1The under-supply of public goods in developing countries has been a well- documented phenomenon. A growing body of empirical research attributes the under-supply of public goods to ethnic diversity, and specifically to the to conflict, political insecurity and corruption that occur in ethnically diverse contexts.

2Several studies looking at multiple countries studies have well established the negative relationship between ethnic diversity and public goods. However, the existing evidence does not adequately address the concern regarding factors endogenous to a locality related to reverse causality, whereby higher public goods provision leads to greater ethnic homogeneity (Tiebout, 1956). This has impeded the establishment of channels to identify the mechanisms that underlie this relationship. Further, existing research predominantly focuses on demand-side factors through group cooperation and preferences, ignoring supply-side mechanisms, which are the most influential factors in determining access to public goods. The existing literature broadly emphasizes two mechanisms to explain the relationship between ethnic diversity and lower supply of public goods. First, ethnic diversity results in heterogeneity in preferences; therefore, voter preferences are further away from chosen policies, resulting in reduced demand for public goods (Alesina et al., 1999). Second, if there are more cooperating individuals belonging to different groups, this would increase competition for non-excludable public goods, resulting in their reduced supply (Miguel and Gugerty, 2005).

3The question of the relationship between ethnic diversity and poor public goods provision is particularly pertinent in India, with its social hierarchy, as well as caste and religious divisions. In 2011, the Census of India reported the number of religions in India to be at 6 million. This religious plurality is mirrored by 7,000 religious conflicts that took place in India from 1950 to 2016.

4In this thesis, I analyze the effects of religious polarization, measured through an ethnic heterogeneity index, on a wide range of public goods. I use the RQ index to measure ethnic heterogeneity because it best captures potential antagonisms when a few sizeable minority groups face a large majority group, as is the case in India with its large Hindu majority and sizeable Muslim, Christian and Sikh minorities. I hypothesize two mechanisms through which religious polarization affects the delivery of public goods. First, I posit that there may be a collective dislike between religious groups in India and that these groups act collectively against each other and hold a preference for discrimination; therefore, political actors who provide public goods focus on identity pol itics1 based around religion rather than improving access to public goods. Second, preferences for goods may differ across groups. However, due to financial and technological constraints, only a few goods can be provided, which results in under-supply of public goods in religiously polarized areas.

5The paper uses geo-referenced spatial panel data disaggregated at the sub-district level measured over 20 years. This allows me to plausibly overcome the potential reverse causality concerns raised by Tiebout (1956), who hypothesized that individuals sort themselves into areas that provide them with the most desired public goods, thereby homogenizing the areas. The empirical strategy of this paper accounts for within-country migration to areas with a higher supply of public goods.

6I use a two-way fixed effects regression to estimate the effect of religious polarization on different public goods across 5,480 sub-districts from 2001 to 2011. Additionally, I adopt an autoregressive spatial regression fixed effects methodology recommended by Baltagi et al. (2011) to account for the impact of spatial characteristics and path dependence of the different variables on my results.

7My reduced-form results suggest a generally negative and statistically significant effect of religious polarization on the supply of public goods, particularly for those supplied by national/provincial government institutions. A unit increase in religious polarization results in a -0.0007 reduction in the number of primary schools per capita, -0.0004 reduction in the number of middle schools per capita, -0.0003 reduction in the number of senior schools per capita, - 0.0004 reduction in the number of colleges per capita, -0.0003 reduction in the number of all-season roads per capita, -0.001 reduction in the number of public toilets per capita, -0.695 reduction in the number of public taps per capita and -0.119 reduction in the number of public tube wells per capita. On average, I do not obtain a statistically significant impact on secondary schools, electricity supply or other public wells. The magnitude of my effects sizes appears smaller on account of my dependent variables being expressed in per-capita terms. Extending the result to absolute numbers, the mean number of primary schools in India is strikingly low, at 0.0011 with an average of 20.59 schools in every sub-district, so a decimal unit increase in religious polarization results in -0.00007 decrease in per capita primary schools which translates to 19.37 primary schools at the sub-district level, i.e. an estimated negative effect size of 6.4% in the number of primary schools.

