1
Inequalities of Opportunities in Health and Natural Reward: a European Perspective
Damien Bricard, PSL, Université Paris-Dauphine, Leda-Legos, France Florence Jusot, PSL, Université de Rouen, CREAM, France
Alain Trannoy, Aix-Marseille School of Economics, EHESS & CNRS, France Sandy Tubeuf, University of Leeds, Academic Unit of Health Economics, United Kingdom
WORK IN PROGRESS – DO NOT CITE
LAGV conference proposition – March. 2013
Abstract
Purpose:
This paper aims to quantify and compare inequalities of opportunities in health in Europe and to assess whether the way the correlation between effort towards health and circumstances empirically matters for the magnitude of inequalities of opportunities.
Methodology:
This paper considers two alternative normative ways of treating the correlation between effort and circumstances championed by Barry and Roemer, and combine regression analysis with inequality measures to compare inequality of opportunities in health within Europe according to the different normative principles. Data from the Retrospective Survey of SHARELIFE, which focuses on life histories of European people aged 50 and over are used.
Findings:
Our results show considerable inequalities of opportunity in health in Germany, Spain, Italy, Denmark, Greece and Belgium whereas Sweden and Switzerland show low inequalities of opportunities in health. The normative principle on the way to treat the correlation between circumstances and effort makes little difference in Austria, France, Czech Republic, Sweden and Switzerland whereas it appears to matter in the Netherlands, Poland, Germany, Spain, Italy, Denmark, Greece and Belgium.
Research limitations/implications (if applicable):
Our results suggest a strong social and family determinism of lifestyles in the Netherlands, Poland, Germany, Denmark, Belgium and the Mediterranean which emphasized the importance of inequalities of opportunities in health within those countries. In terms of public health and social policies, it appears that reducing social and unhealthy lifestyles reproduction across generations would provide important benefits on health. On the other hand Austria, France, and Czech Republic show high inequalities of opportunities in health mainly driven by social and family background affecting adult health directly, and so would require policies compensating for poorer initial conditions.
Keywords (Maximum 6):
2
Introduction
Inspired by the philosophical concept of equality of opportunity developed by Dworkin (1981), Arneson (1989), Cohen (1989), Roemer (1998), and Fleurbaey (2008), a number of recent publications in health economics have focused on drawing the line between legitimate and illegitimate causes of health inequalities (Sen 2002; Fleurbaey 2006; Dias and Jones 2007; Dias 2009; Fleurbaey and Schokkaert 2009; Dias 2010; Trannoy et al. 2010; Tubeuf et al. 2012; Jusot et al. 2013). The main argument is that differences in observed health outcomes are explained by factors for which the individual can be held responsible such as effort and by factors for which the individual should not be held responsible such as circumstances. The distinction between effort and circumstances is at the core of the implementation of equality of opportunities policies and is based on the concept of individual responsibility. Equality of opportunities principles recommend that the impact of individual responsibility, namely effort, on the outcome is respected; this is the principle of natural reward, and that the impact of characteristics independent of individual responsibility, namely circumstances, is compensated; this is the principle of compensation (Fleurbaey 1995). One requires therefore distinguishing the respective contributions of effort and circumstances to overall health inequalities, so that policy-makers are able to identify the effort which should be rewarded and the circumstances that should be compensated. The challenge when doing so is that the two components cannot be assumed to be independent and one needs to decide how the correlation between effort and circumstances should be treated. Two main alternative views have been debated in the literature within this context. According to Roemer (1998) effort should be respected inasmuch as effort is disembodied from the impact of circumstances; in other words the correlation between effort and circumstances is considered as circumstances and is independent from individual responsibility. On the other hand, according to Barry (transcription of Barry’s position according to Roemer 1998, p.21) effort should be entirely rewarded and the correlation of effort and circumstances does not require to be acknowledged. To illustrate the debate, let us consider the case of smokers; would we hold sons of smokers less responsible to smoke than sons of non-smokers? From Roemer’s viewpoint, sons of smokers are less responsible than sons of non-smokers for smoking however from Barry’s viewpoint, the parental circumstances are not relevant and sons of smokers are as responsible as sons of non-smokers for smoking. According to the viewpoint adopted, the magnitude of inequalities of opportunities in smoking will differ and this will have important implications on the implementation of the principle of natural reward and the principle of compensation. Empirical applications of this debate remain scarce (Jusot et al. 2013) and inequalities of opportunities in health have never been considered at the European-level.
