Skip to page content
Employee Benefits Security Administration

Health Insurance Coverage of the Unemployed

Table of Contents

  • Abstract

  • Introduction

  • Overview

  • Health Insurance Coverage

  • COBRA Eligibility

  • Measurement of COBRA Eligibility Using the April 1993 CPS

  • COBRA Eligibility and Health Insurance Coverage

  • Eligibility for Spouse Employer Coverage and COBRA

  • Statistical Models of Health Insurance Coverage and COBRA Eligibility: Theoretical Considerations

  • Binary Logit Analysis

  • Multinomial Logit Analysis

  • Male-Female Differences Among the Unemployed

  • Conclusions

  • References

  • Appendix

  • Endnotes

  • End Credits

Abstract

We use the 1993 April CPS to examine how health insurance coverage rates of the unemployed differ by age, gender, marital status, education, number of children, and length of time unemployed, and other characteristics such as income and disability of the spouse. At the same time, we investigate the effects of COBRA eligibility on health insurance coverage. We estimate binary logit and multinomial logit models of the health insurance outcomes of the unemployed as functions of COBRA and spouse's employer insurance eligibility and worker characteristics. We find that after controlling for these characteristics, COBRA eligibility increases the probability of health insurance coverage among the unemployed by .075. Unemployed women have higher health insurance coverage rates than unemployed men. However, women are less likely to elect COBRA coverage than men, and therefore have lower rates of coverage from their former employers than men.

Introduction

When health insurance is provided through the place of employment, interruptions in the employment relationship disrupt health insurance coverage. The Consolidated Omnibus Reconciliation Act (COBRA), enacted in 1986, contained provisions designed to partly remedy this problem. Most employees are able to purchase health insurance from their former employer for up to 18 months after their employment ends, at a premium not to exceed 102% of the group rate. If workers are reluctant to change jobs because of health insurance considerations, COBRA can improve the efficiency of the labor market. Perhaps more important, COBRA potentially increases the health insurance coverage of the unemployed.

While not universally elected, COBRA does appear to increase the health insurance coverage of the unemployed. Using CobraServ data for 1990-91, Flynn (1992) reports that 21% of workers who qualified, elected COBRA continuation coverage. Others attempt to provide a more precise estimate of the effect of COBRA on the health coverage of the unemployed by holding a number of demographic characteristics and other factors constant. Using data from the Survey of Income and Program Participation (SIPP) and holding constant age, education, and months since job loss, Klerman and Rahman (1992) find evidence of a positive effect of COBRA legislation on the health insurance coverage of the non-employed.

In an important recent study, Gruber and Madrian (1995a) examine health insurance coverage among the non-employed, using longitudinal data from the SIPP for 1983 to 1989 for men aged 25-54. They find that the likelihood of having health insurance drops by approximately 20 percent after a worker is separated from his job. However, they find that state and federal health insurance continuation mandates such as COBRA increase the likelihood of coverage among the non-employed by 6.7 percent. They also find that the estimated effect of continuation mandates varies by the duration of the spell of unemployment. The effect of continuation mandates is insignificant for those with completed durations of one year or less. However, the effects are substantially larger for those with durations of more than one year, presumably the group with the greatest need. For instance, for those with unemployment durations of more than one year, a continuation mandate of one year increases the likelihood of insurance coverage by 9.4 percent (Gruber and Madrian, 1995a, p.23).

COBRA type legislation also appears to have the intended effect on labor market efficiency. Gruber and Madrian (1995a) find that health insurance continuation mandates increase turnover, and are associated with significant wage gains in subsequent jobs. Thus, these mandates appear to reduce job lock and to lead to more productive job search by individuals seeking new jobs. Finally, COBRA type mandates influence workers decisions when to retire. Using SIPP and the March Current Population Survey (CPS) data, Gruber and Madrian (1995b) and Karoly and Rogowski (1994) find that health insurance continuation laws increase retirement probabilities among older workers.

This previous work appears to show that, on average, COBRA legislation has the intended effects on health insurance coverage and labor market transitions. However, there are important differences in health insurance coverage across the unemployed population. These differences may in part be due to different responses to the availability of COBRA. The U.S. Department of Labor (1994, p. F-27) uses the April 1993 Current Population Survey (CPS) to provide some preliminary cross-tabular analysis of the health insurance coverage of unemployed workers by various characteristics. Older workers and higher income workers are more likely to have coverage through their former employer.

Perhaps the most interesting differences are by gender. According to the 1993 April CPS, the health insurance coverage rate of unemployed females exceeded that of unemployed males (48.5% vs. 36.9%). Among those covered, however, 36.2% of the males were covered by a former employer compared with only 18.4% of the females. It is important to disentangle the reasons behind this gender difference in health insurance coverage among the unemployed. In this study, we use the April 1993 CPS to examine the reasons for this gender difference in coverage rates by the former employer. Part of the difference could be due to coverage under spouse's plans. Other studies have not examined the effect of the availability of spouse coverage on COBRA election. Alternatively, there may be higher coverage rates for women from public sources such as Medicaid. Unemployed women may have disproportionately been in jobs before they were unemployed that did not have health insurance and thus not be eligible to take advantage of COBRA.

We also examine differences in health insurance coverage rates and use of COBRA by other characteristics such as age, income, and length of unemployment. In addition to cross-tabular analysis of coverage rates and COBRA eligibility among the unemployed, we estimate logit models of health insurance coverage in order to provide a more precise estimate of the effect of COBRA on coverage.

This project offers an improvement over the existing literature in several ways. First, the April 1993 CPS is the most current available data for studying health insurance of the unemployed. Second, previous studies have been limited in the sample used, the range of determinants of health insurance coverage examined, or in the health insurance outcomes studied. For example, Flynn (1992) is only able to examine COBRA coverage for ages 40-64 using the August 1988 CPS. The April 1993 CPS questions cover ages 25-64. Klerman and Rahman (1992) estimate a model of coverage for the unemployed that includes measures of length of unemployment, age, education, and COBRA eligibility. We examine these differences along with the effects of a much broader range of characteristics than those employed by Klerman and Rahman (1992). Gruber and Madrian (1995a) consider only males in their analysis of health insurance of the unemployed. However, as already mentioned, there are important gender differences in health insurance coverage among the unemployed. Finally, unlike previous studies, the use of the 1993 April CPS data allow us to consider multiple categories of health insurance coverage of the unemployed including coverage by a former employer, coverage through a spouse's employer, and other types of coverage.

Overview

Health Insurance Coverage - The April Current Population Survey (CPS) public use tapes contain the Employee Benefit Supplement as well as the April-March CPS matched records for rotations 2, 3, 6, and 7. Those individuals between the ages of 25 and 64 who were not employed but had previously been employed and were actively looking for work in the last 4 weeks (the experienced unemployed) were asked questions about their pension and health coverage. From the April 1993 CPS, we calculate that 41.62% of the experienced unemployed have some form of health insurance coverage (a_s84=1).(1)

The CPS also contains demographic and other characteristics of the unemployed. For example, the CPS contains information on a worker's gender, age, education, race, number of children, income, amount of time unemployed, and reason unemployed, in addition to many other characteristics.(2) In Table 1, we show the mean proportions of the unemployed that fall into various demographic and unemployment categories by health insurance status. The insured sample has a higher proportion who are female than does the sample without insurance. Similarly, the insured sample has a higher proportion who are over 50 years old, a higher proportion who are white, a higher proportion who have greater than a high school education, a higher proportion who have one or more children, a higher proportion who are married, and a higher proportion who have income more than $25,000 than does the uninsured sample. The differences by income and marital status appear particularly large.

