fedea

Fundación de Estudios de Economía Aplicada

Exploring the Pathways of Inequality in Health, Access and Financing in Decentralised Spain by Joan Costa-Font Joan GIL^** DOCUMENTO DE TRABAJO 2008-13

Serie Economía de la Salud y Hábitos de Vida CÁTEDRA Fedea – la Caixa

Mach 2008

LSE Health and European Institute, London School of Economics (UK) and FEDEA. Departament de Teoria Econòmica & CAEPS, Universitat de Barcelona (Spain) and FEDEA.

Exploring the Pathways of Inequality in Health, Access and Financing in Decentralised Spain

Joan COSTA-FONT a,c and Joan GIL b,c LSE Health and European Institute, London School of Economics (UK) ^b Departament de Teoria Econòmica & CAEPS, Universitat de Barcelona (Spain) ^c FEDEA, Madrid (Spain)

Acknowledgements: We would like to thank Cristina Hernandez-Quevedo, Rosa Urbanos, Ángel López, Guillem López-Casasnovas and Marisol Rodríguez for their comments on and suggestions for a previous, longer version of this document circulated in Spanish. We also wish to express our gratitude for the 2003 Bayer Award and the Ministry of Health and Consumption grant in 2004.

Contact Address: Joan Costa-Font, European Institute & LSE Health, London School of Economics, Houghton Street, WC2A 2AE London. E-mail: j.costa-font@lse.ac.uk

Abstract

The extent to which equality in the access to and the financing of health care reduces inequalities in health is a key question for health-care reform policy-making. Cross-country studies, when they exist suffer from marked comparability limitations due to data heterogeneity and differences between organisational and financing systems. The Spanish devolved national health system offers a “unique field” for exploring these issues, and also for testing the effects of institutional reform, in the form of political decentralisation. The data used is from 2001, the last year before decentralisation was extended to all region states or Autonomous Communities (ACs). This paper contributes to the literature by examining two questions. First, we evaluate the heterogeneity in within regional inequalities in health, health-care access and health financing and we examine whether these are associated with the political decentralisation of health care responsibilities. Second, we explore whether inequalities in health care between regional health services can be explained by inequalities in health-care use and health-care financing, using cross-correlation analysis along with other relevant variables. The results of the study suggest that inequalities in health are not associated with the regional uptake of health-care responsibilities. Instead they appear to be driven by income inequalities and regional health care capacity whilst the influence of inequalities in health-care use depends on quality of life adjustments.

Keywords: health inequality, inequalities in access to health care, inequalities in health care financing, decentralisation, Spain.

1. Introduction

A widely accepted governmental goal in western countries that organise their health system along the lines of a publicly financed health care - and especially national health service’s (NHS), is to improve “equality of opportunities”. This takes place by lowering and ideally removing barriers to health care access. Health equity is at the core of the health-policy agenda and progressively it is possible to evaluate health-policy achievements by the extent to which they attain this goal. Hence, improvements in the degree of equity in the production and maintenance of good health, in the use of different health services and in their financing are normally taken as main outcomes in evaluating the performance of a health system. Moreover, the World Health Organisation (WHO) performance index draws upon a measure of social inequality in health along with a measure of fairness in health-care financing (WHO, 2000). Other things being equal, the lesser the “avoidable” inequalities (the higher equity in health), the better a health system is to perform. Hence, it is possible to circumscribe that ceteris paribus, a health system is argued to perform better, the lesser the “avoidable” inequalities in health it gives rise

To accomplish health equity goals, health systems typically design a set or programs that are intended to curtail existing barriers to health care, most primarily those affecting its financing and generally access -and less so by preventive programmes. Fairness in health financing is addressed by providing comprehensive coverage and limiting the use of direct payments. Other barriers to health care access are normally tackled through programs that improve the delivery of health care and prevention, though public programs not always are capable to curtail pre-existing unequal conditions. Still significant inequalities in health prevail, so that still we find “better health amongst the better off” in spite of public coverage. Among the explanations for the emergence of inequalities in health are the “absolute income” hypotheses which take a materialistic approach, suggesting that the origin of health inequalities lies in the position of individuals in the hierarchy of distribution of goods and services (Marmot, 2000, Wagstaff and van Doorslaer, 2000)¹. Therefore, policies that improve the distribution of material conditions would translate into fairer distribution of health. Another, somewhat competing explanation known as the “relative income” hypotheses, suggest instead a psycho-social explanation whereby social inequalities are ultimately responsible for stress (Cohen et al., 1997) and anxiety (Wilkinson, 1996) which in turn causes poor health in individuals at the bottom of the income distribution (Wilkinson, 1997, 1998). Even though longitudinal studies seem to point towards evidence for the “absolute-income hypothesis” (Gerdtham and Johannesson, 2004), both explanations are not mutually exclusive and suggest that to reduce inequalities in health, it may be important to design interventions that address both psychosocial along with purely materia health production determinants.

¹ Marmot (2002) also puts forward the social participation argument (e.g., enjoying leisure time), according to which people who are poor may enjoy good health if their social participation level is high, and the other way around.

To ascertain whether these theories are empirically sound, an important yet still unresolved question in the literature is whether health inequalities are affected by changes in the access to health care, in income inequalities and the progressivity of health care financing, two of the main institutional responses to curtail health inequalities. This stems from the acceptance that not all inequalities in heath are determined by socio-economic position (LeGrand, 1987). Even when they are, not all of the causes of social inequalities in health can be “avoided” by (usually short-term) public-policy interventions in individual health-production processes. Some inequalities are not under individual or public-authority control, for instance, inequalities resulting from the depreciation of health capital over time; the same would apply to biologically driven gender differences in health² (Wagstaff et al, 1991), or environmental or generic features. Accordingly, research has been increasingly tailored towards examining the methodological processes underpinning the measurement of “avoidable” inequalities (and inequities) in health (Wagstaff et al, 1989, 1999)³. Decomposition approaches disentangle the contribution made by different health-production determinants to the health-inequality indicator from regression techniques. However, little is known about the underlying reasons behind the emergence of such inequalities and this, to some extent, limits policy responses to curbing existing inequalities.

This calls for a better understanding on the underlying causes of health inequalities but to design policy as well as to better evaluate health systems, and its institutional structure. Among this institutional structure, some governments are beginning processes of devolution. Devolution or governmental political decentralisation gives rise to the inclusion of local knowledge as to tailor health polices to local needs. Some argue that devolution might affect equity however very limited evidence is found of such an effect. Accordingly, an empirical question is to examine whether inequalities in the access to health care, it’s financing or its outcomes are affected by government decentralisation.

² This does not include environmentally determinant inequalities that might be gender dependent, such as gender differences in the access to certain health inputs, which could be context dependent.

