9Documentosde Trabajo9Documentos
de Trabajo2007
Pilar García GómezÁngel López Nicolás
Regional Differences in Socioeconomic Health Inequalities in Spain
Gran Vía, 1248001 BilbaoTel.: 94 487 52 52Fax: 94 424 46 21
Paseo de Recoletos, 1028001 MadridTel.: 91 374 54 00Fax: 91 374 85 22
2007-09 19/3/07 12:41 Página 1
Regional Differences in SocioeconomicHealth Inequalities in Spain
Pilar García Gómez 1
Ángel López Nicolás 1,2
1 P O M P E U F A B R A U N I V E R S I T Y2 P O L Y T E C H N I C U N I V E R S I T Y O F C A R T A G E N A
� AbstractThis working paper reports an analysis of income re-lated health inequalities at the autonomous commu-nity level in Spain using the self assessed health mea-sure in the 2001 edition of the Spanish NationalHealth Survey. We use recently developed methods inorder to cardinalise and model self assessed healthwithin a regression framework, decompose thesources of inequality and explain the observed differ-ences across regions. We find that the regions withthe highest levels of mean health tend to enjoy thelowest degrees of income related health inequalityand vice-versa. The main feature characterizing re-gions where income related health inequality is low isthe absence of a positive gradient between incomeand health. In turn, the regions where income relatedhealth inequality is greater are characterized by astrong and significant positive gradient betweenhealth and income. These results suggest that policiesaimed at eliminating the gradient between healthand income can potentially lead to greater reductionsin socioeconomic health inequalities than policiesaimed at redistributing income.
� Key wordsHealth inequalities, decomposition analysis, Spain.
� ResumenEste documento de trabajo presenta un análisis delas desigualdades de salud por motivos económicosen las comunidades autónomas españolas, basándo-se en la edición del año 2001 de la Encuesta Nacio-nal de Salud. Se utilizan métodos recientes para car-dinalizar y modelar el estado de salud autopercibidodentro de un marco de regresión, descomponer lasfuentes de desigualdad y explicar las diferencias ob-servadas entre CC. AA. Se observa que, en términosgenerales, las CC. AA. con los niveles medios de sa-lud más elevados son las que menos desigualdades ensalud sufren por motivos económicos, y viceversa. Elprincipal rasgo que caracteriza a las CC. AA. con pocadesigualdad de salud por motivos económicos es la au-sencia de un gradiente positivo entre renta y salud.A su vez, las CC. AA. que registran más desigualdadesde salud por motivos económicos se caracterizan, deforma marcada, por un gradiente positivo significativoentre salud y renta. Estos resultados apuntan a que laspolíticas enfocadas a eliminar el gradiente entre saludy renta serán más eficaces a la hora de disminuir lasdesigualdades de salud socioeconómicas que las polí-ticas de redistribución de la renta.
� Palabras claveDesigualdades en salud, análisis de descomposición,España.
Al publicar el presente documento de trabajo, la Fundación BBVA no asume res-ponsabilidad alguna sobre su contenido ni sobre la inclusión en el mismo dedocumentos o información complementaria facilitada por los autores.
The BBVA Foundation’s decision to publish this working paper does not imply any re-sponsibility for its content, or for the inclusion therein of any supplementary documents orinformation facilitated by the authors.
La serie Documentos de Trabajo tiene como objetivo la rápida difusión de losresultados del trabajo de investigación entre los especialistas de esa área, parapromover así el intercambio de ideas y el debate académico. Cualquier comenta-rio sobre sus contenidos será bien recibido y debe hacerse llegar directamente alos autores, cuyos datos de contacto aparecen en la Nota sobre los autores.
The Working Papers series is intended to disseminate research findings rapidly amongspecialists in the field concerned, in order to encourage the exchange of ideas and academ-ic debate. Comments on this paper would be welcome and should be sent direct to theauthors at the addresses provided in the About the authors section.
Todos los documentos de trabajo están disponibles, de forma gratuita y en for-mato PDF, en la web de la Fundación BBVA. Si desea una copia impresa, puedesolicitarla a través de [email protected].
All working papers can be downloaded free of charge in pdf format from the BBVAFoundation website. Print copies can be ordered from [email protected].
Regional Differences in Socioeconomic Health Inequalities in Spain© Pilar García Gómez and Ángel López Nicolás, 2007© de esta edición / of this edition: Fundación BBVA, 2007
EDITA / PUBLISHED BY
Fundación BBVA, 2007Plaza de San Nicolás, 4. 48005 Bilbao
DEPÓSITO LEGAL / LEGAL DEPOSIT NO.: M-16.013-2007IMPRIME / PRINTED BY: Rógar, S. A.
Impreso en España – Printed in Spain
La serie Documentos de Trabajo de la Fundación BBVA está elaborada con papel 100% reciclado,fabricado a partir de fibras celulósicas recuperadas (papel usado) y no de celulosa virgen, cumplien-do los estándares medioambientales exigidos por la legislación vigente.
The Working Papers series of the BBVA Foundation is produced with 100% recycled paper made from recov-ered cellulose fibre (used paper) rather than virgin cellulose, in conformity with the environmental stan-dards required by current legislation.
El proceso de producción de este papel se ha realizado conforme a las normas y disposicionesmedioambientales europeas y ha merecido los distintivos Nordic Swan y Ángel Azul.
The paper production process complies with European environmental laws and regulations, and has bothNordic Swan and Blue Angel accreditation.
La serie Documentos de Trabajo, así como información sobre otras publicaciones de laFundación BBVA, pueden consultarse en: http://www.fbbva.es
The Working Papers series, as well as information on other BBVA Foundation publications,can be found at: http://www.fbbva.es
C O N T E N T S
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2. Regional Differences in Health Care Arrangements at the Startof the XXIth Century in Spain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.1. Measurement of health . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.2. Measurement and explanation of inequality . . . . . . . . . . . . . . . . . . . . 123.3. Decomposing inequality between autonomous communities . . . . . . . 143.4. Statistical inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4. Data and Variable Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5. Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175.1. Measuring and decomposing inequality by autonomous community . . 175.2. Decomposing excess inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7. Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
About the Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1. Introduction
THE Spanish health care system has been decentralized to an unprecedent-ed extent in the course of the last 25 years. This process of devolution hascoincided in time with a major overhaul in the nature of its functions at anational level. Two major features of the nation wide reforms are the intro-duction of universal coverage and the development of the primary care net-work as the basic pillar of the system, shifting emphasis away from hospitalcare. The process of devolution has not been homogeneous, however. Someregions were transferred health care responsibilities as early as 1981 while asmany as 10 out of the 17 autonomous regions were transferred in 2002. It iswidely accepted that this fragmented process of devolution has interferedwith the aim of guaranteeing the system’s equity and quality (European Ob-servatory on Health Care Systems, 2000). In this paper we aim to investigatethe degree of income related inequality across regions for the Spanish pop-ulation in the year 2001. For this objective, we use recently developed meth-ods in order to model health status, decompose the sources of inequalityof health over income and explain the observed differences between regions.We shall use data from the 2001 Spanish National Health Survey, a healthsurvey which is representative at the regional level and contains data onhealth status, income and other socioeconomic characteristics. Our conten-tion in this paper is that the heterogeneity of resources and organizationalarrangements across regions might reflect in differences in the joint distribu-tion of health and income after controlling for other correlates of healthsuch as demographic structure, education, activity status etc. In this paper weset out to measure such differences. Our results indeed show that there areimportant geographical disparities: Basque Country, Navarra and La Riojaare the regions with the highest levels of mean health and simultaneously en-joy the lowest degree of income related health inequality. By contrast, Murciais the least favoured region in that its population report one of the lowest lev-els of mean health and suffers the greatest degree of income related healthinequality. Other territories where income related health inequality is highrelative to Basque Country include rich regions such as Madrid, the BalearicIslands and Catalonia. The main feature characterizing regions where in-come related health inequality is low is the absence of a positive gradient
5
between income and health. In turn, the regions where income relatedhealth inequality is greater are characterized by a strong and significantpositive gradient between health and income.
Section 2 briefly summarizes the characteristics of the Spanish healthcare system and provides background references within the Spanish litera-ture. Section 3 presents the methodology that we adopt for the measurementand modeling of health, the measurement of socioeconomic health inequal-ity and the explanation of its changes across space. Section 4 describes thedata set employed throughout the analysis. Section 5 presents the empiricalresults, section 6 discusses the policy implication of the results and section 7concludes.
pilar garcía gómez and ángel lópez nicolás
6
2. Regional Differencesin Health CareArrangements at the Start of the XXIthCentury in Spain
BY the start of the century health responsibilities were devolved to 7 re-gions with governments run by different political parties, with different de-mographic structures and traditions in the industrial organization of healthcare. This is compounded by the fact that Basque Country and Navarra havea distinctive fiscal arrangement which grants them more degrees of freedomin expenditure decisions. These two regions have given public coverage todental care for children since the end of the 80’s, for instance. The remain-ing 10 regions were managed by a central body until 2002, the INSALUD,but this did not guarantee a greater degree of homogeneity. Indeed, onesource of potential differences arose from the calendar of devolution. Cata-lonia (1981), Andalucía (1984), Basque Country and Valencia (1988), Nava-rra and Galicia (1991) and Canary Islands (1994) gained responsibilitiesfirst, but the remaining 10 regions have had a regional government for along period before they have gained health responsibilities. It has been ar-gued (European Observatory on Health Care Systems, 2000) that the co-existence of a central regulating body and a regional government generatedfrictions which have led to an uneven implementation of reforms. The Eu-ropean Observatory on Health Care Systems (2000) cites the case of the pri-mary care reforms in Galicia, which met opposition from the regional gov-ernment from the mid 80’s to the mid 90’s. Galicia finally gained health careresponsibilities in 1991 but the results from these frictions are present in re-cent data. By 2000, 81% of the Spanish population on average were coveredby the new primary care network but the fraction was 50%, the lowest, in Gali-cia. It is important to stress that benefiting from the reformed primary healthcare network is important for equity purposes. The old network consisted ofisolated outlets where general practitioners were typically available for two anda half hours per day (European Observatory an Health Care System, 2000).
7
Unsurprisingly, given the low quality of public primary care, the rich turnedto private outlets except when hospital care was needed. In constrast, thenew network comprises team based practices staffed by doctors and nurseswho have received specific training in family medicine and whose activitiesnot only included curative care, but also preventive care, health promotion,follow up of patients and services targeted to particular population groupssuch as the mentally ill, drug users etc.
The uneven development of the primary health care system reflects inmany indicators of primary health care coverage displaying variation acrossregions in 2001. The Ministry of Health (2000a) provides information for thepercentage of the population covered by specific primary health care pro-grams (these programs include, among others, vaccinations against flu forelderly people, prevention of heart diseases, care for patients with chronicdiseases such as hipertension, COPD, etc.). Heart disease prevention, forinstance, reached 70.6% of the target population in Aragón but less than50% in Murcia or Extremadura. Similarly, vaccination against flu for over65’s reached 65.2 of the target population in Castilla La Mancha but only54.3% in Madrid or 58.4% in Murcia.