This paper aims to quantify and compare inequalities of opportunities in health in Europe and to assess whether it empirically matters to adopt Barry’s or Roemer’s view on the magnitude of inequalities of
3 opportunities. In particular, the paper investigates whether the correlation between effort and circumstances differ from one country to another. We use data from the Retrospective Survey of SHARELIFE, which focuses on life histories of European people aged 50 and over in 2008/2009. Tackling health inequalities in later life and improving the underlying socioeconomic determinants for older people is at the core of the European Union healthy-ageing strategy and health-promotion activity (Jagger et al. 2007; Marmot et al. 2008). Healthy life expectancy has increased in the late decades in Europe (Jagger et al. 2009) but this improvement is not equally distributed among older people (Crimmins and Cambois 2003; Jagger et al. 2008; Masseria et al. 2006) and among European countries. In this context, the analysis of inequalities of opportunity becomes especially relevant. The remainder of the paper are as follows. Section 2 presents the Methods and in particular the econometric model, section 3 describes the data, section 4 presents results on the explanatory factors of overall health inequalities in Europe and focuses on the findings on inequalities of opportunities in health within and between European countries. A discussion and concluding remarks form the final section.
Methods
We empirically assess how Roemer and Barry respective viewpoints matter for the measurement of inequalities of opportunity in health using a two-step methodology as suggested in Jusot et al. (2013). In the first step, a reduced-form model is estimated to measure the association between health status and circumstances and efforts, respectively. In the second step, predicted variables are used to measure the magnitude of health inequalities and to compare inequality of opportunity in health between European countries.
Estimation strategy
Let us assume that individual latent health status H is a function of circumstances C, effort E, demographic variables D and a residual term u:
) , , , (C E D u f H (Eq. 1)
The vector of circumstances C consists of a set of variables beyond individual control related to health status such as childhood conditions and family background on health status in adulthood. The vector of effort E captures individual responsibility for health, such as lifestyles. The vector of demographic variables D captures biological determinants such as age and sex.
We rely on a reduced form model because we are primarily interested in capturing correlations between health and effort; health and circumstances and finally effort and circumstances. In particular, we are not including current socioeconomic status among the regressors because it could be
4 endogenous variables and may be correlated with past health (see for example Adams et al. 2003 and Adda et al. 2003 for discussion on this issue) and with parental characteristics and individual effort. According to Barry, individual effort has to be fully respected whatever the influence of past circumstances on effort decisions. It allows directly regressing circumstances and effort variables on health status to measure the correlation between health status and individual effort in health capital investment on the one hand, and the correlation between health status and circumstances on the other. The health status HiBof individual i within Barry’s context can then be written as follows:
i i B i B i B B B i C E D u H
(Eq. 2)Equation (Eq. 2) allows us to test the condition of equality of opportunity in Barry’s view by testing the equality of
ˆB to zero. Independence between Ci and Ei is not required.Following Roemer (1998), equality of opportunity requires that effort is purged of any contamination coming from circumstances so that it represents pure individual effort. This concept leads us to estimate an auxiliary equation regressing the effort Ei of individual i against her circumstances Ci. It
allows isolation of a residual term ei, the relative efforts, which represent individual efforts purged
from any circumstances:
i i
i C e
E
. (Eq. 3)We then substitute the vector of actual efforts Ei with the estimated relative efforts eˆi in the equation
of health status (Eq. 2) and the health status HiR of individual i within Roemer’s context can be written in as follows: i i R i R i R R R i C e D u H
ˆ
(Eq.4)Equation (Eq. 4) allows us to test the condition of equality of opportunity in Roemer’s view by testing the equality of
ˆR to zero since Ci and ei are independent.As we are in a linear model,
ˆR is the same as
ˆB in (Eq. 2) according to the Frisch-Waugh-Lowell theorem. However and are different because in Roemer’s approach, the coefficient of circumstances incorporate the indirect effect of circumstances on efforts. This indirect term is the product of both the coefficient of efforts and the coefficient of circumstances in Barry’s approach.R
5
Inequality measurement
We estimate equations (Eq. 2) and (Eq. 4) as linear probability models and predict a linearly decomposable measure of health status within each framework. As we are interested in quantifying the magnitude of health inequality related to each of its sources, we argue in favor of the use of the variance because of its natural decomposition (Shorrocks 1982).
