ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?

This paper discusses whether or not policies of positive discrimination, such as reservations, should be deployed in order to reduce the social and economic marginalisation of disadvantaged racial, caste or other ethnic communities. PD policies giving preferences to members of such disadvantaged communities are likely to have significant negative as well as positive consequences. Are there alternative policies - such as class-based preferences - that could deliver similar benefits without significant costs? After reviewing the diversity of ways in which a race-, caste- or ethnicity-based PD policy can be structured, it concludes that the case for deploying such PD policies is strong if the policies are carefully designed to maximise the benefits and minimise the costs.

Special articles

Is Positive Discrimination a Good Way to Aid Disadvantaged Ethnic Communities?

This paper discusses whether or not policies of positive discrimination, such as reservations, should be deployed in order to reduce the social and economic marginalisation of disadvantaged racial, caste or other ethnic communities. PD policies giving preferences to members of such disadvantaged communities are likely to have significant negative as well as positive consequences. Are there alternative policies – such as class-based preferences – that could deliver similar benefits without significant costs? After reviewing the diversity of ways in which a race-, caste- or ethnicity-based PD policy can be structured, it concludes that the case for deploying such PD policies is strong if the policies are carefully designed to maximise the benefits and minimise the costs.

THOMAS E WEISSKOPF

I Introduction

I
n every country in south Asia there are racial, caste, or other ethnic communities whose average welfare – by many social and economic indicators – is significantly below that of the population as a whole. In several of these countries, policies of positive discrimination (PD) have been introduced in an effort to reduce historically persistent lags in the social and economic welfare of relatively poor communities. By “positive discrimination” I mean preferential selection of members of underrepresented ethnic communities1 to desirable positions in society. Where such policies have been implemented, they have most often proven highly controversial. Indeed, they have not only given rise to heated debate but also at times to major public demonstrations, pitting actual or prospective beneficiaries of PD policies against those who are – or believe they are – disadvantaged by the policies.

Should PD policies be deployed to combat the marginalisation of members of ethnic communities that find themselves significantly under-represented in society’s upper strata? Proponents of such policies have suggested many ways in which they may help to bring about a more equal and/or a more vital society:2

  • PD policies compensate for explicit negative discrimination as well as implicit biases that would otherwise unfairly penalise members of under-represented ethnic communities;
  • PD policies redistribute resources and opportunities from relatively well-off ethnic communities to relatively poorly-off communities;
  • PD policies improve the motivation of members of underrepresented ethnic communities to aspire to and work toward more desirable positions in society;
  • PD policies lead to better performance by institutions or
  • organisations where greater diversity of personnel in key positions (resulting from greater representation of under-represented ethnic communities) contributes to productive efficiency;

  • PD policies enable members of under-represented ethnic communities to gain better access to social capital – i e, useful contacts and networks that improve one’s career opportunities – which are currently available mainly to members of other communities.
  • PD policies serve to integrate under-represented ethnic communities into society’s elite, thereby fostering a more legitimate and vital democratic order.
  • Critics of PD policies, on the other hand, have pointed to many reasons why such policies may prove undesirable and possibly counter-productive:
  • PD policies clash with the liberal principle of evaluating every person on the basis of their own individual abilities rather than on ascribed group characteristics;
  • PD policies benefit mainly the best-off members of underrepresented ethnic communities (the “creamy layer”) rather than those members who most need help;
  • PD policies arbitrarily penalise those members of nonbeneficiary communities who happen to be displaced by preferential access for members of beneficiary communities;
  • PD policies lead to poorer performance on the part of beneficiaries than would have been provided by those who would otherwise have been selected;
  • PD policies lead to under-appreciation of the accomplishments of members of beneficiary communities, whose successes may be attributed to PD policies rather than to their own abilities;
  • PD policies, as a consequence of all of the above, have a strong potential for exacerbating inter-community tensions and divisions.
  • Given all these potential drawbacks of PD policies, one is led to ask: aren’t there better ways to achieve the prospective benefits, by means of policies that would not give rise to so many possible costs?

    II Alternatives to Ethnicity-Based Preferential Selection

    Why Not Use Other Policies to Aid Disadvantaged Ethnic Communities?

    One obvious set of alternatives to PD policies based on preferential selection of members of under-represented ethnic communities consists of policies that directly protect members of such communities from negative discrimination. A second alternative set of policies consists of programmes in which resources are redistributed to members of such communities. In practice, both these types of policies often accompany policies of preferential selection. Given the potential negative consequences of PD policies, however, one might well ask if it is not better to pursue only anti-discrimination policies and resource-transfer programmes as a means of aiding marginalised ethnic communities.

    The experience of the US in the 1960s with “affirmative action” displays the limitations of a purely anti-discrimination strategy. The initial thrust of US affirmative action policies was precisely to neutralise negative discrimination via a set of new antidiscrimination measures. The landmark 1964 Civil Rights Act was based on the equal protection clause of the US Constitution; it formally outlawed discriminatory practices in almost all public spheres of American life. The Civil Rights Act was followed by a series of executive orders to promote equal opportunity in employment and education. It was hoped and expected that the assertion of formal legal equality of all citizens, the removal of overtly discriminatory barriers, and a much wider diffusion of relevant information, would lead to significant increases in the representation of African Americans in desirable jobs and educational institutions. In this context the term “affirmative action” was applied to active outreach efforts made by organisations and enterprises to assure that members of under-represented communities would have the same kind of access to jobs and educational opportunities as their more advantaged fellow citizens – without any regard for their ethnic identity.

