Local Community Composition and School Provision in India

Abstract

We study provision of schools in the Indian setting and the challenges associated with a heterogeneous society when local communities play an active role in provision. Based on the political economy literature we expect more fragmented communities to have weaker collective action. We hypothesise that this weak tendency to act collectively impacts different schools differently depending on the extent of their reliance on local community action. Consistent with our expectation, we show that there are fewer schools financed by local community (private and local government schools) in fragmented districts. Presence of public schools, provided by state and central government, for which the community has little discretionary financing powers, is not impacted. However, since public schools rely on active community action for monitoring of schools, they are found to be of poor quality. Exhaustive empirical tests have been performed to support the mechanism and discount alternative explanations.

KEYWORDS:

• Public schools
• private schools
• caste fragmentation
• collective action
• local community
• schools
• school provision

Acknowledgements

We would like to thank Rachel Gisselquist, Miguel Nio-Zaraza, Arne Bigsten, Alok Kumar, seminar participants at UNU-WIDER, DSE winter school and University of Gothenburg for helpful comments. We would also like to show our gratitude to UNU-WIDER, where a part of this research was conducted. The usual disclaimer applies.

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/10.1080/00220388.2021.1971651

Notes

1. Here concerns have been raised about the ability of local communities to lobby even for placement of central and state government schools.

2. We do not use more recent data since we need data temporally close to the 2011 census to perform robustness checks.

3. Schools are normalised by school going population (5 to 19) in the district.

4. However, results show that caste fractionalisation has a positive impact on enrolment.

5. Even though caste might differ in principle in many respects from ethnicity, it follows basic features about ethnicity that underlie the diversity-debit hypothesis. One, membership to caste, just like ethnic groups, is inherited (Chandra, 2007). Two, there are visible socio-economic differences between the ethnic groups. There is ample evidence that disadvantaged castes in India are discriminated against in the labour market, have lower representation in the political sphere, have worse education and health outcomes. Therefore, it is reasonable to relate the study of caste fragmentation in the Indian context, generally, to a broad literature on the impact of ethnic fragmentation elsewhere in the world.

6. The finding of this paper diverges from ours since we look at actual public good provision. The relationship between willingness to pay in experimental settings and public good provision in real settings is far from established.

7. Before 2012-13, DISE data collected information on only elementary schools.

8. To get caste level data, some studies have used the Survey of Living Conditions data conducted by the World Bank for two Indian states for 1997 and 1998 Lee (2018). However, since this paper studies all major Indian states, we make use of Census of 1931.

9. Note that this exercise was done for a subset of districts in two north Indian states, Bihar and Uttar Pradesh and the caste data for the year 1997-98 was collected for the seven castes. So essentially, caste proportions were matched for seven castes in two states over the two data sets.

10. Rather than conducting village census, data on caste is reported based on information provided by key village informants.

11. Apart from the urbanisation rate, we also performed the analysis with a dummy which indicates if the district has more than 50 per cent of its population living in urban areas. All our results remain same with this control variable as well. We have also done the analysis separately for rural and urban areas and all our results hold.

12. Note that we use the data on district GDP for the year 2010-11. However, the data is available only for 12 out of the 18 states in our sample. For the remaining six states, we use the latest available district GDP data, which leads to some inconsistency in the timing of district GDP data.

13. NSS thick rounds are generally considered to be representative at the district level for almost all states except for a few smaller North eastern states, Jammu Kashmir and union territories. Since our analysis anyway focuses on the major 18 Indian states, we believe our inequality estimates to be representative of district inequality.

14. The reasons suggested for the negative association is low preference for education because of high poverty levels amongst these groups and also lower ability of this group to mobilise funds for private school provision.

15. However it is less clear why would political competition only affect private schools and not public schools.

16. The unit at which parliamentary elections happen, parliamentary constituencies, are different from districts, which is the unit of our analysis. We match constituencies to districts using delimitation commission reports of the Election Commission of India. Authors can be contacted for more details on the procedure of matching.

17. We also employ other measures of political environment, namely, vote share of the winner party, vote share of national and state parties and the proportion of voters per population in a district. However, our result remains robust to all the measures of political environment (results can be obtained from authors on request).

18. We also check the impact of fragmentation on other quality variables like number of classrooms, number of toilets, availability of library, and electricity facility in a school. However, we do not find any impact on these infrastructure variables (table not reported).

19. SMC represent community participation platform for elementary schools while SMDC is meant for secondary and higher secondary schools. If the school has classes above elementary level, then it is suggested that SMC be the primary community participation platform.