Spain stands as one of the most suitable institutional settings where to examine regional inequalities. Globally, Spain ranks 11th out of 191 countries in attainment of equity in health, 26th out of 191 in fairness in financing, and overall 19th in goal attainment (WHO, 2000). This study attempts to go one step further than previous studies, by taking advantage of the decentralised institutional structure in Spain and examining the evidence from the 17 regional health services in order to explore potential links between structural determinants (income, income inequality etc.) and procedural ones (access to health inputs, etc.) as influencing inequalities in health. The progressively decentralised governmental structure and availability of data – from the Spanish National Health Survey (SNHS) and the Continuous Household Budget Survey (CHBS) –avoids problems of health-system specificity. Moreover, it makes use of the decentralised structure of the Spanish national health system to examine data on existing inequalities inside each regional health service. This is an especially interesting feature given that most studies deal with inter-regional inequalities (López-Casasnovas et al., 2005) and only one study has addressed intraregional inequalities in health (Costa-Font,2005) but not health care and health financing.

The objectives of this paper are twofold. First, we explore whether decentralization has had an effect on inequalities in health, access to health care and health care financing within Spanish regions states (RQ1). A recent study found that inequalities in health were mainly explained within region states, and that the degree of decentralisation had no effect on the generation of health inequalities (Jiménez-Rubio et al, 2007). Second, in the light of these results, we test what stands behind as an explanation for differences in health inequalities. Particularly, we focus on the effects of theoretically relevant variables such as income inequality and health care resources available in each region state (RQ2).

³ Which are clearly distinguished from preference-based measures of altruism (Wagstaff and Van Doorslaer, 2000)

Previous studies that have addressed this issue, primarily Doorslaer et al. (1997), focus exclusively on country-based data drawn from different surveys that have substantially different wording. They deal mainly with inequalities in health and leave open the question of whether inequalities in health result from other inequalities in financing or health care-delivery is left open. Indeed, some studies take this association for granted although there is no reason for such an assumption. Spain is an interesting case because it makes it possible to examine whether changes in the way countries finance or organise the health system affect in any way the development of inequalities in health. The Spanish General Health Bill of 1986 already defined the “equal access to equal need” principle behind the organisation of the Spanish health service, and the 2003 Cohesion and Quality Act reinforced such equity principles, (López-Casasnovas et al 2004).

The structure of the study is the following. Section 2 contains a discussion of the underlying determinants of health and health inequalities in the light of existing literature, and describes the data limitations and the institutional setting in Spain. Section 3 briefly presents the data and the empirical methodology. Section 4 reports the results and Section 5 contains the conclusions and discussion.

2. Background

2.1 Pathways to Health Inequalities

In the light of these underlying differences, this study conceptualises the existence of inequalities in health as the results of inequalities in the structure of, and in the access to the health-production process. Indeed, inequalities in health result from:

\[D (H _ {t}) = f (D (P _ {t}), D (U _ {t}) D (F _ {t}), A _ {t}, Y _ {t}, D (Y _ {t}), G _ {t})\tag{2}\]

where [D(P)] are inequalities in lifestyles and prevention, [D(U)] inequalities in the access and use of health-care services, [D(F)] inequalities in financing, [(A)] represents differences in demographic composition, [(Y)] the distribution of goods and services, [D(Y)] differences in the disposal of goods and services , and [G] is gender.

The importance of fair distribution of health financing is that it may contribute to better health, by reducing the risk that people who need care do not get it because it would cost too much, or pay for health care but become impoverished and exposed to more health problems (WHO, 2000).

2.2 The Institutional Setting

The Spanish institutional setting before 2002 is especially interesting because since 2002 the organisation of health care has been totally devolved to the 17 different ACs. In fact the ACs have taken on the responsibility for health-care delivery, but financing is still mainly in the hands of the central government. However, ACs differ in several features affecting the delivery of care: the role of the private sector (e.g., in Catalonia 70% of hospitals are privately owned); culture and political preferences, such as the priority given to equity; supplementary health insurance (e.g., more than 20% purchases private health insurance (PHI) in Catalonia, the Balearic Islands and Madrid)⁴. ACs also differ in the organisation of health care. One would expect these differences to have some effect on equity-relevant features of their financing and delivery systems. Indeed, territorial health-care financing takes place through capitation formulas that do not take into account risk-sharing amongst different ACs (López et al., 2004). Therefore, some ACs might be better prepared to undertake pro-poor health policies than others.

Although the possibility of introducing mild co-payments has been discussed in some ACs, visits to GPs and specialists are still free at the point of delivery. Recent studies have examined horizontal inequalities in the use of health care in Spain (Urbanos, 2001, Abásolo et al., 2000). The first study focuses on the whole of Spain and the second on a specific AC. Due to the decentralisation process that has taken place in Spain, the question of whether different ACs are equally successful in eradicating inequalities has become a key policy issue. As GPs act as gatekeepers to health-care access, it is likely that some inequalities are the result of accessing a health-care layer. Furthermore inequalities are probably connected by some specific features.

3. Data and Methods

3.1 Data

The calculation of income-related inequalities in health status and health-care access is based on the Spanish National Health Survey 2001 (hereafter SNHS) drawn up by the Ministry of Health and Consumption. The survey, which consists of 21,120 interviews made during May and June 2001, has been widely used and is fully representative of each AC. The SNHS follows a multi-stage, stratified sampling system, with the basic sampling units being urban districts. It contains the information needed to elicit inequality indexes for the relevant variables, namely access and the final outcome, health status. Given the absence of a visual analogue scale (VAS) in the survey, the cardinal values of a VAS from the Catalan Health Survey 2002 were used. (This was the only one that existed at the time the study was made.) This involved some adjustments such as simplifying the self-reported-health scales; the study had to group “excellent” and “very good” health responses together and do the same with “bad” and “very bad” responses (see Table A1), in order to estimate an interval regression model that followed previous specifications (Fonseca and Jones, 2003). The question used to measure self-reported health status was the following:

⁴ People with PHI might not support policies expanding access to primary and specialist care that is already covered by PHI.

“Let us talk about your health status, in the last 12 months, would you say your health status could be defined as very good, good, fair, bad or very bad?”.

Besides this, the health-production determinants used by the study in the regression model included income, professional status, educational attainment (as conveying positive effects on health production), civil status (to cover the personal-interaction effects arising from marriage), and AC. Age and gender were included to account for specific effects that cannot be modified, (known as unavoidable inequalities). Income was measured as a cardinal variable from interval regression specifications and was compared to the average income obtained from the Spanish Household Budget Continuous Survey.

To examine inequalities in health-service use the study used the same survey (2001 Spanish National Health Survey) and the question used was:

“Have you visited a physician for any health problem or illness in the last couple of weeks?”.

