The Burden of Caste on Social Identity in India

Abstract

This paper uses the World Values Survey to investigate the determinants of perceived social status in India. Caste is still the largest determinant, yet not the only one, as income, education and occupation are all relevant factors. However, only unlikely improvements in those economic attributes could offset the burden of being from a low caste or tribe on perceived social rank. This study is part of the literature that shows how the internalisation of prejudice and long-lasting discrimination may have impaired individuals’ self-esteem. The results stress the need to account for self-depreciation when assessing the efficiency of affirmative action policies.

Acknowledgements

I am grateful to Ashwini Despande, Samuel Bowles and Jean-Claude Berthélémy for useful comments. I also thank Richard Palmer-Jones, as well as two anonymous referees, who helped improve the article with useful suggestions.

Notes

1. See, for instance Benjamin, Choi, and Strickland (2010), Benjamin, Choi, and Geoffrey (2010), Croizet and Claire (1998), Dee (2009) and Spencer, Steele, and Quinn (1999). For a more extensive review of these experiments, please refer to Akerlof and Kranton (2010), Chen and Li (2009) or Steele (1997).

2. Data and code are available upon request.

3. Constitution of India; Directive Principles of State Policy; Article 46.

4. Other rounds of the survey are not workable, since information about caste is either unavailable or the sample is largely unbalanced.

5. In theory, caste does not exist outside Hinduism, while in practice it does. Moreover, caste is fixed by birth among Hindus, while many non-Hindus in India converted from Hinduism in order to escape low-caste status. Whether a convert succeeds in changing status may depend on his class. Thus, the kind of analysis undertaken here, in which caste is fixed, is difficult for non-Hindus.

6. According to the 2001 Census of India, SCs represent 16 per cent of the population and STs 8 per cent.

7. It is important to stress that this question was translated into thirteen local languages and that the connotation associated with the classes’ labels may be quite different from what is commonly understood in English. It is very likely that the classes are perceived by the interviewees as some sort of ranking on a scale from 1 to 5 and the labels devoid of the usual sociological meaning, as well as the word ‘class’. Unfortunately, further information on this particular topic is lacking. If the assumption according to which individuals felt they were asked to rank themselves, irrespective of the actual meaning of the classes is correct, then this leaves room for subjectivity. This issue is further addressed in Section 5.1. In the article, I stress the fact that what is dealt with is perceived social class, this latter word being most probably understood as a rank rather than a class in a Marxist sense.

8. Table A1 in the Online Appendix presents the distribution of the assumed main determinants, besides castes, of social classes across castes.

9. Educational outcomes in the survey sample and in the 2001 Census of India have been compared and, generally speaking, they are quite similar.

10. Results from these tests are available upon request.

11. Table A2 in the Online Appendix presents the distribution of income across occupations and clarifies occupational ranking. Being employed as an agricultural worker is the lowest earning category.

12. The household is defined as urban if it is located in a town with more than 5,000 inhabitants. This is the criteria retained by the Census of India.

13. Estimates are available upon request.

14. χ² test of difference in parameters are available upon request.

15. Source: Central Statistic Organization, Government of India.

16. The exact number is 87,633 rupees a year. Four million individuals fall into the top percentile.

17. For instance, 80 per cent of the individuals who did not achieve middle school earn less than 3,000 rupees per month, while only 35 per cent of those who achieved middle school do.

18. Table A3 in the Online Appendix sets out the occupational distribution across education levels

19. Marginal probabilities show that lower-middle class is the threshold as likelihoods to rank in this class and higher turn negative for SCs and STs. Figure 2 is based on estimates and marginal probabilities from a probit model that sets as success self-ranking into either lower- or upper-middle class or upper class. Coefficients and marginal probabilities for this model are set out in Table A4 in the Online Appendix.

20. Unfortunately, observations for STs are too scarce (94) to perform an analysis that goes beyond the scheduled nonscheduled partition.

21. The happiness variable is ordinal and is maximised when the interviewee declared plain unhappiness, which explains the coefficient’s negative sign.

22. Results were estimated, including a set of dummies for the language used. Results are little changed in this specification compared to the basic one. They are available upon request. Moreover, given that states were defined in India according to linguistic criteria, such a specification is very close to one that would include state fixed effects, which is presented in Section 5.2.

23. The North dummy is equal to one if the most widely spoken language belonging to the Indo-European family. Languages of the southern states belong to the Dravidian family.

24. Source: 2001 Census of India.