    It soon became apparent, however, that affirmative action of this kind would not have an immediate and significant impact on the numbers of African Americans in most professions and educational institutions in which they had always been greatly underrepresented. On the premise that continuing under-representation was due in significant part to continuing (if on the whole subtler) forms of discrimination against African Americans, the US labour department began to measure progress in ending discrimination in terms of quantitative increases in the percentage of African Americans in various fields and workforces. Public organisations, as well as non-profit institutions and private companies supported by or otherwise linked to the federal government came under pressure to increase the representation of African Americans in order to demonstrate that they were operating in a nondiscriminatory manner. Such pressure increased as courts began to accept statistical information on the low percentage of African Americans employed, in relation to their proportion of the population presumably qualified for a position, as evidence of racial discrimination. By the late 1960s these developments had led many government agencies, as well as some private organisations, to discriminate in favour of African American candidates for jobs or contracts. Moreover, many educational institutions – both public and private – and some other non-profit organisations sought to increase the representation of African Americans by applying preferences in admission or selection. In this context the term “affirmative action” took on a stronger new meaning – which it has retained ever since – of positive discrimination via preferential selection in favour of under-represented communities.3

    Why not, then, pursue policies involving resource transfers instead of preferential selection in favour of under-represented ethnic communities? Such resource transfers could be directed to members of those communities in such a way as to enable them to develop the skills needed to qualify for better jobs or to acquire the capital needed to launch business enterprises, thus contributing directly to economic development as well as social uplift. Preferential selection policies are problematical because, among other things, they concentrate losses on those applicants displaced by the admission of beneficiaries; and this leads not only the applicants thus displaced, but also many other rejected applicants, to attribute their rejection to the unfairness of preferential policies. A programme of developmental aid to members of under-represented communities, on the other hand, involves costs that are spread much more broadly among the general public – and, if the revenue-raising system is reasonably progressive, the cost burden falls largely on those who can most afford to pay.

    Developmental aid policies, however, tend to be significantly more expensive than preferential selection policies, since they involve substantial resource transfers from government agencies or private institutions to the beneficiaries. Such organisations have limited budgets and many demands on their largesse; and there is only so much that they are able and willing to devote to help members of particular ethnic communities. From the perspective of any organisation seeking to provide such aid, it is far less costly to mandate preferential selection – and thereby to deflect most of the costs of the aid away from the organisation itself – than it is to transfer resources.

    In any case, preferential policies and resource-redistributional policies in favour of any given group are not mutually exclusive ways of reducing group disparities. First of all, a significant element of preference is involved in directing resource transfers to members of a specific group. Second, as I will argue below, a certain amount of resource transfer is likely to be very helpful

    – if not indispensable – to the success of any preferential policy favouring a disadvantaged group.

    Why Not Base Preferential Selection Policies on Characteristics Other Than Ethnicity?

    Critics of PD policies based on preferential selection frequently allude to the fact that such policies tend to favour the best-off members of under-represented beneficiary communities rather than the worst-off, since it is the former who are best placed to qualify – with a little help – for selection to top educational institutions, influential jobs, etc. To assure that PD policies help those who are most disadvantaged, many observers have proposed that preferential selection should be geared not to members of under-represented ethnic communities but to individuals of low socio-economic status (SES). Such class-based preferential selection policies, as compared with ethnicity-based preferential selection policies, would lead to the selection of (1) many more poorly-off and far fewer better-off applicants from underrepresented ethnic communities and (2) many poorly-off applicants from well-represented communities. The result would be much greater socio-economic diversity and much less ethnic diversity. There are several reasons for which this would be an undesirable outcome.

    First of all, for any given number of preferential selection beneficiaries, the number of beneficiaries from under-represented ethnic communities would obviously be smaller. Indeed, this number would be significantly smaller except in the (unlikely) event that there were relatively few members of other ethnic communities to be found in the nether reaches of the society’s socio-economic strata. Yet one of the central purposes of PD policies is to reduce disparities specifically between ethnic communities, independently of any reduction of socio-economic disparities. This is so because disparities among ethnic communities are especially likely to be attributed (with much factual basis) to past or present mistreatment of under-represented communities by other communities – a source of much continuing tension in heterogeneous societies. Moreover, members of ethnic communities are particularly vulnerable to damaging negative stereotyping, to stigmatisation,4 and to residential segregation.5 These kinds of phenomena tend to affect adversely even the bestoff members of under-represented ethnic communities.

    Second, class-based preferential selection is very likely to lead to significantly poorer performance on the part of its beneficiaries than is ethnicity-based preferential selection. Many of the benefits of PD policies depend on successful performance by its beneficiaries; likewise, many of the costs are correlated with rates of failure on the part of the beneficiaries. A good rate of beneficiary success is therefore essential to the overall success of a PD policy. In all societies, however, there is – regrettably but understandably

    – a high correlation between SES and preparedness to function well in educational institutions and job settings. By substituting as preferential selection beneficiaries low-SES members of all ethnic communities for higher-SES members of underrepresented communities, class-based preferential selection brings into demanding environments many more people who are bound to find it very challenging to perform well than do the prospective beneficiaries of ethnicity-based preferences. This means that a class-based PD policy will either require much more in the way of resources to assist challenged beneficiaries to be successful or – more likely – it will simply result in a much higher rate of failure.

    It is not only the need for substantially greater assistance to challenged beneficiaries that will tend to make a class-based PD policy a much more expensive proposition than a comparablysized ethnicity-based PD policy. Beneficiaries from low-SES families and (typically) low-SES neighbourhoods also need more financial support than do higher-SES beneficiaries in order simply to persist in educational institutions or job settings. It follows that, even if other things (including success rates) are equal, classbased preferences will be more costly than ethnicity-based preferences – and thus capable of being implemented only on a smaller scale.

    The above reasoning suggests that socio-economic disadvantage cannot and should not replace under-represented ethnic community identity as a basis for identifying appropriate beneficiaries of social policy. But the implementation of policies on behalf of under-represented communities does not preclude the simultaneous implementation of policies targeting the lower socio-economic classes. To the contrary, in multicultural societies it will generally be desirable to adopt some policies designed to address ethnicity-group inequalities along with other policies designed to address socio-economic class inequalities.

    Some observers have proposed that PD policies should remain focused on members of under-represented ethnic communities, but that a means test should be introduced in order to channel the benefits of preferential selection to relatively low-SES members of such communities.6 This proposal is on its face attractive, if only because much criticism of PD policies has focused on the way in which they tend to benefit primarily the better-off members of beneficiary ethnic communities. The wisdom of such a test, however, is highly questionable, for it would be subject to some of the same drawbacks as outright class-based preferences. First, it would remove from the ranks of PD beneficiaries precisely those members of under-represented ethnic communities who are likely to be best prepared to perform well in the positions to which PD helps them gain access. A means test would also remove from the ranks of PD beneficiaries precisely those who are likely to require less in resource aid in order to persist and succeed in their new environments.

    An additional weakness of class-based as compared with ethnicity-based preferential selection is that socio-economic status is more difficult to ascertain than ethnic identity. There are a variety of ways in which SES can be and has been measured; indicators such as income, wealth, and educational achievement are all arguably relevant. In determining the SES of any given individual, one must not only decide which indicators should be taken into account, and with what relative weights; one must also decide whether these indicators should be measured for the individual in question, for his/her parents, or for his/her grandparents. Having made these decisions, one is then faced with the daunting task of accurately measuring the extent of people’s income and wealth. The determination of a person’s ethnicity involves only one categorical variable that is generally unambiguous and applicable to the whole family. For some of the same reasons, moreover, SES is easier to alter – or to falsify – than ethnic identity.