In each context j= {B (Barry); R (Roemer)} the decomposition of the variance of health status
)
(
2H
is given by: ) , ˆ cov( ) , ˆ cov( ) , ˆ cov( ) , ˆ cov( ) ( 2 H H H H H H H H H Cj Ej Dj Rj (Eq. 5)Using this decomposition, we propose three different measures to assess the magnitude of inequalities of opportunity in health between European countries.
First, we consider an absolute measure of inequality of opportunities in health using the contribution of circumstances-related health source that is given for Barry and Roemer with j=B, R by:
) , ˆ cov( 1 j j C j H H EOP (Eq. 6)
Then, we propose a first relative measure of inequality of opportunities in health corresponding to the contribution of circumstances-related health source divided by the overall health inequality as measured by the variance within the country:
) ( ) , ˆ cov( ) ( 2 2 1 2 j j j C j j j H H H H EOP EOP (Eq. 7) j
EOP2 is intuitively a measure of inequality of opportunity in health as a proportion of the overall health inequality.
Lastly, we consider a second relative measure of inequality of opportunities in health using the contribution of circumstances-related health source divided by the sum of both relevant sources of inequality (effort and circumstances) within the country:
) , ˆ cov( ) , ˆ (cov( ) , ˆ cov( ) , ˆ cov( 1 1 3 j j E j j C j j C j j E j j j H H H H H H H H EOP EOP EOP (Eq. 8) j
EOP3 is intuitively a measure of inequality of opportunity in health as proportion within the health inequality as explained by effort and circumstances.
6
Data
For the purpose of this study, we mainly used the third wave of the Survey of Health, Ageing and Retirement in Europe (SHARE) called SHARELIFE, the Retrospective Survey focusing on people's life histories. We also used additional information collected at Wave 1 and Wave 2. The SHARE is a multidisciplinary database representative of the European population aged 50 and over in Scandinavia (Denmark and Sweden), Central Europe (Austria, France, Germany, Switzerland, Belgium, and the Netherlands), and the Mediterranean (Spain, Italy and Greece), as well as two transition countries (the Czech Republic and Poland). Additional information about the dataset is available in Börsch-Supan et al. (2005).
We considered a sample of 20,946 individuals (9,447 men and 11,499 women) aged between 50 and 80 years old. The variable of interest is health in adulthood as measured by self-assessed health (SAH). Respondents were asked to rate their own health on a five-point categorical scale ranging from poor to excellent health status. We used SAH as a binary variable taking the value one if the individual rates her health as “good” or better, and zero if she rates her health less than “good”. Self-assessed health has been shown to be a good predictor of mortality, morbidity and subsequent use of health care (Idler and Benyamini, 1997). Table 1 provides the distribution of the sample according to self-assessed health. 62.5% of the European sample reports a good, very good or excellent self-assessed health status. However, we can notice important differences between countries, the proportion of individuals reporting a good health status varies from 34% in Poland to 79.7% in Switzerland.
Three sets of variables were considered in the study: circumstances that matter for health collected in SHARELIFE, effort as measured by health-related behaviours that were collected in the first and second waves of SHARE and demographic characteristics such as age and sex. Table 2 reports descriptive statistics for demographic characteristics, circumstance and effort variables at European-level.
