THE EFFECT OF HEALTH EXPENDITURE ON ECONOMIC GROWTH

THE EFFECT OF HEALTH EXPENDITURE ON

ECONOMIC GROWTH

AUTHOR

JOOST BATS

This paper presents a panel regression on 17 OECD countries to test for relationship between health care expenditure and economic growth using a re-specified Solow growth model. The main results show that health care expenditure has a negative but arbitrarily small effect on economic growth.

JEL-CODES

O47, H51, I15

KEYWORDS

2

I. INTRODUCTION

Following the framework of neoclassical economics, it has traditionally been argued that the main determinant for economic growth and prosperity is the accumulation of physical capital. Any additional long-term growth would considered being exogenous and labeled as technological progress. From the 1980’s onward however, several studies have suggested that long-term growth is endogenous and that not only physical capital accumulation explains economic growth. Whereas the neoclassical model focuses on the growth in physical capital explaining economic growth, other schools of thought claim that it is merely the growth in ideas that determine long-term growth (Lucas, 1988). Referring to this accumulation of ideas, human capital is a just as important determinant for economic growth as physical capital. Not only does the accumulation of human capital create individual economic growth, it also adds to the growth of the social and the national output. Mincer discusses how human capital both grows by the transmission of available knowledge from person to person and by the production of new knowledge (Mincer, 1984). In general, it is said that the production of new knowledge occurs through an increase in the quality and quantity of education for people, through training, or through experience. For this reason, several studies have tested the effect of education on economic growth through an increase in human capital. Keller has concluded for example, that secondary and higher education enrollment rates and expenditures per student in lower education stages have a significant effect on economic growth (Keller, 2006).

3 This paper is organized as follows. Section II provides a review of the existing literature on the topic of health and economic growth. In section III the economic model is specified. Then in section IV the data is presented that will be used for the econometric model in section V. Section VI presents the results and findings followed up by a discussion of their meaning in section VII. The last section provides a conclusion.

II. LITERATURE REVIEW

It is generally believed that economic growth improves a population’s health since economic prosperity causes people to live better and longer. Nonetheless, health also has a causing effect on economic growth. A population’s health may affect its country’s economic growth because when a population’s health improves, it can produce more output with the same amount of knowledge, physical capital, and technological progress. There are in fact, two reasons for why a more healthy population would lead to an increase in producing output. The first is that because people will live longer, they can work longer such that they can produce more output over a lifetime. The second reason is that when people are healthier, it is assumed they will be more productive such that a healthier person will produce more output than a less healthy person during a specific period. For this reason, also health constitutes to human capital. Bloom et al. (2004) show that a population’s good health has a positive statistically significant effect on national output. In their study they assume that output is determined by physical capital, total factor productivity, the labour force and human capital. Human capital subsequently consists of years of schooling, work experience, and health. This study hence concludes that economic growth is endogenously related to human capital, and that human capital does not only depend on education and knowledge but also on health.

4 economic growth. Hence, Fogel suggests that improving health affects economic growth through better nutrition (Fogel, 2004).

Weil (2007) denotes however, that the positive effect of health on economic growth is stronger for relatively poor countries (Weil, 2007). This suggests that poor countries, whose population in general contains a lower level of health, will economically benefit more from increasing the population’s health. Hartwig (2010) supports this conclusion by showing that health in fact has no positive effect on GDP per capita growth for rich countries. He therefore uses data from 21 OECD countries to form a panel granger-causality regression on rich countries. On the contrary, Heshmati has shown that health does have a positive effect on OECD countries their economic growth. Hence, there exists some mixed evidence and conclusions from several studies on how health affects economic growth for rich countries.

Besides the positive effects of increasing health on economic growth, the literature also presents research on the negative effect of worsening health on economic growth. Popkin (2008) for example, proves how the transition of nutrition, physical activity, and an increase in obesity in China had a causing effect on the slowdown of China’s economic growth. He thereby specifies that an economic slowdown does not only result from a decrease in worker’s productivity and health, but also from a potential increase in costs such as for example future health care costs (Popkin, 2008).

