Canadian Health&Care Mall about HOW TAX POLICY SHAPED U.S. HEALTH INSURANCE Using the NORC Data

Posted on 20th January and posted in  Category: HEALTH INSURANCE

Health Insurance Market

Measuring the Impact of the 1954 Tax Changes on the Health Insurance Market

As noted previously, the effect of the 1954 tax subsidy on the demand for health insurance is two-fold. First, it may have increased the number of employers and unions that offered group health insurance plans. Since group insurance is less expensive and is easier to obtain than individual coverage, an increase in access to group insurance would generate an increase in the number of the households with insurance coverage. Second, the tax changes directly made health insurance less expensive for those households with employers who contributed to their health plans because employees did not have to pay income taxes on these contributions. As a result, the “price” of health insurance to the employee fell, which should lead in an increase in the amount of coverage purchased. You do not demand to earn a lot of money to buy drugs via Canadian Health&Care Mall and be afraid that it cannot be covered by your health insurance.

To test these effects, I first examine the impact of the tax changes on the probability that a person had the opportunity to purchase group insurance coverage. If the tax subsidy did lead more employers and more unions to sponsor health insurance plans, then access to group insurance should have increased among employed persons and union members in 1958 relative to 1953. The second effect of the tax subsidy on increasing the amount of coverage purchased can be examined in two ways. First, probit estimation can be used to identify what factors influenced a household’s decision to purchase health insurance coverage or not. This is especially important in determining how access to group insurance affects a household’s decision to purchase health insurance coverage. In addition, it can shed some light on whether the tax changes influenced a household’s decision to purchase health insurance or not. Second, while the tax changes may have induced some people to purchase coverage who would not have otherwise, it is more likely that households who had already made the decision to purchase health insurance purchased a higher level of coverage after the tax changes were enacted. As a result, tobit estimation is used to gauge the effect of the tax changes on the amount of coverage purchased. Estimating the Impact of the Tax Subsidy on Access to Group Insurance Simple comparisons of the samples in 1953 and 1958 reported in Tables 1 and 2 suggest that more people had access to group insurance in 1958 than in 1953. In 1958, 67 percent of households had the opportunity to purchase health insurance through a group plan, compared to 48 percent of households in 1953. The 1954 tax changes resolved uncertainty surrounding the tax-exempt status of employer contributions, so more employers to sponsor plans. In addition, after 1954, union-sponsored plans that employers contributed to were also eligible for the exemption. Thus, access to group insurance should have increased in 1958 relative to 1953.

To specifically test this hypothesis, I estimate a probit model where the dependent variable is equal to one if the household had the opportunity to purchase group insurance, and zero otherwise. Independent variables include socioeconomic and demographic characteristics, including industry, occupation, age, union membership, and employment status. Of particular interest are the variables union and employed. Union is equal to one if the household head is a union member, and employed is equal to one if the head of the household is employed. Union is included because it may have been the case that they exerted pressure on employers to provide group health insurance plans, or themselves sponsored employee insurance plans, particularly after 1954 when tax exemption for group insuranceemployer contributions to such plans was assured. To test whether or not access to group insurance increased among union members and employed persons (the direct beneficiaries of the 1954 tax subsidy), I pool the 1953 and 1958 samples and include a year dummy equal to one in 1958. By interacting the year dummy with union membership and employment status, I can determine whether access to group insurance changed significantly between 1953 and 1958. Specifically, if the estimated coefficients on employed year and union year are statistically significant, then employed persons and union members were more likely to have access to group insurance in 1958 than in 1953.

After eliminating observations with missing values for variables in the regression, the sample in 1953 has 2238 observations, while 2165 observations are in the 1958 sample. Thus, the pooled sample that is used to estimate the probit model contains 4403 observations. Results of the probit estimation, reported in Table 3, support the hypothesis that the 1954 tax changes increased access to employment-based group insurance. The estimated coefficients on union*year and employed*year are statistically significant, suggesting that union members and employed persons in 1958 had a higher probability of having access to group insurance than they did in 1953. A change in the employed*year dummy variable from zero to one leads to a 18.6 percent increase in the probability that the household had access to group health insurance in 1958 compared to 1953. The estimated coefficient on the union*year dummy suggests that nearly a union members in 1958 were roughly 15 percent more likely to have access to group health insurance than their counterparts in 1953. Clearly, more employers and more unions were sponsoring group health insurance plans after 1954. While the 1954 tax subsidy is not the only factor that led to an expansion of group insurance, it certainly provided many employers and unions with an additional incentive to sponsor group plans.