There are also regional differences in the stock of capital available forhospital care. Data from the Ministry of Health (2002b) reveal that in 2001the average number of beds per 100,000 inhabitants was 386 but regionssuch as Andalucía (293.7), Castilla-León (208.75), Valencia (279.09) or Murcia(313) were well below the average. Moreover, the percentage of these bedsbelonging to the public sector varied remarkably around the Spanish averageof 73% reflecting the unequal extent to which the public sectors contractsout the provision of health care. In this sense Catalonia, at 36.8%, had the lo-west ratio of public to total beds. It is worth mentioning that these disparitiesin health care infrastructures across regions are not explained by differing de-grees of need related to demographics or morbidity and mortality. A study byPuig Junoy and López Nicolás (1995) showed that the best regions in terms ofthe ratio of stock of health care capital to health care need were Navarra, Ma-drid, Aragón and Basque Country, while Balearic Islands, Extremadura andGalicia were ranked in the lowest positions. Territorial disparities in the sup-ply of preventive services and high technology have also been found in a re-cent study (González, Urbanos and Ortega, 2004).
Thus the evidence suggests that by 2001 the Spanish health care systempresents a good degree of heterogeneity across regions. This does not necessar-ily lead to regional disparities in health outcomes, because differences in themanagement of resources and/or poverty alleviation efforts from other areas ofpolicy making might be more important at generating health differences be-
pilar garcía gómez and ángel lópez nicolás
8
tween populations, as pointed out by García Vargas and del Llano Señaris(2003).
Our contribution to the literature focuses in evaluating the extent towhich health is unequally distributed over income within each of the re-gions, controlling for other covariates of health such as demographic struc-ture, education and activity status. The Spanish literature contains relevantantecedents in the topic. Regidor et al. (1995, 1999, 2002) have found a sig-nificant pro-rich bias in the relationship between socioeconomic class (asdefined by several combinations of education levels and occupation) andoutcomes such as the SF-36 instrument, self-assessed health, prevalence ofchronic diseases, standardised death rates and risky habits. Van Doorslaer et al.(1997) use data from the Spanish National Health Survey 1987 and find thatthere is pro-rich inequality in self-assessed health as measured by the con-centration index. Van Doorslaer and Koolman (2004) again find a signifi-cant degree of pro-rich inequality using data from the 1996 Spanish wave ofthe European Community Household Panel. Thus we know that, on aver-age, there is pro-rich socioeconomic inequality in health outcomes in Spain.What we do not know, however, is how the degree of pro-rich socioeconom-ic inequality varies across regions. Indeed, Van Doorslaer and Koolman(2004) find significant regional effects in the determinants of self assessedhealth and the contributions to health inequality. This suggests that fully dis-aggregated regional analysis is bound to offer interesting evidence. In thissense the evidence contained in this paper is complementary to studies ofregional differences in health and health care such as Abad Díez and Carre-ter Ordóñez (2002), who find important regional disparities in life expec-tancy in a recent study and López-Cassasnovas et al. (2005), who argue thatpolitical decentralization need not raise inequities in health care.
regional differences in socioeconomic health inequalities in spain
9
3. Methods
3.1. Measurement of health
Our measure of health is derived from the respondent’s assessment ofhis/her health status during the year previous to the date of the interview.As in many health surveys, information on self assessed health (SAH) in theSpanish National Health Survey is presented in a categorical variable resultingfrom the following question: “During the last 12 months, would you say thatyour health has been i) very good, ii) good, iii) fair, iv) bad, v) very bad”.There are several methods for the cardinalisation of this measure ofSAH. A first approach (Van Doorslaer et al., 1997; Wagstaff and VanDoorslaer, 1994) would consist in assuming that SAH is an underlying latentvariable with a standard log-normal distribution and then assigning to eachobserved SAH category the mid point of the intervals of a standard log nor-mal as defined by the cumulative distribution of observed SAH categories. Anatural extension of the underlying latent variable approach would consistin modelling SAH with an ordered probit structure (Cutler and Richard-son, 1997; Groot, 2000). Since an ordered probit does not identify the scaleof the latent variable, this procedure requires ex-post rescaling to theinterval within which latent SAH is assumed to vary. The problem of ex-postrescaling can be solved by using external information on a generic healthmeasure in conjunction with categorical SAH. One alternative along thisline consists in using the mean value of generic health per SAH category toscore latent SAH. In a recent paper, Van Doorslaer and Jones (2003) com-pare these alternatives with a new procedure consisting in combining exter-nal information on the distribution of a generic measure of health with thedistribution of observed SAH in order to obtain the thresholds of generichealth that delimit the SAH categories. Given this information, SAH can bemodelled as an interval regression and no ex-post rescaling is necessary. VanDoorslaer and Jones (2003) show that this is the best procedure in terms ofthe ability to mimic the distribution of generic health departing from theSAH categories and the set of covariates used in the interval regression mod-el. Subsequently this procedure has been used by Van Doorslaer and Kool-man (2004) in their analysis of health inequalities in the European Union.
10
We adopt this method for this paper and, in common with their approach,we will use information on the empirical distribution of the Health UtilityIndex (HUI) in the 1994 Canadian National Population Health Survey.Thus, we assume that there is a stable mapping from HUI to the latent var-iable that determines reported SAH and that this applies not only for Ca-nadian, but also for Spanish individuals. Therefore, we compute the cumu-lative frequency of observations for each category of SAH and then find thequantiles of the empirical distribution function for HUI in the NPHS thatcorrespond to these frequencies. Table 3.1 presents the cumulative frequen-cies of the distribution of SAH and the corresponding quantiles in the dis-tribution of HUI.
Therefore, an individual who reports very bad health will be assumedto have a HUI level that belongs to the interval [0,0.34]. Similarly, the inter-vals for the remaining SAH categories are (0.34, 0.68] for the bad category(0.68, 0.86] for the fair category and (0.86, 1] for the good and very good cate-gories.
In short, our procedure to measure health consists in using the pre-dictions for the latent variable in the following econometric model
y*i = b1 + Sk
k = 2bk kki + ui
yi = j if mj – 1 < y*i ≤ mj j = 1,2,3,4 (3.1)
where ui is a standard normal random error term, j = 1,2,3,4 denote the verybad, bad, fair and good or very good SAH categories and mj, are thethresholds whose values are given by the intervals above. Therefore thehealth measure used in the subsequent analysis for the ith individual is giv-en by
y i* = b1 + Sk
k = 2b§k kki (3.2)
regional differences in socioeconomic health inequalities in spain
11
TABLE 3.1: Cumulative frequencies of SAH and quantiles of HUI
SAH Cum. Frequency HUI quantile
Very bad 1.64 0.34
Bad 7.20 0.68
Fair 29.70 0.86
Good 84.77 1
Very Good 100.00 1
The linearity of the resulting health measure, which is expressed inHUI units, is a useful feature at the time of computing and decomposinginequality measures as we will see below.
3.2. Measurement and explanation of inequality
The literature on health inequalities has recently adopted a standard toolfor the measurement of income related health inequalities: the concentra-tion index (CI) of health on income (Wagstaff, Van Doorslaer and Paci,1989). The concentration index has a similar interpretation to the more fa-miliar Gini index for pure health inequality. In fact, the two inequality mea-sures differ in the fact that the ranking variable is income (CI) rather thanhealth (Gini). This means that, unlike the Gini, which takes only non-neg-ative values, the standardized CI ranges between –1 and 1. A value of –1would mean that all health is concentrated in the poorest person, whereas avalue of 1 would result if all health were concentrated in the richest person. Avalue of zero would mean that health is equally distributed over income in thesense that the pth percentage of the population ranked by income has exactlythe pth percentage of total health for any p. Concentration indices have beenused in studies for the Spanish population previously (Rodríguez, Calongeand Reñé, 1993; Urbanos, 2001; Van Doorslaer and Wagstaff, 1999; VanDarslaer et al., 1997). Rodríguez, Calonge and Reñé (1993) and Van Doors-laer and Wagstaff (1999) measure the degree of equity in the financing anddelivery of health care by means of such indices and related measures such asthe Gini and Kakwani indices, while Van Doorslaer et al. (1997) use them forthe measurement of socioeconomic health inequality.
Suppose we are interested in calculating the CI coefficient for a mea-sure of health using individual data in a sample from the population of inter-est. Let yi denote a measure of health for the ith individual, i = 1,2,...N, andR’i denote the cumulative proportion of the population ranked by incomeup to the ith individual (their relative income rank).
Ignoring, for expositional purposes, the fact that in general samplingweights will be necessary, the CI of health on income is given by (see e.g.Van Doorslaer and Jones, 2003),
CI = (2) cov (yi, R’i) (3.3)y
where y- = E(yi). Now let yi be given by the following linear regression model
yi = b1 + Sk
k = 2bk kki + ei (3.4)
pilar garcía gómez and ángel lópez nicolás
12
where k is the number of regressors (x). By substituting this for yi, the CI of ycan be written as (Wagstaff, Van Doorslaer and Watanabe, 2003),
The first term in brackets is the elasticity of y with respect to xk evalu-ated at the sample means (≈ –k and y-) and CI’k denotes the concentration in-dex of xk against income. Thus this inequality measure can be decomposedinto an explained part and an unexplained part. The explained part can be use-fully broken down into the contributions of individual explanatory vari-ables. As for the unexplained part, it is a scaled measure of the covarianceof the residuals in the regression model with the position of the individualin the distribution of income. As such, the unexplained part should bezero if the regression model contains income as an explanatory variable(Gravelle, 2003).
As explained in section 3.1, our health measure is a linear combina-tion of the explanatory variables included in the interval regression model.Given the nature of the dependent variable in the latter model, no residualscan be computed so the decomposition reduces to the deterministic term inequation (3.5). Moreover, if we define the estimated health elasticity withrespect to determinant k as
h§k ≡ bk≈–k (3.6)y
then we can rewrite the decomposition in a way such that the CI is just aweighted sum of the inequality in each of its determinants, with the weightsequal to the elasticities. That is,
CI ≡ Skh§k CIk’ (3.7)
As mentioned by Van Doorslaer and Koolman (2004), the decomposi-tion also clarifies how each correlate of health contributes to total income-related health inequality: this contribution is the result of (i) its impact onhealth, and (ii) how unequally distributed over income it is.
regional differences in socioeconomic health inequalities in spain
13
CI = Sk
k = 2 ( bk≈k ) CI’k + ( 2 ) cov(ei, R’i) (3.5)y y
3.3. Decomposing inequality betweenautonomous communities
As put into practice by Van Doorslaer and Koolman (2004), we have usedthe approach suggested by Wagstaff, Van Doorslaer and Watanabe (2003) inorder to decompose the difference in inequality between autonomouscommunities. The method is a derivation of the well known Oaxaca decom-position whereby the difference between the CI’s of community i and com-munity j can be written as
DCI = CIi – CIj = Skhkj (CI’ki – CI’kj) + S
kCI’ki (hki – hkj) (3.8)
Then, the contribution of any variable to the difference in the in-come-related health inequality is decomposed as:
DCI’k = hkj (CI’ki – CI’kj) + CI’ki (hki – hkj) (3.9)
In practice, for each region, we shall compute the differences in in-equality (and contributions toward such difference) with respect to the re-gion with the smallest level of inequality, Basque Country. Moreover, in or-der to assess the relative importance of the inequality versus the health elas-ticity component in the contribution of each variable, we also compute therelative excess elasticity compared to Basque Country, i.e. (hki – hpv) / hpv,and the relative excess inequality, (CIki – CIpv) / CIpv.
