Equality of Opportunity in Health and Healthcare

From Inequality to Inequality of Opportunity

The first empirical contributions on inequality in health and healthcare provided insight into the determinants of such inequality, especially the relationship between income and other socioeconomic determinants in the context of health and healthcare use or access. The pioneering work of the ECuity² and Eurothine³ collaboration projects has dominated empirical research on the magnitude of inequality and its determining factors. While the ECuity project initiated the use of the concentration index and its decomposition, which was then followed with numerous examples of empirical work measuring inequality in health, healthcare access, and healthcare delivery in developed and developing countries, the Eurothine project aimed at evaluating the effectiveness of interventions and policies implemented to reduce inequality in health and healthcare in various European countries. Following the lead of economics research that had demonstrated the importance of social determinism, empirical research in health economics later evolved toward the understanding of mechanisms that could explain the construction of social-health inequality over the life span. Scholars have especially turned their interest to the role played by childhood conditions in the health of children and adults. At the same time, the issue of responsibility has become more and more predominant in the literature in social choice and political philosophy, through the concept of luck egalitarianism. From a focus on the objective of equalizing health outcomes across socioeconomic groups by means of positive approaches, or ensuring an equal access to care for people with equal needs, research has moved toward normative discussions on the sources of inequality in health and healthcare with a focus on the importance of fairness. The outcome focused upon when analyzing inequalities thus became “opportunities.”

A Definition of Inequality of Opportunity

Based on the philosophical literature on inequality of opportunity (Arneson, 1989; Cohen, 1989; Roemer, 1998, 2002; Rawls, 1999), there is a global agreement in the economic literature on the necessity of taking into account the determinants of any outcome in order to judge the legitimacy of differences in this outcome between members of a given society. The concept of equality of opportunities initiated by Roemer (1998) relies on the distinction between sources of inequality. The share of inequality related to health determinants that are beyond individual responsibility, known as circumstances, are considered as the most unacceptable and are recognized as inequality of opportunity. On the other hand, the share of inequality related to determinants that are freely chosen by individuals according to their preferences, namely those associated with effort, may be considered as legitimate inequality. It is worthwhile underlining that the partition between effort and circumstances is sometimes labeled as legitimate versus illegitimate sources of inequality, or ethically acceptable versus ethically unacceptable, or even equitable or fair versus inequitable or unfair. Disagreement as to which sources are considered as legitimate/effort and illegitimate/circumstances is common. Most studies on inequality of opportunity in health and healthcare consider that social and family background constitute relevant circumstances, since individuals cannot be held responsible for their birth lottery. In particular, they cannot be held responsible for their parents’ educational level, occupation, or living conditions during their childhood, or parental health-related characteristics, such as genetic inheritance, lifestyles, or healthcare habits. On the contrary, effort is difficult to observe and measure; nevertheless, there is a consensus in the health field that individuals’ lifestyle choices, such as not smoking, having a balanced diet, not drinking too much, or using preventative healthcare, can be considered as examples of individual effort being made for health and healthcare. Lifestyles and healthcare habits represent individual choice and effort invested in health capital. It appears relevant here to underline the importance of referring to an “age of consent” that acts as a threshold below which people cannot be held responsible for their efforts (Arneson, 1989). Individuals can only be held responsible for the lifestyles they have consciously and fully chosen, which is likely to be the case when lifestyles are initiated after a certain age (e.g., 15 years old). Brunori (2017) suggests that a good benchmark for such a discussion is the legal literature on criminal responsibility. In the equality of opportunity literature, obesity and healthcare utilization during childhood are more often viewed as children’s health outcomes or as childhood circumstances than as legitimate sources of inequality within a population of children.

While the key idea of responsibility underlying the measurement of inequality of opportunity is easier to conceptualize when the outcome of interest is health status and lifestyles are taken as proxies for effort, the transposition to the concept of inequality of opportunity in healthcare is not straightforward. Equality-of-opportunity theory distinguishes between illegitimate inequality, which is due to nonresponsibility characteristics, and legitimate inequality, due to responsibility characteristics (Trannoy, 2016). The issue is how to define variables that one can be held responsible for in the context of healthcare access or use. Healthcare use is determined by healthcare needs and availability, which both represent circumstances for which individuals cannot be held responsible. However, healthcare use is also likely to express individuals’ healthcare habits and preferences. In social justice theory, there are two competing views on how preferences should be treated. Some philosophers, following Ronald Dworkin (1981a, 1981b), consider that preferences should be fully respected, so that individuals are considered responsible for them, while others, like Gerard Cohen (1989), see individuals as only responsible for what they can control. This debate is of particular importance for equality of opportunity in healthcare outcomes where the crucial question is to investigate whether the factors explaining differences in healthcare use or access are ethically justified (Fleurbaey & Schokkaert, 2011).

