Printer Friendly

Health Insurance Availability at the Workplace.

How Important are Worker Preferences?


Analysts have frequently interpreted the uneven distribution of health insurance across firms of varying size as evidence of insurance market failure in the small group market. We explore an additional explanation by considering the relationship between employee preferences for health insurance and its availability at the workplace. We apply a simple model of job choice to data from the 1987 National Medical Expenditure Survey to examine whether workers with weak preferences for health insurance sort themselves into jobs without coverage. Our results for a sample of single workers are consistent with such sorting behavior.

I. Introduction

The uneven availability of health insurance between small and large firms has been an important focal point of reform efforts to expand coverage to the working uninsured. [1] Analysts have typically interpreted this disparity as evidence of market failure: the unwillingness of insurers, especially those in the small group market, either to provide affordable health insurance or any access to coverage. In this paper, we consider an additional explanation for the failure of some firms to provide health insurance: that the decision of an employer to offer insurance reflects worker preferences and that workers systematically sort themselves into jobs according to their preferences for health insurance. Our results are consistent with such sorting behavior.

As discussed below, the relationship between worker tastes and the form of compensation at the workplace is well-grounded in economic theory and has important policy implications. For example, theory indicates that preferences of the median worker dictate whether certain nonpecuniary benefits are provided, that employers should be indifferent as to the composition of compensation as long as it is commensurate with, and does not adversely affect, worker productivity, and that workers will gravitate to employers that provide a preferred compensation package. On the policy side, initiatives to mandate health insurance are frequently concerned with the efficiency and distributional consequences of requiring groups of workers with heterogeneous tastes for health insurance to purchase coverage (Summers 1989; Melese 1995). In addition, if some workers prefer jobs that do not provide health insurance, reforms to expand coverage in the small group health insurance market may face limited success. Although we do not e xpect the preference hypothesis to dominate market failure as an explanation for the lack of coverage, quantifying its importance can help clarify the efficiency and equity implications of initiatives to expand access to employment-related health insurance.

II. Market Failure, Employee Preferences, and the Availability of Health Insurance

A. Market Failure

The inability of some workers to obtain health insurance from their employers has frequently been attributed to market failure (for example, Cutler 1994a, b; Zellers, McLaughlin, and Frick 1992). This market failure arises from informational asymmetries between employers and insurers regarding workforce health risk and from information costs which preclude insurers from obtaining an accurate actuarial assessment of such risk. Such informational problems have at least two consequences. First, insurers may be unwilling to bear the risk of indemnifying specific groups of employees. Second, insurers may only be willing to provide coverage to certain employers at excessively high premiums which do not reflect the true risk profile of the firm. [2] As a result, individual workers and society suffer welfare losses because both are unable to capture efficiency gains from risk pooling and from externalities associated with the underconsumption of medical care (Arrow 1963; Pauly 1971).

Many employers assert that high premiums are a primary impediment to making coverage available to their workforce and that employee preferences for health insurance are of a second order of importance (Morrisey, Jensen, and Morlock 1994; Sullivan 1990; and McLaughlin and Zellers 1994). Because small firms are unable to take advantage of scale economies from pooling risks over a large and stable employee base, there is some justification for the high premiums they face. However, the notion that market failure has led to prohibitively expensive premiums and/or an inability to obtain coverage is also buttressed by a variety of exclusionary underwriting and rating practices (such as medical underwriting, durational rating, preexisting condition exclusions, and redlining certain firms and employees) that are unique to the small group market (Zellers, McLaughlin, and Frick 1992; Hall 1992; Leibowitz, Damberg, and Eyre 1992). [3]

Although the high premiums associated with such practices may be considered prohibitive, small employers do not appear to be responsive to interventions designed to reduce premiums. In particular, experiments providing subsidized insurance products to small firms have been unsuccessful in inducing employers to offer coverage (McLaughlin and Zellers 1992; Thorpe et al. 1992). [4] As a result, other explanations besides high premiums may warrant consideration.

B. Employee Preferences for Health Insurance

Economic theory and some limited empirical work support the notion that certain workers may prefer wage income to health insurance and may seek employment in firms that fail to provide coverage. In particular, the theory of compensating wage differentials recognizes that workers may be willing to trade specific job attributes, such as fringe benefits or job safety, for increments to wage income (Rosen 1986). In doing so, workers sort themselves among alternative jobs according to their preferences for particular job characteristics, selecting a specific job if the wage differential between preferred and non-preferred jobs exceeds the utility loss from foregoing the job characteristic.

Goldstein and Pauly (1976) first explicitly considered the relationship between employee preferences for health insurance and its availability at the workplace. Comparing health insurance to a local public good, they argued that labor market equilibrium would be characterized by groups of workers with homogeneous preferences selecting jobs consistent with their tastes for health insurance. Pauly (1986) has also argued that labor market sorting with regard to health insurance preferences may also arise due to the exclusion of employer contributions to health insurance from employee taxable income. In addition, Feldman and colleagues' (1997) analysis of small employers' decisions to offer coverage assumes that workers sort themselves among firms according to their insurance preferences. However, none of these studies provide an empirical test of the sorting hypothesis.

Other studies provide some empirical evidence consistent with sorting behavior. Scott, Berger, and Black (1989) found that federal tax laws requiring the nondiscriminatory provision of fringe benefits encouraged employees to sort into firms according to their demand for fringe benefits. Evans and Leighton (1989) and Long and Marquis (1992) found that workers in small firms tend to be young adults, low-wage earners, and part-time workers. Such workers may, at the margin, obtain more utility from a dollar in wage income received with certainty than from the uncertain yield of a dollar allocated to health insurance. Other studies have also emphasized worker demand by noting that large subsidies would be required to induce low-wage workers to participate in either employment-based or nongroup insurance (Marquis and Buchanan 1992; Thomas 1994; Marquis and Long 1995; Chernew, Frick and McLaughlin 1997).

