Sot. Sci. Med. Vol. 25, No. 8,pp.883-888, 1987 0277-9536/87 %3.00+0.00 Printed in Great Britain. All rights reserved Copyright 0 1987 Pergamon Journals Ltd zyxwvutsrq HEALTH STATUS AND MEDICAL EXPENDITURES: MORE EVIDENCE OF A LINK BARBARA WOLFE’ and MARY GABAY~ ‘Department of Economics and Preventive Medicine, University of Wisconsin-Madison, 1180 Observatory Drive, Madison, WI 53706 and 2Abt Associates, Washington, DC 20008, U.S.A. Abstract-The hypothesis that one must include life-style changes in order to accurately capture the true relationship between medical expenditures and health status is explored, using data from 22 countries over a 20-year period. A simultaneous model is estimated using a variety of indicators for life style as well as health status. Changes in life style, aging of the population and changes in occupational risk are modeled as influences on medical expenditures: medical expenditures and changes in life style are modeled as having direct influences on health status. The results are consistent with the existence of a positive link between medical expenditures and health status. Key words-health status, medical expenditures, life style, cross-country comparison, aging of population With few exceptions recent studies of comparative health expenditures have not found a demonstrated relationship between medical expenditures and health status (see for example [l] and [2]). An exception is the 1986 article “Health Status and I:edical Ex- penditures: Is There a Link?” by Barbara Wolfe [3]. In that study a positive relationship was suggested between health status as measured by life expectancy and infant mortality and health expenditures after taking into account changes in life style that have an impact on health. That study used data from only six countries and a deductive analytical approach. This study pursues the same underlying model but is expanded to include data from 22 countries and uses more sophisticated statistical techniques. The general lack of a relationship between medical expenditures and health is a part of the perception held by some that medical care expenditures are a luxury good with little marginal effect on health. A corollary of this view is the argument that cutbacks in health care expenditures are possible with little loss in effectiveness. The various explanations that are offered for the lack of a positive relationship between expenditures and health, include the view that health care ex- penditures are not highly related to health care needs but instead to variations in per capita income [4]; and the view that when there are more physicians they treat fewer patients, so, in order to augment their incomes they tend to provide some medical care whose expected benefit does not exceed expected cost but contributes to physicians’ income. (See [5] for the basic statement of this theory termed supplier- induced demand.) This lower effectiveness of mar- ginal health care spending is consistent with the lack of a relationship between numbers of physicians per 1000 in the population and health status or between expenditures on medical care linked to physicians and health status. The basic thesis of this paper as well as the earlier (1986) Wolfe paper is that changes in life style lead to changes in medical care utilization. These life-style changes also may directly influence health status. In order to understand the link between medical care expenditures and health one must take into account this role of life-style changes. Since a substantial proportion of these life-style changes are negative (i.e. lead to declining health and increasing medical ex- penditures), not taking this link into account is expected to ‘falsely’ suggest the lack of a positive relationship between expenditures and health. Figure 1 compares the two approaches and suggests a vari- ant which also includes the role of life-style changes in influencing medical expenditures, but omits the direct link between life style and health. THE MODEL At the simplest level, we posit a single equation model where health status (H) depends on medical expenditures (M) and life style (L). H = H(M, L). (1) However, we suggest that medical expenditures are partly determined by life style. Therefore a simulta- neous model seems more appropriate H = H(M, L) (2) M = M(L). (3) The discussion in the literature is really not that medical expenditures have no positive effect on health status but that more medical expenditures have no further effect on health status. That is, the discussion refers to the lack of a positive marginal effect. Therefore we restate. the model in change format: dH = dH(dM, dL) (4) dM = dM(dL) (5) A difficulty arises however in that health status and life style are not observed or are imperfectly ob- served. Therefore in order to accommodate these measurement problems, we formulate the problem as a LISREL (Linear Structural Relations, see [6] and the references therein) system for the simultaneous system in (4) and (5). This approach allows us to obtain estimates of these relationships within a 883 S.S.M. 25,&B