Version classiqueVersion mobile

Religious Polarization and Under-Supply of Public Goods

 | 
Pulkit Bajpai

2. Data

Texte intégral

  • 1 Sub-districts (or Tehsils) are administrative units in a district with towns and a large number of (...)

1This paper uses data from Asher et al. (2019) (SHRUG) and the Census of India from 2001 to 2011 to construct a novel panel data set covering a variety of demographic characteristics, public goods and nighttime luminosity. The data cover 5,448 sub-districts or Tehsils1 spread across 28 Indian states and 7 union territories. The census is carried out with a 100% sample with a universal coverage across India. The unit of analysis for the study is at the sub-district level, while sub-districts have been merged across time using the unique identifiers prescribed by Asher et al.(2019). Administrative changes with regard to administrative area names and boundaries have been accounted for.

2The panel data set plausibly tries to solve the endogeneity concerns in the literature on ethnic fractionalization. It allows me to distinguish the supply-side channel from the demand-side channel, linking diversity in preferences in heterogeneous populations to the under-supply of public goods. Despite my best attempts to use techniques to minimize the problem of omitted variable bias, I cannot rule out the possibility of it affecting my results.

2.1 Religious Demographics

3Indian society has long preserved the modern concept of secularism, well represented by its strong demographic plurality today. The country is a cultural melting pot with multiple languages, ethnicities and religions. The Census of India reports 32 languages having more than a million native speakers and 12 major ethnicities. The number of religions followed in India has been reported at over 6 million and these have been grouped into seven broad religions: Hinduism (79.80% of the population); Islam (14.23% of the population); Christianity (2.30% of the population); Sikhism (1.72% of the population); Buddhism (0.70% of the population); Jainism (0.37% of the population); and other religions and faiths including tribal religions (0.78% of the population).

4Notwithstanding the secular fabric of the country, religious conflict in India has a long history that has persisted from the 12th century until the present. Religious conflict in India is an example of a broader category of ethnoreligious conflict – defined as conflict that involves ethnic groups distinguished from each other on the basis of their religions (Iyer and Shrivastava, 2018). More recent religious conflict in India is rooted in the partition of the country in 1947 and has continued through the second half of the 20th century, accounting for over 7,000 deaths and 30 religious riots each year in India over the period from 1950 to 2006 (Mitra and Ray, 2014). Hasan (2010) estimates that there have been over 10,000 causalities resulting from Hindu-Muslim conflict since 1950.

  • 2 Hindus do not consume beef becaues cows are considered sacred by followers of the faith.

5The issue of religious conflict is pertinent now more than ever with existing social antagonisms being reignited and exacerbated by political actors, resulting in more sectarian violence and reduced focus on governance and capacity building by political actors. India’s current political climate revolves around the idea of Hindu nationalism with anti-minority sentiments cropping up as a by-product of this religious populism. The ruling government has pushed forward controversial anti-Muslim citizenship laws and anti-conversion bills for inter-faith relationships, and has been complicit in the steep rise in mob violence, lynchings and beef vigilantism2 against religious minorities.

6Minorities, especially Muslims, Scheduled Castes and Scheduled Tribes, face non-violent discrimination in areas including employment, education and housing (Pandya et al., 2010). They also encounter obstructions to achieving political power and wealth and lack access to health care and essential services. This non-violent conflict directed toward minorities, which is manifested in the form of discrimination, has long-term adverse impacts on individuals and communities belonging to a minority group. Ahmed et al. (2011) find that religious discrimination can cause depression, anxiety and suicidal ideation. The psychosocial impact of religious discrimination poses a more significant, graver concern, which has primarily been undocumented in India. While I can quantify the impact of violent religious conflict, I cannot document the more considerable psychosocial impact of religious discrimination.

