The influence of health care financing systems on health expenditure growth: a
comparison of the Bismarck model and the Beveridge model
Max Go June 2018
Abstract
European countries have experienced substantial health care expenditure growth during the past few decades. The purpose of this paper is to determine whether the type of health financing system that is used in a country has influence on health expenditure growth. This is accomplished by comparing the Bismarck model and the Beveridge model. The income elasticity of demand for health care is estimated for 16 OECD countries using panel data analysis. The results indicate that health care is more income elastic in countries that use the Bismarck model to finance health care. This finding indicates that the Beveridge model may be more cost efficient, but additional research on the non-financial aspects of health care is required in order to reach a definitive conclusion on this matter. The proposed strong relationship between health expenditure and per capita GDP is confirmed and this paper also finds a significant effect of population aging on health expenditure growth.
Bachelor thesis
University of Amsterdam Economics
Author: Max Go
Student number: 10760598 Supervisor: Timo Klein
1
Statement of Originality
This document is written by Student Max Go who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no
sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of
2
1. Introduction
Health care expenditure in OECD countries has been rising rapidly for the last decades and it is expected that expenditure will continue to rise. In 2016, the United States has spent the most on health care of all OECD countries. It is estimated that the United States has spent USD 9 892 on health care per capita, which is more than twice the average amount spent on health care (OECD, 2017). European countries have also experienced a large increase in health care spending. Because health care spending is such a large fraction of GDP, it is important for policymakers to understand what factors actually drive this growth in health care expenditure. Health care competes with other important expenses in a country, such as spending on education, the military, infrastructure and other parts of the economy in which we tend to invest a large fraction of our income. A thorough understanding of the determinants of health care expenditure growth could help manage the costs in the future. Previous literature has repeatedly found GDP growth to be a driver of health expenditure growth. One issue on which researchers have not reached consensus yet is how strong this effect of GDP growth on health expenditure actually is. This can be estimated by calculating the income elasticity of demand for health care. The problem is that research on the income elasticity has led to contradictory conclusions. There are researchers who estimate the elasticity to be around unity, while others estimate the elasticity to be well above unity. When a good has an income elasticity that is greater than unity, then that good can be defined as a luxury good. We can expect health care to become an increasingly larger share of GDP if health care can indeed be classified as a luxury good. One goal of this paper is therefore to estimate the income elasticity of demand for health care and determine whether health care is a luxury good.
Another factor that might have influence on the national level of health spending is the type of health financing system that is implemented in a country. There are two dominant health financing systems that are used in European countries. The Beveridge model uses general taxation to finance health care while the Bismarck model uses an insurance system to finance health care. It may be expected that the Beveridge model promotes greater cost control as the government is the sole payer in the health care system. The purpose of this paper is to estimate the income elasticity of demand for health care with the use of panel data and to determine whether the elasticity is dependent on the type of financing system that is used. The regression results indicate that the income elasticity is lower in countries that use the
3
Beveridge model to finance health care. This may be indicative that health care can be more cost efficient when it is paid for by a government through general taxation. One problem with this analysis however, is that only the financial aspects of health care are taken into consideration. Additional research is thus required to reach a definite conclusion on the cost effectiveness of health financing systems. This paper first provides an overview of previous literature on health expenditure growth in section 2. A description of the research method and data is given in section 3. The results can be found in section 4 and a discussion of these results is provided in section 5. Section 6 concludes.
2. Literature review
This section provides an overview of previous literature on health expenditure growth. The first paragraphs describe the most commonly proposed factors that could influence health expenditure. These factors are the effect of GDP growth, the effect of aging and the effect of technological progress. The following paragraphs provide a brief description of the Bismarck model and the Beveridge model as well as a summary of previous research that was concerned with the differences between these financing systems.
Previous literature on health care expenditure growth has consistently found a strong positive relationship between per capita GDP and per capita health expenditure. Newhouse (1977) collects data about health care expenditure and income for 13 developed countries in the early 1970s. Using this data, Newhouse estimates the income elasticities of demand for health care for these countries. He finds that income elasticities are in the range 1.15 - 1.31 and that 90 per cent of the variation in health expenditure can be explained by variation in GDP. Because the income elasticities exceed one, Newhouse argues that medical expenditure is technically a luxury good. That is, the fraction of income spent on health care should become increasingly larger as income rises. This is exactly what we observe forty years after Newhouse first published his findings.