Table 1 - Means by Health Insurance Status

Variable

Insured

Not Insured

Definition

female

.4760664

.3608687

  =1 if gender is female, =0 otherwise

over50

.2408537

.1588362

  =1 if age over 50, =0 otherwise

nonwhite

.1857875

.2814059

  =1 if race is black or other, =0 otherwise

gthseduc

.5131650

.3366575

  =1 if education > high school, =0 otherwise

children

.8807072

.7242192

  =1 if have 1 or more children, =0 otherwise

married

.7298426

.3911447

  =1 if married, =0 otherwise

lowinc

.3706291

.8162309

  =1 if income < 25K, =0 otherwise

look26wk

.2412858

.2584080

  =1 if unemployed > 26 weeks, =0 otherwise

quitjob

.0992253

.1143216

  =1 if quit job, =0 otherwise

Source: 1993 April Current Population Survey

COBRA Eligibility - In 1987, the Consolidated Omnibus Budget Reconciliation Act (COBRA) went into effect, which contained provisions allowing certain former employees, spouses and dependent children to buy temporary health insurance at group rates. "Employees" eligible for COBRA can be full-time or part-time workers, agents, independent contractors, directors, and certain self-employed individuals eligible to participate in a group plan. A qualified employee is anyone who was covered by a group health plan the day before a "qualifying event." Such events include voluntary or involuntary termination of employment for reasons other than gross misconduct, or a reduction in the number of hours worked. Spouses and dependent children qualify for coverage by demonstrating that either of these events were applicable to the covered employee (who was either their spouse or parent), because of death or divorce of the covered employee, or in the case of dependent children, if they lose dependent child status under the plan's rules. If a covered employee becomes eligible for Medicare benefits, their spouse and/or dependents qualify for COBRA coverage. After a qualifying event, beneficiaries have up to 60 days to elect COBRA coverage (U.S. Department of Labor, 1990, pp. 4-5, 9).

Employees can receive coverage for up to 18 months at rates of up to 102 percent of the cost of the plan to similarly situated individuals who have not incurred a qualifying event. This coverage can be extended for up to 11 more months if a qualified beneficiary is determined under Title II or XVI of the Social Security Act to have been disabled at the time of termination or reduction in hours. The cost for the additional 11 months of coverage can be increased to 150 percent of the plan's cost. Spouses and dependent children are also eligible for 18 months of coverage if the coverage employee terminates employment or suffers a reduction in hours. Spouses and dependent children can obtain up to 36 months of coverage if they become eligible through the death or divorce of the covered employee, or if the child loses his or her dependent status under the plan (U.S. Department of Labor, 1990, pp. 6-7,15).

Certain employers are exempt from providing COBRA benefits. The law generally covers group health plans of employers with 20 or more employees during the previous year. The law covers plans provided in the private sector and by state and local governments. The law does not apply to Federally sponsored health plans or the plans of certain church-related organizations (U.S. Department of Labor, 1990, p. 2)

Measurement of COBRA Eligibility Using the April 1993 CPS - While the April 1993 CPS does not directly ask about COBRA eligibility, the series of questions on health insurance coverage allow us to construct a measure of COBRA eligibility for the unemployed. The most direct measure uses questions a_s84, a_s85, and a_s88.(3) All of the experienced unemployed are asked this series of questions. COBRA eligible could then be estimated to be those who indicate that they have coverage from a previous employer (a_s84=1 and a_s85=1) or that they had coverage on their last job (a_s88=1). This potential measure indicates that 45.13% of workers are eligible for COBRA coverage.

This is an upper bound measure of the number of COBRA eligibles among the experienced unemployed for a number of reasons. First, some may have been unemployed for more than 18 months and thus no longer eligible for COBRA.(4) Firms with less than 20 employees in the last year are not subject to COBRA. COBRA also does not apply to plans sponsored by the federal government and certain church-related organizations. Therefore, we adjust the COBRA eligibility variable in the following way. Those individuals who are in an unemployment spell longer than 78 weeks (a_wkslk>78) are coded as COBRA ineligible (COBRA=0), as are those whose last job was in the federal government (a_clswkr=2), or in a religious organization (a_ind=880). The number of employees is not available as part of the April 1993 CPS for the unemployed (it is available for the currently employed). However, as part of the April-March CPS match, we have data for most of the sample on the number of workers of the employer for the longest job held in 1992. We code as COBRA ineligible those who worked for employers with less than 25 employees in the longest job held last year (noemp=1 or 2). This revised measure indicates that 33.59% of the unemployed are eligible for COBRA. We use this measure of COBRA eligibility for the remainder of the paper.

What are the characteristics of the COBRA eligible among the unemployed? In Table 2, we show the means of various characteristics among those eligible and not eligible for COBRA. A higher proportion of the eligible sample are female than are the sample of those not eligible for COBRA. Similarly, the eligible sample has a higher proportion who are over 50 years old, a higher proportion who have greater than a high school education, a higher proportion who have one or more children, a higher proportion who are married, a higher proportion who have annual income greater than $25,000, a higher proportion who have been unemployed less than 26 weeks, and a higher proportion who quit their job than the sample of those who are ineligible for COBRA. There are virtually no differences in the race makeup of the eligible and ineligible groups.

Table 2 -- Means by COBRA Eligibility

Variable

Eligible

Not Eligible

Definition

female

.4342093

.3959584

=1 if gender is female, =0 otherwise

over50

.2308315

.1738128

=1 if age over 50, =0 otherwise

nonwhite

.2407589

.2420465

=1 if race is black or other, =0 otherwise

gthseduc

.5024079

.3634186

=1 if education > high school, =0 otherwise

children

.8217148

.7729649

=1 if have 1 or more children, =0 otherwise

married

.5419932

.5270878

=1 if married, =0 otherwise

lowinc

.4885505

.7027527

=1 if income < 25K, =0 otherwise

look26wk

.2244352

.2648648

=1 if unemployed > 26 weeks, =0 otherwise

quitjob

.1388566

.0924484

=1 if quit job, =0 otherwise

Source: 1993 April Current Population Survey

COBRA Eligibility and Health Insurance Coverage - How many of the current unemployed take advantage of COBRA eligibility? We answer this question in Table 3 by showing various cross tabulations of COBRA eligibility and health insurance coverage. Table 3A shows that 50.35% of those who were eligible to elect COBRA at the beginning of their unemployment are currently insured. This is substantially higher than the 37.20% coverage rate among those not eligible for COBRA. In Table 3B we see that 40.64% of those with health insurance were eligible for COBRA compared to 28.57% of those without health insurance. In order to see how many of those eligible actually elected COBRA, Table 3C shows the breakdown by type of coverage into four categories: 1) former employer coverage (a_s84=1 & a_s85=1), 2) spouse's employer coverage (a_s84=1 & a_s85=2 & spouse a_s62=1), 3) other coverage (a_s84=1 & a_s85=2 & spouse a_s62!=1), and 4) no coverage (a_s84=2).

Table 3A - Health Insurance Status of Individuals Eligible and Not Eligible for COBRA

Health Insurance Coverage

COBRA Eligibility

No

Yes

Total

No

 62.80

 49.65

 58.38

Yes

 37.20

 50.35

 41.62

Total

100.00

100.00

100.00

Table 3B - COBRA Eligibility Status for Individuals with and without Health Insurance

Health Insurance Coverage

COBRA Eligibility

No

YES

Total

No

71.42

28.57

100.00

Yes

59.36

40.64

100.00

Total

66.41

33.59

100.00

Table 3C - Coverage Categories by COBRA Eligibility

Health Insurance Coverage

COBRA Eligibility

No

Yes

Total

Employer

  4.96

 24.57

 11.55

Spouse

 15.41

 12.59

 14.46

Other

 16.83

 13.18

 15.60

None

 62.80

 49.65

 58.38

Total

100.00

100.00

100.00

Among eligibles, 24.57% have coverage from a former employer, presumably through COBRA. Even this estimate is slightly higher than the 21% coverage election rate reported by Flynn (1992). However, we are not calculating a rate for a sample of new "qualifying events." Instead, we have a sample of the stock of unemployed at a point in time. From a policy point of view it may be the stock of the unemployed that is more relevant. It is not clear whether the a stock of unemployed persons at a point in time or a flow of new "qualifying events" would produce a higher coverage rate. The coverage rate could be higher in the flow sample because over time those in the stock sample may drop coverage due to the cost of maintaining coverage or as individuals begin to obtain spouse coverage. On the other hand, Gruber and Madrian (1995a) results suggest that the effect of continuation mandates are highest for those who are in longer spells of unemployment. In any event, it may be more important to know how much COBRA is being used among unemployed at a point in time than among the newly unemployed.

