Choosing To Be Uninsured:

Determinants and Consequences of the Decision to Decline Employer-Sponsored Health Insurance

October 1999
Peter J. Cunningham



ABSTRACT

Objective. (1) To identify the factors associated with the decision to decline employer-sponsored health insurance, and (2) to compare health care access and use of uninsured persons who declined employer-sponsored insurance with health care access and use of other uninsured and insured persons.

Data Source.1996-97 Community Tracking Study Household Survey.

Study Design. Potential determinants of the decision to decline employer-sponsored coverage were identified as (1) those relating to the cost of insurance to the individual (e.g., family income, job characteristics, types of plans offered); (2) availability of free sources of health care in the community (e.g., public hospitals, community health centers); and (3) factors that reflect individuals’ need and preferences for insurance (e.g., health status, demographics, attitudes toward risk). Measures of access to care and health care use include usual source of care, difficulty obtaining needed medical services, the volume and pattern of ambulatory medical care use and the rate of inpatient utilization.

Data Collection. A survey, primarily by telephone, of households in 60 communities, defined as metropolitan statistical areas and nonmetropolitan areas.

Principal Findings. Factors that reflect the cost of insurance to individuals were strong determinants of the decision to decline coverage in favor of being uninsured, while the availability of free sources of health care in the community had little effect. Health status was not significantly related to the decision to decline coverage, although other individual characteristics, such as age, race/ethnicity, education and attitudes toward risk had significant effects. Uninsured persons with access to employer-sponsored insurance were similar to other uninsured persons in their health care use and access.

Conclusions. Individuals are not motivated to decline employer-sponsored insurance because they are able to achieve relatively good access to care without insurance, but appear to be motivated mostly by concerns about the affordability of insurance and, to some extent, individual preferences. The decision to decline insurance (in favor of being uninsured) appears to have serious consequences regarding access to care and health care use. Policy implications regarding the use of tax incentives to encourage individuals to purchase insurance are discussed.

Key Words. Uninsured, employer-sponsored health insurance, take-up, access

INTRODUCTION
Employer-sponsored private insurance is the most common source of health coverage for non-elderly persons. While the vast majority of persons who are offered coverage by their employers accept that coverage, concern about the relatively small number of persons who refuse coverage may be justified for at least two reasons. First, the "take-up" rate (percentage of workers who accept coverage when offered and eligible) decreased from 88 percent in 1987 to 80 percent in 1996, and appears to be a major reason for the overall decline in employer-sponsored coverage during that period (Cooper and Schone 1997). Second, even a low rate of declining coverage among all workers may still result in a substantial number of uninsured persons who have access to employer-sponsored coverage coverage, either through their own employer or as a dependent of a family member who is offered coverage (Thorpe and Florence 1999). If more of these individuals could be encouraged to accept insurance when offered, then a sizeable reduction in the number of uninsured persons might be achieved at lower public cost than other options.

Nevertheless, little is known about what causes individuals to decline employer-sponsored coverage–especially when the result of that decision is to be uninsured–and what are the consequences of that decision regarding access to care and health care use. The few studies that exist on the topic have been concerned primarily with estimating the rate at which individuals take up or decline employer-sponsored health insurance (Cooper and Schone 1997; Thorpe and Florence 1999; Gabel et al. 1999; Long and Marquis 1993). These studies have also shown that take-up rates tend to be lower for young adults (age 19-24), blacks and Hispanics, low-wage workers, part-time workers and workers employed in smaller firms (Cooper and Schone 1997; Thorpe and Florence 1999; Gabel et al. 1999).

However, these analyses were descriptive only; they did not include multivariate analyses showing the unique effects of a much broader range of individual and employer characteristics that reflect individual preferences, need for health care and the affordability of coverage to the individual. In addition, the availability of free sources of health care in the community–such as public hospitals and community health centers–may be a significant deterrent to enrolling in employer-sponsored health insurance coverage. In other words, some individuals may decline employer-sponsored health insurance if other sources of health care are available in the community, enabling them to achieve reasonably good access to care without health insurance. This would not only imply that the phenomenon of declining employer-sponsored coverage is a less serious concern than has been assumed, but it would also suggest that efforts to expand direct care services to the uninsured via the health care safety net may have the unintended consequence of increasing the number of uninsured (and those who depend on the public sector for health care rather than the private sector).

This study has two basic objectives. Based on a multivariate analysis that includes a broad range of factors, we first identify the factors that determine whether individuals decline employer-sponsored coverage. In addition to those factors that reflect costs, individual preferences and the need for health care (i.e., health status), we explicitly examine whether the availability of free sources of care in the community (e.g., public hospitals, community health centers) affects the decision to enroll in employer-sponsored health insurance.