Theoretically, achieving EOP implies respecting two basic principles (Fleurbaey, 2008): the compensation principle, which demands that inequality due to circumstances be eliminated or compensated for, and the liberal reward principle, which necessitates that any efforts made by individuals be rewarded, and respected, when drawing up redistribution policies. Despite this general agreement, the measurement of equality of opportunity still entails addressing many theoretical, methodological, and empirical questions (Brunori et al., 2013; Fleurbaey & Schokkaert, 2011; Ramos & Van de Gaer, 2016; Roemer & Trannoy, 2014).

Two Approaches for Measuring of Inequality of Opportunity

Following the typology proposed by Fleurbaey and Peragine (2013) and used by Ramos and Van de Gaer (2016), Li Donni, Peragine, and Pignataro (2014), and Roemer and Trannoy (2014), two different approaches have been proposed to the formulation of the compensation principle: the ex ante and the ex post approaches. The ex post approach consists of looking at differences in the actual outcome between individuals with the same responsibility (effort) characteristics and finds that there is equality of opportunity if all those who exert the same effort obtain the same outcome. The ex ante approach, in contrast, suggests that there is equality of opportunity if all individuals face the same set of opportunities, regardless of their circumstances.

The ex post approach requires the observation of the efforts that are to be rewarded or the imposition of very restrictive assumptions about the relationship between responsibility characteristics and outcomes. Conversely, the ex ante approach only requires the observation of circumstances, since inequality of opportunity is identified by comparing outcome distributions between types of circumstances. This second approach is less data hungry since it only allows consideration of a limited set of relevant factors, independent of individual responsibility. Fleurbaey and Peragine (2013) have shown the incompatibility of the ex post and the ex ante approaches with the compensation principle. While ex post compensation and liberal reward are often found to be inconsistently related to the timing of the choice to make an effort by the individual, ex ante compensation and liberal reward are consistent. The choice between the ex ante and ex post approaches is driven either by data availability or ethical viewpoint. Effort could be regarded as being unobservable by nature, hence when measurements of effort are used, the choice of those variables must be well justified and it will be necessary to demonstrate that individuals are fully responsible for that effort.

The Correlation Between Effort and Circumstances

An issue in this context was highlighted in Jusot, Tubeuf, and Trannoy (2013); this brought to the forefront the open debate in the literature on the relationship between effort and circumstances, which cannot be assumed to be independent of each other. The key matter in this context is the precise definition of effort, which should be rewarded, and the definition of circumstances, which should be compensated for. The original debate on the correlation between effort and circumstances was between John Roemer and Brian Barry in the field of education (Barry, 2005; Roemer, 1998, p. 22). The context was a debate about the case of Asian students. In Roemer’s words, such students “generally work hard in school and thereby do well because parents press them to do so. The familial pressure is clearly an aspect of their environment outside their control.” Roemer argued that an ideal equal-opportunity policy would respect individual effort through an approach whereby “we could somehow disembody individuals from their circumstances” (Roemer, 1998, p. 15). As a consequence, the extra effort of Asian students should not be rewarded because it is determined by a circumstance outside their control. Conversely, Barry argued that nevertheless, “the fact that their generally high levels of effort were due to familial pressure does not make their having expended high levels of effort less admirable and less deserving than it would have been absent such pressure.” From this point of view, which is the mainstream view in the literature on incentives (Yellen, 1984), the extra effort of Asian students should be fully rewarded and the lack of familial pressure on other types of students should not be compensated for. When transposing this debate to the field of health, lifestyles—such as having a balanced diet, doing exercise, not smoking, or not drinking too much—are often considered as relevant efforts, as they constitute causal determinants of health status. The idea that lifestyles are freely chosen individual behaviors is debatable, in particular because they are likely to be influenced by the family and social environment during childhood as well as by genetic characteristics and preferences. The Barry/Roemer debate in health would then ask about the legitimacy of, for instance, holding sons of smokers less responsible for their own smoking than the sons of nonsmokers. While, for Barry, this distinction is irrelevant, Roemer considers that the share of smoking, which can be attributed to family and social background, is a circumstance and not an effort. This distinction is typically meaningful for the ex post approach; however, an ex ante approach, whereby only circumstances are observed, must adopt a Roemerian viewpoint.

