Summarize the attached article

Applied Economics, 2002, 34, 59762 Rm Estimating a health production function for the US: some new evidence JAMES THORNTON Dcpartmcnl of El'amlillil'x, Ear/cm Michigan University, Yprilnnli, Ml 48197 , USA E-mail: eco_lli0rm0l1(ui ()nlinr'.llm'li.edu This study reports some new evidence on the impact of medical care. socioeconomict lifestyle and environmental factors on the health status of the population of the USA. The results show that additional medical care utilization is relatively ineffective in lowering mortality and increasing life expectancy. The most important factors that inuence death tat ' are related to socioeconomic status and lifestyle. The results suggest that health care policy which focuses primarily on the provision of medical care services and ignores larger economic and social considerations may do little to benet the nation's health. I. INTRODUCTION Over the past several decades, medical care expenditures in the US have risen rapidly, increasing from $26.9 billion in 1960 to $1.03 trillion in 1996. During the same period of time, the share of gross domestic product (GDP) allocated to medical care increased from 5.1 to 13.6% (Levit er al., 1998). This trend is expected to continue in the future. The Health Care Financing Administration projects that spend- ing on medical care will rise to $2.1 trillion by the year 2007 and account for 16.6% of the GDP (Smith c1111.. 1999). An important policy question, which is currently the subject of much debate, is whether medical care should be allowed to consume an increasing portion of goods and services pro. duced in the US? If a goal of public policy is to maintain and improve the nation's health status, then knowledge of the contribution of additional medical care utilization to the health of the population can serve as valuable input in illuminating this debate and inrorming policy decisions. Health economists have long been interested in the impact of medical care and other factors on health out- comes. To investigate these relationships, empirical studies have adopted a health production function analytical framework. where health is viewed as an output that is produced by a set of inputs. A relatively large number of studies have examined the marginal contribution of selected environmental, socioeconomic, behavioural, and medical inputs on various measures of health outcomes using the individual as the unit of analysis (see. for ex- ample, Newhouse and Friedlander, 1980: Rosen and Taubman, 1982; Taubman and Rosen, 1982: Leigh. 1983; Berger and Leigh, 1989: Kenkel, 1991), Only a few studies have attempted to estimate an aggres gate. multifactor health production function for the US. A major advantage of estimating an aggregate health produc- tion function is that an estimate of the over-all effect of medical care utilization on the health status of the popula- tion can be obtained. This information can help shed light on the benefit the nation as a whole is getting from increased spending on medical care services and indicate whether investments in alternative health programmes may have a larger return than medical care. Prior empirical studies of the aggregate health produc tion function for the elderly and non-elderly population of the US by Auster cl til. (1969), Silver (1972) and Hadley (1982) use data for the period 1972 or earlier, and nd that the marginal contribution of medical care in reducing mor- tality is relatively small, with a 1% increase in medical care expenditures resulting in an approximate 0.10 to 0.15% decrease in the death rate. These studies suggest that socio- economic status and lifestyle are important factors inuen- cing mortality. but nd both positive and negative associations between aggregate measures of income and health. While these studies are historically important, many changes have taken place in the health care system in the US. and it is possible that the nature of the aggregate mortality production function has also changed. The pur- pose of this study is to update earlier research on the aggre- A/ililwrl Er'ivinlim \\ ISSN omsenlm print ISSN 1466742X3 olillne t, 2002 Taylor 64 Francis Ltd 59 hup, nun r df.co.uk Journals DUI l0 Inll GOOSRM'KlImZSoS D 60 gate health production function for the US employing data tor the year 1990. The basic approach follows the seminal study by Auster at al. (1969) and uses the state as the unit of observation, Several modications are made to general- ize and improve the anal pecically, additional health related factors are incorporated in the model. both medical care and income are treated as endogenous variables. and the non-white population is incltided in the analysis. II. VARIABLE SPECIFICATION AND MEASUREMENT This section discu s the variables included in the health production function and their measurement. limit/l .r/ams The dependent variable in the health production function is the health status of the population measured by the age- adjusted death rate. Medical care Following prior studies, a measure of medical care expen- ditures per capita is employed as a proxy for the quantity of medical care services. Hadley (1982) argues that this measure is more desirable than using stocks of providers. because variations in expenditures across geographic regions better reect differences in the quantity and quality of services. States with higher levels ofmedical care use are expected to have lower death rates, all else the same. Srlcicaminniit- sin/us Education and income are included to represent socioeco- nomic status. Education is measured as the percent of