The vector of circumstances included social conditions in childhood, parents’ longevity and parents' health-related behaviours. Social conditions in childhood include the main breadwinner occupation during childhood, the number of books at home as a proxy of parents' educational level, indicators of living conditions at home and indicators of period of difficulties during childhood. Main breadwinner occupation during childhood is described with the ISCO classification (International Standard Classification of Occupations) into six groups (i) “senior managers and professionals”, (ii) “technicians and associate professionals and armed forces”, (iii) “office clerks, service and sales workers”, (iv) “skilled agricultural and fishery workers”, (v) “craftsmen and skilled workers” and, (vi) “elementary occupations and unskilled workers”, and an additional category is added if individuals reported no breadwinner figure at home during childhood. Regarding main breadwinner occupation,
7 most of the respondents in Europe have a parent who was skilled agricultural or fishery worker (26.8%), or craftsman or skilled worker (26.2%) whereas only 8.1% of the sample is born from a father who was manager or professional. The number of books at home is a four categories variable starting from a first category with individuals declaring to have none or very few books (0-10 books) to a last category with individuals describing to have enough to fill two or more bookcases (more than 100 books). We use two variables measuring living conditions. The first variable measures the number of rooms per household member at home when the respondent was 10 and the second variable indicates the number of facilities available in the accommodation when the respondent was 10 such as having cold running water supply or central heating for example. Lastly, we use two indicators of financial difficulties during childhood: individual report of economic hardships and report of hunger episodes before the respondent was aged 16. Parents' longevity is a measure for each parent stating whether they are still alive at the time of the survey and if not, their age at death. We divided the group of deceased parents into two groups with respect to the median age at death of their own country: those who died earlier and those who died at the median age or later. Considering the age of the respondents, it is not surprising to find that only 10.4% of the fathers are still alive and 26.3% of the mothers. Regarding parents’ health-related lifestyles, we use a parental smoking indicator taking the value one if at least one of the two parents was reported to be a smoker, an indicator of parental problem concerning alcohol and an indicator concerning parental health care attitudes which indicates the lack of regular visits to the dentist for their children.
As effort variables, we considered three lifestyles binary variables: smoking status, obesity status and sedentary lifestyles. Smoking informs if the respondent reported to be a current smoker in at least one of the past waves; it equals one if she did currently smoke and zero otherwise. Obesity status is constructed using the reported height and weight and calculating individuals’ body mass index (BMI), it takes the value one if the respondent’s BMI was higher than 30 in one of the past waves and zero otherwise. Sedentary lifestyle measures whether the respondent was hardly ever or not at all engaged in activities requiring a moderate level of energy in one of the past waves; it equals one if she was never engaged and zero otherwise.
Results
The main results of interest of the paper are the comparisons of the magnitude of inequality of opportunity and the differences according to the two alternative viewpoints across European countries. In the first part we will give an overview of the determinants of health inequalities in Europe as a whole commenting on the regression analysis results for the health equation in the two alternative viewpoints. We will then focus on the results on cross-country differences in inequality of opportunity.
8
Barry and Roemer regressions at European-level
The results of both linear probability models are presented in Table 3 and are provided as marginal effects of circumstances and efforts on the probability of reporting excellent, very good or good health at European-level within each scenario (columns 2 and 3). Marginal effects differ across scenarios as the variables in each model are different from a mathematical point of view: actual effort is measured as dummy variables in Barry's model whereas relative effort is measured as continuous variables in Roemer's model. However, as we use linear probability models in auxiliary equation, the marginal effects of effort variables are identical in both scenarios while the marginal effects of circumstance variables differ in Roemer's scenario according to the extent circumstances are correlated to efforts. There are clear differences in magnitude of the marginal effects of circumstance variables between Barry’s scenario and Roemer’s scenario; the marginal effects of circumstances being larger in Roemer’s scenario. However results remain similar in terms of signs and relatively close in terms of significance considering both specifications. It appears that any circumstances included in the model are significantly associated with the probability of reporting good health in Europe. Higher social background is strongly and significantly associated with the probability of reporting a good health status. Individuals born in a family where the main breadwinner was a senior manager or professional worker significantly report a better health status than individuals born of an elementary occupation or an unskilled worker. The number of books at home during childhood is also found to be related to a better health status in adulthood as individuals reporting to have had enough book to fill at least one shelf significantly report a better health status than those reporting none or very few books at home. Moreover, we note a significant and positive effect of housing characteristics during childhood; the probability of reporting a good health status is increasingly associated with the number of room per household members and the number of facilities at home. Period of difficulties during childhood also significantly contributes to the probability of reporting a good health status with a 0.12 point decrease in the case of economic hardships and 0.06 point decrease in the case of hunger episodes. Parents’ health also drives health disparities: having a father or a mother who died in older ages or who is still alive at the time of the survey is associated with a higher probability of good health status in adulthood. Finally, we also find a negative and significant effect of parents' poor health-related behaviours such as the lack of regular visits to the dentist for their children, parents' smoking and parents' alcoholic consumption during childhood.