    In recognition of the weaknesses of class-based preferential policies, some critics of ethnicity-based preferential policies have suggested the alternative of geography-based PD policies – providing preferences in selection to individuals living in disadvantaged localities (neighbourhoods, villages, towns or regions) or to individuals attending disadvantaged schools.7 The most appealing feature of such plans is that, while steering benefits to groups consisting disproportionately of the under-represented communities to whom ethnicity-based PD policies are oriented, they eschew direct reference to race, caste, or any other kind of ethnic characteristic. Geography-based PD policies are therefore less likely than ethnicity-based PD policies to exacerbate inter-community tensions and divisions.

    On closer examination, however, this alternative to ethnicitybased PD policies also has some very significant weaknesses – some of them much the same as the weaknesses of class-based PD policies. First of all, policies favouring people with particular geographic characteristics can obviously not do as good a job of increasing opportunities for members of under-represented ethnic communities, or members of low socio-economic classes, as policies directly targeted to the disadvantaged communities or classes. Second, the direct beneficiaries of geography-based preferences are likely to be less well prepared and hence less capable of performing well in their new and more challenging environments than those who would have been admitted under an ethnicity-based preference policy, because most of them come from relatively impoverished families who generally cannot provide their children with high quality education. Therefore the rate of failure of the beneficiaries of geography-based PD policies will be relatively high. Finally, geographic location and/or school attendance is relatively easy to alter, as compared to ethnicity. Thus geography-based preferences are likely to encourage people who are in no way disadvantaged to change their geographic location in order to become eligible for preferential selection – thus undermining the PD policy objective of increasing opportunities for members of under-represented groups.

    In sum, class- or geography-based preferences are inadequate substitutes for ethnicity-based preferential selection. Concern about socioeconomic inequalities in a society can best be addressed by policies oriented directly to the disadvantaged, rather than by changing the scope of eligibility for positive discrimination. Thus there is a strong case for adopting simultaneously ethnicity-based policies of preferential selection and class-based policies of resource transfer.

    III Optimal Structuring of Ethnicity-Based Preferences

    Sphere of Applicability of Positive Discrimination Policies

    The primary spheres in which ethnicity-based policies of positive discrimination have been applied in practice are: political representation (e g, reserved parliamentary seats); educational institutions (mainly at the level of higher education); employment (more often in public organisations than in private firms); and government contracting with private firms. Of these four spheres, political representation is the one for which the ‘prima facie’ case for positive discrimination is strongest. As Andre Beteille has emphasised,8 a central function of politicians is to represent their constituencies; and a political system operates more effectively if all relevant constituencies are indeed represented in relevant political bodies. Thus the potentially worrisome trade-off between adequate representation of disadvantaged groups, on the one hand, and efficient performance, on the other, is greatly muted in the political arena.

    Turning to the other three primary spheres of PD policies, the trade-off between representation and performance can potentially be quite salient. In this context, there are a number of reasons for which it would be most desirable to focus PD policies on access to educational opportunities and less desirable to focus them on access to jobs or government contracts. First of all, the difficulties that PD beneficiaries face in achieving good performance in a challenging setting tend to be less daunting when the beneficiaries are younger and have not yet suffered as much from the cumulative effects of group stigmatisation and socioeconomic disadvantage. The earlier in the life cycle that PD beneficiaries are asked to catch up with their less disadvantaged peers, the more important are their potential capabilities as opposed to their realised capabilities, and the greater their chances of success are likely to be.9

    Second, the negative consequences of poor performance in educational institutions are likely to be confined largely to the individual PD beneficiary; whereas in job or contracting settings poor performance can hurt other parties as well. To the extent that PD policies generate a conflict between justice/equity and merit/efficiency, this conflict is likely to be smaller in magnitude and easier to manage in an educational setting than in job or contracting environments. Indeed, PD policies in higher education provide a way of defusing the tension between representation and expertise in responsible positions, because attending more selective universities enables members of disadvantaged communities to gain the skills necessary to qualify for – and fulfill the responsibilities of – such high-status positions. In this way the advantages flowing from a more integrated professional elite can be gained without much loss in the competence of those filling the positions.

    Another reason to prefer PD policies in the educational sphere to such policies in the employment sphere is that passions and resentments about the actual and/or perceived unfairness of the selection processes involved are likely to run higher in the latter than the former. This is because the stakes for the individuals involved – whether they are preferred or displaced by PD – tend to be greater when it comes to a job as compared to admission to an educational institution (unless the latter serves as a perfectly straightforward channel to a good job). Thus the costs of positive discrimination associated with inter-group tension and conflict may be less serious in the case of educational admissions.10

    Everyone can agree that it would be desirable to find ways to enable under-represented community members to increase their numbers in high-quality educational institutions and in esteemed occupations by becoming more successful in a fully competitive selection process, rather than by receiving PD preferences. This has led some observers to propose that policies of “preferential positive discrimination” be replaced by policies of “developmental positive discrimination.” Developmental PD would encompass programmes to enable applicants from disadvantaged communities to perform better on standardised tests, whether for admission to educational institutions or for selection to jobs; and, more generally, programmes to improve the physical and social environments of the relatively poor neighborhoods and regions in which members of such communities typically live, so as to make these environments more conducive to their advancement.

    Certainly developmental PD, as defined in this way, is much to be desired. It cannot, however, be considered an alternative to preferential PD. On the one hand, developmental PD involves the channeling of resources disproportionately – if not exclusively – to members of certain under-represented communities, so it entails a significant element of preference. On the other hand, preferential PD often enables its beneficiaries to attain positions where they are likely to be better able to develop their skills and abilities, so it becomes partly developmental in character.11 Furthermore, a developmental programme typically calls for a greater commitment of resources than does a purely preferential PD policy. This is the reason why it has proven much more difficult to get developmental programmes going on an adequate scale than to enact preferential policies.