Estimating the Demand for Health Insurance Coverage

Given that the tax changes in 1954 increased access to group plans, how did this affect the probability that a household purchased health insurance coverage? To examine the effect of the tax subsidy on the demand for health insurance coverage, I first estimate a probit model to determine the factors that determine whether or not a household purchases health insurance. A tobit model is then used to examine the factors that influence the amount of insurance coverage purchased by a household.

The demand for health insurance coverage is a function of socioeconomic and demographic characteristics, as well as the costs associated with obtaining insurance coverage. The costs associated with obtaining insurance coverage can be divided into two parts, the loading factor (load) on a policy, and the access costs associated with obtaining coverage. The loading factor on a policy is the percentage increase in the actuarially fair premium that insurance companies charge to cover expenses associated with the policy. The load on group insurance policies is lower than on individual policies because there are significantly lower administrative expenses associated with group insurance than with individual insurance. Employers may deduct the premium from employees’ paychecks (eliminating the need for insurance companies to bill individuals), and insurance agents do not have to meet with people individually.

In addition to the load on a policy, other costs associated with obtaining coverage include the access costscosts of seeking out insurance, as well as the insurability of a person of obtaining coverage. Thus, even after controlling for the lower explicit load on group insurance, the transactions costs associated with enrolling in group insurance are much lower than those associated with having to seek out an insurance company and obtain coverage individually. Unlike an individual plan which requires that consumers contact an agent, complete a detailed medical history, and possibly undergo a physical examination, a typical group insurance application requires only that an individual fill out a simple enrollment card. Households who have to merely sign up with their employer are probably more likely to have coverage than households who do not have the same opportunity. At a time when health insurance premiums were fairly low relative to other goods, these transactions costs may have been a significant portion of the total cost of insurance. In addition, group coverage is “… written without evidence of individual insurability when the individuals who comprise the group are bound together by some broad interest” (Faulkner, p. 199). Thus, it may be the case that some individuals who are considered “uninsurable” individually may be able to obtain group coverage since they do not have to undergo a rigorous screening process.

To illustrate how individual households (denoted by the subscript i) may choose to allocate their budgets between spending on other goods, (ci) and spending on health insurance, (h), I offer a straightforward model of utility maximization which provides a basis for the empirical relationship between consumption and spending on health insurance. Suppose a household has a probability n of suffering a sickness which causes an income loss of $S, either due to medical expenditures or due to lost wages. To protect against this loss, the household can purchase a health insurance policy which pays $hi in the event of illness. The cost of the policy is equal to the actuarially fair premium multiplied by (1 + Xi), where Xi e [0, is the loading fee on the policy, or the percentage increase in the actuarially fair premium that an insurance company charges to cover its administrative costs. Thus, the total premium of the policy is pi hi where pi = n(1 + Xi). In addition to the premium, an individual who purchases health insurance coverage faces a fixed transaction cost, F, associated with obtaining coverage. The transactions costs associated with obtaining individual coverage are much greater than those associated with obtaining group coverage. While F may not affect the level of insurance coverage purchased at an interior solution, it will affect the size of the household’s feasible set. Thus, a household which purchases health insurance if transactions costs are low may opt not to purchase insurance if transactions costs are high.

In an expected utility framework, a household purchases health insurance if and only if its expected utility with health insurance coverage exceeds its expected utility without health insurance coverage. The price of the composite consumption good has been normalized to unity for convenience. The household’s decision problem can be written as:

Formule 1

Note that in this framework, the household is assumed to be risk-averse (so that they may purchase an actuarially unfair premium), but utility is well-defined even in the event that h = 0. In addition, a can be thought of as the individual’s reservation price for insurance, and is a function of the individual’s socioeconomic and demographic characteristics such as income, age, marital status and education.

For an interior solution, the model yields:

Formule 2

I specify the reservation price as at = exp(ziY + ut) , where zt is a vector of these socioeconomic and socioeconomicdemographic characteristics, and u is distributed normally with a mean of zero and a variance of o2.

