On the Determinant of Human Health: An Economic Perspective

3
INTRODUCTION
Since Arrow’s (1963) seminal paper, interest in the economic aspect of health
issues and in economists’ opinion about how to deal with problems related to health and
the health care sector has strongly grown. Arrow’s paper announced the entry of health
economics as a new discipline and became one of the most widely cited articles in the field
of health economics. Today, health is considered one of the most valuable personal and
universal rights:
[The enjoyment of the highest attainable standard of health is one of the funda-
mental rights of every human being without distinction of race, religion, political
belief, economic or social condition.],
as stated in the preamble to the World Health Organization (WHO) Constitution;
the determinants of health and health inequalities have became two issues of vital impor-
tance to health policy in establishing the extent to which the effect of various non-medical
inputs such as income, education, social position, age, ethnicity, health-related behaviors
and environmental quality contribute to modify the individual health status.
Grossman’s (1972) contribution was the first formal economic model of the deter-
minants of health and represented the first relevant theoretical and empirical work after
Arrow’s theoretical approach. Treating health as endogenous was a major difference be-
tween Grossman’s model and the health models that preceded it. Drawing on the theory
of human capital formulated by Becker (1965), Grossman constructed a model where in-
dividuals use medical care and their own time to produce health. Grossman interpreted
a person’s health as a capital stock that exogenously deteriorates at an increasing rate
with age. To counteract this health deterioration, he assumed that individuals invest a
4
portion of their assets into health production each period. Hence, he did not consider
individuals’ health status as totally exogenous but as dependent, at least in part, on the
resources allocated to its production.
In the last decades there has been increasing concerns about the possible adverse
impact of pollution on human health. Public awareness on environmental health issues
and their important economic repercussions has led to a pattern of substantial literature
developmentontheoreticalandempiricalaspectsoftheeconomicvaluationofenvironment-
related health costs with the aim of promoting policies to improve environmental quality
and human health.
The health production approach, first introduced by Grossman, has been success-
fully employed in the evaluation of health pollution related damages: subsequent contri-
butions to Grossman’s paper, by analyzing the decisions consumers make concerning the
resources allocated to health production, tried to infer the value of health to the consumers
and derived and estimated econometrically a measure of individual willingness to pay for
a reduction in pollution (Gerking and Stanley, 1986; Cropper, 1981; Dixie and Gerking,
1991). An important contribution in this area was Cropper (1981). Cropper presented
a simple model of preventive health care, similar to that of Grossman (1972). She takes
changes in environmental conditions to influence the rate at which an individual’s stock of
health depreciates and used her model to define what a person would pay for a change in
air quality.
The health production function approach is one of the several methodologies that
environmental and health economists have developed to measure the value of pollution
5
health related damages. In the first chapter of our work, we critically review these meth-
ods and the research efforts that have been devoted to estimating pollution health re-
lated impacts. First, starting with Weitzman’s (1974) seminal paper, we provide a short
but comprehensive overview of the key literature on the choices faced by policy makers
concerning price-based versus quantity-based instruments to regulate pollution and protect
human health; then, we review the methods employed in estimating pollution abatement
costsandpollutionrelatedhealthdamageswhosecomparison(withreferencetoWeitzman’s
theoretical rule) form the basis for the choice among the price-based and quantity-based
regulation instruments.
Theaimof chapter1istoshedlightontheissueconcerningthedifficultiesthatthe
analysts face in estimating the value of (marginal) pollution health damages or (marginal)
benefits from reduced pollution. In fact, while measuring control costs seems relatively
straightforward (market exists in principle in which pollution control equipment can be
bought, and such equipment will reduce pollution by measurable levels), health damages
(or benefits) are much harder to measure. First, analysts should accurately measure the
health effects of pollution. Once they have determined those effects, they have to put a
monetary value on them. However, valuing health is obviously controversial because each
person may place a different value on it.