We next examine how COBRA eligibility and health insurance coverage varies across the unemployed. In Table 4, we show health insurance and COBRA eligibility cross tabulations by gender, age category, education category, number of children, marital status, race, income, time unemployed, and reason unemployed.

Coverage and COBRA Eligibility by Worker Characteristics

Table 4A - By Gender

-> female=

Male

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  6.67

 27.47

 13.35

Spouse

 12.06

  8.23

 10.83

Other

 13.31

 11.42

 12.70

None

 67.97

 52.89

 63.12

Total

100.00

100.00

100.00

->female=

Female

Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  2.35

 20.80

  8.94

Spouse

 20.53

 18.27

 19.72

Other

 22.20

 15.49

 19.80

None

 54.92

 45.44

 51.54

Total

100.00

100.00

100.00

Table 4B - By Age

->Over 50

<50 Yrs

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  4.07

 20.80

  9.43

Spouse

 15.48

 12.78

 14.61

Other

 15.71

 13.82

 15.11

None

 64.74

 52.60

 60.85

Total

100.00

100.00

100.00

-> over50=

50+ Yrs

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  9.17

 37.14

 20.41

Spouse

 15.12

 11.96

 13.85

Other

 22.13

 11.06

 17.68

None

 53.38

 39.84

 48.06

Total

100.00

100.00

100.00

Table 4C - By Race

-> nonwhite=

White

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  5.44

 26.79

 12.62

Spouse

 17.59

 13.67

 16.27

Other

 16.46

 14.46

 15.79

None

 60.51

 45.09

 55.32

Total

100.00

100.00

100.00

-> nonwhite=

Nonwhite

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  3.45

 17.59

  8.18

Spouse

  8.60

  9.19

  8.80

Other

 17.97

   9.16

 15.02

None

 69.98

 64.06

 68.00

Total

100.00

100.00

100.00

Table 4D - Education Level

->gthseduc=

<=HS

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  3.44

 21.69

  8.61

Spouse

 13.51

 10.50

 12.66

Other

 14.98

  8.28

 13.08

None

 68.07

 59.54

 65.65

Total

100.00

100.00

100.00

-> gthseduc=

>HS

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  7.62

 27.43

 15.77

Spouse

 18.75

 14.66

 17.07

Other

 20.06

 18.04

 19.23

None

 53.57

 39.87

 47.93

Total

100.00

100.00

100.00

Table 4E - By Number of Children

-> children=

No Children

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  6.68

 22.87

 11.28

Other

 11.67

 13.84

 12.29

None

 81.65

 63.29

 76.43

Total

100.00

100.00

100.00

-> children=

1+ Children

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  4.45

 24.94

 11.62

Spouse

 19.94

 15.32

 18.32

Other

 18.34

 13.04

 16.49

None

 57.26

 46.70

 53.57

Total

100.00

100.00

100.00

Table 4F - By Marital Status

-> married=

Not Married

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  3.32

 20.83

  9.08

Other

 16.02

 12.77

 14.95

None

 80.66

 66.40

 75.97

Total

100.00

100.00

100.00

-> married=

Married

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  6.43

 27.74

 13.72

Spouse

 29.24

 23.23

 27.18

Other

 17.56

 13.53

 16.18

None

 46.78

 35.50

 42.92

Total

100.00

100.00

100.00

Table 4G - By Income Level

-> lowinc=

>25K

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  9.76

 34.15

 21.11

Spouse

 44.38

 22.62

 34.25

Other

 18.85

 11.81

 15.57

None

 27.01

 31.41

 29.06

Total

100.00

100.00

100.00

-> lowinc=

<=25K

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  2.93

 14.55

  5.95

Spouse

  3.16

  2.09

  2.88

Other

 15.97

 14.62

 15.62

None

 77.94

 68.75

 75.55

Total

100.00

100.00

100.00

Table 4H - By Number of Weeks Unemployed

-> look26wk=

=26 Wks

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  4.10

 25.60

 11.59

Spouse

 15.52

 11.51

 14.13

Other

 17.56

 14.40

 16.46

None

 62.82

 48.48

 57.83

Total

100.00

100.00

100.00

-> lookwk

> 26 Wks

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  7.33

 21.01

 11.44

Spouse

 15.11

 16.31

 15.47

Other

 14.80

  8.97

 13.05

None

 62.76

 53.70

 60.04

Total

100.00

100.00

100.00

Table 4I - By Whether the Worker Quit Job

-> quitjob=

Other

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  5.46

 25.06

 11.82

Spouse

 15.89

 11.66

 14.52

Other

 16.96

 13.04

 15.69

None

 61.69

 50.24

 57.97

Total

100.00

100.00

100.00

-> quitjob=

Quit

Health Insurance Category

COBRA Eligibility

No

Yes

Total

Employer

  0.00

21.52

  9.29

Spouse

 10.73

 18.36

 14.03

Other

 15.54

 14.07

 14.90

None

73.74

46.05

61.78

Total

100.00

100.00

100.00

Source: 1993 April Current Population Survey

There exist several interesting differences in health insurance coverage and COBRA usage by worker characteristics. Males are less likely to be covered by health insurance, but are more likely to elect COBRA than are females. Females are more likely to be covered through the spouse or by other coverage than are males. Among those eligible for COBRA, workers over 50 years old, who are white, have more than a high school education, have one or more children, are married, or earn more than $25,000 are more likely to elect COBRA coverage. Whites, those with more than a high school education, and those with more than $25,000 income are more likely to have spouse and other forms of coverage as well.

Eligibility for Spouse Employer Coverage and COBRA - Clearly spouse coverage comes into play when individuals decide whether or not to exercise their COBRA rights. However, in Tables 5 and 6, we only observe health insurance outcomes, one of which is spouse coverage. What we would like to observe is eligibility for spouse coverage and not the outcome of spouse coverage. We construct an eligibility for spouse coverage variable (a_s61=1 or 3) and use it in addition to COBRA eligibility in our cross tabulations. We construct an eligibility variable with the categories: 1) COBRA eligibility, 2) eligible for coverage through spouse's employer, 3) eligible for both COBRA and spouse coverage, and 4) eligible for neither COBRA nor spouse coverage.

Table 5 shows eligibility for COBRA and spouse coverage for the entire sample of unemployed and for various demographic groups.

COBRA And Spouse Health Insurance Eligibility

Table 5A - Full Sample

COBRA and Spouse Insurance Eligibility

Percent

Cum.

