Elsevier

Journal of Health Economics

Volume 27, Issue 6, December 2008, Pages 1582-1593
Journal of Health Economics

Progressivity, horizontal inequality and reranking caused by health system financing: A decomposition analysis for Switzerland

https://doi.org/10.1016/j.jhealeco.2008.07.009Get rights and content

Abstract

This paper presents an application of the Duclos et al. [Duclos, J.-Y., Jalbert, V., Araar A., 2003. Classical horizontal inequity and reranking: an integrated approach. Research on Economic Inequality 10, 65–100] decomposition to an analysis of the 1998 Swiss health system financing. We see that in addition to measuring horizontal inequality in the classical sense, this decomposition is more efficient and flexible than earlier ones. It is also pointed out that methods involving a nonparametric estimation lead to asymptotically biased vertical and horizontal effects. A procedure to estimate this bias is given. Finally, it is shown that despite a major reform, health system financing is still very regressive and social health insurance is more regressive than direct financing.

Introduction

The Swiss health system experienced a major reform when the Federal Law on Sickness Insurance came into effect in 1996. This reform has promoted competition between health insurers within a strictly regulated framework. Some substantial changes were that basic health insurance became compulsory, premiums stopped being risk-related within cantons, the insured gained the right to freely change their insurer and the range of services covered was significantly extended. Such a substantial reform might have an influence on equity and it is essential to analyze potential changes in that important end-goal of the health system. Leu and Shellhorn, 2004a, Leu and Shellhorn, 2004b have studied equity in health status and health-care utilization in Switzerland with data from 1982, 1992, 1997 and 2002. Their results indicate the usual positive relationship between income and health, but also that the distribution of health is among the least unequal in Europe. Concerning health care utilization, they measured little or no inequity except with respect to specialist visits, which are clearly “pro-rich” distributed as in most other OECD countries. Equity in health system financing has also been analyzed in Switzerland with data from 1982 and 1992 Wagstaff and van Doorslaer, 1992, Wagstaff et al., 1999, van Doorslaer et al., 1999, but no comprehensive study has been carried out since the reform. This is a substantial lack because, in a country where the population health state is good1 and fairly equitably distributed and where health care quality is satisfactory (FSIO, 2001) and to a large extent accessible to everyone, the main concern is the steady increase in cost of the health system and its financing. This paper attempts to fill this gap with an analysis of equity in post-reform health system financing.

The view that the health system should be financed by households according to their ability-to-pay has widespread support amongst policy-makers in OECD countries (Wagstaff and van Doorslaer, 2000). Therefore health system financing should not be linked to utilization, and the distribution of household contributions has to be seen as an independent policy choice, consequences of which should be examined separately. It is thus of primary interest to policy-makers to measure to what extent, and how, financing is related to household ability-to-pay. Wagstaff et al. (1989) were the first to quantify the progressivity or regressivity of health system financing sources by using Kakwani’s index (Kakwani, 1977). Later, Wagstaff and van Doorslaer (1997) gained further insight by applying the Aronson et al. (1994) (hereafter AJL) decomposition method, which makes it possible to decompose the change in income inequality caused by financing into a vertical, horizontal and reranking effect; each effect corresponding to a different dimension of equity. The vertical effect shows how households with different incomes are affected by the financing, the horizontal effect measures the inequality generated among households with the same pre-financing income, while the reranking effect quantifies the change in the order of income distribution. By simultaneously revealing these three different dimensions of equity, the AJL decomposition constitutes a useful tool for assessing the fairness of health system financing.

It is however important to note that the AJL decomposition requires observing households with the exact same pre-financing income in order to measure horizontal inequality. In practice, there are not enough such exact pre-financing equals and AJL suggest choosing a bandwidth so as to group near-equal households. This grouping constitutes a major drawback of the method, first because horizontal inequality is no longer defined in the classical sense. Moreover, the grouping adversely affects the measurement of all decomposition effects, and estimation of them depends on the arbitrary bandwidth choice. Dependence on bandwidth choice is particularly significant in the case of horizontal inequality and reranking. Furthermore, van de Ven et al. (2001) (hereafter VCL) also showed that the expression of the AJL decomposition is not adequate when near-equal households are grouped, and proposed a criterion for choosing the bandwidth. However, this criterion requires financing to be progressive, which is not the case with many financing sources of the health system. Moreover, VCL do not solve the theoretical problem raised by grouping, which is not to measure the horizontal inequality in the classical sense. In fact, a fully satisfactory measurement of horizontal inequality did not exist until recently when Duclos and Lambert (2000) found a new way of evaluating it by means of a continuous method involving a nonparametric estimation. Duclos et al. (2003) (hereafter DJA) finally presented a decomposition of the redistributive effect encompassing that new way of measuring horizontal inequality. One original feature of this study is the application of the DJA decomposition to the analysis of health system financing.

