minimum wages, employerprovided health insurance, and the

Minimum Wages, Employer-Provided HealthInsurance, and the Non-discrimination Law

MINDY S. MARKS*

This article exploits cross-state variation in minimum wages to investigate theimpact of minimum wage changes on employer-provided health insurance. In

contrast to the existing empirical literature, this article considers an environmentwhere some firms are constrained by non-discrimination laws that govern the pro-

vision of health insurance. For these firms, minimum wage changes do not reducethe probability that workers will receive employer-provided health insurance. Forfirms not covered by the non-discrimination law, and free to tailor their fringe

benefits, low-skilled workers experience a disproportionate reduction in the avail-ability and generosity of health insurance after a minimum wage increase.

Introduction

MOST OF THE EXISTING EMPIRICAL RESEARCH ON THE ECONOMIC IMPACTS of mini-mum wage laws has focused on the effects of such laws on the availabilityof jobs for low-wage workers. Despite decades of study, this vast literature1

has yet to reach a consensus concerning the employment consequences ofminimum wage changes. One drawback of this focus is that it provides toonarrow a basis for policy evaluation. Employment contracts are multidimen-sional agreements which, in addition to providing wages, furnish workerswith numerous non-wage (or fringe) benefits, including flexible workinghours, training, health insurance, pensions, and paid leave. As long as thewage and non-wage portion of total compensation are substitutes, firms canrespond to mandates that increase the wage portion of compensation byreducing the non-wage component. In fact, the state of Nevada recognizesthis trade-off and allows a lower minimum wage for those firms that offerhealth insurance.

* The author’s affiliation is Department of Economics, University of California-Riverside, 4110 SproulHall, Riverside, CA. E-mail: [email protected].

1 See Card and Kruger (1995) for an overview.

INDUSTRIAL RELATIONS, Vol. 50, No. 2 (April 2011). ! 2011 Regents of the University of CaliforniaPublished by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, and 9600 Garsington

Road, Oxford, OX4 2DQ, UK.

241

There is a small but growing empirical literature on the relationship betweenthe minimum wage and the provision of fringe benefits.2 A common approachfor determining the relationship between minimum wage increases and the pro-vision of health insurance is a difference-in-differences methodology wheremid-to-high-wage workers serve as a control group to guard against unmea-sured factors that may be correlated with minimum wage increases and benefitprovision. This work argues that this methodology is flawed due to enactmentof non-discrimination rules in 1978 that govern the provision of health insur-ance. Non-discrimination rules have been shown to limit the within-firminequality in benefit provision (Carrington, McCue, and Pierce 2002).In particular, non-discrimination rules in section 105 of the Internal Revenue

Tax Code state that in order to receive tax-exempt status for the health insur-ance plans of highly compensated employees, self-insured firms must offerhealth insurance plans that do not discriminate in favor of highly compensatedemployees (Collins 1999).3,4 As Gruber (1998:7) notes, for firms coveredunder the non-discrimination rules: ‘‘… it is impossible to selectively offerinsurance to only some employees, without making it a workplace wideoption.’’ This is in contrast with other fringe benefits, such as sick leave oron-the-job training, where the firm can tailor offerings to individual workers.For firms covered by the non-discrimination rule, all employees are potentiallyaffected if the firm alters its availability or generosity of health insurance inresponse to a change in the minimum wage. In this case, it would be incorrectto use employees earning above the minimum wage as a control group as iscommon in the literature on the minimum wage and non-wage compensation.Health insurance is the largest and most expensive of the non-legally

required benefits offered by firms. According to the Employer Costs forEmployee Compensation series, the average hourly cost of providing health

2 The benefit that has received the most attention is on-the-job training. Hashimoto (1982) uses time-series variation in the minimum wage and finds support for the prediction that increases in the minimumwage reduce on-the-job training for white men. Two recent papers have used cross-state variation in theminimum wage to investigate this issue. Neumark and Wascher (2001) conclude that minimum wages signi-ficantly reduce the amount of training received by young workers. Acemoglu and Pischke (2003), incontrast, do not find that minimum wages reduce the extent of training.

3 More specifically, non-highly compensated employees must constitute at least 50 percent of the groupof eligible employees; at least 90 percent of the non-highly compensated employees must be eligible for aplan that is at least 50 percent as valuable as the plan made available to the highly compensated employeewith the most valuable benefit; and the plan must not contain any eligibility provision that suggests discrimi-nation in favor of highly compensated employees. A plan also qualifies if at least 80 percent of non-highlycompensated employees benefit from the plan (Gruber 1998).

4 The distinction between self-insured and third-party-insured health plans is based on risk. Employerswho are responsible for the underlying insurance risk are considered self-insured even if a third-party com-pany administers the health insurance plan. Employers have an incentive to self-insure to be exempt fromtheir state’s health insurance mandates.

242 / MINDY S. MARKS

insurance for workers in the private sector in 2007 was $1.85 or 7.1 percent oftotal compensation (Bureau of Labor Statistics 2007). Furthermore, about 42percent of all private sector workers without a high school education receivehealth insurance through their employer. Thus, there is scope for reducing totalcompensation by cutting the availability or generosity of health insurance inresponse to minimum wage increases.The minimum wage, by constraining the compensation package employers

can offer, may lower the receipt of employer-provided health insurance. How-ever, whether or not workers lose coverage depends critically on if the firm issubject to the non-discrimination law. Using data from the March Current Pop-ulation Survey (CPS) for 1988–1993 and 1998–2005, this article exploitscross-state variation in the minimum wage and finds employer-provided healthinsurance coverage is unchanged in firms covered by the non-discriminationlaw. Firms not covered by the non-discrimination law, however, can tailor theirhealth insurance offerings to individual employees. For these firms, reductionsin the availability and generosity of health insurance are therefore concentratedamong low-skilled workers. Furthermore, the reduction in employer-providedhealth insurance is larger for those workers in predominately low-skilled indus-tries.

Previous Literature

The idea that firms reduce the non-wage portion of compensation in responseto increases in the minimum wages has been investigated previously by Royalty(2000) and Simon and Kaestner (2004). Both papers examine an array of bene-fits. The fringe benefits Royalty investigates are health insurance, pensions, andsick leave. Using two years of data from the 1988 and 1993 CPS’s Survey ofEmployee Benefits and cross-state variation in the minimum wage, she finds noeffect on fringe benefit eligibility for small increases in the minimum wage.However, she demonstrates that minimum wage increases of at least $0.50,decrease the probability that low-skilled workers are eligible for pensions andhealth insurance.5 Simon and Kaestner include health insurance and pensions intheir analysis. They use data from the Annual Demographic Surveys of the CPS.Both cross-state and federal changes in the minimum wage are used to identify

5 Royalty’s analysis focuses on how the minimum wage affects eligibility for fringe benefits. This article,in contrast, focuses on how the minimum wage affects the number of workers who receive (i.e., are eligibleand accept) health insurance from their employers. As such, it implicitly includes potential channels of mini-mum wage influence on employer-provided health insurance that the Royalty paper does not address, suchas requiring the employee to bear a larger portion of health insurance costs or switching to less generousplans.