COBRA

 25.45

 25.45

Spouse

 14.64

 40.09

Both

  8.15

 48.24

Neither

 51.76

100.00

Table 5B - By Gender

COBRA and Spouse Insurance Eligibility

Female

Male

Female

Total

COBRA

 24.68

 26.55

 25.45

Spouse

 13.20

 16.73

 14.64

Both

  7.47

   9.13

   8.15

Neither

 54.65

 47.59

 51.76

Total

100.00

100.00

100.00

Table 5C - By Age

COBRA and Spouse Insurance Eligibility

50 Years Old & Over

<50 Yrs

50+ Yrs

Total

COBRA

 24.07

 31.19

 25.45

Spouse

 15.25

 12.08

 14.64

Both

  7.95

  8.99

  8.15

Neither

 52.73

 47.74

 51.76

Total

100.00

100.00

100.00

Table 5D - By Race

COBRA and Spouse Insurance Eligibility

Nonwhite

White

Nonwhite

Total

COBRA

 24.58

 28.17

 25.45

Spouse

 17.00

  7.23

 14.64

Both

  9.05

  5.31

  8.15

Neither

 49.36

 59.30

 51.76

Total

100.00

100.00

100.00

Table 5E - By Education Level

COBRA and Spouse Insurance Eligibility

More than High School

<=HS

>HS

Total

COBRA

 21.97

 30.44

 25.45

Spouse

 13.59

 16.15

 14.64

Both

  6.37

 10.71

  8.15

Neither

 58.07

 42.70

 51.76

Total

100.00

100.00

100.00

Source: 1993 April Current Population Survey

Table 5F - By Income Level

COBRA and Spouse Insurance Eligibility

Low Income

>25K

<=25K

Total

COBRA

 27.01

 24.53

 25.45

Spouse

 31.42

  4.82

 14.64

Both

 19.52

  1.49

  8.15

Neither

 22.04

 69.16

 51.76

Total

100.00

100.00

100.00

A significant number of the unemployed are eligible for and take advantage of coverage through their spouse's employer. In Table 3C we saw that 14.46% of the unemployed opted for spouse's employer coverage. In Table 5A we find that 14.64% of the unemployed are eligible for spouse coverage (and not COBRA), and 8.15% are eligible for both COBRA and spouse coverage. Thus, over 60% (14.46/(14.64+8.15)) of those eligible for spouse coverage use it, compared to the 24.57% of those eligible for COBRA coverage who elect it. Coverage through one's spouse is clearly an important way for the unemployed to obtain health insurance.

Eligibility for spouse coverage varies across the unemployed. In Table 5, we see that females and those with more than a high school education are somewhat more likely to be eligible for spouse coverage than males and those with less than a high school education. The differences by race and income are striking. Nonwhites are much less likely to be eligible for spouse coverage than are whites, while very few of the unemployed with incomes less than $25,000 are eligible for coverage through an employed spouse. The income variable used in this analysis refers to total husband and wife income in 1992. Husbands and wives with combined incomes less than $25,000 in 1992 would have had relatively low income full-time jobs or a series of part-time jobs. These types of jobs are unlikely to have offered health insurance as a fringe benefit and therefore these jobs are an unlikely source of coverage for an unemployed spouse.

In order to illustrate the differences in the effects of eligibility for COBRA and eligibility for spouse coverage, we present cross tabulations of eligibility and health insurance outcomes in Table 6.

Eligibility and Health Insurance Outcomes

Table 6A - Full Sample

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 23.92

  4.67

 26.62

  5.04

 11.55

Spouse

  0.00

 69.90

 51.90

  0.00

 14.46

Other

 16.26

  5.95

  3.58

 19.90

 15.60

None

 59.83

 19.47

 17.90

 75.06

 58.38

Total

100.00

100.00

100.00

100.00

100.00

Table 6B - By Gender

-> female=

Male

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 26.09

  8.76

 32.01

  6.16

 13.35

Spouse

  0.00

 61.99

 35.43

  0.00

 10.83

Other

 12.87

  6.63

  6.62

 14.92

 12.70

None

 61.04

 22.63

 25.95

 78.92

 63.12

Total

100.00

100.00

100.00

100.00

100.00

-> female=

Female

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 20.99

  0.00

  20.26

  3.18

  8.94

Spouse

  0.00

 78.94

 71.37

  0.00

 19.72

Other

 20.82

  5.19

  0.00

 28.18

 19.80

None

 58.19

 15.88

   8.38

 68.65

 51.54

Total

100.00

100.00

100.00

100.00

100.00

Table 6C - By Age

-> over50=

>50 Yrs

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 19.07

  4.18

  26.04

  4.04

  9.43

Spouse

  0.00

 68.97

 51.47

  0.00

 14.61

Other

 17.33

  5.10

  3.20

 18.78

 15.11

None

 63.60

 21.76

 19.30

 77.18

 60.85

Total

100.00

100.00

100.00

100.00

100.00

-> over50=

50+ Yrs

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 39.54

  7.25

  28.80

  9.66

  20.41

Spouse

  0.00

 74.87

 53.46

  0.00

 13.85

Other

 12.80

 10.48

  5.02

 25.08

 17.68

None

 47.65

  7.40

 12.72

 65.27

 48.06

Total

100.00

100.00

100.00

100.00

100.00

Table 6D - By Race

-> nonwhite=

White

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 26.12

  5.30

  28.61

  5.49

 12.62

Spouse

  0.00

 68.65

 50.77

  0.00

 16.27

Other

 19.08

  5.73

  1.93

 20.16

 15.17

None

 54.81

 20.32

 18.70

 74.35

 55.32

Total

100.00

100.00

100.00

100.00

100.00

-> nonwhite=

Nonwhite

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 17.89

  0.00

  16.00

  3.87

  8.18

Spouse

  0.00

 79.15

 57.93

  0.00

  8.80

Other

  8.54

  7.59

 12.46

 19.23

 15.02

None

 73.570

 13.26

 13.61

 76.90

 68.00

Total

100.00

100.00

100.00

100.00

100.00

Table 6E - By Education Level

-> gthseduc=

<=HS

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 17.78

  2.06

  35.19

  3.76

  8.61

Spouse

  0.00

 71.21

 46.71

  0.00

 12.66

Other

 10.68

  5.04

  0.00

 17.31

 13.08

None

 71.55

 21.69

 18.10

 78.93

 65.65

Total

100.00

100.00

100.00

100.00

100.00

-> gthseduc=

<HS

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 30.29

  7.82

  19.30

  7.55

  15.77

Spouse

  0.00

 68.32

 56.33

  0.00

 17.07

Other

 22.05

  7.06

  6.65

 24.98

 19.23

None

 47.66

 16.79

 17.72

 67.47

 47.93

Total

100.00

100.00

100.00

100.00

100.00

Table 6F - By Income Level

-> lowinc=

25K

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 41.64

  5.89

  23.79

 15.28

 21.11

Spouse

  0.00

 75.51

 53.92

  0.00

 34.25

Other

 17.43

  5.70

  4.05

 37.59

 15.57

None

 40.93

 12.90

 18.24

 47.13

 29.06

Total

100.00

100.00

100.00

100.00

100.00

-> lowinc=

>=25K

Health Insurance Category

COBRA and Spouse Insurance Eligibility

COBRA

Spouse

Both

Neither

Total

Employer

 12.49

  0.00

  48.37

  3.13

  5.95

Spouse

  0.00

 48.52

 36.37

  0.00

  2.88

Other

 15.51

  6.92

  0.00

 16.60

 15.62

None

 72.00

 44.56

 15.26

 80.27

 75.55

Total

100.00

100.00

100.00

100.00

100.00

Source: 1993 April Current Population Survey

Table 6A shows the eligibility status and health insurance outcomes for the full sample of the unemployed. 23.92% of those only eligible for COBRA are observed to have coverage from their former employers while 69.90% of those only eligible for spouse coverage elect to use it. Those eligible for both types of coverage are more likely to elect spouse coverage (51.90% vs. 26.62%). Spouse coverage is more likely to be used among the unemployed than COBRA coverage, perhaps because of lower out-of-pocket expense or because it is potentially permanent coverage (as long as the spouse is employed) while COBRA is not. The net result is that those with eligibility for spouse coverage are much less likely to be uninsured than those with eligibility for COBRA. However, those eligible only for COBRA (and not spouse coverage) are much more likely to be insured than those eligible for neither spouse nor COBRA coverage (40.18% vs. 24.94%).(5)

Gender differences are shown in Table 6B. Among those eligible only for COBRA coverage, the difference in the COBRA election rate is not large. However, women are much more likely than men to opt for spouse coverage if they are eligible for it. When eligible for both spouse coverage and COBRA, men are almost equally likely to elect COBRA as spouse coverage while women are much more likely to opt for spouse coverage. Among those eligible for neither COBRA nor spouse coverage, women are more likely to obtain other coverage, perhaps because they are more likely to be eligible for public programs such as Medicaid. However, as might be expected, this group has the highest rate of non-insurance among the unemployed.

