Essays on Inequality of Opportunity
in
Health and Human Development
João Pedro Cordas da Rosa Dias
Ph.D. Thesis
The University of York
Department of Economics and Related Studies
May 2010
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Guidelines and tips
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Community
Ask the community
Ask the community for help and clear up your study doubts
University Rankings
Discover the best universities in your country according to Docsity users
Free resources
Our save-the-student-ebooks!
Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors
Health Inequality: Analyzing Roemer Model & School Quality, Summaries of Literature
The relationship between equality of opportunity and health, using the Roemer model as a framework. The study focuses on the structural model, data, methods, and results, and also includes discussions and conclusions. Chapter 4 delves into the quality of schooling and its impact on inequality of opportunity in health. various tables and figures to support the analysis.
Typology: Summaries
2021/2022
1 / 143
This page cannot be seen from the preview
Don't miss anything!
Related documents
Partial preview of the text
Download Health Inequality: Analyzing Roemer Model & School Quality and more Summaries Literature in PDF only on Docsity!
Essays on Inequality of Opportunity
in
Health and Human Development
João Pedro Cordas da Rosa Dias
Ph.D. Thesis
The University of York
Department of Economics and Related Studies
May 2010
Abstract
This thesis comprises four essays on inequality of opportunity in health and human development.
Chapter 2 proposes an empirical implementation of the concept of inequality of opportunity in health and applies it to data from the UK National Child Development Study. Drawing on the distinction between circumstance and effort variables in John Roemer's work on equality of opportunity, circumstances are proxied by parental socio-economic status and childhood health; effort is proxied by health-related lifestyles and educational attainment. Stochastic dominance tests are used to detect inequality of opportunity in the conditional distributions of self- assessed health in adulthood. Alternative measures of inequality of opportunity are proposed. Parametric models are estimated to quantify the triangular relationship between circumstances, effort and health. The results indicate considerable and persistent inequality of opportunity in health. Circumstances affect health in adulthood both directly and through effort factors such as educational attainment, suggesting complementary educational policies may be important for reducing health inequalities.
Chapter 3 specifies a behavioural model of inequality of opportunity in health that integrates John Roemer’s framework of inequality of opportunity with the Grossman model of health capital and demand for health. The model generates a recursive system of equations for health and lifestyles, which is jointly estimated by full information maximum likelihood with freely correlated error terms. The analysis innovates by accounting for unobserved heterogeneity, thereby addressing the partial-circumstance problem, and by extending the analysis to health outcomes other than self-assessed health, namely long standing illness, disability and mental health.
Chapter 4 explores the existence of long-term health returns to different qualities of education, and examines the role of quality of schooling as a source of inequality of opportunity in health. It provides corroborative evidence of a statistically significant and economically sizable association between quality of education and a number of health and health-related outcomes that remains valid beyond the effects of measured ability, social development and academic qualifications. The results also establish quality of schooling as a leading source of inequality of opportunity in health.
Chapter 5 exploits a natural experiment provided by the fact that cohort-members attended different types of secondary school, as their schooling lay within the transition period of the comprehensive education reform in England and Wales that commenced in the 1960’s. This experiment is used to explore the impact of educational attainment and of school quality on health and health-related behaviour later in life. A combination of matching methods, parametric regressions, and instrumental variable approaches are used to deal with selection effects and to evaluate differences in adult health outcomes and health-related behaviour for cohort members exposed to the old (selective) and to the new (comprehensive) educational systems.