Newhouse's conjecture that health care is a luxury good has proven to be quite controversial. Research on the income elasticity of demand for health care has often lead to contradictory conclusions. Hitiris and Posnett (1992) use panel data to estimate the income elasticity for 28 OECD countries. Using this larger sample, their analysis confirms that there is a positive relationship between GDP and health expenditure. Although they estimate the income elasticity to be around unity, which is significantly less elastic than Newhouse suggests. More recent examples include Hall and Jones (2007). Their analysis also leads to the conclusion
4
that the rise in health care expenditure is a rational response to increases in income. Hall and Jones argue that the share of income that is spent on health is increasing because the marginal utility of life extension does not decline, whereas the marginal utility of consumption declines rapidly. Hall and Jones find that the demand for health care is income elastic, with a value that is greater than unity. According to their projections, which are based on a quantitative analysis of an econometric model, the share of GDP devoted to health care will exceed 30 per cent by the year 2050 in the United States (Hall and Jones, 2007).
Another variable that has attracted much attention in health expenditure growth research is aging of the population. It may seem reasonable to assume that health expenditure increases as people are getting older. Literature on this topic however, has failed to confirm this assumption. Barros (1998) evaluates the effect of aging on health expenditure for 24 OECD countries. He concludes that the age structure of the population is not a factor that can explain growth in health expenditure. Reinhardt (2003) finds that most of the growth in health care spending was attributed to factors that increase spending for all age groups. He identifies, again, the rise in GDP per capita as an important driver of health expenditure growth as well as the availability of new medical technology.
Okunade and Murthy (2002) find a significant relationship between health expenditure and changes in R&D spending, which they use as a proxy for technological advancement. Their analysis, however, focuses only on health expenditure in the Unites States. This makes it difficult to accept technological progress as a driver for health expenditure growth. Hall and Jones (2007) argue that technological progress in itself cannot explain health expenditure growth because new technologies do not necessarily have to be implemented after their invention. They point out that the share of GDP devoted to health care has increased in almost all developed countries and question why the United States invests that much in new medical technologies (Hall and Jones, 2007). White (2007) decomposes health expenditure growth into economic growth, population aging and excess growth. He argues that there is some evidence that could support the technological argument as he identifies similarities in excess growth among countries that could be explained by the fact that technology is shared fairly freely among developed countries. However, in the last few decades he finds large differences in excess growth between the United States and other countries. These results indicate that institutional factors may also be important determinants of health expenditure growth rates (White, 2007).
5
Most literature about health expenditure growth so far has primarily focussed on the effects of GDP growth and the aging of the population. There has been relatively little discussion about the institutional differences between countries. When analysing health expenditure growth rates, it would make sense to study the way health care is financed in a country. There are multiple models of health care financing, but this paper focuses primarily on health care systems in Europe. Health financing systems in Europe can be divided into two broad categories (Van der Zee and Kroneman, 2007). The first financing system is the Beveridge model. This model was first implemented in the United Kingdom and is also used in Scandinavian countries and southern European countries such as Italy and Spain. In these countries, health care is treated the same as any other public service. The governments of countries that use a Beveridge system are responsible for providing and financing health care. Health care is paid for through general taxation and is supplied by many public providers (Lameire and Wiedemann, 1999). Because the governments of countries that use the Beveridge model are generally the only payers of health care, it is frequently argued that Beveridge systems may be better at containing health care costs (Or et al., 2010).
The second financing system is the Bismarck system, which originated in Germany and is used mostly in central European countries such as the Netherlands, France and Luxembourg. The Bismarck model uses an insurance system to finance health care. It is commonly called a mixed model because there are both public and private providers of health care. There are some minor differences among countries that use the Bismarck model to finance health care. For example, in Germany, health insurance is usually paid for by employees and employers through payroll reduction while in Switzerland citizens are mandated to buy insurance policies by themselves.
Elola (1996) uses data from the OECD to compare European countries on health expenditure, health outcomes and how satisfied citizens are with their national health care system. He finds that the main advantages of using a Beveridge model are superior cost control and a higher level of equity, but the countries from his sample that used the Bismarck model showed higher levels of patient satisfaction. Elola argues that health care system reforms in the future should primarily focus on diminishing this trade-off between cost efficiency and patient satisfaction. One problem to note about this research is that Elola only uses data from 1992. This makes it questionable whether Elola's conclusion is still valid a few decades after he published his findings.