Minimum Wages, Health Insurance, and the Non-discrimination Law / 243

minimum wage effects. In the end they conclude that minimum wages have hadno discernible effect on fringe benefit provision. Under their empirical frame-work, the significant negative relationship they find between health insuranceand the minimum wage in some specifications is due to unmeasured state-yearfactors that are correlated with the minimum wage.This analysis differs from these papers in a key way. Both Royalty and Simon

and Kaestner employ a difference-in-differences methodology. Under thisapproach the difference in fringe benefit receipt for a group of low-wage workerslikely to be affected by an increase in the minimum wage (the treatment group)is compared to the change in fringe benefit provisions for mid-wage workersunlikely to be impacted by the minimum wage (the control group). These authorsargue that the minimum wage has an impact on the provision of non-wage bene-fits if and only if the coefficient on the treatment group is more negative (lesspositive) than the control group. These papers state that the methodologydepends on two key assumptions: ‘‘Obviously, this specification relies heavily onthe adequacy of the ‘control group’ and the assumption that time effects are thesame for high- and low-wage workers’’ (Simon and Kaestner 2004:56) This dif-ference-in-differences framework is employed regardless of whether the outcomevariable is a benefit that the firm may individually tailor, like sick leave, or a ben-efit that is governed by the non-discrimination clause. As mentioned earlier,however, the non-discrimination rule makes near minimum wage workers aninvalid control group when analyzing a benefit that cannot be individually tai-lored, such as employer-provided health insurance. Using data from before andafter the enactment of the non-discrimination rules, Vanness and Wolfe (2002)argue that minimum wage constraints are a more important factor in determiningemployer-sponsored health insurance outcomes after the non-discrimination rulethan before. Accordingly, unlike prior scholarship, this article explicitly takesinto account the role that the non-discrimination law plays in the relationshipbetween the minimum wage and the receipt of employer-provided healthinsurance.

Empirical Specifications

This article examines the effect of the minimum wage on the availability ofemployer-provided health insurance using an approach that has become standardin the ‘‘new’’ minimum wage research.6 This approach exploits cross-state

6 Empirical studies prior to the ‘‘new’’ minimum wage research used time-series data and comparedchanges in the real value of the federal minimum wage to changes in some output measure at the nationallevel.


variation in minimum wages to avoid attributing to the minimum wage influ-ences from unmeasured variables common to all observations in a particularstate or year. States may set minimum wages higher than the federal minimumwage. If firms respond to higher minimum wages by reducing non-wage bene-fits, in any given year a state with an increased minimum wage should have alarger percentage of employees without employer-provided health insurancethan a state whose minimum wage is equal to the federal minimum wage.Furthermore, if two states increase their minimum wage, then the state with thelarger increase should see a larger impact. This hypothesis is tested using thefollowing Probit model:

PðHIijt ¼ 1Þ ¼ Ffaþ b1MWjt þ b2Xijt þ b3Sj þ b4Tt þ eijtg: ð1Þ

In equation (1), HIijt is a dichotomous variable that equals one if individual‘‘i’’ in state ‘‘j’’ in year ‘‘t’’ receives employer-provided health insurance andzero otherwise. MWjt is the real value of the minimum wage in state ‘‘j.’’ Xijt

is a vector of demographic and industry controls which the literature onemployer-provided health insurance (Dranove, Spier, and Baker 2000; Gruberand Lettau 2000; Gruber and Poterba 1994; Vanness and Wolfe 2002;Woodbury 1983) have established to be important. Included as demographiccontrols are: age, age squared, race (black, white), part-time status, and educa-tion (no high school (omitted), high school graduate, some college, collegegraduate, and advanced degree). Also included is a complete set of interactionterms between marital status (married, unmarried), gender (male, female), andpresence of own children (children present, children not present). Thesevariables capture the possible access to insurance through a spouse, andcontrol for the fact that the value of health insurance is increasing with thepresence of children (Buchmueller 1996). The number of workers (0–24,25–99, 100–499, 500–999, over 1000) in the firm and a set of 46 industrycontrols capture benefit provision differences on the firm side. State unemploy-ment rates are also included in Xijt. Sj and Tt are state and year fixed effects,respectively. Equation (1) uses state-level variation in the minimum wage andreceipt of employer-provided health insurance from year to year to identify theeffect of the minimum wage. Federal variation in the minimum wage, which isof larger magnitude than state variation, is subsumed by the year fixed effects.7

The state fixed effects control for any state specific time invariant variables,

7 According to Burkhauser, Couch, and Wittenburg (2000), federal variation in the minimum wage isresponsible for 90 percent of the total variation in the minimum wage between the years 1979 and 1992.Omitting the time-specific effects enables the use of both state-year and federal variation in the minimumwages. In estimates (not shown) that include the federal variation in the minimum wage the results aresimilar but suggest that workers in large firms of all skill categories experience a small reduction in thelikelihood of receiving employer-provided health insurance.


such as insurance mandates, that impact employer-provided health insurance.Included in all specifications are education-specific time trends that allowunmeasured factors such as rising health care costs to differentially impactlow-educated workers.8

Equation (1) can also be estimated with the minimum wage variable inter-acted with a complete set of skill category (high school dropouts, high schoolgraduates, education beyond high school) dummy variables. This alternativeregression equation can be written as follows:

PðHIijt ¼ 1Þ ¼ Ffb1MW % NoHSijt þ b2MW % HSijt þ b3MW % BHSijtþ b4Xijt þ b5Sj þ b6Tt þ eijtg:

ð2Þ

In equation (2) NoHSijt, HSijt, BHSijt are mutually exclusive dummy variablesfor high school dropout, high school graduate, and post-secondary education,respectively. The Xijt vector is the same as in equation (1) and includes a fullset of education dummy variables and education-specific time trends. Equation(2) will investigate if minimum wage increases disproportionately impact thereceipt of employer-provided health insurance for low-skilled workers. Forfirms covered by the non-discrimination law any reduction of health insuranceshould be similar across skill categories since these firms must offer all work-ers the same health insurance. For firms not restricted by the law, low-skilledworkers should be disproportionately impacted.As previously mentioned, the non-discrimination law only affects firms that

self-insure (Collins 1999), thus firms that do not self-insure are legally allowedto provide employees at different compensation levels different healthinsurance plans or different premiums.9 The data do not identify if firms areself-insured; however, firm size has been shown to be the best predictor forself-insurance status. Larger firms are much more likely to self-insure than theirsmaller counterparts because their large size better shields them from risk (Acset al. 1996). Recent data from the Medical Expenditure Panel Survey—Insurance Component finds in 2000, 10.7 percent of firms with fewer than 100workers offered a self-insured plan, 29.6 percent of firms with between 100and 500 workers offered a self-insured plan, while 76.3 percent of firms withmore than 500 workers offered such a plan (Agency for Healthcare Researchand Quality 2000). In 2006 only 25 percent of firms with 100–500 workers

8 The exclusion of the education-specific time trends does little to alter the results.9 There is an active consulting market where law firms advise employers how they can legally provide

different health insurance packages. See http://www.benicompselect.com/company/articles/PDF/Section_105.pdf and http://www.millermartin.com/cgi-bin/aw/acuweb.cgi?s=mmweb&a=search&UID=26&b=articles&t=librarydetail.htm&show=1.


offered self-insured plan whereas 80.5 percent of firms with more than 500workers offered self-insured plans (Agency for Healthcare Research andQuality 2006). Thus, firm size will be used as a proxy for coverage under thenon-discrimination law and equation (2) will be estimated separately for largefirms (at least 500 workers) and small firms (fewer than 500 workers).For those small firms less likely to be covered by the non-discrimination

law we can test the following claim. Assume that the minimum wage onlyaffects the receipt of health insurance for low-skilled workers in firms that arenot subject to the non-discrimination law, and that it has no effect on high-skilled workers in these firms. The effect of the minimum wage will thereforebe captured by the difference between coefficients b1 and b3 in equation (2).Identifying the effect of the minimum wage in this manner guards againstpicking up trends in employer-provided health insurance that affect all workers(in small firms) and are therefore unlikely to be caused by the minimum wagebut are correlated with minimum wage changes. For large firms, governed bythe non-discrimination law, minimum wage changes can directly impact thebenefit provision for high-skilled workers and this interpretation is invalid. Forthese firms any reduction in health insurance should be similar across skill lev-els and therefore the difference between the coefficients b1 and b3 should bestatistically insignificant.