- Abstract
- Acknowledgements
- Declaration
- Chapter
- Introduction
- Chapter
- Inequality of Opportunity in Health: Evidence from a UK Cohort Study
- 2.1 Introduction
- 2.2 Background
- 2.2.1 Equality of Opportunity: the Roemer model
- 2.2.2 Definitions and testable conditions
- 2.2.3 Measures of inequality of opportunity
- 2.2.3.1 The Gini-opportunity index
- 2.2.3.2 An alternative approach
- 2.3. Data
- 2.3.1 The National Child Development Study (NCDS)
- 2.3.2 Variables: health, circumstances and effort
- 2.4 Testing and measuring inequality of opportunity in health
- 2.5 Estimation results
- 2.5.1 Adult health and early life circumstances: direct and indirect effects
- 2.5.2 Circumstances and effort: primary pathways
- 2.6 Conclusions
- Appendix A
- Chapter
- 3.1 Introduction
- 3.2 Equality of opportunity: the Roemer model in the context of health
- 3.3 Outline of the structural model
- 3.4 Data
- 3.5 Methods
- 3.6 Results
- 3.7 Discussion and conclusions
- Appendix B
- Chapter
- Quality of Schooling and Inequality of Opportunity in Health
- 4.1 Introduction
- 4.2 Quality of schooling
- 4.2.1 Primary education
- 4.2.2 Secondary education: the comprehensive reform and equality of opportunity
- 4.3 Data
- 4.3.1 Childhood health, parental background and neighbourhood characteristics
- 4.3.2 Cognitive ability, social development and educational achievement
- 4.3.3 Health-related behaviours, attitudes and outcomes
- 4.3.4 Sample selection and non-response
- 4.4 Methods
- Inequality of opportunity in health
- 4.4.2 Regression analysis
- 4.5 Results
- 4.5.1 Quality of schooling and inequality of opportunity in health
- 4.5.2 Quality of schooling, health and lifestyle: primary schools
- 4.5.3 Quality of schooling, health and lifestyle: secondary schools
- 4.6 Conclusions
- Appendix C
- Chapter
- on Adult Health and Lifestyle The Impact of Childhood Cognitive Skills, Social Adjustment and Schooling
- 5.1 Introduction
- 5.2 Comprehensive schooling reforms and the 1958 cohort
- 5.3 NCDS data and study design
- 5.3.1 Childhood health and parental background
- 5.3.2 Cognitive ability, non-cognitive skills and social adjustment
- 5.3.3 Local area characteristics
- 5.3.4 Educational attainment and quality of schooling
- 5.3.5 Intermediate outcomes: health-related behaviours
- 5.3.6 Main outcomes: adult health
- 5.4 Sample selection and balanced samples
- 5.4.1 Sample selection and non-response
- 5.4.2 Balance of covariates between selective and non-selective schools
- 5.4.3 ‘Coaching effects’: absolute and relative cognitive ability
- 5.4.4 Matched sub-samples
- 5.5 Econometric models and results
- 5.5.1 Pre-schooling characteristics
- observables 5.5.2 The impact of attainment and quality of schooling with controls for
- 5.5.3 Instrumental variables estimates
- 5.5.4 Heterogeneous effects by type of school
- 5.5.1 Pre-schooling characteristics
- 6 Discussion
- Appendix D
- Chapter
- Conclusions
- Bibliography
- Chapter List of tables
- Table 1: Summary statistics
- Table 2: Tests of stochastic dominance between types
- Table 3: Measures of inequalty of opportunity
- Table 4: Adult health and circumstances
- Table 5: Adult health, circumstances and effort
- Table 6: The impact of circumstances on effort
- Chapter
- Table 1: System erros correlation matrix
- Table 2: System estimates
- Chapter
- Table 1: NCDS cohort-members by type of primary school
- Table 2: Secondary school characteristics
- Table 3: Estimation samplevs full sample - behaviours Table 4: Quality of primary schooling, health and health-related - behaviours Table 5: Quality of secondary schooling, health and health-related
- in health Table 6: Stochastic dominance tests for inequality of opportunity
- Chapter
- (as attended by NCDS cohort at age 16) Table 1: Characteristics of different types of schools
- Table 2: Educational attainment: highest secondary qualification
- specific main conditions (ICD-9) Table 3: Breakdown of long-standing illness (LSI) by percentage with
- Table 4: Sample means for outcomes
List of figures
Chapter 2
Figure 1: SAH (age 46) by parental socioeconomic group 39
Chapter 4
Figure 1: NCDS cohort-members by type of school (age 16) 84 Figure 2: Distribution of pupil-teacher ratios by type of primary school 85 Figure 3: Distribution of pupil-teacher ratios by type of secondary school 86 Figure 4: Distributions of cognitive and non-cognitive in the NCDS cohort 87 Figure 5: Stochastic dominance: empirical distributions of SAH (age 46) by type of secondary school 88 Figure 6: Stochastic dominance: empirical distributions of mental illness (age 46) by type of secondary school 89
Chapter 5
Figure 1: Schematic view of study design and NCDS variables 97 Figure 2: Empirical distributions of cognitive ability scores by type of school 100 Figure 3: Empirical density of Bristol Social Adjustment Guide (BSAG) 101 Figure 4: Empirical QQ-plots for cognitive score at 7 and BSAG score: before and after matching 109 Figure 5: Distribution of propensity score over selective (“untreated”) and non-selective (“treated”) schools 111 Figure 6: Empirical distributions of relative ability by type of school 112
Acknowledgements
I would like to extend my gratitude to my thesis advisors Andrew Jones, Nigel Rice and Peter Smith for providing outstanding guidance throughout my PhD. I would also like to thank the Centre for Health Economics, especially the members of the Health Econometrics and Data Group for invaluable encouragement and camaraderie. Funding by Fundação para a Ciência e a Tecnologia is gratefully acknowledged.