The Equality of Opportunity Principles Across Generations

Another challenge in defining effort is identifying “whose” effort it is; as Roemer and Barry respectively observe, “Asian children . . . do well because parents press them to do so” and “their generally high levels of effort [are] due to familial pressure.” The transmission of values through parental effort results in what manifests itself in the effort exerted by the next generation, and if one views familial pressure on children to be educated as a parental effort, the definition of circumstances to be compensated for is less obvious. Jusot et al. (2013) underline the impossibility of respecting the principles of compensation and natural reward for all generations. If one gives precedence to the younger generation in the application of the two equality-of-opportunity principles of compensation and liberal reward, then one should consider that the whole initial background consists of circumstances, including parental effort, independently of the link with children’s effort. On the other hand, if one gives precedence to the past generation in the application of the two equality-of-opportunity principles, then one should consider that parental effort must be respected whatever its consequences to the next generation. This latter position corresponds to that of Swift (2005), who argued, “To the extent that the reproduction of inequality across generations occurs through the transmission of cultural traits, it does so substantially (though not exclusively) through intimate familial interactions that we have reason to value and protect. Preventing those interactions would violate the autonomy of the family in a way that stopping parents doing spending their money on, or bequeathing money to their kids would not” (Sørensen, 2006; Swift, 2005). From Swift’s point of view, family is an association, and following on Rawls’ justice theory, the “basic liberties,” including among them freedom of association, take lexical priority over fair equality of opportunity and the principle of difference (Rawls, 1999). Transposed to the fields of health and healthcare, it is easy to see that the way parents use healthcare, especially in terms of health checks and preventative care for children, makes children more likely to use it in a similar way in adulthood; this might also be the case with diet, physical activity, or smoking and drinking behaviors.

Empirical Approaches

The empirical literature in equality of opportunity in health and healthcare calls upon two types of methodologies: nonparametric and parametric, which are sometimes coupled in studies. The nonparametric approach follows from the methodology proposed by Lefranc, Pistolesi, and Trannoy (2009) in order to identify inequality of opportunity in income. It relies upon the use of dominance criteria and bilateral tests and compares cumulative distribution functions of outcomes conditional to categories of circumstances (types) of individuals and groups with the same effort (tranches). On the other hand, the parametric approach relies on econometric modeling and is used to identify inequality of opportunity by investigating the association between circumstances and outcome. Parametric studies adopting an ex ante approach simply estimate a reduced-form model in order to identify differences in health opportunities related to circumstances, independently from the influence of any unobserved efforts, as suggested by Ferreira and Gignoux (2011) and Trannoy et al. (2010). On the other hand, methodologies used in the ex post approach vary according to the normative position adopted regarding the correlation between efforts and circumstances or legitimate and illegitimate sources of inequality. Health outcomes can simply be regressed on observed circumstances and observed efforts, or auxiliary equations focusing on the efforts on their own or on more complex structural modeling can also be considered. In the ex ante approach, in which effort is by definition not observed, the literature avoids the problem of lacking effort indicators by combining an ex post approach to inequality of opportunity with a concept of relative effort, whereby two people belonging to different types are deemed to have exerted the same effort if and only if they are in the same percentiles of their respective (and different) conditional distributions.