6
Saltman and Figueras' (2004) analysis is similar to the research conducted by Elola (1996) in the sense that they attempt to include all aspects of health care in their analysis. The variables they use to make an assessment of health systems include the health of the population, patient satisfaction, health expenditure and many other variables that can be linked to health. Their results suggest that the type of health system that is used in a country does not have a significant impact on health outcomes. However, they do find that the level of patient satisfaction is slightly higher in countries with a Bismarck system. Their analysis of health expenditure suggests that countries with Beveridge models have lower costs and lower health expenditure growth rates. Their finding that there is a trade-off between satisfaction and cost control is in line with the conclusions of Elola (1996). Because they have discussed all aspects of health care, they find it difficult to make a judgement about which health system performs better. Saltman and Figueras end their discussion with the question whether the additional satisfaction in countries with a Bismarck system is worth the extra resources spent on health care. Health outcomes do not differ under both systems according to their study, so the higher observed level of satisfaction is not a result of better quality of health care (Saltman and Figueras, 2004).
Examples of more recent papers about the differences between the Bismarck and Beveridge model include Van der Zee and Kroneman (2007) and Or et al. (2010). These papers are similar to earlier literature because the performance of both models is evaluated on a broad range of variables. Van der Zee and Kroneman (2007) report that health outcomes in terms of mortality rates are slightly better in countries with Bismarck models. They find that the share of GDP that was spent on health care has increased more in Bismarck countries, which leads them to the conclusion that the Beveridge model is more cost effective than the Bismarck model. Or et al. (2010) also report that the share of GDP spent on health care is lower in countries with Bismarck systems. However, they do not find significant differences in health outcomes.
Most researchers attempt to cover all aspects of health care systems. This induces them to include as many variables as they can find in order to try to see the whole picture. As a result, a lot of papers include variables that are probably not that relevant when comparing health financing systems. For example, Or et al. (2010) include data about prostate cancer survival rates in their analysis. Although it may be possible to argue that these are important statistics when discussing health, it is questionable whether the inclusion of such variables improves our knowledge of health financing systems. Health outcomes should not be dependent on the
7
way health care is financed and are possibly more influenced by other country characteristics. For instance, the quality of education or the eating habits of the population. The main difference between the Bismarck model and Beveridge model is the way in which health care is financed. This implies that analyses of differences between these models should primarily focus on the financial aspects of health care. Similarly, both Elola (1996) and Saltman and Figueras (2004) put a lot of emphasis on consumer satisfaction. This is also not a variable that is easily attributable to financing systems and could possibly be explained better on account of cultural differences. Researchers tend to include many aspects of health care that are outside the field of economics. These practices are apprehensible as there are so many aspects of national health care, but it does shift our focus away from the financial aspect of health financing systems.
The purpose of this paper is to provide a comparison of the financial outcomes of the Bismarck model and the Beveridge model. The central question that arises after reviewing all literature is whether the Beveridge model is indeed superior in terms of cost containment. This question will be answered by reassessing Newhouse's original conjecture that health care is a luxury good. By estimating the income elasticities of health care of countries with either a Bismarck or a Beveridge model using the most up to date relevant data and comparing them, it should become clear whether there is a model that promotes greater health care cost control. One goal of this paper is to eventually reach a clear conclusion about the cost effectiveness of health financing systems. For that reason, aspects of health care other than financially related variables, such as health outcomes and patient satisfaction, will not be taken into consideration.
3. Data and method
All data is collected from the OECD database. The variables that are used for this paper are health expenditure per capita (HE), GDP per capita (GDP) and the fraction of the population that is older than 65 years, measured in percentage of the total population (Pop65). Health expenditure is defined by the OECD as the final health consumption of the population. Examples of variables that are not part of final health consumption include investments, research and development spending as well as any spending on health related education (OECD, 2011).
8
The sample consists of 16 European countries that use either a Bismarck system or a Beveridge system to finance health care. Luxembourg is omitted from the analysis due to too much missing data. The data is collected for the years 1990 until 2016. For this period, all data is available for each country except for one missing year for Germany1. Greece, Portugal, Italy and Spain have transitioned from the Bismarck model to the Beveridge model. The last transition occurred in Spain in the year 1985 (Van der Zee and Kroneman, 2007). By restricting the analysis to the period 1990 - 2016, possible transition effects are mitigated. An overview of the included countries and descriptive statistics are given in Table 1.