Continuing on the earlier studies of the health effects on economic growth, this paper will specialize into the effects of health care expenditure on economic growth. Although much research has confirmed that health has a positive effect on economic growth, far less research has presented the actual effect of a country’s health care expenditure on economic growth. There are nonetheless, several studies that present the effects of health care expenditure on economic growth of national economies. Odubunmi et al. suggest, using a re-specification of the Solow model (that includes human capital), that increasing the proportion of Nigeria’s national budget allocated to health care services would foster economic growth for Nigeria (Odubunmi et al., 2012). Furthermore, Bukenya concludes that the relationship between health care expenditure and economic growth is positive for the southeast of the United States (Bukenya, 2009). Mehrara and Maysam present no support however, for the effect of health care expenditure being positive on economic growth for Iran (Mehrara and Maysam, 2011).

5 (Wang, 2011). He concludes that a country’s health care expenditure growth would stimulate a growth in a country’s GDP. In contrast to this conclusion, Hartwig analyzes 21 OECD countries and suggests that health care expenditure does not have a positive effect on economic growth (Hartwig, 2010). The difference between these conclusions may be explained by the fact that Wang uses a cluster of relatively poor and relatively rich countries, whereas Hartwig performs a panel regression on only relatively rich countries.

The models used in Wang and Hartwig their studies however, are from this paper’s point of view too simplistic. Firstly, Wang bases his conclusion on a model in which output is only explained by total health care expenditure, autonomous expenditure, and the income elasticity of health care expenditure. Next to that, Hartwig uses five year averages of health care expenditure to explain economic growth. Clearly, following earlier work of Solow, Lucas, and Weil, economic growth is explained by much more than a population’s health. Furthermore, Wang uses a model in which health care expenditure explains the aggregate national output of a country. He therefore, does not base his study on the actual growth rate of a country’s national output.

Based on the above review, this paper will test the effect of health care expenditure per capita on economic growth by using an econometric model that incorporates more explanatory variables than earlier panel studies have done. Also, this study will in particularly focus on growth rates and not on absolute aggregate values.

In the contrary to the studies of Odubunmi et al, Bukenya, and Mehrara and Maysam, this paper will not study the effect of health care expenditure on one specific country but will instead conduct a panel regression analysis focusing on developed OECD countries. Using a different regression model from Hartwig’s and Wang’s study, and due to the mixed evidence of the health effect on economic growth for richer countries, this paper will test whether there is an actual positive affect of health care expenditure on economic growth for the relatively rich OECD countries.

III. ECONOMIC MODEL

6

(1)

Where = Aggregate real output, = Capital Stock, = Human Capital, = efficiency Factor, = labour force, and where the subscript i represents the ith country and the subscript t represents the tth year of observation. Following Weil, human capital is assumed to depend on a population’s education and health and therefore, this paper specifies human capital as:

(2)

Where = return to education multiplied by the average number of years of education per person and = total health care expenditure per capita. The rate of return to education converts years of education into a measure of human capital. Continuing on Weil’s approach, a rate of return to education of 10 % will be applied to the education variable in this regression (Weil, 2007). By taking the natural log of equation (1) and (2), the model transforms to a multiple linear regression model:

(3)

Since this paper studies the effect of health care expenditure per capita on economic growth, the aggregate Cobb-Douglas model transforms to a model represented by growth rates and includes a variable representing the initial value of output. Also, a lagged variable of health care expenditure will be included since it is assumed that the effect of health care expenditure on GDP growth needs time to be present.:

(4)

Where = the natural log of output’s growth rate, ln = the natural log of the efficiency factor, = the natural log of the initial value of output, ln = the natural log of the growth of capital, and where ln = the natural log of population growth. The

7 assumed that the population growth corresponds to the labour force growth. Therefore, it is assumed that population growth has a similar effect on the growth of GDP.