3.4. Statistical inference
Many of the statistics that we are going to report are non-linear functions ofthe data whose sampling distributions are hard to obtain. For this reason weshall use bootstrapping methods in order to derive standard errors. Thebootstrap estimates for standard errors are computed following the five-stepapproach used by Van Doorslaer and Koolman (2004). The number of rep-lications has been set to 500.
pilar garcía gómez and ángel lópez nicolás
14
4. Data and VariableDefinitions
WE use the 2001 edition of the Spanish National Health Survey. This is na-tion wide survey collecting information on health and socioeconomic character-istics of individuals. The survey contains separate adults (16 + ) and chil-dren samples. The analysis in this paper is based on the adult sample. Thesampling scheme is a complex multi-stage stratified process whereby pri-mary strata are autonomous communities and, within the latter, sub-strataare defined according to residence area population size. Within substrata,municipalities (primary sampling units) and sections (secondary samplingunits) are selected according to a proportional random sampling scheme.Finally individuals are randomly selected from the sections. The survey doc-umentation includes weighting factors that correct for the fact that thenumber of observations within the primary strata is not proportional to ac-tual population. We use these weights whenever a nationwide statistic iscomputed.
The information contained in the data files do not allow the identifi-cation of all the primary sampling units (because municipalities with a pop-ulation below 100,000 are not identified). Similarly, information about thesecondary sampling units is omitted so it is impossible to control for clustereffects at either the municipality level or the section level.
The ranking variable is total monthly income earned by the house-hold. In the ENS this is measured as a categorical variable with 6 responsecategories. The midpoint of each income group was attributed to all house-holds in the category and this is subsequently divided by an equivalence fac-tor equal to (number of household members)0.5, to adjust for differences inhousehold size.
The initial ENS sample included 26,265 individuals from all the auton-omous communities, although the 399 observations from Ceuta and Meli-lla were dropped. From the remaining 25,866, we have dropped 66 becauseself assessed health was not reported, 6,532 whose household income wasmissing, 3,954 whose age was missing. A further 38 individuals with missingvalues for marital status, job status or education are dropped from the sam-
15
ple. As a result, the estimating pooled sample contains 15,276, which are di-vided across autonomous communities as follows: 1,488 are from Andalu-cía, 756 from Aragón, 683 from Asturias, 664 from Balearic Islands, 787from Canary Islands, 547 from Cantabria, 820 from Castilla-La Mancha,1,134 from Castilla-León, 1,324 from Catalonia, 1,220 from Valencia, 827from Extremadura, 1,045 from Galicia, 1,484 from Madrid, 641 from Mur-cia, 472 from Navarra, 820 from Basque Country and 564 from La Rioja.
pilar garcía gómez and ángel lópez nicolás
16
5. Empirical Results
5.1. Measuring and decomposing inequalityby autonomous community
As discussed in section 3.1, we specify and estimate an interval regressionmodel for the level of SAH inspired in the specification used by Van Doors-laer and Koolman (2004). It is useful to stress that this is not a structuralmodel for health and therefore its estimates cannot be given a causal in-terpretation. However, it might be interpreted as a reduced form staticmodel of demand for health whose estimates provide an indication of howexogenous changes in health determinants can affect the degree of so-cioeconomic inequality in health. The explanatory variables in this modelare:
i) the logarithm of equivalent household income;ii) 14 age-gender categories corresponding to age groups 16-19, 20-24,
25-29, 30-34, 35-39, 40-44, 45-49, 5-54, 55-59, 60-64, 65-69, 70-74, 75-79, 80 +for men and women (the omitted category corresponds to a woman agedbetween 16 and 19);
iii) 5 educational categories: university (omitted category), secondaryschool, primary school, reads and writes and illiterate;
iv) 4 marital status categories: single (omitted category), married, di-vorced, widowed;
v) 5 activity categories: family care (omitted category), employed, pen-sioner, unemployed and student.
The first row of table 5.1 contains the mean predicted values for HUIfor each of the regions. Note that there are important variations: Navarra,La Rioja and Basque Country are the three regions with the top scores formean HUI, while at the bottom of the league there are Canary Islands andMurcia. In its second row, table 5.1 also shows that the richest regions (interms of mean equivalised household income) are Asturias, Basque Coun-try, Balearic Islands, Madrid, Navarra and Catalonia, while the poorest re-gions are Extremadura, Andalucía, Castilla-La Mancha and Canary Islands.
17
pilar garcía gómez and ángel lópez nicolás
18
TA
BL
E5.
1:M
ean
pred
icte
d H
UI,
mea
n lo
g in
com
e an
d H
ealth
equ
atio
ns: i
nter
val r
egre
ssio
n co
effi
cien
ts p
er r
egio
n
And
aluc
íaA
ragó
n A
stur
ias
Bal
eari
c C
anar
yC
anta
bria
Cas
tilla
Cas
tilla
C
atal
onia
Val
enci
aE
xtre
mad
ura
Gal
icia
Mad
rid
Mur
cia
Nav
arra
Bas
que
La
Rio
jaIs
land
sIs
land
sL
a M
anch
aL
eón
Cou
ntry
Mea
n
pred
icte
d
HU
I0.
852
0.86
90.
861
0.85
50.
843
0.86
30.
871
0.86
30.
850
0.86
50.
851
0.85
40.
859
0.84
80.
891
0.88
00.
889
Mea
n L
og
Inco
me
11.2
7611
.505
11.6
9911
.574
11.3
3111
.455
11.3
1711
.387
11.5
4411
.403
11.2
2211
.413
11.5
7311
.377
11.5
4411
.654
11.4
39
Con
stan
t0.
4357
0.80
630.
5911
0.14
940.
6607
0.85
280.
7168
0.74
030.
4615
0.70
250.
7121
0.48
750.
2090
0.43
170.
9998
0.86
350.
8704
Log
Inc
ome
0.03
960.
0084
0.02
780.
0579
0.01
720.
0059
0.01
630.
0157
0.03
800.
0159
0.01
220.
0329
0.05
530.
0379
–0.0
053
0.00
160.
0049
F20-
24–0
.023
40.
0099
–0.0
142
0.02
080.
0032
–0.0
275
–0.0
324
–0.0
011
–0.0
072
–0.0
105
0.00
920.
0165
–0.0
277
–0.0
004
–0.0
705
–0.0
211
0.01
03
F25-
29–0
.009
1–0
.002
7–0
.028
9–0
.024
70.
0383
–0.0
143
0.01
48–0
.011
0–0
.038
8–0
.013
80.
0462
0.00
90–0
.008
60.
0102
–0.0
024
0.01
280.
0064
F30-
34–0
.020
10.
0107
–0.0
377
–0.0
124
0.02
90–0
.006
0–0
.006
4–0
.005
0–0
.020
4–0
.012
10.
0238
0.00
15–0
.008
70.
0116
–0.0
392
0.00
260.
0235
F35-
39–0
.049
6–0
.029
2–0
.029
4–0
.012
90.
0212
–0.0
632
0.00
86–0
.031
3–0
.041
2–0
.003
90.
0358
0.01
16–0
.046
20.
0023
0.00
06–0
.029
4–0
.018
2
F40-
44–0
.066
10.
0030
–0.0
176
–0.0
555
0.01
34–0
.081
1–0
.016
0–0
.027
7–0
.063
2–0
.063
0–0
.003
9–0
.009
0–0
.033
90.
0360
–0.0
320
–0.0
279
–0.0
142
F45-
49–0
.063
4–0
.040
5–0
.047
2–0
.062
40.
0188
–0.0
339
–0.0
299
–0.0
170
–0.0
445
–0.0
402
0.01
12–0
.021
8–0
.074
8–0
.038
2–0
.109
1–0
.052
7–0
.023
9
F50-
54–0
.070
8–0
.043
5–0
.074
7–0
.046
5–0
.027
2–0
.067
4–0
.005
7–0
.045
5–0
.077
4–0
.046
6–0
.062
5–0
.020
5–0
.073
9–0
.055
1–0
.066
6–0
.033
5–0
.030
4
F55-
59–0
.146
1–0
.054
0–0
.070
9–0
.025
3–0
.085
1–0
.130
4–0
.088
6–0
.046
0–0
.124
1–0
.053
6–0
.005
7–0
.044
3–0
.038
9–0
.073
4–0
.030
6–0
.053
5–0
.029
6
F60-
64–0
.143
0–0
.059
0–0
.033
90.
0148
–0.0
502
–0.0
833
–0.0
873
–0.0
631
–0.0
471
–0.0
744
–0.1
093
–0.0
540
–0.0
836
0.00
45–0
.052
6–0
.037
7–0
.051
0
F65-
69–0
.093
9–0
.037
9–0
.072
2–0
.088
9–0
.061
0–0
.092
5–0
.063
6–0
.115
6–0
.019
1–0
.068
5–0
.039
6–0
.069
1–0
.075
50.
0230
–0.0
660
–0.0
028
–0.0
581
F70-
74–0
.149
1–0
.102
8–0
.067
7–0
.058
40.
0119
–0.0
813
–0.1
511
–0.1
603
–0.0
652
–0.1
066
–0.0
539
–0.0
496
–0.1
295
–0.0
065
–0.1
143
–0.1
017
–0.0
516
F75-
79–0
.051
8–0
.058
4–0
.183
4–0
.032
6–0
.091
1–0
.070
6–0
.050
3–0
.099
8–0
.093
4–0
.089
2–0
.069
7–0
.118
6–0
.121
3–0
.128
1–0
.053
4–0
.051
2–0
.122
8
F80
–0.0
989
–0.1
309
–0.0
817
–0.0
219
–0.0
788
–0.0
626
–0.1
905
–0.1
434
–0.1
044
–0.0
818
–0.0
659
–0.0
917
–0.1
294
–0.0
589
–0.1
058
–0.0
380
–0.0
730
M16
-19
–0.0
049
–0.0
100
0.00
460.
0427
0.02
08–0
.046
60.
0074
0.01
10–0
.004
1–0
.011
60.
0206
0.02
94–0
.022
9–0
.003
6–0
.016
5–0
.011
40.
0116
M20
-24
–0.0
018
–0.0
059
–0.0
106
0.01
490.
0596
–0.0
255
–0.0
009
0.00
210.
0023
–0.0
215
0.02
06–0
.004
10.
0115
0.02
090.
0010
0.00
300.
0065
M25
-29
–0.0
208
–0.0
078
–0.0
128
0.02
770.