If empirical studies in equality of opportunity in health initially tested the existence of inequality of opportunities in health, recent studies go beyond assessment and propose a quantification of the inequality of opportunities in health. Two broad types of measures are used, direct and indirect ones, as emphasized in the literature (Brunori et al., 2013; Fleurbaey & Schokkaert, 2011; Ramos & Van de Gaer, 2016). On one hand, direct measures assess how great the inequality is when only the share of inequality that is attributable to circumstances remains. Empirically, it consists of estimating the inequality using a counterfactual outcome distribution in which all inequalities due to differences in effort have been eliminated. The direct unfairness proposed by Fleurbaey and Schokkaert (2009) is a typical direct measure of inequality of opportunities, which is consistent with the ex ante approach to compensation. Direct unfairness evaluates the level of inequality that would exist if all individuals chose to exert the same reference level of effort. On the other hand, indirect measures assess how much inequality remains after opportunities have been equalized. Empirically, this indirect approach often consists of estimating the level of inequality of opportunity by comparing inequality in the actual outcome distribution to inequality in a counterfactual outcome distribution whereby all individuals would face the same circumstances. The use of this measurement method is, however, debatable (Schokkaert, 2018), since different distributions could lead to the same measure of inequality of opportunities. The fairness gap (Fleurbaey & Schokkaert, 2009) is a preferable indirect measure of inequality of opportunities and quantifies the inequality in the distribution as the distance between the observed outcome and the outcome that would exist if all individuals had the same reference set of circumstances. The fairness gap corresponds to the ex post compensation principle.

Ex Ante Inequality of Opportunities in Health and Healthcare

The 15 studies that measured inequality of opportunities in health and healthcare using an ex ante perspective are shown in Table 1.

More than half of the studies focused on European countries: two from the United Kingdom (Jones, Rice, & Rosa Dias, 2012; Li Donni et al., 2014), two from France (Bricard, 2013; Trannoy et al., 2010), one each from Italy (Gigliarano & D’Ambrosio, 2013), Spain (Pinilla, Lopez-Valcarcel, & Urbanos-Garrido, 2017), and Norway (Ovrum & Rickertsen, 2015), while one study used data from 14 European countries that participated in the Survey on Health, Ageing and Retirement in Europe (SHARE) and the English Longitudinal Study of Ageing (ELSA; Pasqualini, Lanari, Minelli, Pieroni, & Salmasi, 2017). Outside Europe, one study focused on data from the United States (Chen, 2015), and one on Israel (Lazar, 2013). Finally, only five studies focused on low- or middle-income countries: four on South America, two on Columbia (Fajardo-Gonzalez, 2016; Rivera, 2017), one on Brazil (Barbosa, 2016), one on Chile (Gallardo, Varas, & Gallardo, 2017), and one on Indonesia (Jusot, Mage, & Menendez, 2014).

Most studies considered health status as the outcome of interest (87%) while only two papers used healthcare outcomes, such as healthcare habits, physician visits, and preventative care, for this. Studies of inequality of opportunity in health mainly focused on self-assessed health (SAH; 73%) while four studies considered various scores as health outcomes. The Colombian study considered the EQ-5D health-related quality-of-life score (Rivera, 2017); the U.S. study used both the physical and mental summary scales of the SF-12 health questionnaire (Chen, 2015); one of the British studies used mental and chronic diseases and disability (Jones et al., 2012); and the Indonesian study used a synthetic health score built on the basis of several self-reported and objective health measures (Jusot et al., 2014). Regarding the type of circumstances considered, most studies considered various characteristics as a proxy for parental social background; these included parental education (53%), parental social class (33%), family financial situation or income (33%), and, more rarely, the number of books in the household at age 10 (Pasqualini et al., 2017), living conditions (Chen, 2015), or assets ownership (Fajardo-Gonzalez, 2016). Some other sociodemographic characteristics were also considered, such as ethnicity (27%), religion and language spoken (Jusot et al., 2014), or family structure (Rivera, 2017). While individual education level is sometimes viewed as an effort variable in the ex post approaches, Barbosa (2016) considered education as a source of illegitimate inequality in healthcare insofar as a child has limited responsibility for parental decisions about which school to attend; and Jones et al. (2012) used cognitive and noncognitive abilities and happiness, measured while at school, as childhood circumstances.

Geographical characteristics were also used as circumstances; this included the country or region of residence at birth or during childhood (33%), the country or region of residence as an adult (27%), the rural/urban status of the location (20%), parents’ country of birth in one study (7%), or socioeconomic characteristics of the country of residence (7%).

Beyond socioeconomic background, numerous studies considered health-related circumstances, such as parents’ health or vital status (40%), and one used both the parents’ health habits and childhood physician density (Bricard, 2013). Finally, one study considered childhood health status as a circumstance (Jones et al., 2012), and the U.S. analysis treated accidents, health shocks, and disability related to health limitations (Chen, 2015) as circumstances determining the level of SAH.