To take into account technological progress and the depreciation of capital, equation (4) is transformed in the following way:

(5)

Following the difference between equation (4) and (5), is changed to , where n captures the growth rate of the population and where the sum of technological progress and the depreciation of capital is assumed to be 0.05 for every country and constant over time. It is assumed that technological progress represents a country’s efficiency and therefore, is removed from the equation.

In section V, this paper will use equation (5) as the foundation for the econometric model. The econometric model will be further adjusted by adding country- and time fixed effects and by further testing the econometric model on its eligibility. This paper will first however, describe the data that will be used as a proxy for the dependent and explanatory variables in the following section.

IV. DATA

A panel of 17 developed OECD countries is constructed over a period from 1980-2010 using 5 year averages as to control for business cycle changes. This leads to a total of 102 observations. The data for this panel comes from the following countries: Australia, Austria, Belgium, Canada, Switzerland, Germany, Denmark, Finland, France, the United Kingdom, Ireland, Iceland, Luxembourg, the Netherlands, Norway, Sweden, and the United States. France currently has the lowest GDP per capita with a value of 38,163 US dollars in 2014. Therefore, the sample of OECD countries chosen includes relatively rich and developed countries. A descriptive statistics of the data is given at the end of this section and presents the number of observations, mean value, standard deviation, minimum value, and maximum value.

8 appendix, shows that GDP growth increased for every country during the early 1980’s, and that it decreased after 2000 (except for the Netherlands and Switzerland). As shown by the descriptive statistics at the end of this section, the GDP growth data presents to be fluctuating over an average of 2.5 %.

The initial value of output is represented by the aggregate value of GDP per capita. This data is valued in constant 2005 US dollars and is also obtained from the World Bank. There are no missing values for this data.

Capital growth is presented by a country’s gross capital formation as a percentage of GDP. The data representing a country’s gross capital formation is also obtained from the World Bank. There are no missing values for this data.

The fourth variable represents the mean years of total schooling for adults aged 25 and over, and is multiplied by the rate of return to education (10%) so that the variable becomes . The data is obtained from the data set constructed by Robert Barro and Lee Jong-Wha (2010). Robert Barro and Lee Jong-Wha have presented the mean years of total schooling for every five years such that there is data for the years 1980, 1985, 1990, 1995, 2000, 2005, and 2010. Since this analysis will use 5 year averages for the panel regression, the data given corresponds to the right years. The data shows a clear increasing trend for all countries over time.

Total health care expenditure data are from OECD statistics (2014). The data represents the expenditure of both individual and collective health care per capita given in US dollars based on 2005 PPP rates and will be used as a proxy in the regression for the variable

. For this data, Luxembourg shows missing values until 1999 and Belgium contains

missing values until 1995. Also Sweden contains missing values for the first five years from 1980 on. There is however, following figure 2 in the appendix, a clear increasing trend for all 17 countries in their total health care expenditures. Therefore, linear series is applied to make up for the missing values such that Luxembourg, Belgium, and Sweden also contain increasing values from the start of 1980. By comparing figure 1 and figure 2 from the appendix, there is no clear correlation visible between the two variables. It seems in no particular way that health care expenditure per capita behaves in the same pattern as the growth of GDP. This will be confirmed in table 1, which presents the correlation matrix.

9 Following upon the variables presented above, the descriptive statistics are given in table 1 below.

Variable N Mean Std. Dev. Min Max

GDP growth 102 0.254 0.015 -0.011 0.096 Initial value of GDP 102 36737.070 11753.890 17648.840 82378.980 Capital growth 102 0.230 0.028 0.178 0.306 Average Years of Schooling 102 9.815 1.469 5.960 13.130 Population growth 102 0.007 0.005 -0.001 0.022 Health Care

Expenditure per capita

102 2442.075 963.375 943.414 6760.615

To determine to which extent all of the above variables are related, a correlation matrix is presented. This correlation matrix is denoted as table 2.

