The issue of access to hospital care by uninsured persons has
increased in importance in the United States over the past decade.
Between the late 1970s and 1987, persons without insurance grew
markedly, from 13.0 to 17.5 percent of the nonelderly population. By
1987, the number of uninsured Americans had reached 37 million (Swartz
1989). Currently, a significant fraction of the population continues to
be uninsured (Families USA Foundation 1993). At the same time, the
ability of hospitals to provide uncompensated care has been impaired by
the spread of cost containment and other procompetitive forces.
Concern for the uninsured and access to medical care has led to the
formulation of a number of policy options including the expansion of
public health insurance coverage; the. advancement of private insurance
alternatives such as risk pools or employer-mandated insurance; and
state regulation of hospital payment to finance care for the uninsured.
While the first two options increase coverage and target individuals,
the latter approach uses direct payments to hospitals and targets
In 1980, as part of a wider reform of hospital payment, the state
of New Jersey chose payment regulation as a way to promote care for the
uninsured and assist hospitals who provide this care. Under this
provision, payment rates for all payers, public and private, were set to
include the costs of the uncompensated care provided by a hospital. In
this way, patients who paid wholly subsidized the care of those who did
In this article, we evaluate three effects of New Jersey's
uncompensated care provision: (1) Has the level of uncompensated care
changed in response to the payment reform? (2) Has the provision of
uncompensated care been shared to a greater extent across hospitals in
the state? and (3) Are the hospitals that provide a disproportionately
larger share of uncompensated care better off financially since
implementation of the reform? The answers to these questions provide
useful insights into the effects of New Jersey's approach on access
to hospital care by the uninsured.
We begin by providing background on uncompensated care and hospital
payment in New Jersey. We then review evidence from previous studies of
the impacts of uncompensated care payment in the state. Next, we
describe a conceptual model and discuss the data used in the study.
Finally, we present our results and offer some general conclusions.
UNCOMPENSATED HOSPITAL CARE AND THE UNINSURED IN NEW JERSEY
Uncompensated hospital care can be separated into two components:
charity and bad debt. Charity care includes services to persons who are
"medically indigent" at the time of admission and not expected
to pay their hospital bills. Bad debt is that accrued by persons who are
expected to pay but do not. It is estimated that uninsured patients
account for about 75 percent of the uncompensated care provided by
hospitals (Coye 1989).
In 1986, about 11 percent or 843,000 of New Jersey's 7.6
million residents were without health insurance. This compares with an
estimate of 15 percent nationwide (State of New Jersey Department of
Health 1989). As is the case nationally, New Jersey's uninsured are
young: more than half are below age 25, and about one-fourth are
children under age 18. Two-thirds of the state's uninsured are
employed, or are in a family with at least one working adult. Following
national trends, the number of persons without health insurance in the
state rose over the past two decades (State of New Jersey Department of
1980 REFORM OF HOSPITAL PAYMENT IN NEW JERSEY
In 1980, in response to rising hospital costs, the deteriorating
financial condition of inner-city hospitals, and a growing volume of
uncompensated care, New Jersey undertook a major reform of hospital
payment. The new system, based on prospective diagnostic-related group
(DRG)-based per case payment, replaced a prospective per diem payment
regulation known as the Standard Hospital Accounting and Rate Evaluation
(SHARE) program. SHARE had covered Blue Cross and Medicaid patients,
while the new payment system covered all payers. Hospitals were phased
onto the system over a three-year period beginning in 1980; 26, 36, and
35 hospitals entered the system in 1980, 1981, and 1982. The SHARE and
the 1980 DRG-based payment systems are compared in Figure 1.(1)
A key element of the payment reform was the introduction of a
provision to fund hospitals' charity and bad debt care by setting
payment rates to explicitly include the cost of this care. This system
replaced one where the costs of uncompensated care were either not
covered, financed through philanthropy or government grants, or
reimbursed through increased charges to commercial insurers and self-pay
patients. Between 1980 and 1986, institutions in the state received a
hospital-specific markup to their payment rates to cover the reasonable
costs, verified by audit, of their own uncompensated care. These markups
ranged from 1 to 25 percent across hospitals, varying with the amount of
uncompensated care provided by individual institutions (Rosko 1990). In
1987, to assist hospitals with disproportionately high numbers of
uninsured and low-income patients, a statewide markup of payment rates
UNCOMPENSATED CARE PAYMENT IN NEW JERSEY: PREVIOUS STUDIES
Two studies have focused on how uncompensated hospital care payment
in New Jersey has affected the level of this care provided. Using
inpatient data for the years 1979 to 1982, Mann, Melnick, and Kominski
(1989) found that the number and proportion of self-pay admissions
increased following the introduction of uncompensated care payment.