Empirically, most of the ex ante analyses directly used circumstances in their models while two studies, one in France (Trannoy et al., 2010) and the other in Spain (Pinilla et al., 2017), proposed a pathway analysis to distinguish the direct impact of circumstances on health outcomes from the indirect impact they made through their role in the determination of individual socioeconomic status.

Four studies focused on the impact of circumstances on lifestyles; two looked into the impact of circumstances on healthcare use (one in Brazil [Barbosa, 2016] and one in France [Bricard, 2013]), and two focused on the impact those circumstances had on various unhealthy behaviors, such as smoking, alcohol consumption, diet, obesity, physical activity, and teenage pregnancy (one in the United Kingdom [Jones et al., 2012] and one in Norway [Ovrum & Rickertsen, 2015]). These studies did not measure the consequences of those lifestyles on health, nor discuss their potential impact in terms of the natural-reward principle. This was the reason why those studies could be more easily attached to the ex ante approach since they considered, implicitly or explicitly, that intergenerational transmission of lifestyles is one of the pathways explaining inequality of opportunity in health.

Regarding the methodology used, 11 studies were based on parametric modeling (73%) and one used the Fleurbaey and Schokkaert framework (Barbosa, 2016), whereas ten investigations relied on a non-parametric approach (67%): six used first- or second-order stochastic dominance tests (40%) in order to compare health-outcome distribution by group of circumstances, and three proposed an analysis of equality of opportunity by types (20%) or rank analysis (7%).

When measuring inequalities, several indices and methods are used to quantify the inequality of opportunity: these include the direct unfairness and fairness gap (Barbosa, 2016); the Shapley measure (Fajardo-Gonzalez, 2016); variance decomposition (Jusot et al., 2014) or R-squared decomposition (Pasqualini et al., 2017); indices decompositions, including the Gini index (Fajardo-Gonzalez, 2016; Ovrum & Rickertsen, 2015; Rivera, 2017; Trannoy et al., 2010); the Erreygers index (Trannoy et al., 2010), or the Atkinson index (Li Donni et al., 2014); and the concentration index (Barbosa, 2016; Ovrum & Rickertsen, 2015).

Ex Post Inequality of Opportunities in Health and Healthcare

The 18 studies that measured inequality of opportunity in health using an ex post perspective are described in Table 2; none of them considered a healthcare outcome.

Around half of the studies focused on the United Kingdom or England, using either the data from the 1958 National Child Development Study (Jones, Roemer, & Rosa Dias, 2014; Rosa Dias, 2009, 2010), the British Household Panel Survey (Li Donni, Peragine, & Pignataro, 2009; Li Donni et al., 2014), the Health and Life Survey (Balia & Jones, 2011), or the Health Survey of England (Carrieri & Jones, 2018). Five other studies focused on single European countries (France [Jusot et al., 2013], Luxembourg [Deutsch, Alperin, & Silber, 2017], the Netherlands [Garcia-Gomez, Schokkaert, Ourti, & Bago D’Uva, 2015]) or cross-country comparisons (13 different countries [Bricard, Jusot, Trannoy, & Tubeuf, 2013], or 14 countries [Pasqualini et al., 2017]). Outside Europe, studies with an ex post approach included two in the United States (Asada, Hurley, Norheim, & Johri, 2015; Chen, 2015), one in Canada (Asada, Hurley, Norheim, & Johri, 2014), one in Israel (Lazar, 2013), one in Chile (Carranza & Hojman, 2015), and one in China (Sun, Ma, Song, & Gu, 2013).

A large majority of studies focused on inequality of opportunity in health, using SAH as the main health outcomes of interest (67%). Four studies considered various scores as health outcomes: The two North American studies considered the Health Utility Index (HUI), which is a quality-of-life score (Asada et al., 2014, 2015); two studies used mental health scores (Chen, 2012; Rosa Dias, 2010); and one used a physical score from the SF12 questionnaire (Chen, 2015). Other health-related outcomes were also considered; these included health events (Garcia-Gomez et al., 2015), chronic diseases or disability in two studies (Pasqualini et al., 2017; Rosa Dias, 2010), various biomarkers (Carrieri & Jones, 2018), and mortality in two studies (Balia & Jones, 2011; Garcia-Gomez et al., 2015). One study used BMI (Pasqualini et al., 2017), although such health-related lifestyle factors are often considered as effort variables to reward in other studies. Another study considered healthcare used as a proxy of health status (Sun et al., 2013). There were no studies where inequality of opportunity in healthcare use per se was measured; this is probably related to the fact that ex post studies require an actual measurement of effort, which is conceptually always difficult to obtain, especially in the case of healthcare.