The second objective is to examine the consequences of declining health insurance regarding access to care and the use of health services. This is done by comparing the health care utilization and access of uninsured persons who decline coverage with other uninsured persons (i.e., those who do not have access to employer-sponsored) and with insured persons. Better access to care and higher health care use among those uninsured who decline employer-sponsored coverage (compared to the uninsured who are not offered coverage) would suggest a rational decision to opt out of coverage because these individuals are apparently able to obtain health care either through the free care system or by paying out of pocket. In this event, it is doubtful that policy interventions are necessary or would do any good. However, similar or worse access, compared with uninsured who are not offered insurance, would suggest that the decision to decline coverage has serious consequences regarding access to care. In this case, policies that provide incentives for individuals to enroll in coverage may be warranted and effective.

DATA AND METHODS

Data Source

The Community Tracking Study (CTS) is a major initiative of The Robert Wood Johnson Foundation to track changes in the health care system over time and to gain a better understanding of how health system changes are affecting people. A more detailed discussion of the design and scope of the study is provided elsewhere (Kemper et al. 1996; Metcalf et al. 1996). Data collection is focused on 60 randomly selected communities, or sites, nationwide. Sites were defined as counties or groups of counties based on Metropolitan Statistical Areas (for metropolitan sites) and Bureau of Economic Analysis Economic Areas (for nonmetropolitan sites). Nonmetropolitan sites include areas contiguous with MSAs and isolated sites clustered around economic centers too small to be designated as MSAs. The 60 sites were randomly selected with probability in proportion to population to ensure representation of the U.S. population. Sites were stratified by region and size to ensure diversity in these areas.

Households were randomly selected within each of the 60 CTS study sites. While random-digit dialing was the primary sampling method, a small field sample was also included to represent households with no telephones or with intermittent telephone service. Information was obtained about all adults in the household and one randomly selected child within each family in the household (all families within a household were interviewed separately). Interviews were conducted in Spanish for family respondents who were not fluent in English. For more detail on survey methods and procedures, see Strouse et al. (1998).

The final sample for the 60 sites include 30,787 families and 54,371 individuals. The overall response rate was 65 percent for families. The potential for bias in estimates resulting from survey nonresponse cannot be assessed directly since no information was collected on families that refused to participate in the survey. Person-level weights used for making population estimates were post-stratified to correct for any differences in nonresponse based on age, gender, race/ethnicity and education.

All estimates are weighted to be representative of the civilian noninstitutionalized population of the continental U.S. and for each of the 60 communities. Standard errors used in tests of statistical significance were computed using the SUDAAN software and take into account the complex survey design, including the clustering of the sample in the 60 sites, the inclusion of multiple families within a household, sampling multiple adults within families and the random selection of one child. (Shah et al. 1996).

Methodology for Examining the Determinants of Declining Coverage

The first part of this analysis involves a logistic regression model to identify the factors associated with an individual declining employer-sponsored insurance and ending up as uninsured. The sample for this analysis includes all employed persons between 18 and 64 years of age who are offered and eligible for health insurance coverage at their workplace (ascertained during the interview) (n=19,324).

Workers who are offered and eligible for coverage essentially have three options: (1) They can accept the coverage offered by their employers; (2) they can decline coverage in favor of other private or public coverage (e.g., accept coverage offered through a spouse’s job, purchase a nongroup policy or enroll in Medicaid or other public coverage if eligible); or (3) they can decline coverage and end up as uninsured. For this analysis, we are primarily interested in examining the factors that influence the decision to decline coverage specifically when the result of that decision is to be uninsured (i.e., the third group). Thus, the dependent variable is coded as 1 if a worker declined coverage and was uninsured (the third group), and is coded 0 for all other workers who were offered and eligible for coverage (the first and second groups). While there may be differences between workers who accept coverage offered by employers (the first group) and workers who decline coverage in favor of various other types of insurance (the second group), modeling the full array of health insurance options (both employer-sponsored and others) is much more complex and is not central to the objectives of this study, which is to understand why individuals opt out of employer-sponsored coverage in favor of being uninsured.

Independent variables were selected that correspond to three general categories: (1) factors associated with the relative cost to the individual of employer-sponsored coverage; (2) variables that reflect individuals’ need and preferences for being insured; and (3) availability of free sources of health care in the community.