Table 1 suggests the existence of significant differences in self-reported health status (SRHS) between ACs. The study found that, after transforming SRHS using the VAS, very small differences were observed in the order: the Spearman rank correlation was above 0.9, and it did not modify the position of the ACs at the ends of the distribution. The correlation with income improved after this transformation, which possibly indicates that income is correlated with healthrelated quality of life. Finally, the correlation between the predictions of an ordered probit model and an interval-regression model was approximately 0.7.

[Insert Table 1 about here]

Finally, to examine inequalities in financing two types of data are required in these calculations. On the one hand, individual-level data are needed to calculate individual health-care payments and, on the other hand, macroeconomic data to find out the weights to be assigned to each kind of payment. Although other micro-data sets (e.g., the European Union Household Panel) were available, and contained more complete income information, the study finally opted for the Spanish Household Budget Continuous Survey 2000 (HBCS) carried out by the Spanish National Statistics Institute (INE)⁵ since it offered a great deal of information on householdconsumption expenditures including indirect-tax payments.

3.2 Inequalities in Health Methods

Researching horizontal inequalities in health involves examining whether the income distribution of health care compares to that of need. As in previous studies, need was examined using variables from the SRHS as well as other variables such as disability and morbidity which are commonly available in health surveys such as the one employed in this study.⁶ Previous studies exploring inequalities in health suggest the existence of clear-cut inequalities (Urbanos, 2000; Abasolo et al., 2001). However, more recent studies employ different methodologies and raise some methodological questions (García and López, 2004a, 2004b and Costa-Font, 2005) such as the need to obtain a cardinal measure of health to examine inequalities further. Indeed, recent research proposes cardinalising the SRHS by applying the Health-Related Quality of Life (HRQoL) values used by van Doorslaer and Jones (2003), Jones and Fonseca (2004). Following van Doorslaer and Jones (2003), the equivalent cardinal value of the cut-off point of each response to the ordinal question was obtained so as to estimate the cardinal value of self-reported health using an interval-regression approach.

The survey methodology is available at the site: http://www.ine.es/en/daco/daco43/notecpf8597_en.htm

⁶ Self-reported variables considered suitable because they correlate with measures of morbidity and health-care use (Idler y Benyamini, 1997). Although the existence of potential biases in individual perceptions of own health status is well known, they reflect aspects of health that might not be immediately observable using measures of physical health.

This study used VAS values taken from the Catalan Health Survey which were attributed to the other ACs. Ideally a cardinal measure of health should be obtained for each region, but this information was not available. On the other hand, one should not expect significant intraregional variability in the way health is valued. Other studies simply attribute the values found in a survey from British Columbia in Canada or other Health-Related Quality of Life (HRQoL) scales. We estimate health status using a linear index based on rescaling the ordered variable to obtain a normalised health index, as in Cutler and Richardson (1997). However, this still implies accepting some arbitrary assumptions on the value and distribution of individual health status. The underlying reasons for an individual categorisation into a specific health scale are still not accounted for. Therefore, some research claims that SRHS can be interpreted instead as individual categorisation into an interval, which can be ascertained by finding a link between selfreported measures of health and some health-utility indexes (van Doorslaer and Jones, 2003). This allows the use of interval regression so as to generate a continuous measure of self-reported health. In the Spanish case, the only possible measure of such an index was the VAS, used in the Catalan Health Survey, in order to obtain the inferior and superior intervals for each self-reported health response (see Appendix). On the other hand, social position can be measured in a rank drawn from a socio-economic reference variable, namely individual income.⁷ The method to estimate the inequalities in health follows the standard decomposition methodologies described in the next section.

3.3 Inequalities in Access Methods

Health-care-utilisation data such as visits to the doctor are known to have a highly-skewed distribution; the majority of survey respondents report no visits or very few visits, and only a very small proportion of individuals report frequent use. The negative binomial model, which allows for over-dispersion, has often been shown to be an adequate choice in studies of health-care utilisation (see Urbanos, 2001, for an examination of health inequalities in the Spanish context.) The present study used a conservative estimate of health care access by defining it as access to any physician regardless of specialisation (as in Atella et al., 2004) and used the same methodology as van Doorslaer et al. (2004). The underlying hypothesis was that use or access (U) depends on need (H), and can also depend on income (Y) and other variables (X) as follows:

⁷ It is assumed that income is well measured and adequately proxies permanent and absolute income, and is associated with other proxies of socio-economic status including individual and family wealth. The extent to which

\[U _ {i} = f (Y _ {i}, H _ {i}, X _ {i})\tag{4}\]

so that a health system can be evaluated by the extent to which patients have equal access for equal need. This implies observing a measure of access, through a probit or a linear probability model, and decomposing inequalities as required. Indeed, after estimating the utilisation specification, following a decomposition method using a linear-regression model, linking the variable of interest to a set of k exogenous determinants is used:

\[y _ {i} = \alpha + \sum_ {k} \beta_ {k} x _ {k i} + \varepsilon_ {i}\tag{5}\]

it is possible to apply Rao’s theorem for income inequality, so that the concentration index (CI) of the probability of visiting a doctor can be decomposed by factors:

\[C I = \sum_ {k} \left(\frac {\beta_ {k} \mu_ {k}}{\mu}\right) C I _ {k} + \frac {G C _ {\varepsilon}}{\mu}\tag{6}\]

where is the mean of the k variable, is the concentration index of the k variable and the last term is a generalised concentration index for the residuals. Equation (6) shows that the CI of the probability of contacting a physician can be thought of as the sum of two components. The first term is the deterministic component, equal to a weighted sum of the concentration indexes of the k repressors, where the weights are the elasticities of y with respect to each variable evaluated at the sample mean. The second term is a residual component that reflects the inequality in utilisation of health care that cannot be explained by systematic variation across income groups in the The main drawback of the decomposition method is the requirement of a linear-regression model and the need of relying on the fact that y is additive in its components. Van Doorslaer, Koolman and Jones (2004) propose an approximation based on the partial effects representation for the decomposition analysis, which has the advantage of being a linear additive model of utilisation. Once total inequality has been broken down into components, the inequity index can be calculated by the difference between the actual utilisation inequality and the estimated incomeinequalities of certain variables that are considered unavoidable, such as need, age and gender:

this holds or whether the introduction of additional controls available in the databases makes any difference to the estimates is a matter of future research.

\[H I = \hat {C} I \text { actual } - \sum_ {n} \hat {C} I \text { need }\tag{7}\]

Hence, to obtain a measure of inequity rather than inequality, it is common practice to subtract unavoidable components, such as gender and age (CI*) from the inequality measure.

