The data utilized in this study provides information on 440,000 individuals in 26 European countries. The European Union Statistics on Income and Living Conditions (EU-SILC) is specifically aimed at providing data that are comparable between countries. It is anchored in the European Statistical System (ESS) and collected by Eurostat. EU-SILC was launched in 2004 in 13 European Union (EU) countries, but has since grown, and data collected in 2007 include 26 European countries. The countries included are: Austria (n = 13,391), Belgium (n = 12,322), Cyprus (n = 8,470), Czech Republic (n = 19,384), Denmark (n = 11,610), Estonia (n = 11,971), Finland (n = 21,773), France (n = 20,357), Germany (n = 26,291), Greece (n = 12,346), Hungary (n = 18,490), Iceland (n = 6,567), Ireland (n = 10,892), Italy (n = 44,629), Lithuania (n = 10,913), Luxembourg (n = 7,913), Latvia (n = 9,270), Netherlands (n = 19,623), Norway (n = 11,706), Poland (n = 34,888), Portugal (n = 9,947), Spain (n = 28,656), Sweden (n = 14,204), Slovenia (n = 24,730), Slovakia (n = 12,573), and the United Kingdom (n = 17,484).
The reference population of EU-SILC is all private households and their current members residing in the territory of the relevant countries at the time of data collection. Persons living in collective households and in institutions are generally excluded from the target population. This data contains information about income and socio-economic status, as well as health-related variables and detailed labor-market information. The data are collected from different sources and by different modes; constructed, deducted from sample frame, deducted from sample design, settled by interviewers, collected from household respondent, collected from household members or collected from a proxy. It is thus a complex dataset in terms of data origin. We will now discuss specifically the health and income variables used.
In a survey, the traditional five-level self-assessed health variable (SAH), ranging from “very good” to “very poor” was obtained. The literature shows that SAH predicts mortality and morbidity, even in the presence of additional controls [26–36]. Furthermore, this measurement is frequently used and will thus increase the chances for other researchers to calculate comparable health CIs for other countries that can be held up against the current estimates. In all instances, the numeric values of the SAH variable are such that a higher number indicates worse health, and thus ill-health CIs are calculated. Weighted average ill-health in the sample is 2.260 (SE 0.0024). When comparisons are made across countries it is important that the same health measure, SAH, is used when CIs and ACIs are calculated as previous results have shown estimates to be sensitive to the health measures used . There are pros and cons to using the SAH variable in continuous form or dichotomized. Results in this study are qualitatively robust to such changes. However, especially when comparing other results to those reported here, it should be kept in mind that the reported results are those from estimations using the full information in SAH.
The main income measure used in the calculations is equivalized disposable household income, but calculations were made using individual gross employee cash or near cash income as well. Both of these measures were available for subjects of all countries, while most other income measures were only available for a few of the 26 countries and would thus have rendered cross-country comparisons less meaningful. What is furthermore important about those measures is that one measures the income rewards for individuals’ labor-market efforts, while the other one measures the resulting access to finances. This difference may be important, especially in countries with extensive income transfers, such as the Nordic countries. Individual gross employee cash or near cash income is the simpler of the two income measures. It includes the value of any social contributions and income taxes payable by an employee or by the employer on behalf of the employee to social insurance schemes or tax authorities . It may be argued that equivalized disposable household income includes fuller information about individuals’ access to finances than individual gross employee cash income. It is constructed from total disposable household income multiplied with a within-household non-response inflator factor used to correct the effect of non-responding individuals within a household. Countries using the factor are Germany, Greece, Latvia, Portugal, Slovakia and Spain, while other countries imputed missing personal interviews. This multiple is then divided by equivalized household size which assigns the value 1 to an adult in a household, 0.5 to each additional household member aged 14 and over and 0.3 to each household member aged 13 or less. Equivalized disposable household income thus measures access to finances through own and family-members income, taking into account economies of scale in household production.
Total disposable household income is the sum for all household members of gross personal income components (gross employee cash or near cash income, gross non-cash employee income, gross cash benefits or losses from self-employment (including royalties), unemployment benefits, old-age benefits, survivor benefits, sickness benefits, disability benefits and education-related allowances), plus gross income components at household level including income from rental of a property or land, family/children related allowances, social exclusion not elsewhere classified, housing allowances, regular inter-household cash transfers received, interests, dividends, profit from capital investments in unincorporated business, income received by people under 16, minus regular taxes on wealth, regular inter-household cash transfer paid, tax on income and social insurance contributions . However, some of these components are missing for a few countries (company car for France and Norway, regular taxes on wealth for Norway, sickness benefits for Italy, and non-cash employee income for the Netherlands), and therefore this measure is not entirely complete. However, those differences should not be expected to dramatically affect results. Income measures used were scaled to thousands of euros.
