MEDICAL CARE EXPENDITURES AND GDP GROWTH IN OECD NATIONS

MEDICAL CARE EXPENDITURES AND GDP GROWTH IN OECD NATIONS

R. Mark Wilson, University of South Florida

A nation’s medical care spending (per person) is examined as a part of a model of economic growth. Two-stage-least-square estimation results indicate that medical care expenditure influences the rate of economic growth and that the rate of economic growth increases spending for medical care in OECD countries. These results have two implications for the prediction of medical care spending. First, include the per capita GDP growth rate as an explanatory factor, and second, use a statistical method that allows the relationship to be interdependent.

Introduction

The United States spends approximately 14% of its gross domestic product (GDP) on medical care; France spends 10% and Norway 8% of their respective GDPs. Why does the U. S. spend so much on medical care, or Norway so little? Are the people in the U. S. healthier than people in these other nations? Apparently not. The U. S. population has the lowest expected lifespan and the highest infant mortality rate of these three, but is healthy relative to most of the rest of the world. Worldwide, nations with the least healthy populations don’t spend the most on medical care, measured either as expenditures per person or as a percent of the nation’s GDP. If not the healthiness of the people, what does determine the amount a nation spends on medical care? More to the point of this paper, how important are economic factors?

A growing literature in the field of health economics has centered on the role of GDP in determining a nation’s level of medical care expenditures. Following the influential work of Newhouse (1977), the central question has been whether a percent increase in per capita GDP in a nation will bring about a greater, an equal or a lower percentage increase in medical care expenditures per person. The preponderance of studies (Gerdtham and Jonsson, 1991; Hitiris and Posnett, 1992; Moore, Newman and Fheili, 1992; Newhouse, 1977; O’Connell, 1996; Parkin, McGuire and Yule, 1987) find that medical care spending rises by at least the percentage that GDP does in developed economies, and close to the same percentage in developing nations (Gbesmate and Gerdtham, 1992). The studies vary by years covered, method of converting from national currencies to a common dollar unit for comparison (exchange rates versus purchasing power parity), type of statistical technique, and what other factors are held constant. Every study finds GDP to be the major determinant of the medical care expenditures of a nation. However, none consider that the strong relationship between GDP and medical care expenditures might be due in part to continued large medical expenditures over recent years leading to current high GDP per capita.

A separate avenue of economic research has focused on the causes of economic growth, i.e. exploring the determinants of GDP growth across nations. The studies are grounded in macroeconomic growth theory and, when data is available, empirical tests of hypotheses concerning the rate of growth of GDP include a proxy for the healthiness of the population as an explanatory variable (Barro and Sala-I-Martin, 1995). The implication is that medical care expenditures improve health and health raises GDP growth rates. It is not clear whether the direction of causation is from health to GDP, as in the growth literature, or from GDP growth to increased medical care spending to improved health, as would be consistent with the health economics literature and the findings of Aturupane, Glewwe and Isenman (1994).

In this study, I use concepts and procedures from each of the two research fields to identify determinants of medical care expenditures in developed nations. I begin with simple correlations between medical care expenditures (per person) and potential determinants such as GDP level, GDP growth rate, as well as the age structure of the population and the method of funding health insurance in the country. I then estimate a joint model of medical care spending (per person) and GDP growth rate (per person). Hypothesis tests provide insights into the strengths of the determining factors and into the interdependent relationship between medical care spending and economic growth.

Models and Methods

Conceptual Framework

Economists have not developed a formal theory to explain or to predict the per capita medical care expenditure of a nation. Theories of countries’ preconditions that lead to spending for medical care do not exist, and as Parkin, McGuire and Yule (1987) point out, formidable problems exist when researchers attempt to construct a theory of national medical care spending by aggregating models of the purchasing behavior of individual households. In the absence of a theory, empirical work in this field has necessarily been based on ad-hoc reasoning and data availability. The models and methods used in this study to predict a nation’s medical care spending (per person) are therefore based on the results of previous studies, the hypothesized two-way relationship between health spending and economic growth and data availability.

