How important is caste identity for village level networks? We try to explore answers to this question by trying to understand the drivers of village network formation. We use the data from The Rural Economic and Demography Survey (REDS), which contains detailed information at the household level about social networks and Jatis. The data are derived from surveys carried out across 17 states and 248 villages and includes 120000 households. Detailed Jati level data was collected for each of the households ensuring the richness of data. Each household in these surveyed villages was asked questions about their network. Specifically, each household was asked to name three households it could borrow food from, three households it could borrow a small amount of money from and three households it would prefer to have as its neighbour. The data allows us to understand networks of three different kinds - the first is likely driven by the female, the second by male and the third by household preferences.
Each time a household names another household for one of the questions about the network, we say the household forms a link with that other household. For each of these networks, we observe not only how the household forms links but also how the household receives links. The household is said to receive a link each time another household forms a link with this household. The links formed and received by each household forms the basis of how we study the network.
We first look at the degree of the total links formed and received by the household, and try the drivers for that. We try to explain the degree of the household in each of the three networks using household-specific variables, caste-specific variables and village-level variables. We find that caste and village-level characteristics are important drivers of connectivity in the village. We find that characteristics of the caste determine how connected the members of the caste are and similarly, there are village-level characteristics which drive the level of connectivity of the village.
We find that household-level variables do not always impact link formation in the three kinds of networks at the same level. Two important determinants of household status are whether the household is headed by a male and the number of panchayat elections the household has contested in. These determinants work to increase the degree of connectivity of that household. We find that the absolute level of education matters only for the third network of a most preferred neighbour where more educated households have higher connectivity. We also find that variables such as wealth and income matter not in their absolute value but in the relative values. The relative position of the household with respect to its own caste group and the village as a whole matter in determining the connectivity for the household. We find that households with greater than average income have a higher degree in the network for borrowing money and preferred neighbour but not for the network of borrowing food.
In the borrowing money and preferred neighbour network, we find that having more land than average income increases the degree of the household. At the same time, being too rich or having too much land reduces the degree. Similarly, being too educated reduces the degree of the household for each of the three networks.
The characteristics of the caste determine the connectivity of its member households. The proportion of male-headed households in the caste increase the degree for all the households in all three networks. The total migration out of the village from that caste group increases the degree for the households for the borrow money network and also the preferred neighbour network. As the caste gets more dispersed by location, the connectivity of the households belonging to that caste decreases. Households belonging to SCST groups have lower degree in all the networks.
One of the most interesting findings relates to the size of caste group. There is an inverted-u relationship between the size of the caste group and connectivity of the member households. As the size of the caste group increases, more links are formed when the caste group is small, but fewer links are formed as the caste group becomes a bigger proportion of the village. Till the caste group is about 40% of the population, the degree of each household is increasing and then decreasing.
Village-level variables which depend on the caste composition have an important impact on the connectivity of the village. As the proportion of SC and ST groups increase in the village, there is higher connectivity. Note that this does not necessarily mean more connections across groups but might mean greater connectivity within caste groups. Fractionalization measures how the village fractures along caste lines with a higher fractionalization indicating more diversity along caste. Higher fractionalization decreases the connectivity of the village.
The village-level variables impacting the overall level of link formation include variables such as average age, average education and number of landed households in the village. The rise in variables increases the connectivity in the village. On the other hand, as the average time for the households in the village increases, there is lower connectivity. Inequality in income has an impact on connectivity but, the direction of the impact is not clear as two variables measuring inequality impact the connectivity differently. Inequality as measured by the dispersion of income seems to increase connectivity while higher income inequality as measured by the Gini coefficient for income, decreases the connectivity of the village. Similarly, more physical dispersion of the village is correlated with lower connectivity. Similarly, as the proportion of Hindu households grows in the village, connectivity increases.
Overall, these results indicate that networks in the village are driven by village-level and caste-level dynamics.
The results described above are from a joint project with Prof Hari Nagarajan
About The Author
Prof. Pritha Dev
Ph.D. in Economics, New York University, 2008 MS in Quantitative Economics, Indian Statistical Institute, Delhi, 2002 BA(Hons) in Economics, University of Rajasthan, 2000