Natural Log GDP growth Natural Log Initital GDP Natural Log Capital growth Return on Years of Schooling Health Care Expenditure per Capita Lagged Health Care Expenditure per Capita Natural Log of the Population growth Natural Log GDP growth 1.0000 Natural Log Initital

GDP -0.0758 1.0000 Natural Log Capital growth 0.0314 -0.0391 1.0000 Return on Years of Schooling -0.1821 0.4064 0.0186 1.0000 Health Care Expenditure per Capita -0.0138 0.5897 -0.0597 0.6207 1.0000 Lagged Health Care Expenditure per Capita -0.1689 0.1387 0.0232 0.1965 0.4810 1.0000

Natural Log of the Population growth

0.1440 0.3727 0.2026 0.4035 0.4420 0.1372 1.0000

The correlation matrix shows that overall, no variables are highly correlated (>0.8) so that it is not expected to encounter difficulties with studying the separate effects of the variables. Interesting is that the natural log of initial GDP, the return on average years of schooling, the health care expenditures per capita, and the lagged health care expenditures are all negatively correlated with GDP growth. Furthermore, as expected the natural log of initial GDP is

Table 2: Correlation Matrix Studied Variables

Note: the variables are presented such that they exactly relate to the variables in model 5.

Table 1: descriptive statistics of the data described

10 positively correlated with the return on average years of schooling, and the health care expenditures per capita. Since we study the effect of health care expenditure per capita on the natural log of GDP growth, it is interesting to note that there is a substantial low correlation value between these two variables. Since the lagged health care expenditure per capita presents a higher correlation value with the natural log of GDP growth, we may expect this variable to be more significant than the current health care expenditure per capita variable.

V. ECONOMETRIC MODEL

For the foundation of the econometric model, a linear panel data model is presented:

(6)

Where = the unknown intercept representing a country’s average, = the “between-country” error term, and = the “within-country” error term. Equation (6) is defined as a random effects model such that represents the country’s average intercept and where random country differences from the country’s average intercept are presented by . In case of country-specific heterogeneity however, it is assumed that the behavioral differences between countries are not random such that is not constant but instead, can be different for every country. Following this reasoning, a county fixed effects model is denoted by adding subscript “i” to such that the error term can be left out of the equation. This is represented by equation (7)

(7)

11 Since the data forms a regression over time, the intercept could also differ over time due to time differences. Therefore, having both country fixed effects and time fixed effects, the model changes to:

(8)

Where can now be different for both every country and every year. To estimate model (8) this paper uses a least squares dummy variable estimator (LSDVE) that includes intercept dummy variables for every five years. To test whether time fixed effects are needed, a Wald test is conducted on the intercept dummy variables for every five years. The test shows that the probability of the F-statistic is 0.0275 and therefore smaller than five percent. Hence, the null hypothesis that the coefficients for all years are jointly equal to zero is rejected and therefore the econometric model uses both country- and time fixed effects.

Following upon model (8), it may be expected for serial correlation to exist since this paper consults cross sectional data over time series. To test for serial correlation, this paper used the Wooldridge test (Wooldridge, 2002). The Wooldridge test for serial correlation gives an F-statistic (1,16) of 3.348. This F-statistic has a probability of 0.0860 to be greater than the critical value of the F-distribution. With regards to a 0.05 significance level, we fail to reject the null hypothesis that there is no serial correlation. Therefore, the standard errors of model (8) are not biased such that the efficiency of the coefficients are not affected. Although the null hypothesis of no serial correlation would have been rejected regarding a 0.10 significance level, the model will not be adjusted for serial correlation because of the short macro panels (N=6 due to the five year averages).