As in other economic studies, the expected finding was a dominating positive relationship between national medical care expenditures and gross domestic product, both in per capita terms. Unlike previous studies, it was also expected that the expenditures would be correlated with the growth rate of GDP per capita. The hypothesis is that the extra national income generated by economic growth is spent differently when the growth is rapid rather than slow or negative, other factors held constant. The empirical evidence on this topic is virtually non-existent. Van der Gaag and Barham (1998) found that government spending for health care declines when countries experience recession, i.e. negative growth rates. Evidence of the effects on all medical care spending or of positive growth rates remains unexamined. Gyimah-Brempong and Traynor (forthcoming) find that the rate of economic growth positively influences expenditures on physical capital; I hypothesize the same relationship applies to human capital investments. Large expenditures are made in every nation every year for two major types of human capital, education and health. This paper’s focus is spending to build health human capital. The hypothesis of Gyimah-Brempong and Traynor (forthcoming) and others is that the amount of capital will keep pace proportionately with production in the economy. Rapid economic growth implies increases in capital, and therefore investments in the health of the population.

Previous researchers have tested the significance of a variety of non-economic background factors in determining a nation’s medical expenditures and most exert very little influence. However, Hitiris and Posnett (1992) and O’Connell (1996) find that the percent of the population over the age of 65 is correlated significantly with a nation’s per capita medical care spending. I include the same age distribution variable to represent high potential medical care use in the country.

Leu (1986) found that the proportion of medical care spending that is public and the presence of a centralized system of medical care are positively related to expenditures. Since then, Hitiris and Posnett (1992) and Moore, Newman and Fheili (1992) convincingly argued against Leu’s conclusions about the share of public spending. Also since then, several national health systems have changed with respect to centralization, dating Leu’s findings. The dominant current view is that funding and provision of medical services in a country is endogenously determined. That is, funding and distribution methods are decided concurrently with determining the level of medical care spending; they do not determine that level. According to this reasoning, a proxy variable concerning the centralization of the health care system would contain information about the level of medical care spending, but it would not be a truly exogenous variable. Results both with and without a variable representing the degree of centralization of funding for the national medical care system are reported. If a centralized funding method brings greater government control, per capita medical care expenditures will be reduced. If it brings few government controls, low personal out-of-pocket spending for medical care and wide availability of services, expenditures will rise. The empirical results will provide a clue to the role of factors associated with funding method.

In the above discussion, the rate of economic growth raises medical care spending. In the economic growth literature, however, human capital (such as health) influences the rate of growth. Following Barro and Sala-I-Martin (1995), a nation’s economic growth is a function of two types of factors, called state and environmental effects. State variables represent the "state of the economy" upon entering the time frame of the dataset and are generally proxies for the amount, or stock, of human and physical capital available. Environmental variables represent changes occurring in the economy during the time under study, i.e. the current "economic environment". Appropriate concepts span a range of economic and political measures including, but certainly not limited to, current investment in human and physical capital, the economic and political environment for foreign trade and political stability in the country. Of special importance to this study is that expenditure on medical care is a form of investment in human capital and is expected to increase the rate of economic growth.

Neo-Classical theory of economic growth (see Solow, 1956) predicts that simultaneous proportional increases in the state variables, human and physical capital, will cause a slower growth rate of per capita GDP, other factors held constant. In this situation, diminishing marginal returns to capital reduce the rate of economic growth. However, given a constant level of physical capital, a higher level of human capital will increase the growth rate of GDP toward the model’s long-run equilibrium. This implies that a statistical analysis that holds other factors constant will find a positive relationship between human capital and the rate of economic growth. Additional medical care investments would build human capital and spur growth.

The newer "endogenous" growth models, where technological progress is determined within the model (see Nelson and Phelps, 1966 and Romer, 1990), describe the interdependence between economic growth and medical care spending. Additions to human capital (including health) facilitate the creation and/or adoption of new technologies and lead to economic growth. The growth in GDP brings investments in human capital for continued growth. Therefore, medical care expenditures increase the rate of economic growth and economic growth raises medical care spending.

In summary, per capital spending on medical care is a function of a nation’s rate of economic growth and other economic and non-economic influences. A nation’s rate of economic growth is itself a function of several factors, including investment in, i.e. expenditure on, health. This study will investigate determinants of a nation’s level of health care spending while explicitly accounting for this interdependence.

