Abstract
In a political system characterized by strong loyalty to coethnic politicians and parties, why do some voters support a non-coethnic party? Scholars theorize that voters support coethnics in order to access patronage benefits. However, for some groups, particularly minorities, coethnics may never win power, forcing voters into a second-best choice. I examine the factors that shape that choice, and the choices that voters make. This paper shows that voters are more likely to support non-coethnic parties when such parties use direct outreach to overcome credibility concerns and when their chances of victory are high.
Acknowledgements
The funding for the survey was generously provided by the Yale Macmillan Center International Dissertation Research Grant. This work was also supported by the Yale South Asian Studies Council, the Princeton Institute for International and Regional Studies, and the University of New Hampshire Graduate School Summer Faculty Fellowship. I am grateful to Steven Wilkinson, Tariq Thachil, Susan Stokes, Elena Gadjanova, and Sanjay Kumar; to participants at panels at the International Studies Association, Elections in South Asia workshop at Yale, the Yale Comparative Politics workshop, and the Princeton South Asia Graduate Workshop; and to two anonymous reviewers for their feedback on the project and earlier versions of the manuscript.
Supplemental Data
Supplemental data for this article can be accessed at https://doi.org/10.1080/17449057.2019.1594558.
Notes
1 Ichino and Nathan argue that voters in Ghana support a non-coethnic party when living in an area with a large population affiliated with that party because they expect that party to target their own coethnics in that area with local public goods i.e. public goods like roads or health facilities available to locals from which non-coethnic residents cannot be excluded.
2 The term ‘constituency’ is used in India and the UK to refer to legislative/electoral districts while countries like the US use the term ‘district.’ I use the term constituency throughout the paper when referring to electoral districts and use the term ‘administrative district’ to refer to a separate set of boundaries that have limited overlap with electoral boundaries. All constituencies discussed in the paper are assembly constituencies or in other words, state-level electoral districts rather than parliamentary constituencies which are larger.
3 See Section 2 for explanation.
4 See Appendix.
5 Interview with Maulana Khalid Rasheed of Firangi Mahal, Lucknow, March 2012.
6 Author fieldwork.
7 Data presented as support from Muslim low castes, high castes, and unknown castes.
8 Interview, December 9, 2011, Saharanpur City.
9 Interviews conducted April–May 2012.
10 15% includes the 4% of constituencies where the BSP and SP tied for most popular party.
11 Name changed.
12 As a caveat, sample sizes in each constituency are small and therefore the data from survey questions are only suggestive of dynamics that I test for in the survey.
13 Name changed.
14 Name changed.
15 Interview with Sanjay Garg, BSP MLA candidate in Saharanpur City. December 12, 2011.
16 Mentioned as a BSP programmatic benefit by Maulana Khalid Rasheed of Firangi Mahal, Lucknow, March 2012.
17 Interview with Swami Prasad Maurya, Uttar Pradesh state president of the BSP. May 22, 2012.
18 See Appendix for description of sampling method.
19 See discussion in Appendix.
20 Name withheld to maintain confidentiality.
21 Some may argue data from a Hindu nationalist party is less reliable as the Hindu right has employed the threat of a Muslim population explosion in their mobilization efforts. However, this is data not just on Muslims but on all ethnic groups in every constituency. The purpose of this data is not for public consumption but rather for designing party strategy which incentivizes maintaining accurate numbers.
22 I was given access to the data estimates of the Muslim population from CSDS, which was not sufficient to test the impact of the size of other ethnic groups on Muslim behavior. However, this data allowed me to check the reliability of other demographic data I obtained.
23 I discuss this in the appendix.
24 I do not use income to ascertain class status. Responses to a question about monthly income were categorized by surveyors into nine possible responses, mirroring surveys used by CSDS. In robustness checks, I used a binary variable in which respondents were categorized by whether they were above or below the per capita income of UP and the results are similar to education levels results – there is no class difference between SP and BSP supporters. However, income obscures significant heterogeneity in responses. Most importantly, family sizes vary significantly among respondents and therefore income categories cannot capture the differences in economic circumstances between a family of five and a family of 15.