The Data

The data for this article are drawn from the Annual Demographic Survey ofthe CPS, a national survey conducted by the Census Bureau that gathers infor-mation on health insurance, labor force participation, and other socioeconomicvariables (King et al. 2009). Appended to the CPS records is information onthe state minimum wage as well as information on the state unemploymentrate drawn from the Bureau of Labor Statistics Current Local Area Unemploy-ment Statistics. Since the CPS health insurance questions concern the previousyear’s employment, each year’s minimum wage as of January 1st is mergedwith the CPS survey in year t + 1. For instance, the 2000 state minimumwages are applied to the 2001 CPS responses (which ask respondents aboutinsurance in 2000).10 To be included in the sample, the respondent must be ofworking age (18–64), currently working with reported earnings above zero and

10 There is some debate over the reference year for the health insurance questions in the CPS (Currieand Yelowitz 1999). This article treats them as if they refer to the prior year’s health insurance status. Thesemodels are equivalent to assuming that respondents provide information about current year health insurancebut that firms are only able to respond to increases in the minimum wage with a lag. Results are robust (butlower in magnitude) to treating respondents as if their answers refer to the current calendar year.


not be self-employed, a student, or in the public sector.11 Respondents fromHawaii and Nevada are excluded because Hawaii mandates employer-providedhealth insurance and Nevada allows a lower minimum wage if the firm offershealth insurance.Analysis is conducted to focus on two separate time periods 1988–1993 and

1998–2005 where there were numerous state-level minimum wage changesand complete information about health insurance coverage.12 Lee (1999),Orrenius and Zavodny (2008), Royalty (2000), and Simon and Kaestner(2004) each provide evidence that the minimum wage was binding for a subsetof workers during these time periods. In addition, there was considerablevariation in the state minimum wages during these years. Table 1 shows thenominal minimum wage in each state that had a state minimum wage abovethe federal minimum wage at some point in the sample period. In all instancesthe higher minimum wage (state or federal) is used. The number of states witha minimum wage above the federal minimum wage ranges from 3 to 14 forthe analysis years. The set of states that alter their minimum wage is notidentical across sample periods. For instance, Delaware, Illinois, and NewYork had minimum wages equal to the federal minimum wage in the earlyperiod, but had state minimum wages above the federal minimum wage in thelater period. On the other hand, Iowa, Minnesota, New Hampshire, NewJersey, Pennsylvania, and Wisconsin had minimum wage changes in the earlyperiod but adhered to the federal minimum wage in the later period. Thus,these states will appear as control states in the later time period but astreatment states in the early time period.The key health insurance questions in the CPS are as follows: ‘‘Was [the

employee] covered by a private health insurance plan in [the employee’s] ownname?’’ If and only if the respondent answered affirmatively to this question,

11 The non-discrimination law only applies to full-time workers. One potential adjustment mechanism toan increase in the minimum wage is to shift low-skilled workers to part-time status where benefits can bemore easily altered. Carrington, McCue, and Pierce (2002) and Wolaver, McBride, and Wolfe (1997) findevidence consistent with firms using part-time work to skirt non-discrimination provisions. While the desireto reduce expenditures on benefits may cause an increase in the use of part-time labor following an increasein the minimum wage, the direction of the change is theoretically ambiguous. If part-time and full-timeworkers are substitutes in production (but wages for part-time workers are lower than wages for full-timeworkers) then a minimum wage increase will raise the relative price of part-time work, and part-time workshould decline (McKee and West 1984). Results, not shown, suggest that state-level changes in the minimumwage are associated with around a 2 percentage point decrease in the likelihood of part-time work for low-skilled workers in both large and small firms. Since part-time work is potentially endogenous the author hasincluded all workers in the sample and a control for part-time status.

12 In 1994 only Washington increased its minimum wage. There were no state minimum wage changesin 1995. In 1996 only Massachusetts and Vermont increased their minimum wage by 50 and 25 cents,respectively. In addition to this, in 1994, the CPS did not collect information about what share of the healthinsurance was paid for by the employer.


TABLE

1

STA

TEM

INIM

UMW

AGES

1988

1989

1990

1991

1992

1993

1998

1999

2000

2001

2002

2003

2004

2005

Average

minim

umwage

⁄median

wage

Early

Late

Federal

minim

umwage

3.35

3.35

3.35

3.80

4.25

4.25

5.15

5.15

5.15

5.15

5.15

5.15

5.15

5.15

Alaska

3.85

3.85

3.85

4.30

4.75

4.75

5.65

5.65

5.65

5.65

5.65

7.15

7.15

7.15

0.473

0.423

California

4.25

4.25

4.25

5.75

5.75

6.25

6.75

6.75

6.75

6.75

0.612

0.542

Connecticut

3.75

4.25

4.25

4.25

5.65

6.15

6.40

6.70

6.90

7.10

7.10

0.477

0.395

Delaw

are

5.65

6.15

6.15

6.15

6.15

6.15

0.474

0.414

Illin

ois

5.50

6.50

0.439

0.423

Iowa

3.85

4.25

4.65

4.65

0.421

0.501

Maine

3.65

3.75

3.85

3.85

5.75

6.25

6.25

6.35

0.475

0.481

Massachusetts

3.65

3.75

3.75

5.25

5.25

6.00

6.75

6.75

6.75

6.75

6.75

0.488

0.408

Minnesota

3.55

3.85

3.95

4.25

0.396

0.458

New

Ham

pshire

3.55

3.65

3.75

3.85

0.386

0.406

New

Jersey

5.05

0.412

0.392

New

York

6.00

0.456

0.416

Oregon

4.25

4.75

4.75

4.75

6.00

6.50

6.50

6.50

6.50

6.90

7.05

7.25

0.601

0.519

Pennsylvania

3.70

0.421

0.428

Rhode

Island

3.65

4.00

4.25

4.25

4.45

4.45

5.65

6.15

6.15

6.15

6.75

6.75

0.504

0.483

Vermon

t3.55

3.65

3.75

3.85

5.75

6.25

6.25

6.25

6.75

7.00

0.494

0.453

Washing

ton

3.85

4.25

4.25

5.70

6.50

6.72

6.90

7.01

7.16

7.35

0.532

0.426

Wisconsin

3.65

0.397

0.418

All

0.474

0.468

Num

berabov

efederalminim

umwage

810

1411

45

36

99

1010

1112

NOTE:Allentriesreflecttheminim

umwageas

ofJanuary1stof

that

year.

Allentriesareequalto

thefederalminim

umwageunless

otherw

isenoted.