Very special thanks go to Dimitra for her loving support and understanding.
I dedicate this thesis to my parents.
Chapter 1
Introduction
This thesis consists of a collection of four essays on inequality of opportunity. It is motivated by recent advances in the theory of distributive justice and contributes towards an integrated normative analysis of inequalities in health, education and other aspects of human development.
As asserted by Roemer (2005), equality of opportunity is to be contrasted with equality of outcomes. The Achilles' heel of the advocacy of equality of outcomes has traditionally been its failure to hold individuals accountable for their choices. In light of this, the greatest recent progress in the egalitarian theory of justice, as Cohen (1989) puts it, is arguably the co-option of the sharpest idea in the anti– egalitarian arsenal: the notion of responsibility. By compensating for the impact of circumstances beyond individual control, yet holding individuals responsible for the consequences of their choices, equality of opportunity is an appealing compromise between strict equality of outcomes and mere equity of formal rights. It has thus attracted growing attention in the economics literature and is being increasingly advocated by policy makers, as is made clear in The World Bank Development Report 2006, Equity and Development , which focuses on the inequality issue (World Bank, 2005).
This conceptual progress is the culmination of a series of developments in political philosophy. Rawls’ (1971) pioneering work is credited with reinventing egalitarian justice. Together with Amartya Sen’s concept of equality of capabilities , Rawls’ equality of social primary goods replaces subjective utility with an objective criterion. Once these goods and capabilities are equally distributed, any residual inequality is deemed a legitimate consequence of individual choice, hence of individual responsibility. As Barry (1991) makes clear, between the polar extremes of the choicist position, which attributes every individual outcome to free and unconstrained choice, and the anti-choicist argument, which views outcomes as the reflection of differences in the circumstances that determine choices, there are infinite
intermediate positions. Dworkin (1981; 2000) proposed a solution to this dilemma by treating responsibility as the corner-stone of distributive justice. Like Rawls and Sen before him, Dworkin rejects equality of welfare as a valid criterion since people differ through dissimilar circumstances and handicaps, which determine, at least in part, choices and outcomes. The problem thus becomes one of finding the distribution of resources that appropriately compensates individuals for these circumstances and handicaps. This approach leads to Dworkin’s widely debated concept of equality of resources , which has attracted important criticisms, such as those raised by Arneson (1989) and Cohen (1989) who address the intractable separation between preferences and resources. This debate prompted key progresses in social choice theory, rendering these new ideas operational within the analytical framework known as the equal-opportunity approach.
Equality of opportunity has been given different formal expressions in the social choice literature, such as those of Fleurbaey (1994) and Bossert (1995). These contributions proved too abstract for empirical application, however, hence the vast majority of the applied work on inequality of opportunity is based in the model proposed by Roemer (1996; 1998; 2002). The four essays in this thesis are empirical implementations of this version of the concept of equality of opportunity in the field of health economics.
Arguably, inequality of opportunity is already the implicit equity concept in some earlier contributions in health economics, such as Williams’ fair innings argument (Williams, 1997) and the Rawlsian approach to the measurement of health inequalities proposed in Bommier and Stecklov (2002). However, this normative crucial shift in emphasis, from outcomes to opportunities, is still very scarcely reflected in the latest empirical work on health inequalities. This thesis contributes towards narrowing this gap in the health economics literature.
The relevance of the analysis of inequality of opportunity in health extends well beyond its normative appeal. At the heart of the inequality of opportunity concept lies the interaction between circumstances beyond individual control and effort variables, for which individuals are at least partly responsible. In a health context, early childhood circumstances, parental background, cognitive and non-cognitive
The results indicate the existence of considerable and persistent inequality of opportunity in health among NCDS cohort-members. Part of the effect of childhood circumstances is a direct one and thus only amenable to policy during the early years of life. However, a significant part of this effect is channelled through behavioural choices regarding education and lifestyle. This suggests an important role for complementary policies to reduce health inequalities outside the health care system, in particular, in the education sector.