All amounts are converted by the OECD to U.S. dollars using Purchasing Power Parities (PPP) exchange rates. Hitiris and Posnett (1992) argue that the use of PPP is generally preferred over the use of exchange rates because exchange rates do not equalize the purchasing power of different currencies. All amounts have been adjusted for inflation using 2010 as the base year. The variables HE, GDP and Pop65 were converted into natural logs. The resulting new variables are called lnHE, lnGDP, and lnPop65. The variable of interest is the income elasticity of demand for health care. Because both variables are expressed in natural logarithms, the resulting coefficients can be interpreted as elasticities.
The purpose of this paper is to estimate the income elasticity of demand for health care in Europe and to determine whether this elasticity is dependent on the type of health financing system that is used in a country. Previous research on the income elasticity has yielded contradictory results and Newhouse's conjecture that health care is a luxury good is still controversial. A good is defined as a luxury good if the income elasticity of that good has a value above unity. The first hypothesis is therefore that the income elasticity is greater than unity.
The first hypothesis will be tested by running a panel data regression of lnHE on lnGDP, lnPop65 and country specific dummy variables. The country dummy variables are added in order to mitigate omitted variables bias that could arise from unobserved variables specific to
1
9
each country. This regression model is outlined in equation (1). The coefficient β1 in equation
(1) can be interpreted as the income elasticity.
( )
Previous literature on the difference between the Bismarck and the Beveridge model regarding cost containment has frequently found that the Beveridge model promotes greater cost control. This implies that health care should be more income elastic in countries with a Bismarck system. The second hypothesis is therefore that the income elasticity is higher in countries with a Bismarck system than in countries with a Beveridge system.
The second hypothesis will be tested by running an OLS regression of lnHE on lnGDP, lnPop65 and a dummy variable Bismarck. This dummy variable has a value of 1 if a country uses the Bismarck model and a value of 0 otherwise. This regression model is outlined in equation (2). The coefficient β1 in equation (2) can again be interpreted as the income
elasticity.
( )
Both equation (1) and equation (2) can additionally be extended by controlling for year fixed effects. Similar to the country dummy variables in equation (1), the addition of year dummy variables could mitigate omitted variables bias because there could be unobserved variables specific to each year. The regression models are modified in order to determine whether controlling for year fixed effects changes the estimated relationship between health expenditure and GDP. The resulting new models are outlined in equations (1a) and (2a),
10
where α and λ represent the set of country dummy variables and year dummy variables, respectively.
( )
( )
Table 1: Health Expenditure and GDP per capita in U.S dollars, 1990 and 2016
4. Results
Table 2 shows the regression output of equation (1). In the first column, the output of the simple regression without country dummy variables is given. These dummy variables are added to the regression in the second column. The asterisks after the coefficients of lnGDP indicate the p-value of the one-tailed test that the coefficient is equal to one. All other coefficients in the following tables are tested for the null hypothesis that the coefficients are equal to zero. The simple regression of lnHE on lnGDP and lnPop65 yields an income
11
elasticity estimate of 1.323. This estimate becomes significantly higher at 1.518 when the country specific dummy variables are added. Both estimates are significantly higher than one with p<0.01. If the income elasticity is 1.518, we would expect an increase of 15.2% in health expenditure if GDP increases by 10%. The hypothesis that the income elasticity is greater than unity cannot be rejected based on this regression output. These results imply that health care is technically a luxury good and confirm the findings of Newhouse (1977) and Hall and Jones (2007). The country dummy variables are all statistically significant except for the dummy for Greece. A country dummy for Austria was excluded from this regression in order to avoid the dummy variable trap. As a result, Austria serves as a reference value and the country dummy coefficients reflect changes from this reference value. For example, the model predicts that health expenditure is 100*.112=11.2 per cent higher in France relative to Austria when accounting for GDP per capita and the age structure of the population. Although the dummy variables coefficients are not that straightforward to interpret, the statistical significance of these estimates does show that there are unobserved variables other than GDP and age structure that influence health expenditure growth.
12
Table 3 shows the regression output of equation (2). The regression of lnHE on lnGDP, lnPop65 and Bismarck estimates that the income elasticity is 1.251 with p<0.01 in all countries. This regression is thus also supportive of the first hypothesis that the income elasticity is greater than unity. The coefficient of the Bismarck dummy is 0.134 and is
significant at the one per cent level. Based on this model, we would expect health expenditure to be 13.4% higher in countries with a Bismarck system after accounting for variation in GDP per capita and the age structure of the population. This implies that the demand for health care is more income elastic in countries with a Bismarck system. The second hypothesis can thus not be rejected based on these results. The findings of Elola (1996) and Saltman and Figueras (2004) regarding cost efficiency of health financing systems are confirmed by this analysis of a more recent sample.