Their results indicated that the concentration of self-pay admissions
across hospitals also diminished. Since they did not control for other
factors, it is unclear whether the payment program, changes in the
hospitals' market conditions, or the underlying distribution of the
uninsured population, or some combination, was responsible for their
Rosko used hospital level data for the years 1979 to 1985 and
multivariate regression to examine the payment program's effects on
inpatient discharges (Rosko 1990). Controlling for unemployment,
population, recipients of Aid to Families with Dependent Children
(AFDC), and teaching status, he found that payment for uncompensated
care increased the number of self-pay discharges in the state. Rosko
also concluded that the payment program promoted the financial status of
hospitals that typically provided a disproportionately high level of
uncompensated care. He did not examine changes in the distribution of
this care across hospitals in the state.
METHODS AND DATA
Our investigation builds on this previous research. We extend the
period of study beyond the initial years of the payment program and
include outpatient hospital care in our analysis. Further, we model more
fully the factors that influence the overall amount of uncompensated
care provided by hospitals in the state. We also provide a more detailed
analysis of the distribution of uncompensated care across hospitals.
We analyze uncompensated care in New Jersey over the period 1979 to
1987. We use descriptive statistics to summarize trends in important
variables as well as a multivariate regression analysis to isolate
effects. To evaluate changes in the distribution of this care across
hospitals, we employ a Herfindahl index, a measure of concentration.
Finally, to gauge hospitals' financial status before and after the
program, we examine their surplus of revenues over expenses.
As the measure of uncompensated care, we use total inpatient and
outpatient hospital charges attributed to charity and bad debt
care.(3,4) This variable incorporates two components: the number of
episodes (inpatient admissions or outpatient visits) and the service
volume per episode, as measured by charges. This allows us to test for
the effect of factors on not only the level, but also the intensity, of
uncompensated care services. This approach contrasts with other studies
that have employed inpatient discharges or admissions as the dependent
We measured the effect of uncompensated care payment on services
using the following reduced-form model:
Uncompensated [Care.sub.it] = f ([NJREG.sub.it], [DEMVARS.sup.it],
[SUPPVARS.sup.it]) where i indexes hospitals and t indexes time. NJREG
is a binary variable measuring when a hospital entered the regulated
system and began receiving uncompensated care payments. For each
hospital, NJREG is equal to 0 prior to the year of entry (1980, 1981, or
1982) and 1 for all subsequent years. DEMVARS and SUPPVARS describe
vectors of variables reflecting the supply and demand factors
influencing the amount of uncompensated care provided. Definitions and
descriptive statistics for the variables included in the model are given
in Table 1. We discuss these variables and their expected signs in the
[TABULAR DATA 1 OMITTED]
We would expect payment for uncompensated care to affect both the
supply and demand for this care. For hospitals, these payments greatly
increase expected revenues from providing care to the uninsured. Prior
to the new system, hospitals financed indigent care, when possible,
through philanthropy, grants, or increased charges to other patients.
Further, prior to 1980, the patient base to which hospitals could
increase charges was limited by the SHARE regulation, which set
prospective payment rates for Blue Cross and Medicaid patients. The new
system covered all payers and built the full costs of uncompensated care
into each hospital's rates.
For patients, as their perception of hospitals' willingness to
provide free care increases, they will expect to be able to pay less. As
a result, they may increasingly seek this care.
