Fairness in education – a normative analysis of OECD policy documents

Educational policy depends on assumptions about fairness in education, whether they are made explicit or kept implicit. Without a view of fairness, one would be in the dark as to what should be done about the reproduction of social inequality through education, or whether or not anything should be done at all. The aim of this paper is to uncover the view of fairness in Organisation for Economic Co-operation and Development (OECD) education policy. It is based on an analysis of the normative argumentation concerning educational fairness in a set of policy documents from the last seven years, with special emphasis on the association between social background and educational achievement. The main result of the analysis is that the OECD explicitly operate with a loose idea of equal opportunity, compatible with even a merely formal equality, but implicitly with a meritocratic variant of fair equality of opportunity. In the final section, I argue that the OECD approach to fairness suffers from a limitation in that it considers educational justice in isolation from social justice in general.

influence of international organisations. Prime among these, at least in the affluent world, is the OECD. Although the OECD do not make binding decisions but only policy recommendations, they are one of several organisations that have contributed to 'policy convergence' across nations (Grek 2009). No longer a mere think-tank or data centre, the OECD are now a significant policy actor, spearheading what has been called 'a global education policy field' (Lingard 2011, 368).
Within the sociology of education policy, the OECD are often seen as the main proponent of 'the global education reform movement' (Sahlberg 2011, 99), a neoliberal approach prescribing privatisation, competition, and accountability through testing. Framing this development is a human capital perspective. On this view, the economy has (and should) become more knowledge-based, and education is accordingly seen as a part of economic policy. This leads to what Sellar and Lingard (2013b, 5) succinctly phrase as both the 'economisation' of education policy and the 'educationising' of economic policy. The main vehicle of OECD's influence on education policy has been its assessment programmes, which facilitate comparison of educational performance across nations. PISA, in particular, now commands a significance far beyond that of a mere informative basis for decision-making. Rather, quantitative comparison has become a mode of governance itself (Nóvoa and Yariv-Mashal 2003), constituting what Grek (2009) calls 'governing by numbers'. True, this movement has been most pronounced in the Anglo-American countries and to a lesser degree elsewhere, such as the Nordic countries, but its influence has been felt even there.
Yet, the OECD have also been more and more concerned with equity in education, both in dedicated reports and as part of PISA. A critical question is whether this concern complements or conflicts with the general neoliberal thrust of OECD education policy. How does the emphasis on equity fit with the picture of the OECD as one of the main agents of the marketisation of education? The OECD themselves see equity in education both as an 'end in itself' and as a harmonious complement to the pursuit of economic efficiency (Field, Pont, and Kuczera 2007, 32). Some scholars seem to agree. Chapman and Aspin (2013, 150), for instance, regard the OECD's work on equity as an attempt to balance the economic perspective on education with 'broader human and social concerns'. In this perspective, equity is considered necessary to optimise economic growth, since inequities imply that human resources are left unused.
Others are more critical. Savage, Sellar, and Gorur (2013) acknowledge that PISA has highlighted issues of inequity, such as the persistent association between social background and educational performance, but claim that the OECD's equity discourse has been co-opted into economic rationality. They see the growing social inequality in member countries as evidence that the OECD's promotion of equity 'sits uneasily' with their neoliberal agenda of market reform (163). Rizvi and Lingard (2009, 448) likewise argue that within the OECD framework, equality is no longer regarded as a moral value in its own right, but only as 'a means to human capital development', and that the OECD notion of equity is therefore 'too narrow'. In the same vein, Rizvi (2013) describes this development as a shift from a social democratic to a market conception of equity. Equity has become separated from traditional ideas of social justice and re-articulated as formal access to education and participation in markets.
My own contribution to this debate is to shed light on the OECD notion of equity by way of a close textual analysis of selected policy documents. To be sure, textual analysis is far from sufficient to determine actual policy, since policy documents are always embedded within a larger political, ideological, and historical context, but it is a necessary prerequisite. As will become clear, the conclusion of this study will be a partial concession to the critical view. On one hand, there is a quite demanding conception of fairness in the OECD, and a stronger one than they sometimes get credit for. On the other hand, the critics are right that the OECD conception of fairness is defined in isolation from social justice in general.