III. Some Descriptive Findings

In Table 1, we present data from the 1987 National Medical Expenditure Survey (NMES) (described in detail below) to examine the relationship between worker preferences and the availability of employment-related health insurance. We consider this relationship for single workers, thus avoiding the complexities associated with disparities in the preferences of married workers and with the process of joint decisionmaking regarding health insurance and employment choices. To measure preferences for health insurance, we construct three variables from responses to the following statements that were collected via a self-administered questionnaire. The first variable is based on the statement "I'm healthy enough that I really don't need health insurance"; the second on "health insurance is not worth the money it costs"; and the third on "I'm more likely to take risks than the average person." For each variable, workers are considered to have "weak" preferences for health insurance if they "agree strongly" or "agree s omewhat" with the specific statement. Note that these variables differ with regard to the basis for preferences for coverage (for example, respondent perceived health, value of insurance given the costs of coverage, and general risk-taking behavior). Finally, responses to these statements were obtained independently of questions about respondent health insurance status. This minimizes the potential bias from correlations between the attitudinal variables and insurance status which can arise when responses to the former questions are a rationale for not having coverage.

Table 1 provides some evidence consistent with the hypothesis that single workers select jobs that reflect their preferences for health insurance. Only 71.8 percent of all single workers who agree that they "don't need health insurance" obtain jobs that offer coverage compared to 78.1 percent of those who disagree (p [less than] .05). This difference is significant for single males (p [less than] .10) but not for single females. The disparity is somewhat sharper with regard to attitudes toward health insurance costs. Some 69.3 percent of single workers who agree that health insurance is not worth the cost are offered coverage compared to 79 percent who disagree (p [less than] .01). Again, the difference is significant for single males (p [less than] .01) but not for single females. Finally, note that our results are statistically significant only when we consider questions that deal directly with health insurance: we find no difference in the percent of workers offered coverage when we compare workers with v arying attitudes toward risk-taking in general.

Our data also suggest that sorting among job opportunities is far from complete and that a sizable proportion of single workers fail to obtain jobs consistent with their preferences for coverage. Using "I'm healthy enough that I really don't need insurance" as our preference measure (data not shown), we find that 29 percent of single workers (7.5 million such workers in 1987) are "mismatched" with regard to their health insurance preferences, either obtaining a job that provides coverage when they report weak preferences (10 percent) or obtaining a job without coverage when they express strong preferences (19 percent). Although this tabulation cannot reveal whether "mismatched" workers have traded insurance for other job attributes, it suggests that some workers may face search costs or other barriers to jobs that meet their tastes for coverage.

IV. Model and Empirical Approach

A. Model of Job Choice

The data presented above provide some support for the notion that workers sort themselves among jobs according to their preferences for health insurance but also reveal that some workers are unsuccessful in obtaining employment consistent with their preferences. To explain why such sorting may be imperfect, we apply a model that considers the relationship between worker preferences and job choice and also recognizes that search costs may preclude some workers from obtaining their desired jobs.

Consider an individual seeking employment among jobs that do and do not provide health insurance. Each job is characterized by a level of wage income, W, search costs, C, and out-of-pocket medical expenses, M, that depend on whether or not the job provides insurance. Let U([W.sub.i], [M.sub.i], [C.sub.i], T) represent a worker's utility from job i where i = 0, n (jobs that offer and do not offer insurance, respectively). Let T represent an individual's preferences for health insurance which affect the position and shape of the utility function (Pudney 1989). In empirical work, T is typically unobserved and proxied by the individual's demographic characteristics. As noted, however, we have direct information on an individual's preferences for insurance.

The worker's problem is whether to seek a job that does or does not provide health insurance. In deciding between job types, a worker will compare the utility from each type of job and obtain one with insurance if:

(1) [U.sub.o]([W.sub.o], [M.sub.o], [C.sub.o]; T) [greater than] [U.sub.n]([W.sub.n], [M.sub.n], [C.sub.n]; T)

To operationalize (1), let [U.sub.i][.] be a linear function of wages, [W.sub.i], search costs, [C.sub.i], two mutually exclusive indicators of tastes, [T.sub.s] (equal to 1 for strong preferences for insurance, 0 otherwise) and [T.sub.w] (equal to 1 for weak preferences, 0 otherwise), and [e.sub.i] a stochastic error term. Let the job with health insurance provide full coverage for medical expenses so that a worker does not face any out-of-pocket medical costs ([M.sub.o] = 0). In contrast, let a job offer without insurance require the worker to pay all medical expenditures out of pocket, and define [M.sub.n] as a vector of factors affecting such expenses. As regards T, we assume that workers gain utility from jobs that match their health insurance preferences so that a worker with [T.sub.s] = 1 gains utility from a job with coverage and a worker with [T.sub.w] = 1 gains utility from a job without coverage. Thus,

(2) [U.sub.o] = [alpha][W.sub.o] + [beta][T.sub.s] - [gamma][C.sub.o] + [e.sub.o] (job with health insurance)

(3) [U.sub.n] = [alpha][W.sub.n] + [beta][T.sub.w] - [gamma][C.sub.n] - [gamma][M.sub.n] + [e.sub.n] (job without health insurance).