Factors other than uncompensated care payment are likely to affect
the amount of this care provided. In addition to expected price, the
demand for this care is a function of the number of uninsured or
underinsured persons in a hospital's area. Since there are no
detailed measures of the health insurance coverage for a particular area
over time, we approximate demand using several variables: the level of
government funding for public health insurance (e.g. Medicaid), as
measured by the population receiving AFDC and general assistance
payments; the share of the labor force employed in sectors historically
providing different levels of insurance coverage (low coverage, high
coverage, and unemployed); per capita income; the area's elderly
population; and the hospital's location (inner-city, urban,
suburban, or rural). We would expect AFDC coverage, employment in
high-coverage sectors, and real per capita income to decrease demand for
uncompensated care. This demand is also expected to fall with an
increased proportion of elderly persons, through their eligibility for
Medicare coverage. Hospitals' location may further describe the
characteristics of the area population.
Studies have found that hospital ownership, teaching status, and
size are important in explaining the level of uncompensated care
provided by an institution (Thorpe and Spencer 1991, Sloan, Morrisey,
and Valvona 1988, and Feder, Hadley, and Mullner 1984). We model these
supply attributes using hospital ownership (private, public, or owned by
a religious organization); teaching status (major, minor, or
nonteaching); and hospital bed size. Historically, public-owned and
teaching hospitals have provided a disproportionate share of the
uncompensated care in an area, while hospitals owned by religious
organizations may also be more likely to provide such care.
The presence of particular types of institutions in a
hospital's area may also affect the uncompensated care provided
(Thorpe and Brecher 1987). We include the share of beds in a
hospital's area in public-owned and teaching hospitals. We expect
the supply of free care provided by a hospital to decline with the
increased presence of these alternative sources.
Finally, the characteristics of physicians in the hospital's
area may influence the amount of uncompensated hospital care provided.
We model these effects using physician supply and primary care mix. As
the physician supply increases, we expect that the overall demand for
hospital care, including uncompensated care, will increase. Since
services provided by primary care physicians may substitute for hospital
outpatient care, we expect the supply of uncompensated care to decrease
with the increased presence of these physicians.
We use data from a sample of New Jersey hospitals for the years
1979 through 1987 to complete our analysis. This sample includes 80 of
the 88 short-term, acute care hospitals in the state during this time.
We excluded the remaining hospitals due to incomplete data. All 88
hospitals considered were nonprofit institutions. An examination of the
key variables for the omitted hospitals suggests that the results were
not sensitive to their exclusion. To remove the effects of general price
inflation, all financial variables were adjusted to 1987 dollars using
the general consumer price index (U.S. Department of Labor 1980-1988).
The pooled cross-section-time series nature of the data
necessitates that we examine different model specifications for our
multivariate analysis. We evaluated three approaches: (1) ordinary least
squares (OLS) estimation without fixed effects; (2) OLS with fixed
effects for hospitals and years (Kmenta 1971); and (3) a random effects
model estimated using generalized least squares (Wallace and Hussain
1969). We use Hausman's test to select the proper specification
Uncompensated care costs in New Jersey during the 1980s were
substantial, totaling more than $282 million in 1984, or about 7 percent
of total gross hospital revenue (Coye 1989). Table 2 shows the mean
levels of uncompensated care and other important variables for the years
1979 through 1987. As shown, total uncompensated care has increased
significantly during this period, growing at an annual compound rate of
10 percent in constant dollars. Both inpatient and outpatient
uncompensated care grew, at similar rates. Table 2 also shows that this
care increased as a percentage of total revenues.
[TABULAR DATA 2 OMITTED]
Variables likely to affect the amount of uncompensated care include
public insurance coverage and the sectoral composition of the labor
force. Table 2 shows that participation in programs related to public
insurance, as measured by those receiving AFDC or general assistance
payments, decreased during the study period. Further, the share of the
labor force employed in sectors historically providing lower levels of
health insurance coverage increased, while the share in high-coverage
sectors decreased. These trends would be expected to increase demand for
Table 2 also presents the Herfindahl index of total uncompensated
care charges for hospitals in the state for the years 1979 through 1987.