The rest of the paper is organized as follows. Section 2 presents AJL, VCL and DJA decomposition s and a discussion of the main links and differences between them. Section 3 gives an overview of the Swiss health system by presenting the economic agents financing it. Section 4 gives the definition of the 10 basic financing sources taken into account as well as the computational methods used. Particular attention is paid to the issue of DJA nonparametric method choice, the estimation of its asymptotic bias, and of the resulting one on estimating vertical and horizontal effects. Section 5 starts by illustrating the influence of bandwidth choice on the measurement of AJL and VCL decomposition effects. The analysis of the Swiss health system financing by means of the DJA decomposition is then presented. Finally, Section 6 discusses the empirical results and methodological implications of using DJA method instead of AJL’s when studying equity in health system financing.

Section snippets

The AJL decomposition of the redistributive effect

The overall influence of a given financing on income distribution can be measured by means of the redistributive effect RE which is the difference between the pre- and post-financing inequality indexes (Reynolds and Smolensky, 1977). AJL showed that it can be decomposed as follows:RE=VHR,where the horizontal effect H and reranking effect R respectively correspond to classically defined horizontal inequality and reranking caused by the financing. These two effects can never be negative. As for

Swiss health care financing system

In 1998, the total cost of the Swiss health system was CHF 40.3 billion,5 representing 10.5% of GDP and placing Switzerland second only to the USA in the ranking of OECD countries. Social insurances contribute 39.7% and comprise social health (SHI), old age and invalidity, accident, and military insurance. SHI is the main financing source with 31.8% of total costs.

Definition and computation of health system financing sources

The microdata come from the 1998 Swiss Household Income and Expenditure Survey (SHIES) undertaken by SFSO. There are more recent surveys available but the 1998 one has the advantage of recording health care expenditures and health insurance reimbursements over a period of 1 year. This survey gives information about income, consumption and other characteristics of 9295 Swiss households. SHIES, like all other such surveys, does not enable us to directly observe all financing sources, and proxy

Influence of bandwidth choice on measurement of AJL and VCL effects

As discussed in Section 2, AJL and VCL decompositions require choosing a gross income bandwidth which has an incidence on measurement of the effects. An illustration of this influence is provided here for total health system financing which has the particularity of being regressive. Fig. 1 shows the dependence of AJL and VCL vertical effects on bandwidth choice.9

Discussion

This paper starts by presenting AJL, VCL and DJA decompositions and then discusses the advantages of the latter for the analysis of health system financing. It appears that the DJA decomposition is the first to measure horizontal inequality – defined in the classical sense – in the absence of exact pre-financing equals, and also makes possible the setting of realistic social preferences in terms of equity. Here it is shown, both in theory and by an empirical illustration, that the VCL bandwidth

Acknowledgements

I thank Theo Gasser, Walter Koehler, Alois Kneip and Eva Herrmann for their Fortran algorithm “lokern” which estimates the expected net income function and its 4th derivative. I am also most grateful to Jean-Paul Chaze for his helpful comments as well as to Gabrielle Antille Gaillard and Yves Flückiger for their ongoing support. Finally, I also would like to thank two anonymous reviewers for their useful comments on previous versions of this paper.

References (30)

  • J.-Y. Duclos

    Gini indices and the redistribution of income

    International Tax and Public Finance

    (2000)
  • J.-Y. Duclos et al.

    A normative approach to measuring classical horizontal inequity

    Canadian Journal of Economics-La Revue canadienne de science économique

    (2000)
  • J.-Y. Duclos et al.

    Classical horizontal inequity and reranking: an integrated approach

    Research on Economic Inequality

    (2003)
  • V. Epanechnikov

    Nonparametric estimates of a multivariate probablity density

    Theory of Probability and its Applications

    (1969)
  • J. Fan

    Design-adaptive nonparametric regression

    Journal of the American Statistical Association

    (1992)
  • Cited by (0)

    This paper derives from the joint IRIS research program of the universities of Geneva and Lausanne and the federal polytechnic school of Lausanne. The program is funded by the CUS and by said institutions.

    View full text