To continue, the model has also been tested on heteroskedasticity. Since the regression is formed over several countries, it may occur that the model’s variance differs across countries. Therefore, a modified Wald statistic is calculated to determine whether there exists groupwise heteroskedasticity in the error terms of the model. The χ² = 34761.20 which gives a probability of the χ² of 0.0% to be greater than the critical value of the χ²-distribution. Therefore, the null hypothesis of constant variance is rejected such that the model is assumed to be heteroskedastic. To solve for this heteroskedasticity problem, the regression is formed by using white robust standard errors.

12 Variables Lags average (chosen by

akaike information criterion (AIC)

Adjusted T-statistic Probability-value

Natural Log GDP growth 0.59 -3.0311 0.0012

Natural Log Initital GDP 0.82 -9.0520 0.0000

Natural Log Capital growth

0.59 -35.0135 0.0000

Return on Years of Schooling

0.52 -3.9652 0.0000

Health Care Expenditure per Capita

0.53 -13.3985 0.0000

Natural Log of the Population growth

0.47 -22.1672 0.0000

For every variable, the lags average has been estimated based on the Akaike Information Criterion (AIC). This is presented in the second row of the table. All variables show a probability value less than 0.01. Therefore, the null hypothesis that panels contain unit roots is rejected for every variable and model (8) can be considered stationary.

VI. RESULTS

We begin by estimating equation (8). The results are shown in table 3. Due to the occurrence of heteroskedasticity in every regression, all six regressions are formed using white robust errors.

Explanatory variables (2) (3) (4) (5) (6) (7)

Natural Log Initital GDP -0.0323 (0.981) -0.0375 (0.978) 0.0055 (0.997) 0.4071 (0.744) 0.4114 (0.732) 0.3271 (0.795) Natural Log Capital growth 0.2085

(0.851) 0.2111 (0.848) 0.0389 (0.970) 0.2811 (0.770) 0.1030 (0.908) 0.0923 (0.918) Natural Log of the Population

growth 0.8619 (0.744) 0.8570 (0.741) 1.0894 (0.672) 0.6045 (0.799) 0.7093 (0.759) 0.6965 (0.763) Health Care Expenditure per

Capita 6.24e-06 (0.973) -0.0001 (0.480) 0.0001 (0.581) Five-year Lag of Health Care

Expenditure per Capita

-0.0001 (0.205) -0.0002*** (0.002) -0.0002*** (0.008)

Average Years of Schooling -3.4302***

(0.004) -3.7759*** (0.002) -3.7168*** (0.002) R-squared (within) 0.3019 0.302 0.306 0.352 0.365 0.366

Prob > F (Model’s F-statistic) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Table 2: Levin-Lin-Chu unit-root test on all variables using lags determined by AIC

13 The first six rows present all six explanatory variables used in the regressions. The last two rows present the values for the R-squared (within) and the F-test’s probability for all the model’s coefficients. At a first sight it can be seen that all probabilities for every model’s F-test are less than 1 percent. Hence, for every model all coefficients used are significantly different from zero.

Starting in column (2), the results only include the natural log of initial GDP, the natural log of capital growth, and the natural log of population growth. All three explanatory variables show to be insignificant having p-values greater than a significance level of 10 percent. Subsequently, adding health care expenditure per capita in column (3) shows to have no effect on the significance of the other explanatory variables. Health care expenditure also provides to have no significant effect on GDP growth with a p-value of 0.973. Therefore, by only adding health care expenditure per capita we cannot derive any new conclusions. Instead of adding health care expenditure per capita, the lagged variable of health care expenditure per capita has been added in column (4). Although the natural log initial GDP changes of sign, all variables (including the lag of the health care expenditure per capita) remain insignificant at a 10 percent significance level. Continuing to column (5), by adding the return on years of schooling to the regression together with the health care expenditure per capita we find that the R-squared (within) increases to 0.352. Hence, by including the variable average years of schooling, the model’s goodness of fit improves by about 16%. Nevertheless, the first four variables in column (5) remain highly insignificant with regards to their effect on the natural log of GDP growth. The return on average years of schooling however, shows to be significant. With a p-value of 0.004, the variable has a significant explanatory effect at the 0.01 significance level. Different from expected however, is that the effect of average years of schooling shows to be negative on the natural log of GDP growth. In column (6), the only variable that is left out is the health care expenditure per capita. Instead, the lag of the health care expenditure has been included which leads to a small increase of the R-squared (within) to a value of 0.365. Next to the average years of schooling, the lag of health care expenditure now also shows to have a significant effect at the 0.01 significance level. Although the effect is negative, the coefficient estimate shows to be particularly low such that that the negative effect is negligibly small. Lastly, column (7) shows the regression output by using all six explanatory variables.