The index is a measure of the evenness of a distribution; the lower the
index, the more even the distribution. As shown, the distribution of
uncompensated care became markedly less concentrated. The Herfindahl
index decreased from 0.071 in 1979 to 0.028 in 1987. In comparison, the
same index for gross charges, compensated and uncompensated, grew
slightly from 0.016 to 0.017 over the same period.
Table 3 summarizes changes in uncompensated care over the study
period by hospital ownership, teaching status, location, and bed size.
With the exception of public-owned hospitals, the level of uncompensated
care increased for all hospital groups. Further, this care increased at
a greater rate after 1981.(5) The results for public-owned versus other
hospitals and teaching versus nonteaching hospitals suggest that the
volume of uncompensated care was redistributed between hospitals during
[TABULAR DATA 3 OMITTED]
Table 4 presents the results of the multivariate regression
analysis. In estimating our models, we corrected for heteroscedasticity
by taking natural logarithms of all continuous independent and dependent
variables.(6) This approach also allowed the estimated coefficients to
be interpreted as constant elasticities. Based on Hausman's
specification test, we selected the fixed effects model, using a
separate binary variable for each hospital, as the most appropriate
specification. An analysis evaluating the contribution of fixed-effects
for time in the model indicated that they were not necessary.(7)
Table 4 (Model I) shows that the implementation of the 1980 payment
reform had little effect on the overall amount of uncompensated care.
The coefficient on NJREG is small and insignificant. Hospitals did not
provide more uncompensated care in the aggregate than would have been
predicted without the payment regulation.
The estimates of the remaining coefficients in Model I show some
interesting results. Table 2 indicated a shift in the labor force over
the study period to sectors with historically low levels of medical
insurance. The negative coefficients for HIGHCOV and UNEMRT in Table 4
suggest that this shift is associated with increased uncompensated
care.(8) In other words, employment in low-coverage industries has a
more positive effect on this care than employment in high-coverage
industries or unemployment. The finding for HIGHCOV is as expected. The
result for UNEMRT may be due to the availability of public insurance
programs for the unemployed poor or to COBRA (Consolidated Omnibus
Budget Reconciliation Act) provisions that often allow workers to retain
health insurance benefits for a limited period after becoming
unemployed. These coefficients support our model: the amount of
uncompensated care provided by New Jersey hospitals is sensitive to
demand factors in the local economy.
The negative coefficient for the share of public hospital beds in
an area (PUBSHR) is as expected; the increased presence of public
institutions decreases uncompensated care in other area hospitals. The
coefficient on the minor teaching hospital share variable (MINSHR)
suggests a similar effect. However, the coefficient for the share of
major teaching hospital beds (MAJSHR) is opposite from that expected,
and is insignificant. Finally, the coefficients on AFDC and general
assistance coverage (AFDCGAP) are of the expected sign (negative) but
We used the multivariate model to explore further any
redistribution of uncompensated care across hospital groups. To do this,
we included variables in our model that represent the relative
postregulation experience of hospitals by ownership and teaching status.
These results are presented as Model II in Table 4. The variables
PUBLIC*REG, CHURCH* REG, MAJOR*REG, and MINOR*REG are the interaction
between the variable NJREG and the hospital ownership and teaching
status variables. These variables are measured relative to private,
non-religious-owned, non-teaching hospitals (the reference group), which
are represented by the result for the variable NJREG in Model II. A
positive coefficient for these variables suggests a greater response to
the regulation than the reference hospitals. As shown, public-owned and
minor teaching hospitals experienced a relative decrease in
uncompensated care following the regulation, while hospitals owned by
religious organizations experienced a relative increase. Major teaching
hospitals, although maintaining their high level of uncompensated care,
showed no significant relative change.(9)
ANALYSIS OF HOSPITAL FINANCIAL CONDITION
We assessed the effects of uncompensated care payment on
hospitals' financial condition using the margin of total revenues
over total expenses. We first divided the hospitals into groups based on
the percentage of their total revenue composed of uncompensated care in
1979. We selected three groups: light uncompensated care burden (< 4%
of revenues); medium burden (4-8% of revenues) and heavy burden (> 8%
of revenues). We identified 25, 30, and 25 hospitals in the three
Figure 2 shows that uncompensated care payment had a positive
effect on the financial condition of hospitals providing a
disproportionately larger share of this care. The surplus margins of
these hospitals increased following the regulation and were similar to
those of other hospitals after 1983.
