ABSTRACT
This paper studies the long-run relationship between health care expenditure and income using a panel data set of emerging economies over the period 1995–2012. The results show that expenditure on health care and income are non-stationary and cointegrated. After controlling for cross-sectional dependence and unobserved heterogeneity among different countries, we find that the income elasticity of health care is less than 1, indicating that health care is a necessity and not a luxury. Government expenditure and out-of-pocket expenditure turn out to be important determinants of health care expenditure. Among non-monetary factors, results show that old age dependency and female education seem to have significant bearings on health care expenditures. Policy recommendations suggest that government should increase spending on health care in emerging economies since higher incomes may not automatically translate into higher health care spending by the people of these countries.
Acknowledgements
We would like to thank the anonymous referee for very helpful comments on a previous version of this paper.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 See Pedroni (Citation1999, pp. 659–62) for the testing procedure and complete formulation of test statistics.
2 The within-dimension statistics are constructed by summing numerator terms over the N dimension in the test statistic separately, whereas the between-dimension statistics are constructed by dividing the numerator by denominator prior to summing over N dimension.
3 For detail description of the STATA procedures, see Persyn and Westerlund (Citation2008).
4 For details on the test statistics and their derivation refer to the above reference.
5 The Levin, Lin, and Chu (Citation2002), Maddala and Wu (Citation1999) tests for unit roots results are not reported but are available from the authors. All these tests reject the null of a panel unit root for health care expenditures and income by including both constant and trend.
6 The null hypotheses are same for both IPS and Breitung tests but not the alternative hypothesis. Breitung (Citation2000) proves that his test has better power than IPS in the presence of individual-specific trend.
7 Banerjee, Marcellino, and Osbat (Citation2004) have shown that panel cointegration tests can be largely oversized in the presence of cross-unit long-run relationships. Not accounting for such relationships may provide results in favour of cointegration, which may not be true. It is therefore, appropriate to perform the Westerlund test which accounts for cross-sectional dependencies and provides us correct information about the cointegration.