Regarding the types of circumstances or illegitimate factors considered, most of the studies considered various characteristics as a proxy for parental social background, including: parental education (39%); parental social class or employment (44%); family financial situation and adverse life events (28%); grandfather’s education (11%); in a few studies, the number of books in the household at age 10 (11%); parental literacy (6%); or living conditions (6%). Some other sociodemographic characteristics considered included ethnicity (11%) and family structure (11%).

Finally, whereas one’s own education level could be viewed as an effort variable that is under one’s responsibility, several studies considered some circumstances related to educational achievement, such as: achieved education (28%); cognitive abilities; happiness at school, as well as parental support to stay at school (6%); math test score at 11 years old (6%); or social development at 11 years old (6%). Two other studies considered marital status or socioeconomic status as illegitimate sources of inequality in health as well (17%).

Geographical dimensions were also often taken into account as circumstances: these included the country of birth and year of immigration (11%); parents’ country of birth (6%); socioeconomic characteristics of the country or region of residence (11%); and, in one study, health-related regional averages (6%).

Beyond socioeconomic background, numerous studies considered health-related circumstances, such as: parents’ health or vital status or longevity (22%); parental smoking (22%), or mother’s smoking during pregnancy and breastfeeding (6%); diabetes, epilepsy, or heart disease in the family (11%); accident (6%); parents’ alcohol consumption (11%); and dental visits for children (6%). Finally, some studies considered respondents’ own health status in childhood or birthweight as a circumstance (17%), as well as current disabilities related to health limitations (11%).

Regarding effort variables, most studies (94%) considered that lack of smoking represented an effort to be rewarded. The next most used variables for effort were body mass index (39%), physical activity (39%), diet (17%), and alcohol consumption (17%). Few studies also considered socioeconomic factors (income, education, employment, home ownership, religion, area of residence, or marital status) as effort variables (28%), especially in order to test different ethical positions on the difference between effort and circumstances. Finally only two studies considered the taking of medication (Carrieri & Jones, 2018) or treatment preferences (Sun et al., 2013) as measures of individual effort.

Regarding the methodology, most studies (94%) used parametric modeling, except one, which used a semiparametric approach (Balia & Jones, 2011). Four of those studies specifically referred to the Fleurbaey and Schokkaert structural framework (28%). Only two of them relied on nonparametric approaches, including stochastic dominance tests (Chen, 2015; Rosa Dias, 2009), and equality of opportunity analysis by types (Carrieri & Jones, 2018; Chen, 2015; Jones et al., 2014) or by tranches (Lazar, 2013; Li Donni et al., 2009; Li Donni et al., 2014). While most of the studies implicitly adopted a Barry viewpoint (Barry, 2005) rewarding the full effort independently from the correlation with circumstances, four studies used a Roemerian approach to assess the full impact of circumstances on health outcomes, including their indirect effect on effort variables (Bricard et al., 2013; Carranza & Hojman, 2015; Deutsch et al., 2017; Jusot et al., 2013).

Several indices and methods were used for the quantification of inequalities of opportunity: direct unfairness and fairness gap (17%); variance decomposition (22%); Gini index (28%); Atkinson decomposition (17%); and counterfactual decomposition (11%). Rarer were decompositions of a rho-squared (Pasqualini et al., 2017), Theil (Carranza & Hojman, 2015), Sen welfare (Balia & Jones, 2011), or dissimilarity index (Jones et al., 2014), as well as a generalized Lorenz curve (Balia & Jones, 2011) and a Shapley decomposition (Deutsch et al., 2017).

Inequality of Opportunity in Child Health and Healthcare

The 15 studies that measured inequality of opportunity in child health and healthcare are described in Table 3.