If we now turn our attention to the marginal effects of the three individual efforts variables, smoking, being obese and lack of activity are found significantly and negatively associated with good health.
9
Cross-country comparison of inequality of opportunity
Using the estimated coefficients of the health equations, we can assess how the magnitude of legitimate health inequalities and illegitimate health inequalities, namely inequalities of opportunity in health changes within the alternative views. Roemer’s view would a priori maximise the magnitude of inequalities of opportunities in health.
The first column of Table II is based on Europe as a whole and gives the magnitude of health inequalities using the variance of the predicted latent health status and then various insights on the differences in magnitude of inequalities of opportunities in health within each scenario. We find significant inequalities of opportunity in both Barry and Roemer scenarios. In particular, inequality of opportunity stands for about almost 50% of the relevant observed inequality (inequality due to circumstances and efforts) in Barry’s scenario and 57.5% in Roemer’s scenario; the difference between Roemer and Barry being significant and representing about 14.4% of the inequality measured in Roemer scenario.
When focusing on each European country more specifically, we find significant inequalities of opportunity in health within all of them. More particularly, inequalities of opportunity in health are larger in both absolute and relative terms in Austria, Germany, Spain, France, Denmark, Greece, Belgium, Italy and Czech Republic with significance levels for inequality measurement in those countries at less than one per cent. Inequalities of opportunity in health are however lower in Sweden, the Netherlands and Poland with significance levels at five per cent and in Switzerland with significance levels at ten per cent.
If we turn our attention now to the magnitude of share of inequalities of opportunities in health relative to the share of overall health inequality explained by both circumstances and effort in the country as measured by j
EOP3 , the ranking within inequalities of opportunities is considerably changing. Inequalities of opportunities in health represent 30% in Belgium and the Netherlands whereas they represent more than 70% in Spain, Greece and Czech Republic. There are therefore considerable differences between the share of legitimate health inequalities (inequalities related to effort) and illegitimate inequalities (related to circumstances) between European countries.
Differences between Roemer’s scenario and Barry’s scenario are also observed and found significant between most the countries (except in Sweden and Switzerland). Those differences can be interpreted as a strong link between effort and circumstances within those countries. In other words, people’s effort as measured by lifestyles is likely to be strongly determined by circumstances as measured by family and social background. This is particularly true in Germany, The Netherlands, Spain, Italy, Denmark, Greece, Belgium and Poland and to a lesser extent in Austria, France and Czech Republic.
10
Discussion
This study attests the existence of inequalities of opportunity in health in Europe. The comparison of the magnitude of inequalities of opportunity in health across European countries provides interesting results. The preliminary results show a general pattern of combined low inequalities of opportunity in health within Europe and weak differences between Roemer’s and Barry’s viewpoints in Sweden and Switzerland and combined high inequalities of opportunity in health and strong differences between Roemer’s and Barry’s viewpoint in Germany, Spain, Italy, Denmark, Greece and Belgium. Some countries are not fitting with the general patterns, however. This the case of Austria, France and Czech republic where high inequalities of opportunity in health are observed but the two alternative viewpoints do not appear to matter much. Inversely, in the Netherlands and Poland, inequalities of opportunities in health are not marked but differences between the two scenarios are considerable. These considerations open the debate on the determinants to be tackled for the reduction of inequality of opportunity in health and more globally health inequalities: health-related behaviours or poor effects of past conditions. As social background, parents' health and parent's health related behaviours both represent factors beyond the realm of individual responsibility (Roemer 1998; Fleurbaey 2008; Fleurbaey & Schokkaert 2009; Trannoy et al. 2010), they are socially or morally unacceptable sources of inequality. Furthermore, the recent report of the World Health Organization’s Commission on the Social Determinants of Health (Marmot et al., 2008) highlights the role of childhood conditions as primary sources of inequality in health. Given the magnitude of the effect of social background, reducing social reproduction across generations would also provide important benefits on health in the course of life. Consequently, improving childhood conditions and equality of opportunity in income acquisition appears to be first-rate candidate for a policy aiming at reducing inequality in health. The distinction between the alternative normative viewpoints about the correlation between efforts and circumstances seems to matter much in some European countries than in others and this will suggest differences in the underlying public health policies that could be put in place to fight inequalities of opportunities in health. Our results suggest a strong social and family determinism of lifestyles in the Netherlands, Poland, Germany, Denmark, Belgium and the Mediterranean which emphasized the importance of inequalities of opportunities in health within those countries. In terms of public health and social policies, it appears that reducing social and unhealthy lifestyles reproduction across generations would provide important benefits on health. On the other hand Austria, France, and Czech Republic show high inequalities of opportunities in health mainly driven by social and family background affecting adult health directly, and so would require policies compensating for poorer initial conditions.