The normative problem
Next, a few words about the underlying normative problem: one of the most wellknown and well-established findings from the sociology of education is that educational performance is highly correlated with social background (Esping-Andersen 2004). 1 Most people will intuitively feel that there is something morally fishy about such correlations. Yet, it is hard to say exactly why it is wrong, unfair, or unjust. Hence, it is also hard to determine exactly when correlations of this kind are evidence of injustice, and when, if ever, they are not. We might for instance agree that it would be deeply unfair if skin colour in itself systematically influenced educational performance and thereby later status in society. We might also agree, although this is more controversial, that it would not be unfair if study effort influenced educational achievement; that, in many people's view, is exactly as it should be. But where, in between these two poles, do we cross the line from fair to unfair? Which factors can legitimately influence achievement? Innate talents? Parental income? Parental education? Homework support?
This normative problem is debated within political philosophy, for instance by Rawls (1971), Sen (1997), Roemer (1998), Arneson (1999) and Swift (2006, 2009). Much of this debate has centred on the notion of equality of opportunity, which, in various forms, figures prominently in theories ranging from traditional meritocratic views to more recent 'luck-egalitarian' views (Mason 2001). Regrettably, this philosophical debate has largely proceeded in splendid isolation from actual education policy. 2 This is unfortunate, since education policy depends (among many other things, of course) on assumptions about fairness in education, whether they are made explicit or kept implicit. Without a view of fairness, one would be in the dark as to what should be done about the reproduction of inequality through education, or whether or not anything should be done at all. The aim of this paper is to uncover the view of fairness in OECD educational policy documents. How do the OECD answer, explicitly or implicitly, the questions above? As we shall see in the analysis to come, the OECD documents also operate with a notion of fairness as equality of opportunity. We must therefore begin by introducing some terminology and clarifications regarding this troublesome concept.
When asked, most people support the idea of equal opportunity. However, that agreement is largely verbal, since 'equal opportunity' can take on very different meanings. A common distinction is that between formal and fair equality of opportunity, but here I shall instead adapt a tripartite division made by Swift (2006, 98-106). Note, though, that Swift is concerned with equal opportunity in general, whereas this paper is restricted to educational opportunity. Obviously, there are close links between educational success and occupational success, and those links are part of what makes educational fairness important. Nevertheless, the question of social justice in general, including the fairness of correlations between social background and social status (e.g. income differences), raises additional problems that would complicate my analysis here to the point of becoming unmanageable. In this paper, therefore, equality of opportunity will mean equality of educational opportunity.
Minimal equality of opportunity includes what is often called 'formal' equality, that is, the absence of legal restrictions on educational access, but is already broader. Minimal equality refers to the type of equality that is undermined not only by discriminating laws, but also by prejudice and informal discrimination. The key normative idea is that class, gender, or ethnicity should neither determine whether one receives a good education, e.g. access to higher education, nor how one is treated within educational institutions, e.g. in grade-setting. Minimal equality of opportunity can thus be summarised as the view that a competent student from a poor family should have the same educational chances as a similarly competent child from a wealthy family.
Fair equality of opportunity goes a step further. The difference between fair and minimal equality of opportunity can be captured by a distinction between competence (actual or realised knowledge and skills) and talent (natural or potential). Whereas minimal equality is the view that performance (length of education and grades achieved) should be a function of competence, fair equality is the view that performance and competence should be a function of talent (and effort). The crucial implication is that according to the principle of fair equality of opportunity, one's opportunities to acquire competence should be independent of family background. To achieve this, the elimination of prejudice and discrimination is insufficient; rather, the playing field must be levelled by compensating disadvantaged children. To sum up, a talented child from a poor family should have the same educational chances as a similarly talented child from a wealthy family.
Finally, the principle of radical equality of opportunity says that opportunities should exclusively be a function of people's choices, in particular how much effort they choose to exert. The key idea behind this principle is the following: since the talent you are born with is just as much a matter of luck as the family you are born into, one should, for the sake of consistency, compensate for talent as well. Educational opportunities should be a function of choice and effort alone. The implication is that the influence of natural talent should also be equalised or neutralised. Radical equality of opportunity can thus be summarised as the view that a less talented child should have the same educational opportunities as a more talented child.