the population 25 years of age or older that has schooling beyond a high school degree States with more educated populations are expected to have lower death rates, all else the same. Grossman (1972) and other economists hypothe- size that higher levels of education increase the efciency with which individuals are able to produce health by increasing health knowledge and improving the ability to process health information. This reasoning implies a direct causal elfeet of education that works independent of income and unobservable characteristics that may be cor- related with both health and years of schooling. Berger and Leigh (1989), Rosen and Taubman (1982) and others have provided empirical evidence in support of this hypothesis. Income is measured as personal income per Capltil. The expected relationship between income and mortality is ambiguous. Higher levels of income can nance improve- ments in health that come with higher living standards: better sanitation and nutrition are often cited examples J. Thornton However. beyond some threshold level ofaluence increas ing income may no longer buy reduced mortality, and may well lead to the sort ofstressful and unhealthy lifestyle that could adversely effect the health status of the population. (Fuchs, 19941Auster (-1 (1]., 1969). As stated above, empiri~ cal evidence on the impact of income on mortality is mixed. Liter/yiefaum-s The impact on mortality of ditferences in lifestyles across geographic areas is captured by cigarette consumption, measured by packs sold per capita. alcohol consumption. measured by gallons sold per capita. and marital status measured by percent of married households. The positive relationship between cigarette consumption and mortality is well established: however, the health effects of alcohol consumption are less clear. While heavy alcohol consump- tion is known to lead to cirrhosis of the liver and other alcohol related diseases that negatively effect health, recent studies suggest that moderate consumption may have benecial health eects. Thus, the net effect of alcohol consumption on the death rate in population groups is uncertain. States with a larger proportion or married households are expected to have lower ageadjusted death rates. It has been hypothesized that married persons receive better home care and place a higher value on health relative to other market goods and risky activities than unmarried persons, and avoid the negative health errects of grier and stress associated with divorce and death ofa spouse. While it is possible that individuals in poorer health are less likely to get married so that causation runs from health to mar~ tial status, Rosen and Taubman (1982) nd no empirical support for this reverse causation hypothesis. Enrirmuimlml [zit loi's Environmental inuences on mortality are represented by three variables: urbanization. crime and manufacturing. Urbanization is measured as the percent of the population that resides in a standard metropolitan statistical area. It is a proxy for a collection of potential negative and positive health related factors. such as pollution, congestion. and access to medical care, and therefore its net effect on mor tality is uncertain. Crime is measured as the number of violent crimes per 100 our) population and is included pri- marily as a control variable: however, the relative magni- tude of its effect on mortality is of policy interest. Manufacturing activity, measured as the percent of employed persons in manufacturing. is an index of indus- trialization, and may capture the health risks associated with exposure to hazardous agents or other adverse health effects associated with manufacturing type work. Health production function for the US 61 62 J. Thornton Control variables Table 1. Estimation results for health production function model Overall, the results of this study suggest that the medical Grossman, M. (1972) The Demand for Health: A Theoretical and Empirical Investigation, National Bureau of Economic Race and gender variables, not reported below, are Variables Coefficient -Statistic care system is but one component in a much larger health care system. If the primary goal of health care policy is to Research: New York. included to control for personal characteristics of the Hadley, J. (1982) More Medical Care, Better Health? Urban Constant 6.59 4.61 population. maintain and improve the health status of the population, -0.065 Institute: Washington, DC. Medical care expenditures 0.43 Income -0.179 1.86 then greater recognition should be given to the role of Kenkel, D. (1991) Health behaviour, health knowledge, and schooling, Journal of Political Economy, 99, 287-305. Education -0.200 2.42 socioeconomic and lifestyle factors in preventing disease 0.077 and improving life expectancy. Policies that strengthen Leigh, J. (1983) Direct and indirect effects of education on health, Cigarette consumption 3.17 III. EMPIRICAL FRAMEWORK AND Alcohol consumption 0.038 0.96 the nation's education system by improving quality and Social Science and Medicine, 17, 227-34. Urbanization -0.025 1.00 increasing access, foster sustained income growth shared Levit, R., Lazenby, H. and Braden, B. (1998) National health RESULTS pending trends in 1996, Health Affairs, 17 (1, January/ Manufacturing 0.013 0.69 February), 35-51. Married households -0.572 2.94 by all, reduce crime, and promote social stability may The analysis uses state level data for the year 1990. The Crime 0.038 Newhouse, J. and Friedlander, L. (1980) The relationship between 2.87 well be highly effective in improving the nation's health data