11
Acknowledgements
We gratefully acknowledge the financial support of the Risk Foundation (Health, Risk and Insurance Chair, Allianz). Damien Bricard also benefited from a PhD studentship from the Régime Social des Indépendants (RSI). We thank the participants to the to the 2013 PhD Seminar on Health Economics and Policy (Grindelwald, Switzerland) for their helpful comments.
References
Adams P, Hurd M, Mc Fadden D, Merril A, Ribeiro T. Healthy, Wealthy, and Wise ? Tests for direct causal paths between health and socioeconomic status. Journal of Econometrics 2003; 112 (1); 3-56.
Adda J, Chandola T, Marmot M. Socioeconomic status and health: causality and pathways. Journal of Econometrics 2003; 112; 57-63.
Arneson R. J. Equality and equal opportunity of welfare. Philosophical Studies 1989; 56; 77-93. Börsch-Supan,A., Brugiavini,A., Jürges,H., Mackenbach,J.P., Siegrist,J., & Weber,G. (2005). Health, ageing and retirement in Europe - First results from the Survey of Health, Ageing and Retirement in Europe. Mannheim.
Cohen G.A. On the Currency of Egalitarian Justice. Ethics 1989; 99; 906-944.
Crimmins,E., & Cambois,E. (2003). Social inequalities in health expectancy. In J.-M. Robine, C. Jagger, C. Mathers, E. Crimmins, & R. Suzman (Eds.), Determining Health Expectancies (pp.111-126). Chichester.
Dworkin R. What is equality? Part I: Equality of Welfare. Philosophy and Public Affairs 1981 ; 10 ; 185-246.
Fleurbaey,M. Equality and responsability. European Economic Review 1995, 39 683-689.
Fleurbaey M. Health, Equity, and Social Welfare. Annales d’Économie et de Statistique 2006; 83/84; 21-59.
Fleurbaey M. Fairness, Responsibility, and Welfare. Oxford University Press. Oxford; 2008. Fleurbaey M, Schokkaert E. Unfair inequalities in health and health care. Journal of Health Economics 2009; 28(1); 73-90.
Jagger,C., Robine,J.-M., Van Oyen,H., & Cambois,E. (2007). Healthy ageing - a challenge for Europe.
Jagger,C., Gillies,C., Cambois,E., Van Oyen,H., Nusselder,W., Robine,J.-M., & EHLEIS team (2009). Trends in disability-free life expectancy at age 65 in the European Union 1995-2001: a comparison of 13 EU countries. EHEMU Technical report, 5.1.
12 Jagger,C., Gillies,C., Moscone,F., Cambois,E., Van Oyen,H., Nusselder,W., Robine,J.-M., & EHLEIS team (2008). Inequalities in healthy life years in the 25 countries of the European Union in 2005: a cross-national meta-regression analysis. The Lancet, 372(9656), 2124-2131.
Jusot F., Tubeuf S., Trannoy A. Circumstances and Effort: How important is their correlation for the measurement of inequality of opportunity in health?, Health economics 2013, DOI: 10.1002/hec.2896.
Marmot,M., Friel,S., Bell,R., Houweling,T., taylor,S., & the Commission on Social Determinants of Health (2008). Closing the gap in a generation: health equity through action on the social determinants of health. The Lancet, 372 1661-1669.
Masseria,C., Mossalios,E., & Allin,S. (2006). Measurement of socioeconomic inequality of health in 10 European countries: an exploratory analysis of SHARE using three approaches. European Commission - Research Note LSE.
Roemer J. Equality of opportunity. Harvard University Press; Cambridge; 1998.
Rosa-Dias P. R. Inequality of Opportunity in Health: evidence from the UK cohort study. Health Economics 2009; 18(9); 1057-1074.
Rosa-Dias P. R. Modelling opportunity in health under partial observability of circumstances. Health Economics 2010; 19(3); 252-264.