0171
–0.0
050
0.00
37–0
.001
7–0
.014
8–0
.007
20.
0173
0.01
410.
0010
0.01
22–0
.011
40.
0115
0.01
52
M30
-34
–0.0
145
–0.0
114
–0.0
513
0.01
540.
0347
–0.0
067
0.00
03–0
.008
9–0
.022
8–0
.026
60.
0280
–0.0
009
–0.0
123
0.02
64–0
.022
2–0
.032
20.
0146
M35
-39
–0.0
483
–0.0
068
–0.0
304
–0.0
122
0.01
32–0
.009
20.
0039
–0.0
139
–0.0
093
–0.0
164
0.02
35–0
.029
2–0
.022
60.
0447
0.00
07–0
.042
50.
0184
regional differences in socioeconomic health inequalities in spain
19
TA
BL
E5.
1 (c
ontin
uatio
n):
Mea
n pr
edic
ted
HU
I, m
ean
log
inco
me
and
Hea
lth e
quat
ions
:in
terv
al r
egre
ssio
n co
effi
cien
ts p
er r
egio
n
And
aluc
íaA
ragó
n A
stur
ias
Bal
eari
c C
anar
yC
anta
bria
Cas
tilla
Cas
tilla
C
atal
onia
Val
enci
aE
xtre
mad
ura
Gal
icia
Mad
rid
Mur
cia
Nav
arra
Bas
que
La
Rio
jaIs
land
sIs
land
sL
a M
anch
aL
eón
Cou
ntry
M40
-44
–0.0
240
–0.0
352
–0.0
157
0.02
38–0
.002
9–0
.041
70.
0062
0.00
54–0
.045
5–0
.006
60.
0236
–0.0
076
–0.0
100
0.03
14–0
.030
1–0
.026
20.
0053
M45
-49
–0.0
454
–0.0
283
–0.0
932
–0.0
571
–0.0
376
–0.0
430
0.00
27–0
.034
5–0
.035
1–0
.026
80.
0347
0.00
01–0
.058
60.
0415
–0.0
076
–0.0
090
0.00
45
M50
-54
–0.0
657
0.00
76–0
.067
7–0
.029
50.
0065
0.00
10–0
.032
6–0
.024
4–0
.022
8–0
.075
2–0
.030
7–0
.021
3–0
.021
50.
0080
–0.0
484
–0.0
508
–0.0
316
M55
-59
–0.0
405
–0.0
291
–0.0
952
–0.0
608
–0.0
038
0.03
96–0
.013
6–0
.041
4–0
.050
0–0
.072
80.
0471
–0.0
487
–0.0
381
0.02
19–0
.115
6–0
.066
1–0
.021
5
M60
-64
–0.0
710
–0.0
364
–0.0
738
0.02
340.
0401
–0.0
245
0.00
37–0
.054
90.
0017
–0.0
586
–0.0
103
–0.0
145
–0.0
575
–0.0
467
–0.0
122
0.05
52–0
.026
2
M65
-69
–0.0
089
–0.0
782
–0.1
213
0.02
95–0
.049
80.
0465
0.02
87–0
.014
6–0
.036
9–0
.065
50.
0671
0.00
61–0
.063
6–0
.005
2–0
.087
4–0
.017
50.
0077
M70
-74
–0.0
899
–0.0
185
–0.0
730
0.04
68–0
.043
80.
0398
0.01
32–0
.084
0–0
.002
9–0
.056
70.
0444
–0.0
874
–0.0
320
–0.0
271
–0.0
531
–0.0
134
–0.0
586
M75
-79
–0.0
469
–0.0
809
–0.0
901
–0.0
130
0.11
400.
0205
–0.0
330
–0.0
877
0.01
93–0
.145
60.
0096
–0.0
724
–0.0
574
–0.0
031
–0.0
406
0.02
66–0
.016
0
M80
–0.0
355
–0.1
356
–0.0
416
0.04
81–0
.138
8–0
.019
00.
0400
–0.1
328
–0.0
377
–0.0
932
–0.1
050
–0.0
875
–0.0
959
–0.1
345
–0.0
892
0.02
02–0
.159
0
Illit
erat
e–0
.034
7–0
.230
40.
0761
–0.0
502
–0.1
053
–0.2
411
–0.0
004
–0.0
764
–0.1
277
–0.0
621
–0.1
452
–0.0
322
–0.0
803
–0.1
149
0.00
58–0
.000
9–0
.144
2
Rea
ds a
nd
wri
tes
–0.0
057
–0.0
407
–0.0
032
0.02
30–0
.075
90.
0531
–0.0
258
–0.0
352
–0.0
492
–0.0
498
–0.0
134
–0.0
537
0.00
99–0
.044
1–0
.085
60.
0431
–0.0
447
Prim
ary
Scho
ol0.
0212
–0.0
181
0.00
370.
0217
–0.0
315
–0.0
107
0.01
49–0
.016
3–0
.014
70.
0126
0.00
43–0
.010
00.
0181
0.01
52–0
.001
20.
0129
–0.0
196
Seco
ndar
y
scho
ol0.
0294
–0.0
026
0.01
500.
0459
0.00
230.
0174
0.00
68–0
.017
8–0
.000
60.
0095
0.01
600.
0060
0.01
310.
0175
–0.0
113
–0.0
037
–0.0
054
Mar
ried
0.01
11–0
.000
5–0
.019
20.
0145
0.00
02–0
.012
0–0
.001
90.
0064
0.00
130.
0047
–0.0
023
0.00
800.
0222
–0.0
081
0.00
130.
0104
0.00
00
Div
orce
d/
Sepa
rate
d–0
.024
9–0
.040
5–0
.089
1–0
.001
1–0
.031
4–0
.085
6–0
.021
8–0
.030
00.
0076
–0.0
321
–0.0
785
–0.0
015
–0.0
120
–0.0
533
0.01
21–0
.044
6–0
.006
5
Wid
ow0.
0110
0.00
99–0
.068
30.
0564
0.01
86–0
.037
5–0
.008
70.
0167
0.02
43–0
.008
90.
0952
0.02
820.
0076
–0.0
098
–0.0
036
0.02
090.
0495
Em
ploy
ed0.
0122
0.02
790.
0207
0.02
960.
0178
0.00
570.
0063
0.01
210.
0091
0.02
870.
0280
0.02
910.
0379
0.01
290.
0068
0.03
43–0
.010
6
Pens
ione
r–0
.048
3–0
.013
1–0
.001
1–0
.068
4–0
.031
4–0
.076
4–0
.071
3–0
.030
6–0
.060
10.
0030
–0.0
790
–0.0
148
0.00
79–0
.075
3–0
.015
3–0
.047
9–0
.031
1
Une
mpl
oyed
0.00
20–0
.004
30.
0196
0.03
24–0
.035
7–0
.041
2–0
.002
9–0
.011
5–0
.012
60.
0241
0.02
680.
0047
0.03
360.
0190
–0.0
484
0.02
49–0
.029
5
Stud
ent
0.00
380.
0301
–0.0
051
0.01
380.
0319
–0.0
044
0.02
380.
0115
0.02
650.
0420
0.04
080.
0268
0.05
990.
0342
0.02
890.
0586
0.00
23
Not
e:va
lues
sig
nific
antly
diff
eren
t fro
m z
ero
(at P
< 0
.05)
in b
old
type
face
.
The data also shows differences in the demographic structure acrossregions. The age pyramid is widest at the base in Balearic Islands, Canary Is-lands, Andalucía, Valencia and Murcia whereas mean age is greater in Casti-lla León, Castilla La Mancha, La Rioja and Aragón. There are important dis-parities in the education levels of the population. Concerning education thedata show that in Castilla-La Mancha, Canary Islands, Extremadura and An-dalucía more than 13% of the population have not completed primaryschool. At the other extreme, Basque Country has the highest proportion ofuniversity graduates followed by Madrid, Murcia, Asturias and La Rioja.Concerning marital state, there are important differences too. In CanaryIslands 32% of the population are single, but in Catalonia the proportion is10% smaller. Another important difference is found in employment rates.In Balearic Islands, Catalonia and Madrid the proportion of population whodeclares to be in employment exceeds 50% whereas in Andalucía, Asturias,the two Castillas and Extremadura the proportion is below 40%. The figuresfor these descriptive statistics for demographics, education, marital statusand activity are available from the authors.
The results for Spain from the ECPH reported in Van Doorslaer andKoolman (2004) reflect a positive and significant association between the loga-rithm of equivalised household income and health. However, as can be seenin table 5.1, where the interval regression results for the separate regionalmodels are presented, in this case the estimates show a somehow heteroge-neous pattern. For Navarra, Basque Country, La Rioja, Cantabria, Aragón,Extremadura and Canary Islands the partial (log) income effect is not signif-icantly different from zero at conventional levels. The concentration of in-significant impacts along the east Cantabric coast (Basque Country, Canta-bria) and neighboring regions (Navarra, Aragón and La Rioja) wouldsuggest a sort of common geographical effect. As reported above, these arealso the regions with the highest mean HUI scores so this would suggest aconcave relationship between health and income, with the healthiest re-gions situated at points where the profile is flat. For the two Castillas, Valen-cia and Asturias the partial effect of income is significant at the 10% leveland the point estimates are small. In contrast, for Galicia, Murcia, Catalonia,Andalucía, Madrid and Balearic Islands the income effect is greater andclearly significant. The point estimates for Madrid and Balearic Islands havethe greatest absolute value. This is a striking result in the sense that Madridand Balearic Islands are rich regions. Thus, unlike the results reported inVan Doorslaer and Koolman (2004), the data do not generally support a neg-ative relationship between the strength of the (log) income effect and thelevel of regional income per capita.
pilar garcía gómez and ángel lópez nicolás
20
The patterns of health variations by demographics are similar to theevidence found by Van Doorslaer and Koolman (2004) for the 13 Europeancountries. In general women report less health than men all else held equaland for both genders the level of health decreases with age. However, inAragón, Asturias, Canary Islands, Valencia and Basque Country there is nota clear association between gender and health reported. Similarly, individ-uals within the two lowest educational categories (illiteracy and no formalqualifications) report a significantly lower level of health than those with sec-ondary or university education. Also, divorcees tend to report a lower levelof health than the rest of individuals. A surprisingly common feature formost of the regions, is the fact that, else equal, widows report a greater levelof health than other individuals. Concerning activity status, there are two sa-lient features. On one hand, those in employment tend to report betterhealth than the rest of individuals, although this effect is not significant atconventional levels for quite a few regions, it is particularly strong in BasqueCountry and Madrid. On the other hand, pensioners tend to report a signif-icantly lower level of health in Basque Country, Murcia, Andalucía, Extre-madura, Catalonia, Castilla-La Mancha, Cantabria and Balearic Islands.