Substituting (2) and (3) into (1), a worker will take a job with health insurance if:

(4) [alpha][W.sub.o] + [beta][T.sub.s] - [gamma][C.sub.o] + [e.sub.o] [greater than] [alpha][W.sub.n] + [beta][T.sub.w] - [gamma][C.sub.n] - [gamma][M.sub.n] + [e.sub.n]

Rearranging terms and assuming that ([e.sub.n] - [e.sub.o]) is logistically distributed, the probability that a worker obtains a job with insurance is given by:

(5) - [alpha]([W.sub.n] - [W.sub.o]) + [gamma][M.sub.n] + [beta]([T.sub.s] - [T.sub.w]) - [gamma]([C.sub.o] - [C.sub.n])

This probability is negatively related to the wage difference between jobs without and with insurance ([W.sub.n] - [W.sub.o]) and positively related to factors that increase expected out-of-pocket medical expenditures, [M.sub.n]. Note that workers with strong preferences for jobs with coverage ([T.sub.s] = 1, [T.sub.w] = 0) will be more likely to take a job with coverage while workers with weak preferences for insurance ([T.sub.w] = 1, [T.sub.s] = 0) will be less likely to take a job with insurance. The likelihood of obtaining a job with coverage also decreases as the cost of finding such a job increases compared to that of finding a job without coverage. We estimate Equation 5 as a reduced form logit equation and include exogenous regressors for the terms described above.

B. Data

The data used in this analysis are from two components of the 1987 NMES: the Household Survey and the Health Insurance Plans Survey (HIPS). [5] The NMES household component surveyed approximately 15,000 households (36,000 individuals) from the civilian, noninstitutionalized population during 1987 and provides detailed information on their medical care use and expenditures, sources of payment, demographic and employment characteristics, income, health insurance, and health status. The HIPS component verified information on private insurance reported by household respondents and provided data on the availability and costs of coverage. For our analysis, the key variable from HIPS is whether an employer offered coverage to the NMES household respondent.

Our empirical work relies on the three preference variables used in our descriptive tables: "I'm healthy enough that I really don't need health insurance," "health insurance is not worth the money it costs," and "I'm more likely to take risks than the average person." [6] A negative relationship between weak preferences and the likelihood that a worker obtains an insurance offer would provide evidence consistent with sorting behavior.

As our model suggests, the greater the wage differential between jobs without and with health insurance, ([W.sub.n] - [W.sub.o]), the less likely a worker will take a job that offers coverage. [7] We assume that this wage differential (the gain from taking a job without coverage) is positively associated with the

following instruments that reflect the costs of insurance: the worker's occupation, state health insurance taxation rates, and medical care costs in the worker's county (measured by the Medicare Prevailing Charge Index from the Area Resource File). [8] Variables for region and urban or rural locale are also included to account for any unobserved geographic factors that may be correlated with premiums.

Other instruments for the wage differential include worker characteristics (for example, age, gender, race/ethnicity, and education). Certain workers, such as those who are older or better educated, may receive more generous health benefits than others. As a result, such workers may obtain a larger wage gain from taking a job without coverage than will others, thereby reducing their likelihood of obtaining a job with coverage. However, we would also expect older and better educated workers to have greater access to jobs with health insurance, and therefore, to face lower search costs for such jobs relative to younger and less educated workers. As a result, increases in age and education may be positively associated with offer probabilities. Thus, the reduced form coefficients on such worker characteristics will reflect these competing influences.

As noted in our model, factors associated with higher expected out-of-pocket medical care costs increase the likelihood of obtaining a job with insurance. These costs are assumed to be positively associated with the presence of household members in fair or poor health or with chronic conditions (for example, cancer, heart disease, diabetes), with household size, the presence of children less than five years of age, worker age, and with the Medicare Prevailing Charge Index. In contrast, out-of-pocket costs are expected to be lower for workers with other exogenous work-related health insurance. [9]

We also control for the county unemployment rate (a proxy for labor market conditions) and for the proportion of jobs in small establishments in the worker's county of residence. The latter variable serves as a proxy for the cost difference between searching for job offers with and without health insurance ([C.sub.0] - [C.sub.n]). In particular, as the proportion of jobs in small establishments rises, ([C.sub.0] - [C.sub.n]) increases since more effort and resources are required to find a job with coverage and less effort is required to find a job without coverage. Thus, the likelihood that a worker obtains a job with insurance is reduced. We derived two measures of the proportion of jobs in small establishments from the Census Bureau's 1987 data on County Business Patterns (the proportion of jobs in establishments of less than 10 employees and of 10 to 19 employees).

Our sample includes wage and salaried workers who are single (including persons never married, widowed, divorced, or separated), nonagricultural employees between the ages of 18 and 64. We excluded workers enrolled in Medicaid or other indigent care programs since their characteristics (that is, skill-levels and work experience) and the constraints on their earnings capacity make them unlikely candidates for jobs that offer health insurance. [10] We also excluded workers who were dependents on another employment-related health plan, viewing such joint decision making between dependent and policyholder as endogenous and beyond the scope of our analysis. Our resulting sample consists of 2,219 single workers. Finally, we note that a small proportion of workers in our sample (9.6 percent using sample weights) are students and thus may have access to health insurance through schools or through their parents. Because access to such alternative coverage could affect employment decisions of such workers, we control for student status in our empirical work. [11]

V. Empirical Results

In Table 2 we present logit estimates of the reduced form job choice equation using alternative measures of health insurance preferences. [12] Our results support the hypothesis that workers select jobs that reflect their tastes for health insurance. Entering each preference variable separately, we find that single workers who report themselves as "healthy enough and not in need of health insurance" (Specification 1) and those who agree that "health insurance is not worth the cost" (Specification 2) are less likely to take a job with coverage than workers who disagree (p [less than] .05 and [less than] .01, respectively). As in Table 1, we find no significant effect on job choice by workers who report themselves more generally to be risk-takers (data not shown). In Specification 3, we find the coefficients of both preference variables to be jointly significant (p [less than] .02). However, we find that only "health insurance is not worth the cost" remains statistically significant (p [less than] .02).