Unfortunately, there are no direct measures of the cost of employer-sponsored coverage in the survey, such as the amount of the employees’ share of premiums for employer-sponsored health plans. However, the survey includes a number of variables that directly or indirectly reflect the relative cost to the individual of employer-sponsored coverage. Chief among these is the family income of the worker. Because the relative cost of coverage to the individual is much higher for low-income workers, we expect that they will be more likely to decline coverage when offered. The cost of insurance to workers is also higher among certain types of employers, either because there is a smaller group of workers with whom to spread medical risk (e.g., in the case of small firms), or because firms with certain characteristics have traditionally been less generous with respect to health benefits (and therefore have higher employee cost-sharing). Small employers, firms that employ primarily low-wage workers and industries like agriculture, forestry, retail sales and certain services are less likely to offer any health benefits (Cooper and Schone 1997; Long and Marquis 1993; Cantor et al. 1995; Swartz et al. 1993; Gabel et al. 1999), and hence may be less generous when they do offer health insurance. In fact, employee premium contributions have been shown to be higher in small firms and firms that employ primarily low-wage workers (Gabel et al. 1997; Gabel et al. 1999).

Having a choice of health plans may also indirectly reflect differences in the cost of health insurance to the worker, since having a choice of more than one plan increases the likelihood that at least one plan will be attractive to the employee, whether it has lower premium costs or offers benefits that suit their needs. Firms with multiple plan offerings may also be inherently more generous with respect to employee health benefits than firms that offer only a single plan, which may also influence workers to accept coverage. While there is only limited information in the survey on characteristics of plans offered by the employer, we distinguish among workers who are offered (1) a choice of HMO and non-HMO options, (2) a choice of multiple HMO options, (3) a choice of multiple non-HMO options, (4) a choice of one HMO plan and (5) a choice of one non-HMO plan.

Economic considerations in deciding whether to enroll in employer-sponsored coverage are also important to the extent that free sources of health care are available to uninsured persons in the community. These are typically referred to as "safety net providers" and include public and teaching hospitals, community health centers and other free clinics and hospital emergency rooms. The number and type of safety net providers varies considerably across communities (Baxter and Feldman 1999; Lipson and Naierman 1996), and it is expected that individuals would be more likely to decline employer-sponsored coverage in communities where safety net providers are more extensive. In this study, data from the American Hospital Association annual survey are used to measure the supply of public and teaching hospitals in the county of residence (measured as the number of beds), as well as the number of hospital emergency departments (EDs), while data from the Bureau of Primary Health Care are used to measure the number of physicians practicing at federally funded community health centers in the county. These measures are standardized relative to the number of low- income persons in the county.

Aside from strictly economic considerations, the decision to forgo coverage also depends on how salient health care and health insurance coverage is to the individual and his or her family, and how risk-averse the individual is with respect to having to incur large health care expenses because of a sudden illness or accident. We identify a number of factors that reflect individuals’ need and preferences for having insurance coverage. These include a measure of the extent to which the individual perceives himself/herself as being more of a risk-taker than the average person. A measure of perceived general health is included to reflect the need for health insurance (because of higher than average health care use). We include a measure to reflect both the health status of the individual and a measure that indicates whether any other member of the individual’s immediate family (i.e., those who could be covered as a dependent) is in fair or poor health. Need and preferences for health insurance may also be related to other characteristics of the individual, including age, gender, educational attainment, race/ethnicity and family composition (i.e., marital status and whether there are children in the family).

Controls for the metropolitan status of the area and the nine census divisions are also included in the model.

Methodology for Examining the Consequences of Declining Coverage

The second objective of this study is to compare the health care utilization and access of uninsured individuals who decline employer-sponsored coverage with other uninsured individuals (i.e., those who do not have access to employer-sponsored coverage) and insured persons. Because the potential consequences of declining employer-sponsored coverage also apply to those dependents (i.e., spouses, children) of workers who decline coverage, we compare all uninsured persons who have access to employer-sponsored coverage (defined below) with uninsured persons who do not have access to employer-sponsored coverage. The sample for this analysis includes all persons under the age of 65 (n=53,270).

The key independent variable in this analysis–insurance status–is defined for each individual using the following categories: (1) any form of private insurance (employer-sponsored or purchased on the individual market) or military insurance (CHAMPUS); (2) any form of public coverage, including Medicaid, Medicare or other state program; (3) uninsured with access to employer-sponsored coverage; and (4) uninsured with no access to employer-sponsored coverage. Uninsured persons are considered to have access to employer-sponsored coverage if they are offered and eligible for coverage through their own job, or they are a dependent (i.e., spouse, child) of someone in the family who is offered and eligible for employer-sponsored coverage.

Differences between the four insurance groups on commonly used measures of access to care and utilization are examined. These include whether or not the individual has a usual source of care, which is considered important for facilitating entry into the health care system (Starfield 1992). We also use direct measures of whether the individual reported some difficulty in obtaining needed medical care in the previous year, including whether they were not able to get needed care (unmet need), whether they put off or delayed getting medical care, and whether they reported any difficulty (unmet need or delayed care-seeking) specifically due to the cost of care. Ambulatory health care use (from all sources) is also examined, including the likelihood of having an ambulatory care visit, the average number of ambulatory care visits (for persons with any use), and the proportion of ambulatory care visits made to hospital emergency rooms. Relatively greater use of the hospital ED would suggest less access to more appropriate sources of primary care, such as a physician’s office or clinic. Differences in the likelihood of having an inpatient hospital stay in the past year are also examined. Consistent with other research on these measures, we expect the findings to reflect lower access and health care use for uninsured persons as a group (Cunningham and Kemper 1998; Donelan et al. 1996; Berk et al. 1995; Cunningham and Whitmore 1998; Krauss et al. 1999).