SOURCES:State

Labor

Law

s,Monthly

Labor

Review

(Nelsonvariousyears),March

CPS.


they were asked the follow-up question: ‘‘Was this health insurance planoffered through [the employee’s] current or former employer or union?’’ Thusa positive response to both questions implies the worker has health insurancefrom her employer in her own name (as opposed to being covered by a spou-sal plan). The nesting of the questions means it is impossible to distinguish ifa negative response to the first question implies the firm did not make healthinsurance available to that employee or the firm offers health insurance but theemployee declined the firm’s coverage.13

Table 2 contains summary statistics for the entire sample (column 2), work-ers in state-years that have a minimum wage above the federal minimum wage(column 3), and workers in state-years where the minimum wage equals thefederal minimum wage (column 4). State-years with minimum wages greater

TABLE 2

MEANS (AND STANDARD DEVIATIONS) BY MINIMUM WAGE

Entiresample

State-yearsabove federalminimum wage

State-yearsat federal

minimum wage

Real minimum wage 2.97 (0.279) 3.41 (0.267) 2.88 (0.165)Minimum wage ⁄ average hourly wage 0.472 (0.069) 0.540 (0.075) 0.456 (0.057)Percent with health insurance 61.6% (0.486) 60.9% (0.488) 61.7% (0.486)Percent with health insurance (high school dropouts) 41.5% (0.492) 38.5% (0.487) 42.3% (0.494)Percent with health insurance fully paid by employer 16.5% (0.371) 17.6% (0.381) 16.2% (0.369)Percent with pension plan 46.3% (0.499) 44.8% (0.497) 46.7% (0.499)State unemployment rate 5.48% (1.407) 5.55% (1.196) 5.47% (1.450)Age 38.56 (11.321) 38.58 (11.27) 38.56 (11.33)Wage 21,990 (21,373) 23,212 (23,155) 21,716 (20,942)Percent with public insurance 5.1% (0.220) 5.8% (0.233) 5.0% (0.217)Percent married 59.8% (0.490) 57.2% (0.495) 60.3% (0.489)Percent with children 48.4% (0.500) 47.1% (0.499) 48.7% (0.500)Percent African American 10.8% (0.310) 5.5% (0.228) 11.9% (0.324)Percent female 45.8% (0.498) 45.1% (0.498) 46.0% (0.498)Percent high school dropout 12.8% (0.334) 14.2% (0.349) 12.4% (0.330)Percent some college or higher 51.1% (0.500) 54.9% (0.498) 50.2% (0.500)25–99 employees 15.4% (0.361) 16.3% (0.369) 15.2% (0.359)100–499 employees 15.9% (0.365) 16.0% (0.367) 15.9% (0.365)500–999 employees 6.1% (0.239) 5.9% (0.235) 6.1% (0.240)Over 1000 employees 37.8% (0.485) 35.4% (0.478) 38.4% (0.486)Percent working part time 11.8% (0.323) 12.7% (0.333) 11.6% (0.320)Observations 690,377 138,737 551,640

NOTES: All monetary values are expressed in real (1982–1984 = 100) terms. The samples are weighted using populationweights. Public insurance means the workers received Medicaid, Medicare, or Champus.

13 There are datasets that allow one to distinguish between these two cases. For instance, the Employ-ment Benefit Survey conducted periodically by the CPS contains this information. Unfortunately, this datasetis unavailable for the years under study.


than the federal minimum wage account for 20.1 percent of the sample. Thecoverage of employer-provided health insurance is similar across the two sub-samples (60.9 percent vs. 61.7 percent). This would appear to be inconsistentwith the argument that a higher minimum wage decreases the likelihood ofreceiving employer-provided health insurance. However, high school dropoutsare significantly less likely to have employer-provided health insurance in highminimum wage state-years (38.5 percent vs. 42.3 percent). Thus, there is preli-minary support for the hypothesis that firms respond to higher minimum wagesby reducing the availability or generosity of health insurance. It is clear thatworkers in states with high minimum wages are not identical to their federalminimum wage counterparts. States with higher minimum wages also tend tohave higher average wages, fewer married individuals, fewer minorities, moreeducated workers, smaller firms, and more part-time workers. All of these fac-tors, except firm size and part-time status, are correlated with higher rates ofemployer-provided health insurance.

Empirical Results—Employer-Provided Health Insurance

Table 3 shows the results from estimating equation (1) for the two time peri-ods. Since the magnitude of raw Probit coefficients are not interpretable, all tablespresent marginal effects for the continuous variables and discrete changes in pre-dicted probability for the dummy variables holding all other variables at theirmeans. There is little evidence that state-level increases in the minimum wage areassociated with a reduced likelihood that the average worker will receiveemployer-provided health insurance. In both time periods (Panels A and B) andfor both small and large firms the estimated impact of minimum wage changesare small and statistically insignificant. However, this specification assumes ahomogenous response to an increased minimum wage across skill levels.As shown in column 1, the receipt of employer-provided health insurance is

much more common for workers who work in large firms. In both time periods aworker working for a firm with over 1000 employers has almost a 30 percentpoint increase in the predicted probability of receiving employer-provided healthinsurance relative to his counterpart who works for a firm with fewer than 25employees. The year fixed effects (not shown) indicate that the likelihood ofreceiving employer-provided health insurance declined over both sample periods.The relationship between education and the likelihood of receiving

employer-provided health insurance varies noticeably by firm size. As shownin Panel A: column 3, in small firms, which have more freedom to tailorhealth insurance plans to their employees, education is a strong predictor ofreceipt of health insurance. In the early time period, moving from high school


dropout (the omitted category) to having an advanced degree increases thelikelihood of receiving employer-provided health insurance by 24 percentagepoints. For large firms (column 2) moving from the lowest education categoryto the highest education category increases the likelihood of receivingemployer-provided health insurance by only 11 percentage points. The fact thatin both periods, additional educational attainment implies a larger increase inthe receipt of health insurance in small firms than in large firms is consistentwith non-discrimination laws placing additional restrictions on the ability oflarge firms to offer workers with different skill levels different health insuranceofferings.14

TABLE 3

ESTIMATES OF THE EFFECTS OF MINIMUM WAGES ON EMPLOYER-PROVIDED HEALTH INSURANCE

All firms Large firms Small firms

Panel A: 1988–1993Minimum wage )0.006 (0.015) )0.018 (0.013) 0.005 (0.016)High school 0.120 (0.018)** 0.070 (0.012)** 0.140 (0.022)**Some college 0.145 (0.018)** 0.085 (0.011)** 0.173 (0.025)**College 0.188 (0.014)** 0.118 (0.009)** 0.223 (0.023)**Beyond college 0.189 (0.015)** 0.107 (0.009)** 0.241 (0.020)**25–99 employees 0.156 (0.004)** 0.183 (0.004)**100–499 employees 0.219 (0.003)** 0.263 (0.004)**500–999 employees 0.238 (0.004)**>1000 employees 0.308 (0.004)** 0.028 (0.005)**Observations 261,611 112,246 149,365

Panel B: 1998–2005Minimum wage )0.001 (0.005) 0.006 (0.011) )0.006 (0.010)High school 0.147 (0.024)** 0.144 (0.026)** 0.134 (0.029)**Some college 0.207 (0.025)** 0.175 (0.024)** 0.218 (0.031)**College 0.248 (0.022)** 0.210 (0.020)** 0.261 (0.029)**Beyond college 0.257 (0.019)** 0.199 (0.016)** 0.288 (0.024)**25–99 employees 0.158 (0.004)** 0.176 (0.005)**100–499 employees 0.213 (0.005)** 0.243 (0.007)**500–999 employees 0.229 (0.005)**>1000 employees 0.284 (0.006)** 0.027 (0.004)**Observations 428,760 183,870 244,890

*Statistically significant at the 0.05 level; **at the 0.01 level.Estimates were obtained using Probit model and standard errors have been corrected for clustering at the state level.Regression models include the following additional variables: age; age squared; indicators for race, part-time status,seven controls for family structure, forty-five industry controls, state unemployment rates, state and year fixed effects,and year times education time trends. Results are weighted to reflect population averages. Large firm means more than500 employees.