14 In every regression, the first four explanatory variables remain highly insignificant, such that only the lag of the health care expenditure and the average years of schooling have a significant effect on the natural log of GDP growth. These significant effects are however, for both variables negative which is different from what was expected. Nonetheless, it follows that the negative effect of the lagged health care expenditure can be considered negligibly small since it obtains a coefficient value of -0.0002. Also, as follows from the regression output, the lag of the health care expenditure per capita only becomes significant when the average years of schooling is included.

Since the average years of schooling shows to have a negative effect on the natural log of GDP growth, it is also interesting to see whether health care expenditure per capita has an effect on a population’s average years of schooling. Hence, this paper presents a second regression output model with instead, the return on the population’s average years of schooling as the dependent variable. This regression output is given in table 4 and is explained by the following four variables: the natural log of initial GDP, the natural log of capital growth, health care expenditure per capita, and the lag of the health care expenditure per capita. Table 4 presents the regression output with both country- and time fixed effects. Different from model (8) in table 3 however, is that there exists significant serial correlation between the return on average years of schooling and the four explanatory variables used in the regression. This has been given by forming a Wooldridge test. Hence, the regression output in table 4 has been formed using a first-order autoregressive model.

Explanatory variables (2) (3) (4) (5) (6) (7)

Natural Log Initital GDP -0.0280

(0.724) -0.0113 (0.884) -0.0441 (0.680) 0.0123 (0.911)

Natural Log of Capital growth 0.0131

(0.814)

-0.0171 (0.760) Health Care Expenditure per

Capita -0.0000 (0.121) -0.0000 (0.138) -0.0000 (0.150) Five-year Lag of Health Care

Expenditure per Capita

-0.0000** (0.044) -0.0000* (0.057) -0.0000* (0.059) R-squared (within) 0.7330 0.7548 0.7373 0.7567 0.7385 0.7573 Prob > F (Model’s F-statistic) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

Although the five-year lag of health care expenditure per capita shows to be negative and significant in column (3), the effect is very small on a population’s average years of Table 4: regression output model; dependent variable: return on average years of schooling;

country- and time-fixed effects regression; first-order autoregressive model

Estimated p-values (for the coefficients t-tests) are given in the parentheses below the coefficient estimates *Significant at the 10% level

15 schooling. Adding the natural log of initial GDP in column (4) and (5) and adding the natural log of Capital growth in column (6) and (7) does not have any effect on the coefficients of both the lagged and the non-lagged health care expenditure per capita. Overall, table 4 shows that only the five year lag of health care expenditure per capita shows to have significant effect on a population’s average years of schooling. The effect is however, arbitrarily small.

Overall, the regression outputs have shown that health care expenditure per capita has an arbitrarily small effect on both the natural log of GDP growth and the average years of schooling of a population in the context of the OECD sample that has been studied. Furthermore, different from what was expected, table 3 has shown that a population’s average years of schooling shows to have a large negative effect on GDP growth. The other three explanatory variables (natural log of initial GDP, natural log of capital growth, and the natural log of the population growth) have shown to present no significant effect on the natural log of GDP growth. This is different from what economic theory suggest. Following upon these results, this paper will continue to discuss the meanings of these results in the following section.