Because health care utilization and access are strongly related to other characteristics of individuals, estimates of access and use for each of the insurance groups were computed while adjusting for other individual factors. These include age, gender, health status, race/ethnicity, family income and education. These adjustments were made by estimating regression equations in which the access and use variables were included as dependent variables, and the insurance groups and other individual characteristics were included as independent variables. Logistic regressions were used for dichotomous dependent variables, while ordinary least squares analysis was used for the average number of ambulatory visits and the percentage of visits to a hospital ED.

Predicted values for each insurance group were computed while holding constant the values of all other independent variables equal to their population mean. For the logistic regressions, the predicted values were then transformed back into probabilities to reflect percentages.

RESULTS

Number and Percentage Who Decline Employer-Sponsored Coverage

Overall, 80.3 percent of workers who are offered coverage through their employer take up that coverage, which is consistent with other studies (Table 1). Most workers who decline coverage do so in favor of other health insurance, including private insurance offered through a spouse’s employer, private insurance purchased in the individual market or public coverage. Only 4 percent of workers (3.2 million) who are offered coverage decline that coverage and end up as uninsured.

However, when uninsured dependents of these workers are also included (i.e., spouses or children who could be covered through a family policy), findings show that 7.3 uninsured persons have access to employer-sponsored insurance, either through their own job or through a spouse’s or parent’s job. These individuals represent about one-fifth of all uninsured persons, according to the CTS Household Survey data.

Table 1. Number and percentage of workers who decline employer-sponsored coverage.

  Number Percent

All workers (age 18-64) offered and eligible for coverage

79,557,000

100.0

Accept coverage

63,877,000

80.3

Decline coverage in favor of

other coverage1

12,513,000

15.7

Decline coverage and are

Uninsured

3,167,000

4.0

1Includes employer-sponsored coverage obtained through a spouse’s employer, direct purchase of nongroup private insurance, Medicaid, Medicare, CHAMPUS and other public coverage.

Factors Associated with the Decision to Decline Coverage (in Favor of Being Uninsured)

Table 2 shows the results of the logistic regression analysis for the likelihood of declining employer-sponsored coverage in favor of being uninsured. For this analysis, we focus on the sample of workers from Table 1 who are offered and eligible for coverage through their employer. As expected, factors related to the relative cost of employer-sponsored coverage to the individual appear to be highly salient in the decision to accept or decline coverage. The likelihood of declining coverage (and being uninsured) is much greater for poor and low-income workers, and decreases sharply as the family income of the worker increases. The likelihood of declining coverage is also much higher for low-wage workers (independent of income), which may reflect either an additional income effect, or the fact that health benefits are typically less generous in firms that employ primarily low-wage workers (Gabel et al. 1999). Workers in traditionally "high uninsurance" industries (e.g., retail sales, some services, agriculture, construction) were more than twice as likely to decline coverage as workers in other industries, which also suggests that these industry types offer less generous health benefits.

The number and types of employer offerings have substantial effects on the decision to enroll in coverage. Compared with persons who have multiple plan offerings that include at least one HMO plan, those offered only one plan or a choice of only non-HMO plans are two to three times more likely to decline coverage. Thus, the key aspect of choice appears to be multiple plan offerings that involve at least one HMO plan, which may reflect the fact that HMO plans have generally lower premiums and are frequently offered by employers to decrease their overall health benefit costs and to give employees a lower-cost alternative. Alternatively, employers that offer a diverse set of health plans may be inherently more generous with respect to employee health benefits than firms that offer only a single plan.

The effects of firm size on the decision to decline were not statistically significant, despite the fact that descriptive results from the CTS data and other studies do show that workers in smaller firms are more likely to decline coverage (Cooper and Schone 1997). Further analysis revealed that hourly wage, industry and the number and types of plans offered by the employer appear to account for the association between firm size and the decision to decline coverage. When these variables were excluded from the model, the odds ratio for firm size decreased from the .92 shown in Table 2 to .75 (an 18 percent decrease) and was highly significant at the .01 level. Thus, workers in smaller firms are more likely to decline coverage, not because of the size of the firm per se, but because they are more likely to be low-wage employees, offered less choice of plans and employed in typically "high uninsurance" industries.

