The targeting challenge in India’s welfare programs6 min read . Updated: 08 May 2019, 10:21 PM IST
The socio-economic caste census is being increasingly used for targeting in welfare schemes but there are discrepancies with other measures of deprivation
The noisy and often-polarized campaign for the ongoing Lok Sabha elections seems to have masked one emerging consensus in India’s polity cutting across regional and ideological divides: that targeted transfers to selected beneficiaries are here to stay. Where politicians and parties differ is on the mechanics of selecting beneficiaries and the most effective means to reach them.
After the National Democratic Alliance (NDA) government unveiled an income-support scheme for small and marginal farmers, the Pradhan Mantri Kisan Samman Nidhi (PMKSN), the principal opposition party, Congress has promised its own variant, the Nyuntam Aay Yojana (NYAY) for India’s bottom quintile, cutting across rural and urban areas. These announcements follow similar initiatives by regional parties in their respective states. Most of these initiatives depend on either land records — which are often patchy — or on a dated database based on 2011 numbers: the Socio-Economic Caste Census (SECC). The SECC is also being used in other central schemes such as the Ayushman Bharat and Pradhan Mantri Awas Yojana to identify beneficiaries. The SECC will also be used to roll out the NYAY scheme if the Congress comes to power, according to the party’s spokespersons.
But how reliable is the SECC database?
A Mint analysis of the SECC numbers suggests that there are reasons to be cautious about it. A district-wise comparison with data from the last census conducted in 2011 and numbers from the more recent National Family Health Survey (NFHS 2015-16) suggests that while there are some common patterns in all three databases, there are considerable differences when it comes to identification of the most backward districts.
For this analysis, all districts were ranked in five quintiles according to the percentage of deprived households, as identified by the deprivation criteria laid down in the SECC. A similar ranking exercise was undertaken according to the asset ownership data from the Census, and the percentage of households identified as poor on the basis of the Multidimensional Poverty Index (MPI), created by the Oxford Poverty and Human Development Initiative, based on the NFHS data.
Of all the districts classified as the most deprived by the MPI, 48% of them are found to be the most deprived according to the census. The overlap between these two databases is the strongest although they were conducted five years apart. In contrast, there is a smaller overlap between the SECC and the census, which were both conducted in the same year. Merely 40% of the districts ranked as the most deprived according to the SECC rank as most deprived when we use the census data. The match between the MPI and the SECC is a dismal 25% when it comes to the most deprived districts (districts in the bottom quintiles), showing large discrepancies.
One reason why there is a greater match between the census and the MPI databases could be that neither of them were designed or introduced to measure poverty. The SECC was conducted to replace the old below-poverty-line (BPL) lists to identify the potential beneficiaries of government schemes better. As a result, it is likely that the SECC overestimates deprivation, at least in some parts of the country, while the other two databases don’t.
However, it is worth noting that both the SECC and MPI data share one common pattern. They both suggest that the most backward districts are concentrated towards the eastern and north-eastern regions of the country, with the least backward districts in the south and the west. Interestingly, ranking the districts on the basis of asset ownership according to the 2011 Census reveals a slightly different pattern of deprivation. The Census of India, which captures household ownership of major assets (like TVs and cars), shows that the poorest districts are concentrated in central and eastern India.
The other reason for these differences of course lies in the differing nature of the deprivation or asset criteria used. None of these databases are meant to be, or could be, identical.
The SECC was conducted by the rural development ministry, primarily to determine beneficiaries of welfare schemes. The survey captured various aspects of poverty, such as access to amenities and asset ownership, and the results are increasingly being used as the basis for targeting welfare schemes.
The SECC can also be used to exclude households from accessing schemes through certain exclusion criteria, such as owning a car. But the crux of the SECC are seven deprivation criteria, such as no literate adult member above 25, which are used as criteria to identify beneficiaries for specific schemes. In addition, households can be automatically included if they meet certain other inclusion criteria. A household is considered deprived if it meets any one of the seven deprivation criteria.
While the SECC database is considered by most economists as an improvement compared to the old below-poverty-line lists, the striking differences with the other databases when it comes to the identification of the most deprived districts suggests that India’s old problem of identifying beneficiaries has not been fully solved yet.
NC Saxena, whose recommendations led to the SECC argues that there might be inherent biases in the data. Respondents had a vested interest to overstate the extent of their deprivation in order to be identified as beneficiaries of welfare schemes. He also argues that errors in enumeration may have led to under-counting of the poorest sections. Further, with regard to the caste information, there were 81 million errors reported, that are still being rectified. According to a former census commissioner, M Vijayanunni, the SECC exercise was deliberately undermined by policymakers while it was being planned.
Solely relying on SECC for targeting may lead to inefficient outcomes. One solution could be to use SECC data at a state-level using state-specific criteria, as being done with the PDS, according to the development economist Jean Dreze. In a few states including Jharkhand, West Bengal and Bihar, the state government are using the SECC as the basis for the identifying PDS beneficiaries.
The other challenge in using the SECC database is that it is already eight years old in an economy which is transforming fast, and where some people have climbed up the income ladder while others have fallen down. This means that a SECC-type exercise needs to be repeated at frequent intervals to ensure that it matches current reality. But the more the database is mined for such use, the greater the chances of reporting biases creeping in, as people learn how to game the database to remain within the ‘right’ cutoff limits.
Most developing countries in the world rely on proxy-means tests—based on directly verifiable and observable information on household assets or amenities (such as roof and wall material) rather than on self-reported incomes—to classify and target households. But the exact formula to calculate the proxy-means score is often kept secret because if it is known, households (perhaps in cooperation with better-informed agents) may strategically misreport or hide assets to make sure they fall under the cutoff, as the economists Rema Hanna and Benjamin Olken pointed out in a 2018 National Bureau of Economic Research (NBER) working paper. Applying such secret formulae however robs the process of transparency, and may invite charges of political favouritism in a keenly-contested democracy such as ours.
The inherent challenges in any targeting exercise suggests that quasi-universal schemes with simple exclusion criteria based on regular and professionally conducted censuses may be a better bet for a country such as ours.