    In general, it is desirable to ensure that PD policies will serve a developmental function, not merely one of preferential selection. Except in circumstances where the selection process identifies under-credentialed applicants who are actually better prepared to do the work, the PD beneficiary will face the challenge of catching up with peers selected without any preference. His/her chances of success will depend significantly on the extent to which the new position that PD has enabled the beneficiary to attain has a development component – one that will help the beneficiary to develop his/her capabilities and overcome the initial disadvantage that he/she faces. This seems more likely to be the case when PD is applied to the selection of applicants for an educational programme than for employment. But within the latter sphere, PD policies can be expected to work better, the more significant is the role of on-the-job training.

    Another consideration relevant to the choice of sphere for positive discrimination is the degree to which the institution involved is elite as opposed to run-of-the-mill in character. Since integration of the societal elite is a principle objective of PD policies, the net benefits from a PD policy are likely to be especially high in the case of elite institutions – whether these are schools from which graduates move directly into high-status professions, or major enterprises with a high degree of economic or political power. Moreover, such institutions are bound to be relatively well-endowed, so they will be correspondingly better able to mount programmes that can increase the prospects of PD beneficiary success.

    A final important question with respect to the sphere for positive discrimination is whether or not there is reason to differentiate between the public and the private sector in applying PD policies. In principle, there is no reason to expect that the benefits and costs of positive discrimination will be any different (ceteris paribus), whether the institutions, organisations or enterprises involved happen to be in the public or the private sector. The same arguments and considerations affecting the net benefits realisable from a PD policy apply irrespective of the sector. Thus the case for focusing PD policies on educational institutions in preference to job settings, on programmes with a developmental emphasis, and on relatively elite institutions, remains strong – whether the institutions are public or private.

    In practice, one must recognise that society-specific historical, constitutional, and legal traditions will strongly affect whether PD policies are likely to emanate from governmental or nongovernmental decision-makers. Such factors will also influence also how well PD policies can be implemented and enforced, and what role government should be expected to play in this regard. One can therefore adequately address the issue of public and/or private sector PD policies only in the context of particular societies and historical situations.

    Choice of Beneficiary Communities Eligible for Preferences

    The desirability of pursuing a PD policy will vary from one under-represented ethnic community to another. Indeed, the identification of communities to be favoured by preferential selection processes is the most critical decision to be made in adopting policies of positive discrimination. Since integration of the societal elite is arguably the most important objective of PD policies, the eligibility of any particular ethnic community for PD should be positively related to the degree to which the community is in fact under-represented in society’s most desirable positions. This is likely to be highly correlated with the extent to which the average socio-economic status of community members is below the societal average. But relatively low average socioeconomic status does not alone justify positive discrimination on the basis of ethnic identity. An ethnicity-based PD policy is warranted to the extent that members of a disadvantaged community have been – and continue to be – mistreated and stigmatised on the basis of their group identity.

    In principle, one should try to estimate the overall net benefits associated with each distinct community for whom a case for PD seems plausible on the above grounds, in order to determine which ones should be made eligible. In practice, it would make sense to limit such an effort to communities whose members have a strong sense of common ethnic identity, so that many will be able to share – materially as well as psychically – in the benefits that will go in the first instance to a relatively advantaged few. Unless there is considerable solidarity among community members, the immediate beneficiaries of PD policies will not be inclined to use their enhanced positions in ways that expand opportunities for many less favoured members of the community, nor will the former serve as compelling role models for the latter.

    Deciding which communities should be eligible for PD benefits is likely to be an especially difficult task when a prospectively eligible community contains distinct sub-groups, for this gives rise to the question: should all of these subgroups, or only some of them, be eligible?12 A related question is whether different communities eligible for PD policies in any given country should be accorded the same amount and kind of benefits. The optimal magnitude of preference is likely to differ from one community to another, so one should in principle establish different preference magnitudes for different eligible communities.13 I will discuss the issue of multiple preference magnitudes further on.

    Determining on a rational basis which under-represented ethnic communities should be eligible for positive discrimination is, to be sure, a great challenge. In principle, it should be done on the basis of meticulous and dispassionate social scientific research into the extent of the net benefits of PD policies applied to potentially eligible communities.14 Given the political stakes in the outcome of such research, it is bound to be very difficult to keep politics from interfering with the process.15 In advocating reconsideration of the eligibility of communities currently favoured by PD, one must recognise that there is an asymmetry between the adoption and removal of a PD policy. The removal of PD eligibility from a community is a political blow of considerably greater magnitude than the failure to grant such eligibility in the first place. Thus the removal of eligibility from some communities or community subgroups, while others continue to be eligible, must be limited to situations where the case for doing so is strongly supported by widely respected research findings.

    Given all the difficulties and complications associated with the determination of which under-represented ethnic communities should be eligible for PD, there is a very good case for simplicity in several respects: (1) simplicity in limiting the number of eligible communities to the most compelling cases; (2) simplicity in defining communities by simple criteria that clearly distinguish those subgroups and individuals who are members of the community from those who are not; and (3) simplicity in grouping together into one community subgroups that might have an argument for being considered separate communities – unless that argument is extremely compelling.

    Configuration of Positive Discrimination Policies

    As noted earlier, the overall net benefits of positive discrimination policies are significantly and positively correlated with the (average) quality of performance by beneficiaries in the institutions and organisations to which they gain preferential access. One can therefore view the detailed configuration of a PD policy as confronting a type I/type II error problem. In selecting formally underqualified but potentially successful applicants from under-represented beneficiary communities, one hopes to minimise errors of two types:

    Type I: failure to select someone who would have been successful (a true null hypothesis is incorrectly rejected).

    Type II: actually selecting someone who will be unsuccessful (a false null hypothesis fails to be rejected).

    Among the key elements of choice in configuring PD policies are the magnitude of the preference given to beneficiary groups, the sensitivity of the selection process, the identifiability of the PD beneficiaries, and the extent of support for PD beneficiaries. After addressing briefly the issue of quotas versus preferences, I will examine each of the above four PD policy elements with a view to determining how a PD policy can be best designed to minimise errors of type I and II and thus to maximise net benefits.

    Quotas vs Preferences

    Whether a PD policy is based on a system of quotas or a system of preferences is often assumed to be its most important defining characteristic. A quota system involves the establishment of a certain number of positions reserved for applicants from beneficiary communities, thus dividing the overall competition for selection into two or more separate competitions: one for members of each beneficiary community and the other open to all other applicants. Whatever the criteria for ranking applicants, these criteria are applied separately to each group of applicants; and the highest-ranking applicants in each group are selected until the available positions are filled. The principal alternative to such a quota system is one in which there is a single general competition for selection, but PD beneficiary communities are given preferences – more precisely, preferential boosts – in the form of more favourable consideration in the determination of the ranking of candidates. In the case of quantitative selection procedures, in which an applicant’s qualifications are summarised in an overall point score to determine his/her position in the rank order, the preferential boost could take the form of a certain number of additional points credited to a beneficiary community’s applicants. In the case of qualitative selection procedures, in which a variety of applicant qualifications are taken into account but not formally aggregated into a single overall point score, the preferential boost would take a less precise form – for example, applications from members of relevant communities could be viewed in a rosier light or given extra credit for signs of unrealised potential.