Rosa-Dias P. R, Jones A. Giving equality of opportunity a fair innings. Health Economics 2007; 16; 109-112.
Sen A.K. Why health equity? Health Economics 2002; 11; 659-666.
Shorrocks A.F. Inequality Decomposition by Factor Components. Econometrica 1982; 50(1); 193-211.
Trannoy A, Tubeuf S, Jusot F, Devaux M. Inequality of Opportunities in Health in France: A First Pass. Health Economics 2010; 19 (8); 921-938.
Tubeuf S., Jusot F., Bricard D., (2012) Mediating role of socioeconomic factors and lifestyles in the relationship between early-life conditions and health: Evidence from the 1958 British cohort. Health Economics, 21 (Issue Supplement S1): 129-150.
13
Tables
Table 1 - Distribution of “good” health status across European countries
Percentage Europe 62.5 Austria (AT) 58.0 Germany (DE) 56.7 Sweden (SE) 70.2 Netherlands (NL) 68.9 Spain (SP) 46.7 Italy (IT) 56.1 Percentage France (FR) 62.1 Denmark (DK) 72.3 Greece (GR) 73.3 Switzerland (SW) 79.7 Belgium (BE) 69.4 Czech Republic (CZ) 56.4 Poland (PL) 34.0
Table 2 - Descriptive statistics at European-level (20,946 observations)
Percentage Sex Men 45.1 Women 54.9 Age 50-54 11.5 55-59 21.1 60-64 21.0 65-69 17.9 70-74 15.0 75-80 13.5
Main breadwinner occupation
Senior managers and professionals 8.1
Technicians, associate professionals and armed forces 6.1
Office clerks, service workers and sales workers 13.5
Skilled agricultural and fishery workers 26.8
Craftsmen and skilled workers 26.2
Elementary occupations and unskilled workers 17.7
No main breadwinner 1.6
Number of books at home :
None or very few (0-10 books) 43.2
Enough of fill one shelf (11-25 books) 22.6
Enough to fill one bookcase (26-100 books) 21.5
Enough to fill two or more bookcases (more than 100 books) 12.7 Number of rooms/household members 0.72 Number of facilities at home:
None 26.7
One 19.7
Two or three 29.0
14 Period of difficulties during childhood
Economic hardships 2.3
Hunger 5.9
Parent's longevity
Mother prematurely deceased 38.6
Mother deceased in later ages 35.2
Mother alive 26.3
Father prematurely deceased 47.6
Father deceased in later ages 42.0
Father alive 10.4
Parent's health-related behaviours
No regular dentist visits for their children 47.9
Parents' smoking 63.6
Parents' alcohol consumption 8.4
Lifestyle/Effort variables
Smoking 21.3
Obesity 18.9
15
Table 3 - Marginal effects from Barry and Roemer specification at European-level
Barry specification Roemer specification Main breadwinner occupation (ref : Elementary occupations and unskilled workers)
Senior managers and professionals 0.0536*** 0.0606***
Technicians, associate professionals and armed forces 0.0192 0.0251
Office clerks, service workers and sales workers 0.0291* 0.0325**
Skilled agricultural and fishery workers 0.00621 0.0128
Craftsmen and skilled workers 0.00973 0.0116
No main breadwinner 0.0283 0.0273
Number of books at home (ref: None or very few (0-10 books))
Enough of fill one shelf (11-25 books) 0.0488*** 0.0559***
Enough to fill one bookcase (26-100 books) 0.0600*** 0.0710***
Enough to fill two or more bookcases (more than 100 books) 0.0499*** 0.0580*** Number of room/household member 0.0257** 0.0369*** Number of facilities (ref: None)
One 0.00476 0.0153
Two or three 0.0251* 0.0321**
Four or five 0.0372** 0.0461***
Period of difficulties during childhood
Economic hardships -0.117*** -0.119***
Hunger -0.0564*** -0.0570***
Mother's longevity (ref: mother prematurely deceased)
Mother deceased in later ages 0.0184* 0.0238**
Mother alive 0.0294*** 0.0355***
Father's longevity (ref: father prematurely deceased)
Father deceased in later ages 0.0346*** 0.0409***
Father alive 0.0384*** 0.0466***
Parents' health-related behaviours
No regular dentist visits for their children -0.0288*** -0.0349***
Parents' smoking -0.0172** -0.0192**
Parents' alcohol consumption -0.0658*** -0.0719***
Effort variables/residuals
Smoking -0.0561*** -0.0561***
Obesity -0.130*** -0.130***
Sedentarity -0.206*** -0.206***
Country (ref: Austria (AT))
Germany (DE) -0.0641** -0.0641** Sweden (SE) 0.0252 0.0252 Netherlands (NL) 0.0383 0.0383 Spain (SP) -0.0756*** -0.0756*** Italy (IT) 0.0131 0.0131 France (FR) -0.00154 -0.00154 Denmark (DK) 0.0535* 0.0535* Greece (GR) 0.154*** 0.154*** Switzerland (SW) 0.129*** 0.129*** Belgium (BE) 0.0759*** 0.0759*** Czech Republic (CZ) -0.0689** -0.0689** Poland (PL) -0.202*** -0.202***
Significance levels of test of rejecting the hypothesis of the nullity of the coefficient from 1,000 bootstrapped replications: *** 1%, **5%, *10%. Models are adjusted for age and sex.