All studies focused on one or several countries that are considered low- or middle-income countries according to the World Bank. More than half of the research used data from the Demographic and Health Surveys (DHS) program, which provides nationally representative household data for a wide range of monitoring and impact evaluation indicators in the areas of population, health, and nutrition. The three papers that undertook a comparison of inequality of opportunities in child health used DHS data (Andersen, Griffin, Krause, & Montekio, 2017; Assaad, Krafft, Hassine, & Salehi-Isfahani, 2012; Hoyos & Narayan, 2011). Most of the research focused on health and healthcare outcomes before the age of 5, except for four empirical studies, which used longitudinal survey data. These examined the following age groups: 18 in China (Eriksson, Pan, & Qin, 2014), under 8 in Ethiopia (Hussien & Ayele, 2016), between 2 and 6 in Mexico (Van de Gaer, Vandenbossche, & Figueroa, 2013), and between 10 and 14 in South Africa (Zoch, 2015). Six studies in children considered health outcomes, four studies focused on healthcare-specific outcomes, and five considered both of them. Health outcomes corresponded to World Health Organization (WHO) indicators on child growth and malnutrition, such as height, weight, and a combination of both or either with age via z-scores. Two studies considered slightly different health outcomes: One, in Morocco (El-Kogali et al., 2016), examined a set of early child-development outcomes, including cognitive, emotional, and social development. Another, in Mexico (Van de Gaer, Vandenbossche et al., 2013), looked into the incidence of anemia, stunting, standardized body mass index, and the number of sick days taken from school. When the outcomes were related to access to services, these included basic healthcare, pre- and postnatal care, basic nutrition, immunization, and vaccination, as well as clean water and adequate sanitation.

The vector of circumstances variables varied between studies, probably because of data availability. The family social background was used as a circumstance in each. All of the studies except one (Hoyos & Narayan, 2011) took into account parental education level (the mother only or both parents); household wealth or a proxy of wealth, such as household amenities, was used in 87% of the studies, three of them additionally considered parental occupation (the father only or both parents), and one used the mother’s social class. Another very frequent circumstance was geographical characteristics (80%), such as the region of residence, urban/rural status, and the distance to the nearest health facility. Two studies also considered variables at regional level: smoking prevalence at community level (Eriksson et al., 2014) and regional averages of sociodemographic variables (Singh, 2011), while one study treated access to public services, such as sanitation and clean water, as a circumstance (Hussien & Ayele, 2016). Other variables considered as part of the vector of circumstances included further parental characteristics (e.g., age, gender of the head of household, height, height-for-age, weight-for-age, presence in the household, religion, indigenous background, and caste); the child’s characteristics (e.g., age, sex, birth order, number of siblings, and child of a multiple birth); and characteristics of the household (e.g., size, number of children, and number of adults).

In such a case, where the set of circumstances is potentially very broad, measuring inequality of opportunities nonparametrically is difficult as the types of circumstances become too broad to allow for enough observation in each type. In this case, inequality of opportunity is measured parametrically. Twelve studies (80%) used a parametric approach to measure inequality of opportunity; two studies used both a nonparametric and a parametric approach, while one study only used a nonparametric approach. Most parametric studies (75%) assessed the existence of inequality of opportunity in health or in the healthcare of the child population using the Human Opportunity Index (HOI). The HOI was developed by the World Bank and is an index measuring children’s access to basic services and goods that can be considered prerequisites for childhood development (Paes de Barros et al., 2009). The key services that are included in the HOI vary but generally include access to clean water, immunization, sanitation, basic healthcare, and basic nutrition. The use of the HOI leads to the measuring of an equality of opportunity-sensitive coverage rate, which can rely on various circumstances, outcomes, opportunities, and population groups. Most of the studies using the HOI approach additionally used a dissimilarity index, derived by comparing group means for different combinations of circumstances to the population average to quantify how outcomes differ by circumstances. The dissimilarity index was then completed with a Shapley decomposition to estimate the marginal contribution of each considered circumstance to the inequality of opportunity. One of the parametric studies (Assaad et al., 2012) decomposed the inequality using the Theil-T index while another one undertook an Oaxaca-Blinder decomposition of the inequality between urban and rural areas. The three studies that utilized a nonparametric approach to assess inequality of opportunity used an approach by types (Hussien & Ayele, 2016, Sanoussi, 2018); however only one of the studies (Van de Gaer, Vandenbossche, et al., 2013) compared types with first- or second-order stochastic dominance.