14 The non-discrimination law does not imply complete equality in benefit provision. Even if firmscharge a common premium for health insurance, low-skilled workers may be less willing to pay the pre-mium. In addition, firms may impose waiting periods for coverage that would again have a greater impacton low-skilled workers due to higher turnover.


Table 4 reports the results from estimating models on receipt of employer-provided health insurance where the impact of the minimum wage on healthinsurance is allowed to vary by skill level. Panel A presents results fromestimating equation (2) for the years 1988–1993. For the full sample of firms(column 1), the minimum wage variable is negative and statistically significantfor workers in the lowest skill level only. The coefficient estimates suggest thata $1.00 increase in the minimum wage would reduce the probability of receiv-ing employer-provided health insurance by 5.1 percentage points for the aver-age high school dropout. Minimum wage changes appear to have no impacton the receipt of health insurance for workers in the other skill categories.As discussed earlier, we should expect the impact of minimum wage

changes to differ by coverage under the non-discrimination law. Specifically,for large firms (covered by the non-discrimination law), whatever changes thefirm makes to its health insurance plans must by law affect all employees. Incontrast, small firms (where the non-discrimination law does not apply) havethe freedom to focus the reduction in benefits on the group whose compensa-tion costs are most affected by a minimum wage increase. We can investigate

TABLE 4

ESTIMATES OF THE EFFECTS OF MINIMUM WAGES ON EMPLOYER-PROVIDED

HEALTH INSURANCE BY SKILL LEVEL


Panel A: 1988–1993Minimum wage* no high school )0.051 (0.017)** )0.033 (0.025) )0.044 (0.022)*Minimum wage* high school )0.001 (0.013) )0.004 (0.016) 0.001 (0.017)Minimum wage* beyond high school 0.005 (0.012) )0.025 (0.015) 0.031 (0.017)High school 0.001 (0.058) 0.031 (0.081) 0.015 (0.075)Some college 0.022 (0.057) 0.110 (0.065) )0.027 (0.075)College 0.076 (0.053) 0.139 (0.054)** 0.030 (0.075)Beyond college 0.083 (0.052) 0.126 (0.047)** 0.054 (0.075)Observations 261,611 112,246 149,365

Panel B: 1998–2005Minimum wage* no high school )0.030 (0.011)** )0.019 (0.017) )0.034 (0.014)*Minimum wage* high school )0.007 (0.009) 0.006 (0.012) )0.016 (0.012)Minimum wage* beyond high school 0.008 (0.008) 0.010 (0.010) 0.007 (0.011)High school 0.074 (0.034)* 0.051 (0.051) 0.090 (0.043)*Some college 0.086 (0.031)** 0.078 (0.046) 0.084 (0.041)*College 0.138 (0.029)** 0.127 (0.041)** 0.133 (0.040)**Beyond college 0.161 (0.027)** 0.133 (0.035)** 0.171 (0.038)**Observations 428,760 183,870 244,890

*Statistically significant at the 0.05 level; **at the 0.01 level.Estimates were obtained using Probit model and standard errors have been corrected for clustering at the state level. Allregressions include the control variables listed in Table 3. Results are weighted to reflect population averages.

Large firm means more than 500 employees.


these hypotheses by estimating equation (2) separately for large and smallfirms.15 Column 2 contains the results for the subsample of large firms. Coef-ficient estimates suggest a $1.00 increase in the minimum wage causes a 3.3percentage point reduction for high school dropouts and a 2.5 percentage pointreduction for workers with some post-secondary education. These estimatesare not statistically different from each other or from zero. These resultssuggest that a reduction in health insurance for all employees is too costly amargin of adjustment to minimum wage changes for large firms covered undernon-discrimination laws.In column 3 the sample is restricted to workers in firms with fewer than

500 employees. For these firms, minimum wage changes disproportionatelyimpact low-skilled workers. A $1.00 increase in the minimum wage reducesthe receipt of employer-provided health insurance by 4.4 percentage points forhigh school dropouts and is associated with a 3.1 percentage point increase incoverage for workers with some post-secondary education. The estimatedimpact for the lowest skilled workers is significantly different than that for thehighest skilled worker at the 5 percent level. For those firms who can discrimi-nate in their health insurance provisions, low-wage workers experience areduction in health insurance following minimum wage changes.Panel B repeats the analysis for a later time period. Again, the results for all

firms suggest that low-skilled workers are less likely to receive health insur-ance following an increase in the minimum wage. In large firms the likelihoodof receiving health insurance is unchanged by increases in the state minimumwage. The significant negative relationship between higher state minimumwages and lower receipt of employer-provided health insurance appearsonly for low-skilled workers in small firms who are not subject to the non-discrimination laws.Since the non-discrimination law requires health insurance to be altered at

the firm level, firms whose workforce is predominantly low-wage workersshould find it more attractive to adjust their health insurance plans than firmsthat employ a limited number of low-wage workers. Even for firms not cov-ered by the non-discrimination law, transaction costs may make altering healthinsurance a more attractive option for firms with a large share of low-skilledworkers. To investigate this hypothesis the sample was restricted to workers inindustries where at least 15 percent of the industry workforce consists of high

15 It would not be appropriate to re-estimate equation (2) and include as additional variables theminimum wage variables interacted with firm size dummy variables. This specification would imply that theimpact of additional education attainment is the same in large and small firms. The results in Table 3demonstrate that the impact of education attainment (and other control variables) on the receipt of healthinsurance clearly differs by firm size.


school dropouts.16 This produces a sample where 21 percent of the workforceis a high school dropout. Table 5 presents the relationship between the mini-mum wage and the receipt of employer-provided health insurance for workersin predominantly low-skilled industries. These results parallel the results inTable 4. In large firms, covered by the non-discrimination law, the relationshipbetween the minimum wage and health insurance is statistically insignificant inboth time periods for all workers. In firms exempt from the non-discriminationlaw, low-skilled workers are the only group whose receipt of employer-provided health insurance is negatively affected by an increase in the minimumwage. However, the reduction in the likelihood of receiving employer-providedhealth insurance is consistently larger when the sample is restricted to low-skilled industries. For instance, Panel B: column 3 shows that the averagelow-skilled worker in a low-skilled industry is 6.2 percentage points less likelyto receive employer-provided health insurance following a $1.00 increase inthe minimum wage. The corresponding estimate for the average low-skilled