In 1977, Lave and Seskin published a pioneering work on the physical relationship
between air pollution and health. Their analysis was conducted on aggregate data. The
big problem with these data was that they did not allow to consider the influence of indi-
vidual’s specific choices (nutrition behavior, sport activities, smoking and alcohol habits,
6
sleeping, housingconditionsetc.) andotherseveralimportantfactors. Moreover, theeffects
onhealthwillalsovaryacrossindividualsduetothesocalled”confoundingfactors”: genet-
ics, avoidance behavior, life-style, socioeconomic status. Hence, analysts have to calculate
associated changes in health outcomes by taking into account that pollution could easily be
correlated with other factors that may be just as influential. Based on this insight, in the
secondchapterwhenanalyzingtherelationshipbetweenairpollutionandhealth, wecontrol
for socioeconomic variables and other important individual characteristics; in particular we
control for life-style variables since individuals specific behavior represents another crucial
determinant of the risk of illness. The analysis is focussed on how individual health habits
and outdoor air quality combine to affectthelikelihoodofagoodorbadhealthstatus,ina
second-best world characterized by uncertainty on the level of health, where an individual
may not able to avoid adverse health shocks completely. The framework is built on the
basic concepts and ideas of the demand for health by Grossman (1972) and the subsequent
contribution by Cropper (1981).
Following Grossman (1972) and Cropper (1981), we construct a model of health
accumulation in which we assume that health depreciates at an increasing rate with age
and ambient air pollution. The main differences with respect to Grossman and Cropper’
models are that the level of health is uncertain and, for individuals who suffer or have
suffered from a pathological condition, illness enters directly the rate of health depreciation
too. As in Cropper’s (1981) model, we assume that when pollution increases it becomes
more costly to reduce the probability of suffering from health shocks. Individuals feel less
healthy because they perceive health depreciation rate to be higher. Hence, they may
7
choose to invest less in their health and maintain lower health stock because of the higher
net investment costs. In this sense, a higher pollution concentration may have two effects
on health: a direct effect which consists of an increase of natural rate of depreciation,
and an indirect effect, described by Cropper (1981), by which individuals will invest less in
healthand display ahigher probability of sufferingfromhealthshocks. Wewill analyzethis
aspect in section 4 of this chapter focusing on the relationship between increasing pollution
and health investment decisions. In addition we will examine whether chronic illnesses, by
alteringtherate to whichhealth capital stock deteriorates, haveany influence on individual
health investment decisions too.
To estimate the health accumulation model and investigate the relation between
health status, pollution, and health investment decisions, we use three different measures
of overall health: dichotomous measures of blood pressure and functional limitations and
disability are employed; moreover we take, as an indicator of health, a self-assessed health
measure that is common in empirical research (Contoyannis and Jones, 2004, Balia and
Jones, 2004 etc.). Since we have included life-style variable as regressors in the health equa-
tionaproblemofsimultaneitymayarise. Hencewetrytocorrectedthepotentialendogene-
ity of the behavioral variables by using a recursive multivariate probit model (Contoyannis
and Jones, 2004; Blaylock and Blisard, 1992).
The model is estimated using data based on the 2001 Behavioral Risk Factor
SurveillanceSystem,whichhoweverdoesnotmeasureenvironmentalquality; environmental
informationatmetropolitanarea-levelisavailablefromEPAandcanbeusedinconjunction
with BRFSS data to compare measures of environmental quality and health. Data are
8
merged at the metropolitan area-level which is available in both the BRFSS and EPA data.
In chapter 2, we concentrate our attentionononeofthemainworldwidesources
of air pollution: motor vehicle emissions. The most important standard concerning motor
vehicles pollution is carbon monoxide. CO air concentration is generally high in areas
with heavy traffic congestion therefore we consider carbon monoxide as a proxy for vehicle
emissions (U.S., EPA 2000)
According to our results, a higher concentration of carbon monoxide has respec-
tively a negative impact on the probability of enjoying good health and a positive influence
on healthy habits. Then, concerning vehicular air pollution our results do not support
Cropper’s (1981) model: people living in polluted areas tend to invest more in health prob-
ably to counteract to the deterioration of a higher depreciation rate due to an increasing
pollution. Arguably, people lead a healthy life-style to increase their health stock and build
up resistance against pollution symptoms and future damages.
Grossman’s model has become a cornerstone in the field of health economics.
The model, however, is not undisputed. A key criticism has been that it fails to take into
accounttheuncertaintyonthefuturehealthstatusandtheuncertaintyonthereturnsfrom
investmentsinhealthproduction. Byinvestinginhealth, individualsdonotdeterminewith
certainty their health status: environment and chance are two factors which may interfere.