First, summary statistics and simple linear regressions will describe the data and the correlation of medical care spending with other variables. Next, two-stage-least-square (2SLS) estimates of the interdependent relationship between medical care expenditures and economic growth will be estimated. The system of equations is:

medcare = a₀ + a₁(gdprate) + a₂(gdp) + a₃(age65) + a₄(funding) + a

gdprate = b₀ + b₁(medcare) + b₂ (age65) + b₃(education) + b₄(gdp_t-1) + b₅(labor)

+ b₆(invest) + b

The 2SLS method provides consistent estimates of the coefficients of this interdependent system of equations; ordinary least-square (OLS) and three-stage-least-square (3SLS) procedures would lead to inconsistent estimates, given the probable degree of misspecification in these equations (see Pindyck and Rubinfeld, 1991).

In the first equation, medcare is average medical care spending per person in the nation, gdprate is the rate of growth of GDP per person and gdp is the level of GDP per person in the nation. Age65 is the percent of people in the country who are aged 65 or over. The funding variable is set to 1 if the system of national medical care is funded by general taxation (Denmark, Finland, Greece, Iceland, Ireland, Italy, Norway, Portugal, Spain, United Kingdom) and is zero otherwise.

In the second equation, gdprate is a function of state variables for human and physical capital, age65, education and gdp_t-1, and environmental factors, medcare, labor, and invest. Age65 is a measure of the nation’s health capital; education is the average number of years of education of the population and proxies the nation’s education capital. Measures of the physical capital stocks of nations are often difficult to compare because of different national income accounting rules and depreciation methods. Therefore, I follow Barro and Sala-I-Martin (1995) and use the previous time period’s gdp, gdp_t-1, as a correlate to the stock of physical capital. The reasoning is sound; given the level of human capital, the level of physical capital is directly related to GDP per person. Another reason to include gdp_t-1 as an explanatory variable is to test the convergence hypothesis (technically, conditional convergence in this study), that higher income nations grow at slower rates. The other variables are environmental: medcare; labor, the rate of growth of the labor force; and invest, the ratio of investment to real GDP to represent the current growth of the physical capital stock.

Data is obtained from a variety of sources. Information on medical care expenditure, gross domestic product and economic growth in the Organization for Economic Cooperation and Development (OECD) nations are taken from OECD Health Data files. Almost all of the OECD nations have developed economies and the OECD has made efforts to collect comparable data across the countries. Appendix 1 contains a list of the countries with data available for this study. Method of funding of the system is from Saltman and Figueras (1998) and Folland, Goodman and Stano (1997, p. 517). OECD and World Health Organization (WHO) statistics provide age distribution information. Labor force and investment data are collected by the World Bank and distributed as World Development Indicators. I use the education measure created by Barro and Lee (1996).

OECD, WHO and World Bank information is collected annually. Conceptually, it is not clear how long a time frame for data observation is appropriate. Does GDP growth influence medical care expenditures in a nation immediately and in a lump sum or is the increase in spending phased in over some unspecified amount of time? The data is arranged in five-year averages to allow time for the effects to occur and to reduce the effects of short-term economic disturbances due to such events as destructive weather, business cycles or election years. The final dataset contains 1985-1989 and 1990-1994 average values of the variables for 23 OECD countries. At the time of this writing, 1994 was the most recent year with complete data. Each country has two data points in order to increase the size of the dataset; the points are separated in the estimations by adding a dummy explanatory variable equal to 1 if the point is a 1990-1994 observation, and zero otherwise, to each equation. The lagged variable, gdp_t-1, is GDP per person in 1985 and in 1990 for the two time periods, respectively.

Theory does not indicate a preferable functional form for either equation. Many researchers find, through trial and error, that taking the logarithms of the variables yields relatively high R-squares. Several of the OECD countries suffered recessions (negative growth rates) during the early 1990s, rendering log(gdprate) useless. I follow Barro and Sala-I-Martin (1995) functional forms in the growth equation and the regular, untransformed data for the remaining variables.

Two methods are commonly used to convert national currencies to dollars to allow comparison of expenditures across countries; both methods provide problems. Exchange rate conversion of expenditures valued in foreign currencies to dollars reflects both quantities and prices in the country of origin. Therefore, high medical care spending in a nation may represent more health care or merely high prices. Purchasing power parities (PPP) were developed to identify quantity differences among nations and to eliminate the price effects. The PPPs adjust each nation’s expenditures to reflect approximate average international prices. These are more accurate for traded goods that actually have international prices than for doctor visits and other local medical services. In early studies, the method of conversion may have influenced the empirical results, but the most recent studies (Gerdtham and Jonsson, 1991; Hitiris and Posnett, 1992; Moore, Newman and Fheili, 1992) find little difference. I use values converted with purchasing power parity indices.

