The geography of caste in India I

A theory for which castes locate in which villages

I was reading Oscar Lewis’s wonderful “Village Life in Northern India” and he mentioned that some villages are organized around a single caste — his example Jat farming villages in Haryana. This surprised me. First, my experience — from running surveys in hundreds of villages — is that villages typically have multiple castes, though they are segregated into different hamlets. Second, I had a simple mental model of castes — actually jati’s, a further partition of caste — as occupations. I assumed that the 4 Varnas and the Dalits roughly covered the set of occupations that a society needs. That might imply each caste would be represented in a village.

Of course, my immediate response was to ask the hive mind: Twitter. My main takeaway was that the variety (or not) of castes you find in individual villages varies across the country. For example, there may be more single-caste dominant villages in the North than the South.

Anup Malani @anup_malani

I have a question about caste and villages in India. I’m reading Oscar Lewis’ Village Life in Northern India & he describes towns as organized by caste, Eg, Jat towns. My q is: are castes segregated by villages or by hamlets (clusters of homes) within a village?

A natural question is, why is there geographic variation in caste concentration in villages? My initial hunch was that it related to a conversation I had (also on Twitter!) with Arpit Gupta, about why India had so many more large cities in the South than in the North, especially since the North was more heavily populated than the South. One really neat finding from that conversation was that, in the North, especially the Gangetic plains, there were lots of large villages that were not far from each other. The South by contrast, was cities separated by land that is more sparsely populated with villages. Another way to put it, the high population in the North looked like it was smoothly distributed over the land, while in the South it clustered into peaks that represented cities.

This led to the following theory for the concentration of castes in a village and the distribution of castes across villages, the topic of this post. In this theory, I will take the location of villages as given. That is, I assume villages exist in a location for some reason, e.g., that is where there is some resource, like a river, a road or railroad stop, or a grain silo or a historical market. What I ask is, what castes will be represented in different villages.

This post has a “I” in the title because there are other questions about caste and geography that I am pondering. Another is why the distribution of caste varies across time and states. I am exploring that with Arpit Gupta and Nishaad Rao. I’ll try to write about that in the future.

Economic grounding

There are 3 economic theories motivating my theory of caste distribution. One is Adam Smith’s idea that specialization is limited by the extent of the market. The insight is that specialization — individuals or firms producing 1 product rather than multiple products — has increasing returns. Therefore, the larger is the local market, the more likely there will be specialization by individual producers. A second theory is a corollary: the theory of home production or, as some call it, the make-or-buy decision. This theory basically says that a person compares the cost of producing a good herself versus buying that on the market. Market production entails transactions costs, including information asymmetries and haggling costs. But if prices of market production fall enough, due, e.g., to specialization, it could be better to buy goods than to make them. The third is a gravity model of trade. This says the cost of trade — including the cost of transport and information asymmetry — increases with distance, so trade falls in distance.

Theory of caste distribution

My theory of caste distribution can be summarized by considering 3 classes of villages. One is the isolated small village: a village with small population that is not close to another village or that was found before India’s Independence. This village will likely be dominated by 1 caste that will engage in their associated occupation, usually farming or husbandry, and will make rather than buy or trade for other goods. Think of the most isolated tribal hamlets today or Jat villages in North-central India before 1950.

For practical purposes, the village was a family’s entire economy. Castes that could product food or meat could dominate a village. This meant farming or husbandry castes, and perhaps castes with skills that were adjacent to those occupations. Other castes cannot survive in such villages. They were complementary castes that would have to locate where another caste managed production of food.

Families in these villages produced almost everything aside from food themselves too. The reason was that trade was difficult. There might be the occasional oxcart or bus ride to a distant town. But that was a day or more’s investment of time, so uncommon.

Unsurprisingly, these villages may have been almost average hundreds or thousands of years ago, but looked impoverished compared to large villages or villages that were close to one another over time.

The second class of villages were ones that were small but close to other villages. These villages could trade. So villages might have multiple castes — though not a wide array of them. The regional economy could support some specialization, but each village could not. Farmers were likely the most common occupation, thus jati, across villages. But one village might contact a second caste of wood workers while another might contain a second caste of priests or metal workers. Indeed, the idea of a panchayat of villages suggests that even governance, often by Kshatriyas or Brahmins, captures a local economy with trade among a cluster of villages that enables some specialization. The result was villages with a small number of castes, one of a common set (farmers, husbandry) and a small number of others that would differ across villages that were close to one another.

Because of regional trade, there was more buying and less making in villages. People traded for goods more often than in the small village. The extent to which this happened depended on the population of the cluster of villages that traded with one another. One should be able to test this in time use data.

The third class of village are large villages, or, more accurately, towns. Towns could sustain specialization without much travel. As a result, they would have a large number of jatis. Moreover, there should be far less making relative to buying, i.e., time use should show much more specialization too.

An important variant that arises, but is beyond the 3 economic theories above, is that farming or husbandry requires land, which limits population density. The solution could be a small town with a nearby cluster of villages where farmers reside. In other words, a town is a natural extension of the second category of villages where you see non-farming jatis all locate in the same close but “big” village (i.e., town) rather than each co-locating with a farmer castes in other local villages.

Testing this theory

There are papers that examine caste segregation within villages. The reason is that scholars are naturally interested in inequality and segregation shows inequality. Inequality is important. However, it is not the focus of this post — or my theory. My theory asks what sorts of caste distribution (and, by implication, time use) we will see in villages. I do not know of papers that explore this.

What data would be required to explore this? Data on villages and the jati’s located within them. One would also want to know how close other villages are and the time-cost of transport between villages. This can be done with knowledge of road/railroad networks or by looking at historical data when transport costs were much higher. Finally, one would want to have time use data to examine make or buy decisions.

One data set that does this the Consumer Pyramids Household Survey (CPHS). These data have jati, though may not capture all jati’s in a village, as they sample only 16 households per village, starting on the main street. Thus they may not reach each hamlet in the village. CPHS does have time use, though not clearly what you make with your time. The main thing missing from CPHS is what villages they sample and how close other villages are.

There are few data sets that measure jatis, and fewer still that examine a number of villages in a region. Possible candidates are REDS, ICRISAT, the second round of IHDS. If you are willing to replace caste with occupation, you could find other data sets. But none of these include time use.

Ideally, one would test this theory with a new survey that examines a few regions across the country, surveys all villages in a block or cluster of blocks, maps out distance between villages and transport networks, and also captures time use. Adding incomes and prices would be a bonus!

If any PhD student or faculty is seriously interested in pursuing this, I would be open to collaborating! And, as always, if you have anything to add to the discussion — or think I have made any mistakes — please email me!