Socioeconomic and demographic characteristics that contribute to a household’s insurance decision include family size, age, income, region, race, occupation, and whether or not the individual has access to group insurance. Whether or not the person has a high school degree is also included since an individual’s level of education may also be important (see Dewar, 1998). As noted previously, access to group insurance is important for three reasons. First, group insurance is less expensive than individual insurance, due to lower administrative costs. Second, people who may not have been able to purchase insurance in the heavily screened individual insurance market could have more easily purchased insurance through a group. To control for access to group insurance, I include the dummy variable group in the estimation. Group is equal to one if the person had access to group insurance, regardless of whether group insurance was actually purchased.

The marginal tax rate of a household may also have affected the household’s health insurance coverage decision. In effect, the tax subsidy lowered the cost of employer provided health insurance, and this benefit increased with a household’s marginal tax rate. Assuming a full compensating differential, the employee pays for the entire premium through lower wages regardless of the amount of employer contribution. However, while the employee pays the entire premium, he or she enjoys a tax subsidy only on the explicit employer contribution. This tax subsidy offsets the loading factor, thus making the “price” of one dollar of coverage less expensive. The insurance premium faced by the employee is thus equal to

[5] Premium = (1 – t)E + e,

where t is the employee’s marginal tax rate, E is the employer’s contribution to the premium and e is the amount of the premium paid by the employee. As a family’s marginal tax rate rises, employer provided health insurance becomes less expensive, leading to an increase in the amount of coverage purchased. To capture the effect of the tax subsidy on the demand for health insurance, I include a measure of the household’s marginal tax rate (mtr) in the estimated equation. A priori, the estimated coefficient on mtr should be positive, and given the substantial amount of uncertainty surrounding the tax policy of employer contributions to employee health insurance premiums prior to 1954, the effect of the marginal tax rate variable in 1953 should be small relative to its effect in 1958. To test this directly, I also include the dummy variable mtr*year, which is equal to the value of the household’s marginal tax rate in 1958 and zero otherwise. If the tax subsidy did increase the probability that a household purchased insurance, the coefficient on mtr*year should be positive and statistically significant.

An additional effect of the tax changes in 1954 may have been to encourage more employers to not only sponsor group health insurance plans for their employees, but to contribute to (or contribute more to) employee health plans. As shown in table 2, employers more frequently contributed to employee coverage and more frequently paid the entire premium in 1958 compared to 1953. Only seven percent of households in 1953 had an employer who paid their entire health insurance premium. Seventeen percent had partial contributions to their insurance premiums, and 77 percent of people with insurance did not have any employer contributions. In contrast, 18 percent of households in 1958 had their entire premium paid by their employers, 25 percent had partial contributions, and 58 percent paid their entire premium by themselves. After the tax change, more employers contributed to the health care plans suggested by Canadian Health&Care Mall of their employees, and employers who had already been contributing seem to have contributed more.

The ideal variable to gauge the effect of increasing contributions is a measure of increasing contributions. Unfortunately, the NORC samples report employer contributions to employee health plans as “none,” “some,” or “all.” Further, these indicator variables cannot be used in the probit estimation because the presence of employer contributions perfectly predicts insurance purchase. To help capture the effect of the tax changes in increasing the amount of employer contributions, I include the interactive dummy group*year, which equals one in 1958 for a household with access to group insurance. If the estimated coefficient on this variable is statistically significant, it means that households with access to group insurance in 1958 are more likely to purchase insurance than households with access to group insurance in 1953. One reason for this may have been additional employer contributions that occurred after the implementation of the tax subsidy in 1954.