3.4 Inequalities in Financing Methods

The purpose of this section is to analyse the extent to which individual public and/or private payments to finance health-care services in each Spanish AC are related to ability to pay. In other words, to quantify whether this relationship is proportional, progressive or, alternatively, regressive. A financing scheme is said to be “progressive” when the ratio of health payments to income increases as income grows. It is considered “regressive” when the opposite is true and “proportional” when the ratio is constant through all income levels. In a markedly progressive financing scheme the proportion of the financial burden of health-care payments borne by the lowest income group is lower than its total income share, while the reverse is true for the richest part of the society. This study made use of progressivity indices, particularly the Kakwani index (Kakwani, 1977), to measure the degree to which the different health-care financing payments in each AC are progressive.

This study followed a two-stage procedure. Firstly, the degree of progressivity with respect to each particular kind of payment is scrutinised and, secondly, overall progressivity is assessed by weighting the Kakwani indices calculated for each type of health-care payment. To that end it is necessary to examine all the different kinds of payments (public and private) used by individuals to buy health services. This means considering not just direct or “out-of-pocket” payments but also private medical-insurance premiums and direct and indirect taxes, given that the Spanish national health system is financed through general taxation. In practice, in order to derive Kakwani indices for each payment source and AC, before-tax Gini coefficients and Concentration indices were computed for all payments by using micro-data and what is known as the “convenient covariance method” (Jenkins, 1988).

3.5 Health Care Payments and Weights

The study used the HBCS 2000 to derive public and private payments such as income taxes, VAT taxes, excise taxes, property taxes, out-of-pocket payments and private medicalinsurance premiums. However, before calculating income taxes it was necessary to transform the income variable. As long as total income was measured in the Survey as household earnings, net of taxes and in interval terms, the study first derived a continuous measure by performing an interval regression model and using the characteristics of the head of household. Secondly, equalised net income was found by applying the modified OECD equivalence scale. Finally, gross annual equalised income was deduced after adjusting the effective income-tax rates by income brackets which were obtained from the Spanish Tax Administration (AEAT, 1999). From this earnings measure, individual income taxes were easily computed by applying the effective tax rates.

As for VAT payments we grouped goods and services consumed by households and subjected to the indirect tax into three categories, given the three legally VAT tax rates in Spain (4, 7 and 16%). Individual consumption payments were then derived by using these legal tax rates and the above mentioned equivalence scale.⁸ Hence, we implicitly assume that the tax is totally translated into higher market prices. To calculate excise taxes or duties, we followed admittedly simplistic imputation methods given the lack of sufficient information. On the one hand, for both the beer and the alcohol and alcoholic beverages (wine, spirits, liquors, cava, “sherry”…) tax payments were deduced applying to equalized consumption the share of total revenues in 2000 (taken from official statistics as MAT/AEAT, 2000) to total aggregate consumption from the Survey. On the other hand, tobacco taxation was imputed in a different manner: cigarettes and cigar spending (at market prices) were assimilated to the tax base and after deflating these amounts by their duty rates (54% for cigarettes and 12.5% for cigars) we were able to calculate individual tobacco payments.⁹ Finally, we also assigned individual payments for the energy excise tax which represents the most important duty tax in Spain. Although the fiscal tax base is determined in physical units, from declared spending on petrol (leaded and unleaded gasoline, gas-oil….) and liquid combustibles for housing (gas-oil, fuel-oil…) we applied a weighted tax duty rates for petrol (41%) and combustible liquid (9.84%) to derive individual payments.¹⁰ As for the local property tax, the imputation was easier as long as the CHBS includes one question regarding this type of payments. Hence, we simply had to transform household payments for both the principal and secondary house (if any) into individual tax payments.

Hence, we could impute up to four classes of public payment (income, VAT, excise and local-property taxes) at the individual level. As Table 2 reports, these taxes represented 71.4% of the total tax revenues collected by government bodies in Spain in the year 2000. Table 2 also compares allocated taxes vs. revenues collected; making it possible to know to what extent the public payments assigned to individuals in the sample are representative. Although VAT payments were quite satisfactorily assigned, since almost 70% of total VAT revenues were allocated to individuals, unfortunately duties and local property taxes were poorly imputed (20% and 11%, respectively). The imputation of income tax is, certainly, poor. However, since almost 80% of total income-tax revenues in Spain are attributed to labour income and almost all incomes in the HBCS-2000 are of the same nature (capital incomes are very poorly measured), income payments allocated in our sample represented 40% of total labour income-tax revenues. This indirectly demonstrates the existence of large-scale income sub-reporting in the HBCS, even when information is only declared through income intervals.

⁸ We took into account that consumption spending is measured at market prices, that is, VAT and, in some cases, special taxes are included.

⁹ Other smoking tobaccos were excluded from the estimations.

10 From information on i) average monthly sell retailing prices of different energy products in the year 2000 −disentangling between price before taxes, VAT and excise taxes− offered by the “Oil Bulletin. Year 2000-2001”, European Commission (http://europa.eu.int/comm/energy/en/oil/bulletin_en.html#Monthly%20Prices%202000) and ii) aggregate consumption of petrol products in Spain (Energy national Committee, http://www.cne.es/mercados.html) we constructed the above mentioned weighted average energy duty rates as the share of excise duties to selling prices.

[Insert Tables 2 and 3 about here]

In addition, two types of private health-care service acquisition payments were considered.¹¹ On the one hand, direct or out-of-pocket payments for medical services, paramedical services, hospitalisation care, drugs, therapeutic material and devices, and, on the other hand, indirect payments or private medical-insurance premiums.

According to the Spanish Ministry of Health and Consumption¹² total public health-care expenditure in the year 2000 amounted to €32,671m, or 5.7% of the GDP. It is estimated that 76.1% of total health-care expenditure corresponds to the national health system, while the remaining 23.9% (1.7% of GDP) corresponds to the private health-care sector.¹³ This macroeconomic data togwether with the information in Table 2 made it possible to calculate the “macroeconomic weights” (Table 3), which are used to aggregate health-care payments. Under the rubric of “Public or Tax Payments” the study aggregated income, VAT, excise and local property taxes. Besides, under the heading of “Private Payments” it grouped all direct and indirect payments devoted to financing private health care services. Hence, “Total Payments” was calculated by summing up public and private payments.

4. Empirical Results

4.1 Income-related Inequities in Health

From the different specifications for health production in Spain the study found that income-related inequalities in health were moderate, CI was 0.017 and the inequity index was

¹¹ The survey collects information on effective health spending incurred by households and excludes any imputed spending by use of public health care services.

¹² “Estadística del Gasto Sanitario Público” del Ministerio de Sanidad y Consumo

Out-of-pocket payments represent approximately 82.5% of total private-health spending, while insurance payments amount to 17.5% (Gil, 2004).