Those over 50 are more likely to elect COBRA and use spouse coverage and have higher overall coverage rates. However, when eligible for both spouse and COBRA coverage, those under 50 are more likely to elect COBRA coverage than those over 50 (Table 6C). Similar patterns are observed for those with a high school education or less relative to those with more than high school, and for those with incomes less than $25,000 relative to those with incomes more than $25,000 (Tables 8E and F). Perhaps those under 50, with a high school education or less, or with incomes less than $25,000 are more likely to elect COBRA because spouse coverage is less likely to be a permanent alternative for them. Whites are more likely to elect COBRA and be covered than nonwhites (Table 8D). However, the rate of non-coverage is similar for whites and nonwhites who are not eligible for COBRA or spouse coverage.

The tabulations shown in Tables 1 through 8 provide us with an initial determination of the effects of COBRA eligibility on health insurance coverage. However, they do not hold other variables constant that could be affecting health insurance coverage of the unemployed. In order to provide more precise estimates of the COBRA and spouse effects, we turn to the estimation of binary and multinomial logit models of health insurance coverage in the next section.

Statistical Models of Health Insurance Coverage and COBRA Eligibility: Theoretical Considerations - In this section, we construct statistical models to estimate the effects of COBRA eligibility on the health insurance coverage of the uninsured. In the previous section, we saw that those eligible for COBRA were more likely to be covered by health insurance. However, we were only able to partially control for other factors that may influence whether or not a given unemployed worker is covered. Here we attempt to parameterize a COBRA effect within a multivariate framework.

In many ways the observed health insurance outcome of the unemployed can be thought of as a standard consumer choice problem. The unemployed worker chooses whether to purchase health insurance, given his or her income, prices, and his or her tastes. The problem is complicated somewhat by the fact that because we are dealing with health, outcomes are uncertain, and an expected utility framework is usually employed. Within an expected utility framework, the unemployed will have a higher demand for health insurance if the expected loss from the event being insured against is higher, holding constant the probability of the event. The expected loss could be monetary or non-monetary. It could be lost earnings if the individual is unable to work, although presumably for an unemployed worker this is less of a problem than for an currently employed worker. Or the expected loss could be the loss of well-being associated with bad health and untreated health problems. Similarly, for relatively low probability events, the unemployed worker's demand for insurance increases as the probability increases for an event with a loss of a given magnitude. Thus, we need to consider prices, income, tastes, expected loss, and probability of a loss when considering whether an individual chooses to obtain health insurance.(6)

The problem is that we do not observe health insurance premiums (prices), expected losses, or probabilities of loss in the April CPS. However, we do observe a number of variables that are correlated with these variables. The premium is effectively reduced for those eligible for COBRA or coverage through their spouse's employer plan. In either case the premium an individual faces is less than if he or she were to go out and purchase an individual health plan. Thus, unemployed workers eligible for COBRA or a spouse's plan are more likely to be covered.

Of course, there are other variables correlated with prices, incomes, tastes, expected losses, and probability of loss that are available in the 1993 April CPS and can be included in our statistical models. Income can be measured by the sum of own and spouse's total income in the previous year, which is available using the April-March CPS match. Assuming that health insurance is a normal good, we would expect that higher income workers would be more likely to be covered by health insurance.

The frequency and severity of health problems increases with age so we would expect health insurance to increase among the unemployed as age increases. Those with children are likely to have a greater frequency of use of health care services and the same may be true of females, leading to a higher demand and a greater likelihood of coverage.(7) Those with disabilities or whose spouses have disabilities would have smaller expected losses from further health problems but a greater frequency of need for health services. Premiums for purchased insurance are likely to be higher among this group or they may be locked out of coverage due to preexisting conditions. The net effect on coverage is uncertain. Those with more education are likely to be more knowledgeable about health and thus need fewer health services. On the other hand, expected losses are higher for this group. Of course, any income losses are likely to be muted in a sample of unemployed. Until eligibility ends or an individual is too sick to engage in job search, they continue to receive unemployment and checks will continue whether sick or not.

In general, health outcomes tend to be worse among blacks than whites, suggesting a greater demand for services. However, because earnings are lower, earnings losses from poor health are lower among blacks. Also, information about health insurance coverage may differ between blacks and whites. In other words, it may be possible that the smaller losses when an individual gets sick are canceled out by the greater demand and use of health services. Thus, the direction of the net effect of race, like many of the demographic variables available in the 1993 April CPS, is uncertain a priori.

We use binary and multinomial logit models to estimate the effect of COBRA eligibility on the health insurance coverage of the unemployed. The binary logit models provide estimates of the determinants of whether an unemployed individual has any form of health insurance coverage, and the multinomial logit models the choice of alternative coverage categories (former employer, other coverage, no coverage). The models use data from the regular portion of the 1993 April CPS, the Employee Benefits Supplement, and the April-March match.

Binary Logit Analysis - More formally, the 1993 April CPS data allow us to estimate the probability of being unemployed and having health insurance, or E(Pt|X,u=1), where Pt is the probability of having insurance after being unemployed for a time t, X is a vector of characteristics, and u=1 if the worker is unemployed and zero otherwise. We can estimate this conditional probability using logit analysis. This allows us to calculate the probability that an unemployed worker will have health insurance given a set of observed characteristics X. The logit model also allows us to determine the marginal effects of changes in the X variables on health insurance coverage.

Included in X are our measure of COBRA eligibility, eligibility for spouse's employer coverage, family income, and demographic characteristics of the unemployed worker: age, education, race, gender, number of children, marital status, and disability status of the unemployed worker and his or her spouse. Formal definitions and means of these variables and the dependent variable are shown in Table 7.(8)

Means And Definitions Of Variables Used In Logit Analysis

Table 7

Variable

Mean

Definition

HI

.4328358

=1 if covered by health insurance

COBRA

.3239684

=1 if eligible for COBRA coverage

Spouse

.2396839

=1 if elig. for spouse coverage

female

.4363477

=1 if gender is female

age

39.55575

age in years

children

1.611062

# own children < 18 years old

married

.5522388

=1 if married spouse present

hsless

.1791045

=1 if has less than high school diploma

somecoll

.1992976

=1 if has some college

assoc

.0597015

=1 if has associate's degree

ba

.1009658

=1 if has bachelor's degree

baplus

.0491659

=1 if has more than bachelor's degree

black

.1395961

=1 if race is black

other

.0553117

=1 if race is other than black or white

disab

.0535558

=1 if wks. worked reduced due to illness or disability

spdisab

.0298507

=1 if spouse's wks. reduced due to ill. or disab.

totinc

25718.41

sum of own and spouse's income in 1992

Source: 1993 April Current Population Survey (n=1139)

In Table 8, we show binary logit estimates of health insurance coverage with A) only the insurance eligibility variables and B) the insurance eligibility variables and demographic characteristics.

Binary Logit Estimates of Health Insurance Coverage

Table 8A - Insurance Eligibility Variables Only

Logit Estimates

Log Likelihood = -670.8202

Number of obs   = 1139

chi2(2)             = 216.73

Prob > chi2      = 0.0000

Pseudo R2       = 0.1391

HI

Coef.

Std. Err.

z

P>|z|

[95% Conf. Interval]

COBRA

.67821

.1399981

4.844

0.000

.4038187

.9526013

Spouse

2.143421

.1683546

12.732

0.000

1.813452

2.47339

_cons

-.9938632

.0906148

-10.968

0.000

-1.171465

-.8162615

Table 8B - Full Specification

Logit Estimates

Log Likelihood = -604.30624

Number of obs   = 1139

chi2(2)             = 349.76

Prob > chi2      = 0.0000

Pseudo R2       = 0.2244

HI

Coef.