VII. DISCUSSION

First of all, it is important to note that any observed significant effect in the last section does not imply causality by itself. GDP growth is explained by many more explanatory factors than are incorporated in the model described in section V. Although time- and country-fixed effects make up for these omitted variables to some extent, this study cannot make any inference with regards to significant explanatory variables having a causing effect on GDP growth. Nonetheless, this section will discuss the meanings of the most important and most remarkable observations. At the end of this section there will follow some limitations with regards to this study.

16 Nonetheless, the results have shown that the effect of health care expenditure on GDP growth is negative and arbitrarily small. This leads to the second observation implying that health care expenditure has no overall effect on GDP growth. This observation has been different from the reasoning in section I and II. There are however, in contrast to this reasoning, two main conclusions that can be derived from these results.

First of all, this study has conducted a panel regression on relatively rich OECD countries (France having the lowest GDP per capita). Generally, the richer a country, the healthier its population such that the marginal benefit of increasing health care expenditure on a population’s health is relatively small compared to poorer countries. Using a sample of relatively healthy populations explains to a fair extent why health care expenditure has a coefficient value close to zero.

A second explanation is that there is a distinct difference between consuming and investing. A country may invest in its population’s health by health care expenditures. These health care expenditures then lead to a healthier and more productive population which in turn may increase a country’s GDP growth. However, when a country’s population is already relatively healthy, health care expenditure may instead be mostly seen as consumption which only increases a population’s utility but has no effect on GDP growth. In the case of consumption, money spend on health care can be considered to be lost investment expenditures such that the effect of health care expenditures on GDP growth is negative.

A third striking observation is that the effect of average years of schooling has a large negative effect on GDP growth. This is inconsistent with what existing literature has suggested, saying that human capital (by means of better and more schooling) should have an overall positive effect on GDP growth. This result may suggest that delaying a person’s participation in the labor force, by means of longer years of schooling, may negatively affect GDP growth. There are however too many factors explaining GDP growth, as suggested in the beginning of this section, that the observation of a negative effect of average years of schooling on GDP growth cannot be supported by this result.

Overall, referring to both table 3 & table 4, the remainder of the explanatory variables all prove to be highly insignificant. This suggests that the results of this paper do not support in favor of the Solow growth model based on the data used for this study (Solow, 1956). There are however, some limitations that have had an impact on these results.

17 variables used in the model may have had little explanatory power on GDP growth. To eliminate the effect of business cycles on GDP growth, it is suggested to study longer time periods as to test for a clearer long-term effect.

This subsequently leads to the second limitation that this paper has used a relatively short time period of only 30 years. Using time periods of over 100 years will create better pictures with respect to factors affecting GDP growth. Nonetheless, since the data of such long time spans is not easily and directly available, this paper has not been able to study the effect over a longer-term than only 30 years.

Thirdly, this paper has not tested the actual effect of health care expenditure on a population’s health but has instead assumed in the model, that a country’s health can be represented by a country’s health care expenditure. This paper is therefore limited, in concluding whether GDP growth is affected by a population’s health.

VIII. CONCLUSION

This paper has presented a panel regression on 17 relatively rich OECD countries to test for the relationship between health care expenditure and economic growth. The model has accounted for economic growth by the initial value of output, capital growth, average years of schooling, health care expenditure per capita, and population growth. Following the main result that the effect of health care expenditure on GDP growth is negative but negligibly small, it is concluded that there is no statistical evidence on the grounds of this study that health care expenditure has either a positive or negative effect on GDP growth. This does not suggest however, that a population’s health will not affect a country’s GDP. This paper has only drawn upon the direct relation between health care expenditure and economic growth, and has not looked at any effects of a population’s health on economic growth. Furthermore, it has been shown that a country’s average years of schooling has a strong negative correlation with economic growth. Since the remainder of the explanatory variables are insignificant however, this paper does not conclude that a country’s average years of schooling has any negative causing effects on GDP growth, since GDP growth is explained by much more than only a population’s average years of schooling. Furthermore, this study has shown that a country’s health care expenditure has an arbitrarily small effect on a population’s average years of schooling.