In table 5.2 we report the concentration indices of predicted HUI andthe explanatory variables. A salient feature is that there is pro-rich healthinequality in all regions, with the bootstrapped standard errors showing thatthe concentration indices are all statistically significant. However, the mostprominent feature concerns the striking differences in the level of incomerelated health inequalities across regions. The regions with the highesthealth levels, i.e Navarra, Basque Country and La Rioja turn out to enjoythe lowest levels of income related health inequalities. At the other extremeMurcia has the highest concentration index, and it is closely followed by Ma-drid, Balearic Islands and Catalonia. Note that there are also differences in thedegree of equivalised household income inequality. The highest level of in-come inequality is found in Canary Islands, followed by Andalucía. At theother extreme Asturias, Navarra and Basque Country enjoy the lowest levelsof income inequality. The concentration indices for the age-sex controls re-veal that older people are concentrated in low income groups with and im-portant difference across genders because, for women, the concentrationinto low income groups starts operates at earlier ages, i.e. while for malesthe age at which concentration into low incomes takes place is 60+, for wom-en it is 45+. As one might expect, individuals with the lowest educationalattainments (illiteracy, basic literacy and primary schooling) are concen-trated into low incomes and those with secondary schooling or universitydegrees are concentrated in high incomes. In all regions there is pro-poor
regional differences in socioeconomic health inequalities in spain
21
pilar garcía gómez and ángel lópez nicolás
22
TA
BL
E5.
2:C
once
ntra
tion
indi
ces
of d
epen
dent
and
inde
pend
ent v
aria
bles
per
reg
ion
And
aluc
íaA
ragó
n A
stur
ias
Bal
eari
c C
anar
yC
anta
bria
Cas
tilla
Cas
tilla
C
atal
onia
Val
enci
aE
xtre
mad
ura
Gal
icia
Mad
rid
Mur
cia
Nav
arra
Bas
que
La
Rio
jaIs
land
sIs
land
sL
a M
anch
aL
eón
Cou
ntry
HU
I pr
edic
ted
0.01
920.
0181
0.01
420.
0235
0.02
100.
0144
0.01
710.
0158
0.02
290.
0141
0.01
690.
0195
0.02
400.
0265
0.00
890.
0055
0.00
94
Log
Inc
ome
0.02
210.
0214
0.01
540.
0207
0.02
460.
0209
0.02
200.
0198
0.02
060.
0204
0.02
060.
0209
0.02
150.
0204
0.01
850.
0190
0.02
07
F20-
240.
0345
0.02
060.
2271
0.07
210.
0672
0.23
51–0
.207
4–0
.016
30.
1535
0.13
890.
0292
–0.0
062
0.14
480.
2358
0.15
12–0
.014
10.
0993
F25-
290.
0330
0.37
790.
1314
0.35
210.
0557
0.06
280.
3388
0.06
320.
1805
0.20
010.
1780
0.13
080.
1527
0.07
200.
4246
0.38
430.
1843
F30-
340.
0662
0.19
990.
0548
0.16
460.
0744
0.06
890.
0612
0.12
190.
1922
0.04
600.
1315
0.19
820.
1959
0.06
990.
1631
0.11
830.
1387
F35-
390.
0625
0.13
080.
0259
–0.0
756
0.10
94–0
.124
70.
0689
0.03
230.
1169
0.07
38–0
.066
7–0
.069
4–0
.130
40.
0928
0.26
700.
2109
0.17
50
F40-
440.
1136
0.15
67–0
.143
90.
1186
–0.1
831
0.04
48–0
.060
40.
0644
0.01
49–0
.045
3–0
.007
10.
0961
0.02
710.
1536
0.13
430.
1009
0.00
04
F45-
49–0
.111
90.
0156
–0.0
903
0.12
56–0
.000
9–0
.035
1–0
.103
0–0
.019
5–0
.046
10.
0471
0.14
290.
0357
0.00
33–0
.212
30.
0793
–0.0
587
0.24
04
F50-
540.
0298
0.00
700.
1588
–0.1
117
–0.0
493
0.09
43–0
.010
1–0
.049
20.
0199
–0.1
124
0.12
35–0
.109
0–0
.081
00.
1077
–0.1
052
–0.1
280
0.02
27
F55-
59–0
.139
8–0
.123
10.
0294
–0.0
268
0.05
24–0
.141
2–0
.059
2–0
.066
7–0
.204
5–0
.088
00.
0575
–0.0
506
–0.1
451
–0.0
796
–0.0
526
0.00
34–0
.090
9
F60-
64–0
.332
7–0
.387
70.
0296
–0.2
158
–0.3
927
–0.3
813
–0.0
476
–0.1
557
–0.1
300
–0.3
110
–0.2
414
–0.1
546
–0.1
029
0.02
19–0
.427
4–0
.498
7–0
.188
7
F65-
69–0
.103
8–0
.264
8–0
.179
4–0
.306
0–0
.418
2–0
.166
5–0
.432
4–0
.341
0–0
.313
2–0
.261
0–0
.130
4–0
.098
2–0
.373
1–0
.260
1–0
.273
3–0
.255
9–0
.493
8
F70-
74–0
.085
3–0
.360
5–0
.200
6–0
.359
1–0
.162
7–0
.161
4–0
.327
6–0
.250
0–0
.385
5–0
.420
1–0
.143
5–0
.223
5–0
.412
4–0
.319
2–0
.444
2–0
.473
8–0
.316
3
F75-
79–0
.214
3–0
.477
8–0
.555
6–0
.350
2–0
.113
7–0
.234
6–0
.357
1–0
.399
7–0
.457
3–0
.218
9–0
.093
3–0
.180
9–0
.403
0–0
.213
7–0
.409
4–0
.147
9–0
.179
4
F80
0.07
16–0
.408
8–0
.641
5–0
.510
2–0
.046
3–0
.345
5–0
.285
6–0
.246
7–0
.344
6–0
.214
3–0
.139
4–0
.259
1–0
.339
9–0
.261
3–0
.342
8–0
.290
6–0
.458
6
M16
-19
0.01
140.
1356
–0.1
255
0.05
290.
0960
0.03
500.
0188
–0.0
031
0.24
360.
0814
0.00
43–0
.107
1–0
.116
80.
0855
0.22
02–0
.164
7–0
.003
5
M20
-24
0.16
470.
1470
0.12
580.
2195
0.15
620.
2216
0.20
170.
2443
0.15
980.
2979
0.19
530.
2322
0.21
190.
1047
0.24
860.
0449
0.21
40
M25
-29
0.17
120.
3172
0.29
320.
0875
0.17
780.
2963
0.24
830.
3521
0.24
170.
2329
0.35
530.
2616
0.18
050.
1042
0.36
080.
2439
0.25
24
M30
-34
0.15
100.
1499
0.20
000.
2373
0.18
550.
3478
0.43
680.
3462
0.40
640.
2400
0.17
980.
1262
0.22
820.
2826
0.35
360.
2268
0.26
37
M35
-39
0.11
430.
2096
0.14
100.
2121
0.10
530.
1520
0.24
040.
1707
0.06
930.
0746
0.18
580.
0892
0.08
69–0
.038
2–0
.006
30.
3330
0.21
67
M40
-44
0.10
290.
2918
–0.0
044
–0.0
382
–0.0
052
0.02
890.
2783
0.09
220.
1779
0.00
290.
1944
0.08
270.
0664
0.34
140.
3277
0.03
890.
1361
M45
-49
–0.0
306
0.19
46–0
.026
00.
0458
0.01
610.
1089
0.32
370.
1535
0.22
940.
0369
–0.1
039
0.06
170.
1378
–0.0
338
0.24
960.
0430
0.15
90
M50
-54
0.07
970.
1806
0.09
900.
2186
–0.0
163
–0.0
938
–0.0
706
0.25
99–0
.061
5–0
.045
50.
0008
0.21
310.
0091
–0.0
666
–0.1
879
–0.0
073
0.18
92
M55
-59
–0.0
465
0.10
180.
2679
0.15
46–0
.028
90.
2344
0.02
070.
2414
–0.1
715
0.13
890.
0811
–0.0
422
0.15
14–0
.217
8–0
.303
60.
2017
–0.1
162
M60
-64
–0.0
658
–0.1
490
0.02
24–0
.178
40.
0480
–0.0
846
0.00
88–0
.048
70.
0771
0.04
27–0
.189
2–0
.162
9–0
.036
1–0
.100
70.
0323
–0.1
161
–0.0
265
regional differences in socioeconomic health inequalities in spain
23
TA
BL
E5.
2 (c
ontin
uatio
n):
Con
cent
ratio
n in
dice
s of
dep
ende
nt a
nd in
depe
nden
t var
iabl
es p
er r
egio
n
And
aluc
íaA
ragó
n A
stur
ias
Bal
eari
c C
anar
yC
anta
bria
Cas
tilla
Cas
tilla
C
atal
onia
Val
enci
aE
xtre
mad
ura
Gal
icia
Mad
rid
Mur
cia
Nav
arra
Bas
que
La
Rio
jaIs
land
sIs
land
sL
a M
anch
aL
eón
Cou
ntry
M65
-69
–0.2
158
–0.2
339
–0.0
289
–0.4
126
–0.1
985
–0.3
639
–0.2
693
–0.1
717
–0.1
941
–0.1
471
–0.3
566
–0.3
075
–0.0
705
–0.4
091
–0.1
619
–0.2
608
–0.1
829
M70
-74
–0.2
456
–0.4
289
–0.2
280
–0.4
077
–0.3
691
–0.2
988
–0.3
407
–0.2
035
–0.1
910
–0.2
507
–0.3
484
–0.2
165
–0.3
347
–0.2
884
–0.3
245
–0.0
933
–0.4
337
M75
-79
–0.2
326
–0.3
879
–0.2
723
–0.3
689
–0.2
368
–0.1
569
–0.2
895
–0.1
267
–0.3
711
–0.0
969
–0.2
860
–0.2
411
–0.0
389
–0.3
670
–0.3
926
–0.1
965
–0.3
414
M80
–0.0
339
–0.4
291
0.03
05–0
.455
8–0
.145
5–0
.062
4–0
.241
6–0
.420
5–0
.605
5–0
.016
4–0
.403
5–0
.221
2–0
.015
7–0
.266
7–0
.500
7–0
.600
8–0
.178
2
Illit
erat
e–0
.457
2–0
.088
3–0
.541
0–0
.475
1–0
.429
8–0
.812
3–0
.462
7–0
.837
7–0
.508
9–0
.379
4–0
.408
8–0
.823
3–0
.640
1–0
.425
6–0
.082
7–0
.510
4–0
.139
7
Rea
ds a
nd
wri
tes
–0.2
568
–0.5
575
–0.3
924
–0.3
998
–0.2
813
–0.3
089
–0.3
124
–0.1
765
–0.4
956
–0.4
545
–0.3
670
–0.3
275
–0.4
877
–0.5
076
–0.7
031
–0.4
344
–0.1
954
Prim
ary
Scho
ol–0
.158
8–0
.263
0–0
.167
3–0
.255
5–0
.248
1–0
.214
4–0
.185
8–0
.171
0–0
.264
7–0
.181
3–0
.063
3–0
.188
3–0
.283
9–0
.196
7–0
.276
5–0
.229
1–0
.236
3
Seco
ndar
y
Scho
ol0.
