Equity in health care finance in Palestine: The triple effects revealed

Abstract

This paper presents an application of the Urban and Lambert “upgraded-AJL Decomposition” approach that was designed to deal with the problem of close-income equals in equity analysis, and as applied to the area of health care finance. Contrary to most previous studies, vertical and horizontal inequities and the triple effects of inter-groups, intra-group and entire-group reranking of various financing schemes are estimated, with statistical significance calculated using the bootstrap method. Application is made on the three financing schemes present in the case of the Occupied Palestinian Territory. Results demonstrate the relative importance of the three forms of reranking in determining overall inequality. The paper offers policy recommendations to limit the existing inequalities in the system and to enhance the capacity of the governmental insurance scheme.

Introduction

Assessing the impact of health care finance on income inequality is a relatively new area of analysis in the context of both developing and developed countries (Wagstaff, 2002). Worldwide empirical evidence has already shown that different health care financing schemes may very differently affect the prevailing income distribution of a country, and consequently, the associated level of overall income inequality (van Doorslaer et al., 1999, Wagstaff, 2002). The distributional impact of health care financing can be inferred from measures of vertical inequity; e.g., progressivity analysis. However, as stated by Wagstaff and van Doorslaer (1997): “depending on the extent of horizontal inequity and reranking involved in health care finance, a progressivity analysis can give a misleading impression about the income redistribution associated with the financing system”. These three dimensions of inequity – relating respectively to violations of the normative principles of “unequal treatment of unequals”, “equal treatment of equals” and “proper treatment of unequals” – are generally addressed by decomposing the redistributive effect (RE) induced by financing into a vertical, horizontal and reranking effect (Aronson and Lambert, 1994). Applied to health care finance, the vertical effect (VE) measures how and to what extent individuals of unequal ability-to-pay are affected by the financing, the horizontal effect (HE) captures the extent to which individuals of equal ability-to-pay make unequal contributions to health care, whereas the reranking (RR) quantifies changes in ranking-order of individuals (by income) following health payments (O’Donnell et al., 2007).

Empirical studies conducted in the context of developed countries (Wagstaff and van Doorslaer, 1997, van Doorslaer et al., 1999) have demonstrated that different forms of health care financing may indeed be associated with both horizontal and reranking effects. This is even more likely in the context of developing countries, where income protection mechanisms are still far underdeveloped (Sekhri and Savedoff, 2005, Pauly et al., 2006), and where high proportions of health care expenditures are funded by households’ direct out-of-pocket payments (Musgrove et al., 2002). Since illness is a stochastic event, the extent of discrepancies in actual payments born by individuals belonging to a similar income group, as well as the extent of changes in income status of individuals due to “catastrophic” health care payments, is likely to be exacerbated in these countries (Wagstaff and van Doorslaer, 2003, Xu et al., 2003, Ichoku, 2005).

A simultaneous assessment of the above three different measures of inequity may therefore be of particular interest to fully reveal the overall RE of health care financing in the context of developing countries. Such assessment can indeed help inform the controversial policy debates about the extent to which reforms aimed at increasing the efficiency of health care systems do not simultaneously increase inequities in health care financing. This may in turn help reduce the possible adverse effects of health care financing on the prevailing income inequalities in a country (Kidson, 1999, McPake and Mills, 2000, James et al., 2006).

The literature on public finance offers various methods to quantitatively measure VE, HE and RR (e.g., Kakwani, 1984, Aronson et al., 1994, Ven et al., 2001, Duclos et al., 2003). The approach that has been previously applied in the specific domain of health care financing (Wagstaff and Doorslaer, 1997, van Doorslaer et al., 1999) is the one initially proposed by Aronson et al. (1994)—hereafter the AJL approach. Theoretically, the AJL approach allows to decompose the total RE of a financing scheme into VE, HE and RR for a population that is composed of groups of true- or exact-income equals, i.e., a situation where the study sample consists of groups of individuals having exactly the same pre-payment income, and for a distribution where the average post-payment income of each group increases with the respective pre-payment income level, i.e., a payment schedule that is supposed not to produce changes in the groups’ ranking-order. However, due to the absence of exact-income equals in real data surveys, empirical implementations of the AJL approach have relied on the principle of close-income equals, i.e., by dividing the study sample into artificial groups of income based on certain definitions of income bandwidths. It has been shown that such practice can lead to misleading results: biases may arise not only due to the arbitrary specification of close-income equals, but also due, in large part, to the possibility of both intra-groups reranking – i.e. the extent to which the payment schedule induces changes in ranking-order of individuals within the specified groups of close-income equals (R_WG) – and entire-groups reranking, i.e., the extent to which the payment schedule induces changes in ranking-order of the whole groups of close-income equals (R_EG) (Ven et al., 2001).