Chapter 3 specifies and estimates a behavioural model of inequality of opportunity in health in which the exertion of effort is the consequence of utility maximising behaviour subject to constraints. The motivation for this is twofold. First, it narrows the gap between the normative literature on health inequalities and the positive economics research on health capital and demand for health. Second, it proposes an empirical solution to a widely debated structural problem of the equality of opportunity framework: in practice, the full set of circumstances affecting health outcomes is typically only partially observable. This analysis contributes to the existing literature by:
- Integrating John Roemer’s framework of inequality of opportunity with the Grossman model of health capital and demand for health, thereby narrowing the gap between the positive and normative dimensions of the relationship between circumstances, effort and health.
- Accounting for the presence of unobserved heterogeneity that simultaneously affects health and each of the effort factors, and hence addressing the problem of partial observability of the set of circumstances.
- Extending the empirical analysis of inequality of opportunity to health outcomes other than self-assessed health, such as the incidence of long standing illness, disability and mental disorder. The results indicate the presence of unobserved factors that impact simultaneously on health outcomes and effort variables, corroborating the empirical relevance of the theoretical problem of partial observability of circumstances. They also show that different health outcomes in adulthood are affected by different subsets of circumstance factors, suggesting that education^1 and social development in
(^1) It should be noted that, from a normative perspective, educational attainment may be treated either as a circumstance or as an effort variable. On the one hand it is strongly influenced by circumstances
childhood have important implications for key lifestyle choices in adulthood, thereby reinforcing the results of Chapter 2. This corroborates the potential for complementary policies in the educational sector as an instrument for the reduction of health inequalities.
Chapters 4 and 5 explore the interaction between education, cognitive skills, social adjustment and health. Chapter 4 exploits well-defined differences in the educational experience of NCDS cohort-members in order to analyse the relationship between quality of schooling and health disparities. While there is a large literature on the association between years of schooling, academic qualifications and health, little is known about the existence of long-term health returns to different qualities of education. This has important policy implications, as evidence of such returns can inform the design of complementary policy interventions linking the education and healthcare sectors. This chapter contributes to the literature by:
- Examining the scarcely studied association between quality of education and various health outcomes and health-related behaviours.
- Investigating the role of a series of potential mediating channels for these relationships.
- Using the stochastic dominance testable conditions proposed in Chapter 1 to assess whether, from a normative standpoint, quality of schooling can be considered a source of inequality of opportunity in health. The results of Chapter 4 provide corroborative evidence for a statistically significant and economically sizable association between quality of education and a number of health and health-related outcomes. This association remains valid over and above the effects of cognitive ability, social development and academic qualifications. The results also establish quality of education as a source of inequality of opportunity in
beyond individual control: primary and secondary school quality are examples of such circumstances. On the other, it is reasonable to assume that, while impacted by external factors, educational attainment is also partly within individual control. Two approaches are thus possible. One may consider that, in practice, the influence of external factors overrides individual volition, hence educational attainment should, in the context of inequality of opportunity in health, be a circumstance. This approach is followed in Chapter 3. In contrast, one may postulate that despite the influence of circumstances, there remains an important element of individual free choice that needs to be accounted for. Since effort factors in the Roemer model are variables that are at least partly within individual control (E(C)), it follows that attainment can then be classed as one such variable. This is done in Chapter 2.
types of schools. Different interpretations of these results are proposed. One possibility is that quality of schooling acts as a catalyst in the relationship between attainment and health. An alternative interpretation is that this asymmetry reflects a non-linearity in health returns of different levels of attainment.
Chapter 6 establishes a nexus between the findings of each chapter, drawing policy implications and identifying avenues for future research.
Chapter 2
Inequality of Opportunity in Health: Evidence from
a UK Cohort Study
2.1 Introduction
Much of the attention traditionally given to equality of outcomes has shifted towards equality of opportunities. This change of emphasis is the consequence of the latest developments in political philosophy, inspired by the work of Rawls and Sen, systematised by Dworkin (1981), and subsequently modified by Arneson (1989) and Cohen (1989). In recent years, equality of opportunity prompted a series of applications in different fields of economic research 2 and attracted growing interest of policy makers, as becomes clear in the World Bank Development Report
- Within health economics, Rosa Dias and Jones (2007) argued that equality of opportunity is the implicit underlying concept of a broad range of inequality studies published over the last decade. Despite this, the number of empirical applications that explicitly apply this concept to health is still scarce 3 ; this paper aims primarily at narrowing this gap.