Grossman’s model, however, did not account for uncertainty as it included neither explicit
acknowledgment of uncertainty nor the description of illness, even though the fundamental
relationship between health and uncertainty was established by economic theory (Arrow,
1963). Because of uncertainty, much of an individual’s demand for health care is not
9
steady, but irregular and unpredictable. This implies that the costs of health care act as
a random deduction from an individual’s income. Therefore, under uncertainty, risk-averse
individuals demand risk-bearing goods, such as health insurance, to safeguard their income
against possible shocks (Cagatay, 2004).
It is well known, however, that health insurance coverage reducing an individual’s
marginal cost of medical care inputs, leads to use additional medical services: an insured
individual, in fact, may consume more medical services and have a greater expenditure
compared to an uninsured one (moral hazard effect, Leibowitz, 2004). Hence, health insur-
ance biases health production decisions toward over-use of curative medical treatment at
the expense of one’s own preventive efforts.
Insurance choice itself may be affected by planned medical expenditure and ex-
pectations about medical care utilization (adverse selection effect). Chapter 3 focuses in
particular on this last effect. In health insurance market, adverse selection may occur when
consumers’truehealth-costriskisprivateinformation: insurancecompaniesmayknowthat
consumers vary in the level of risk, but, in principle, are not able to discern who are high
and who are low risk profile individuals within a group of potential insured. (Akerlof, 1970;
Rothschild and Stiglitz, 1976). Identifying risks accurately is not an easy task and requires
that insurance company incurs some costs. Insured parties are heterogeneous in terms of
expected costs and have more information about their risks. Naturally, high-risk individu-
als are not encouraged to “reveal” their risk to the insurance company; this asymmetry is
a serious problem since may lead insurance company to face large differences in expected
health costs due to heterogeneity in demographics and the incidence of illness.
10
In spite of the extensive theoretical interest on the adverse selection, there is little
empirical evidence on the extent of the problem. The goal of the third chapter is to test
empirically for adverse selection in the U.S. health insurance market. The test is based
on the 2003/2004 Medical Expenditure Panel Survey — Household Component (MEPS-HC)
data used in conjunction with the previous year’s National Health InterviewSurvey (NHIS)
data. The exercise is conducted by estimating the correlation between the completeness of
insurance an individual buys and his ex-post risk experience, conditional on the observable
characteristics which are used in pricing insurance policies.
Completeness of health insurance plan is measured by health insurance reimburse-
ment that is the difference between total health expenditure and out-of-pocket expenditure
on health care paid by consumers. Health insurance reimbursement, however, is only de-
fined for a subset of individuals from the overall population since we observe it only for
those who participate in insurance and face positive health care expenditure. Thus, the
model may suffer from sample selection bias and straightforward regression analysis may
lead to inconsistent parameters estimate. Another problem that arises from the estimation
is the presence of unobserved heterogeneity in the equations of interest. In most of the
studies which test for adverse selection two important estimation issues such as unobserved
heterogeneity and selection bias, are traditionally treated separately. In our model, we
control for selection bias and at the same time for unobserved heterogeneity issue by using
Wooldridge(1995)two-stepestimationprocedure. Weextendthisestimationmethodtothe
case in which selectivity is due to two sources rather than one (participation in insurance
and participation in health care expenditure).
11
We find no systematic relation between illness of individuals and insurance choice.
We think that a possible explanation can be found in the so called ”cream skimming”
practise: health plans may have an incentive to alter their policy to attract the healthy
and repeal the sick (Newhouse, 1996; Ellis, 1997). Then, insurers may practice a kind of
”reverse adverse selection”: they would try attempt to increase their profits by refusing to
write policies for the worst risks in an insurance pool (see Siegelman, 2004 ). This strategic
behavior can take a variety of forms including: designing insurance benefits packages in
such a way as to be more attractive to healthy persons than unhealthy one for instance
by excluding particular prescription drugs, offering numerous pediatrician ( families with
children are better risks) or by excluding cancer specialist visits. In such cases health plan
may also refuse to sell an applicant insurance altogether. If health plans cream healthy
individuals, those who are enrolled in health insurance are relatively healthy people and
this lead to the failure of the correlation test.