Material and method
The OECD policy document to be analysed in this paper is primarily No More Failures: Ten Steps to Equity in Education (Field, Pont, and Kuczera 2007), but also Tertiary Education for the Knowledge Society (OECD 2008), Equity and Quality in Education (OECD 2012a) and Education Today 2013: The OECD Perspective (OECD 2012b) will be examined. These documents overlap to a considerable degree with respect to fairness. The first contains the most thorough treatment (though not as thorough as one could have hoped for), and many of its formulations are repeated in the later documents. I shall not attempt to trace the origin of those formulations, but the OECD's notion of fairness seems to be based on Levin (2003). However, I have not included an analysis of that text here, since it is not an official OECD publication, but a paper commissioned by the OECD. 3 The approach of this study may be unfamiliar to some readers, since it is a combination of philosophical argument and policy document analysis. Philosophers are not used to drawing on empirical material in their discussions; they are usually more at home with imaginary examples, in part for good reasons. Empirical researchers on the other hand will sometimes be tentative about unabashed moral discussion, even though they are often motivated in their work by moral concerns. In particular, researchers who are accustomed to analysing political documents will perhaps be wary of talking about the moral content of policy documents, which are often seen as belonging to the nitty-gritty world of political power games. A normative analysis may seem to them to be naive, as though the documents were taken purely at face value with no notice of darker motives and hidden messages.
Naturally, the question of the role of moral or normative conceptions in policy or policy documentsor for that matter the role of policy documents in policyis a far too complex question for me to attempt to answer here. The basic assumption for this paper is fairly modest: policy documents often include normative elements, and these play an important argumentative role within the documents, even if they may play a less crucial role in the actual implementation of policy. Moreover, it is important to bring these normative elements out into the open, not least in order to confront actual policy with them.
As mentioned, this paper is based on an analysis of the normative argumentation in a set of policy documents. However, argumentative analysis, especially of normative argumentation, and even more so of policy documents, is no mechanical exercise. Rather, it can be described as a logical-hermeneutical art form. Despite its not being a technique, however, we can distinguish some of its characteristic stages. For the sake of simplicity, a full argumentation analysis may be said to involve four main steps.
The first step is identification. Here, we attempt to identify the various parts of the argumentative structure: claims, reasons, and evidence. At this stage, we are exclusively at the explicit level, in the sense that we stick to how the claims and reasons are actually formulated. The second step is reconstruction. Here, we attempt to reconstruct the argument as a whole: how the premises and conclusions are related to each other, as well as the relations between the various arguments and sub-arguments. 4 At this stage, we are partly at an implicit level, since the relations between the various premises, conclusions, and sub-arguments are often not explicitly stated. This means that in the case of complicated arguments, such as those in policy documents, there may be several possible reconstructions, each equally true to the source material. The third step is articulation. Here, we want to get a more precise and specific grasp of the premises by drawing out their implicit content, but also to identify implicit premises in the argument. 'Implicit' does not refer here to something behind the text, for example, the intentions or motivations of its creators, but what can be discerned from the text itself, by considering the context of each premise and the relation between the parts of the argument, especially how the premises function in the argument as a whole. The fourth step is evaluation; that is, the assessment of whether or not the argument or arguments are valid and sound. I will not go into detail here, since my analysis does not move beyond the three first steps.
To perform a full analysis of a complete policy document would be extremely time-consuming; a document contains hundreds, if not thousands, of premises and conclusions, in a complicated web of arguments and sub-arguments. My aim is therefore to focus on only one part of the argumentative structure of the documents.
Nevertheless, some general ideas about what argumentation looks like are still necessary in order to identify what to look for in the documents. In this paper, Douglas Walton's model of ethical argumentation has been used as a tool of analysis (Walton 2003). Yet, since the aim of the paper is not to analyseand even less to evaluatethe entire argumentative structure of the relevant documents, but more modestly to identify and articulate clearly just one premise of that argument, I shall only employ those parts of Walton's model that are relevant to identify and clarify the specifically moral principles involved. I shall not, for instance, extend my interest to the reasons for those principles, or what Walton (2003) calls 'sub-inferences'. Hence, only the first three steps above are relevant here.

The explicit level
The methodological leitmotif of this paper is that public policy documents can be seen as instances of practical ethical reasoning. In its simplest form, ethical argument, as a kind of practical deliberation, has the following structure, which also provides the core of more complex models: P1: Moral principle (It is wrong to lie for personal advantage) P2: Situation description (She's mad at me, but if I tell her a lie, she'll stop) C: Action prescription (Tell her the truth anyway).
