Corruption and the Mahatma Gandhi National Rural Employment Guarantee Act

Corruption continues to strangle India’s public finances. This column presents evidence of embezzlement in India’s flagship rural employment programme and suggests new ways policymakers can test what works in the struggle against corruption.

Social programmes in India are beset with corruption. Former Prime Minister Rajiv Gandhi once suggested that only 15% of spending on such programmes actually reached the intended beneficiaries. The Planning Commission of India in 2005 estimated that a similarly low proportion, 27%, of government transfers actually reach the poor. Such embezzlement adds to the strain on state finances, and may make these programmes more regressive than progressive (Olken and Pande 2012, Niehaus and Sukhtankar 2012b)

Given these problems in the past, it is crucial to understand whether and how corruption is impacting India’s biggest new social venture, the Mahatma Gandhi National Rural Employment Guarantee Act (MNREGA). The act is supposed to guarantee 100 days of paid employment to rural Indian households, with no eligibility criteria other than being an adult willing to work. The scale of this programme is enormous: in 2010-2011, the central government allocated Rs.401 billion, or $8.9 billion, or 3.6% of government spending, to the programme. This puts it on par with other large developing country social programmes such as Oportunidades (formerly PROGRESA) in Mexico. It is the largest ‘workfare’ programme in the world, dwarfing comparable programmes in Argentina or even the Earned Income Tax Credit Scheme in the US in the number of people it reaches.

In light of India’s well-publicised recent public finance difficulties, reducing the additional burden imposed by corruption in costly programmes like MNREGA takes on massive importance. But, as I discuss here, this task will be far from easy.

Understanding corruption in MNREGA

How does corruption from MNREGA occur? Theft from the materials budget has not been carefully studied, but by law materials costs can constitute at most 40% of the total and in practice they are even lower. Theft from the labour budget has been studied by a number of authors. To understand how it works, consider the following hypothetical scenario. Suppose the statutory wage set by the government is Rs.100 a day. The labourer works for 5 days, but instead of paying him Rs.100 daily, the official pays him only Rs.90, keeping the remaining Rs.50 ((100 – 90) x 5) for himself. Let us call this form of corruption ‘underpayment’. In addition to underpayment, the official also reports to the government that the labourer worked for 10 days rather than 5, and pockets the additional Rs.500 (100 x 5). We’ll call this ‘over reporting’. Under this breakdown, work done by ‘ghost’ workers (people who don’t exist) is also considered over reporting, as would any collusion by workers and officials to extract payments from the government for work that is never done.

Underpayment and over reporting differ markedly in terms of whom they hurt and hence who might monitor these types of corruption. Underpayment directly hurts workers, and hence we might expect workers to complain if they are underpaid. Over reporting does not directly hurt workers – in fact workers are probably unaware of it – but it does hurt taxpayers and increase the overall fiscal burden of the programme. Hence monitoring over reporting falls under the jurisdiction of officials in charge of supervising the programme.

Both monitoring from above and below have been shown to be effective in combating corruption in other contexts. For example, Olken (2007) finds that increasing the probability of top-down audits reduced theft from a road construction programme in Indonesia by 30%. Bjorkman and Svensson (2009) find that a bottom-up community based monitoring programme was effective in reducing shirking by public health officials. The evidence in India, however, is not so promising. Banerjee et al (2008) find that a programme to monitor absenteeism in nurses failed due to the reluctance of the administration to impose penalties on flouting nurses.

The Bad

To measure the extent of over reporting and underpayment in MNREGA, we conducted a survey of about 3,000 households who were reported as having worked on the programme in three districts in Orissa and one district in Andhra Pradesh, and compared household reports of work done and payments received to official reports of these quantities. Our results indicate that at least at the time of the survey (2007-08), corruption was very high, on the order of 75%-80% of reported expenditures, with the vast majority of it coming from over reporting (Niehaus and Sukhtankar 2012a). Underpayment, while initially close to zero on average, increased with an increase in the statutory wage in May 2007, as none of this increase was passed on to workers. (Niehaus and Sukhtankar 2012b)

These results must be interpreted with a large dose of caution, however. Our figures are from somewhat remote districts of Orissa and Andhra Pradesh and also focus on a particular period in time, hence cannot be viewed as representative of the programme in general. Moreover, since we relied on recall data we cannot determine the absolute levels of corruption with precision. Nonetheless, our figures are in the same ballpark as those found by studies conducted by NGOs (Center for Science and Environment, for example) as well as the Planning Commission report cited earlier. Moreover, a quick search of the Comptroller and Auditor General of India’s website (cag.gov.in) reveals numerous accounts of irregularities found in MNREGA implementation across nearly every state in India, and newspaper articles noting various accounts of fraud are too numerous to mention.

The Good

The focus on corruption may sometimes obscure the fact that MNREGA does provide employment and a source of income for a large number of India’s poor, serving the act’s screening and social insurance goals (Engler and Ravi 2012, Gaiha 2011). Is it possible to reduce corruption and make the achievement of these goals cost effective? A number of pieces of evidence suggest so.

First, we find that officials are sensitive to people’s concern for their own careers. When opportunities for corruption tomorrow increase, officials are cautious about being corrupt today so that they may have the opportunity to be corrupt tomorrow. This, along with the fact that over reporting dramatically declines on public holidays, suggests that officials are sensitive to monitoring from above (Niehaus and Sukhtankar 2012a).
Second, we find that although on average none of the increase in the statutory wage passed through to workers in Orissa, in places where NGOs were active there was indeed some pass-through. This suggests that pressure from below could perhaps also be successful (Niehaus and Sukhtankar 2012b).

Also encouraging is the promise shown by ‘social audits’, public meetings where expenditures and accounts on the programme are scrutinised. Afridi (2008) describes how these have worked in Rajasthan and Andhra Pradesh and suggests that results from case studies are promising.

The Ugly

Unfortunately, however, the responses to community monitoring across India provide a cautionary tale. For starters, it seems as though the punishment for exposing corruption is worse than the punishment for being found to be corrupt: an activist who showed up fraud in MNREGA was murdered in Jharkhand (Vanaik 2008), while the worst punishment meted out to perpetrators seems to be having to return the money and be suspended from their jobs. Depressingly such stories seem to be common across programmes in India – Bihar and Maharashtra have also seen cases where whistle blowers in road construction and the Targeted Public Distribution System have been murdered. Moreover, powerful lobbies have ganged up against monitoring, as the Rajasthan example, where a group of local officials brought a court case against social audits being run by NGOs, has shown. Hence, unless governments show that they are serious about punishing corrupt activity, it is likely to continue with impunity.

So what can we do?

One possible solution involves making payments directly into beneficiary accounts, using the technological infrastructure provided by a biometrically-authenticated system like Aadhar. This has the potential to reduce underpayment, and also over reporting that takes place without the knowledge of the beneficiary or through ghost workers. However, this solution is not fool proof: it may simply encourage collusion, or displace corruption into other aspects of the programme or other programmes. Moreover, it may have negative impacts on participation if beneficiaries do not enrol for the biometric IDs or lose them.

Whether technological solutions can be successful will only be known after a careful evaluation of such interventions; the government of Andhra Pradesh is collaborating with Karthik Muralidharan, Paul Niehaus, and I on a rigorous randomised evaluation trial to determine the effect of Smartcards on MNREGA payments. Insights from this evaluation will inform the implementation of the programme across India, and possibly prevent the need to start rationing jobs for the rural poor (Ravallion et al 1993).

The author thanks Paul Niehaus for comments and Anuraag Girdhar for research assistance.