0722
0.17
030.
0697
0.05
390.
0523
0.07
620.
1422
0.10
890.
1035
0.11
800.
0984
0.08
070.
0250
0.11
960.
1488
0.08
030.
1032
Mar
ried
–0.0
241
–0.0
447
0.01
25–0
.065
4–0
.018
4–0
.041
1–0
.044
2–0
.038
8–0
.007
4–0
.013
1–0
.069
8–0
.029
8–0
.051
7–0
.043
9–0
.076
0–0
.023
7–0
.040
0
Div
orce
d/
Sepa
rate
d–0
.043
00.
0978
–0.2
552
0.22
35–0
.018
6–0
.134
70.
2161
0.14
970.
0316
–0.1
470
0.22
650.
0360
–0.0
023
0.02
54–0
.129
40.
0729
0.06
44
Wid
ow–0
.057
5–0
.313
0–0
.400
2–0
.350
7–0
.200
7–0
.244
9–0
.142
3–0
.228
9–0
.365
8–0
.228
0–0
.058
6–0
.211
9–0
.254
8–0
.194
2–0
.276
9–0
.257
9–0
.202
0
Em
ploy
ed0.
2430
0.26
270.
1551
0.19
980.
1931
0.20
460.
3017
0.25
350.
1939
0.21
960.
2125
0.19
630.
1898
0.15
750.
2609
0.22
610.
2332
Pens
ione
r–0
.199
3–0
.303
5–0
.166
6–0
.332
9–0
.197
3–0
.247
7–0
.246
9–0
.188
7–0
.270
6–0
.180
3–0
.214
4–0
.247
0–0
.199
2–0
.277
5–0
.306
7–0
.218
9–0
.283
2
Une
mpl
oyed
–0.2
247
–0.2
619
–0.1
817
–0.0
762
–0.3
063
–0.3
758
–0.2
091
–0.0
459
–0.0
745
–0.2
578
–0.0
994
–0.1
475
–0.1
831
–0.2
972
–0.1
631
–0.2
383
–0.0
755
Stud
ent
0.08
040.
2528
0.11
220.
1319
0.07
250.
1452
0.00
660.
0534
0.16
910.
1163
0.09
390.
1297
0.04
490.
2731
0.20
73–0
.095
5–0
.070
8
inequality in the distribution of widowhood, as it might be expected fromthe fact that many individuals in this collective have a non-contributorypension as their main source of income. Finally note that pensioners andthe unemployed are concentrated within low incomes, whereas, as expected,employment is concentrated among high incomes.
Next we analyse the contributions of the explanatory variables to thedegree of income related health inequalities. These contributions arecontained in table 5.3, where results are aggregated by groups of variables(results for the full set of variables are available upon request). Part of the interregional differences in the degree of income related health inequality aredue to differences in the age-gender structure of the population and thefact that there is heterogeneity across regions in both the joint distributionsof age and gender with equivalised household income and the partial ef-fects of age and gender on health. We can standardize the concentration in-dex by age and gender by substracting the contributions of age and genderfrom the raw concentration index. The resulting figures are presented inthe second row of table 5.3. In general the standardized indices reveal thesame pattern as the raw counterparts, with Balearic Islands, Catalonia, Ma-drid and Murcia among the greatest levels of standardised inequality andNavarra, Basque Country and La Rioja at the opposite extreme. In the caseof Balearic Islands, the standardized index is greater than the raw one. Aswe mentioned before, the population in this region is younger than on av-erage, so this result suggests that the degree of income related health inequal-ity would be greater if Balearic Islands had a population with the averageSpanish age-gender distribution. On the contrary, the standardized indicesfor Madrid, Castilla-León, Castilla-La Mancha, Navarra, La Rioja and Aragónare notably smaller than their raw counterparts. There are striking varia-tions in the contributions of the age-gender structure to the overall level ofincome related health inequality. For instance, it accounts for more than64% of the raw index in La Rioja and more than 50% in Castilla-León andAragón. On the other hand, it barely accounts for about 15% of the raw in-dex in Murcia, Extremadura, Catalonia and Cantabria. The distribution ofeducational attainments accounts for a substantial part of income relatedhealth inequalities in some regions. In Canary Islands, Murcia, Extremaduraand La Rioja they contribute to roughly 20% of the raw concentration in-dex. Note that these are regions where the distribution of education is moreunequal: Canary Islands and Extremadura have a high proportion of individ-uals with less than primary schooling and Murcia and La Rioja have a highproportion of university graduates. At the other extreme, in Andalucía, As-turias, Balearic Islands, the two Castillas, Madrid and Navarra, education
pilar garcía gómez and ángel lópez nicolás
24
regional differences in socioeconomic health inequalities in spain
25
TA
BL
E5.
3:H
ealth
ineq
ualit
y co
ntri
butio
ns o
f gr
oups
of
expl
anat
ory
vari
able
s pe
r re
gion
And
aluc
íaA
ragó
n A
stur
ias
Bal
eari
cC
anar
yC
anta
bria
Cas
tilla
Cas
tilla
C
atal
onia
Val
enci
aE
xtre
mad
ura
Gal
icia
Mad
rid
Mur
cia
Nav
arra
Bas
que
La
Rio
jaIs
land
sIs
land
sL
a M
anch
aL
eón
Cou
ntry
C H
UI
pred
icte
d0.
0192
40.
0181
20.
0142
30.
0235
50.
0209
50.
0143
50.
0171
00.
0158
20.
0229
40.
0141
40.
0168
90.
0194
70.
0240
20.
0265
10.
0088
80.
0054
60.
0094
4
I* =
C-C
*0.
0160
00.
0097
10.
0102
60.
0240
80.
0169
10.
0123
00.
0107
90.
0071
60.
0195
10.
0094
50.
0144
70.
0153
80.
0182
20.
0232
60.
0014
20.
0041
60.
0033
3
Log
Inc
ome
0.01
157
0.00
238
0.00
583
0.01
626
0.00
571
0.00
163
0.00
465
0.00
411
0.01
066
0.00
426
0.00
331
0.00
916
0.01
603
0.01
036
–0.0
0128
0.00
040
0.00
132
As
%60
.12
13.1
640
.93
69.0
527
.25
11.3
727
.21
25.9
846
.46
30.1
519
.57
47.0
566
.75
39.0
8–1
4.39
7.27
13.9
9
Dem
ogra
phic
s0.
0032
40.
0084
10.
0039
8–0
.000
530.
0040
40.
0020
50.
0063
00.
0086
60.
0034
30.
0046
90.
0024
30.
0040
90.
0057
90.
0032
50.
0074
60.
0013
00.
0061
1
As
%16
.85
46.4
327
.96
–2.2
419
.28
14.3
036
.87
54.7
314
.95
33.1
614
.36
20.9
924
.12
12.2
683
.98
23.8
164
.70
Edu
catio
n0.
0011
10.
0025
70.
0003
40.
0003
90.
0064
60.
0020
40.
0006
10.
0007
20.
0035
50.
0020
60.
0045
00.
0027
6–0
.000
620.
0049
2–0
.000
15–0
.001
480.
0019
8
As
%5.
7814
.16
2.37
1.68
30.8
614
.23
3.54
4.54
15.4
814
.55
26.6
714
.20
–2.5
918
.57
–1.7
0–2
7.14
20.9
7
Mar
ital S
tatu
s–0
.000
21–0
.000
530.
0028
5–0
.002
62–0
.000
280.
0016
30.
0000
8–0
.000
74–0
.000
940.
0003
4–0
.000
87–0
.000
82–0
.000
920.
0003
80.
0000
0–0
.000
88–0
.001
07
As
%–1
.08
–2.9
420
.05
–11.
11–1
.34
11.3
40.
48–4
.66
–4.1
02.
42–5
.18
–4.2
3–3
.85
1.45
–0.0
1–1
6.04
–11.
39
Job
Stat
us0.
0035
30.
0052
90.
0012
40.
0100
40.
0050
20.
0070
00.
0054
50.
0030
70.
0062
40.
0027
90.
0075
30.
0042
80.
0037
40.
0075
90.
0028
50.
0061
20.
0011
1
As
%18
.34
29.2
08.
7042
.63
23.9
548
.76
31.9
019
.42
27.2
019
.73
44.5
821
.99
15.5
728
.65
32.1
111
2.10
11.7
2
accounts for a small share of the concentration index. Moreover, in the caseof Basque Country, education contributes negatively to income-related healthinequality. Although the contribution of marital status is small in general, it isrelatively high in some regions such as Asturias—20% of the CI—and BalearicIslands or Basque Country and Navarra , among others, where inequality inmarital status actually reduce the CI.
By far the most important contributors to income related health in-equality are equivalised household income itself and activity status. In Anda-lucía, Balearic Islands and Madrid the contribution of income exceeds 60%.For Catalonia, Galicia, Asturias and Murcia the contribution is in line withthe Spanish average. For some the regions where we cannot reject that thepartial effect of income is zero such as Aragón, Cantabria, Navarra, La Riojaand Basque Country the point estimate of the contribution is small (graph5.1 plots the elasticity of HUI with respect to log income against the Gini in-dex of log income in order to gauge the strength of the two components forthe contribution of income). Concerning the contribution of employment
pilar garcía gómez and ángel lópez nicolás
26
GRAPH 5.1: Elasticity of HUI to log income and income inequality
Balearic Islands
Madrid
Galicia
AndalucíaMurcia
Catalonia
Canary IslandsC. Mancha
Asturias
Navarra
Basque Country
C. LeónValencia
Extremadura
AragónCantabria
La Rioja
Gini of log income
Ela
stic
ity
of H
UI
to lo
g in
com
e
0.015 0.02 0.025
0.8
0.6
0.4
0.2
0
status, income related inequalities in the distribution of employment andpensioner status are the main drivers. In Basque Country these two factorstogether account for more than income-related health inequality itself. Thatis, if the rest of covariates had their effect neutralized, the CI for BasqueCountry would be a greater. In Balearic Islands, Castilla La Mancha, Extre-madura, Catalonia, Murcia, Basque Country, La Rioja and Cantabria, theunequal distribution of pensioner status accounts for a large fraction of in-come related inequality on predicted HUI.