My aim is not to evaluate the argumentative structure of the policy documents, but rather to identify it, and moreover to identify only one part of it, namely P1. Hence, I shall not address the difficult question of whether or not such arguments can or are even meant to be deductively valid. According to Walton (2003, xiii), ethical argumentation is best seen as 'abductive'. Unlike a deductive argument, the truth of the premises in an abductive argument does not guarantee the truth of the conclusion, but gives us at best good reasons to accept it. Thus, even valid ethical arguments are not proofs. There are no guarantees that they will be met with unison agreement.
A policy document is obviously much more complex than the simplified example above suggests. Nevertheless, even in policy documents, the same core structure can be discerned; we come to a conclusion about what to do on the basis of how we think things actually are and a general moral rule. In the case of the OECD, it can be laid out like this: P1: Moral principle: the equity principles P2: Empirical studies on educational performance and effects of background C: Policy recommendations: the ten steps to equity in education.
The conclusion (C) in the OECD documents, then, amounts to the so-called 'Ten Steps', which involve among other things limitation of early tracking, management of school choice, strengthening of links between school and home, and the directing of resources to the students with the greatest needs (Field, Pont, and Kuczera 2007, 9). This conclusion is based on the empirical studies (P2). Most of these fall into two groups: comparisons of educational performance among OECD member states, and studies on the relation between educational performance and background variables (sex, ethnicity, class, and so on, but also national policies such as school choice and public expenditure on education).
Clearly, P2 alone does not imply C. The empirical data themselves are practically inert unless combined with principles that judge the present state of affairs as inefficient or unjust, such as for instance the fact that some children leave school without being able to read properly, or that children from middle-class homes do better in the educational system than do those from working-class homes. The question that this paper sets out to answer is: what exactly are these principles in the case of the OECD? What amounts to P1 in the relevant policy documents? At the first and explicit level, the role of P1 in the OECD policy documents is played by the notion of equity. In OECD parlance, equity has two dimensions, fairness and inclusion, that correspond to two moral principles. The principle of fairness states that 'personal and social circumstancesfor example gender, socio-economic status or ethnic originshould not be an obstacle to achieving educational potential' (Field, Pont, and Kuczera 2007, 11). The principle of inclusion states that there should be 'a basic minimum standard of education for allfor example that everyone should be able to read, write and do simple arithmetic' (Field, Pont, and Kuczera 2007, 11).
It must be said that although OECD argue that these two dimensions of equity are closely intertwined in that they tend to reinforce each other in practice, it is still puzzling that they are combined into a single concept (Field, Pont, and Kuczera 2007, 11;OECD 2012a, 16). Fairness and inclusion are logically independent of each other, and it is easy to imagine cases where they pull in different directions. Within political philosophy, there is a debate about which of these two principlesthere called equality and adequacyshould override the other if they conflict (Anderson 2004;Brighouse and Swift 2009). With their terminology, however, the OECD make it hard even to see that there is a problem. 5 The part of P1 that we are concerned with here is the principle of fairness. At this stage of the analysis, this principle is still quite loose. If we restrict ourselves to social background, for instance as measured by parental income and level of education, all we know at this point is that background should not be an obstacle or barrier in education. That formulation is compatible with the most formal equality of opportunity, mandating only the removal of legal obstacles, but also with a very strong principle of equality, that demands that social background should have no statistical effect whatsoever. So far, though, we cannot know which interpretation is correct, and thus what the 10 steps are meant to lead to. It is however difficult to get a more specific idea of P1 in the OECD documents, since the authors seem to suffer from a kind of moral forgetfulness. The moral aspects and elements are there, but they are neither justified, nor elaborated, nor specified, and they take up much less space than do the empirical evidence and policy recommendations, even though (or perhaps exactly because) it is here we find the real controversies. As I shall demonstrate, more sophisticated tools of analysis might however be able to unlock the more specific idea of fairness at play in the OECD documents.

