Chapter 50HEALTH, HEALTH INSURANCE AND THE LABOR MARKETJANET CURRIE*UCLA and NBERBRIGITTE C. MADRIANUniversity of Chicago and NBERContentsAbstractJEL codes1 Overview2 Health and the labor market220.127.116.11.18.104.22.168.8Health as human capitalMeasurement issues: what is health?Effects of health on wages, earnings, and hoursStudies that treat health as an endogenous choiceEvidence regarding health and attachment to the labor marketHealth and type of workChild health and future labor market outcomesHealth and the labor market: summary3 Health insurance and the labor market22.214.171.124.43.5Health insurance provision in the United States: backgroundEstimating the effect of health insurance on labor market outcomes: identi cation issuesEmployer provision of health insuranceThe relationship between health insurance and wagesThe relationship between health insurance and labor force participation: evidence onemployment and hours worked3.6 Health insurance and job turnover3.7 Health insurance and the structure of employment3.8 Health insurance and the labor market: summary4 0000000000000* We thank participants in the Handbook of Labor Economics Conference held in Princeton, New Jersey,September 4 7, 1997 for helpful comments, and we thank Emanuela Galasso for able research assistance.Funding from the National Institute on Aging and the University of Chicago (Madrian) is gratefully acknowledged.Handbook of Labor Economics, Volume 3, Edited by O. Ashenfelter and D. Cardq 1999 Elsevier Science B.V. All rights reserved.3309
3310J. Currie and B. C. MadrianAbstractThis chapter provides an overview of the literature linking health, health insurance and labor marketoutcomes such as wages, earnings, employment, hours, occupational choice, job turnover, retirement, and the structure of employment. The rst part of the paper focuses on the relationshipbetween health and labor market outcomes. The empirical literature surveyed suggests that poorhealth reduces the capacity to work and has substantive effects on wages, labor force participationand job choice. The exact magnitudes, however, are sensitive to both the choice of health measuresand to identi cation assumptions. The second part of the paper considers the link between healthinsurance and labor market outcomes. The empirical literature here suggests that access to healthinsurance has important effects on both labor force participation and job choice; the link betweenhealth insurance and wages is less clear. q 1999 Elsevier Science B.V. All rights reserved.JEL codes: I12; J32; J24¼ that the labor force status of an individual will be affected by his health is anunassailable proposition [because] a priori reasoning and casual observation tell us itmust be so, not because there is a mass of supporting evidence. (Bowen and Finegan,1969)Despite the near universal nding that health is a signi cant determinant of workeffort, the second major inference drawn from [this] review is that the magnitude ofmeasured health effects varies substantially across studies. (Chirikos, 1993)1. OverviewThis chapter provides an overview of some of the literature linking health and labor marketbehavior. The question is important because for groups as diverse as single mothers andolder people, health is thought to be a major determinant of wages, hours, and labor forceparticipation. Thus, an understanding of the effects of health on labor market activity isnecessary for evaluations of the cost effectiveness of interventions designed to prevent orcure disease. Moreover, since the relationship between health and the labor market ismediated by social programs, an understanding of this relationship is necessary if weare to assess the effectiveness and solvency of these programs. In countries with agingpopulations, these questions will only become more pressing over time as more individuals reach the age where health has the greatest impact on labor market outcomes.The two quotations above, one from 1969 and one from 1993, illustrate that a good dealof empirical evidence linking health and labor market activity has sprung up over the last25 years. Indeed, the literature we review suggests that health has a pervasive effect onmost outcomes of interest to labor economists including wages, earnings, labor forceparticipation, hours worked, retirement, job turnover, and bene ts packages. But unfortu-