5.2. Decomposing excess inequality
Which are the factors that generate more income related health inequalityin some regions? Table 5.4 provides the answer by showing the contributionof group of explanatory variables to the excess inequality of each regionwith respect to the region with the lowest CI, Basque Country (results forthe full set of variables are available upon request). We note that an over-whelming fraction of excess inequality, is attributable to income in Andalucía,Asturias, Balearic Islands, Catalonia, Galicia, Madrid and Murcia. Note thatamong the latter there are the top four regions in terms of CI. For the restof regions the contribution of income to excess inequality ranges between37% for Castilla-La Mancha and 14% for Cantabria. The contribution of pop-ulation structure is relatively unimportant in the regions with most inequal-ity. In Murcia it accounts for 9%, in Catalonia 12% and in Balearic Islandsthe population structure actually reduces excess inequality with respect toBasque Country. In contrast, the contribution of population structure is im-portant in regions whose degree of inequality is close to Basque Country. InNavarra it accounts for 180% of the difference, and in Aragón it accountsfor 56%. The contribution of education attainments exceeds 50% in CanaryIslands, and Extremadura and is above 25% for other regions with a high CIsuch as Murcia, Catalonia and Galicia. Note that in another region with ahigh CI, Madrid, the contribution of education to excess inequality is lessthan 5%. When assessing the contribution of employment status to excessinequality, note that income related health inequality in Basque Country isattributable to nearly exclusively (income related inequality in) employmentstatus. Therefore, it is not surprising to find that the contribution to excessinequality is negative for some regions such as Asturias, the two Castillas andValencia, Galicia and Madrid. In contrast, the unequal distribution ofemployment stata exacerbates inequality with respect to Basque Country inBalearic Islands, Extremadura and Cantabria.
regional differences in socioeconomic health inequalities in spain
27
pilar garcía gómez and ángel lópez nicolás
28
TA
BL
E5.
4:C
ontr
ibut
ions
of
grou
ps o
f ex
plan
ator
y va
riab
les
to e
xces
s he
alth
ineq
ualit
ies
per
regi
on
vers
us B
asqu
e C
ount
ry (
in %
of
exce
ss c
once
ntra
tion
inde
x of
HU
I in
fir
st r
ow)
(per
cent
ajes
)
And
aluc
íaA
ragó
n A
stur
ias
Bal
eari
cC
anar
yC
anta
bria
Cas
tilla
Cas
tilla
C
atal
onia
Val
enci
aE
xtre
mad
ura
Gal
icia
Mad
rid
Mur
cia
Nav
arra
La
Rio
jaIs
land
sIs
land
sL
a M
anch
aL
eón
Exc
ess
ineq
ualit
y25
2.4
231.
916
0.7
331.
328
3.7
162.
821
3.1
189.
732
0.1
159.
020
9.4
256.
633
9.8
385.
562
.672
.8
CI-C
I PV
Log
Inc
ome
81.0
15.7
61.9
87.7
34.3
13.9
36.6
35.8
58.7
44.5
25.4
62.6
84.2
47.3
–49.
023
.2
Dem
ogra
phic
s14
.156
.230
.5–1
0.1
17.7
8.5
43.0
71.0
12.2
39.0
9.8
19.9
24.2
9.3
180.
112
0.9
Edu
catio
n18
.832
.020
.710
.451
.339
.617
.921
.228
.840
.852
.430
.34.
630
.438
.987
.0
Mar
ital S
tatu
s4.
82.
742
.5–9
.63.
828
.28.
21.
3–0
.414
.00.
00.
4–0
.36.
025
.6–5
.0
Job
Stat
us–1
8.8
–6.5
–55.
621
.7–7
.19.
9–5
.7–2
9.5
0.7
–38.
412
.3–1
3.1
–12.
87.
0–9
5.6
–126
.1
6. Discussion
LET us now turn to the implications for policy prescriptions that onemight draw from these empirical results. The evidence suggests that, in or-der of importance, income, employment status and education are themost important drivers of differences in income related health inequalityacross regions. For the contribution of each of these factors there are twocomponents: its effect on health as measured by the elasticity and its de-gree of income related inequality. Thus policies aimed at reducing incomerelated health inequality could be directed to either reducing the impacton health of these factors or to altering the distribution of these factors(or both). In the particular case of income, policies could be directed to-wards eliminating the positive gradient between health and income (as itoccurs in Basque Country and other regions, where the gradient is null)or to make income more equally distributed. In order to gauge which ofthe two courses of action would potentially lead to a greater reduction in in-equality it is useful to present the relative differences (with respect toBasque Country) of the elasticity of health with respect to income and thedegree of inequality in the distribution of income (as measured by theconcentration index). The figures in table 6.1 reveal that, for most re-gions, the difference in elasticities is much greater than the difference inthe degree of income inequality. This suggests that, if differences in so-cioeconomic inequalities are to be reduced towards the Basque Countrybenchmark, investigating why the health-income gradient is steeper in therest of regions and correcting the causes can be more effective than mak-ing income more equally distributed. Furthermore, for the other driversof inequality we have also found that the differences in elasticities aremore important than the differences in how unequally distributed are thesefactors (results for the full set of variables are available upon request).While the scope of this paper consists primarily in providing an empiricalaccount of income related inequalities in SAH, it is interesting to suggestways in which the causes for the differences in the gradient betweenhealth and income might be ascertained. In order to do so, we present asimple plot of the elasticity of HUI to income against the regional indicesof public health care infrastructure adjusted by need derived by Puig-
29
Junoy and López Nicolás (1995). Graph 6.1 shows a clear inverse relation-ship among these two magnitudes, suggesting that differences in healthcare infrastructure might play an important role in understanding differ-ences in income related health inequality.
pilar garcía gómez and ángel lópez nicolás
30
TABLE 6.1: Relative excess elasticity and inequality (vs Basque Country)of log income per region(percentage)
Relative excess inequality Relative excess elasticity
Andalucía 16.0 2.413.1
Aragón 12.6 433.7
Asturias –18.9 1.711.4
Balearic Islands 8.9 3.663.8
Canary Islands 29.5 1.011.4
Cantabria 9.7 275.1
Castilla La Mancha 15.4 915.5
Castilla León 4.3 892.9
Catalonia 8.5 2.376.3
Valencia 7.1 903.8
Extremadura 8.0 671.4
Galicia 9.6 2.006.1
Madrid 13.0 3.474.5
Murcia 6.9 2.340.9
Navarra –2.8 –431.5
La Rioja 9.0 205.4
regional differences in socioeconomic health inequalities in spain
31
GRAPH 6.1: Relationship between the elasticity of HUI to equivalised log household incomeand an index of public health care infrastructure adjusted by need
Balearic IslandsMadrid
Galicia
AndalucíaMurciaCatalonia
Canary IslandsC. Mancha
Asturias
Navarra
Basque Country
C. LeónValencia
Extremadura
AragónCantabria
La Rioja
Need adjusted public health care stock
Ela
stic
ity
of H
UI
to lo
g in
com
e
50 100 200
0.8
0.6
0.4
0.2
0
150
7. Summary and Conclusion
IN this paper we have applied recently developed methodologies (VanDoorslaer and Koolman, 2004) to measure and explain the differences inthe degree of income related health inequality across Spanish regions.The results reveal important geographical differences. Basque Country,Navarra and La Rioja are the regions with the highest levels of meanhealth and simultaneously enjoy the lowest degree of income relatedhealth inequality. By contrast, Murcia is the least favoured region in thatits population report one of the lowest levels of mean health and suffersthe greatest degree of income related health inequality. Other territorieswhere income related health inequality is high relative to Basque Countryinclude rich regions such as Madrid, Balearic Islands and Catalonia.
The main feature characterizing regions where income relatedhealth inequality is low is the absence of a positive gradient between in-come and health. Nevertheless, even in these regions there is income relatedhealth inequality operating through inequality of employment status(Basque Country, Cantabria and Extremadura) or age-gender structure(La Rioja, Navarra, Aragón) over the distribution of income. In turn, theregions where income related health inequality is greater are characteri-zed by a strong and significant positive gradient between health and inco-me. In some cases this is reinforced by the effects of education (Catalonia,Galicia and Murcia).
In similarity to the results for 13 European countries reported in VanDoorslaer and Koolman (2004), we do not find substantial differences inthe degree of income inequality across regions, so the differential contribu-tions of income to socio-economic health inequalities are ascribed to het-erogeneity in the elasticities of health with respect to income across regions.This can be generalized to the other drivers of income related health in-equalities. In this sense the policy implications of these results are similar innature to Van Doorslaer and Koolman [17]: policies aimed at eliminatingthe gradient between health and income can potentially lead to greater re-ductions in socio-economic health inequalities than policies aimed at redis-
32
tributing income. Although we find a clear pattern of association betweenthe size of the elasticity of health with respect to income and an indicatorfor need adjusted health care infrastructure, it is necessary to obtain evi-dence on the causal pathways between health and income before beingable to propose specific policies.
regional differences in socioeconomic health inequalities in spain
33
References
ABAD DÍEZ, J. M. and C. CARRETER ORDÓÑEZ (2002): “Indicadores sanitarios por Comunidades
Autónomas”, Informe SESPAS.
CUTLER, D. and E. RICHARDSON (1997): “Measuring the Health of the United States Popula-
tion. Brooking papers on Economic Activity”, Microeconomics, 217-282.
«European Observatory on Health Care Systems», Health Care Systems in Transition, Spain.
GARCÍA VARGAS, J. and J. DEL LLANO SEÑARIS (2003): “El SNS después de las transferencias: ¿se-
guirá siendo un sistema? Fundación Encuentro”, Informe España 2003, Madrid.
GONZÁLEZ, B., R. M. URBANOS and P. ORTEGA (2004): “Oferta pública y privada de servicios sa-
nitarios por Comunidades Autónomas”, Gaceta Sanitaria, 18 (Supl1), 82-89.
GRAVELLE, H. (2003): “Measuring Income Related Inequality in Health: Standardisation and
the Partial Concentration Index”, Health Economics, 12, 803-819.
GROOT, W. (2000): “Adaptation and Scale of Reference Bias in Self-Assessment of Quality of
Life”, Journal of health economics, 19, 403-420.
LÓPEZ CASASNOVAS, G., I. COSTA FOND and I. PLANAS (2005). “Diversity and regional inequali-
ties: assessing the outcomes of the Spanish ‘system of health care services’”. Health Eco-
nomics 14: 5221-5235.
MINISTRY OF HEALTH (2002a): Atención Primaria en el INSALUD: Diecisiete años de Experiencia, Mi-
nisterio de Sanidad y Consumo, Subdirección General de Atención Primaria, Madrid.
— (2002b): Asistencia especializada. Actividad 2001, Ministerio de Sanidad y Consumo, Subdi-
rección General de Desarrollo, Madrid.
PUIG JUNOY, J. and A. López Nicolás (1995): “Assessing Health Care Infrastructure at the Re-
gional Level: a Statistical Approach”, Applied Economics Letters, 2, 463-466.
REGIDOR, E. et al. (1995): “Las desigualdades sociales y la salud en España. La salud y el siste-
ma sanitario en España”, Informe SESPAS, 19-43.
— et al. (1999): “Association between educational level and health related quality of life in
Spanish adults”, Journal of Epidemiology community Health, 53, 75-82.
— et al. (2002): “Comparing Social Inequalities in Health in Spain: 1987 and 1995/1998”,
Social Science and Medicine, 54, 1.323-1.332.
RODRIGUEZ, M., S. CALONGE and J. REÑÉ (1993): Spain in Equity in the Finance and Delivery of
Health Care. An international perspective, F. Rutten, Van Doorslaer and Wagstaff (editors),
New York, Oxford University Press.
URBANOS, R. (2001): “Measurement of Inequity in the Delivery of Public Health Care: Evi-
dence from Spain (1997)”, Fedea. Documento de Trabajo, 15.