The implicit level
Moral argumentation is often circular, in the sense that there is a dialectic interplay between more general and more particular judgements. General judgements are supported by particular judgements, and vice versa. That does not have to be taken as proof of defectiveness. Rawls (1974) famously emphasised the same justificatory interdependence between judgements at different levels of abstraction in moral reasoning, and in effect distinguished good circularity from bad through whether the general and the particular are in reflective equilibrium or not. More important for our purposes, general and particular judgements can at times even constitute each other. What is meant by a general judgement might first become clear when seen together with the particular judgements it is held to imply. 6 The holistic or circular character of moral reasoning means that P2 and C can take part in determining the exact content of P1. Which empirical studies are chosen and which policy measures are recommended may, therefore, play a constitutive role in clarifying the principle of fairness from which OECD work. 7 On this background, we may hope that we can achieve a progressive specification of the content of the OECD's principle of fairness by considering both the role of cited empirical studies and the work that the policy recommendations are supposed to do. In Walton's terminology, we are trying to work out the implicit classificatory assumption regarding the key term in the moral premise: 'obstacles'. Unlike for example 'challenge', obstacle is here a negatively loaded emotive term. 8 The documents do not however tell us much about which personal and social circumstances count as obstacles. Is your genome an obstacle? Are your parents, and if so, when? Such questions are never explicitly answered by the policy documents, but we can come a long way towards reconstructing an answer from their use of empirical evidence and proposed policies. When we attend to contextual elements in this way, a more subtle and stronger conception of fairness is revealed in the OECD documents. What seems at first a vague and undemanding position turns out to be rather egalitarian. Whether or not its policy recommendations are adequate to realise the demands implicit in its conception of fairness, however, is a matter that falls outside the scope of this paper.
As a first step towards limiting the range of possible interpretations of obstacle, we can look at the general use of empirical evidence on correlations between background and performance. We saw above that the explicit OECD principle of fairness is compatible with a mere formal equality of opportunity. However, if the OECD only assumed a formal principle, we could not even tell from the provided statistical evidence regarding the effect of social background that the principle of fairness is violated within the OECD today, given the fact that no OECD country has, as far as I know, laws that for example prohibit working-class students from entering university. Yet, since it is clear from the context that the OECD do treat the statistical material as evidence for a violation of the principle of fairness, the formal interpretation of the principle is excluded (Field, Pont, and Kuczera 2007, 41-42). Hence, by obstacles, the OECD cannot merely mean legal obstacles.
We can limit the range of interpretations further by noting that there is one kind of obstacle that is not regarded as an unfair influence on educational achievement: natural talent. The OECD state that 'accidents' should be overcome in education, but an exception is made for talent. This is partly explicit in the OECD principle of fairness itself, which states that personal and social circumstances should not be an obstacle to achieving educational potential (Field, Pont, and Kuczera 2007, 11). Since one of the measures that the OECD recommend in order to promote fairness is early childhood education and care, we must conclude that 'educational potential' refers to potential at a very early stage (Field, Pont, and Kuczera 2007, 21). Hence, it seems reasonable to interpret educational potential as referring to native talents or abilities. This interpretation gains further support from Equity and Quality, where an equitable education system is described as one 'where individuals can improve their socio-economic situation on a basis of merit' (OECD 2012a, 26). Since merit is commonly defined as talent plus effort, this meritocratic reading is strengthened by Tertiary Education, where it is said (though only with respect to higher education) that in an equitable education system, participation and outcomes are only based on 'individuals' innate ability and study effort' (OECD 2008, 14). If the OECD think that innate ability is what should count even at the tertiary level, the principle of charity leads us to assume that the OECD also think that it is what should count at lower levels, since the latter is, if anything, more reasonable than the former.
Concerning effort, we cannot however reason in the same way. Whereas it may seem reasonable to assume that effort is what should count at the tertiary level, where students are responsible adults, it seems less reasonable to assume that it should be what counts at lower levels. Effort is largely determined by motivation, and motivation for education, at least in smaller children, can be held to be partly or even largely determined by social background. That may be why effort is only mentioned in the analysed documents with regard to tertiary education (OECD 2012a, 26). For that reason, and also because the question of whether or not we are responsible for our effort so easily takes us into general philosophical questions regarding free will, we shall disregard effort here. What is not in doubt is that the OECD documents consider natural talent (or potential) as a factor, indeed, the factor, that should condition educational attainment.
The precise meaning of obstacle can be circumscribed to an even greater extent by considering the treatment of particular studies and policies in the documents. We shall not however go through all the statistics and recommendations, since they overlap to a degree, but pick out those in which the implications for fairness are most salient.
As mentioned, ethical arguments are typically defeasible. One reason is that general moral principles often admit exceptions. Ross (1930, 20) famously used the term prima facie to indicate the conditional character of moral duties. A prima facie duty is a duty at first sight, a duty that may have to be qualified or retracted in light of other considerations (Walton 2003, 9). Depending on other moral principles and/ or particular contextual features, a prima facie or conditional principle may be overridden. This is germane to our topic, for in the OECD documents some background variables are treated as categorically unfair as predictors of educational achievement, whereas others are treated as merely conditionally unfair.