We use the results from Specification 3 to assess the extent to which preferences and other characteristics lead some uninsured workers to take jobs without coverage. We first evaluate specification 3 at the mean values for uninsured workers, assigning weak preferences for both preference measures. This yields a baseline insurance offer probability for these workers. Next, we substitute mean demographic and employment characteristics of all insured workers in Specification 3 holding weak preferences constant. This allows us to assess how differences in human capital and other characteristics between insured and uninsured workers affect offer probabilities. Finally, we hold the employment and demographic characteristics of uninsured workers at their mean values and assign strong preferences (first evaluating the preference variables separately and then jointly). We find that a change from weak to strong preferences yields the following effects relative to the baseline probability that an uninsured worker with weak preferences will obtain a job with coverage (baseline probability of .58): a .04 increase using "I'm healthy enough and really don't need insurance" and a .10 increase using "health insurance is not worth the cost" yielding a joint increase of .14. In comparison, differences in employment and demographic characteristics between uninsured and insured workers yield a .17 increase in offer probabilities. This simulation suggests that preferences (when taken together) may be as important as other characteristics in explaining the insurance status of uninsured workers with weak preferences. [13]

Results for other variables included in Specification 3 are also consistent with our job choice model. We find that the county unemployment rate and proportion of jobs in small establishments (10-19 workers) obtain the expected negative signs (P [less than] .01 and P [less than] .10, respectively). As noted above, the signs of the coefficients on worker age and education cannot be determined a priori. We find that younger workers and those with 12 years of schooling are less likely to obtain jobs with coverage than older workers and those with 16 years of schooling, respectively. These results are consistent with the fact that such workers are less likely to have the requisite skills and experience necessary to obtain jobs with coverage. We also find that workers who are students are less likely than others to obtain a job with insurance (p [less than] .01), perhaps reflecting their access to alternative sources of group coverage.

Although we find no direct effect of correlates of expected medical care costs, such as health status, health conditions, family size, or the presence of small children on the likelihood of obtaining a job with coverage, we find that this probability declines when a worker holds other exogenous work-related coverage which can defray such costs. Finally, we find that workers in the South and West relative to those in the Northeast are less likely to obtain jobs with coverage.

A. Preferences for Coverage and Realized Health Expenditures

Our results provide support for the hypothesis that some workers select jobs based upon their preferences for health insurance. An important issue to consider in this regard is whether workers with weak preferences for coverage are making rational choices when they select jobs without insurance. In particular, do their choices, and hence their preferences for coverage, appear to be consistent with their actual use of health care services? Put differently, does such a job choice reflect a kind of "cognitive dissonance" in which workers choose to believe that they will not incur medical expenditures and decide to forego health insurance despite objective evidence to the contrary? [14] To explore these issues, we briefly consider the relationship among worker preferences, insurance status, and realized health care expenditures.

In Table 3 we report actual health care expenditures for single workers according to their preferences for health insurance and their insurance status. These data reveal some consistency between worker performances and realized expenditures. Overall, workers who report that the they are "healthy enough and don't need insurance" incur expenditures that are on average $423 less than workers who believe that their health status warrants health care coverage (p [less than] .01). This difference remains even after controlling for health insurance status: uninsured and insured workers having weak preferences incur lower health expenditures than their counterparts with strong preferences. In fact, uninsured single workers with weak preferences have the lowest average expenditures ($289, some $689 below insured workers with strong preferences; p [less than] .01). The demographics of single workers with weak preferences are also consistent with their stated preferences. Compared to workers with strong preferences, th ey are younger (31.6 to 34.2 years) and are less likely to be female (40.2 versus 55.3 percent), an important correlate of health care use. Workers with weak preferences are also more likely to report excellent health (42.2 to 33.1 percent), and are half as likely to be in fair or poor health (5.0 to 10.6 percent; data not shown).

Next we consider responses to the statement "health insurance is not worth the money it costs." As noted above, this variable emphasizes the value of health insurance relative to its cost rather than to general health needs. Thus, the relationship between preferences and realized health expenditures may differ from that observed above. Although we do find that workers with weak preferences have lower average expenditures than those with strong preferences, the result is not statistically different. Further, uninsured workers with weak preferences have higher expenditures than their insured counterparts, although again this result is not statistically significant. Only among insured workers do we find that workers with weak preferences have lower average health care expenditures than those with strong preferences (p [less than].01). Considering demographics, workers with weak preferences are younger (32.7 to 34.1 years) and less likely to be female (47.4 to 54.6 percent) than those with strong preferences. Ho wever, we observe no differences in reported health status. Although the above evidence could be used to support the notion that workers who do not believe that health insurance is worth the cost are irrational in their job choice, caution must be used before reaching such a conclusion since this preference variable is based upon the cost of coverage rather than on health needs. It may be that economic circumstances and competing resource demands (for example, for food, shelter, etc.) rather than health status lead these workers to undervalue health insurance.

VI. Conclusions

Market failure has emerged as the most common explanation for why some employers fail to provide health insurance. In this paper, we have considered another reason. We have hypothesized that an employer's decision to offer coverage reflects worker preferences and that workers with weak tastes for health insurance seek employment in firms that do not provide coverage in order to obtain higher wages. Our results for single workers are consistent with this sorting behavior and suggest that preferences may be as important as worker characteristics in explaining the health insurance status of uninsured workers with weak preferences. We also find that workers who believe that their health status does not warrant health insurance do incur lower health expenditures than other workers. In addition, they report themselves to be in better health than workers with strong preferences.

