Caste Discrimination in Provision of Public Schools in Rural India

Abstract

This paper aims to understand discrimination in the provision of public schools in rural India and how it affects the educational outcome of social groups. Using census data, we find that villages with a higher share of marginalised castes viz. Scheduled Castes (SC) and Scheduled Tribes (ST) have a lower probability of having public schools. The negative relationship is non-monotonic as the marginal probabilities get weaker beyond a threshold level of SC/ST share. Though the Sarva Shiksha Abhiyan (Education for All) programme, mainly intended to expand elementary education, has reduced the gaps in the provision of primary schools, the extent of discrimination has increased at the secondary level. A strong association between public schools and educational outcomes highlights the importance of public schools. Finally, we show that the caste-based provisioning of public schools partially explains the disparity in the educational outcomes across social groups.

Acknowledgements

The authors would like to thank Katharina Michaelowa, Rohini Somanathan, Anirban Mukherjee, Debasis Mondal, Abhiroop Mukhopadhyay, Reetika Khera, Jayan Jose Thomas, B. Satheesha, and participants in various conferences and seminars at the University of Zurich, South Asian University, University of Calcutta, Azim Premji University, Indian Statistical Institute Delhi, and Indian Institute of Technology Delhi. We are also grateful to the editor and two anonymous referees for their useful suggestions. The authors are responsible for the remaining errors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary Material

Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2020.1862796

Notes

1. We do not include the state of Jammu and Kashmir because of non-coverage and data reliability issues in several districts of Jammu and Kashmir.

2. More than 40,000 villages in 2011 and 30,000 villages in 2001 have zero population. We delete all such villages from our sample. This leaves us with 586,351 villages in 2001 and 590,599 villages in 2011.

3. This selection leaves us with a sample of 46,771 and 38,061 individuals in 2005 and 2011, respectively. The learning outcome variable is available for the age group 6–14 years.

4. For example, a village is dominated by SC if the share of SC is more than the share of ST and the share of non-SCST in the total village population.

5. Please refer to Figures A1 and A2 of Appendix A for plots of middle and secondary schools.

6. We also check if the discrimination in primary school explains discrimination in secondary school, that is, pattern we see for secondary schools is because of the past discrimination. We find that even after controlling for primary school presence, discrimination in provision of secondary schools still exists.

7. Details regarding specification and results are provided in Supplementary Material 1.

8. We use district level averages because published census data do not have village level disaggregated figures across social groups. For dropout rates, we use IHDS data. These averages do not match with national aggregates as we calculate the district level averages without population weights.

9. Distance to district headquarter is not available for the 2001 census. Therefore, we control for distance to nearest town in 2001.

10. Omitted variable bias can be eliminated or at least mitigated if a proxy variable is available for an unobserved variable. To check the validity of our proxy variable, we run a district level regression from NSSO data, which satisfies both formal requirements of a good proxy.

11. Around a quarter of villages have no SC population at all, around 55 per cent of villages have no ST population, and around 11 per cent of villages are completely inhabited by non-SCST population.

12. One can do the reverse also, by giving attributes of villages without a public school to villages with public schools. We chose not to do it because of the common support problem.

13. Details regarding specifications and results are provided in Supplementary Materials 2 and 3.

14. Details regarding RIF method and its results are provided in Supplementary Materials 5.

15. Results tables are provided in Tables C1 and C2 of Appendix C.

16. The conditional probabilities $P_d(public\mathrm{\_}schoo{l}_v=1|v\in q1)$ and $P_d(public\mathrm{\_}schoo{l}_v=1|v\in q4)$, capture the distribution of schools in a district.

17. Details regarding estimation method and results are given in Supplementary Material 6.