In general, availability of safety net providers in the community appears to have little effect on the decision to decline coverage. Persons living in counties with a relatively large public hospital capacity are significantly more likely to decline coverage, although a one standard deviation increase in public hospital capacity is associated with only a 1.1 percent increase in the likelihood of declining coverage. Even this small effect appears to be offset by the availability of teaching hospitals and hospital EDs (i.e., a higher supply of these providers is associated with a lower likelihood of declining coverage), although these effects were not statistically significant.

Table 2. Logistic regression analysis of the likelihood of declining employer-sponsored private insurance in favor of being uninsured.

  Odds ratios 95% Confidence Intervals

Intercept

0.30**

0.12 — 0.74

Factors that reflect affordability to individuals

   

Family income

LT 100% of poverty

100-149% of poverty

150-199% of poverty

200-299% of poverty

300-399% of poverty

400% and above

1.00

0.56

0.59

0.38

0.30

0.23

 

0.35 — 0.90

0.40 — 0.88

0.25 — 0.60

0.18 — 0.49

0.16 — 0.34

Hourly wage (+1 S.D.)

0.34**

0.21 — 0.44

Firm size (+1 S.D.)

0.92

0.85 — 1.00

High uninsurance industry

2.35**

2.00 — 2.77

Employer offerings (number and type)

Multiple plans, both HMO and non-HMO

Multiple plans, HMO only

Multiple plans, non-HMO only

Single plan—HMO

Single plan—non-HMO

1.00

1.04

2.01**

2.35**

3.20**

 

0.63 — 1.72

1.41 — 2.85

1.78 — 3.09

2.43 — 4.22

     

Availability of free sources of health care

   

Number of public hospital beds in county (+1 S.D.)

1.10*

1.00 -- 1.20

Number of teaching hospital beds in county (+1 S.D.)

0.91

0.80 -- 1.05

Number of CHC physicians in county (+1 S.D.)

1.00

0.97 — 1.03

Number of hospital EDs in county (+1 S.D.)

0.94

0.86 — 1.02

     

Factors related to individual need and preferences

   

Risk-averseness

Not a risk-taker

Somewhat of a risk-taker

Strong risk-taker

Unknown


1.00

1.04

1.34*

0.85

 

0.79 — 1.37

1.05 — 1.72

0.36 — 2.02

Age

19-24

25-34

35-44

45-54

55 and over


1.00

0.85

0.60**

0.40**

0.23**

 

0.64 — 1.12

0.44 — 0.82

0.27 — 0.60

0.14 — 0.36

Race/ethnicity

White (reference group)

Black

Hispanic

Other


1.00

1.59**

1.68*

0.59

 

1.21 — 2.10

1.12 — 2.52

0.35 — 1.00

Gender (1=male)

1.14

0.89 — 1.45

 

Health status of individual

Excellent or very good

Good

Fair

Poor

 

 

0.71

0.72

0.79

1.00

 

 

0.34 — 1.51

0.34 — 1.49

0.31 — 1.98

Other family member in fair or poor health

1.14

0.83 — 1.57

Education

LT 9th grade

9-11th grade

Completed high school

13-15th grade

16th and over


2.26**

1.48*

1.00

0.67**

0.58**

1.08 — 4.72

1.02 — 2.14

0.54 — 0.84

0.42 — 0.79

Family composition

Single

Married, no kids

Single with kids

Married with kids


1.00

0.53**

0.74

0.40**

 

0.37 — 0.75

0.53 — 1.05

0.29 — 0.56

     

Geographic factors

   

Place of residence

Large MSA (> 200,000 persons)

Small MSA

Non-MSA


1.19

0.88

1.00

0.90 — 1.57

0.54 — 1.44

Census Division

New England

Middle Atlantic

East North Central

West North Central

South Atlantic

East South Central

West South Central

Mountain

Pacific


0.80

1.24

1.00

0.60

1.19

0.86

1.63**

1.12

1.04


0.51 — 1.24

0.89 — 1.75

0.29 — 1.26

0.76 — 1.86

0.63 — 1.17

1.20 — 2.20

0.78 — 1.59

0.76 — 1.44

     

* p < .05

** p < .01

Sample includes all employed persons (ages 18-64) who are offered and eligible for employer-sponsored private insurance (n=19,324).

Factors related to individual need and preferences for being insured appear to be significantly related to the decision to decline coverage. While one would certainly expect poor health of the worker or a family member to strongly decrease the likelihood of declining coverage (because of greater need for health care and financial protection from high health expenditures), the results actually show that persons in the best health were least likely to decline coverage in favor of being uninsured, although the results for health status were not statistically significant. It is unclear why the direction of the effect of health status (although not statistically significant) was the opposite of what was expected. It is possible that poor health in this analysis is picking up some unmeasured aspect of wealth, such as high instability of income or the cumulative effects of economic deprivation over a long period of time due to disability or chronic illness. Nevertheless, the notion that those who opt of out health insurance do so because they have very low need for care is not supported by this analysis.