Our analysis provides answers to the three questions we raised.
First, although uncompensated care in New Jersey increased following the
introduction of payment for this care, our multivariate analysis
suggests that the increase cannot be attributed to the payment
provision. Other trends in the state,
including changes in the composition of the labor force and in public
insurance programs, appear to have added to the number of uninsured
persons and resulted in increased demand for free care. Given these
trends, hospitals collectively have not provided more care than that
predicted without the payment regulation.
Second, the adoption of payment for uncompensated care has resulted
in a more even distribution of this care across hospitals in New Jersey.
In particular, the burden of uncompensated care has been shifted from
publicly owned and teaching hospitals to other institutions in the
state. This result suggests that some barriers to care existed at
certain types of hospitals prior to the regulation.
Finally, uncompensated care payment has improved the financial
condition of hospitals providing the largest share of this care. Surplus
margins for the most heavily burdened hospitals became more like those
of other hospitals following the regulation.
The state of New Jersey cited two objectives in adopting payment
for uncompensated hospital care: (1) to promote access to hospital care
for uninsured persons and (2) to ensure the financial viability of
hospitals providing this care. We conclude that the program has met
these objectives. Our findings suggest that, although hospitals did not
provide collectively more free care, opportunities for the uninsured to
receive such care were more widely available than before the regulation.
Further, the financial stability of hospitals providing the bulk of this
care improved, ensuring that patients would have continued access to
State regulation of hospital payment to finance care for the
uninsured has had some success. Our conclusions support those of
previous evaluations of the New Jersey payment program. Further, Thorpe
and Spencer (1991), in analyzing a similar program in New York state,
concluded that it assisted hospitals financially and resulted in an
increased volume of uncompensated care. These programs appear to have
improved access to hospital care by the uninsured.
Despite this success, programs like New Jersey's must be
compared with alternative ways of caring for the uninsured. Compared to
insurance, direct payments to hospitals do not permit patients to choose
care in the most effective and efficient manner. Inefficiencies result
as the uninsured forgo primary and preventive services and seek care
only when a medical condition becomes acute and more expensive. Further,
by focusing solely on hospital services, these programs provide patients
with incentives to initiate care through hospital emergency rooms or
outpatient units--expensive alternatives to a physician's office.
As others have argued (Thorpe and Spencer 1991), expansion of insurance
coverage appears to be a more effective way of providing medical care
for the uninsured.
The authors thank William Hsiao, Eric Latimer, Judy Dernburg, and
Kathleen Adams for their helpful comments on earlier drafts of this
article. We also acknowledge the comments of two anonymous reviewers.
(1.) Several studies have examined the impacts of the New Jersey
SHARE and all-payer DRG payment regulations on overall hospital costs
and utilization (Hsiao et al. 1986, Hsiao and Dunn 1987, Rosko and
Broyles 1986). This article focuses solely on the effects of the
uncompensated care provision of the 1980 payment reform. (2.) Two events
related to uncompensated care payment in New Jersey occurred after 1987.
