Review of the literature on the determinants of healthcare expenditure

Abstract

This study reviews the literature on determinants of healthcare expenditure for the period 1998 to 2007. The methodology combines searches in the MesH database of PubMed with the search in the principal journals of Health Economics. 20 primary studies were found that met the criteria for inclusion. No single pattern of results is clearly identified. Among the 20 articles, four consider income to be the principal determinant of healthcare expenditure, two of them jointly with population ageing. Six highlight population ageing, as against six others that emphasize the proximity to death. The remaining six do not focus on a specific variable, or focus on another variable, e.g. technological progress or territorial decentralization. 11 of the 20 articles calculate the income elasticity of demand, only two of them obtaining a value greater than 1, thus cataloguing healthcare expenditure as a luxury good. There is, therefore, no unanimity in the variables and econometric regressions of healthcare expenditure in Organization for Economic Cooperation and Development (OECD) countries. No solid empirical evidence exists that population ageing is one of the principal determinants of healthcare expenditure, and factors such as technological progress, closeness to death and territorial decentralization of healthcare are increasingly seen as important in the development of explanatory models of healthcare expenditure.

Notes

¹ Wagner's Law or Law of secular growth of the public sector was formulated by its author in 1883, and subsequently in 1911, though it began to arouse interest among economists of the public sector after its rediscovery through the digest Classics in Public Finance by Musgrave and Peacock (1958).

² A disturbance is said to be second-order stationary, if its first- and second-order moments are constant over time. This requires the disturbance to have a constant variance (Guisán, 2002).

³ In the medical area, the different databases generally use the MeSH descriptive terms (controlled vocabulary of the United States National Library of Medicine, which is continually revised and updated). The MesH database includes the ordering of descriptors by major groups of subjects or categories in turn subdivided into more specific subjects or areas (www.pubmed.gov).

⁴ The Boolean search allows complex searches to be created by combining concepts using the Boolean operators ‘OR’, ‘AND’ and ‘NOT’. It is a powerful tool for creating specific searches, and particularly useful when it is desired to expand or limit the search.

⁵ Barros (1998), Di Matteo and Di Matteo (1998), Zweifel et al. (1999), Felder et al. (2000), Karatzas (2000), Roberts (2000), Gianoni and Hitiris (2002), Okunade and Murthy (2002), Herwartz and Theilen (2003), Koening et al. (2003), Clemente et al. (2004), Stearns and Norton (2004), Seshamani and Gray (2004a,b), Di Matteo (2005), Breyer and Felder (2006), Crivelli et al. (2006), Dormont et al. (2006), Mosca (2007), Werblow et al. (2007).

⁶ The Appendix contains a table of these characteristics.

⁷ Regarding the low value of the coefficient of ageing of the population, the author supposes that it may be because the short-term changes in the proportion aged over 65 years are not detected by the model, or because its goodness-of-fit is not good enough.

⁸ See McCoskey and Selden (1998), Karatzas (2000), Roberts (2000) and Herwartz and Theilen (2003).

⁹ Zweifel et al. (1999) introduce the term red herring for the phenomenon that age does not have a significant impact on healthcare expenditure.

¹⁰ In sample 1, for individuals aged over 65, the sex has a coefficient of 1.632, implying that women aged over 65 spend more than men. Having supplementary insurance is significant in both samples, between 0.239 and 0.761, the latter coefficient corresponding to individuals over 65 of sample 1.

¹¹ Other results of the study are as follows: (1) For individuals aged over 65 years, the sex has a coefficient of 1.324. Women aged over 65 years spend more than men. (2) The state subsidy has an indirect relationship with the increase in healthcare expenditure, presenting a coefficient of −0.109 for people over 65 years. (3) Having supplementary insurance is fairly significant, between 0.806 and 0.762 (the latter coefficient corresponding to individuals aged over 65 years).

¹² The discrete, binary and multiple choice models consider the decision-making process as an indirect comparison of utility functions. Two of the models most widely used in the analysis of the demand for healthcare assistance have been the probit and logit models, in which, assuming individuals to be maximizers of their utility, they will choose between two alternatives the one which provides them with the greatest utility. The observed decision reveals which of the alternatives provides most utility, but not their utilities, which are unobservable. For example, let UB_oB, be the utility afforded to the individual by the decision not to demand assistance and UB_1B, the utility of requesting assistance. The probability that Y = 1, i.e. that the individual will decide to demand assistance, will be given by the expression: Pr(Y = 1/x) = Pr(UB_{1 i B} > UB_{o i B}) (Clavero and González, 2005).

¹³ Except Greece, Luxembourg, New Zealand, Portugal and Turkey, as no data exist for these countries.

¹⁴ Not included in the review because it is a working paper on the Canadian healthcare system and therefore does not meet the criteria for inclusion.

¹⁵ To support the thesis of an inefficient use of resources in the public sector, it is mentioned that it is increasingly necessary to devote more resources to this sector without any adequate correlation between the quantity or quality of the services. This phenomenon responds to the so-called ‘Baumol effect’: social sectors are labour-intensive and any increase in productivity in them is therefore lower than in the rest of the economy. In this framework, it is logical to find a growing increase in relative costs.