0.016, all leading significant inequity coefficients by ACs after bootstrapping standard errors (Figure 1). These results are consistent with the view that the Spanish health care system exhibits non pervasive health inequalities, most likely due to its equal access for equal need as some studies predict (Lahelma et al., 2002). The Canary Islands, Murcia, Galicia and Extremadura displayed the most inequality and inequity, and Asturias, Navarre, the Basque Country and Castile-Leon the least. It was therefore not possible to conclude that the ACs with health-care responsibilities exhibited higher inequalities in health. Interestingly, by grouping inequalities by regions with decentralised responsibilities in 2001 the study found an inequity coefficient of 0.015 even though a larger inequality coefficient (0.018) was found in regions with centralised health care responsibilities than in those with transferred health-care responsibilities (0.016).

[Insert Figure 1 about here]

4.2 Income-related Inequities in Access to Health Care

As Table 4 reveals, the probability of a medical visit varied from a high 30% in Madrid to a moderate 13% in Navarre. There was also significant variability between regions that were subject to a common healthcare-management system. The estimation of inequity indices in Figure 2 mostly exhibited a negative coefficient although there were marked differences between them. Hence, inequities in the probability of access to health care were actually pro-poor, showing that individuals with lower incomes use the health system more. Interestingly, some figures were very close to zero and, except in the case of Navarre, no ACs exhibited inequalities. These results are consistent with previous work (García and López, 2004), especially when a decision variable such as the purchase of PHI is not included.

[Insert Table 4 and Figure 2 about here]

4.3 Progressivity in Health-Care Financing

According to Rodríguez et al. (1993) in the 80s the Spanish health-care financing system was regressive, with a negative Kakwani index for total payments of −0.023 (and a Suits index of −0.036). Similar results were found by van Doorslaer et al. (1993). This finding is not strange given that most health-financing resources came from social security contributions (61.7% of total payments), while general taxation barely covered 14% (direct taxes: 7.6% and indirect taxes: 6.4%). However, at the beginning of the 90s there were some advances towards a more progressive heath-financing system. For instance, Wagstaff et al. (1999) found a proportional financing scheme with a Kakwani index for total payments of 0.0004, with public payments (78.3% of total payments) slightly progressive (0.0509) but private payments clearly regressive (−0.1627). The key element behind this change in pattern was the creation of a National Health Service in 1986, which ultimately meant a change in the composition of the financing sources; while social security contributions lost their preponderant role (from 62% in 1980 to 22% in 1990), general taxation increased its contribution (from 14% in 1980 to 56.3% in 1990). Income tax doubled its contribution to total-system progressivity and increased its share in total financing (31% in 1990) but a similar trend was observed in the case of indirect taxation (from 6.4% in 1980 to 25.5% in 1990) especially when Spain introduced the VAT after joining the European Community in 1986.

[Insert Table 5 about here]

Table 5 reports the Gini and Kakwani indices computed. The first row presents the results for the whole country. The Gini index for Spain in year 2000 was 0.3089, indicating significant inequality in the distribution of equivalent gross income, which arguably justified public intervention. For instance, the top decile received almost eight times more income than the bottom one. By contrast, income (labour) taxes were highly progressive (Kakwani index: 0.3811) with the eighth, ninth and, especially, the tenth deciles being the only ones that contributed more than their income share. In fact, several studies indicated that the interplay of the personal and family tax-relief thresholds and labour-income deductions were key elements behind the highly progressive and re-distributive structure of Spanish income tax (Onrubia and Rodado, 2003). As expected, the structure of indirect taxation (VAT and excise taxes) was regressive with a Kakwani figure of −0.1024. Consequently, the Kakwani index for public payments was positive and statistically significant (0.0429). This finding was crucial given that public payments represented 76.1% of total payments. Private health expenditure appeared to be regressive, although this result was the combination of regressive out-of-pocket payments and progressive private-insurance premiums. When the public and private financing sources are added together, the system was slightly progressive with a Kakwani index of 0.0337. Again, the top three deciles were the only ones that paid more than their income-share. Thus, over the last two decades, the Spanish health-care system has gained in terms of vertical equity and the reduction of inequalities in the allocation of the financial burden, moving from a regressive framework in the 1980s to a more progressive system at the beginning of the 21st century.

Table 5 contains another important piece of information. It shows the degree to which each financing source in each AC was progressive or regressive. Great variability in the degree of gross-equivalent-income inequality was observed at the regional level. According to the study’s micro-data, the Gini index ranged from relatively high values, such as 0.3463 (Extremadura), 0.3407 (Canary Islands.) or 0.3338 (Aragon), to low values, such as 0.2656 (Castile-La Mancha), 0.2681 (Cantabria) or 0.2763 (Navarre). Interestingly, the data revealed a non-statistically significant correlation between income inequality and average annual equivalent regional income. Not surprisingly Kakwani indices for income (labour) taxes were positive and significant, indicating a high degree of progressivity in all ACs, although there was noticeable regional disparity. Hence, income taxes contributed to achieving a more even income distribution at regional level, since higher income deciles paid a relatively larger proportion of income tax. As expected, the distribution of indirect tax revenues in relation to income resulted in a negative and significant value for the Kakwani index in all ACs with the exception of Cantabria, conveying a high degree of regressivity for these payments at regional level. Consequently, given the relative share of each financing source the present study found that the public financing of health-care services was progressive in most of the Spanish ACs and proportional in the rest. Interestingly, regarding private health payments (out-of pocket payments and insurance premiums) the results were very mixed: in some ACs they were highly regressive (-0.2005 in Extremadura; -0.1842 in Aragon; -0.1811 in Cantabria; -0.1708 in Navarre; -0.1543 in the Basque Country) but in others there was evidence of proportionality.

The last column of Table 5 shows the overall progressivity of the financing system in each AC, computed by taking a weighted average of the progressivity indices for each individual financing source. The estimated overall Kakwani indices suggested that in 8 of the 17 ACs health-care finance was only modestly progressive: 0.0465 (Valencia), 0.0436 (Navarre), 0.0412 (Andalusia), 0.0397 (Asturias), 0.0396 (Galicia), 0.0382 (Castile-Leon), 0.0382 (Balearic Islands) and 0.0287 (Catalonia). In other words, all the ACs that had assumed health-care responsibilities over the last ten or twenty years (with the exception of the Basque Country) exhibited moderate progressivity in their health-care financing schemes.

4.4 Determinants of Inequalities: a Cross-Correlation Approach

The main reason for undertaking the present study was to find the correlation between the different variables that explain inequalities (see Wagstaff et al., 1997). To explore this question, we made both cross-correlation and regression analyses to search for associations between measures of inequality in health, in access, and in financing so as to explain their potential determinants.