    The difference between a quota system and a preferential-boost system is not as great as it may first appear. Corresponding to a quota system that selects any given number of beneficiary community applicants for a particular position, there is bound to be some amount of preferential boost that leads to the same outcome. In the case of a selection process in which applicants’ qualifications are summarised in a single point score, the amount of preferential boost that would do so is the number of points needed to bring the marginal beneficiary community applicant’s score up to the level that would make him/her the last applicant admitted in the general competition. In the case of a qualitative selection process, the amount of the preferential boost would be more difficult to calculate; but the same principle holds.

    In practice, quota systems are most often constrained by specification of minimum conventional qualifications (e g, a minimum qualifying score) below which beneficiary community applicants will be rejected, even if their quota is not filled. Whenever such a minimum conventional qualifications requirement serves to keep the number of selected applicants below or equal to the quota, this kind of constrained-quota system has the same effect as a preferential-boost system in which the size of the preferential boost is equal to the gap between the minimum conventional qualifications required of a successful applicant in the limited competition and the conventional qualifications of the last applicant admitted in the general competition.

    An unconstrained-quota system, by focusing on a target number of beneficiary community applicants to be selected, ignores problems likely to arise if the conventional qualifications gap between PD beneficiaries and other applicants selected is substantial. To try to fill quotas of reserved seats with applicants from disadvantaged communities, irrespective of their conventional qualifications, is to invite poor performance by those selectees whose conventional qualifications are well below the qualifications required of applicants in a general competition. This is especially likely to be the case if the size of the quota is determined by the community fraction of the overall population, as opposed to the community fraction of a smaller population consisting of plausible recruits to the positions in question. An unconstrained-quota system thus makes little sense.

    Since a constrained-quota system does not differ in its essence from a preferential-boost system, the choice between the two is immaterial for the success or failure of a PD policy.16 Several other characteristics of the selection system employed by a PD policy are likely to have much more significant effects on the consequences of the policy.

    Magnitude of the Preference

    An obviously important dimension of any PD policy is the extent to which it intervenes into and perturbs the allocation of available positions between beneficiaries and other applicants, as compared with the allocation that would obtain in the absence of the policy. A useful way to measure the impact of a PD policy focuses on the magnitude of the preference extended to PD beneficiaries. This magnitude is easy to measure for selection procedures that are quantitative, in the sense that applicants’ qualifications are summarised in an overall point score. In the case of a quantitative preferential-boost system, it is simply the number of additional points granted to beneficiary community members. In the case of a quantitative constrained-quota system, it is the difference between the point score of the marginal applicant selected in the general competition and that of the marginal applicant selected in the quota competition. When a PD selection procedure is qualitative, involving consideration of a variety of qualifications that are not scored and aggregated into a single overall point score for each applicant, the average amount of preference extended to PD beneficiaries is implicit in the process and much harder – but not impossible – to measure.17

    The preference magnitude is an especially important characteristic of a PD policy because of its impact on type I and II errors and hence on the net benefits of the policy. The smaller is the magnitude, the fewer will be the number of beneficiary community applicants selected (thus more type I errors), but the greater will be the proportion of them who are likely to perform as well as their non-beneficiary peers (thus fewer type II errors). The larger the magnitude, the greater will be the number of PD beneficiaries (thus fewer type I errors), but the smaller will be the proportion who perform as well as their peers (thus more type II errors). There is bound to be some level of the preference magnitude at which the costs of marginal unsuccessful candidates selected will start to exceed the benefits of marginal successful ones. Where the optimal preference magnitude lies depends on the relative frequency of each type of error at each level and on the costs associated with each type of error.18

    In principle, with full information about the consequences of different preference magnitudes, and with a common standard for evaluating the net benefits from a good and a bad selection decision, one could estimate both the number of PD beneficiaries and the overall net benefits associated with each preference magnitude. For any given under-represented community the net benefits (ceteris paribus) from PD would presumably rise initially, as the magnitude of the preference was raised from zero, because at low magnitudes PD beneficiaries could be expected to perform almost as well as other applicants. After a certain point, however, the additional net benefits from a higher preference magnitude would turn negative because, at ever higher magnitudes, an ever smaller proportion of additional PD beneficiaries selected would be able to perform well.19 The magnitude of the preference at that turning point could therefore be identified as the optimal one, which maximises the expected net benefits from a PD policy favouring the given community. PD policies would then be judged worthy of adoption in favour of each community for which the maximum expected net benefits are in fact positive, with the magnitude of the PD preference set at the optimal level for each eligible community.

    In reality, of course, decision-makers will never have access to sufficient information to determine in such a systematic and precise manner the optimal preference magnitudes and the corresponding expected net benefits for each community under consideration for PD-eligibility. Instead, they will have to mix available information with educated guesses and rely on their own best judgment to determine which communities should best be made PD-eligible and at what level the preference magnitude for each eligible community should be set. In the interests of simplicity (as advocated earlier), the preference magnitudes should be set at the same level unless there is a very compelling reason to do otherwise for a particular community.

    Sensitivity of the Selection Process

    PD policies vary greatly with respect to the sensitivity – as opposed to the rigidity – of the process whereby applicants are preferentially selected. The most rigid, mechanical type of process involves a quantitative procedure in which all applicants take some kind of standardised test and are ranked simply by scores on that test. In a quota system, the test rankings are used to select applicants in separate competitions; in a preferentialboost system, a certain number of points are added to beneficiary applicant scores before all candidates are ranked for selection in a single competition. A somewhat less rigid selection process would take account of several different qualification criteria, not just a test score. This kind of process would assign scores on each criterion to every applicant and then aggregate every applicant’s scores into a composite quantitative index for purposes of ranking, prior to the selection of beneficiary applicants via a quota or preferential-boost system.