Table 1: Descriptive Statistics, 1953 and 1958

Variable Mean,
1953
Std. Dev, 1953 Min,
1953
Max,
1953
Mean,
1958
Std. Dev, 1958 Min,
1958
Max,
1958
Age 49.240 14.863 19 90 46.292 14.649 18 90
Age2 2645 1536.341 361 8100 2358 1436.505 324 8100
Covg 18.27 35.251 0 197.93 49.09 62.97 0 348.45
Education 0.338 0.473 0 1 0.289 0.453 0 1
Employed 0.810 0.393 0 1 0.861 0.347 0 1
Family Size 3.113 1.678 1 13 3.363 1.742 1 14
(Family Size)2 12.508 14.323 1 169 14.341 16.076 1 196
Female Head 0.193 0.394 0 1 0.117 0.322 0 1
Group 0.478 0.500 0 1 0.667 0.471 0 1
Income 4187 2824 100 50000 5694 4108 37 57000
Insured 0.629 0.483 0 1 0.764 0.424 0 1
Lodge Member 0.197 0.398 0 1 0.255 0.436 0 1
Married 0.763 0.425 0 1 0.821 0.384 0 1
MTR 0.173 0.102 0 0.75 0.179 0.081 0 0.38
Overten 0.061 0.240 0 1 0.103 0.304 0 1
Premium 6.703 4.107 0 35.27 6.688 5.270 0 31
Black 0.103 0.304 0 1 0.086 0.280 0 1
Trunc. Income 4156 2584 100 10000 5282 2720 37 10000
Union Member 0.331 0.471 0 1 0.332 0.471 0 1
Occupation Dummies
Professional 0.150 0.357 0 1 0.112 0.316 0 1
Farmer 0.106 0.308 0 1 0.06 0.237 0 1
Clerical 0.053 0.224 0 1 0.093 0.290 0 1
Sales 0.045 0.207 0 1 0.050 0.219 0 1
Craftsman 0.174 0.379 0 1 0.209 0.407 0 1
Service 0.077 0.266 0 1 0.076 0.265 0 1
Laborer 0.061 0.240 0 1 0.073 0.259 0 1
Operative 0.169 0.375 0 1 0.193 0.394 0 1
Industry Dummies
Agriculture 0.121 0.326 0 1 0.079 0.270 0 1
Mining 0.105 0.307 0 1 0.026 0.159 0 1
Manufacturing 0.252 0.434 0 1 0.315 0.464 0 1
Finance 0.022 0.148 0 1 0.038 0.192 0 1
Service Industry 0.115 0.319 0 1 0.155 0.362 0 1
Public 0.041 0.199 0 1 0.061 0.239 0 1
Trade 0.134 0.341 0 1 0.144 0.351 0 1
Transportation 0.071 0.256 0 1 0.092 0.289 0 1
Regional Dummies
Northeast 0.260 0.439 0 1 0.279 0.449 0 1
Northcentral 0.283 0.450 0 1 0.244 0.43 0 1
Southeast 0.146 0.353 0 1 0.149 0.356 0 1
Southcentral 0.168 0.374 0 1 0.171 0.377 0 1
Mountain 0.043 0.203 0 1 0.035 0.184 0 1
Pacific 0.100 0.300 0 1 0.118 0.322 0 1

Table 2: Percent of Households with Employer Contributions to Health Insurance

Household Reports: 1953 1958
No Contribution by Employer 77% 58%
Partial Contribution by Employer 17% 25%
Full Contribution by Employer 7% 18%

Table 3: Results from Pooled Probit Analysis

Variable Marginal Effect Std.Error P-Value
Family Size 0.0152 0.0173 0.3800
(Family Size)2 -0.0028 0.0019 0.1440
Age 0.0137 0.0038 0.0000
(Age)2 0.0002 0.0000 0.0000
Income 0.0001 0.0000 0.0000
Employed 0.0794 0.0393 0.0420
Union Member 0.1015 0.0257 0.0000
Lodge Member -0.0357 0.0226 0.1130
Black 0.1222 0.0314 0.0000
Professional 0.1681 0.0292 0.0000
Clerical 0.2368 0.0297 0.0000
Sales 0.1130 0.0385 0.0060
Craftsman 0.0862 0.0297 0.0050
Service 0.1276 0.0345 0.0010
Laborer 0.0806 0.0380 0.0400
Operative 0.1746 0.0292 0.0000
Agriculture -0.1250 0.0410 0.0020
Manufacturing 0.2729 0.0236 0.0000
Finance 0.0139 0.0543 0.7990
Service Industry -0.0251 0.0338 0.4560
Public 0.0625 0.0428 0.1550
Trade -0.0281 0.0339 0.4060
Transportation 0.1929 0.0309 0.0000
Education 0.0340 0.0205 0.0980
Northeast 0.0021 0.0447 0.9620
North Central -0.0719 0.0452 0.1100
Southeast -0.0245 0.0475 0.6040
South Central -0.1198 0.0477 0.0110
Pacific -0.1167 0.0501 0.0190
Overten -0.2626 0.0384 0.0000
Union*Year=58 0.1498 0.0360 0.0000
Employed* Y ear=5 8 0.1860 0.0477 0.0000
Year=58 -0.0612 0.0474 0.1980

All Rights Reserved.