In my last column, I reported some work Laveesh Bhandari and I have done in identifying backward districts in India. The identification is based on variables linked to the millennium development goals, constrained by the problem of the lack of availability of data at district-level. The variables used are poverty, hunger, infant mortality, immunization, literacy and gross enrolment rates. There is a set of around 70 districts that are most backward and these are geographically concentrated. There is a set of another 70 districts that are just above the most backward category and these tend to be spread out more spatially. Understandably, any such identification is dependent on variables used for identification. In a little more detail, here is broadly what we find.

Poverty ratios show backward districts not only in the undivided BIMARU (Bihar, undivided Madhya Pradesh, Rajasthan, undivided Uttar Pradesh) states, but also in Gujarat, Maharashtra, Karnataka, Tamil Nadu, Andhra Pradesh, Orissa, West Bengal and the Northeast. Hunger (defined in National Sample Survey terms) exhibits a similar spatial distribution, but is less universal than poverty and is also more concentrated towards the East and the Northeast. Other than a few neighbouring districts of Karnataka and Andhra Pradesh and the Northeast, the infant mortality rate identifies undivided BIMARU and Orissa. Lack of immunization is geographically a more serious problem, with a clear North versus South divide. Low literacy rates are spread throughout the country and expectedly, this is also mirrored in gross enrolment rates.

While there is some correlation, it is not surprising that identification of backward districts depends on which of the six indicators are used. For example, under the poverty ratio criterion, India’s worst districts are located in Bihar, UP, Jharkand, Orissa, MP, Assam, Maharashtra, West Bengal and Chhattisgarh, with a few districts from Arunachal, Karnataka and Tamil Nadu thrown in. Hunger has a broader geographical spread, with hungry districts also existing in Andhra, Goa, Haryana, Kerala, Manipur, Nagaland, Pondicherry, Rajasthan, Tripura and Uttaranchal. The contrast in spreads between the poverty criterion and the hunger criterion, with the former more concentrated, is a finding that needs emphasis. And this also has a policy angle.

Moving on to infant mortality, the worst districts are in UP, Orissa, MP, Chhattisgarh and Rajasthan. The worst-off districts under immunization have a broader spread, with Bihar, Jharkhand, Arunachal, Karnataka, Assam, Gujarat, West Bengal and the Northeast also included. A somewhat surprising finding is the fact that none of Bihar’s districts figure in the worst-off list under infant morality, but several are included under immunization. Conversely, Orissa has an extremely good record in immunization, but a poor record in infant mortality.

Immunization and infant mortality ought to move together. Backwardness under the literacy criterion is concentrated in Orissa, undivided BIMARU, Arunachal, Karnataka, Andhra, Assam, Gujarat, Himachal, Jammu and Kashmir, Punjab, West Bengal and the Northeast. The enrolment criterion broadly mirrors this, with a large concentration in Bihar and UP. Perhaps one should mention that several of Gujarat’s districts are backward under the enrolment criterion.

For each indicator, the identification therefore throws up a set of districts. Is it possible to identify backward districts across the six indicators? The arbitrary cut-off — that a district is backward if it classifies as backward in at least 4 out of the 6 indicators — is used. Certainly, this selection is arbitrary, but one cannot completely avoid arbitrariness. This throws up a list of 69 backward districts — 26 in Bihar, 13 in UP, 10 in Jharkhand, 10 in Orissa, 6 in MP, 3 in Arunachal and 1 in Karnataka. These are India’s most disadvantaged districts. But one should mention that these 69 districts are the most backward. Geographically, they are contiguous to another 70-odd districts that are also fairly backward.

What policy interventions are required to ensure that these backward districts do not continue to be marginalized? This is a difficult question to answer and also concerns appropriate targeting of anti-poverty programmes such as food security, employment guarantee schemes and education and health-sector interventions. Backwardness in Bihar, UP, Jharkhand and Orissa is more broad-based. But there is a case for specifically targeting Gulbarga in Karnataka, Changlang, Tirap and Lohit in Arunachal and Chhatarpur, Damoh, Dhar, Jhabua, Panna and Tikamgarh in MP.

This exercise does not explain backwardness. We stop at identification. Explaining backwardness will come later and will involve an econometric exercise, seeking answers in connectivity, governance, institutions, historical legacies and even land tenures. Having said this, there is clearly a correlation between backwardness and connectivity. For example, backward districts in Orissa, Arunachal, Karnataka and MP are off the national highway network and it is doubtful that satisfactory feeder roads exist. This impression is reinforced when one notices that many villages in backward districts are not connected to pucca roads.

However, this argument does not hold for Bihar, UP or Jharkhand. Barring Arunachal, rail networks exist in these backward districts. That does not, of course, mean that feeder roads exist. So do rivers, although existence of rivers does not necessarily imply that navigable waterways exist. Rather interestingly, in Bihar, UP and Orissa, many of these backward districts tend to be flood-prone. One knows this impressionistically and there are implications for the river-linking project. But there is not much of a correlation between backwardness and districts affected by drought. This is contrary to what many people believe and requires further probing.

At a somewhat simplistic level, a two-point agenda then emerges. First, improve road-connectivity in backward districts in Orissa, Arunachal, Karnataka and MP. Second, solve the flood problem in Bihar, UP and Orissa. Although not used for identification, some other variables are also mapped. First, there is no obvious correlation between access to safe drinking-water and backwardness. Non-availability of safe drinking-water is almost a universal phenomenon, spanning all states. Second, there is no obvious correlation with safe delivery, the problem of women not receiving skilled attention during delivery also being a near-universal problem.

Third, there is not much correlation between gender disparities, measured by the 0-6 sex ratio and backwardness. If one ignores the Western parts of UP, the correlation is almost negative. Fourth, gender disparity measured by the ratio of female to male literacy leads to the same kind of conclusion. Gender disparities cut across backwardness. Fifth, there is some correlation (with the exception of Arunachal) between low female work-participation rates and backwardness. Sixth, there is a finding that is again contrary to popular perception. There is not much correlation between backwardness and existence of tribal populations per se. While this is a valid correlation if one restricts oneself to the central Indian core of backwardness, it is not a correlation that is evident at an all-India level.

This is by no means the last attempt at identifying pockets of backwardness in India. At a broad-based level, going beyond poverty ratios, this is certainly the first. Hopefully, this exercise will stimulate interest in taking the development debate beyond state boundaries and thereby impart focus to districts. That is where the deprivation is. And that is where the development needs to be. As I said in the earlier article, state boundaries are administrative ones. Development and deprivation do not necessarily follow these boundaries.