    A selection process becomes progressively more sensitive the greater is the variety as well as the number of criteria involved in ranking applicants. Even more important, a selection process becomes more sensitive the more the process of evaluating an applicant’s standing with respect to relevant criteria is a qualitative rather than a quantitative one, involving considered judgment by selection personnel rather than mechanically determined scores fed into a composite index. A highly sensitive and nuanced PD selection process would not only include qualitative evaluation of the extent to which a beneficiary community applicant satisfies various relevant criteria, but also treat disadvantaged community status as a signal to look especially hard for evidence of additional applicant characteristics suggesting a strong potential for good performance.20

    It is clearly desirable to make the PD selection processes as sensitive and nuanced as possible, so as to maximise the potential for determining which community applicants – who are underqualified in terms of conventional indicators – have the greatest potential to be successful, and thereby to minimise both type I and type II errors. This implies, first of all, that one should not try to fill quotas of reserved seats with under-represented community applicants, irrespective of their conventional qualifications; rather, one should give each such applicant the same magnitude of preference in competition with non-community applicants. This implies, further, that one should make use of a variety of relevant qualification criteria, and that evaluations of the extent to which an individual applicant is qualified should be based as much as possible on the exercise of qualitative judgment by selection decision-makers rather than on the input of mechanistically determined scores into a quantitative composite overall index.21

    Sensitive selection processes are, to be sure, harder and more costly to administer than rigid ones, since they require that more information of various kinds be gathered from applicants and that more people be employed to implement the selection process. The extent to which a PD selection process can be made sensitive therefore depends on the availability of resources to finance the process; and the costs of raising such resources must be weighed against the benefits expected from a more sensitive process. There is a difficult decision to be made as to how far to invest limited resources into processes of selection; but it is a safe guess that such investment is usually not carried as far as it should be, since some of the benefits of better selection will not accrue to the organisation doing the selection. Government subsidies may therefore be warranted to achieve the desired end.

    Identifiability of the Beneficiaries

    The existence of a PD policy means, of course, that some individuals selected in a competition will owe their selection to that policy. Depending on how the PD policy is implemented, however, the identity of the beneficiaries – as well as their total number – may or may not be known to anyone other than those administering the selection process (and sometimes not even to them). This is by no means a trivial matter, because both the self-esteem of a beneficiary and his/her treatment by others is likely to be adversely affected by knowledge that he/she would not have been selected in the absence of the PD policy. Thus one characteristic of a PD policy relevant for its consequences is the extent to which PD beneficiaries are made identifiable.

    One might at first presume that a quota system necessarily makes PD beneficiaries identifiable and a preferential-boost system does not. The fact that a quota system channels members of certain communities into separate competitions for reserved positions, however, does not require that those chosen in each competition be separately identified. And even when the number of positions reserved for a particular community is known, that does not mean that an equivalent number of community members owe their positions to PD; for it is always possible that some of the successful applicants did not need PD to be selected and/ or that some of the reserved positions went unfilled. Conversely, the fact that a PD policy is based upon a preferential boost does not always preclude the selection administrators from revealing who required such a boost in order to be selected. Under a preferential-boost system it is not always clear precisely which selected beneficiaries owe their selection to the PD policy, so identification of PD beneficiaries is less likely to occur under such a system than under a quota system. But whether or not it occurs under any selection process is really a policy variable.

    It is clearly best that a PD policy not render its beneficiaries easily identifiable as such – and hence stigmatised – by others; this reduces the prospect of type II errors. One of the most common ways in which PD beneficiaries are in practice rendered identifiable is by including them in remedial and developmental programmes designed to facilitate their adjustment to the demands of the competitive environment to which they have gained access. The objective of such programmes is certainly a worthy one (as I discuss just below). But to avoid the negative fallout of making PD beneficiaries easily identifiable, it is desirable – wherever possible – not to confine these programmes exclusively to PD beneficiaries but to include in them also non-PD beneficiaries who are likely to benefit from them.

    Extent of Support for under-prepared Beneficiaries

    Whether a PD beneficiary is able to meet the challenges of the position to which PD has provided access is likely to depend significantly and positively on the extent to which support is made available after the person is selected; such support thus reduces type II errors. Policies of positive discrimination may be confined to the selection process itself (costing no more than the administrative expenses of that process), or they can involve various forms of subsequent support for PD beneficiaries (costing a good deal more). There are several kinds of support at the organisational or institutional level that can be helpful for PD beneficiaries, whether in an educational or a workplace setting. These include both human resources, such as favourable attitudes on the part of supervisors and mentoring on the part of colleagues, and financial resources made available for programmes and activities that help PD beneficiaries adjust to their new settings and work productively in them. Resources of these kinds will no doubt be most plentifully available in well-endowed elite institutions. The provision of such support to PD beneficiaries either requires that they be publicly identified as beneficiaries or that the support be extended to a broader group, including non-beneficiaries as well, which is correspondingly more costly.

    Just as it takes resources to improve the sensitivity of selection processes, so it takes resources to provide developmental support for underprepared beneficiaries. Decisions about how much to invest in support of PD beneficiaries will therefore have a significant impact on the overall net benefits of the PD policies. Resources are always limited, and the payoffs to this kind of investment are not always obvious – nor are they all captured by the investors. Here too there will probably be a need for subsidies to encourage organisations practising PD to make sufficient investments of this kind.

    I have argued above that the prospective success of a PD policy can be significantly enhanced by financial resource commitments to selection programmes, which increase the ability of select promising beneficiaries, and to developmental programmes, which improve the performance of beneficiaries. There is one further kind of financial resource commitment that can make a significant contribution to the success of a PD policy: direct financial aid to needy beneficiaries. Student PD beneficiaries may not be able to afford all of the (unsubsidised) expenses associated with attending an educational institution, so they may require financial aid just to enroll. Moreover, for some PD beneficiaries, economic insecurity is a potentially significant source of poor performance. Students from socio-economically disadvantaged families are much more likely to have dropped out of an educational programme

    – temporarily or permanently – in order to help their families by contributing their labour to work within the family or by earning additional income. Employees from socioeconomically disadvantaged families are more likely to absent themselves from work – if not to quit their job – in order to address family crises. Financial aid to PD beneficiaries who are students, and loans to those who are employees, can help to prevent such problems from compromising the success of PD policies.

    This latter context is one where a means test is perfectly appropriate and wholly desirable. As noted above, excluding well-to-do PD beneficiaries from access to PD preferences would mean foregoing some important benefits of PD policies. It would be perfectly appropriate, however, to exclude well-to-do PD beneficiaries from financial aid programmes designed to enable socio-economically disadvantaged PD beneficiaries to remain active in the positions to which PD has given them access.