First, Medicare stopped contributing payments to hospitals for
outpatient and inpatient uncompensated care in 1988 and 1989,
respectively. Second, for the first half of 1991 (January to June), the
payment method for uncompensated care shifted back to hospital-specific
markup rates. In 1992, the statewide markup was 19 percent. The
uncompensated care provision was eliminated in January 1993 (Frankel
1993). These changes are outside the scope of this article, but
highlight the potential political and financial instability of this
approach to dealing with hospital care for the uninsured. (3.) Discounts
on charges were not considered as part of uncompensated care for the New
Jersey regulation. As a result, only charity and bad debt care were
included in this analysis. (4.) It can be argued that hospitals have an
incentive under the provision to more "accurately" or
"inaccurately" code cases as charity or bad debt care, or to
let up on collection procedures. In order to curb this incentive and
standardize the coding process, New Jersey adopted a definition for
charity care and a collection procedure for bad debt (how many times a
hospital must try to collect a bill over a period of time). If hospitals
do not follow this procedure, they cannot claim the unpaid bill as bad
debt. There is no evidence that hospitals' coding of charity cases
or efforts to collect bills have changed, or that these factors have
influenced the level of uncompensated care reported. (5.) Since
hospitals were phased onto the new payment system over a three-year
period beginning in 1980, the year 1981 was chosen as marking the
"before" and "after" period for the analysis shown
in Table 3. (6.) The OLS results are available from the authors. (7.)
Subsequent use of the White test failed to reject the null hypothesis of
homoscedasticity (White 1980). (8.) Since the logarithms of the share of
the labor force in low- and high-coverage sectors, and unemployed, sum
to some relatively constant amount (the unlogged values sum to 100
percent for an area), we omitted the variable LOWCOV from the
regression. The parameters for HIGHCOV and UNEMRT can be interpreted as
the effects of these variables on uncompensated care relative to that of
LOWCOV. (9.) We also considered a potential caveat to our findings using
the multivariate regression model. In addition to payment for
uncompensated care, the 1980 reform of hospital payments in New Jersey
included other provisions, such as the introduction of prospective
all-payer DRG-based payment. It can be argued that if the new payments
provide a strict financial constraint to hospitals, the level of
uncompensated and other care provided could be affected. As a result,
the response to the new payment rates will be confounded with that from
the introduction of uncompensated care payment.
We examined the potential effect of fiscal pressure on the
hospitals' postregulation response using our multivariate
regression model. For each of the years following the introduction of
the regulation, we included in regression Model I a hospital's
previous year's surplus margin. This approach assumes that the
hospitals make decisions about the provision of uncompensated care based
on the previous year's financial experience. The estimated
coefficient for this variable can be interpreted as the effect of the
interaction between the financial constraint and the new payment
regulation on the provision of uncompensated care. We found the
coefficient for this variable to be .001 with a t-statistic of .50. The
sign, significance, and magnitude of the other variables in the model,
including NJREG, were not affected. We conclude that our results are not
confounded by this potential effect of the new payment regulation.
Coye, M. J. The Uncompensated Care Trust Fund: Assuring Universal
Access to Hospital Care in New Jersey. Trenton, NJ: State of New Jersey
Department of Health, November 1989. Families USA Foundation. "Half
of Us: Families Priced Out of Health Protection." Washington, DC:
Families USA Foundation, April 1993. Feder, J., J. Hadley, and R.
Mullner. "Falling Through the Cracks: Poverty, Insurance Coverage,
and Hospital Care for the Poor, 1980 and 1982." Milbank Memorial
Fund Quarterly 62, no. 4 (1984): 544-66. Frankel, D. H. "USA:
Health Care Reform in New Jersey." Lancet 269 (1993): 153. Hausman,
J. A. "Specification Tests in Econometrics." Econometrica 46,
no. 6 (November 1978): 1251-71. Hsiao, W. C., and D. L. Dunn. "The
Impact of DRG Payment on New Jersey Hospitals." Inquiry 24, no. 3
(1987): 203-11. Hsiao, W. C., H. M. Sapolsky, D. L. Dunn, and S. C.
Weiner. "Lessons of the New Jersey Payment System." Health
Affairs 5, no. 2 (1986): 32-45. Kmenta, J. Elements of Econometrics. New
York: McMillan Publishing Co., Inc. 1971. Mann, J. M., G. A. Melnick,
and G. F. Kominski. Uncompensated Care Payments to Hospitals: Does
Access for the Uninsured Improve? Santa Monica, CA: RAND Corporation,
November 1989. Morrisey, M. A., F. Sloan, and J. Valvona. "Defining
Geographic Markets for Hospital Care." Law and Contemporary
Problems 314 (February 1986): 522-57. Rosko, M. D., and R. W. Broyles.