Std. Err.

z

P>|z|

[95% Conf. Interval]

COBRA

.415881

.1541821

2.697

0.007

.1136897

.7180723

Spouse

1.45283

.2025905

7.171

0.000

1.05576

1.8499

female

.4058274

.1438808

2.821

0.005

.1238262

.6878287

age

.0335475

.0074741

4.489

0.000

.0188986

.0481963

children

.0720561

.0631101

1.142

0.254

-.0516373

.1957496

married

.4608835

.1958698

2.353

0.019

.0769857

.8447813

hsless

-.1231223

.2064497

-0.596

0.551

-.5277563

.2815177

somecoll

.3254417

.1911921

1.702

0.089

-.0492879

.7001712

assoc

.5209284

.2972679

1.752

0.080

-.061706

1.103563

ba

.5919875

.2531611

2.338

0.019

.0958009

1.088174

baplus

1.462033

.3818037

3.829

0.000

.713711

2.210354

black

-.0720482

.2191163

-0.329

0.742

-.5015083

.357412

other

.1630545

.310763

0.525

0.600

-.4460298

.7721388

disab

-.6363293

.3384751

-1.880

0.060

-1.299728

.0270697

spdisab

.0494727

.3947527

1.125

0.900

-.7242285

.8231739

totinc

.0000175

3.98e-06

4.406

0.000

.9.73e-06

.0000253

_cons

-3.25259

.3598647

-9.038

0.000

-3.957912

-2.547269

Source: 1993 April Current Population Survey

In Table 8A, we see that COBRA and spouse eligibility increase the likelihood of health insurance coverage, mirroring the results from our earlier cross tabulations. The estimated COBRA coefficient in Table 8A is .678, which when converted to a probability is .140.(9) In Table 8B, after controlling for a number of observed characteristics, the estimated COBRA effect is reduced substantially. The estimated COBRA coefficient is .416, which when converted to a probability and evaluated at the means of the other variables is .075. Our estimated COBRA effect of .075 is similar to the .067 estimated effect of Gruber and Madrian (1995a) using SIPP data.

Several observable worker characteristics in Table 8B are significantly related to health insurance coverage. Age and income are significantly positively related to coverage as expected. Females, married individuals, and those with higher levels of education are more likely to be covered than males, single individuals, and those with lower levels of education. There is some evidence that disabled individuals have lower coverage rates.

Our estimated COBRA effects appear to be somewhat larger than that estimated by Klerman and Rahman (1992). They estimate a COBRA coefficient of .052 using a probit specification and SIPP data. In order to provide a comparison with Klerman and Rahman (1992), we estimate a probit model with a similar specification to theirs using the April 1993 CPS.(10) In our estimated probit model (results not shown), the COBRA coefficient is .228 (std. error= .083). Evaluated at the means, the estimated increase in the probability of being covered is .090. While we are unable to translate the Klerman and Rahman (1992) estimate into a comparable probability without their sample means, it appears from the magnitudes of the estimated coefficients that our COBRA effect is larger than theirs.

Multinomial Logit Analysis - The binary logit estimates shown in Table 8 only consider whether an unemployed individual is covered or not and does not distinguish among health insurance coverage from a former employer and other sources. Therefore we next estimate multinomial logit models that include these alternatives along with having no insurance coverage at all. Let P1t denote the probability that a worker receives health insurance from a former employer after being unemployed for a time t. A worker may have coverage from an alternative source such as the spouse's employer, the government or private health insurance, denoted by P2t. The probability that a worker has no health insurance coverage is P3t. We use multinomial logit to estimate E(Pit|X, u=1), i=1,2,3. This approach not only allows us to track who has health insurance when unemployed, but also to examine the determinants of the source of coverage.(11)

As before, we show two specifications in Table 9. In Table 9A we show the multinomial logit estimates that include only the insurance eligibility variables and in Table 9B we show the full specification, including a set of observable worker characteristics. The multinomial logit estimates show the effect of a change in the X variable on the natural log of the probability of each outcome relative to the base outcome, in this case no insurance.

Multinomial Logit Estimates of Health Insurance Coverage

Table 9A - Insurance Eligibility Variables Only

Multinomial Regression

Log Likelihood = -900.6442

Number of obs   = 1139

chi2(4)             = 327.94

Prob > chi2      = 0.0000

Pseudo R2       = 0.1540

hcat

Coef.

Std. Err.

z

P>|z|

[95% Conf. Interval]

COBRA

1.884316

.2129647

8.848

0.000

1.466913

2.301719

Spouse

1.012408

.2680134

3.777

0.000

.487111

1.537704

_cons

-2.660824

.176637

-15.064

0.000

-3.007026

-2.314622

Other

COBRA

.044253

.1651787

0.268

0.789

-.2794913

.3679974

Spouse

2.424341

.1741311

13.923

0.000

2.083051

2.765632

_cons

-1.237432

.0980825

-12.616

0.000

-1.42967

-1.045193

(Outcome hcat==None is the comparison group)

Table 9B - Full Specification

Multinomial regression

Log Likelihood = -815.62127

Number of obs   = 1139

chi2(32)             = 497.98

Prob > chi2       = 0.0000

Pseudo R2        = 0.2339

hcat

Coef.

Std. Err.

z

P>|z|

[95% Conf. Interval]

COBRA

1.588409

.2295178

6.921

0.000

1.138563

2.038256

Spouse

.1672063

.3196103

0.523

0.601

-.4592184

.793631

female

-.1940584

.2270073

-0.855

0.393

-.6389846

.2508678

age

.0480965

.0113131

4.251

0.000

.0259232

.0702698

children

-.0714158

.1102133

-0.648

0.517

-.28743

.1445983

married

.4595111

.3024614

1.519

0.129

-.1333023

1.052325

hsless

-.5268852

.3827794

-1.376

0.169

-1.277119

.2233487

somecoll

.5057127

.2838788

1.781

0.075

-.0506794

1.062105

assoc

.2733529

.4502701

0.607

0.544

-.6091604

1.155866

ba

.4839196

.3612337

1.340

0.180

-.2240854

1.191925

baplus

1.364486

.4950886

2.756

0.006

.3941298

2.334842

black

-.178763

.3540869

-0.505

0.614

-.8727605

.5152345

other

-.4102342

.5561079

-0.738

0.461

-1.500186

.6797173

disab

-.6368233

.5640805

-1.129

0.259

-1.7422401

.4687541

spdisab

.071244

.6243361

0.114

0.909

-1.152432

1.29492

totinc

.0000225

4.81e-06

4.684

0.000

.0000131

.000032

_cons

-5.125466

.5776516

-8.873

0.000

-6.257643

-3.99329

Other

COBRA

-.1367382

.1773388

-0.771

0.441

-.4843158

.2108395

Spouse

1.779716

.2138468

8.322

0.000

1.360483

2.198748

female

.6214455

.1581918

3.928

0.000

.3113952

.9314958

age

.0273117

.0083285

3.279

0.001

.0109881

.0436353

children

-.1107654

.0676498

1.637

0.102

-.0218259

.2433567

married

.4848019

.219375

2.210

0.027

.0548348

.9147689

hsless

.0294785

.2253034

-0.131

0.896

-.471065

.4121081

somecoll

.2322449

.213849

1.086

0.277

-.1868915

.6513812

assoc

.6329582

.3254872

1.945

0.052

-.0049849

1.270901

ba

.6353141

.2772659

2.291

0.022

.091833

1.178745

baplus

1.518018

.4109095

3.694

0.000

.7126499

2.323386

black

-.0155238

.2421146

-0.064

0.949

-.4900598

.4590122

other

.2888596

.3292701

0.877

0.380

-.3564978

.9342171

disab

-.6338952

.3783407

-1.675

0.094

-1.375429

.1076389

spdisab

.0918728

.427573

0.215

0.830

-.7461549

.9299005

totinc

.0000146

4.27e-06

3.421

0.000

6.24e-06

.000023

_cons

-3.405367

.4006411

-8.500

0.000

-4.190609

-2.620125

(Outcome hcat==None is the comparison group)

Source: 1993 April Current Population Survey

As might be expected, COBRA eligibility does not significantly affect the outcome of other coverage relative to no coverage, and spouse coverage does not significantly affect the outcome of employer coverage relative to no coverage. As with the binary logits, the addition of demographic variables reduces the magnitude of the COBRA and spouse insurance effects. Table 10 shows the estimated effect of COBRA eligibility on the three outcomes evaluated at the means of the variables.