    IV Conclusion

    Are policies of positive discrimination a good way to enhance the well-being of disadvantaged and under-represented ethnic communities? PD policies have the potential to generate significant benefits – perhaps most importantly in facilitating greater integration of beneficiary communities into society’s elite, which can contribute to a more vital democracy and more effective societal institutions as well as greater equity across communities. On the other hand, PD policies also have the potential to give rise to significant costs – such as poorer performance by individuals in key positions and heightened tensions between ethnic communities.

    In this paper I have argued that the likelihood that a PD policy will generate net benefits depends on a host of factors, including the nature and circumstances of the prospective beneficiary community, the sphere(s) to which the policy is applied, and the way in which the policy is configured. Key elements of PD policy configuration include the magnitude of the preference afforded to members of a beneficiary community and the extent of support provided for underprepared beneficiaries.

    A critical determinant of the success of a PD policy is the ability of its beneficiaries to perform well, once selected preferentially to a desired position. An inherent obstacle to the success of PD policies is that the likelihood of good performance by PD beneficiaries tends to vary inversely with the strength of the need for giving preference to a particular under-represented ethnic group – since most members of groups with a strong case for positive discrimination will ipso facto be poorly prepared to compete with and perform as well as members of more advantaged groups (except perhaps in the arena of representative politics). This puts a premium on the capacity of PD policies to identify the most promising applicants among members of underrepresented ethnic communities and to ensure support for PD beneficiaries in the positions that they are enabled to attain.

    A PD policy is more likely to be successful in generating net benefits the greater its conformity with the following conditions:

  • the beneficiary community is fairly homogeneous; its members have been and continue to be subject to mistreatment and stigmatisation by other communities; and they are significantly under-represented in esteemed positions in society;
  • the PD policy is applied to a sphere of activity that has a significant developmental component, so that beneficiaries can acquire human and social capital that will significantly enhance their abilities and opportunities;
  • the PD policy applies to high-quality and relatively wellendowed institutions and organisations that provide access to the upper strata of society;
  • the PD policy is applied to a sphere in which the quality of performance of a beneficiary affects mainly the individual beneficiary and not other parties;
  • the magnitude of the PD preference granted to beneficiaries is not so large as to make it difficult for them to succeed in the environment to which they gain preferential access;
  • the process of selection of PD beneficiaries is sensitive, nuanced, and capable – at least to some extent –of identifying those formally underqualified applicants who are most likely to succeed if given the opportunity;
  • the PD policy does not render its individual beneficiaries clearly identifiable as different from all of their peers;
  • the PD beneficiaries are afforded significant human and financial support after being preferentially selected.
  • The greater the extent to which a PD policy is structured to meet the above desiderata, the more likely it will prove to be a good way of enhancing the well-being of disadvantaged and under-represented ethnic communities.

    Finally, a good policy of positive discrimination is one that is ultimately time-limited. To the extent that it is successful, a PD policy will reduce the degree of a disadvantaged community’s under-representation that provided the rationale for preferential selection of its members in the first place. The conditions giving rise to the need for a PD policy, however, are of long duration and very deep-seated. One cannot therefore expect that the need for such a policy will disappear within just a generation or two; it is more likely a matter of at least a century. While it remains in effect, there is every reason to revisit a PD policy periodically and to adjust it so that it functions effectively under ever-changing circumstances.

    EPW

    Email: tomw@umich.edu

    Notes

    [An earlier version of the paper was presented to the conference on ‘Religiousand Social Fragmentation and Economic Development in South Asia’ held at Cornell University in October 2005. I am grateful to Robert Travers, Neema Kudva, Farhana Ibrahim and other conference participants for their helpful comments.]

    1 I will use the general term “ethnic community” to refer to a group of people defined in terms of their commonality of such characteristics as race, caste, tribe, religion or native language, which are for the most part ascribed rather than voluntarily chosen, determined at birth and rarely alterable. It is the fact that such groups are under-represented in the positions to which they are given preferred access that makes the discrimination in their favour “positive”.

    2 See Weisskopf (2004), chapter 3, for a comprehensive discussion of the arguments for and against PD policies.

    3 A successful anti-discrimination policy is in any case likely to lead to implicit preferences for members of the community being protected. As Fryer and Loury (2005: 149) have shown in a highly plausible theoretical analysis, “given a legal environment that eschews affirmative action and requires only non-discrimination, employers will nevertheless behave as if they faced an “implicit quota”; that is, they will adhere to a self-imposed hiring target that can be understood as their equilibrium response to incentives created by imperfect auditing.”

    4 See Loury (2002). 5 See Anderson (2002). 6 See, for example, Patwardhan and Palshikar (1992); they recommend

    (p 108) that “As individuals climb several rungs of the social ladder with the assistance of reservation and other facilities and become part of the educated upper middle class, their children should be deemed ineligible for such benefits as they are no longer socially handicapped”. Indeed, when the Supreme Court of India in 1992 gave its approval to nationallevel public sector job reservations in favour of other backward classes (as recommended by the Mandal Commission), it imposed a “creamy layer test” of individual eligibility, in order to eliminate persons from affluent or professional families; see Sivaramayya (1996).

    7 A prominent example of a geography-based PD policy is the “10 per cent plan” adopted by the University of Texas (UT) after its race-based preferences in admissions were ruled unconstitutional by a US federal appeals court. Under this plan, UT is committed to admitting any student whose grade-point average is among the top 10 per cent of his/her high school graduating class (see Montejano 1998). Because of the typically high degree of racial residential segregation in the US African American and Hispanic American students tend to be concentrated in particular localities where they constitute the vast majority of the graduating class in local high schools. The 10 per cent plan therefore assures that a significantly higher proportion of such minority groups are admitted to UT than would otherwise have been the case.

    8 See, for example, Beteille (1981).

    9 This argument implies also that it is desirable to apply PD policies in educational admissions at the lowest possible level. There is, however, a powerful argument going the other way: higher educational institutions are critically important in training students for responsible jobs and in providing the social capital that leads to good job opportunities. Moreover, as a practical matter, it is likely to be more difficult to arrange for younger than older children from disadvantaged ethnic communities to attend firstrate schools, since such schools are likely to be located far from the children’s homes.