Our findings are consistent with other research that has emphasized worker preferences as a reason why coverage may not be available at the workplace. As has been documented, workers in firms that do not provide health insurance tend to be young and low-wage and may prefer an additional dollar of wage income to the uncertain yield of a dollar allocated to health insurance. These workers also appear to benefit little from the tax subsidy for employer contributions to health insurance (Monheit, Nichols, and Selden 1996). Thus, the relatively low expected payoff from insurance, along with strong preferences for wage income and the availability of free or low cost public health care may explain why some workers prefer jobs without coverage.

As regards public policy, our study is perhaps most relevant for the debate over mandated health insurance. Requiring certain employers to provide coverage is likely to result in private welfare losses for those workers who prefer wage income to health insurance and who, therefore, may be uninsured by choice. Such losses could be reduced by restricting the mandate to certain kinds of low-cost coverage (i.e., catastrophic coverage) instead of more comprehensive and expensive health plans. Wage subsidies could also be provided to specific groups of low-wage workers in an effort to target those who may undervalue the benefits from an insurance mandate.

Our findings also suggest that the limited success achieved to date by small group market reform (Jensen and Morrisey 1997; Markus, Ladenheim, and Atchison 1995) may reflect its emphasis on altering insurer behavior. Recognizing that sorting behavior may be partly responsible for the lack of coverage at small firms could refocus efforts on creating incentives which raise the value of health insurance to the targeted group of small firm employees. Finally, our analysis of the role of heterogeneous preferences for health insurance also has implications for discussions of the relative merits of mandates versus direct tax-supported efforts to expand coverage (Summers 1989). If some workers do not fully value health insurance benefits, then the efficiency gain from imposing a mandate may not be forthcoming.

Alan C. Monheit is the director of the Division of Social and Economic Research, and Jessica Primoff Vistnes is a senior economist at the Center for Cost and Financing Studies, Agency for Health Care Policy and Research (AHCPR). The views expressed in this paper are those of the authors; no official endorsement by AHCPR or the Department of Health and Human Services is intended or should be inferred. The authors wish to thank Philip Cooper, Barbara Schone, Kathy Swartz, two anonymous referees, and seminar participants at AHCPR and the Georgetown University Health Policy Center for helpful comments. Chao-Sung Yu and Jane Faulman of Social and Scientific Systems, Bethesda, Maryland provided excellent computer programming support. The data used in this analysis can be obtained beginning June 2000 through May 2003 from Alan C. Monheit, (, Center for Cost and Financing Studies, AHCPR, 2101 East Jefferson Street, Suite 500, Rockville, MD 20852. [Submitted February 1998; accepted March 1999]

(1.) Monheit and Vistnes (1997) report that in 1996, 30 percent of wage earners in establishments of fewer than ten workers were uninsured as were 22 percent in establishments with between ten and 24 workers. Only 7 percent of those in establishments of 500 or more workers lacked coverage.

(2.) Leibowitz, Damberg, and Eyre (1992) note that in any one year, a large share of the observed high use of medical care by small firm employees may be random.

(3.) Such practices were common prior to small group market reform legislation passed by many states in the early 1990s. Such legislation included measures to regulate premiums, to guarantee issue and renewal of policies, and to ensure continuity of coverage (U.S. General Accounting Office 1995). The Health Insurance Portability and Accountability Act of 1996 (HIPAA) prohibits employers and insurers from using pre-existing health conditions to deny employment-based or nongroup coverage to the previously insured who have changed or lost jobs.

(4.) See Morrisey, Jensen, and Morlock (1994) and Feldman et al. (1997) for contrary views.

(5.) Edwards and Berlin (1989) discuss the NMES household survey and Emmons and Hill (1994) describe HIPS.

(6.) We also considered several other variables reflecting respondent attitudes toward medical care (for example, "I can overcome most illnesses without help from a medically trained professional," home remedies are better than drugs prescribed by a doctor," etc.). We did not find a significant association between such variables and whether a worker obtained a job offer with insurance.

(7.) We assume workers bear the full premium costs of coverage so that [W.sub.n] [greater than] [W.sub.o].

(8.) Since employment-based group coverage is experience rated, industry and firm size are obvious correlates of premiums. However, we view these variables as endogenous to job choice. We consider worker occupation to be an exogenous correlate of coats in this context since certain occupations may be differentially associated with health risk.

(9.) Such coverage was mainly obtained directly from a labor union and held by less than two percent of our sample. We hypothesize that the decision to obtain union coverage is independent of the decision to seek employment at a particular job. Our data do not permit us to examine whether union coverage began prior to the start of the current job.

(10.) See Yelowitz (1995) for a similar argument. The proportion of individuals with Medicaid in our sample with jobs offering such coverage was substantially lower than for other workers.

(11.) Most of these students had strong labor force attachment, were older, and were all beads of households: roughly three-quarters worked full-time (more than 35 hours per week) and two-thirds were over age 25 (and unlikely to have access to coverage through a parent).

(12.) NMES data have been obtained from a multi-stage probability sampling design which resulted in a clustering of households and unequal probabilities of sample selection. We use STATA software to apply sampling weights to adjust parameter estimates and standard errors for these design effects. Failure to apply sampling weights will yield parameter estimates that are not representative of the population under consideration, while failure to adjust for possible correlation of responses within a cluster will understate coefficient standard errors.