2.2 Religious Polarization

7The literature on ethnic diversity uses the popular Ethnic Fractionalization (EF) Index from Alesina et al. (2003) to measure ethnic heterogeneity. This is computed as 1 minus the Herfindahl index of ethnic group shares and reflects the probability that two randomly selected individuals within a population will belong to different groups. The EF index can be written as follows:

πij i i .., N j

8The measures designed by Esteban and Ray (1994) define polarization as the sum of interpersonal antagonisms that come to fruition as a result of a sense of group identification (measured by α) and a sense of alienation with respect to other groups measured by distance (conventionally normalized to 1). The normalized Polarization Index as defined in Montalvo and Reynal-Querol (2005) can be written as follows:

πi i i .., N jK > α , .α

9This study measures Religious Polarization (RQ) with the median value of polarization sensitivity (α = 0.8) in accordance with Montalvo and Reynal-Querol (2005) and Bros (2010). Therefore, the RQ index for the median value can be simplified as proposed by Montalvo and Reynal-Querol (2005), written as follows:

πij i i .., N j

10Secondly, according to Horowitz (1985), ethnic conflicts occur in countries where a few sizeable ethnic minorities face an ethnic majority. Therefore, the presence of a large ethnic group is close to being a necessary condition for a high probability of ethnic conflict but is not sufficient. The RQ index is closest to the maximum value in the presence of a significant undivided ethnic minority. Lastly, Collier (2001) emphasizes the non-monotonic relationship between ethnic diversity and conflict. Highly heterogeneous societies have an even lower probability of conflict than homogeneous societies, and the highest probability of conflict is associated with the middle range of ethnic diversity. Using this index, I substantiate that the Indian demography is greatly polarized. My measure of ethnic heterogeneity ranges from 0.01 to 0.94 and has a mean of 0.38 in 2011, compared to the mean value of 0.25 in the multiple countries analysis for countries demographically comparable to India: i.e., with a large ethnic majority and minority groups (Bhavnani and Miodownik, 2009).

11I illustrate the relationship between the RQ index and incidents of religious conflicts (riots, arson etc.) in Figure 2.1 below. I use crime data from the National Crime Records Bureau from 1991 to 2011 at the district level and find the correlation between incidents of religious conflict with the RQ index. I observe a positive, poorly fitting correlation between the RQ index and religious conflict:, i.e. the higher the RQ in a district, the larger the number of religious conflicts. I propose that the illustrated relationship should not be interpreted as a causal relationship.

Figure 2.1: Correlation between RQ Index and Religious Con flict

Figure 2.1: Correlation between RQ Index and Religious Con flict

2.3 Public Goods

12This paper hypothesizes a negative impact of religious polarization on public goods provision due to religious antagonisms supported by political leaders. I focus on two types of public goods: one supplied by national/regional government institutions, the other supplied by local governments through collective action. This distinction between public goods allows me to identify the channels that determine the supply of public goods by exploring the channels of social antagonisms through political leadership or local collective action. I have included public goods provided by a national/regional government institution, such as primary schools, middle schools, senior schools, senior sec ondary schools, colleges, electrical supply and accessibility to regional roads. I have also included facilities that may be produced locally through collective action such as the availability of toilets, taps, wells and tube wells.

2.3.1 Explanatory Variables

13Population Growth: Total population sizes within each district and sub-district are obtained from Asher et al. (2019) (SHRUG) for the census years 1991, 2001 and 2011. I perform a logarithmic transformation of the population at the sub-district level to account for the level effect of population growth on the supply of public goods (Tamai, 2018).

14Caste Fractionalization: Indian society is deeply fragmented along historically defined castes (Jatis). There is low upward social mobility for individuals born into the lower castes, as well as social barriers on customs, marriages and social interactions. Caste plays a significant role in determining economic and political outcomes; hence I include caste fractionalization in the form of the traditional Ethnic Fractionalization (EF) index to distinguish between religious and caste heterogeneity. The EF index is theoretically motivated by virtue of a large number of sub-divided castes present in India (Banerjee et al., 2007).