Results

Appendix 2 contains the means and standard deviations, along with the countries with the highest and the lowest values for each variable. The United States has the highest medical care expenditures per person. Part of the reason is the strong correlation of medical care spending with GDP; the U. S. also has the highest GDP per person. This strong relationship is shown in Figure 1, a plot of GDP and medical care spending (both per capita). The point above the linear regression line on the upper right of the graph is the U.S. Notice that the points are close to the line and GDP looks like a good predictor (indeed, the adjusted R² = .80) of expenditures. A contrast to the close fit of the data to the regression line in the top graph is the relationship between two other variables in the medical care spending equation, the rate of per capita GDP growth and the percent of the population aged 65 or more, and expenditures. Their graphs are not shown because the data points are not well predicted by the lines and the correlations are not statistically significant.

The column of points along the vertical axis in the bottom graph of Figure 1 represents the spending of countries with decentralized insurance systems. The U. S. is the top point on the axis. The column toward the right of the graph is a plot of medical expenditures by nations with medical care systems funded by central taxation. The slope of the line connecting the mean value of each group is significantly negative at a = .05.

Table 1 contains the results of the simultaneous estimation of medical care spending and the rate of economic growth in a nation, both in per capita terms. Gross domestic product is the best single predictor of expenditures; the growth rate of the economy is also statistically significant, but a distant second. Given a level of GDP per capita, faster growing economies spend more on medical care per person. The other variables are not significant. It made little difference to the remaining coefficients when Funding and/or Log(Age65) were omitted or when Age65 was entered in place of Log(Age65).

The second equation in Table 1 shows that medical care spending is a significant predictor of the rate of growth of an economy. The hypothesis from the theories of economic growth that investments in health human capital increase the rate at which GDP grows is supported by the data. The significant negative correlation of lagged per capita GDP, gdp_t-1, is consistent with the economic growth literature. It shows diminishing marginal returns to physical capital when holding human capital constant, and demonstrates the convergence of GDP per person across nations. Larger economies grow more slowly than smaller economies, other things equal. The other significant coefficient is on the time period dummy variable. Recessions in many nations in the early 1990s dramatically lowered growth rates.

Discussion

Medical care expenditure influences the rate of economic growth and the rate of economic growth increases spending for medical care. The new empirical finding is that among nations with identical per capita GDPs, per capita medical care spending will be highest in the fastest growing country. These results have two implications for the prediction of medical care spending. First, include the per capita GDP growth rate as an explanatory variable, and second, use a statistical method that allows the relationship to be interdependent.

In other ways these estimates agree with those of existing studies. GDP is the primary determinant of medical care spending in a nation, and not much else matters. Real per capita GDP converges over time across countries. The percentage change in medical expenditures due to a one percent increase in GDP, the income elasticity of demand, is estimated to be 1.79% in this study. This elasticity is in line with previous cross-national estimates. It is higher than some (many are in the 1.0-1.3% range), but lower than the 2.8-3.2% calculated from linear estimates in Moore, Newman and Fheili (1992). The major statement regarding the elasticity estimate is that the interdependent estimation procedure leads to the same conclusion as did single equation models: medical care is a luxury good. As a nation becomes wealthier it spends proportionately more on medical care.

Two factors in this study may influence this percentage change calculation. The first is the use of untransformed data. Parkin, McGuire and Yule (1987) show that R²s are not necessarily maximized with the linear setup, and that elasticity calculation is subject to pitfalls for any of a variety of functional forms. All elasticity estimates, past and present, should be viewed as approximations. Although these results are quite reasonable, other functional forms should be tried so that a consensus can emerge from a variety of reasonable estimates.

The second is the timing of the dataset. This data is newer than in any of the cited studies, and with the new data comes an unusual situation. The early 1990s were recession years for over half of the OECD nations. The theories are based on behavior in growing economies, not contracting ones. High unemployment, extra plant capacity and government budget problems all may have an effect on medical care expenditures. The elasticity estimate may be entirely accurate; a 1% decrease in the GDP variable may truly lead to a 1.79% decrease in medical care spending. In addition, spending will be reduced further because of the negative GDP growth rate. Among nations with identical per capital GDPs, per capita medical care spending will fall the farthest in the most recessionary economy.