Simple correlation analysis among all possible variables suggested that inequalities in health were positively and significantly associated with gross-income inequalities at the regional level (0.672, p<0.05). A positive correlation was also found between inequities in health and income inequalities (0.675, p<0.05). Interestingly, an even a stronger association was found for the ACs with health-care responsibilities (0.810, p<0.05 and 0.805, p<0.05, respectively). On the other hand, there was evidence of a negative association between inequalities in the probability of access to health care and gross-equalised-income inequalities (-0.601, p<0.05), although this pattern was only observed in those AC with health-care responsibilities centralised at national level (0.720, p<0.05). A negative association between inequities in the probability of use and income inequality was also found for these ACs (-0.698, p<0.05). Finally, no association was found between total progressivity in health financing and inequalities and inequities in health status and access to health care.

4.5 The Determinants of Health Inequalities Using Regression Methods

In Table 6 we report the results of a set of regressions exploring the determinants of the income-related inequalities and inequities in health status among Spanish ACs. A set of possible explanatory variables taken from theoretical debates were considered, including financing progressivity, inequalities in use, income inequalities, health-care spending, and number of physicians. For obvious colliniarity issues, the entire set of variables could not be included in the same equation, particularly those referring to inequalities/inequities in the probability of use and income inequality. Results suggested that inequalities (inequities) in health status among ACs were explained by income inequalities, consistent with the absolute-income-hypothesis approach. As expected, when income inequality was excluded the impact of inequalities in the probability of health-care use was negative and statistically significant, suggesting that improving the propoor access inequality would result in a fairer distribution of the level of health. Finally, the number of available physicians let to a statistically significant reduction in health inequalities (inequities) which would mean that health inequalities (inequities) are, at least to some extent, a problem of heath-care resources.

[Insert Table 6 about here]

5. Discussion

This paper attempted first to obtain an empirical estimate of the extent to which different regional health systems in Spain exhibit inequalities in health, health-care access and financing and the extent to which region states that enjoyed health care responsibilities exhibit higher inequalities in health. This was done using representative data at AC level for 2001 so that seven region states enjoyed health responsibilities whilst the remaining ten did not. Secondly, we used the estimated coefficients to examine the connection between income-related inequalities in health, health care access and health financing by using cross-correlation analysis from homogeneous units to provide further insight. Previous studies by van Doorslaer et al. (1997) employed country-based data from different surveys, so that suffered from significant institutional and survey specific heterogeneity. This evidence allows us to examine whether inequalities in financing and in the use of health care explain inequalities in health. Unlike previous studies this research covered health-related inequalities in the whole of Spain, using region state data for all the different types of inequalities.

The study found evidence of some intra-territorial income-related inequalities (inequities) in Spain, and consistent with other studies that examine Canadian data (Jimenez-Rubio et al., 2007) we found that decentralisation did not appear to be the variable that accounted for the rise of inequalities in health or in health care access; if anything it seemed to curtail health inequalities (RQ1). Inequities in the probability of access were very slight and the study found overall health financing to be progressive and equitable. With the remarkable exception of the Basque Country and Madrid, most of the remaining ACs showed moderate levels of progressivity in their health financing schemes, suggesting that financing followed equitable patterns. Income inequalities were the main variable explaining both income-related inequalities and inequities in health along with health care capacity. It was also found that (pro-poor) inequalities in access had a statistical negative impact on health inequality, showing that enhancing the access of poorer people to health services would mean a fairer distribution of health. This explains why inequalities in access are connected to inequalities in health (RQ2).

The relevance of this study lies in that unlike other goods, health care cannot be distributed directly (Hausman et al, 2002). Health equity can mainly be indirectly promoted through few health and social policies but primarily with fiscal instruments that transfer income from the relatively affluent to the relatively poor, policies to invest in poor neighbourhoods and that improve environmental determinants behind health production. This is usually combined with long-term policies to cut down inequalities in health such as education programmes to ensure that children receive adequate health information and lead healthy lifestyles regardless of the socio-economic status of their parents. On the other hand, in the case of inequalities in access to health care, policies can be introduced to improve the conditions of care delivery, or to make the financing more pro-poor. However, in undertaking active social policies it is important to remember that a pro-poor distribution of income that does not reduce inequalities in health is not necessarily welfare-improving (Contoyannis and Forster, 1999). This is due to the fact that some allocation alternatives make better use of existing resources and therefore greater improvements in welfare.

Some of the areas not covered in this study are the extent to which inequalities in health are explained by factors such as inequalities in the access to drugs (in Spain drugs are one of the few health-care inputs that are subject to cost sharing). The main caveats of this study are that it only examines a cross-section. Longitudinal data is gradually being made available to examine changes in health inequalities over time, which could enable researchers to obtain further insights into what lies behind them. Deaton (2002) found that societies with lesser income inequalities were also those with lesser health inequalities. However, it is important to focus on specific studies of individual conditions, given that evidence at the individual and at the aggregate level might be significantly different (Evans, 2002). Indeed, it is necessary to include individual heterogeneity when looking for explanations for health inequalities. This can be done by exploring longitudinal databases and including the environmental factors underlying healthproduction determinants.