In the section called 'Where are the big problems?', the OECD illustrate 'the problem of unfairness' by pointing out that across the OECD area, 'children from poorer homes are between three and four times more likely to be in the lowest scoring group in mathematics at age 15' (Field, Pont, and Kuczera 2007, 13). 9 The influence of parental economy on educational performance is here treated as categorically unfair; it is neither a matter of degree nor conditional upon other factors. This type of correlation is regarded as unfair tout court. True, a lower odds ratio would probably signify a lower degree of unfairness, but the unfairness is not seen as a matter of degree in the sense that a sufficiently lower odds ratio would no longer signify unfairness at all. It would perhaps have been less unfair if poor children were, for example, only twice as likely to be in the lowest-scoring group, but it would still be unfair. Nor is the unfairness of the effect of parental economy on educational performance treated as conditional upon other factors. This can be seen from a comparison with other influences on achievement: parental education and home support.
In their second policy recommendation for equity in education, the so-called 'Step 2', the OECD advocate careful management of school choice, since school choice poses a risk to equity in general and fairness in particular. One reason is that well-educated (and better-off) parents have the resources to exploit that opportunity and make better choices for their children, something which will 'accelerate the progress of those who have already gained the best start in life from their parents' (Field, Pont, and Kuczera 2007). Parental education is therefore seen as at least a risk to fairness, though it is not stated or implied that any correlation between parental education and school performance is unfair. That risk means that the influence of parental education should be reduced through policy changes, for instance through limiting school choice and academic selection, which will lessen well-educated parents' opportunities for making their education count as much.
Something similar applies to the OECD's ' Step 6'. The evidence shows that parental support, such as helping with homework, is associated with school performance. However, since resourceful parents involve themselves more with their children's education, parental support also poses a risk to fairness (Field, Pont, and Kuczera 2007, 100-101). Once again, it is not stated or implied in the documents that any correlation between home support and school performance is unfair. Unlike the former factor (parental education), however, this risk is not sought to be reduced by policies that make it more difficult for resourceful parents to exploit their advantage, but rather by raising the level of support for children of less resourceful parents. This can for instance be done through after-school homework clubs and the reinforcement of links between school and home (Field, Pont, and Kuczera 2007, 108).
In these documents, then, unlike parental income, which is treated as categorically unfair as an influence on educational achievement, parental education and home support are regarded as 'merely' conditionally unfair. However, the defeating condition in the two cases is not, as it often is in these debates, parental rights, but contextual features: more specifically, school policies. The effect of parental education is seen as unfair if it is being used to gain an (additional) advantage through school choice and selection (Field, Pont, and Kuczera 2007, 78). As I interpret the documents, the OECD position can therefore be summed up thus: the advantage one gains, in a myriad of ways, by having well-educated parents is not automatically unfair, but it is unfair if that initial advantage is allowed to turn into an additional advantage through school choice and selection, which will in turn lead to disadvantaged students acquiring an inferior education. Similarly, the influence of home support is unfair if there are no compensatory measures to support disadvantaged students. For instance, the OECD do not recommend abolishing homework, but instead advise schools to set up things like homework clubs in order to compensate disadvantaged students (Field, Pont, and Kuczera 2007, 102-104). In this manner, they also seek to avoid the levelling-down argument against educational fairness.
To sum up: the OECD explicitly operate with a loose idea of equal opportunity, compatible with even a merely formal equality, but implicitly with a meritocratic variant of fair equality of opportunity. The OECD principle of fairness can thus seem very similar to John Rawls' principle of fair equality of opportunity, which says that people with equal talent (and motivation) should have equal chances of succeeding: More specifically, assuming that there is a distribution of natural assets, those who are at the same level of talent and ability, and have the same willingness to use them, should have the same prospects of success regardless of their initial place in the social system. In all sectors of society there should be roughly equal prospects of culture and achievement for everyone similarly motivated and endowed. The expectations of those with the same abilities and aspirations should not be affected by their social class. (Rawls 1971, 63) We shall return to Rawls' position in the next section, but considered on its own, the principle that Rawls formulates here for the social system as a whole seems similar to the one with which the OECD implicitly operate for the educational system. They diverge, however, in significant ways.