18 example refer to changes in a population’s health and changes in a country’s available capital for investment. Additionally, the sample studied can be increased by not only looking at OECD countries, but by also adding developing countries, and other relatively rich countries such as Japan. Increasing the sample with these extra countries may have a significant effect on the model studied.

REFERENCES

Barro, Robert and Jong-Wha Lee, April, "A New Data Set of Educational Attainment in the World, 1950-2010." Journal of Development Economics 104 (2010): 184-98.

Bhargava, Alok, Dean T. Jamison, Lawrence J. Lau, and Christopher J.l Murray. "Modeling the Effects of Health on Economic Growth." Journal of Health Economics 20.3 (2001): 423-40. Bloom, David E., David Canning, and Jaypee Sevilla. "The Effect of Health on Economic Growth: A

Production Function Approach." World Development 32.1 (2004): 1-13. R.

Bukenya, James. "Do Fluctuations in Health Expenditure Affect Economic Growth?" The Open

Economics Journal 2 (2009): 31-38.

Fogel, Robert W. "Health, Nutrition, and Economic Growth." Economic Development and Cultural

Change 52.3 (2004): 643-58.

Hartwig, Jochen. "Is Health Capital Formation Good for Long-term Economic Growth? – Panel Granger-causality Evidence for OECD Countries." Journal of Macroeconomics 32.1 (2010): 314-25.

Heshmati, A. “On the causality between GDP and health care expenditure in augmented Solow growth model.” Stockholm School of Economics Working Paper in Economics and Finance No. 423 (2001)

Keller, Katarina R. I. "Investment In Primary, Secondary, And Higher Education And The Effects On Economic Growth." Contemporary Economic Policy 24.1 (2006): 18-34.

Lucas, Robert E. "On the Mechanics of Economic Development." Journal of Monetary Economics

22.1 (1988): 3-42.

Mehrara, Mohsen, and Maysam Musai. "The Causality Between Health Expenditure and Economic Growth in Iran." Internatonal Journal of Economic and Research 2.4 (2011): 13-19.

Mincer, Jacob. "Human Capital and Economic Growth." Economics of Education Review 3.3 (1984): 195-205.

Odubunmi, Ayoola Sunkanmi, Jimoh Olakunle Saka, and David Mautin Oke. "Testing the Cointegrating Relationship between Health Care Expenditure and Economic Growth in Nigeria." International Journal of Economics and Finance 4.11 (2012): n. pag.

OECD (2014), "Labour Force Statistics: Population and labour force", OECD Employment and Labour Market Statistics (database).

19

Popkin, B. M. "Will China's Nutrition Transition Overwhelm Its Health Care System and Solow Economic Growth?" Health Affairs 27.4 (2008): 1064-076.

Solow, Robert M. "A Contribution to the Theory of Economic Growth." The Quarterly Journal of

Economics 70.1 (1956): 65.

The World Bank." Data | The World Bank

Wang, Kuan-Min. "Health Care Expenditure and Economic Growth: Quantile Panel-type Analysis."

Economic Modelling 28.4 (2011): 1536-549.

Weil, D. N. "Accounting for the Effect Of Health on Economic Growth." The Quarterly Journal of

Economics 122.3 (2007): 1265-306.

Wooldridge, J. M. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT Press. APPENDIX 0 5 10 0 5 10 0 5 10 0 5 10 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010

Australia Austria Belgium Canada Denmark

Finland France Germany Iceland Ireland

Luxembourg Norway Sweden Switzerland United Kingdom

United States the Netherlands

G D P gr ow th (per ce n tag e)

Figure 1: GDP growth given in percentages over time

20 Healt h Car e E xp en d iture p er Capi ta in US d oll ar s

Figure 2: Health Care Expenditure per Capita given over time