Can Baumol's model of unbalanced growth contribute to explaining the secular rise in health care expenditure? An alternative test

Abstract

In a recent paper, I argued that Baumol's model of ‘unbalanced growth’ offers a ready explanation for the observed secular rise in Health Care Expenditure (HCE) in rich countries. Baumol's model implies that HCE is driven by wage increases in excess of productivity growth. I tested this hypothesis empirically, using data from a panel of 19 Organization for Economic Cooperation and Development (OECD) countries and found robust evidence in favour of Baumol's theory. An alternative way to test Baumol's theory is to check whether its implication that variations in the relative price of medical care contribute significantly to explaining variations in health expenditure in the same direction has an empirical grounding. Earlier studies, although mostly not in an explicit attempt to test Baumol's theory, have occasionally rejected this hypothesis. Despite poor data quality of the available medical price indices, I perform the alternative test using data for nine OECD countries. My findings suggest that the relative price of medical care is in fact a statistically significant explanatory variable for health expenditure, thus lending support to Baumol's theory.

Acknowledgements

I wish to thank Sheila D. Smith, Jan-Egbert Sturm and Philippe Ulmann for their thoughtful comments on earlier work that have inspired this article. Also, I would like to thank two referees of Applied Economics for their proposals on how to improve the present article. The usual disclaimer applies.

Notes

¹ Cf Gerdtham and Jönsson (2000) for a review of the literature up to the late 1990s and Okunade et al. (2004) for a recent contribution.

² Recently, the OECD (2006) as well as Nixon and Ulmann (2006) have considered Baumol's approach, but, again, without testing it.

³ A technical discussion of Baumol's model can be found in Hartwig (2008). It need not be repeated here.

⁴ Baumol affirms this with respect to computers. He writes: ‘Despite the use of the computer in medicine …, there is no substitute for the personal attention of a physician…’ (Baumol, 1967, p. 423).

⁵ Okunade and Karakus (2001) find HCE to be I(2) for some countries, but this is probably due to the noninclusion of a deterministic time trend in their Augmented Dickey–Fuller (ADF) regressions. Hansen and King (1998) advocate the inclusion of a time trend since HCE are clearly trended. For the four variables inspected in this article, panel unit root tests reject the hypothesis that they are I(2). Test results are documented in the Appendix. For a detailed account of panel unit root tests, cf. Breitung and Pesaran (2008).

⁶ Zweifel et al. (1999) have disputed the nexus between population ageing and HCE growth. Their results have been challenged in turn by Seshamani and Gray (2004a, b).