There are three differences, which nevertheless are closely related. First, nowhere in the OECD documents do we find an explicit or implicit advocacy of a complete eradication of the influence of social class. Whereas, Rawls states explicitly that equally talented students should have the same prospects for success, the OECD do not commit themselves to the ideal of eradicating the influence of class, only that it should be of less importance. The aim, therefore, is not equality of opportunity, but less inequality of opportunity. 10 Second, the conditional unfairness of the influence of parental education and home support introduces a significant indeterminacy in the question of educational justice. Statistics showing correlations between social background and educational attainment cannot then be taken ipso facto as a sign of systemic unfairness. Third, rather than referring to equal chances or prospects or probabilities, the OECD prefer the vague and metaphorical 'obstacle'. When exactly a certain variable, for example home support, goes from being an influence to being an obstacle is neither stated nor easy to fathom. The combined effect of these three differences is a significant tempering of the OECD principle compared with that of Rawls. Whether that is a strength or weakness is not for me to decide here.

Concluding discussion
One may speculate as to why the stronger notion of fairness is kept implicit in the OECD documents. The OECD attempt to explain (or excuse) their lack of a clear explicit notion of fairness by citing Levin, who claimed, in a paper commissioned by the OECD, that while 'we cannot define what [equity] is, we know when we are far from it' (2003,5). Yet that rings false, since the OECD implicitly operate with a more specific idea of fairness than the one they explicitly define, and also since without such a specific idea, we could not even 'know when we are far from it'. It seems to me that a simpler and more reasonable explanation for the disparity between the implicit and the explicit level is that the OECD have to cater to a variety of states, ranging from Scandinavian social democracies to Anglo-Saxon liberal welfare states, to use Esping-Andersen's famous typology (1990). That may have led the OECD to soften a principle of fairness that in some political circles in some of these states would have been seen as too radical.
Although it has not been my intention with this paper to evaluate or criticise the normative argumentation in the OECD documents, I shall end by mentioning one serious problem with it, since it bears on the relation between these documents and others, and thus between education policy and other areas of public policy. This will align the conclusion of the present paper with other critical perspectives on the OECD equity discourse; it will also reveal a deeper difference between the positions of Rawls and the OECD.
To a large degree, the analysed documents isolate the question of educational justice from the question of social justice in general. There is hardly any attempt to relate fairness within the educational system to fairness within the social system in general, be it on the micro-level of individual action or the macro-level of social structure. Although there is an emphasis on equality of opportunity as a means through which to achieve positions in the social hierarchy, there is on the one hand no discussion of the rightfulness of that social hierarchy itself, nor, on the other, of whether or not students from different social classes may fairly and legitimately simply choose different educational trajectories.
This criticism may seem unfair. After all, the documents with which I have been concerned are on education, not on society as a whole. There are other white papers in the OECD for that. Moreover, even the documents treated here do from time to time highlight the links between educational equity and equity in general: 'Fair and inclusive education is one of the most powerful levers available to make society more equitable' (Field, Pont, and Kuczera 2007, 11) and 'success in improving equity in education also depends upon other policies (e.g. health, housing, welfare, justice, social development)' (OECD 2012a, 18). Nevertheless, this is not sufficient to parry the criticism made above, since these links are causal in nature whereas the criticism under consideration here trades on the idea that educational justice is partly constituted by general social justice. The point is that one simply cannot determine, from looking at the educational system alone and the correlations between its inputs and outputs, whether or not that system is fair. Hence, one cannot distinguish fairness in education from fairness in society as a whole and treat them as though they were two separate questions, belonging to two different departments.
Other scholars have arrived at similar conclusions. Sellar and Lingard (2013a, 723) note that while PISA focus on the equity of schooling systems, the OECD often omit consideration of 'structural inequality in society beyond schooling' as explanation of educational inequities. As part of his argument that egalitarianism is beneficial to educational excellence, Condron (2011, 54) likewise contends that 'it is problematic to think that schools can reduce achievement disparities by themselves', since schools are 'embedded within the economic systems of their societies'. The same conclusion is drawn by Reay (2012, 592-593): 'Educational inequalities are inextricably bound up with social inequalities and cannot be addressed in isolation from them'. Yet these scholars do not distinguish clearly or explicitly between a causal and a constitutive relation between educational and social justice. By distinguishing these, my study has allowed a more precise understanding of the way in which the OECD isolate educational fairness from social justice in general.