This study merges two, formerly separate, avenues of economic research to examine common determinants of medical care spending across countries. Results show the topics of economic growth and investment in the health human capital of a nation are interdependent. Future studies of growth or national health spending would gain by reflecting this two-way relationship. Expansions of the interdependent analysis can be made by adding countries, perhaps including lesser developed nations, by addressing growth versus recession differences and by experimenting with functional forms, estimation techniques and additional years of data. The result will be greater insight into explaining the difference in medical care expenditures among nations.

Acknowledgement: I thank Kwabena Gyimah-Brempong for his helpful comments and constructive advice. However, I cannot hold him responsible for the final product; all remaining errors and omissions are my own.

References

Aturupane, H., Glewwe P. & Isenman P. (1994). Poverty, human development, and growth: an emerging consensus? American Economic Review. 84, 2, 244-249.

Barro, R. J. & Lee, J. W. (1996). International measures of schooling years and schooling quality. American Economic Review. 86, 2, 218-223.

Barro, R. J. and Sala-I-Martin X. (1995). Economic Growth. New York: McGraw-Hill, Inc.

Folland, S., Goodman, A. C. & Stano, M. (1997). The Economics of Health and Health Care. Upper Saddle River, New Jersey: Prentice-Hall.

Gbesmate, K. P. & Gerdtham, U-G. (1992). Determinants of health care expenditure in Africa: a cross-sectional study. World Development. 20, 2, 303-308.

Gerdtham, U-G. & Jonsson, B. (1991). Conversion factor instability in international comparisons of health care expenditures. Journal of Health Economics. 10, 227-234.

Gyimah-Brempong, K. and Traynor, T. L. (forthcoming). Political instability, investment, and economic growth in Sub-Saharan Africa. Journal of African Economies.

Hitiris, T. and Posnett, J. (1992). The determinants and effects of health expenditure in developed countries. Journal of Health Economics. 11, 173-181.

Leu, R. E. (1986). The public-private mix and international health care costs. In Culyer, A. J. and Jonsson, B. (eds.) Public and Private Health Services. Oxford: Basil Blackwell.

Moore, W. J., Newman, R.J. and Fheili, M. (1992). Measuring the relationship between income and NHE’s. Health Care Financing Review. 14, 1, 133-139.

Nelson, R. R. and Phelps, E. S. (1966). Investment in humans, technological diffusion, and economic growth. American Economic Review. 56, 2, 69-75.

Newhouse, J. (1977). Medical care expenditure: a cross-national survey. Journal of Human Resources. 12, 115-125.

O’Connell, J. M. (1996). The relationship between health expenditures and the age structure of the population in OECD countries. Health Economics. 5, 573-578.

Parkin, D., McGuire, A. and Yule, B. (1987). Aggregate health care expenditures and national income. Journal of Health Economics. 6, 109-127.

Pindyck, R. S. and Rubinfeld, D. L. (1991). Econometric Models and Economic Forecasts. New York: McGraw-Hill, Inc.

Romer, P. M. (1990). Endogenous technological change. Journal of Political Economy. 98, 5, II, S71-S102.

Saltman, R. B. and Figueras, J. (1998). Analyzing the evidence on European health care reforms. Health Affairs. 17, 2, 85-108.

Solow, R. M. (1956). A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics. 70, 1, 65-94.

van der Gaag, J. and Barham, T. (1998). Health and health expenditures in adjusting and non-adjusting countries. Social Science and Medicine. 46, 8, 995-1009.

Appendix 1: OECD Countries in the Study

Australia 1985-89, 1990-94 Japan 1985-89, 1990-94

Austria 1985-89, 1990-94 South Korea 1990-94

Belgium 1985-89, 1990-94 Netherlands 1985-89, 1990-94

Canada 1985-89, 1990-94 New Zealand 1985-89, 1990-94

Denmark 1985-89, 1990-94 Norway 1985-89, 1990-94

Finland 1985-89, 1990-94 Portugal 1985-89, 1990-94

France 1985-89, 1990-94 Spain 1985-89, 1990-94

Germany 1990-94 Sweden 1985-89, 1990-94

Greece 1985-89, 1990-94 Switzerland 1985-89, 1990-94

Iceland 1985-89, 1990-94 United Kingdom 1985-89, 1990-94

Ireland 1985-89, 1990-94 United States 1985-89, 1990-94

Italy 1985-89, 1990-94

Appendix 2: Summary Statistics