TABLE 5

ESTIMATES OF THE EFFECTS OF MINIMUM WAGE CHANGES ON EMPLOYER-PROVIDED HEALTH

INSURANCE IN LOW-SKILLED INDUSTRIES


Panel A: 1988–1993Minimum wage* no high school )0.083 (0.024)** )0.052 (0.038) )0.076 (0.028)**Minimum wage* high school )0.003 (0.020) 0.009 (0.029) )0.012 (0.024)Minimum wage* beyond high school 0.050 (0.022)* 0.011 (0.031) 0.069 (0.028)*High school )0.062 (0.080) )0.009 (0.126) )0.037 (0.094)Some college )0.219 (0.084)** )0.036 (0.137) )0.245 (0.087)**College )0.152 (0.086) 0.027 (0.126) )0.190 (0.087)*Beyond college )0.171 (0.086)* 0.013 (0.130) )0.204 (0.082)*Observations 115,117 43,916 71,201

Panel B: 1998–2005Minimum wage* no high school )0.049 (0.021)* 0.014 (0.039) )0.062 (0.023)**Minimum Wage* high school )0.009 (0.019) 0.022 (0.034) )0.020 (0.021)Minimum wage* beyond high school 0.006 (0.020) 0.015 (0.033) 0.006 (0.022)High school 0.027 (0.063) 0.116 (0.112) 0.012 (0.069)Some college 0.054 (0.064) 0.176 (0.091) 0.000 (0.071)College 0.091 (0.063) 0.201 (0.070)** 0.027 (0.072)Beyond college 0.093 (0.064) 0.180 (0.060)** 0.038 (0.075)Observations 89,509 23,592 65,917

*Statistically significant at the 0.05 level; **at the 0.01 level.Estimates were obtained using Probit model and standard errors have been corrected for clustering at the state level.See notes in Table 3 for a list of control variables. Results are weighted to reflect population averages.

Large firm implies more than 500 employees.Low-skilled industries are those where at least 15 percent of the workforce is not a high school graduate.

16 Results are robust to an alternative cut-off of 20 percent.


worker in all industries (Table 3: Panel B: column 3) is a 3.4 percentage pointreduction.In results (not shown) that focus on the subsample of high-skilled industries

(where only 8 percent of the sample is a high school dropout), the minimumwage variable is insignificant at the 0.05 level in all specifications. Alteringthe health insurance of employees is too costly a margin of adjustment inindustries without a significant low-wage population. The fact that the mini-mum wage variables are insignificant in these specifications suggests that thereare no unmeasured factors that are correlated with minimum wage changesand receipt of employer-provided health insurance.The CPS also contains information on the share of the health insurance plan

paid by the employees (all, part, or none). One mechanism for reducing expen-ditures on employees following minimum wage increases is to require workersto directly bear part (or all) of the cost of health insurance. In Table 6 thedependent variable is an indicator variable that equals one if the workerreceives a health insurance plan with no employee contribution from her firm

TABLE 6

ESTIMATES OF THE EFFECTS OF MINIMUM WAGES ON CONTRIBUTION FREE

EMPLOYER-PROVIDED HEALTH INSURANCE

Allfirms

Largefirms

Smallfirms

Panel A: 1988–1993Minimum wage* no high school )0.049 (0.014)** 0.011 (0.026) )0.062 (0.016)**Minimum wage* high school 0.002 (0.009) )0.003 (0.016) 0.005 (0.012)Minimum wage* beyond high school 0.006 (0.009) 0.001 (0.014) 0.004 (0.011)High school )0.084 (0.039)* 0.059 (0.077) )0.115 (0.045)*Some college )0.075 (0.034)* 0.050 (0.077) )0.083 (0.039)*College )0.065 (0.034) 0.051 (0.079) )0.063 (0.039)Beyond college )0.054 (0.033) 0.059 (0.082) )0.047 (0.040)Observations 261,611 112,246 149,365

Panel B: 1998–2005Minimum wage* no high school )0.025 (0.008)** )0.020 (0.013) )0.020 (0.009)*Minimum wage* high school )0.010 (0.006) 0.002 (0.008) )0.016 (0.007)*Minimum wage* beyond high school 0.001 (0.005) 0.006 (0.007) )0.003 (0.006)High school 0.033 (0.033) )0.012 (0.053) 0.082 (0.043)Some college 0.021 (0.030) )0.014 (0.050) 0.079 (0.042)College 0.028 (0.032) )0.018 (0.048) 0.104 (0.047)*Beyond college 0.051 (0.035) )0.001 (0.051) 0.141 (0.054)**Observations 428,760 183,870 244,890

*Statistically significant at the 0.05 level; **at the 0.01 level.Estimates were obtained using Probit model and standard errors have been corrected for clustering at the state level. Allregressions include the control variables listed in Table 3. Results are weighted to reflect population averages. Large firmimplies more than 500 employees.


and zero otherwise.17 The results from this analysis are similar to thoseobtained when the dependent variable is simply the receipt of health insurance.In firms covered by the non-discrimination laws, changes in the state minimumwage have no impact on likelihood of receiving generous health insurancecoverage in either period. In small firms that have the flexibility to requireincreased contributions from low-skilled workers, the likelihood of having ahealth insurance plan without an employee contribution is reduced for onlylow-skilled workers. For instance, the final column in Panel A implies that a$1.00 increase in the minimum wage has no impact on the generosity of healthinsurance for high-skilled workers but reduces the likelihood of receiving nocost health insurance by 6.2 percentage points for high school dropouts insmall firms. The results in Panel B are weaker and suggest that in the latertime period firms were only slightly less likely to reduce the generosity of theinsurance offered to their low-skilled employees.

Robustness Checks

The analysis thus far has been predicated on the assumption that changes inthe minimum wage are uncorrelated with unmeasured factors that vary bystate-year and influence the receipt of fringe benefits.18 There could be unmea-sured state-year factors, like economic conditions, Medicaid ⁄AFDC eligibility,or state policies, which may be correlated with state minimum wage increasethat drive down the generosity of the benefit package being offered by smallfirms but not their large counterparts.To correct for this potential correlation, a second minimum wage measure

was created. This new measure is the ratio of the minimum wage in each statein a given year to the median wage for workers with a high school degree orless in that state in that year (relative minimum wage).19 This variable captures

17 Information about generosity of workplace health insurance is only provided for individuals thataccept health insurance from their employers. Thus, the appropriate sample is all workers (as opposed to therestricted sample of workers with employer-provided health insurance). It is possible that in response to anincreased minimum wage, employers reduce the generosity of health insurance and as result some workersnow reject their firm’s offer of insurance. This channel would be omitted if the sample were restricted toonly those with health insurance from their employers.

18 There is a small literature on the political economy of state minimum wage laws. This literatureconcludes that ideological factors are the dominant force behind state minimum wage laws (Waltman andPittman 2002).