"The Impact of the New Jersey All-Payer DRG System." Inquiry
23, no. 1 (1986): 67-75. Rosko, M. D. "All-Payer Rate-Setting and
the Provision of Hospital Care to the Uninsured: The New Jersey
Experience." Journal of Health Politics, Polity and Law 15, no. 4
(Winter 1990): 815-31. Sales and Marketing Management. Survey of Buying
Power. New York: Bill Publications, 1980-1988. Sloan, F. A., M. A.
Morrisey, and J. Valvona. "Hospital Care for the |Self-Pay'
Patients." Journal of Health Politics, Polity and Law 13 (Spring
1988): 83-102. State of New Jersey Department of Health. Access to
Health Care for the Uninsured in New Jersey: Issues and Answers.
Trenton: New Jersey Department of Health, 1989. Swartz, K. Strategies
for Assisting the Medically Uninsured. Washington, DC: The Urban
Institute, August 1989. Thorpe, K. E., and C. Brecher. "Improved
Access to Care for the Uninsured Poor in Large Cities: Do Public
Hospitals Make a Difference?" Journal of Health Politics, Policy
and Law 12, no. 2 (1987): 313-24. Thorpe, K. E., and C. Spencer.
"How Do Uncompensated Care Pools Affect the Level and Type of Care?
Results from New York State." Journal of Health Politics, Policy
and Law 16, no. 2 (Summer 1991): 363-81. U.S. Department of Labor,
Bureau of Labor Statistics. Handbook of Labor Statistics. Washington,
DC: U.S. Government Printing Office, 1980-1988. U.S. Department of
Labor, Bureau of Labor Statistics. Employment and Earnings.
Establishment Data: State and Area Employment. Washington, DC:
Government Printing Office, 1980-1988. U.S. General Accounting Office.
Health Insurance: An Overview of the Working Uninsured. Report No.
GAO/HRD-89-45. Washington, DC: GAO, February 1989. Wallace, T. D., and
A. Hussain. "The Use of Error Components Model in Combining Cross
Section with Time Series Data." Econometrica 37 (January 1969):
55-72. White, H. "A Heteroscedasticity Consistent Covariance Matrix
and a Direct Test for Heteroscedasticity." Econometrica 48 (May
This research was supported by a grant from the Robert Wood Johnson
Foundation. Address correspondence and requests for reprints to Daniel
L. Dunn, Ph.D., Research Economist, Department of Health Policy and
Management, Harvard School of Public Health, 1350 Massachusetts Avenue,
Cambridge, MA 02138. Michael Chen, Ph.D. is Associate Professor,
Department of Social Welfare, National Chung-Cheng University, Chia-yi,
Taiwan. This article, submitted to Health Services Research on November
30, 1992, was revised and accepted for publication on August 26, 1993.
Table 4: Multivariate Regression Results
Dependent Variable. UC Model I Model II
Constant 11.017(**) 12.772(**)
NJREG 0.015 (0.038
AFDCGAP -0.085 (0.109
HIGHCOV -3.417(*) -3.559(*)
UNEMRT -0.203(*) -0.211(*)
RCAPINC 0.550(**) 0.483(**)
POP65 2.652(*) 2.435(*)
BEDS 0.507(*) 0.453(*)
PUBSHR -1.118(**) -1.043(**)
MAJSHR 0.416 0.469
MINSHR -0.921(*) -0.902(*)
SPMD -0.169 -0.204
PHYSPOP -0.351 -0.351
No. observations 720 720
Adjusted [R.sup.2] 0.923 0.925
Note: All continuous variables were transformed using natural logarithms. Due to
effects specification of the final models, the binary variables for hospital cha
(CHURCH, MAJOR, MINOR, INNER, URBAN, SUBUR, and RURAL) were omitted. Results
for these variables for the OLS specification are available from the authors.
(*) p [less than or equal to] .01.
(**) p [less than or equal to] .05.
(***) p [less than or equal to].10.
([dagger] Standard errors of parameter estimates shown in parentheses.