15Nighttime Luminosity as a Proxy for Regional GDP: Nighttime luminosity has become an important proxy for economic activity, particularly due to the paucity of data measuring GDP in developing countries. Given the absence of regional-level data on GDP in India, I use nighttime luminosity as a measure for regional GDP following Hodler and Raschky (2014). They establish a linear relationship between GDP and nighttime luminosity at the country and regional level. Prakash et al. (2019) find that night light data exhibit a reasonably robust correlation with regional GDP in India.

16This paper uses the DMSP-OLS (Defense Meteorological Satellite Program, DMPS; Operational Linescan System, OLS) annual measures of nighttime luminosity from Asher et al. (2019) and Henderson et al. (2009) compiled for the years from 2001 to 2011. I use the mean of the luminosity values (ranging from 0 to 63) at each sub-district that is merged through a unique identifier. However, given the low level of economic activity in my sample, the data show a deficient nighttime activity level, which has been accounted for using a logarithmic transformation. I find a strong correlation (0.78) between nighttime luminosity and bootstrapped per capita consumption values (Asher et al., 2019) at the sub-district level. A potential endogeneity concern arises in my analysis from the relationhip between electricity supply and nighttime luminosity. However, I find a low correlation (-0.006) between the two variables from 1992 to 2011.

17This study also uses area size, literacy levels, population share of religious groups and geographical variables in the form of rainfall and temperature as additional explanatory variables. The RQ index, caste EF, religion-specific population shares and per-capita public goods form the core variables of my analysis. My control variables, such as population growth, area size, literacy levels, nighttime luminosity, temperature and rainfall, form my secondary variables.

18I report the variation in core variables between 2001 and 2011 using mean differences reported in Table 2.1. I notice that the religion RQ and caste EF increase from 2001 to 2011: i.e., religious polarization and caste fragmentation increased over this time period, albeit insignificantly.

19My variables of interest – i.e., number of public goods at the sub-district level, including primary schools, middle schools, senior schools, senior secondary schools, colleges, roads, toilets, taps, wells and tube wells – had a statistically significant increase between 2001 and 2011. For ease of interpretation, this table reports numbers that are not expressed in per capita terms. I report summary statistics for my panel data in Table 2.2 below. My RQ index ranges from 0.01 to 0.94 with a mean of 0.38, whereas the EF index ranges from 0.00 to 0.66 with a mean of 0.36.

Table 2.1: Mean difference between variables (2001-2011)

Table 2.1: Mean difference between variables (2001-2011)

Table 2.2: Descriptive statistics of main variables (per capita) across 5,448 sub-districts (2001-2011)

Table 2.2: Descriptive statistics of main variables (per capita) across 5,448 sub-districts (2001-2011)

Source: Census of India

Notes

1 Sub-districts (or Tehsils) are administrative units in a district with towns and a large number of villages.

2 Hindus do not consume beef becaues cows are considered sacred by followers of the faith.

Table des illustrations

URL http://books.openedition.org/iheid/docannexe/image/8742/img-1.png
Fichier image/png, 18k
URL http://books.openedition.org/iheid/docannexe/image/8742/img-2.png
Fichier image/png, 24k
URL http://books.openedition.org/iheid/docannexe/image/8742/img-3.png
Fichier image/png, 22k
Titre Figure 2.1: Correlation between RQ Index and Religious Con flict
URL http://books.openedition.org/iheid/docannexe/image/8742/img-4.png
Fichier image/png, 40k
Titre Table 2.1: Mean difference between variables (2001-2011)
URL http://books.openedition.org/iheid/docannexe/image/8742/img-5.png
Fichier image/png, 137k
Titre Table 2.2: Descriptive statistics of main variables (per capita) across 5,448 sub-districts (2001-2011)
Crédits Source: Census of India
URL http://books.openedition.org/iheid/docannexe/image/8742/img-6.png
Fichier image/png, 593k

CC-BY-NC-ND-4.0

Le texte seul est utilisable sous licence CC BY-NC-ND 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.

Acheter

Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search