(13.) Among single uninsured workers, 15.3 percent agree that their health needs do not require health insurance, 20.6 percent agree that health insurance is not worth the cost, and 8.3 percent agree with both statements.

(14.) In this regard, the psychological benefits to a worker of believing that he/she won't need health insurance (possibly because of an unpleasant association between insurance and the likelihood of serious illness or use of the health care system) exceed the costs associated with the probability of incurring medical expenditures. Akerlof and Dickens (1982) discuss cognitive dissonance in an economic context with regard to worker safety and Meltzer et al. (1998) apply the concept with regard to patient's reported versus actual choice of treatment for diabetes.


Akerlof, George, and William Dickens. 1982. "The Economic Consequences of Cognitive Dissonance." American Economic Review 72(3):307-19.

Arrow, Kenneth. 1963. "Uncertainty and the Welfare Economics of Medical Care," American Economic Review 53(5):941-73.

Cutler, David. 1994a. "A Guide to Health Care Reform," Journal of Economic Perspectives 8(3):13-29.

_____. 1994b. "Market Failure in Small Group Health Insurance." NBER Working Paper No. 4879. Cambridge, Mass.: National Bureau of Economic Research, October.

Chernew, Michael, Kevin Frick, and Catherine Mclaughlin. 1997. "The Demand for Health Insurance Coverage by Low Income Workers: Can Reduced Premiums Achieve Full Coverage?" Health Services Research 32(4):453-70.

Edwards, W. Sherman, and Martha Berlin. 1989. Questionnaire and Data Collection Methods for the Household Survey (Pub. No. 89-3450): National Medical Expenditure Survey Methods 2, Agency for Health Care Policy and Research, Public Health Service, Rockville. Md.

Emmons, Carol, and Craig Hill. 1994. Questionnaire and Data Collection Methods for the Health Insurance Plan Survey (Pub. No. 94-0016) National Medical Expenditure Survey Methods 5, Agency for Health Care Policy and Research, Public Health Service, Rockville, Md.

Evans, David, and Linda Leighton. 1989. "Why Do Small Firms Pay Less?" Journal of Human Resources 24(2):299-317.

Feldman, Roger, Bryan Dowd, Scott Leitz, and Lynn Blewett. 1997. "The Effect of Premiums on the Small Firm's Decision to Offer Health Insurance." Journal of Human Resources 32(4):635-56.

Goldstein, Gerald, and Mark Pauly. 1976. "Group Health Insurance as a Local Public Good." In The Role of Health Insurance in the Health Services Sector. ed. Richard Rosett, 73-110. New York: Neale Watson Academic Publications for the National Bureau of Economic Research.

Hall, Mark 1992. "The Political Economics of Health Insurance Market Reform," Health Affairs 11(2):7-27.

Jensen, Gall, and Michael Morrisey. 1997. "Small Group Reform and Insurance Provision by Small Firms, 1989-1995." Unpublished manuscript. Detroit: Wayne State University.

Leibowitz, Arleen, and Michael Chernew. 1992. "The Firm's Demand for Health Insurance." In Health Benefits and the Workforce, 77-84. U.S. Department of Labor, Pension and Benefits Administration. Washington, D.C.: GPO.

Leibowitz, Arleen, Cheryl Damberg, and Kathleen Eyre. 1992. "Multiple Employer Arrangements." In Health Benefits and the Workforce, 55-76. U.S. Department of Labor, Pension and Benefits Administration. Washington, D.C.: GPO.

Long, Stephen, and M. Susan Marquis. 1992. "Gaps in Employment-Based Insurance: Lack of Supply or Lack of Demand?" In Health Benefits and the Workforce, 37-42. U.S. Department of Labor, Pension and Benefits Administration. Washington, D.C.: GPO.

Markus, Anne, Kala Ladenheim, and Lisa Atchison. 1995. Small Group Market Reforms: States' Experiences. New York: The Commonwealth Fund.

Marquis, Susan M., and Stephen Long. 1995. "Worker Demand for Health Insurance in the Nongroup Market." Journal of Health Economics 14(1):47-64.

Marquis, Susan M. and Joan Buchanan. 1992. "Subsidies and National Health Care Reform: the Effect on Workers Demand for Health Insurance Coverage." In Health Benefits and the Workforce, 85-92. U.S. Department of Labor, Pension and Benefits Administration Washington, D.C.: GPO.

McLaughlin, Catherine. 1993. "The Dilemma of Affordability: Private Sector Health Insurance for Small Businesses," In American Health Policy: Critical Issues for Reform, ed. Robert Helms, 133-51. Washington, D.C.: AEI Press.

McLaughlin, Catherine, and Wendy Zellers. 1992. "The Shortcoming of Voluntarism in the Small Group Market." Health Affairs 11(2):28-40.

McLaughlin, Catherine and Wendy Zellers. 1994. Small Business and Health Care Reform: Understanding the Barriers to Employee Coverage and Implications for Workable Solutions. Ann Arbor: University of Michigan School of Public Health.

Melese, Francois, 1995. "Government-Mandated Benefits, Taxes, and Wages, Southern Economic Journal 62(1):53-70.

Meltzer, David, Tamar Polonsky, Brian Egelston, and Janet Tobian. 1998. "Do Quality Adjusted Life Years Reflect Patient Preferences? Validation using Revealed Preference for Intensive Therapy in Insulin-Dependent Diabetes Mellitus. Unpublished manuscript. Chicago: University of Chicago.

Monheit, Alan, Len Nichols, and Thomas Selden. 1996. "How are Net Health Insurance Benefits Distributed in the Employment-Related Insurance Market?" Inquiry 32(4):379-81.