8I also extend my model to smaller city sizes and areas with larger minority group sizes and obtain a comparable negative and statistically significant relationship. My findings are robust to other explanatory variables, a quadratic transformation of group sizes, and alternative degrees of religious polarization. Overall, I find that religious polarization has a negative impact on the supply of public goods, particularly for goods supplied by national/provincial political institutions. My panel and spatial econometric approach contributes to the rich literature on the relationship between ethnic heterogeneity and economic outcomes by plausibly resolving the persistent issue of endogeneity while establishing a causal relationship between ethnic heterogeneity and economic outcomes. To the best of my knowledge, this research is the first to focus on religious heterogeneity and its impact on economic outcomes in India.

Literature Review

9The literature on the impact of ethnic heterogeneity on public goods is predominantly comprised of several studies looking at multiple countries. While the negative relationship is well established, a major drawback of existing studies has been the issue of endogeneity in the form of reverse causality and understanding the mechanisms that underlie the relationship between ethnic diversity and public goods provision.

10This relationship was first brought to light by Alesina et al. (1999). They focused on the United States using county, metro and city data and found that more fragmented cities spent proportionally less on schooling, roads and trash pickup but more on health and police. However, Banerjee and Somanathan (2007) critique these findings for the authors’ use of data on public goods that are contemporaneously correlated to fractionalization data, and, given the high mobility environment of US cities, their analysis suffers from the reverse causality issues due to Tiebout’s sorting, whereby localities with a greater access to public goods were on-average more ethnically homogenous.

11Miguel and Gugerty (2005) and Banerjee and Somanathan (2006) explore ways to overcome the endogeneity problem discussed above. Miguel and Gugerty (2005) exploit historically determined patterns of ethnic land settlement that have remained stable for a long period of time. They find ethnic diversity to be associated with lower primary school funding and worse school facilities along with poor water well maintenance. In a similar vein, Banerjee and Somanathan (2007) tackle the issue by using ethnic fragmentation data from 1931 to explain the provision of public goods in 1971 and 1991, given the low level of mobility in rural India. Their analysis provides mixed results wherein ethnic fragmentation in the form of caste has a significant negative impact on 10 out of 26 public goods and a positive significant impact on 3.

12Other important studies in the field have been carried out by Alesina et al. (2016), Dayton-Johnson (2000) and Khwaja (2009). Alesina et al. (2016) identify a causal relationship between the level of deforestation in Indonesia and the degree of ethnic fractionalization at the district level by using the exogenous creation of districts in the country. They find ethnic fractionalization to positively affect the degree of deforestation. Dayton-Johnson (2000) looks at irrigation systems and maintenance indicators in Mexico and finds that social heterogeneity has both a direct negative effect by lowering cooperative effort and a positive indirect one by making a group less likely to select the poorly performing allocation rule such that its indirect effect on cooperation is positive. Economic inequality is found to lower cooperative effort, although its impact is U-shaped. Further, Khwaja (2009) also hypothesizes a U-shaped effect of land inequality on the maintenance of projects in rural communities in Pakistan along with a negative effect of ethnic fragmentation.

13An important addition to the literature has been the paper by Habyarimana et al. (2007). While most literature focuses on quasi-experimental strategies, Habyarimana et al. (2007) is unique in terms of establishing the relationship between ethnic diversity and public goods through experimental games. Their results suggest that successful public goods provision in homogeneous ethnic communities can be attributed to a strategy selection mechanism where individuals within co-ethnic groups cooperate whereas those in non-co-ethnic groups do not. In addition, they find evidence that co-ethnic groups are more closely linked on social networks and thus plausibly better able to support cooperation through the threat of social sanction. This result is in line with evidence on cooperation within groups being more efficient than coordination across groups presented by Berge et al. (2019). Lastly, Munshi and Rosenzweig (2016) find that cooperation is restricted to within an ethnic group and that the supply of public goods increas es in relation to the group’s size. Within-group cooperation is also matched by intra-group competition for public goods which results in lower public goods access in ethnically diverse areas.