Other individual-level ill-health determinants used in the analyses are marital status, education, activity status, risk of poverty, as well as age and gender. Marital status dummy variables include married (reference category), divorced/separated, widowed and never married. Education levels are classified according to the International Standard Classification of Education (ISCED) and grouped into three categories: tertiary education (ISCED 5–6), upper secondary stage education or post-secondary non-tertiary education (ISCED 3–4), and lower secondary education or less (ISCED 0–2) (reference category). Activity status includes working full time (reference category), working part time, unemployed, student, retired, permanently disabled, in compulsory military community or service, fulfilling domestic tasks and care responsibilities, and other inactive. The risk of poverty threshold is set at household equivalized disposable income being less than 60% of its within-country median. All analyses are weighted using the cross-sectional personal weights provided within the EU-SILC.
Aggregate measures of income-related health inequalities are derived for the countries involved. The CI is based on the Lorenz curve, a cumulative frequency curve, which compares the distribution of a specific variable with the uniform distribution that represents equality. The ill-health-income concentration curve is a plot of the cumulative proportion of ill-health against the cumulative proportion of the population ranked by income. As such, it allows for examination of variations in one variable relative to variations in another variable. The income dimension is captured by the ranking of observations by income on the horizontal axis (with the least advantaged furthest to the left). The cumulative proportion of the ill-health variable is then represented on the vertical axis. The concentration curve can be compared with a diagonal line representing a uniform distribution, or perfect equality. The greater the deviation of the concentration curve from this line, the greater is the inequality.
The numeric representation that goes with the concentration curve is the CI or the concentration coefficient and corresponds to twice the area between the concentration curve and the diagonal line. The CI provides a measure of socio-economic inequality in health. It ranges from -1 to 1, with 0 representing perfect equality and -1 and 1 representing total inequality. The CI can be computed straightforwardly with individual-level data using a formula proposed by Kakwani, Wagstaff and van Doorslaer [38
(i = 1, …, n) is the ill-health score of individual i, μ is the mean level of ill-health, R
is the relative rank by income of individual i, σ
2 is the variance of R
, β is the CI, and ϵ
is the error term.
The CI has a number of advantages as a measure of income inequalities in health. Most important, it reflects the experience of the entire population and not just those of two extreme socio-economic groups, as measures frequently used by non-economists do. The CI would thus change if the sizes of various groups changed, even if their mean health did not. One limitation of the CI is the fact that if everyone’s health were to double, the value of the index would not change. Such a difference would be captured in the ACI, which scales the relative CI by the mean of the health variable used . That number would obviously increase if everyone’s health was enhanced and is thus a summary measure for ill-health and its distribution. That measure can thus be taken as an overall measure of the extent to which the overarching goals of a health care system have been reached.
The CI does not take into account the fact that demographics play a role in generating inequality in health. However, these factors can be taken into account by partitioning the CI into avoidable and unavoidable (age-gender) health inequality. Using the indirect method of standardization the inequality due to the age and gender composition of the sample can be computed and subtracted from the CI to obtain a standardized CI. This standardized CI then shows the health inequality that is potentially avoidable and thus relevant for policy purposes. Another approach, which is used here, is to decompose the relative CI into contributions of its various determinants, both unavoidable and avoidable, using the following linear regression model:
is the ill-health measure for individual i
is an ill-health determinant for regressor k
and individual i
is the error term. Given the relationship between y
, the CI for y
can be written as follows:
where μ is the mean of y, is the mean of x
is the CI for x
is the generalized CI for the error term. CI is thus equal to the sum of all the CIs of the k regressors weighted by the elasticity of y with respect to x
is the residual and reflects health inequality not explained by systematic variation across income in the x
Finally results are compared across countries with regard to varying social, cultural, and institutional conditions. It is natural to start with public health expenditures (data missing for Greece, Ireland, Italy and the UK), as they are partially intended to mitigate the health-income relationship. Similarly, we focus on general income inequality, as it may affect health inequality directly or through the design of the CI . A recent theoretical publication suggested that more equality in opportunity might cause greater health inequality , and thus we use public expenditures on education (data missing for Greece) as a proxy for equality of opportunity. Furthermore we use two measures of GDP, GDP expressed in euros per capita and GDP expressed in purchasing power standards (PPS) per capita, both scaled to million. All these aggregate measures come from Eurostat Statistics Database . Finally we use dummy variables for neighboring countries that share similar social and cultural attributes. Countries are categorized into the following areas: Nordic countries (Denmark, Finland, Iceland, Norway and Sweden), Eastern Europe (Czech Republic, Estonia, Latvia, Lithuania, Hungary, Poland, Slovakia and Slovenia), North-Western Europe (Austria, Belgium, Netherlands, Luxembourg, Germany, Ireland and UK) and Southern Europe (Cyprus, Greece, Portugal, Spain, Italy and France). All associations are examined using bivariate linear regressions. We emphasize that those bivariate regressions are not intended to be direct tests, but rather as a way to organize the data and present the patterns within it. All data were analyzed with Stata 11.0 software .