Monheit, Alan, and Jessica Vistnes. 1997. "Health Insurance Status of Workers and their Families: 1996." Rockville, MD: Agency for Health Care Policy and Research. MEPS Research Findings No. 2. AHCPR Pub. No. 97-0065.

Morrisey, Michael, Gail Jensen, and Robert Morlock. 1994. "Small Employers and the Health Insurance Market," Health Affairs 13(5):149-61.

Pauly, Mark. 1971. Medical Care at Public Expense: A Study in Applied Welfare Economics. New York: Praeger.

Pauly, Mark. 1986. "Taxation, Health Insurance, and Market Failure in the Medical Economy," Journal of Economic Literature 24(2):629-75.

Pudney, Stephen. 1989. Modelling Individual Choice: The Econometrics of Corners, Kinks, and Holes. Oxford, U.K.: Basil Blackwell Ltd.

Rosen, Sherwin. 1986. "The Theory of Compensating Differences," In Handbook of Labor Economics Volume 1, ed. Orley Ashenfelter and Richard Layard, 642-92. New York: Elsevier Science.

Scott, Frank, Mark Berger, and Dan Black. 1989. "Effects of the Tax Treatment of Fringe Benefits on Labor Market Segmentation," Industrial and Labor Relations Review 42(2):216-29.

Sullivan, Cynthia. 1990. "Why Employers do not Offer Coverage." In Providing Employee Health Benefits: How Firms Differ. Washington, D.C.: Health Insurance Association of America.

Summers, Lawrence. 1989. "Some Simple Economics of Mandated Benefits." American Economic Review 79(2):177-82.

Thomas, Kathleen, 1994. "Are Subsidies Enough to Encourage the Uninsured to Purchase Health Insurance? An Analysis of Underlying Behavior." Inquiry 31(4):415-24.

Thorpe, Kenneth, Anne Hendricks, Deborah Garnick, Karen Donelan, Joseph Newhouse, et al. 1992. "Reducing the Number of Uninsured by Subsidizing Employment-Based Health Insurance," Journal of the American Medical Association 267:945-48.

U.S. General Accounting Office. 1995. Health Insurance Regulation: Variation in Recent State Small Employer Health Insurance Reforms. Washington, D.C., June.

Yelowitz, Aaron. 1995. "The Medicaid Notch, Labor Supply, and Welfare Participation: Evidence from Eligibility Expansions." Quarterly Journal of Economics 1 10(4):909-39.