19 Median wage was calculated from a weighted average of the median hourly wage for full-time andpart-time workers in each state assuming full-time workers work 2000 hours a year (40 hours worked perweek times 50 weeks worked per year) and part-time workers work 1000 hours a year (20 hours per weekfor 50 weeks).


state-year variation in the ‘‘bindingness’’ of the minimum wage and can beseen in Table 1. The closer the measure is to one the closer the minimumwage is to the state’s median wage and it is more likely that employers withlow-skilled workers will change their behavior in response to changes in theminimum wage. The set of states with the highest minimum wage differs fromthe set of states with the most binding minimum wage. The six states with themost binding minimum wages are California, Oregon, Mississippi, NewMexico, Texas, and Arkansas. Of these only California and Oregon have aminimum wage greater than the federal minimum wage at some point in thesample period. Replacing the original minimum wage variable with this rela-tive minimum wage variable is equivalent to estimating the main specificationwith a different set of control states (those states with relatively non-bindingminimum wages). It is unlikely that whatever omitted factors are correlatedwith living in a high minimum wage state are also correlated with living in astate whose minimum wage has become more binding.As a robustness check the main model (equation (2)) was re-estimated with

this new minimum wage variable for the sub-sample of low-wage industries.20

The coefficients on the key variables from this model can be found in Table 7.The same patterns of findings emerge under this specification of the state mini-mum wage. For small firms, exempt from the non-discrimination law, as theminimum wage becomes more binding, low-skill workers experience adecrease in their likelihood of receiving employer-provided health insurancerelative to their higher-skilled counterparts. This is true in both time periodsunder study. In contrast, when the relative minimum wage variable is used, thehealth insurance of workers in large firms governed by the non-discriminationlaw is unaffected when the minimum wage becomes more binding.21 That theresults are robust to an alternative specification of the minimum wage suggeststhat there are no unmeasured factors that are correlated with state-level mini-mum wage changes and lowered receipt of fringe benefits in small firms.A final robustness check investigates the receipt of employer-provided pen-

sion benefits. A weaker version of the non-discrimination law covers pensionplans but this law applies equally to small and large firms (Carrington, McCue,and Pierce 2002). Thus, unlike health insurance, small firms are constrained toalter the rules governing inclusion in their pension plans in a way that should

20 There is limited state-year variation in the bindingness of the minimum wage—thus results are pre-sented for the subsample of low-wage industries where the minimum wage is most likely to have an impact.Results for the full sample are similar in pattern but smaller in effect. In both periods there is a statisticallysignificant difference in the availability of health insurance between low- and high-skilled employees insmall firms.

21 Results (not shown) suggest that workers in small firms also receive less generous health insurance, asthe minimum wage becomes more binding.


not discriminate against low-skilled workers. The results from estimating equa-tion (2) with a dichotomous variable that reflects inclusion in the employer-provided pension plan as the outcome variable can be found in Table 8. Over-all these results support a binding non-discrimination law in the provision ofpension benefits. In small firms, both high-skilled and low-skilled workersexperience a 3 percentage point decrease in likelihood of being included intheir firms’ pension plan in response to a $1.00 increase in the minimum wagein the early period (Panel A: column 3). The corresponding reduction in likeli-hood of being included in the firm’s pension plan is larger for large firms(Panel A: column 2) but, again, there is not a statistically significant differenceacross the highest and lowest skill category. In Panel B there appears to be nosystematic relationship between higher state-level minimum wages and inclu-sion in the firm’s pension plan during the later sample periods. Unlike healthinsurance, pensions—which are governed by a non-discrimination law forsmall firms—show a similar change in benefit provision for both low- andhigh-skilled workers following an increase in the minimum wage.

Conclusion

This article argues that rising state minimum wages create incentives forfirms to substitute between wage and non-wage benefits. Indeed, the fact thatemployers may respond to increases in the minimum wage by decreasing theavailability of non-wage benefits may explain why empirical researchers find it

TABLE 7

ESTIMATES OF THE EFFECTS OF MINIMUM WAGES ON EMPLOYER-PROVIDED HEALTH INSURANCE IN

LOW-SKILLED INDUSTRIES CONTROLLING FOR MEDIAN HOURLY WAGE


Panel A: 1988–1993Relative minimum wage* no high school )0.327 (0.126)** )0.119 (0.169) )0.349 (0.123)**Relative minimum wage* high school )0.005 (0.109) 0.097 (0.120) )0.060 (0.128)Relative minimum wage* beyond high school 0.276 (0.126)* 0.162 (0.129) 0.321 (0.155)*Observations 115,117 43,916 71,201

Panel B: 1998–2005Relative minimum wage* no high school )0.419 (0.103)** 0.036 (0.184) )0.526 (0.103)**Relative minimum wage* high school )0.176 (0.103) 0.127 (0.159) )0.258 (0.103)*Relative minimum wage* beyond high school 0.089 (0.091) 0.340 (0.184) )0.009 (0.103)Observations 89,509 23,592 65,917

*Statistically significant at the 0.05 level; **at the 0.01 level.Estimates were obtained using Probit model and standard errors have been corrected for clustering at the state level. Allregressions include the control variables listed in Table 3. Results are weighted to reflect population averages.


difficult to detect an economically significant relationship between the mini-mum wage and employment.Using data from the 1988–1993 and 1998–2005 waves of the March CPS,

this article demonstrates that state minimum wages influence the number ofindividuals who receive employer-provided health insurance but the impactdepends on coverage under the non-discrimination law. For firms subject tothe non-discrimination law, it appears that the non-discrimination law is bind-ing. Large firms do not reduce the availability of health insurance following aminimum wage change. For uncovered small firms only low-skilled workersexperience a significant reduction in employer-provided health insurance. Thisfinding is in contrast to research that does not focus on the non-discriminationlaw and concludes that minimum wages have no effect on the provision ofhealth insurance. In addition, it appears that industry composition matters.Workers in predominantly low-skilled industries experience the largest declinesin the likelihood of receiving employer-provided health insurance following aminimum wage increase, while the minimum wage has no impact on healthinsurance for workers in high-skilled industries.

TABLE 8

ESTIMATES OF THE EFFECTS OF MINIMUM WAGES ON EMPLOYER-PROVIDED

PENSION PLAN BY SKILL LEVEL


Panel A: 1988–1993Minimum wage* no high school )0.064 (0.029)* )0.091 (0.026)** )0.030 (0.030)Minimum wage* high school 0.003 (0.014) 0.011 (0.014) )0.002 (0.019)Minimum wage* beyond high school )0.036 (0.012)** )0.037 (0.017)* )0.030 (0.014)*High school )0.066 (0.094) )0.168 (0.069)* 0.013 (0.101)Some college 0.047 (0.093) )0.034 (0.075) 0.097 (0.098)College 0.092 (0.098) 0.023 (0.073) 0.127 (0.106)Beyond college 0.105 (0.100) )0.006 (0.074) 0.180 (0.117)Observations 261,611 112,246 149,365

Panel B: 1998–2005Minimum wage* no high school )0.026 (0.030) )0.054 (0.036) )0.000 (0.023)Minimum wage* high school 0.007 (0.011) )0.018 (0.012) 0.024 (0.011)*Minimum wage* beyond high school 0.009 (0.010) )0.005 (0.010) 0.017 (0.009)High school 0.033 (0.094) 0.016 (0.129) 0.035 (0.068)Some college 0.065 (0.101) )0.005 (0.133) 0.104 (0.073)College 0.140 (0.099) 0.078 (0.127) 0.165 (0.075)*Beyond college 0.164 (0.097) 0.075 (0.123) 0.217 (0.077)**Observations 428,760 183,870 244,890

*Statistically significant at the 0.05 level; **at the 0.01 level.Estimates were obtained using Probit model and standard errors have been corrected for clustering at the state level. Allregressions include the control variables listed in Table 3. Results are weighted to reflect population averages. Large firmimplies more than 500 employees.


