How bad is the SECC data?

How bad is the SECC data?

It’s very bad. It proves the adage — garbage in, garbage out.

Urban India, Urban Indian households, SECC, SECC data, Tendulkar Committee, Socio Economic and Caste Census, Rural poverty,
The grim-faced Congress and its feigned ‘concern’ for the poor are both funny and tragic.

Lalu Prasad: We demand the release of the caste data as per the Socio-Economic and Caste Census (SECC).

Mainstream media (English) and the Congress party: More than 50 per cent of the households in rural areas are landless.

The grim rural picture revealed by the SECC data shows that the NDA has contempt for the poor.

Congress spokesperson Rajeev Gowda (Zee News, July 4): SECC 2011 reveals that there is tremendous work to be done in rural India; further, past Congress governments had made huge efforts in transforming rural India via programmes like the green revolution, white revolution and MGNREGA.


cessAs Twitterati would say, LOL. The Modi government, which the Congress is attacking for the SECC “revelations” came to power in May 2014. One has come to accept Lalu Prasad’s clownishness, and he sometimes manages to be funny. But the grim-faced Congress and its feigned “concern” for the poor are both funny and tragic. Operations for the census in question started in June 2011, when the Sonia Gandhi-led Congress was in power. The census was to be completed by December 2011, but only got completed sometime in 2013. Over these two years, the survey operations were done with handheld computers, using state-of-the-art technology.

So questions arise: Why were the data not released by the Congress government? The reason handheld computers were used was to allow faster processing and the faster release of data. Second, why was the census done by the ministry of rural development (MRD) rather than the registrar general and census commissioner, or the NSS? Both organisations have been doing survey/ census work for the last 65 years. The MRD is late to this game, and has a reputation akin to the CBI rather than the NSS — that is, it’s prone to be a political organisation and not an objective quasi-academic unit.

But one need not jump to conclusions. Before proceeding with the Congress’s hyperbole and hyper-ventilating reactions, let us try and first assess the accuracy of the SECC data. There is a close benchmark NSS survey year to the SECC census. Comprehensive data by the NSS were collected between July 2011 and June 2012 in two surveys — the Consumer Expenditure Survey (CE) collected detailed data on consumer expenditures while the Employment-Unemployment Survey (EU) collected detailed data on landholdings, individual wages and earnings, as well as the age and education structure of the population.

A detailed comparison of the SECC and the NSS is done in the table. For all indicators except education, the SECC data seem to be compellingly bad — that is, not worth discussing, let alone deriving any policy conclusions. But judge for yourself. A caveat: The released SECC data are based on a total of 299 out of 640 districts. According to the MRD and its advisors, the data are only released after thorough vetting, cross-checking, verification, etc (see Abhijit Sen’s ‘Neither BPL nor APL’, The Indian Express, July 22, for details). So even though incomplete, one can assume that the SECC data are representative and robust.

Education: This is the good news. The SECC educational attainment data almost exactly match NSS data. For example, the rural illiterate percentage (SECC 35.7, NSS 39), or percentage of graduates in the rural population (SECC 3.5 vs NSS 2.4). Since the Congress has been ruling India and formulating its education policy for 55 of the last 68 years, including the last 10, this is about as damning an indictment as one can obtain for the in-the-name-of-the-poor party. Fortunately for the Congress, the situation is not so bad — only 10.6 per cent of rural households have zero literacy (NSS data). And 46.6 per cent of rural households have at least one member of the family with greater than zero and less than five years of education, and 11.1 per cent households have at least one member who has at least 14 years of schooling.

Given that the population-education numbers match, one would expect that data on strong correlates of education — incomes — would match. This is where the SECC data fail big time. In its enthusiasm, the MRD decided to go for the big enchilada — incomes of rural India rather than the more prosaic consumption data collected by the NSS. The NSS does collect income data, but the data collected are only for wages and salaries, and data omitted are for profit, rent, interest income, pension, etc. Hence, a conservative interpretation is that the NSS understates true rural income for the people whose incomes are reported.

In contrast, the SECC data are likely to overstate household income because it reports only the earnings of the highest earning member of the household. One further overstatement in the SECC relative to the NSS: the former is an average for the period of July 2011 to 2013; the latter is for the agricultural year July 2011-June 2012. On average, the SECC 2011-13 income data are likely to be 14 per cent higher (9 per cent inflation and 5 per cent real growth) than the NSS 2011-12 data.Despite the considerable overstatement involved in the SECC, it still reports lower rural incomes than the NSS. On a comparable basis (like with like, household with household, and all in 2012-13 prices), mean average household income is about 70 per cent of mean NSS income. (If we go into comparison with household income as compared to national accounts, NA, a rough guess-timate is that SECC incomes are only a quarter of those revealed by NA!). One other clue to the poor quality of the SECC data: Mean rural SECC household incomes are only 85 per cent of mean rural NSS consumption.

Land distribution: According to the SECC, 56 per cent of the house olds are landless, that is, they have ownership of land less than 0.01 hectare. The NSS estimate — a considerably lower percentage of 26.3. Actually, the SECC estimate is almost half that of the NSS!

Casual labour: The NSS and SECC are very far apart on landless casual labour. Under the SECC, the major source of income of more than half of the rural population is casual labour — the NSS has 35 per cent.

The SECC has 30 per cent of households with income from casual labour and not possessing any land (less than 0.01 hectare), while the NSS has only 1.8 per cent of such households.

The only fair conclusion is that the MRD’s SECC data are not worth discussing, let alone analysing. What about the caste data? Release it you must, but only after processing it and, more importantly, validating it. A true picture of Indian caste distribution will have to await a Census-conducted census, and not one processed by the MRD.


Bhalla is a contributing editor for ‘The Indian Express’ and co-author, with Ankur Choudhary, of ‘Criconomics’.