16
Table 4 - Inequality of opportunity measurement according to Barry and Roemer scenario across European countries
Europe AT DE SE NL ES IT FR DK GR SW BE CZ PL Variance 0.234 0.244 0.246 0.209 0.214 0.249 0.246 0.236 0.200 0.196 0.162 0.212 0.246 0.225 Barry scenario 0.0086*** 0.0227*** 0.0129*** 0.0088** 0.0057** 0.0140*** 0.0094*** 0.0143*** 0.0114*** 0.0101*** 0.0043* 0.0068*** 0.0127*** 0.0072** 3.66*** 9.32*** 5.25*** 4.22** 2.66** 5.62*** 3.82*** 6.08*** 5.70*** 5.15*** 2.66* 3.21*** 5.15*** 3.21** 49.17*** 63.73*** 44.40*** 54.94*** 33.90*** 70.04*** 42.22*** 65.60*** 50.73*** 71.54*** 40.91*** 31.11*** 78.25*** 56.58*** Roemer scenario 0.0100*** 0.0249*** 0.0147*** 0.0092** 0.0068*** 0.0154*** 0.0109*** 0.0152*** 0.0129*** 0.0110*** 0.0045* 0.0082*** 0.0134*** 0.0082** 4.27*** 10.21*** 5.99*** 4.38** 3.15*** 6.17*** 4.40*** 6.45*** 6.42*** 5.61*** 2.76* 3.84*** 5.45*** 3.66** 57.42*** 69.80*** 50.69*** 57.03*** 40.09*** 76.82*** 48.65*** 69.58*** 57.19*** 77.85*** 42.48*** 37.18*** 82.92*** 64.52*** Difference between Roemer and Barry
0.0014*** 0.0022* 0.0018** 0.0003 0.0010** 0.0014** 0.0014*** 0.0009* 0.0015** 0.0009*** 0.0002 0.0013*** 0.0008* 0.0010*
0.61*** 0.89* 0.74** 0.16 0.49** 0.54** 0.58*** 0.37* 0.73** 0.45*** 0.10 0.63*** 0.31* 0.45*
8.25*** 6.07** 6.30*** 2.09 6.19*** 6.78*** 6.43*** 3.99** 6.46*** 6.31*** 1.57 6.07*** 4.67*** 7.94***
In % of Roemer 14.37*** 8.70* 12.42*** 3.66 15.44*** 8.83*** 13.22*** 5.73* 11.30*** 8.10*** 3.70 16.32*** 5.63** 12.31***
N 20946 648 1550 1193 1794 1439 2094 1800 1746 2466 1032 2250 1514 1420
- Significance levels of test of rejecting the hypothesis of the nullity of the coefficient from 1,000 bootstrapped replications: *** 1%, **5%, *10%.
- 1 cov( ˆ , ) j j C j H H EOP ; ) ( ) , ˆ cov( ) ( 2 2 1 2 j j j C j j j H H H H EOP EOP ; ) , ˆ cov( ) , ˆ (cov( ) , ˆ cov( ) , ˆ cov( 1 1 3 j j E j j C j j C j j E j j j H H H H H H H H EOP EOP EOP