As one would expect, individuals who considered themselves strong risk-takers were 1.3 times more likely to decline coverage than persons who did not consider themselves to be risk-takers. The likelihood of declining coverage decreases monotonically with age, with young adults (age 18-24) being the most likely to decline coverage among all age groups. The much higher likelihood of declining coverage for young adults may reflect both lower expected health care use (across all levels of health status) and having fewer financial assets to protect.

Even after controlling for factors related to employment and socioeconomic status, blacks and Hispanics were more than 1.5 times as likely to decline coverage as whites. Persons with lower educational attainment were more likely to decline coverage than more highly educated persons. Differences in the likelihood of declining coverage by race/ethnicity and education–controlling for other socioeconomic and employer characteristics–may reflect differences in the relative importance of health benefits among these groups, although it is difficult to determine the specific reasons for these different preferences. It is also important to point out that these preferences may reflect some selection into the types of jobs for which health benefits tend to be more or less generous. For example, employers with a large number of highly educated workers (e.g., a college or university) may respond to employee preferences by offering more generous benefits, thereby increasing the likelihood that these benefits will be accepted.

Family composition also plays a role in the decision to decline coverage, as married families with children are the least likely to decline coverage in favor of being uninsured and single individuals are the most likely to decline coverage. These findings suggest that health benefits are perceived as more salient to married families and families with children. However, the marital status effect appears to be stronger than the effect of having children in the family, given that the odds ratio for single-parent families was not as strong in magnitude as it was for married families and was not statistically significant. It is possible that the somewhat higher rate of refusal among single parents (compared with married families) is due to having fewer options (i.e., they do not have spouses who are also offered employer-sponsored coverage), or that single mothers tend to be employed in jobs that are not typically generous with respect to health benefits.

There were virtually no significant differences by size of MSA or Census Division in the decision to decline coverage. Individuals in the East South Central region were 1.6 times more likely to decline coverage in favor of being uninsured than persons in the East North Central region.

Consequences of Being Uninsured When Coverage is Declined

Table 3 shows the adjusted means for health care utilization and access by insurance status. As expected, uninsured persons in general (compared with both private and publicly insured) are less likely to have a usual source of care, more likely to have difficulty getting health care, have generally lower utilization of all types of services and have a higher proportion of ambulatory care visits in hospital emergency departments.

However, the findings provide little evidence that uninsured persons with access to employer-sponsored coverage differ fundamentally from other uninsured in their access to and use of health care services, and that, in fact, the decision to decline coverage appears to have consequences for their ability (and the ability of other family members) to get medical care. This is seen most explicitly by the fact that uninsured persons with access to employer-sponsored coverage are just as likely to have difficulty getting medical care (either not getting or delaying care) as other uninsured persons, and have much greater difficulty than insured persons. There are no significant differences in any type of health care use between uninsured adults with access to employer-sponsored coverage and other uninsured adults.

The same is largely true for uninsured children with access to employer-sponsored coverage, although their proportion of ambulatory visits at hospital emergency rooms is similar to privately insured and publicly insured children, and significantly less than children with no access to employer-sponsored private insurance. This suggests that children with access to employer-sponsored private insurance are better able to obtain care in appropriate primary care settings than other uninsured children, possibly because children are often the focus of direct care services provided through publicly sponsored community and school-based clinics. However, while this may influence to some extent the parent’s decision to decline insurance, overall service use and access of these children is still significantly below that of insured children.

Table 3. Differences in access to care and health care use for the nonelderly, by insurance status.

   

Difficulty getting care in past year

Ambulatory care use in past year2

 
 

% with no usual source of care

% with unmet need

% who delayed care

% with any difficulty due to cost1

% with any use

Average number of visits3

% of visits at hospital ED3

% with any inpatient stay

Adults (18—64)

               

Private coverage

10.1*

4.4*

17.4*

7.7*

83.1*

5.2*

8.6*

8.4*

Public coverage

11.1*

4.9*

15.3*

6.5*

86.3*

7.6*

10.3*

10.6*

Uninsured—no access to ESPI

28.0

13.2

34.6

31.0

64.2

4.1

17.1

6.7

Uninsured—has access to ESPI

27.3

13.3

33.7

32.0

64.3

3.7

15.9

6.5

                 

Children (LT 18)

               

Private coverage

3.2*

1.9*

4.0*

2.2*

87.1*

4.6*

8.9

4.7

Public coverage

1.8*

2.5*

5.3*

1.7*

92.1*

5.1*

9.9

6.3

Uninsured—no access to ESPI

13.5

8.6

13.1

12.4

71.7

3.1

12.8*

3.2

Uninsured—has access to ESPI

11.0

7.5

14.2

13.3

74.0

3.8

8.6

2.5

*Difference with "uninsured–has access to ESPI" is statistically significant at .05 level.

