A tale of two numbers
It is possible to overcome the issues by improving the survey methodology.
The only India-wide survey we can use to measure actual MGNREGS employment is the NSS.
For my first field visit to study the Mahatma Gandhi National Rural Employment Guarantee Scheme (MGNREGS) a few years ago, Nikhil Dey took me from Jaipur to Rajsamand, where I met a team from the Mazdoor Kisan Shakti Sangathan (MKSS) and the block officers they worked with. The block officers explained how the details of each day of work provided under the MGNREGS was entered online at nrega.nic.in. This seemed to me to be a fantastic effort to promote transparency in public programmes. At that point, Shankar Singh, an MKSS member, quipped, “Bhains ko chahiye chaara, sarkar ko chahiye aankda (The government needs numbers like cattle need fodder)”.
Six years later, the debate around corruption in the MGNREGS is still raging, and it is all about numbers. In his article ‘Move from NREGA to cash transfers’ (IE, December 12), Surjit S. Bhalla suggests that two-thirds (one-third under “generous assumptions”) of MGNREGS employment may be fake and concludes that the programme deserves the axe. In a more nuanced approach, economist Abhijit Banerjee in “What’s the plan for MNREGA? Modi needs to explain” (Hindustan Times, November 26) quotes a much lower figure, of 20 per cent leakage, down from 50 per cent in 2007, to argue that the government’s efforts to reduce corruption in the MGNREGS have been successful.
A tale of two numbers can explain the origin of these differences. The first number is the official number of MGNREGS days, based on data entries on the official website. This number reflects what the Central government, or the Indian taxpayer, pays for. But part of these days are fake — they are “ghost days” used by corrupt officials to siphon off MGNREGS funds. How much of it is fake? It is hard to tell, unless one compares it to a second number, a measure of “actual days”, which can only come from an independent survey done on the ground.
The only India-wide survey we can use to measure actual MGNREGS employment is the National Sample Survey (NSS) employment-unemployment survey, carried out in 2007, 2009 and 2011. It asks people what they have done each day of the last week, and among others, includes two categories: “NREGA work” and “Public works other than NREGA”. How do interviewers and respondents tell the difference between the two? My experience is that it is difficult: in front of a respondent who had no clue about “Narega kaam”, I often ended up asking about “mitti ka kaam” or “sadak ka kaam”.
Because of this, we end up with not one but two numbers to compare official figures to. One is probably too low, as it counts only the days that have been reported as days of MGNREGS work. The other is likely to be too high because it counts employment on all public works, including, but not limited to, the MGNREGS. These numbers yield very different estimates of the proportion of MGNREGS days that are “real”. For 2011, the first is close to 33 per cent, and the second close to 80 per cent. Bhalla uses the lower bound, while Banerjee uses the higher bound. So they come to radically different conclusions.
There is no simple way to know what the true number is. To take one example, in Bihar, the number of public works days in the NSS is higher than the official number of MGNREGS days. This is due to an ambitious rural road construction programme that has nothing to do with the MGNREGS. In this case, the higher bound estimate is clearly too high (no leakages). But the number of MGNREGS days in the NSS is close to zero. Awareness about the MGNREGS is so low in Bihar that workers do not know that they are on an MGNREGS worksite. Hence, in this case, the lower bound is far too low (as it suggests 100 per cent leakage).
It is possible to overcome these issues by improving the survey methodology. Puja Dutta, Rinku Murgai, Martin Ravallion and Dominique van de Walle carried out a survey on the MNGREGS in a representative sample of Bihar villages. They obtained a list of MGNREGS works from local officials and asked respondents about how much they worked on specific projects. They find that 70-80 per cent of MGNREGS employment reported in official data is independently confirmed by their survey. However, using the same method in the NSS would be a formidable challenge.
Another possible way out would be to compare employment on all public works in the NSS with employment on all public works programmes according to official sources. This would allow us to compute an estimate of fake employment in all public works, including other national schemes, such as the Pradhan Mantri Gram Sadak Yojana, and state schemes such as the Bihar road construction programme. Unfortunately, this data is not readily available, because unlike the MGNREGS, most public works programmes involve private contractors, and we do not know how much labour they hire.
This tale of two numbers does not lead to the conclusion that this exercise is useless. The government, and the public debate, needs numbers like cattle need fodder. But it is important for the government to know what it is being fed.
The writer is a research fellow at the Department of Economics and Nuffield College, Oxford University
Share
