality administrative data, Dube et al. conduct an omnibus difference-in-difference test of the impact of minimum wage differences on county-level employment changes. Armed with such geographically specific data, they are able to show that by aggregating across labor markets, many national studies are unable to control for heterogeneous differences in employment trends. Once they do so, most negative elasticities associated with minimum wages fall to zero.
BL should at least take greater note of these doubts and note the impact that the new research would have on their findings. Their assertion that their results are a 'lower bound' on the disemployment effects is called into question.
OTHER WAYS THE MANDATE MIGHT BE ABSORBED
These findings, of course, raise a good question-one that is relevant to this article tooas to how minimum wage increases, or potential health mandates, are absorbed. While many analysts, including BL, are quick to assume that lower employment or fewer hours is the sole absorption mechanism, at least three other mechanisms must be considered: prices, profits, and productivity.
Suppose employers push the mandate forward to consumers in higher prices. Since the mandate is universal, no firm is at a competitive disadvantage. In fact, in some local debates, firms that provide their employers with health care have been supportive of mandates like the one BL examine for this very reason: to level the playing field and preclude any competitive advantage by firms that choose not to provide coverage.
There is also an evolving view, especially, in regard to low-wage firms, that society ends up paying the costs of health care for uncovered low-wage workers from low-income families (and such care is often inefficiently provided, say through emergency rooms). In this regard, the universal mandate internalizes an external cost. Finally, as we argue later, good health care reform includes cost containment, so higher employer costs may be offset in part by more efficient care.
Actually, the second mechanism-reduced profits-is already embedded in BL's model. It is just that in their model, firms do not sacrifice any profits until they are forced to do so by the minimum wage mandate, which prevents a full wage/benefit trade-off. But any mandate can be paid for through diminished profits, especially if workers have the clout to enforce that trade-off, or if profit margins are especially high and employers may be more willing to sacrifice at this margin.
Finally, there is some evidence that the cost of higher rninimum wages has been paid for through productivity gains, primarily through lower turnover and fewer vacancies, problems that can be quite common and disruptive at low-wage firms. More to the point regarding BL's research, evidence suggests that a higher quality job, such as one that provides decent health insurance benefits, will better attract and retain more productive employees, will reduce absenteeism and turnover, and raise productivity through improved morale and worker loyalty (Obrien, 2003).
An interesting piece of anecdotal evidence on this point recently surfaced in regard to Wal-Mart, one of the nation's largest employers of low-wage workers. The behemoth retailed decided to considerably upgrade the quality of their health care plans, including shorter wait times for coverage and the offer of more generous plans. The company spokesperson argued that 'if workers 'are healthy, they will do a better job at work, they'll be more productive, they'll be happier, nicer to our customers' . . . all of which results in less absenteeism and turnover, a longstanding problem in retailing.'1
In addition, a survey of small employers found that the majority of employers reported that offering health benefits affected recruitment, helped retain employees, improved employees' attitudes and performance, improved the health of employees and reduced absenteeism (see Employee Benefits Research Institute, 2000).
Their observations are important for BL to consider for two reasons. First, as noted, they provide alternative absorption mechanisms for the costs of the mandate, but equally importantly, they introduce dynamic cost savings into their static model. It is now widely recognized that there are considerable inefficiencies in our current system of health care provision, and that these inefficiencies drive part of the increase in health care costs (Orszag and Ellis, 2007). For example, the costs of the uninsured are typically picked up by businesses and their employees in the form of higher premiums for their insurance. Given that outcome, a mandate that led to higher coverage among currently uninsured workers would likely have some cost offsets by reducing these externality costs generated by the current system (Fronstin, 2000). While such dynamics are beyond the scope of BL's simple model, they would be worth mentioning as a potential offset to the employers costs of the mandate.
PUBLIC POLICY ALTERNATIVES
While these next comments go beyond the scope of BL's analysis, we wanted to use this opportunity to comment briefly on the policy that they simulate: a fairly generous employer mandate, where employers pay 100 percent of premiums and full coverage of family premiums is required.
There are other, and to our thinking, better ways to mandate health coverage for workers. For example, a 'pay or play' mandate, where employers either provide coverage (at a specified level) or pay into a public insurance fund that has some desirable qualities (see Hacker, 2007). The public option, for example, creates a large risk pool that can be advantageous to small firms in terms of cost savings, and such firms have disproportionate shares of uninsured workers. Also, pay or play options can be financed through a fixed share tax on payrolls, which means smaller dollar payments for low wage firms. This contrasts with the BL plan, which charges employers a fixed cost for premium coverage regardless of size or payroll.
In other words, more dynamic models than BL's find that health care reform can have positive benefits: the creation of large risk pools, cost containment, low administrative costs, purchasing power, board oversight, lower costs from insurance due to preventive care, regular doctor visits, chronic disease management, and the reduction of avoidable hospitalizations. All of these benefits have the potential to feed back into the system, leading to lower costs than would be the case under the current system.
Of course, employers who currently do not insure their workers will face higher costs under any insurance plan that mandates their participation. But as long as America is going to stick to providing health insurance to working age families through their jobs, and if we are truly moving toward a political consensus that 47 million (and rising) uninsured persons is a problem that must be dealt with, then some type of employer mandate is almost surely forthcoming.
BL provide a sensible and clear, though static, framework for looking at the impact of mandates. We encourage them to continue to pursue this work and to expand their model by (1) adding caveats and sensitivity analysis regarding both the wage/benefit trade-off, and the minimum wage elasticity, (2) considering other mechanisms through which mandate price increases get absorbed, and (3) considering some 'second round' benefits of health reform that have the potential to significantly contain costs.
1 http://www.nytimes.com/2007/11/13/business/13walmart.html?hp
REFERENCES
Buchmueller, T. C, and M. K. Lettau, 1997, Estimating the Wage-Health Insurance Tradeoff: More Data Problems?, Unpublished paper, University of California at Irvine.
Dube, A., W Lester, and M. Reich, 2007, Minimum Wage Effects Across State Borders: Estimates Using Contiguous Counties, Institute for Research on Labor and Employment, University of California, Berkeley. Available at http://repositories.cdlib.org/iir/iirwps-157-07/
Employee Benefits Research Institute, 2000, Small Employer Health Benefits Survey (Washington, DC: EBRI).
Fronstin, P., ed., 2000, The Economic Costs of The Uninsured: Implications for Business and Government (Washington, DC: Employee Benefit Research Institute).
Hacker, J., 2007, Health Care for America, EPI Briefing Paper #180 (Washington, DC: Economic Policy Institute).
O'Brien, E., 2003, Employers' Benefits From Workers' Health Insurance, The Milibank Quarterly, 81(1): 5-43.
Orszag, P. R., and P. Ellis, 2007, The Challenge of Rising Health Care Costs-A View From the Congressional Budget Office, New England Journal of Medicine, 357:18.
Reber, S., and L. Tyson, 2004, Rising Health Insurance Costs Slow Job Growth and Reduce Wages and Job Quality (Los Angeles: University of California).
Schwabish, J. A., 2004, Accounting for Wages and Benefits Using the ECI, Bureau of Labor Statistics: Monthly Labor Review, September: 26-41.
Simon, K. I., 2001, Displaced Workers and Employer-Provided Health Insurance: Evidence of a Wage/Fringe Benefit Tradeoff?, International Journal of Health Care Finance and Economics, 1(3/4): 249-271.
Jared Bernstein is an economist with Economic Policy Institute; e-mail: jbernstein@epinet.org. Elise Gould is an economist with Economic Policy Institute; e-mail: egould@epi.org.