Access to ESPI— Employment-sponsored private insurance is offered by the person’s employer or the spouse’s employer (for adults), or through a parent’s employer (for children).

1Includes persons who reported that they were not able to get needed services in the past year, or they had to put off getting services in the past year due to concerns about the cost of care.

2Includes visits to physicians and nonphysicians at all sites, including physician offices, hospital outpatient clinics and EDs, clinics and urgent care centers.

3Includes only persons with one or more ambulatory visits.

All estimates were adjusted to control for differences across insurance groups on the following factors: age, gender, health status, race/ethnicity, family income and education. The adjusted means and percentages were computed using multiple regression analysis.

CONCLUSION
To what extent should policy makers be concerned about individuals who decline employer-sponsored coverage? On the one hand, very few with access to employer-sponsored coverage (through their own job or a spouse’s) decline this coverage, and most who opt out do so in favor of other private or public coverage. While previous research indicates that the percentage of persons declining employer-offered coverage is increasing (Cooper and Schone 1997), it is unclear whether the growing number of those who decline coverage are able to find other health insurance or they become uninsured.

On the other hand, those uninsured with access to employer-sponsored coverage (more than 7 million) comprise one-fifth of the total number of uninsured. Thus, 100 percent acceptance of employer-sponsored coverage would do more to decrease the number of uninsured persons than the incremental insurance expansions that have been passed in recent years or are under consideration, such as the Health Insurance Portability and Accountability Act, the State Children’s Health Insurance Program and the proposed Medicare buy-ins for the near-elderly.

One could also argue that declining employer-sponsored coverage–even if the result is to be uninsured–is not a problem when it represents a rational choice. That is, persons decline coverage either because they don’t need health care or because they are able to obtain needed care without insurance. However, the findings from this study strongly refute this. First, persons in the best health (i.e., those with presumably the fewest health care needs) were actually less likely to decline coverage than those in the poorest health, although this finding was not statistically significant. Second, uninsured persons with access to employer-sponsored coverage reported as much difficulty in obtaining needed care as other uninsured persons, and both groups of uninsured have much greater difficulty getting care than privately or publicly insured persons. For the most part, levels of service use were also comparable for both groups of uninsured, suggesting neither greater access nor lower demand among those uninsured with access to employer-sponsored insurance. Thus, the decision to decline employer-sponsored coverage in favor of being uninsured appears to have the same negative consequences regarding access to care as it does for all other uninsured persons.

The findings also provided very little evidence that a large safety net capacity in an area induces individuals to decline employer-sponsored coverage, presumably because free sources of health care mitigate the need for insurance. While much concern has been expressed about the potential for private insurance "crowd-out" stemming from expansions in Medicaid and the State Children’s Health Insurance Program, similar concerns about "safety net crowd-out" of private insurance are not warranted by these findings. In fact, safety net providers often serve as the focal point of outreach efforts in communities to enroll low-income persons in health insurance coverage for which they are eligible, although much of this effort is directed at public programs such as Medicaid. Given their tenuous dependence on public funds and the financial pressures many are currently experiencing in providing uncompensated care, safety net providers have a vested interest in getting their patients enrolled in private insurance plans that would give providers important new sources of revenue.

Thus, declining employer-sponsored coverage is an important matter for policy makers to consider. But given the already high rates of acceptance of employer-sponsored coverage, is it reasonable to expect that 100 percent enrollment could be attained? Without additional financial incentives and/or mandates, probably not. As indicated by the findings in this study, most of the strongest determinants of whether coverage was declined reflected the cost to the individual or family, and low-income and low-wage workers were far more likely to decline coverage than other workers. While policy makers are currently considering the use of tax credits to provide financial incentives for uninsured persons to purchase coverage, it is notable that many low-income persons decline employer-sponsored coverage despite the already significant financial incentives of employer-subsidized premiums, tax exclusions on the employer share of the premium and group rates. Nevertheless, refundable tax credits would provide an additional incentive to low-income persons to enroll in employer-sponsored insurance, since they currently pay little or no tax and, therefore, they currently benefit little from the exclusions from taxes of employer-paid premiums.

One important limitation of this study is that we were unable to directly assess the effects of premium costs or plan benefits on the decision to decline coverage, although the fact that workers were three times more likely to decline coverage when offered only a single non-HMO plan (compared with workers offered a choice of HMO and non-HMO plans) suggests that the mix of benefits and level of employee cost-sharing is relevant. Future research should examine this more explicitly, since refusing plans with high deductibles and limited benefits may be understandable for some workers, especially low-income and young adult workers who have few assets to protect and limited income with which to pay the potentially substantial out-of-pocket costs. From a policy perspective, it is unclear how much better off these workers would be (in terms of access and health care use) by accepting such plans, and providing financial incentives to workers to take up these types of plans may not be particularly meaningful if workers perceive them to be of limited value.