Finally, these findings contribute to the debate about why so many work-ing Americans lack health insurance (Farber and Levy 2000; Field andShapiro 1993). The United States is unique among developed nations in itsreliance on employers to provide health insurance for the majority of itscitizens. This insurance has been implicitly subsidized by the federal gov-ernment since a 1943 Internal Revenue Service ruling made compensationin the form of health insurance excludable from taxable income. Despitethe favorable tax treatment of health insurance, the majority of uninsuredadults are in families with at least one employed worker (EBRI 2000). Theresults of this article are consistent with the argument that the interactionof the minimum wage and the non-discrimination provisions of the taxcode have the unintended consequence of reducing the receipt of employer-provided health insurance. If individuals in our society—especiallylow-income individuals—are denied the advantages of subsidized healthinsurance, it may be time to reexamine our reliance on employer-providedhealth insurance.

REFERENCES

Acemoglu, Daron, and Jorn-Steffen Pischke. 2003. ‘‘Minimum Wages and On-the-Job Training.’’ Researchin Labor Economics 22: 159–202.

Acs, Gregory, Stephen H. Long, M. Susan Marquis, and Pamela Farley Short. 1996. ‘‘Self-Insured EmployerHealth Plans: Prevalence, Profile, Provisions, And Premiums.’’ Health Affairs 15(3): 266–78.

Agency for Healthcare Research and Quality, Center for Cost and Financing Studies. 2000. ‘‘Medical Expen-diture Panel Survey Summary Data Tables.’’ Table I.A.2.a. Available at: http://www.meps.ahrq.gov/mepsweb/data_stats/summ_tables/insr/state/series_2/2000/tiia2a.pdf, accessed April 5, 2010.

Agency for Healthcare Research and Quality, Center for Financing, Access and Cost Trends. 2006. MedicalExpenditure Panel Survey Summary Data Tables. Table I.A.2.a. Available at: http://www.meps.ahrq.gov/mepsweb/data_stats/summ_tables/insr/state/series_2/2006/tiia2a.pdf, accessed April 5, 2010.

Buchmueller, Thomas. 1996. ‘‘Marital Status, Spousal Coverage and the Gender Gap in Employer-Spon-sored Health Insurance.’’ Inquiry 33: 308–16.

Bureau of Labor Statistics. 2007. ‘‘Employer Costs for Employee Compensation Summary.’’ WashingtonDC: Economic News Release. Available at: http://www.bls.gov/news.release/archives/ecec_06212007.pdf, accessed April 5, 2010.

Burkhauser, Richard, Kenneth Couch, and David Wittenburg. 2000. ‘‘A Reassessment of the New Econom-ics of the Minimum Wage Literature with Monthly Data from the Current Population Survey.’’Journal of Labor Economics 18(4): 653–80.

Card, David, and Alan Kruger. 1995. Myth and Measurement: The New Economics of the Minimum Wage.Princeton, NJ: Princeton University Press.

Carrington, William, Kristin McCue, and Brooks Pierce. 2002. ‘‘Nondiscrimination Rules and the Distribu-tion of Fringe Benefits.’’ Journal of Labor Economics 20(2): Part 2, S5–33.

Collins, Michael. 1999. ‘‘A Primer on the Self-Insured Health Plan Nondiscrimination Rules.’’ Journal ofPension Planning and Compliance 25(2): 1–15.

Currie, Janet, and Aaron Yelowitz. 1999. ‘‘Health Insurance and Less Skilled Workers.’’ NBER WorkingPaper No. 7291. Cambridge, MA: National Bureau of Economic Research.

Dranove, David, Kathryn E. Spier, and Laurence Baker. 2000. ‘‘‘Competition’ Among Employers OfferingHealth Insurance.’’ Journal of Health Economics 19: 121–40.


Employee Benefit Research Institute. 2000. ‘‘The Working Uninsured: Who They Are, How they haveChanged, and the Consequences of Being Uninsured.’’ EBRI Issue Brief.

Farber, Henry, and Helen Levy. 2000. ‘‘Recent Trends in Employer-Sponsored Health Insurance Coverage:Are Bad Jobs Getting Worse?’’ Journal of Health Economics 19(1): 93–119.

Field, Marilyn J., and Harold Shapiro. 1993. Employment and Health Benefits A Connection at Risk.Washington DC: National Academy Press.

Gruber, Jonathan. 1998. ‘‘Health Insurance and the Labor Market.’’ NBER Working Paper No. 6762.Cambridge, MA: National Bureau of Economic Research.

—––—, and Michael Lettau. 2000. ‘‘How Elastic is the Firm’s Demand for Health Insurance?’’ NBERWorking Paper No. 8021. Cambridge, MA: National Bureau of Economic Research.

—––—, and James Poterba. 1994. ‘‘Tax Incentives and the Decision to Purchase Health Insurance: Evidencefrom the Self-Employed.’’ The Quarterly Journal of Economics 109(3): 701–33.

Hashimoto, Masanori. 1982. ‘‘Minimum Wage Effects on Training on the Job.’’ American Economic Review72(5): 1070–87.

King, Miriam, Steven Ruggles, Trent Alexander, Donna Leicach, and Matthew Sobek. 2009 IntegratedPublic Use Microdata Series, Current Population Survey: Version 2.0. [Machine-readable database].Minneapolis, MN: Minnesota Population Center.

Lee, David. 1999. ‘‘Wage Inequality in the United States During the 1980s: Rising Dispersion or FallingMinimum Wage?’’ Quarterly Journal of Economics 64: 977–1023.

McKee, Michael, and Edwin G. West. 1984. ‘‘Minimum Wage Effects on Part-Time Employment.’’Economic Inquiry 22: 421–8.

Nelson, Richard R. ‘‘State Labor Legislation Enacted in 1988.’’ Monthly Labor Review, various years.Neumark, David, and William Wascher. 2001. ‘‘Minimum Wages and Training Revisited.’’ Journal of Labor

Economics 19(3): 563–95.Orrenius, Pia, and Madeline Zavodny. 2008. ‘‘The Effect of Minimum Wages on Immigrants’ Employment

and Earnings.’’ Industrial and Labor Relations Review 61: 544–63.Royalty, Anne Beeson. 2000. ‘‘Do Minimum Wage Increases Lower the Probability that Low-Skilled

Workers Will Receive Fringe Benefits?’’ Joint Center for Poverty Research Working Paper 222.Simon, Kosali Ilayperuma, and Robert Kaestner. 2004. ‘‘Do Minimum Wages Affect Non-Wage Job

Attributes? Evidence on Fringe Benefits.’’ Industrial and Labor Relations Review 58(1): 52–71.Vanness, David, and Barbara Wolfe. 2002. ‘‘Government Mandates and Employer-Sponsored Health Insur-

ance: Who Is Still Not Covered?’’ International Journal of Health Care Finance and Economics2(2): 99–135.

Waltman, Jerold, and Sarah Pittman. 2002. ‘‘The Determinates of State Minimum Wage Rates: A PublicPolicy Approach.’’ Journal of Labor Research 23(1): 51–56.

Wolaver, Amy M., Timothy D. McBride, and Barbara L. Wolfe. 1997. ‘‘Decreasing Opportunities forLow-Wage Workers: The Role of the Nondiscrimination Law for Employer-Provided HealthInsurance.’’ Institute for Research on Poverty, Discussion Paper No. 1124–97.

Woodbury, Stephen A. 1983. ‘‘Substitution Between Wage and Nonwage Benefits.’’ American EconomicReview 73(1): 162–82.


minimum wages, employerprovided health insurance, and the

Documents