Diversity Deficit and Scale-Flip

Abstract

We present a comprehensive multi-scale test of the diversity-deficit hypothesis that posits a negative association between diversity and development. We develop a ‘scale-flip hypothesis’ that formalises how the political salience of diversity is contingent on the level of analysis. We also contribute to the political economy of public goods literature using the largest dataset used to date – n ≈ 1.2 million village-year points from a two-period panel of all villages in the Indian national census data. We find evidence for ‘scale-flip’ so that there is a robust positive association between diversity and public goods at the local level.

Acknowledgements

We thank Adam Auerbach, Christian Davenport, Satish Deshpande, James Fearon, Devesh Kapur, Alexander Lee, Vijayendra Rao, Hillel Soifer, and Emmanuel Teitelbaum for useful comments and suggestions. We also thank the two anonymous reviewers for constructive criticism that helped improve the paper. Bharathi acknowledges support from Harvard University's South Asia Institute. Mishra received funding for this research from the International Center of Advanced Studies: Metamorphoses of the Political (ICAS:MP), in its TM6 (political economy) module. Part of the research reported here was carried out when Malghan was affiliated with Princeton University. He thanks the Princeton Institute for International and Regional Studies for funding support. The usual disclaimers apply.

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/10.1080/00220388.2020.1802011

Notes

1. For instance, inter-group violence has been studied across countries (Fearon, 2003), within sub-national boundaries (Wilkinson, 2006), cities (Varshney, 2003), counties (Tolnay & Beck, 1995), or even grid-cells (Pierskalla & Hollenbach, 2013). Similarly, studies of the association between diversity and public good provision have used nation-states (Baldwin & Huber, 2010), legislative assemblies (Banerjee et al., 2008), cities (Alesina et al., 1999), or villages (Miguel & Gugerty, 2005; Munshi & Rosenzweig, 2016).

2. Gerring et al. (2015), however, do include country fixed effects. Cf. Gerring, Thacker, Lu, and Oncel (2017) for a multi-scale test of status of minorities as a function of geographic unit of analysis. The authors use a subset of global Demographic Health Survey (DHS) data from nine different countries where DHS identifies two sub-national aggregations – regions and districts. However, this cross-national dataset relies on definitions of regions and districts that can vary across countries. We are grateful to an anonymous reviewer for pointing us to this article as an example of a true multi-scale test in a cognate field.

3. Scholars in the field of comparative politics have also made a strong case for diversity-dividend at the sub-national level, albeit through multiple channels (Gao, 2016; McDonnell, 2016; Singh, 2015b; Soifer, 2016). However, these sub-national studies also suffer from problems related to theoretical and empirical concerns of scale-dependent group behaviour and spatial aggregation effects (Soifer, 2019).

4. The pertinent point here is not whether contact is possible at larger aggregations such as the district, but that contact is at the very least potentially contingent on scale. While sustained contact is more likely at local scales, the relationship between contact and scale is ultimately an empirical question. We are grateful an anonymous reviewer for helping clarify this linkage between contact theory and scale.

5. Spatial variation in public goods also occur on account of regional partisanship and the nature of urban concentration, as in Ecuador and Colombia (Soifer, 2016). Also, cf Trounstine (2016).

6. See https://www.washingtonpost.com/graphics/2018/national/segregation-us-cities/(accessed, 30 March 2020).

7. Also cf. Gerring et al. (2015) who report a scale-flip in a diverse set of countries around the world.

8. While administratively classified as SCs, or ‘Scheduled Castes’, many of these groups self-identify themselves as Dalits, or the oppressed.

9. Revisiting the same question, albeit with updated district-level census information, Balasubramaniam et al. (2014) found that households in districts with higher caste diversity have lower access to tap water while religious diversity exhibits an opposite effect.

10. Measuring the psychological effect of local reservations for the SCs and STs, Chauchard (2014) finds that while upper-caste individuals may continue to harbour biases against marginal groups that avail quotas, inter-group interactions are amiable and there is a reduction in discriminatory practices.

11. The 73rd Amendment to the Indian constitution (1992) that extended political quotas to local government elections has further transformed the nature of redistributive local political preferences, thereby affecting the provision of public gods (Besley et al., 2004; Parthasarathy, 2017).

12. Highly aggregate category data that combines caste and religion is available at the sub-district level, too.

13. For meaningful representation of the distribution at all three levels, we dropped states with less that thirty sub-districts (administratively, sub-districts are most commonly referred to as Talukas, or Thesils). Across India, $\approx 600,000$ villages are embedded in $\approx 6000$ sub-districts, that in turn are contained within $\approx 600$ districts.

14. We calculate similar coefficients across each of the Indian states and the correlations continue to be very high. Also cf. $\mathcal{\S }$ D of Supplementary Materials.

15. For example, cf. Gershman and Rivera (2018) for data from Africa.

16. $\mathcal{\S }$ C, Supplementary Materials.

17. $\mathcal{\S }$ D, Supplementary Materials.

18. $\mathcal{\S }$ E, Supplementary Materials.