References

1. Abásolo, I. (1998) Equidad horizontal en la distribución del gasto público en sanidad por grupos socioeconómicos en Canarias. Un estudio comparado con el conjunto español. Hacienda Pública Española, 147: 3-27
2. Abásolo, I. Manning, R. and Jones, A. (2001) Equity in utilization of and access to public-sector GPs in Spain. Applied Economics, 33: 349-364
3. AEAT (1999) Memoria de la Administración Tributaria 1999, Agencia Estatal de la Administración Tributaria (http://www.aeat.es)
4. Atella, V., Brindisi, F., Deb, P. , Rosati, F.C. (2004) Determinants of access to physician services in Italy: a latent class seemingly unrelated probit approach. Health Economics 13, 657- 668.
5. Atkinson, AB. (1970) On the measurement of inequality. Journal of Economic Theory 1970; 2: 244-263.
6. Besley, T. and Coate, S. (2003) On the public choice critique of welfare economics. Public Choice, 114: 253-273.
7. Bommier, A and Strecklov, G. (2002) Defining health inequality: why Rawls succeeds when social welfare theory fails. Journal of Health Economics, 21: 497-513.
8. Brooks, R., Rabin, R. and de Charro, F. (eds) (2003) The measurement and valuation of health status using EQ-5D: a European perspective. Kluwer. Broome, J. (1991) Weighting goods. Basil Blackwell, Oxford.
9. Cohen, S., Line, S., Manuck, S., Rabin, B. Heise, E. and Kaplan, J. (1997) Chronic social stress, social status, and susceptibility to Upper Respiratory Infections in Nonhuman Primates. Psychosomatic Medicine, 59 (3): 213-221.
10. Contoyannis, P. and Forster, M. (1999) The distribution of health and income: a theoretical framework. Journal of Health Economics, 18: 605-622.
11. Costa-Font, J. (2005) Inequalities in self reported health within Spanish Regional Health Services: devolution re-examined? International Journal of Health Planning and Management 20(1): 41-52
12. Culyer, A. (1980) The Political Economy of Social Policy, Oxford, Martin Robertson
13. Cutler, D. and Richardson, E. (1997) Measuring the health of the United States Population. Brookings Papers of Economic Activity, 1997 (2): 217-271.
14. De Vogli1R., Mistry R., Gnesotto R. and Cornia, G.A. (2004) Has the relation between income inequality and life expectancy disappeared? Evidence from Italy and top industrialised countries. Journal of Epidemiology and Community Health 2005;59:158-162.
15. Deaton, M (2002) Policy Implications of the Gradient of Health and Wealth. Health Affairs 21(2).
16. Duclos, J-Y and A. Araar (2004) Poverty and equity: measurement, policy and estimation with DAD, Preliminary version, CIRPÉE, Quebec, Canada.
17. Evans, R (2002) Interpreting and addressing inequalities in health: from Black to Acheson to Blair to,…? Office of Health Economics, London.
18. Fonseca, L. and Jones, A. (2003) Inequalities in self-assessed health in the Health Survey of England. Equity Working Paper nº 11.
19. Fuchs, V.R. (1986) Schooling and Health: the Cigarette Connection. In Fuchs, VR (ed) pp 243- 254. The Health Economy, Harvard University Press, Massachusetts.
20. García-Gómez, P. and López, A. (2004) The evolution of inequity in the access to health care in Spain: 1987-2001. Working Paper, Universitat Pompeu Fabra.
21. Garcia-Gómez, P. and López, A. (2004a) Regional differences in socio-economic health inequalities in Spain. Working paper CRES, Universitat Pompeu Fabra& an updated version in un the BBVA Working Paper Series
22. García-Gómez, P. and López, A. (2004b) The evolution of socio-economic inequalities in Spain: 1987-2001. Working paper CRES, Universitat Pompeu Fabra.
23. Gerdtham, U-G, Johannesson, M. (2004) Absolute income, relative income, income inequality and mortality. Journal of Human Resources, 39 (1): 228-244.
24. Gil, J. (2004) El gasto de las familias en servicios sanitarios privados, in Anexo III.1 of Informe Anual del Sistema Nacional de Salud 2003, Observatorio del Sistema Nacional de Salud, Ministerio de Sanidad y Consumo.
25. Grossman, M. (1972) On the concept of health capital and demand for health. Journal of Political Economy, 80(2): 223-255.
26. Hausman, D.H., Asada, Y. and Hendemann, T. (2002) Health Inequalities and why they matter. Health Care Analysis, 10:1777-1791.
27. Idler, E.L. and Benyamini, Y. (1997) Self-rated health and mortality: a review of 27 community studies. Journal of Health and Social Behaviour, 38(1): 21-37.
28. Jenkins, S. (1988) Calculating income distribution indices from micro-data, National Tax Journal, 41: 139-142.
29. Jimenez D., Smith, P. and van Doorslaer, E. (2007) Equity in health and health care in a decentralized context: evidence from Canada. Health Economics (in press).
30. Kaplan, G., Pamuk, J. and Lynch, R. et al. (1996) Inequality in Income and Mortality in the United States: Analysis of Mortality and Potential Pathways. British Medical Journal, 312: 999-1003.
31. Kakwani, N.C, Wagstaff, A. and van Doorslaer, E. (1994) Socioeconomic inequalities in health: measurement, computation and statistical inference. Journal of Econometrics, 77(1):87- 104.
32. Kakwani, N.C. (1977) Measurement of tax progressivity: an international comparison. Economic Journal 87: 71-80.
33. Kenkel, D. (1991) Health behaviour, health knowledge and schooling. Journal of Political Economy,99(2): 287-208.
34. Kunst, A.E. and Mackenbach, J.P. (1994) International variation in the size of mortality differences associated with occupational status. International Journal of Epidemiology, 23: 1-9.
35. Lahelma, E., Kivelä, K., Roos, E., Tiouminen, T., Dahl, E., Diderchesen, F., Elstad, J.I., Bissau, I., Lundberg, O., Rahkonen, O., Rasmussen, N.K. and Aber, M. (2002) Analysing changes of health inequalities in the Nordic welfare status. Social Science and Medicine, 55: 609- 625.
36. Lambert, P.J. (1996) La distribución y la redistribución de la renta: un análisis matemático, 2ª edición. Colección de Estudios de Hacienda Pública, Instituto de Estudios Fiscales, Ministerio de Economía y Hacienda.
37. LeGrand, J. (1987) Inequalities in health: some international comparisons. European Economic Review, 31: 182-191.
38. LeGrand, J. (1991) "Editorial: The distribution of health care revisited: a commentary on Wagstaff, van Doorslaer & Paci, and O'Donnell & Propper", Journal of Health Economics, 10: 239-245.
39. López-Casasnovas, G., Costa-Font, J. and Planas, I. (2005) Diversity and regional inequalities: assessing the outcomes of the Spanish system of health care services, Health Economics 14(S): S221-S235.
40. Marmot, M. (2002) The influence of income on health: views of an epidemiologist. Health Affairs, 21(2).
41. Moloughney, B. (2004) Housing and Population Health. The State of Current Research Knowledge. Canadian Population Health Initiative, Toronto.
42. Onrubia, J. and Rodado, M.C. (2003) Estructura del IRPF y redistribución de la renta: un análisis comparativo de la reforma de 1999, in: J. Onrubia and J.F. Sanz, eds., Redistribución y bienestar a través de la imposición sobre la renta personal, Estudios de Hacienda Pública, Instituto de Estudios Fiscales (Madrid):131-150.
43. Rawls, J. (1971) A Theory of Justice. Harvard University Press, Cambridge, MA.
44. Rodríguez, M., Calonge, S. and Reñé, J. (1993) Spain, in: Equity in the finance and delivery of health care: an international perspective (eds. Van Doorslaer, Wagstaff and Rutten) Health Services Research Series, Commission of the European Communities, OUP 1993.
45. Sen, A. (2002a) Why health equity? Health Economics 11: 659-666.
46. Sen, A. (2002b) Health: perception vs. observation, British Medical Journal, 324: 860-861.
47. Serrano-Mancilla, A. (2004) Impacto redistributivo desagregado del IVA actual in España: posibles vías de reforma, mimeo Universidad Autónoma de Barcelona.
48. Smith, J. (1999) Healthy Bodies and Thick Wallets: The Dual Relation between Health and Socio-Economic Status. Journal of Economic Perspectives, 13(2): 145-166.
49. Urbanos, R. (2001). Explaining inequality in the use of public health care services: evidence from Spain. Health care Management Science, 4: 143-157.
50. Urbanos, R. (1999) Análisis y evaluación de la equidad horizontal interpersonal en la prestación pública de servicios sanitarios. Un estudio del caso español para el período 1987-1995. Doctoral thesis presented in Department of Applied Economy VI of la Universidad Complutense, Madrid.
51. Urbanos, R. (2000) Desigualdades sociales en salud y efectividad potencial de las políticas públicas: un estudio aplicado con datos españoles. Hacienda Pública Española, 154: 217- 237.
52. Van Doorslaer, E., Koolman, X. and Jones, A.M. (2004) Explaining income related inequalities in doctor utilisation in Europe. Health Economics, 13: 629-647.
53. Van Doorslaer, E. et al. (1999) The redistributive effect of health care finance in twelve OECD countries, Journal of Health Economics 18: 293-315
54. Van Doorslaer, E., Wagstaff, A. and Rutten, F. (1993) Equity in the finance and delivery of health care: an international perspective, Health Services Research Series, Commission of the European Communities, OUP 1993.
55. Van Doorslaer, E., Waggstaff, A., Bleichtdrodt, H. et al. (1997) Income-related inequalities in health: some international comparisons. Journal of Health Economics, 16: 93-112.
56. Van Doorslaer, E., Koolman, X. and Puffer, F. (2001) Equity in the use of physician visits in OECD countries: has equal treatment for equal need been achieved? OECD conference
57. proceedings Measuring Up: Improving Health Systems Performance in OECD Countries. OECD Health Conference on Performance Measurement and Reporting
58. Wagstaff, A. and van Doorslaer, E. (2000) Measuring and testing for inequity in the delivery of health care. Journal of Human Resources, 35(4): 716-733
59. Wagstaff, A., van Doorslaer E. and Paci, P. (1989) Equity in the finance and delivery of health care: some tentative cross-country comparisons. Oxford Review of Economic Policy 5(1), 89-112.
60. Wagstaff, A., van Doorslaer, E. and Paci, P. (1991) On the measurement of horizontal inequity in the delivery of health care. Journal of Health Economics, 10:169-205
61. Wagstaff, A. and van Doorslaer, E. (1994) Measuring inequalities in health in the presence of multiple-category morbidity indicators. Health Economics, 3: 281-291.
62. Wagstaff, A. and van Doorslaer, E. (2000) Equity in health care financing and delivery, In: AJ Culyer and JP Newhouse (Eds.) Handbook of Health Economics, North Holland, 1803-6
63. Wagstaff, A. et al. (1992) Equity in the finance of health care: some international comparisons, Journal of Health Economics 11: 361-387
64. Wagstaff, A. et al. (1999) Equity in the finance of health care: some further international comparisons, Journal of Health Economics 18: 263-290.
65. Wagstaff, A. and van Doorslaer, E. (2001) What makes the personal income tax progressive? A comparative analysis for fifteen OECD countries, International Tax and Public Finance, vol. 8 (3), :299-316.
66. Weinstein, N.D. (1987) Unrealistic optimism about susceptibility to health problems: conclusions from a community-wide sample. Journal of Behavioural Medicine, 10, 481-500.
67. Wilkinson, R (1998) Low relative income affects mortality. British Medical Journal, 316 (7144): 1611.
68. Wilkinson, R. (1997) Health inequalities: relative or absolute material standards. British Medical Journal, 314(7080): 591-95.
69. World Bank. Quantitative Techniques for Health Equity Analysis: Technical Notes #4, 7,16 (http:/www.worldbank.org/poverty/health/wbact/health_eq.htm).
70. WHO (2000). The World Health Report 2000 – Health systems: Improving performance. The World Health Organization, Geneva, Switzerland