The need to consider the potential impact of R_WG and R_EG, as well as the sensitivity of the empirical estimations of VE and HE to the choice of income bandwidth for close-equals, has been advocated for the assessment of RE of tax and transfer systems (Ven et al., 2001, Urban and Lambert, 2008). These aspects may also be relevant with regard to the assessment of RE of different health care financing schemes. A methodological extension to the earlier work in tax literature has latterly been provided by Urban and Lambert (2008)—hereafter the UL approach. In contrast to the classical AJL approach and its previous applications, the UL approach reset the measurement system of VE, HE and RR using a conceptual model that is specifically designed to accommodate close-income equals setting. The UL approach presents two complementary advantages: it is able to capture all possible reranking effects, and it provides a more convenient identification of vertical and horizontal inequities by smoothing the actual effect of payments within each close-income equals group. In such approach, the VE is measured by allocating to each individual the average payment paid by the respective group of close-income equals, while HE is estimated directly based on person-by-person comparisons of actual and counterfactual; i.e., smoothed, post-payment incomes within close-equals groups. Lastly, although there is no consensus in the empirical literature on an optimal procedure to identify the income bandwidth of close-equals (Ven et al., 2001, Duclos et al., 2003), the UL approach, while computationally involves direct estimates of VE, HE, and RR as sample statistics, advocates an assessment of the relative importance of these effects given different choices of income bandwidth. This may, indeed, facilitate an appropriate specification of close-income groups for policy purpose. The UL approach has been recently applied to investigate the RE of taxation in the USA (Kim and Lambert, 2009). However, to our knowledge, there has been no previous attempt to apply such methodological improvement in the specific area of health care financing.

Another limitation of previous work on inequality measurement in health care finance, which may also fuel unnecessary misinterpretations, is related to the fact that most of the previous studies have rarely assessed the statistical significance of inequality measures. The very few studies (Klavus, 2001, Cissé et al., 2007) that have attempted to do so have used the classical asymptotic method, which has its own limitations (Mills and Zandvakili, 1997). However, statistical inference based on bootstrap methods were shown to lead to more subtle treatment for statistical problems associated with the measures of inequality (Andres and Calonge, 2005, Moran, 2006). The bootstrap method takes into account the specific bounds of inequality measures while no underlying function of distribution is being imposed. Besides comparing standard errors and probability intervals, an obvious advantage of the bootstrap method is that it allows for testing the relationship between two interdependent curves according to the dominance criterion, and therefore, reduces the risk of biased interpretations due to sample structure (Davidson and Flachaire, 2007, Abu-Zaineh et al., 2008).

The purpose of this paper is to apply the above methodological advances initially developed for inequality measurement of taxation to the specific domain of health care financing, and to illustrate how these developments can significantly help clarifying debates about health care policies in the context of developing countries, using the particular case of the Occupied Palestinian Territory (OPT). The financing structure of health care in the OPT is expected to be associated with a major risk of exacerbation of inequalities: the country lacks a universal system of health care financing and a substantial share of health care expenditures is funded through households’ direct out-of-pocket payments PCBS, 2004. The main public financing scheme is the one known as the Governmental Health Insurance (GHI). Initially, this was on a compulsory basis for public sector employees. However, the scheme has been opened up to others – in the private and informal sectors – on a voluntary basis. By 2002, over 60% of the Palestinian households were covered by public scheme, a little less than half of these being covered on a voluntary basis (MoH-PHIC, 2006). On the other hand, the ongoing political crisis in the region, which has brutally increased the proportion of the population living under the poverty line (PCBS, 2006b), tends to aggravate existing inequities in access to health care (Mataria et al., 2006). Consequently, the Ministry of Health has started to provide an almost free of charge coverage to the mostly affected classes of the population. This insurance scheme was lately known as “Al-Aqsa Intifada Insurance”. However, the indigent performance of the local economy, the freeze of financial support and tax transfers to the Palestinian Authority (PA) threaten to negatively affect the current health care delivery system. Furthermore, due to the lack of effective and sufficient redistributive polices (e.g., proper system of tax transfer), which may help reduce inequality and ensure more equitable distribution of income, the redistributive effect of the current health care financing structure are expected to aggravate the global existing inequality in the prevalent income distribution in the country.

The empirical analysis of this paper is based on data from a recent household health expenditure survey, originally designed for initiating a system of National Health Accounts for the OPT. The survey, which was carried out in 2004 by the Palestinian Central Bureau of Statistics PCBS, 2004, covered a national representative sample of Palestinian households residing in the West Bank (WB) and Gaza Strip (GS). Collected data provide detailed information on households’ incomes and expenditures, individuals’ health care seeking behaviours, health care expenditures, and government and private insurance premiums. Consequently, the survey offers a unique opportunity to assess some equity features of health care financing, and to discuss the extent to which recent developments in inequality measurement, and its statistical inference, may add to our understanding of equity issues in the Palestinian context. Some of the methodological developments that we try to transfer to the field of equity measurement in the case of health care financing in the OPT may also be worthwhile for other contexts in developing countries.

The remainder of the paper is organized as follows. Section 2 discuses some methodological issues in measuring and decomposing the total RE into VE, HE, and RR, using the measurement approach proposed by Urban and Lambert (2008); this is followed by describing the estimation procedures and statistical inferences for inequality measures. Section 3 describes our data and variables definitions used for the empirical analysis. Results are reported in Section 4. The last two sections contain our discussion and conclusions.