All conceptions of equal opportunity draw on some distinction between fair and unfair sources of inequality. Environmental factors such as parental income are largely seen as illegitimate sources of health inequalities. On the contrary, the differences in health status that are due to lifestyles, are often seen as ethically justified by individual choice. These contrasting sorts of factors have been studied independently by two well developed strands of research: the literature on the impact of childhood conditions on adult health and that concerned with health and lifestyles. The interaction between the two is much less explored. Furthermore, both strands were developed in relative isolation from the literature on health
(^2) For example Betts and Roemer (2001), Le Grand et al. (2002), Lefranc et al. (2004) and Bourguignon et al. (2005). 3 Zheng (2006) and Devaux et al. (2008) are two of the very few papers focused on inequality of opportunity in health.
devoted to the impact of childhood circumstances on health outcomes: Currie and Stabile (2004), Case et al. (2005) and Lindeboom et al. (2006) are recent examples. Using different datasets, these studies appraise conflicting theories about the channels by which childhood conditions influence long-term health. The most prominent among these theories are: the fetal-origins hypothesis (Barker (1995), Raveli et al (1998)) according to which parental socioeconomic characteristics influence the in utero conditions for fetal growth which, in turn, condition long term health; the life course models (Kuh and Wadsworth (1993)) which emphasise the impact of deprivation in childhood on adult health and longevity; the pathways models (Marmot et al. (2001)) which suggest that health in early life is important mainly because it will condition the socioeconomic position in early adulthood, which explains disease risk later in life.
This paper follows this strand of research: it considers as circumstances the parental socioeconomic characteristics, spells of financial hardship during the cohort members’ childhood and adolescence, proxies of congenital endowment such as the prevalence of chronic conditions in the family and birth weight, as well as incidence of acute conditions, chronic illnesses and obesity in childhood and early adolescence. All these factors affect the cohort members before the age of 16, reflecting conditions and choices that are largely beyond individual control.
There is also considerable work done on the relationship between health and lifestyles; examples include Mullahy and Portney (1990), Kenkel (1995), Contoyannis and Jones (2004) and Balia and Jones (2008). Lifestyles, such as cigarette smoking, alcohol consumption, and diet are at least partially within individual control, hence they constitute the primary effort factors. While the literature has established that educational outcomes are impacted very strongly by childhood circumstances, it remains plausible to postulate that a degree of educational attainment lies within individual control. Because of this, and given that it is a potential explanatory factor of health in adulthood, it is also taken here for an effort factor.
The Roemer model defines social types consisting of the individuals who share exposure to the same circumstances. The set of observed individual circumstances
allows the specification of these social types in the data. It is assumed that the society has a finite number of T types and that, within each type, there is a continuum of individuals. A fundamental aspect in this setting, is the fact that the
distribution of effort within each type ( F t ) is itself a characteristic of that type;
since this is beyond individual control, it constitutes a circumstance.
In order to make the degree of effort expended by individuals of different types comparable, Roemer proposes the definition of quantiles of the effort distribution (in this case, the number of cigarettes per day or number of units of alcohol consumed per week) within each type: two individuals are deemed to have exerted
the same degree of effort if they sit at the same quantile (^) ( π (^) )of their type’s
distribution of effort. When effort is observed, this definition is directly applicable. However, if effort is unobservable, an additional assumption is required: by assuming that the average outcome, health in this case, is monotonically increasing in effort, i. e. that healthy lifestyles are a positive contribution to the health stock, effort becomes the residual determinant of health once types are fixed; therefore,
those who sit at the π th quantile of the outcome distribution also sit, on average, at
the π th quantile of the distribution of effort within his type.
The definition of equality of opportunity used in this paper also follows from the Roemer model: equality of opportunity in health attains when average health outcomes are identical across types at fixed levels of effort. This means that, on average, all those who adopt identical lifestyles should be entitled to experience a similar health status, irrespective of their circumstances. Such a situation corresponds to a full nullification of the effect of circumstances, keeping untouched the differences in outcome that are caused solely by effort.
When aggregating over different effort levels Roemer (2002) employs the Mean of Mins social ordering criterion, as defined by Fleurbaey (2008, p. 201). This criterion consists of maximizing the average (health) outcome of the whole population that would result if each individual outcome were put at the minimum observed in its own responsibility class. The model is nevertheless compatible with many