were obtained from various publications of the US R unadjusted 0.80 as well as achieving other important social goals. It is medical resources and measures of health: some additional Department of Health and Human Services, US Census R adjusted 0.74 here, rather than within the medical care system, that the evidence, Journal of Human Resources, 15, 200-18. Rosen, S. and Taubman, P. (1982) Some socioeconomic deter- Bureau, National Institute on Alcohol Abuse and roots of true health care reform might be found. Notes: Instrumental variable measures used for medical care minants of mortality, in Economics of Health Care (Eds) Alcoholism, and the Advisory Council on Intergovern- expenditures and income. The instrument set contains, education, . van der Gagg, W. B. Neeman and T. Tsukahara, Praeger mental Relations. cigarette consumption, alcohol consumption, urbanization, Publishers: New York, pp. 255-71. Silver, M. (1972) An econometric analysis of spatial variations in The empirical aggregate health production function is manufacturing, married households, crime, race, gender, unem- REFERENCES given by: ployment rate, birth rate, percentage of the population 65 years of mortality rates, in Essays in the Economics of Health (Ed.) age and older, percentage of population without health insurance, Auster, R., Leveson, I. and Sarachek, D. (1969) The production V. Fuchs, National Bureau of Economic Research: New of health, an exploratory study, Journal of Human Resources, In Di = Bo + BM In M; + Bs In S; + BL In L; percentage of population below federal poverty level and Medi- York, pp. 161-227. care recipients per capita. The R" statistics for the first-stage 4, 411-36. Smith, S., Freeland, M., Heffler, S. and McKusick, D. (1998) The + BE In E; + BcCit Hi (1) regressions for medical care expenditures and income are 0.72 Berger, M. and Leigh, J. (1989) Schooling, self-selection, and next ten years of health spending what does the future hold? and 0.83, respectively. health, Journal of Human Resources, 24, 433-55. Health Affairs, 17 (5, July/August), 86-95. where In denotes natural logarithm, D; is the age-adjusted Enthoven, A. (1980) Health Plan. Addison-Wesley. Reading, MA. Taubman, P. and Rosen, S. (1982) Healthiness, education, and death rate in state i, M; is medical care expenditures, S; is Fuchs, V. (1994) The Future of Health Policy, Harvard University marital status, in Economic Aspects of Health (Ed.) V. Fuchs, Press: Cambridge. University of Chicago Press: Chicago, pp. 121-42. the vector of socioeconomic variables, L; is the vector of lifestyle variables, E; is the vector of environmental vari- For the non-medical variables, the results indicate that ables, C; is the vector of control variables, Bo and BM are states with higher levels of income, education, and percent scalar parameters, Bs, BL, BE and Bc are parameter vectors, of married households have significantly lower death rates, and u; is a mean zero, finite variance error term. The theory while states with greater cigarette consumption and crime of health demand (Grossman, 1972) suggests that not only have significantly higher mortality, all else the same. does health depend on medical care and income, but med- Marital status has the largest impact on mortality ical care and income also depend on health: healthier popu- (-0.572), followed by education (-0.200), income lations demand less medical care and generate greater (-0.179) and cigarette consumption (0.077). All of these money earnings. To obtain consistent estimates of the par- variables have plausible signs and are significant at the 5 ameters of the health production function, the two-stage and 1% levels, with the exception of income which is least squares (2SLS) estimation procedure is employed. significant at the 10% level. States with higher levels of Table 1 presents the 2SLS results for the health produc- alcohol consumption and manufacturing activity have tion function model. This model explains 80% of higher age-adjusted mortality rates, while more urbanized the interstate variation in age-adjusted death rates. The states have lower death rates, but the estimates for these estimated coefficient of medical care is -0.065 and its three variables are highly insignificant, with t-statistics of t-statistic is 0.43. The results indicate that the marginal one or less. contribution of medical care utilization in lowering mortal- ity in the US is quite small. Like Auster et al. (1969) there is no evidence that medical care is a significant factor explain- ng variation in age-adjusted death rates across states. The IV. CONCLUSIONS point estimate indicates that a 1% increase in medical care use will decrease mortality by 0.065%. This estimate The results of this paper indicate that additional medical obtained using 1990 data is somewhat smaller than the care utilization has little if any impact on mortality. This estimate of -0.10 to 0.15 produced by prior studies, and suggests that the expected doubling of medical care is consistent with the argument that the US is experiencing expenditures between 1996 and 2007 will contribute little diminishing returns to medical care use and may well be to the goal of reducing mortality in the US. The findings operating in an environment of 'flat-of-the-curve' medicine, suggest that lifestyle factors and socioeconomic status where additional resources devoted to medical care yield are important determinants of mortality, and have a little if any improvement in lowering mortality (Enthoven, substantially larger marginal impact on death rates than 1980). medical care utilization