The regression output of both equation (1) and equation (2) show that there is a significant effect of aging of the population. Equation (1) produces a coefficient estimate of 0.534 for lnPop65 and equation (2) produces an estimate of 0.565. Both coefficients are statistically significant at the one per cent level. These models predict that health expenditure increases by roughly 0.5% if the proportion of the population that is older than 65 years increases by one per cent. This finding is rather surprising given that previous research has frequently failed to find a significant relationship between health expenditure growth and aging of the population.
²
In order to test whether these regression results are robust, the specifications are adjusted to account for year fixed effects. The inclusion of year fixed effects could possibly mitigate omitted variable bias if there are unobserved variables specific to each year. The results of these regressions are given in table 4. The regression of equation (1a) produces an elasticity estimate of 1.056 with p=0.138. This estimate is not significantly higher than one and the first
13
hypothesis would thus be rejected based on this regression output. The regression of equation (2a) produces an elasticity estimate of 1.057 with p=0.008. This estimate is significantly higher than one and is thus supportive of the first hypothesis. The income elasticity estimates become significantly lower when controlling for year fixed effects. This indicates that there are unobserved variables specific to each year that possibly leads to unobserved variable bias when using the other specifications to estimate the income elasticity. The model that includes both year fixed effects and country fixed effects (1a) seems to explain the variation in health expenditure better than the other specifications. This regression produced the highest R² with a value of 0.966. The question that arises here is whether it is an appropriate measure to include both year fixed effects and country fixed effects. According to Roberts (1999), health expenditure research is an area in which theory provides very little guidance in specifying the correct models. The fact that the estimated income elasticity becomes substantially lower when controlling for year fixed effects confirms her finding that income elasticity estimates are very sensitive to model specification and sample selection (Roberts, 1999).
²
5. Discussion
All countries that are included in this analysis have experienced a vast increase in health expenditure. There has been a long going debate about what factors actually contribute to this growth in health expenditure. The results from this paper can confirm that there is a strong positive relationship between health expenditure and per capita GDP. There is some evidence that the age structure of the population has influence on health expenditure growth, although research on this matter is far from conclusive. It should be noted that evaluating the effect of aging on health expenditure is not a primary goal of this paper and as a result, only one
14
simple proxy has been used to include the effect of aging in the analysis. Additional research is required in order to answer the question whether aging is a driver of health expenditure growth.
The purpose of this paper is to estimate the income elasticity of demand for health care in Europe and to identify differences in the cost efficiency of health financing systems. Similar to previous research, the regressions in this paper produce contradictory elasticity estimates because of the use of multiple model specifications. The overall elasticity is estimated to be in the range 1.056 - 1.518. The model that includes both year fixed effects and country fixed effects produces an elasticity estimate that is not significantly greater than unity. While all other models provide evidence that the income elasticity is greater than unity. The regression output of these models support the hypothesis that health care is a luxury good. This means that there are reasons to expect that the share of GDP devoted to health care will become increasingly larger if these models are correctly specified. It is, however, not clear which specification is appropriate to use and the question whether health care is a luxury good cannot be answered due to the significant differences in elasticity estimates. This paper confirms Robert's (1999) assertion that income elasticity estimates are very dependent on model specification. This could provide an explanation of the contradictory income elasticity estimates of previous research. One of the challenges of future research on the income elasticity of demand will be to determine which model is appropriate to use.
One problem with this research is that it is likely that there is omitted variable bias. Omitted variable bias affects the expected value of the coefficient of lnGDP if there are unobserved variables that are correlated with both lnHE and lnGDP. There are probably many variables that are correlated with national health expenditure and national income as these variables cover the expenses for the whole population. An example in this case would be spending on education because the educational level of an individual is likely to affect both his or her income and health. Another problem is the possibility of multicollinearity. Multicollinearity leads to an increase in the standard error for the coefficient of lnGDP if there are variables included that affect lnGDP. A consequence of multicollinearity could be that the larger standard errors lead to a false rejection of the null hypothesis. This could be a possible explanation of the contradictory regression output of equation (1a).
Another problem has to do with the direction of causality. This paper is mainly concerned with the effect of GDP on health expenditure. The problem is that economic theory also
15
provides arguments that there could be a causal relationship in the other direction. That is, GDP could also be a function of the level of health expenditure. For example, an increase in health expenditure could lead to better health and subsequently higher productivity, which could raise GDP. No measures were taken to test for reverse causality for this paper. Future research might want to focus on determining whether health expenditure has a causal effect on GDP. This could be accomplished by performing a Granger causality test or via the instrumental variables approach.