We are now in a position to see the deeper difference between Rawls' view of equality of opportunity and that of the OECD. For Rawls, fair equality of opportunity is only one part of the story. Social inequalities must also fulfil the condition set by the so-called difference principle: they must be to the greatest benefit of the worst off (Rawls 1971, 302). This implies, according to Rawls, that the degree of inequality must not become so large so as to undermine the self-respect of the least advantaged (107). The distributive implications of Rawls' theory therefore go far beyond that of the principle of fair equality of opportunity alone. When equality of opportunity is seen in tandem with the difference principle, it also becomes clear that Rawls' theory does not imply meritocracy: if there are to be positions of advantage, then the justification cannot be that the advantaged deserve their positions, seeing that their advantage largely depends on natural and social fortune (103-104). Although Rawls does not go as far as to support radical equality of opportunity, since he favours a social structure where the effects of differences in talent are not redressed but rather 'work for the good of the least fortunate' (102), neither does he support the meritocratic variant of equality of opportunity that we find in the OECD documents.
Rawls' position also implies that the fairness of the educational system ultimately depends on the fairness of the structure of society as a whole and cannot be determined by the principle of equal educational opportunity alone. Whether a particular pattern of correlations between social background and educational performance is fair or unfair will in this view depend on the degree to which society in general, and its distributive pattern in particular, is fair or unfair. For instance, in a fully just society, it might not matter or might even be a good thing if parents transmit their skills, knowledge, values, and good habits in such a way that their children will tend to take up the same occupations as they have. However, in a society marked by large and unjust inequalities, that very same kind of reproduction is morally problematic. For this kind of egalitarian, the real problem is not that the occupational structure is partly reproduced through education, but that this structure is one of great inequality in status, power, and other resources. This perspective is absent from the analysed policy documents: While the OECD insist strongly that coming from a disadvantaged home should not be a disadvantage in education, they do not question whether there should be disadvantaged homes at all. Notes 1. I shall start out by using 'social background' in an open and unspecified sense, since part of the aim of the article is to discover which factors OECD think should and should not be allowed to influence educational performance. Furthermore, 'performance' and 'achievement' will, unless specified, refer to both length, results (grades), and type (prestige) of education. 2. An important exception is the work of Adam Swift and Harry Brighouse. 3. Unfortunately, space does not allow me to compare this set of documents with earlier OECD policy documents. Thus, I am no position to decide whether or not the position of fairness I outline here is evidence of change or stability in the OECD's position. For a treatment of the OECD educational policy before 2007, see Rizvi and Lingard (2006). 4. The first two steps are not strictly separated, since merely to identify something as a claim and something else as a reason for that claim is already to have some idea of the structure as a whole. 5. Marginson (2011) also focuses attention on the tension between fairness and inclusion.
He thinks (26) that of those two, inclusion should be the primary value and target. He also associates the two values, fairness and inclusion, with two differing philosophical traditions with regards to social justice, a utopian and a realist, respectively, but I find this part of his analysis less convincing. 6. At this point, I depart from Walton's 'chained' or 'layered' model (2003). 7. Of course, one possibility is that the particular policy is in outright conflict with the normative principle; in that case, the argument is a bad one. As mentioned, though, my aim here is not to assess the quality of the argument, but only to identify one premise of it. 8. Walton argues convincingly that there is nothing inherently wrong with or fallacious about using loaded terms in argumentation (2003). 9. 'Poor' is here defined relatively as those belonging to the lowest socio-economic group.
The relation between relative and absolute definitions of poverty is important for the question of how the OECD understand the normative significance of poverty and social inequality. Unfortunately, space does not allow me to go into this topic here, since it will take us too far into issues of general social policy (OECD 2011). 10. As explained above, I have not considered effort and motivation in this article, since the OECD documents simply do not give us enough to go on in that respect. Furthermore, I have also disregarded the possibility that natural talent is not equally distributed among social classes. Rawls seems to assume that both natural talent and effort/motivation (at least one aspect of it) are independent of class.

Notes on contributor
Steinar Bøyum is an associate professor at the Department of Education, Faculty of Psychology, University of Bergen, Norway. He holds a doctorate in philosophy, and has published widely in both philosophy and education. His main interests are in education policy, political philosophy and philosophy of education.