    10 There is one consideration that could put a premium on positive discrimination preferences in the sphere of employment. Some important prospective benefits of PD – e g, the integration of the societal elite – require for their realisation that members of disadvantaged communities occupy high-status positions. If there is any doubt that qualified members of these communities will be hired into and/or promoted up to high-status positions, then positive discrimination policies may be needed at the employment stage of the life cycle.

    11 Indeed, preferential PD in the sphere of higher education serves as a form of developmental PD with respect to the sphere of employment, for it enables community members (who would otherwise not have had this opportunity) to gain skills that make them more competitive in the job market.

    12 In the Indian context, this question is especially relevant to the determination of which “other backward classes” should be eligible for reservations.

    13 The same logic would suggest that distinct sub-groups within an eligible community should each have their own preference magnitude. There are good reasons, however, for maintaining a uniform magnitude of preference for all sub-groups within an eligible community. As a practical matter, it will be very difficult to estimate optimal preference magnitudes with any precision for different sub-groups. Moreover, the capture by a strong sub-group of the lion’s share of PD benefits should not be a cause for concern, if the community has been defined in a way that groups together people with a genuinely common ethnic identity. This is because most of the benefits anticipated from a PD policy have to do with better representation of a community in a set of esteemed positions; and from this perspective what matters is that those community members with the best chance of succeeding in such positions get that opportunity, and that other community members have reason to identify with them. If members of one community sub-group do not in fact feel that they are well represented by members of another sub-group from the same community, then the net benefits from a PD policy favouring that community will be significantly diminished; and one should reconsider whether the two sub-groups belong in the same community – or, indeed, whether they belong in any eligible community.

    14 See Cunningham et al (2002) for similar proposals.

    15 The history of positive discrimination in India makes abundantly clear what a strong role politics has played in the identification of PD-eligible groups to date. In the 1970s India’s Mandal Commission was mandated to undertake precisely the kind of research advocated here; the nature of its work, and the process whereby many of its recommendations were adopted, provides a sobering reminder of how difficult it is to keep political considerations at bay when eligibility for PD is at issue [Radhakrishnan 1996].

    16 If and when the preferential boost implied by a minimum qualifications requirement in a constrained-quota system is not binding – i e, if it alone would allow more applicants to be selected than the size of the quota

    – then there is of course a difference in outcome as between the constrainedquota system and the corresponding preferential-boost system.

    17 In such situations one can only try to estimate the magnitude of the preference by ex post facto construction of a numerical scoring framework within which relevant applicant qualifications can be quantified and aggregated into a single overall attributed point score for each applicant. (An applicant’s standardised test score alone could serve as a simple, if crude, measure of such a point score.) The magnitude of the preference extended to PD beneficiaries can then be approximated by the difference between the average attributed point score of the lowest-scoring subset (say 20 per cent) of beneficiary community applicants selected and that of the lowest-scoring subset (of the same number) of general applicants selected. The reason for choosing subsets of applicants rather than marginal individual applicants is that, in the context of a qualitative PD selection procedure, there is bound to be a good deal of overlap between the bottom ends of the distributions of attributed point scores for the two sets of applicants.

    18 See Chatterjee (1983) for a rigorous decision-theoretic analysis of preferential selection policies, which addresses precisely the question of how to determine the optimal magnitude of the preference.

    19 One cannot rule out a priori the possibility that a PD policy – by recognising hidden capabilities of community applicants – will actually improve the accuracy of applicant assessment. In this case PD beneficiaries will, on average, be capable of performing as well as non-community peers with stronger conventional qualifications. It stands to reason, however, that even in this case there will be some magnitude of preference beyond which the performance level of PD beneficiaries will fall below that of peers selected without any such preference.

    20 Identifying high potential in low-credentialed applicants is a difficult but critical challenge for any PD policy. To achieve this objective it may be helpful to look at the extent to which an applicant has overcome previous disadvantages and to give special weight not to “absolute” but to “relative” test scores, i e, test scores residualised by controlling for background variables reflecting family, neighbourhood or community socio-economic status. Careful analysis of past selection decisions and outcomes may also help, by shedding light on what factors correlate with success of lowcredentialed applicants.

    21 The superiority of sensitive over mechanical PD selection processes was at the heart of the US Supreme Court 2003 rulings on the University of Michigan affirmative action cases, in which the court upheld the law school’s qualitative and individualised admissions procedures but struck down the undergraduate point-based procedures (see Gratz et al vs Bollinger et al, 529 US 2003, and Grutter vs Bollinger et al, 539 US 2003).

    References

    Anderson, E A (2002): ‘Integration, Affirmative Action and Strict Scrutiny’, NYU Law Review, 77, pp 1195-1271.

    Beteille, A (1981): ‘The Problem’, Seminar, 268, pp 10-13.

    Chatterjee, B B (1983): ‘Social Costs of Reservation in Higher Education: A Decision Theoretic View’, Journal of Higher Education, 9, pp 77-89.

    Cunningham, C D, G C Loury and J D Skrentny (2002): ‘Passing Strict Scrutiny: Using Social Science to Design Affirmative Action Programmes’, Georgetown Law Journal, 90, pp 835-82.

    Fryer, R G and G Loury (2005): ‘Affirmative Action and Its Mythology’, Journal of Economic Perspectives, 19-3, pp 147-62.

    Loury, G C (2002): The Anatomy of Racial Inequality, Harvard University Press, Cambridge, MA.

    Montejano, D (1998). ‘Maintaining Diversity at the University of Texas’ in R Post and M Rogin (eds), Race and Representation: Affirmative Action, Zone Books, New York, pp 359-69.

    Patwardhan, V and V Palshikar (1992): ‘Reserved Seats in Medical Education: A Study’, Journal of Education and Social Change, 5, pp 1-117.

    Radhakrishnan, P (1996): ‘Mandal Commission Report: A Sociological Critique’ in M N Srinivas (ed), Caste: Its Twentieth Century Avatar, Viking Press, New Delhi.

    Sivaramayya, B (1996): ‘The Mandal Judgment: A Brief Description and Critique’ in M N Srinivas (ed), Caste: Its Twentieth Century Avatar, Viking Press, New Delhi.

    Weisskopf, T E ( 2004): Affirmative Action in the United States and India: A Comparative Perspective, Routledge, London, printed and distributed in South Asia by Foundation Books.

    To read the full text Login


    To know more about our subscription offers Click Here.

    Comments

    (-) Hide

    EPW looks forward to your comments. Please note that comments are moderated as per our comments policy. They may take some time to appear. A comment, if suitable, may be selected for publication in the Letters pages of EPW.

    Back to Top