Nonetheless, individual differences in how much value and importance is placed on health insurance–regardless of the benefits package–also plays a role, and there will always be some who opt out as long as health insurance is strictly voluntary. Even individual assessments of whether they can afford health insurance depend in part on how willing and able they are to forgo other goods and services that could be purchased with the money used to pay premiums. But while individuals differ in this economic calculation–which is especially difficult for low-income persons–those who opt out still face the same problems that all other uninsured persons have, which is greatly diminished access to health care services.

REFERENCES

Baxter, R.J., and R.L. Feldman. 1999. Staying in the Game: Health System Change Challenges Care for the Poor. Research Report (March). Center for Studying Health System Change.

Berk M.L., C.L. Schur and J.C. Cantor. 1995. "Ability to Obtain Health Care: Recent Estimates from The Robert Wood Johnson Foundation National Access to Care Survey." Health Affairs 14: 139-146.

Cantor, J.C., S.H. Long and M.S. Marquis. 1995. "Private Employment-Based Health Insurance in Ten States." Health Affairs 14 (Summer): 199-209.

Cooper, P.F., and B.S. Schone. 1997. "More Offers, Fewer Takers for Employment-Based Health Insurance: 1987 and 1996." Health Affairs 16 (November/December): 142-149.

Cunningham, P.J., and P. Kemper. 1998. "Ability to Obtain Medical Care for the Uninsured: How Much Does it Vary Across Communities?" JAMA 280 (10): 921-927.

Cunningham, P.J., and H. Whitmore. 1998. How Well Do Communities Perform on Access to Care for the Uninsured? Center for Studying Health System Change. Research Report #1 (September).

Donelan K., R.J. Blendon, C.A.Hill, C. Hoffman, D. Rowland, M. Frankel and D. Altman. 1996. "Whatever Happened to the Health Insurance Crisis in the United States?" JAMA. 276:1346-350.

Gabel, J., P.B. Ginsburg and K.A. Hunt. 1997. "Small Employers and Their Health Benefits, 1988-1996: An Awkward Adolescence." Health Affairs 16 (September/October): 103-110.

Gabel, J. K. K. Hurst, H. Whitmore and C. Hoffman. 1999. "Class and Benefits at the Workplace." Health Affairs 18 (May/June): 144-150.

Kemper P., D. Blumenthal, J.M. Corrigan, S.M. Felt, J. Grossman , L.T. Kohn, C.E. Metcalf, R.F. St. Peter, R.C. Strouse and P.B. Ginsburg. 1996. "The Design of the Community Tracking Study." Inquiry 33: 195-206.

Krauss, N.A., S. Machlin and B.L. Kass. 1999. "Use of Health Care Services, 1996." MEPS Research Findings No. 7. Agency for Health Care Policy and Research, Rockville MD. AHCPR Pub. No. 99-0018.

Lipson, J.D., and N. Naierman. 1996. "Effects of Health System Changes on Safety Net Providers." Health Affairs (15): 33-48.

Long, S., and M.S. Marquis. 1993. "Gaps in Employer Coverage: Lack of Supply or Lack of Demand?" 12 (Supplement): 282-293.

Metcalf C.E., P. Kemper, L.T. Kohn and J.D. Pickreign. 1996. Site Definition and Sample Design for the Community Tracking Study. Technical Publication No. 1. Center for Studying Health System Change.

Shah B.V., B.G. Barnwell and F.S. Bieler. 1996. SUDAAN User’s Manual, Release 7.0. Research Triangle Park, NC: Research Triangle Institute.

Starfield B. 1992. Primary Care: Concept, Evaluation, and Policy. New York: Oxford University Press.

Strouse, R., J. Hall, F. Potter, B.L. Carlson, P. Cunningham, J. Pascale and R. Stapulonis. 1998. Report on Survey Methods for the Community Tracking Study’s 1996-1997 Round One Household Survey. Center for Studying Health System Change. Technical Publication #15 (November).

Swartz, K., J. Marcotte and T. McBride. 1993. "Personal Characteristics and Spells Without Health Insurance." Inquiry 30: 64-76.

Thorpe, K.E., and C.S. Florence. 1999. "Why Are Workers Uninsured? Employer-Sponsored Health Insurance in 1997." Health Affairs 18 (March/April): 213-218.

The author would like to thank Paul Ginsburg, Peter Kemper, Chris Hogan, and Sally Trude for reviewing an earlier version of the manuscript and providing helpful comments. Beny Wu of Social and Scientific Systems, Bethesda, Md., provided excellent programming assistance.