Table 1. Self-reported Health Status and VAS level by AC

Note: SRHS=Self-reported health status. The best SRHS (‘very good’) takes value 1 and the worst one (‘very bad’) takes value 5. VAS=Visual analogue scale, which is computed from the predictions of an interval regression against a set of independent variables (age, gender, income, educational level and cohabitation).

Table 2: Macroeconomic Aggregates and Tax Payment Imputations (in millions of pesetas)

Note: 1. The revenue figures were obtained from “Cuentas Consolidadas de las Administraciones Públicas, Año 2000” computed according to the European system of national and regional accounts (ESA-95). 2. Total taxation refers to national accounts of a) import and production taxes and b) wealth and income taxes. (*) Fiscal revenues do not include taxes on some types of transport and minor taxes. Source: own elaboration. (**) HBCS = Household Budget Continuous Survey

Table 3: Structure of Health-Care Payments

Source: Own elaboration.

Table 4: Probability of a visit to a physician across ACs

Figure 2. Income-related Inequities in Access to Health Care

And=Andalusia, Arag=Aragon, Ast=Asturias, Can=Canary Islands, Cant=Cantabria, Cast LM= Castile la Mancha, Cast Le=Castile Leon, Cat=Catalonia, Val.=Valencia, Extrem=Extremadura, Gal=Galicia, Mad=Madrid, Mur=Murcia, Nav=Navarra, PV=Basque Country, Rioj=La Rioja.

Table 5: Gini and Kakwani Indices for Health-Care Payments (2000)

Note: Robust standard errors were computed for the statistical inference analysis. Coefficients statistically significant at 5% (10%) are in bold (italic) typeface.

Table 6: Ordinary Least Square estimation of Health Inequality and Inequity by AC

Note: Robust standard errors were computed for the inference analysis. Coefficients statistical significant at 5% (10%) are in bold (italic) typeface. Overall health-care financing progressivity is assessed through the Kakwani index. Inequality in use is measured as the inequality in the probability of visiting a physician. The Gini coefficient is used to measure equivalent gross-income inequality. Health-care spending is measured as the log of average spending per capita in each AC, and the number of physicians is expressed per 1000 inhabitants.

Figure 3. Equity Performance Index by AC

Andalucia=Andalusia, Baleares=the Balearic Islands, Canarias=the Canary Islands, Catalunya=Catalonia, Castilla LM=Castile la Mancha, Castilla Le=Castile Leon, Navarra= Navarre, País Vasco= the Basque Country.

Table A1. Matching the ESCA(*) and the Spanish National Health Survey 2001

Source: Encuesta de Salud de Catalunya (ESCA), 2002. * The Catalan Health Survey (ESCA) ** Visual Analogue Scale

Table A2. Descriptive statistics

* National Institute of Public Health Care (Spain)

Table A3 Specification and Decomposition of Health Inequalities