The results of the second regression indicate that health care expenditure is less income elastic in countries in which a Beveridge system is used to finance health care. One problem with this paper however, is that aspects of health care other than the financial aspect are not included in the analysis. As a consequence, it is still unknown whether the higher costs of Bismarck systems have positive implications such as better health outcomes or a higher level of patient satisfaction. It is therefore not clear whether the Beveridge model is actually more cost efficient than the Bismarck model. Another shortcoming of this research has to do with the collection of data. The OECD uses a number of different methodologies to estimate health expenditure depending on the data that is available about each country. The possibility that this inconsistency in expenditure estimation has affected the results should not be rejected. A suggestion for future researchers who wish to identify differences between the Bismarck and the Beveridge model would be to collect data from countries in which a transition from the Bismarck model to the Beveridge model has occurred. A separate analysis of these countries could provide an interesting comparison of both financing models as other country related variables can be held fixed. An alternative approach would be to use a difference-in-differences estimator, which could mitigate biases that arise from permanent differences between countries.
6. Conclusion
The main conclusion that emerges from this analysis of health care expenditure is that the demand for health care is more income elastic in countries that use the Bismarck model to finance health care. The question whether the overall income elasticity is greater than unity can unfortunately not be answered because the elasticity estimates have proven to be susceptible to model specification. This analysis of 16 developed European countries does confirm that there is a strong relationship between health care expenditure and per capita GDP. Furthermore, there is some evidence that the level of national health spending is partly
16
dependent on the age structure of the population. Only the financial aspects of health financing systems were considered for this study. Variables such as health outcomes and patient satisfaction were omitted from the analysis. It is therefore not possible to declare one health system to be the best performer overall in Europe. Although this paper and other literature of health financing systems may provide guidance to policymakers in developing countries where national health care systems are not yet well established. Future research on health financing systems might focus on countries where a transition from one system to the other has occurred.
17
References
Barros, P. P. (1998). The black box of health care expenditure growth determinants. Health economics, 7(6), 533-544.
Elola, J. (1996). Health care system reforms in western European countries: the relevance of health care organization. International journal of health services, 26(2), 239-251. Figueras, J., Saltman, R. B., Busse, R., & Dubois, H. F. (2004). Patterns and performance in
social health insurance systems. Series editors’ introduction, 81.
Hall, R. E., & Jones, C. I. (2007). The value of life and the rise in health spending. The Quarterly Journal of Economics, 122(1), 39-72.
Hitiris, T., & Posnett, J. (1992). The determinants and effects of health expenditure in developed countries. Journal of health economics, 11(2), 173-181.
Lameire, N., Joffe, P., & Wiedemann, M. (1999). Healthcare systems—an international review: an overview. Nephrology Dialysis Transplantation, 14(suppl_6), 3-9.
Newhouse, J. P. (1977). Medical-care expenditure: a cross-national survey. The Journal of Human Resources, 12(1), 115-125.
OECD (2011). A System of Health Accounts 2011, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264270985-en.
OECD (2017). Health at a Glance 2017: OECD Indicators, OECD Publishing, Paris, http://dx.doi.org/10.1787/health_glance-2017-en.
Okunade, A. A., & Murthy, V. N. (2002). Technology as a ‘major driver’of health care costs: a cointegration analysis of the Newhouse conjecture. Journal of health economics, 21(1), 147-159.
Or, Z., Cases, C., Lisac, M., Vrangbæk, K., Winblad, U., & Bevan, G. (2010). Are health problems systemic? Politics of access and choice under Beveridge and Bismarck systems. Health Economics, Policy and Law, 5(3), 269-293.
Reinhardt, U. E. (2003). Does the aging of the population really drive the demand for health care?. Health Affairs, 22(6), 27-39.
Roberts, J. (1999). Sensitivity of elasticity estimates for OECD health care spending: analysis of a dynamic heterogeneous data field. Health Economics, 8(5), 459-472.
Van der Zee, J., & Kroneman, M. W. (2007). Bismarck or Beveridge: a beauty contest between dinosaurs. BMC health services research, 7(1), 94.
White, C. (2007). Health care spending growth: how different is the United States from the rest of the OECD?. Health Affairs, 26(1), 154-161.
18
19