Ch. 50: Health, Health Insurance and the Labor Market3311nately there is no consensus about the magnitude of the effects or about their size relativeto the effects of other variables. We will, however, be able to shed some light on factorsthat cause the estimates to disagree.Much of the best work linking health and labor market outcomes focuses on developingcountries. This may be because the link between health and work is more obvious insocieties in which many prime age adults are under-nourished and in poor health, andalso because the theory of ef ciency wages provides a natural starting point for investigations of this issue. However several excellent recent surveys of health and labor markets indeveloping countries already exist (see Behrman and Deolalikar, 1988; Strauss andThomas, 1998). In order to break newer ground, this survey will have as its primaryfocus papers written since 1980 using US data, although we will refer to the developingcountry literature where appropriate.2. Health and the labor market2.1. Health as human capitalIn his pioneering work on human capital, Becker (1964) drew an analogy between investment'' in health capital and investment in other forms of human capital such as education.This model was further developed by Grossman (1972). A simple version of his modelfollows. First, consumers are assumed to maximize an intertemporal utility function:TXt 1Et 1 1 1 d t Ut 1 B AT11 ;1 where d is the discount rate, B( ) is a bequest function, A denotes assets, and Ut isgiven byUt U Qt ; Ct ; Lt ; Xt ; u1 ; 11t ;2 where Q is the stock of health, C is consumption of other goods, L is leisure, X is avector of exogenous taste shifters, u1 is a vector of permanent individual speci c tasteshifters, and 1 1 denotes a shock to preferences. Utility is maximized subject to thefollowing set of constraints:Qt Q Qt21 ; Gt ; Vt ; Zt ; u2 ; 12t ;3 Ct Yt 1 Pt Gt 2 At11 2 At ;4 Yt It 1 wt Ht 1 rAt ;5 Lt 1 Vt 1 Ht 1 St 1;6 St S Qt ; u3 ; 13t ;7
3312J. Currie and B. C. Madrianwhere G and V are material and time inputs into health production, Z is a vector ofexogenous productivity shifters, u2 are permanent individual speci c productivity shifters, 1 2t is a productivity shock, Y is total income, P represents prices, I is unearnedincome, w is the wage, r is the interest rate, S is sick time, u3 are permanent individualspeci c determinants of illness and 1 3t are shocks that cause illness. Endowments ofhealth and assets, Q0 and A0, are assumed to be given.This model has several features. First, the stock of health today depends on past investments in health, and on the rate of depreciation of health capital (which is one of theelements of u2). Health is valued by consumers both for its own sake and because beingsick is assumed to take time away from market and non-market activities. Non-markettime is an input into both health production and the production of other valued non-marketgoods (e.g., leisure activities). This model can be solved to yield a conditional labor supplyfunction in which labor supply depends on the endogenous health variable. From anempirical point of view, the main implication of the model is that health must be treatedas an endogenous choice.In principle, the stock of education is also determined by endogenous choices. Buteducation is often treated as predetermined since the optimal investment pro le dictatesthat most investment should occur early in the lifecycle (see Weiss, 1986). This is not thecase for health since workers typically start with a large health endowment that must becontinuously replenished as it depreciates and many investments in health occur later inlife. Thus, the endogeneity of health may be a greater potential source of bias than theendogeneity of education in many applications.Still, health is similar to general human capital in more traditional models, since it isvalued by employers and employees take it with them from job to job. One implication isthat individuals will bear the costs of investments in their health so that the costs ofemployer-provided health insurance, for example, should be passed on to employees inthe form of lower wages. On the other hand, if there are complementarities betweenreturns to health and returns to speci c human capital, then employers may be willingto bear some of the costs of investments in health.The simple model outlined above treats wages and all other prices as parametric.However, one of the major foci of the health and labor markets literature is measuringthe effect of health on wages, usually by adding health measures to a standard Mincerianwage function (Mincer, 1974). Thus, a more complete model of the choices faced byindividuals would recognize that investments in health may alter wages. Conversely,wages can affect investments in health, just as they affect educational decisions (Willisand Rosen, 1979). Thus, health is determined endogenously with both wages and laborsupply.An additional possibility is that wages and labor market activity have a direct effect onhealth. There is a large literature examining the effects of labor market activity on health,some of which is surveyed in Ruhm (1996). 1 In principle, exogenous changes in employment or wages can in uence health by directly affecting the probability of workplaceinjury, stress and risk-taking behaviors, by changing the opportunity costs of investments