Estimated Effect of COBRA Eligibility on Health Insurance Outcomes

Table 10

Model Specification

Health Insurance Outcome

Employer

Other

No Insurance

Total

A. Health Vbls. Only

.202

-.063

-.139

0.00

B. Full Specific.

.154

-.081

-.073

0.00

Source: 1993 April Current Population Survey (uses estimates from Table 9)

The estimated increase in the probability of employer coverage from COBRA eligibility is .202 from the estimates in Table 9A and is .154 from the full specification including demographic characteristics in Table 9B. The estimated effect on employer coverage and no coverage is reduced by the addition of individual characteristics to the model.

The effects of the individual characteristics are in many ways similar to those estimated using the binary logit model. Increases in age, income, and education significantly increase the probabilities of both employer and spouse insurance relative to the probability of no coverage. Being female or married significantly raises the probability of other coverage relative to no coverage. The magnitudes of the effects vary across the two insurance choices. Increases in age and income produce bigger effects on the relative probability of employer coverage, being female produces a bigger effect on other coverage, and the effects for being married are almost identical for the two choices.

Male - Female Coverage Differences Among the Unemployed - As mentioned in the Introduction, gender differences in health insurance coverage among the unemployed remain an interesting puzzle. We return to this issue now. What do we know so far? Unemployed women are more likely to have some type of coverage than unemployed men (Table 4A). Among COBRA eligible, men are more likely to have employer coverage than are women (Table 4A). On the other hand, women are more likely to have coverage from their spouse's employer or some other type of coverage among both those eligible and ineligible for COBRA (Table 4A). Further, women are more likely to choose spouse's employer coverage if eligible for it (Table 6A). Women also are somewhat more likely to be eligible for COBRA coverage and spouse's employer coverage than are men (Table 5A). Assessing the relative strengths of these various factors will go a long way toward understanding male-female differences in coverage.

As a first step, we estimate separate binary logit and multinomial logit models for males and females and present the estimated COBRA and spouse coverage effects on the various outcomes in Table 11. The estimates shown use models with the same specification as Tables 8A and 9A and are evaluated at the mean of the other health eligibility variable.(12)

Gender Differences in COBRA and Spouse Eligibility Effects

Table 11

Model, Gender, Variable

Health Insurance Outcome

Any Coverage

Employer

Other

No Ins.

Total

Logits

Males

COBRA Elig

.156

 

 

-.156

0.00

Spouse Elig

.454

 

 

-.454

0.00

Females

COBRA Elig

.120

 

 

-.120

0.00

Spouse Elig

.490

 

 

-.490

0.00

Mult. Logits

Males

COBRA Elig

 

.212

-.056

-.156

0.00

Spouse Elig

 

.013

.441

-.454

0.00

Females

COBRA Elig

 

.184

-.065

-.119

0.00

Spouse Elig

 

-.056

.544

-.488

0.00

Source: 1993 April Current Population Survey

Table 11 shows that the estimated effect of COBRA eligibility on employer coverage is greater for males than for females. The estimated effect of spouse insurance eligibility on spouse coverage is greater for females. In fact, eligibility for spouse coverage actually reduces the probability that they will be observed with employer coverage. These results suggest that women's responses to spouse eligibility outweigh men's responses to COBRA eligibility, producing higher overall coverage rates for women and higher employer coverage rates for men. This would be the end of the story if eligibility rates for COBRA and spouse coverage were the same for men and women. However, we know from Table 5A that eligibility rates are somewhat higher for women.

The relative contributions of differences in eligibility rates by gender and gender differences in the estimated responses to COBRA and spouse insurance eligibility are shown in Table 12. We perform "Blinder-Oaxaca" decompositions on the observed unweighted raw difference in health insurance coverage for unemployed men and women. The same set of logit estimates from Table 11 are also used here. This exercise allows us to decompose the raw differences into amounts due to differences in X's (in this case differences in COBRA and Spouse eligibility rates) and differences in estimated parameters, or responses to eligibility.(13)

Decomposition of Male - Female Differences in Outcomes

Table 12

Model, Portion of Raw Diff.

Health Insurance Outcome

Any Coverage

Employer

Other

No Ins.

Logits

Due to X's

-.010

 

 

.010

Due to b's

-.058

 

 

.058

Total Raw

Difference (Male-Fem.)

-.068

 

 

.068

Mult. Logits

Due to X's

 

.002

-.013

.011

Due to b's

 

.048

-.105

.057

Total Raw

Difference

 

.050

-.118

.068

Source: 1993 April Current Population Survey

The decompositions illustrate that gender differences in eligibility rates (X's) only explain a small portion of the raw differences in health insurance outcomes. Instead, it is gender differences in the responses to eligibility for COBRA and Spouse coverage. However, we are still left with explaining why gender differences in the responses to COBRA and spouse insurance exist. Perhaps women have higher responses to the possibility of spouse coverage than men because the expectation is for longer job tenure for men. Thus, spouse coverage is a more attractive alternative for women than for men, all else equal. Why are men more likely to elect COBRA coverage, even holding constant spouse coverage? The answer may be that unemployed women are more likely to qualify for public sector health insurance such as Medicaid than are men. Among some lower income women, public sector alternatives would be preferable than paying for COBRA coverage out of their own pocket.

Conclusions

When confronted with a medical catastrophe, lower- or middle-income households that lack health insurance face possible financial devastation. A drawback of our system of employer-provided health insurance is that labor force transitions may have the unintended side effect of eliminating the worker's health insurance coverage. COBRA was enacted to provide workers and their dependents with a safety net in the event of a job interruption. Health insurance coverage can be maintained under COBRA during changes in employment status.

Using the 1993 April CPS, we estimate that 33.59% of the stock of unemployed are or were eligible to elect COBRA. Of those eligible, 24.57% are observed to have health insurance from a former employer. After controlling for worker characteristics and eligibility for insurance from a spouse's employer, our binary logit models imply that COBRA eligibilty increases the probability of health insurance coverage by .075. Our multinomial logit models provide a more detailed estimate of the COBRA eligibility effect: COBRA eligibility increases the probability of employer coverage by .154 reduces the probability of other coverage by .081, and reduces the probability of no coverage by .073. While many unemployed workers turn to other sources of coverage such as through a spouse's employer, our results clearly indicate that COBRA is an important part of the safety net for unemployed workers.

References

  • Flynn, Patrice. 1992. "Employment-Based Health Insurance: Coverage Under COBRA Continuation Rules." In Health Benefits and the Workforce, Pension and Welfare Benefits Administration, U.S. Department of Labor, Washington, DC, pp. 105-116.

  • Folland, Sherman, Allen C. Goodman, and Miron Stano. 1993. The Economics of Health and Health Care. Prentice Hall, Englewood Cliffs, NJ.

  • Gruber, Jonathan and Brigitte C. Madrian. 1995a. "Non-Employment and Health Insurance Coverage." NBER Working Paper No. 5228, August.

  • Gruber, Jonathan and Brigitte C. Madrian. 1995b. "Health Insurance Availability and the Retirement Decision." American Economic Review 85:4 (September), pp. 938-948.

  • Karoly, Lynn A. and Jeannette A. Rogowski. 1994. "The Effect of Access to Post-Retirement Health Insurance on the Decision to Retire Early." Industrial and Labor Relations Review 48:1 (October), pp. 103-123.