34
VAN DOORSLAER, E. and A. JONES (2003): “Inequalities in Self-Reported Health: Validation of a
New Approach to Measurement”, Journal of Health Economics, 22, 61-87.
— and X. KOOLMAN (2004): “Explaining the Differences in Income-Related Health Inequali-
ties Across European Countries”, Health Economics, 13, 609-628.
— and A. WAGSTAFF (1999): “Equity in the Finance of Health Care: Some Further Internation-
al Comparisons”, Journal of Health Economics, 18, 263-290.
— et al. (1997): “Income-Related Inequalities in Health: Some International Comparisons”,
Journal of Health Economics, 16, 93-112.
WAGSTAFF, A. and E. VAN DOORSLAER (1994): “Measuring Inequalities in Health in the Pres-
ence of Multiple-Category Morbidity Indicators”, Health Economics, 3, 281-291.
— VAN DOORSLAER E. and P. PACI (1989): “Equity in the Finance and Delivery of Health
Care: some Tentative Cross-Country Comparisons”, Oxford Review of Economic Policy, 5,
89-112.
— E. VAN DOORSLAER, N. WATANABE (2003): “On Decomposing the Causes of Health Sector
Inequalities with an Application to Malnutrition Inequalities in Vietnam”, Journal of Econo-
metrics, 112, 207-223.
regional differences in socioeconomic health inequalities in spain
35
A B O U T T H E A U T H O R S *
PILAR GARCÍA GÓMEZ holds an MSc in economics from the University
of York and is a researcher at Centre de Recerca en Economia i Sa-
lut (CRES). She is currently researching for her PhD thesis at Pom-
peu Fabra University. She has written several articles in the area of
health economics.
E-mail: [email protected]
ÁNGEL LÓPEZ NICOLÁS holds a PhD in economics from the European
University Institute (Florence). He is an associate professor of ap-
plied economics at Pompeu Fabra University, currently in second-
ment at the Polytechnic University of Cartagena, and a researcher at
CRES. He has published several articles and book chapters in the
areas of demand analysis, health economics and applied microecono-
metrics.
E-mail: [email protected]
Any comments on the contents of this paper can be addressed to ÁngelLópez at [email protected].
* This paper derives from the project “La dinámica del estado de salud y losfactores socieconómicos a lo largo del ciclo vital. Implicaciones para las po-líticas públicas”, which is supported by the BBVA Foundation. We are grate-ful to Guillem López, Vicente Ortún, David Casado, Andrew Jones, XanderKoolman, Eddy van Doorslaer and participants at the 2004 ECuity II meetingin Helsinki for useful comments and suggestions. The views expressed in thispaper are those of the authors and not necessarily those of the funders orthe authors’ employers.
D O C U M E N T O S D E T R A B A J O
NÚMEROS PUBLICADOS
DT 01/02 Trampa del desempleo y educación: un análisis de las relaciones entre los efectosdesincentivadores de las prestaciones en el Estado del Bienestar y la educaciónJorge Calero Martínez y Mónica Madrigal Bajo
DT 02/02 Un instrumento de contratación externa: los vales o cheques.Análisis teórico y evidencias empíricasIvan Planas Miret
DT 03/02 Financiación capitativa, articulación entre niveles asistencialesy descentralización de las organizaciones sanitariasVicente Ortún-Rubio y Guillem López-Casasnovas
DT 04/02 La reforma del IRPF y los determinantes de la oferta laboralen la familia españolaSantiago Álvarez García y Juan Prieto Rodríguez
DT 05/02 The Use of Correspondence Analysis in the Explorationof Health Survey DataMichael Greenacre
DT 01/03 ¿Quiénes se beneficiaron de la reforma del IRPF de 1999?José Manuel González-Páramo y José Félix Sanz Sanz
DT 02/03 La imagen ciudadana de la JusticiaJosé Juan Toharia Cortés
DT 03/03 Para medir la calidad de la Justicia (I): AbogadosJuan José García de la Cruz Herrero
DT 04/03 Para medir la calidad de la Justicia (II): ProcuradoresJuan José García de la Cruz Herrero
DT 05/03 Dilación, eficiencia y costes: ¿Cómo ayudar a que la imagen de la Justiciase corresponda mejor con la realidad?Santos Pastor Prieto
DT 06/03 Integración vertical y contratación externa en los serviciosgenerales de los hospitales españolesJaume Puig-Junoy y Pol Pérez Sust
DT 07/03 Gasto sanitario y envejecimiento de la población en EspañaNamkee Ahn, Javier Alonso Meseguer y José A. Herce San Miguel
DT 01/04 Métodos de solución de problemas de asignación de recursos sanitarios Helena Ramalhinho Dias Lourenço y Daniel Serra de la Figuera
DT 01/05 Licensing of University Inventions: The Role of a Technology Transfer OfficeInés Macho-Stadler, David Pérez-Castrillo y Reinhilde Veugelers
DT 02/05 Estimating the Intensity of Price and Non-price Competition in Banking:An Application to the Spanish CaseSantiago Carbó Valverde, Juan Fernández de Guevara Radoselovics, David Humphrey
y Joaquín Maudos Villarroya
DT 03/05 Sistemas de pensiones y fecundidad. Un enfoque de generaciones solapadasGemma Abío Roig y Concepció Patxot Cardoner
DT 04/05 Análisis de los factores de exclusión socialJoan Subirats i Humet (Dir.), Ricard Gomà Carmona y Joaquim Brugué Torruella (Coords.)
DT 05/05 Riesgos de exclusión social en las Comunidades AutónomasJoan Subirats i Humet (Dir.), Ricard Gomà Carmona y Joaquim Brugué Torruella (Coords.)
DT 06/05 A Dynamic Stochastic Approach to Fisheries Management Assessment:An Application to some European FisheriesJosé M. Da-Rocha Álvarez y María-José Gutiérrez Huerta
DT 07/05 The New Keynesian Monetary Model: Does it Show the Comovement between Output and Inflation in the U.S. and the Euro Area?Ramón María-Dolores Pedrero y Jesús Vázquez Pérez
DT 08/05 The Relationship between Risk and Expected Return in EuropeÁngel León Valle, Juan Nave Pineda y Gonzalo Rubio Irigoyen
DT 09/05 License Allocation and Performance in Telecommunications MarketsRoberto Burguet Verde
DT 10/05 Procurement with Downward Sloping Demand: More Simple EconomicsRoberto Burguet Verde
DT 11/05 Technological and Physical Obsolescence and the Timing of AdoptionRamón Caminal Echevarría
DT 01/06 El efecto de la inmigración en las oportunidades de empleode los trabajadores nacionales: Evidencia para EspañaRaquel Carrasco Perea, Juan Francisco Jimeno Serrano y Ana Carolina Ortega Masagué
DT 02/06 Inmigración y pensiones: ¿Qué sabemos?José Ignacio Conde-Ruiz, Juan Francisco Jimeno Serrano y Guadalupe Valera Blanes
DT 03/06 A Survey Study of Factors Influencing Risk Taking Behaviorin Real World Decisions under UncertaintyManel Baucells Alibés y Cristina Rata
DT 04/06 Measurement of Social Capital and Growth:An Economic MethodologyFrancisco Pérez García, Lorenzo Serrano Martínez, Vicente Montesinos Santalucía
y Juan Fernández de Guevara Radoselovics
DT 05/06 The Role of ICT in the Spanish Productivity SlowdownMatilde Mas Ivars y Javier Quesada Ibáñez
DT 06/06 Cross-Country Comparisons of Competition and Pricing Powerin European BankingDavid Humphrey, Santiago Carbó Valverde, Joaquín Maudos Villarroya y Philip Molyneux
DT 07/06 The Design of Syndicates in Venture CapitalGiacinta Cestone, Josh Lerner y Lucy White
DT 08/06 Efectos de la confianza en la información contable sobre el coste de la deudaBelén Gill de Albornoz Noguer y Manuel Illueca Muñoz
DT 09/06 Relaciones sociales y envejecimiento saludableÁngel Otero Puime, María Victoria Zunzunegui Pastor, François Béland,
Ángel Rodríguez Laso y María Jesús García de Yébenes y Prous
DT 10/06 Ciclo económico y convergencia real en la Unión Europea:Análisis de los PIB per cápita en la UE-15José Luis Cendejas Bueno, Juan Luis del Hoyo Bernat, Jesús Guillermo Llorente Álvarez,
Manuel Monjas Barroso y Carlos Rivero Rodríguez
DT 11/06 Esperanza de vida en España a lo largo del siglo XX:Las tablas de mortalidad del Instituto Nacional de EstadísticaFrancisco José Goerlich Gisbert y Rafael Pinilla Pallejà
DT 12/06 Convergencia y desigualdad en renta permanente y corriente: Factores determinantesLorenzo Serrano Martínez
DT 13/06 The Common Agricultural Policy and Farming in Protected Ecosystems:A Policy Analysis Matrix ApproachErnest Reig Martínez y Vicent Estruch Guitart
DT 14/06 Infrastructures and New Technologies as Sources of Spanish Economic GrowthMatilde Mas Ivars
DT 15/06 Cumulative Dominance and Heuristic Performancein Binary Multi-Attribute ChoiceManel Baucells Alibés, Juan Antonio Carrasco López y Robin M. Hogarth
DT 16/06 Dynamic Mixed Duopoly: A Model Motivated by Linux versus WindowsRamon Casadesus-Masanell y Pankaj Ghemawat
DT 01/07 Social Preferences, Skill Segregation and Wage DynamicsAntonio Cabrales Goitia, Antoni Calvó-Armengol y Nicola Pavoni
DT 02/07 Stochastic Dominance and Cumulative Prospect TheoryManel Baucells Alibés y Franz H. Heukamp
DT 03/07 Agency RevisitedRamon Casadesus-Masanell y Daniel F. Spulber
DT 04/07 Social Capital and Bank Performance:An International Comparison for OECD CountriesJosé Manuel Pastor Monsálvez y Emili Tortosa-Ausina
DT 05/07 Cooperation and Cultural Transmission in a Coordination GameGonzalo Olcina Vauteren y Vicente Calabuig Alcántara
DT 06/07 The Extended Atkinson Family and Changes in Expenditure Distribution:Spain 1973/74 – 2003Francisco J. Goerlich Gisbert, María Casilda Lasso de la Vega Martínez
y Ana Marta Urrutia Careaga
DT 07/07 Análisis de la evolución de la dependencia en la tercera edad en EspañaDavid Casado Marín
DT 08/07 Designing Contracts for University Spin-offsInés Macho-Stadler, David Pérez-Castrillo y Reinhilde Veugelers
9Documentosde Trabajo9Documentos
de Trabajo2007
Pilar García GómezÁngel López Nicolás
Regional Differences in Socioeconomic Health Inequalities in Spain
Gran Vía, 1248001 BilbaoTel.: 94 487 52 52Fax: 94 424 46 21
Paseo de Recoletos, 1028001 MadridTel.: 91 374 54 00Fax: 91 374 85 22
2007-09 19/3/07 12:41 Página 1
Top Related