In a first, accounting for jati and not just govt-defined castes gives new insight about India’s ethnic heterogeneity. 

The fiasco around the NRC Assam draft shows that the fear of migrants and ‘inter-mixing’ have often caused tensions, or worse, violence in India. But does ethnic diversity really harm the economy of a place? Or does it boost development?

Well, it depends.

The relationship between diversity and development is an artefact of where diversity is measured, how diversity is measured, and what diversity is measured. It jointly determines an ethnic-geographic continuum.

To understand the relationship between diversity and development in India, we studied data from 1,641 jatis, 27,000 villages and 175 sub-districts in Karnataka, the seventh largest state in the country.

What we found

There is no systematic relationship between diversity and development. Both the direction and strength of any such association is contingent on particular ethnic and geographic scales.

We combined data from various sources including ‘night lights’ that can be seen on the ground from satellites, village-level administrative data on human development, census data on village-level provision of public goods, and independent India’s first census-scale enumeration and coding of detailed endogamous caste groups (jati).

While at the village-level we saw broad support for the theory that diversity negatively impacts development or the diversity-debit hypothesis, the negative association largely disappears at the sub-district level.

Even at the village-level, ethnic inequality is positively associated with growth in per-capita night-lights, an effect that is not statistically significant at the sub-district level.

We were unable to rule out the possibility that both diversity-debit and diversity-credit relationships are spurious statistical artefacts. The only stable association is a negative relationship between caste polarisation and development.

Caste polarisations have proved to be bad for development in an area.

However, this negative association is not statistically significant at the sub-district level. Similarly, the positive association between religious, linguistic diversity and development is not statistically robust across dependent variable specification or geographic scales.

How is this new?

Earlier studies have mainly assumed that the relationship between ethnic diversity and political economy is negative. Which means, greater the ethnic diversity in an area, lesser the economic and political benefit. This is called the diversity debit theory.

However, these studies used nations as the principal geographical and political unit and did not take into account sub-national politics or contexts. The sub-national context becomes especially important in a country like India where a study using government-defined castes will not have the same nuances as those addressing sub-castes or jatis.

Limited evidence from sub-national settings in developing countries has generally not supported the diversity debit hypothesis.

The nature of governance regimes, as well as political processes at the subnational level, are different from those at the national level, confounding the potential effect of diversity and leading to greater empirical ambiguity.

Indeed, in modern urban centres supporting a complex economy, diversity has a positive effect on both wages and productivity.

Also, our caste coding represents the first such census-scale attempt since the colonial decennial census of 1931. The rich micro-data allowed us to study caste with other diversity variables such as religion, language, and even economic class.

Caste is the most important social cleavage in agrarian India and influences economic outcomes.

What it means

The relationship between diversity and development is nothing but a statistical artefact.

We used 17 different diversity metrics across different levels of ethnic and geographic aggregation to reach this conclusion.

Not only did we use existent parameters, we also brought in household landholding data to measure ethnic inequality. The dominant landowning castes own nearly 45 per cent of all agricultural land held by rural residents of Karnataka.

And finally, for the first time since 1931, jati data was used in a study about diversity. While castes are government defined, jatis give a much more realistic picture of what is going on in the ground.

This is an edited extract from “More Heat than Light: Census-scale Evidence for the Relationship between Ethnic Diversity and Economic Development as a Statistical Artifact” published in August 2018. Read the full paper here.