  • Klerman, Jacob Alex and Omar Rahman. 1992. "Employment Change and Continuation of Health Insurance Coverage." In Health Benefits and the Workforce, Pension and Welfare Benefits Administration, U.S. Department of Labor, Washington, DC, pp. 93-104.

  • Sindelar, Jody L. 1982. "Differential Use of Medical Care by Sex." Journal of Political Economy 90:5 (October), pp. 1003-1019.

  • U.S. Bureau of the Census. 1995. Statistical Abstract of the United States, 115th edition. Washington, DC.

  • U.S. Department of Labor. 1990. Health Benefits Under the Consolidated Omnibus Budget Reconciliation Act (COBRA). Pension and Welfare Benefits Administration, Washington, DC.

  • U.S. Department of Labor. 1994. Pension and Health Benefits of American Workers: New Findings from the April 1993 Current Population Survey. Washington, DC.

Appendix

Current Population Survey Questions Used in the Analysis:

  1. Health Insurance Questions from the April 1993 CPS

a_s84 Are you covered by a health insurance plan?

1 = yes
2 = no
3 = don't know

a_s85 Is this plan provided by a former employer?

1 = yes
2 = no
3 = don't know

a_s88 Were you covered under a health insurance plan provided by your employer on your last job?

1 = yes
2 = no
3 = don't know

a_s61 Does your employer offer a health insurance plan to any of its employees?

1 = yes, employee and family coverage offered
2 = yes, employee coverage only
3 = yes, employee coverage, don't know family coverage
4 = no
5 = don't know

a_s62 Are you covered by this insurance plan?

1 = yes for myself and family members
2 = yes - for myself only
3 = no
4 = don't know

  1. Other Questions from the April 1993 CPS

a_age Age (years)

a_maritl Marital Status

1 = married, civilian spouse present
2 = married, armed forces spouse present
3 = married, spouse absent
4 = widowed
5 = divorced
6 = separated
7 = never married

a_sex Sex

1 = male
2 = female

a_hga Educational Attainment

00 = none
31 = less than 1st grade
32 = 1st - 4th grade
33 = 5th or 6th grade
34 = 7th or 8th grade
35 = 9th grade
36 = 10th grade
37 = 11th grade
38 = 12th grade - no diploma
39 = high school graduate
40 = some college but not degree
41 = associate degree - occupational/vocational program
42 = associate degree - academic program
43 = bachelor's degree
44 = master's degree
45 = professional degree
46 = doctorate degree

a_race Race

1 = white
2 = black
3 = American Indian, Aleut, Eskimo
4 = Asian or Pacific Island
5 = other

a_whylk Why did you start looking for work; was it because...

1 = lost job 2 = quit job
3 = left school
4 = wanted temporary work
5 = change in home or family responsibilities
6 = left military service

a_wkslk Weeks unemployed (# of weeks)

a_ind Industry of Employment (3 digit Census industry code)

a_clswkr Class of Worker

1 = private
2 = federal government
3 = state government
4 = local government
5 = self employed - incorporated
6 = self employed - unincorporated
7 = without pay
8 = never worked

a_pfnocd Number of own children <18 in primary family

0 = not in primary family
1 = no children
2 = 1 child
3 = 2 children
4 = 3 children
5 = 4 children
6 = 5 children
7 = 6 children
8 = 7 children
9 = 8+ children

a_supwgt April Supplement Sample Weight

  1. Questions from the March 1993 CPS

rsnnotw What was the main reason you did not work in 1992?

1 = ill or disabled
2 = retired
3 = taking care of home
4 = going to school
5 = could not find work
6 = other

pyrsn What was the main reason you were not working or looking for work in the remaining weeks of 1992? (i.e. the weeks other than those spent working or looking for work)

1 = ill or disabled
2 = taking care of home
3 = going to school
4 = retired
5 = no work available
6 = other

noemp Counting all locations where this employer operates, what is the total number of persons who work for your employer?

1 = under 10
2 = 10-24
3 = 25-99
4 = 100-499
5 = 500-999
6 = 1000+
7 = other

ptotval Total income in 1992 (dollars)

Endnotes

  1. All of the counts and tabulations in this section through Table 6 are weighted using the 1993 April CPS supplemental weights (a_supwgt). The weights are designed to improve the accuracy of population counts and distributions.

  2. All of these variables are constructed from questions in the April 1993 CPS except family income which is taken from the March 1993 CPS and is included in the April-March match set of questions. The income variable is constructed from questions concerning total income in 1992. The April CPS includes data on current earnings but the question only applies to currently employed workers and not the unemployed.

  3. The Appendix provides a listing of the CPS questions used in our analysis.

  4. Those who qualify as disabled under the Social Security Act may be eligible for 29 months. However, those who qualify as disabled will not be looking for work and will not be counted among the unemployed, and thus will not be in our data. Also, we do not consider the coverage status of spouses or dependents of the unemployed. Therefore, the 36 month eligibility period for spouses and dependents after qualifying events is also not a factor in our data.

  5. The rates of coverage are obtained by summing across the three coverage categories in Table 6A. For those eligible for COBRA coverage, overall coverage is the sum of 23.92% and 16.26% and for those eligible for neither spouse nor COBRA coverage, overall coverage is the sum of 5.04% and 19.90%.

  6. See Folland, Goodman, and Stano (1993, Chapter 11) for a more extensive treatment of determinants of the demand for health insurance.

  7. See Sindelar (1982) and U.S. Census Bureau (1995, Tables 178, 179, 189, 190, 191, 192, 194) for evidence on greater frequency of use of health care services for females.

  8. The results presented in Tables 7 through 12 use unweighted data. These tables show means, parameter estimates, and calculations for behavioral models rather than population summary statistics. In this case, weights serve primarily to correct for heteroscedasticity which we have no reason to believe is present in these data.

  9. Our earlier cross tabulations produced similar results. In Table 3A, the implied COBRA effect on the probability of any form of coverage is 13.15% (50.35% - 37.20%).

  10. Our specification includes weeks unemployed, a dummy for the first month of unemployment, age, age squared, a series of education completion dummies, and a COBRA eligibility dummy. Thus, our specification is identical to Klerman and Rahman (1992) except that they use months instead of weeks unemployed and years of schooling instead of a series of schooling dummies.

  11. While we used four health insurance outcomes in our cross tabulation analysis (employer, spouse's employer, other, no coverage), we were forced to reduce the number of outcomes to three in the multinomial logit analysis for identification reasons. In particular, only those eligible for spouse coverage are observed to have spouse coverage. No one eligible for neither spouse nor COBRA coverage is observed to have spouse coverage. Thus, it is not possible to include spouse coverage as a separate choice in the multinomial logit. Instead we aggregate spouse and other coverage and there is no longer an identification problem. This is not a problem with COBRA eligibility and employer coverage. Individuals are observed with and without employer coverage in all eligibility categories. For example, some of those not eligible for COBRA are still be observed to have employer coverage.

  12. We also estimated multinomial logit models with specifications similar to those in Table 11B and obtained similar though in general somewhat smaller estimated effects. The specifications used differed from that used in Table 9B in that the female, spdisab, and disab variables were dropped. The spdisab and disab variables could not be used in the gender specific analysis because no females with disab=1 or spdisab=1 were observed to have employer provided insurance.

  13. Similar results are obtained when the full set of explanatory variables from Table 9B (except for female, disab, and spdisab) are used. The portions of the raw differences explained by differences in X's increase slightly; however, the major portions of the raw differences are still due to differences in the estimated b's.

End Credits

Mark C. Berger
Dan A. Black
Frank A. Scott
Carolyn Looff and Associates
1635 Ashwood Road
Lexington, KY 40502

The opinions expressed in this study are the sole responsibility of the authors and do not represent the views of the U.S. Department of Labor. Amitabh Chandra provided capable research assistance.