Ch. 50: Health, Health Insurance and the Labor Market3313in health capital, or by changing the return to health. In this case, the health measure maybe correlated with the error in the wage equation, again suggesting that health ought to betreated as an endogenous choice.In fact, most of the literature surveyed below treats health as an exogenous, if oftenmismeasured, variable. The implicit assumption is that exogenous shocks to health are thedominant factor creating variation in health status, at least in developed countries. Thismay not be an unreasonable assumption given that current health depends on past decisions and on habits that may be very dif cult to break (e.g., smoking, or a preference for ahigh fat diet), and the fact that individuals often have highly imperfect information aboutthe health production function at the time these decisions are made. 2 However, relativelylittle research has been devoted to assessing the empirical importance of the potentialendogeneity bias.One of the main differences between health and other forms of human capital is thathealth capital is often subject to large negative shocks. 3 If variation in current health isdominated by shocks, then uncertainty about the return to investments in health will bevery important, and insurance should play a large role in mediating the relationshipbetween health and the labor market. In his survey of the importance of education ashuman capital, Willis (1986) notes that researchers tend to focus on the supply of education rather than on the determinants of demand for education. An examination of theemployer side of the market is especially important in the health and labor marketsliterature because of the key role of employer provided health insurance in the UnitedStates.2.2. Measurement issues: what is health?The concept of health'' is similar to the concept of ability'' in that while everyone hassome idea of what is meant by the term, it is remarkably dif cult to measure. Failure toproperly measure health leads to a bias similar to ability bias'' (Griliches, 1977) instandard human capital models. That is, if healthier individuals are likely to get moreeducation, for example, then failure to control for health in a wage equation will result inover-estimates of the effects of education. Similarly, if healthier individuals have lowerlabor supply elasticities, then failure to control for heterogeneity due to health in a labor1Most studies of the effects of labor market participation on health have either used micro-data to compare thehealth of the employed and the unemployed, or used aggregate time-series data to look into the responsiveness ofhealth measures such as mortality rates to aggregate economic conditions. Studies using micro-data tend touncover a link between unemployment and various health problems, but these studies generally do not controlfor the potential endogeneity of employment status. Inferences drawn from aggregate data tend to be sensitive tothe exact empirical speci cation chosen. Thus the link between exogenous changes in employment and healthremains controversial.2On the other hand, models of rational addiction'' show that people may start smoking cigarettes for example,even if they realize that the likely consequence is that they will become addicted (Becker and Murphy, 1988).3Altonji (1993) explores the implications of uncertainty in the returns to education and shows that there can belarge differences between ex ante and ex post rates of return.
3314J. Currie and B. C. Madriansupply equation will lead to smaller estimates of the elasticity of labor supply with respectto wages.In one of the rst papers to make this point, Lambrinos (1981) shows that in a sample of18,000 disabled and non-disabled adults from the 1972 Social Security Survey of Disabledand Non-disabled Adults, the estimated elasticity of labor supply (with respect to wages)depends on whether a health variable is included and also on whether or not disability isused to exclude individuals from the sample. 4 The substitution elasticities range from 0.71with no health controls, to 0.59 with a control for disability, to 0.48 in a sample thatexcludes the disabled. Including a health index constructed using data on activity limitations also improved model t by 28%. The size of this health bias'' is likely to vary withthe health measure used, and the exact magnitude may prove as dif cult to pin down as thesize of ability bias'' has been.Ideally we would like some summary measure of health as it pertains to the ability anddesire to work. Such a measure might be called work capacity''. In practice the types ofmeasures usually available can be divided into eight categories: (1) self-reported healthstatus (most often whether someone is in excellent, good, fair or poor health); (2) whetherthere are health limitations on the ability to work; (3) whether there are other functionallimitations such as problems with activities of daily living (ADLs); (4) the presence ofchronic and acute conditions; (5) the utilization of medi
Health is valued by consumers both for its own sake and because being sick is assumed to take time away from market and non-market activities. Non-market time is an input into both health production and the production of other valued non-market goods (e.g., leisure activities). This model can be solved to yield a conditional labor supply