Zellers, Wendy, Catherine McLaughlin, and Kevin Frick. 1992. "Small Business Health Insurance: Only the Healthy Need Apply." Health Affairs 1l(1):174-80.
        Worker Preferences and Employment-Related Health Insurance:
       Percent of Single Workers with Jobs Offering Health Insurance
                                        Percent of Each Group Offered
                                              Health Insurance
                                                    Total             Male
All single workers                                  77.2%             76.1%
Worker Preferences
 Don't really need health insurance
   Agree                                            71.8              70.9
   Disagree                                         78.1              77.2
 Health insurance is not worth the cost
   Agree                                            69.3              66.4
   Disagree                                         79.0              78.7
 More likely to take risks
   Agree                                            76.4              78.4
   Disagree                                         77.5              74.5
Number of observations [a]                           2,098              919
All single workers                      78.2%
Worker Preferences
 Don't really need health insurance
   Agree                                73.1
   Disagree                             78.8
 Health insurance is not worth the cost
   Agree                                72.6
   Disagree                             79.3
 More likely to take risks
   Agree                                72.8
   Disagree                             79.5
Number of observations [a]              1,179
Source: 1987 National Medical Expenditure Survey:
Household Survey and Health Insurance Plans Survey.
(a.)Sample subset to observations with nonmissing preference variables.
             Estimates from the Reduced Form Logit Equations to
                          Predict Offered Insurance
                    Specification                                   Means
Independent                                                       (Standard
Variables                (1)           (2)            (3)        Deviations)
Intercept             -1.6740 [*]    -1.5836 [*]    -1.620l [*]      --
                      (0.9363)       (0.9435)       (0.9522)
Age                    0.1447 [**]    0.1423 [**]    0.1438 [**]   33.82
                      (0.0366)       (0.0370)       (0.0368)      (11.00)
Age squared           -0.0015 [**]   -0.0015 [**]   -0.0015 [**] 1264.90
                      (0.0005)       (0.0005)       (0.0005)     (857.49)
Healthy don't need    -0.3071 [**]                  -0.1719
 insurance: agree     (0.1559)          --          (0.1574)        0.1213
Insurance not worth      --          -0.4896 [**]   -0.4540 [**]
 the cost: agree                     (0.1860)       (0.1901)        0.1687
Number of kids ages    0.0497         0.0565         0.0497         0.0555
 0-5 in family        (0.2799)       (0.2818)       (0.2825)       (0.2537)
Family size            0.2962         0.3149         0.3405         1.28
                      (0.2182)       (0.2170)       (0.2246)       (0.67)
Number of family
 members in fair/     -0.0649        -0.0429        -0.0637         0.11
 poor health          (0.1738)       (0.1752)       (0.1751)       (0.32)
Number of adult
 family members
 with chronic con-     0.0768         0.0912         0.0833         0.23
 ditions              (0.1671)       (0.1668)       (0.1679)       (0.42)
Number of child
 family members
 with chronic con-    -0.2093        -0.2368        -0.2338         0.09
 ditions              (0.2720)       (0.2732)       (0.2759)       (0.28)
CHAMPUS               -0.3130        -0.2802        -0.3227
                      (0.3361)       (0.3366)       (0.3399)        0.02
Medicare prevailing
 charge index--        0.0007         0.0011         0.0010       281.55
 1984                 (0.0016)       (0.0016)       (0.0016)      (56.32)
Unemployment rate,    -0.0685 [**]   -0.0696 [**]   -0.0697 [**]    5.86
 1987                 (0.0213)       (0.0218)       (0.0219)       (2.32)
Premium tax levels-   -0.0372        -0.0431        -0.0429         1.86
 in state             (0.0899)       (0.0912)       (0.0910)       (0.79)
Gender: 1 = male,     -0.0200        -0.0201        -0.0222
 2 = female           (0.1265)       (0.1232)       (0.1242)        1.53
Region: Midwest       -0.2371        -0.2689        -0.2652
                      (0.2211)       (0.2224)       (0.2231)        0.25
Region: South         -0.4001 [**]   -0.4143 [**]   -0.4144 [**]
                      (0.1900)       (0.1886)       (0.1905)        0.34
Region: West          -0.4829 [**]   -0.5085 [**]   -0.5007 [**]
                      (0.2168)       (0.2159)       (0.2143)        0.21
SMSA: 19 Largest             -0.0082         -0.0286        -0.0269
 SMSA                        (0.2114)        (0.2146)       (0.2146)
SMSA: Other                  -0.0775         -0.0833        -0.0861
 SMSA                        (0.1569)        (0.1602)       (0.1615)
Hispanic                     -0.4047 [**]     0.4234 [**]   -0.4189 [**]
                             (0.1786)        (0.1824)       (0.1834)
Black                        -0.0049          0.0076         0.0151
                             (0.1516)        (0.1537)       (0.1542)
Education: [less than] 12    -0.5800 [**]    -0.5656 [**]   -0.5602 [**]
 years                       (0.1566)        (0.1580)       (0.1569)
Education: 13 to 15           0.2999 [*]      0.2992 [*]     0.2948 [*]
 years                       (0.1744)        (0.1743)       (0.1742)
Education: 16 years           0.6026 [**]     0.6183 [**]    0.6104 [**]
                             (0.2578)        (0.2583)       (0.2573)
Education: [less than] 16    -0.0181         -0.0308        -0.0402
 years                       (0.2692)        (0.2746)       (0.2739)
Percent of establish-
 ments in county:             1.5102          1.5547         1.5207
[less than] 10 employees     (1.8698)        (1.8874)       (1.9065)
Percent of establish-
 ments in county:            -6.5709 [*]     -6.7365 [*]    -6.719 [*]
  10-19 employees            (3.5813)        (3.6578)       (3.654)
Has other employ-
 ment-related insur-         -1.8205 [**]    -1.8574 [**]   -1.8629 [**]
 ance                        (0.4208)        (0.4309)       (0.4291)
Is currently enrolled        -0.4958 [**]    -0.4814 [**]   -0.4806 [**]
 in school                   (0.2020)        (0.2044)       (0.2051)
Log of likelihood
 function                 -1056.51        -1052.02        1048.80
Number of observa-
tions                      2219            2219           2214 [a]
SMSA: 19 Largest
 SMSA                        0.31
SMSA: Other
 SMSA                        0.49
Education: [less than] 12
 years                       0.15
Education: 13 to 15
 years                       0.22
Education: 16 years
Education: [less than] 16
 years                       0.09
Percent of establish-
 ments in county:            0.16
[less than] 10 employees    (0.05)
Percent of establish-
 ments in county:            0.11
  10-19 employees           (0.02)
Has other employ-
 ment-related insur-
 ance                        0.02
Is currently enrolled
 in school                   0.096
Log of likelihood
 function                     --
Number of observa-
tions                     2219

Note: (*.)significant at p [less than].10.

(**.)significant at p [less than].05. The reduced form equations also control for occupation and for missing responses to the preference variables and to the self-administered questionnaire at the person level and family level (for adults). To obtain marginal probabilities, multiply the coefficients in Specifications 1 by .155 and in Specifications 2 and 3 by .154. The marginal probabilities were calculated for each observation and then averaged. Omitted categories include: Preference variables: Disagree, Region: Northeast, SMSA: Not an SMSA, Race/ethnicity: White and other, Occupation: Service Workers, Education: 12 years. Standard deviations are presented for continuous variables. Standard errors and standard deviations are in parentheses. Family members include the single worker and any children under age 18 or children who are full-time single students between the ages of 18 and 23. Adults include the single worker and any children ages 18-23 who are full-time single students.

(a.)Excludes five observations with valid responses to one preference measure but missing responses to the other preference measure.
           Realized Health Expeneditures by Worker Preference and
                          Health Insurance Status
                           All Single  Single Workers   Single Workers
                            Workers   Without Insurance With Insurance
Total                         $833          $476             $935
                              (49)          (63)             (63)
I'm healthy enough I don't
  need insurance:
 Agree                        $472          $289             $543
                              (74)          (87)             (96)
 Disagree                     $895          $561             $978
                              (52)          (90)             (66)
Health insurance is not
  worth the cost
 Agree                        $762          $836             $735
                              (90)          (258)            (82)
 Disagree                     $854          $406             $962
                              (56)          (56)             (71)
Note: The means were obtained using sample weights. The
standard errors are in parentheses and were corrected
for design of the NMES.
COPYRIGHT 1999 University of Wisconsin Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1999 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Monheit, Alan C.; Vistnes, Jessica Primoff
Publication:Journal of Human Resources
Date:Sep 22, 1999
Previous Article:Latent Work Disability and Reporting Bias.
Next Article:The Effect of Child Care Characteristics on Child Development.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters