ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Corruption in the MGNREGS

There is corruption in the Mahatma Gandhi National Rural Employment Guarantee Scheme, no question about that. But simple indices that claim to measure corruption and make an assessment of interstate levels of corruption can end up offering us a wrong understanding.

COMMENTARY

Corruption in the MGNREGS

Assessing an Index

Martin Ravallion

We need to take a closer look at Bhalla’s “corruption index” to see why it is higher in some states than others. His index is the sum of (i) the participation rate for the “non-poor” less that for the “poor”, and (ii) the share of wage expen diture on the scheme going to the non-poor less

There is corruption in the Mahatma Gandhi National Rural Employment Guarantee Scheme, no question about that. But simple indices that claim to measure corruption and make an assessment of interstate levels of corruption can end up offering us a wrong understanding.

These are the views of the author, and need not reflect those of the World Bank or any member country or affiliated organisation. Useful comments were received from Jean Drèze, Rinku Murgai and Dominique van de Walle.

Martin Ravallion (Mravallion@worldbank. org) is director of the World Bank’s research department and is based in Washington.

Economic & Political Weekly

EPW
february 25, 2012

M
uch concern has been expressed in India’s media about corruption on the Mahatma Gandhi National Rural Employment Guarantee Scheme (MGNREGS) – an ambitious national effort, launched in 2005, to fight rural poverty by providing unskilled work at low wages and on demand. Of course, corruption is hardly unique to this scheme. However, the fact that MGNREGS is intended to fight poverty adds to the indignation about corruption.

The relative performance of India’s states in terms of corruption on the scheme is naturally of much interest. Surjit Bhalla (2012) has created an index of state-level corruption on MGNREGS. He claims an “overwhelming presence of non-Congress ruled states in the top half of performance” (i e, the states with less corruption). He points specifically to two Congress-led states, Andhra Pradesh (AP) and Rajasthan, which have a high value of his index.

To those who have studied MGNREGS, Bhalla’s claims are surprising at first glance. To most observers (the author included, based on my fieldwork since 2005), the administrative processes in AP and Rajasthan have appeared to be quite good. So too have related performance measures. The gaps between survey-based estimates of participation in MGNREGS and the numbers recorded in the official administrative data are much lower for these states than for India as a whole – suggestive of lower leakage – although some non-Congress states also do well by this measure, such as Tamil Nadu (Imbert and Papp 2011). The ability to meet the demand for work also appears to be well-above average in AP and Rajasthan, though (here too) there are non-Congress states that also do well (again, Tamil Nadu is an example) (Dutta et al 2012).

vol xlvii no 8

that going to the poor. So we can write the Bhalla index for state i as:

BhallaNon-poor– Pipoor)+(SiNon-poor – SiPoor)

Ci =(Pi Component (i) Component (ii)

Here PiNon-poor is the participation rate in MGNREGS for the “non-poor” (the proportion of the non-poor who participate),

Non-poor

Pipoor is that for the “poor”, while Si and SiPoor are the shares of wage expenditures going to the non-poor and poor in state i respectively.1 The “poor” are defined by Bhalla as those households with consumption per person (as measured in the National Sample Survey for 2009-10) below the Tendulkar poverty lines produced by the Planning Commission, updated for inflation by Bhalla to 2009-10.

Confusing Mistargeting with Corruption

Let us first consider Component (i). We can all agree that a high participation rate on MGNREGS for poor people, relative to those less poor, is desirable. That is what Component (i) measures. In fact Component (i) – minus one – is known as the “Targeting Differential” (TD) in the literature, and it is thought to be a relatively good indicator of performance in reducing poverty (Ravallion 2009). The TD for MGNREGS of 0.12 (on a scale from –1 to +1) is not high. (For example, China’s Di Bao programme – a cash transfer programme targeted explicitly to those with income below the (locally-determined) poverty lines – has a TD of 0.22; see Ravallion 2009.)

However, there is nothing “corrupt” about people living above the Tendulkar poverty line participating in MGNREGS. The Act that created the scheme does not bar those living above any poverty line from participating. Rather it says that anyone who wants work at the stipulated wage rate should get it (up to 100 days per household).

COMMENTARY

Figure 1: Bhalla’s ‘Corruption Index’ Plotted against the Rural Poverty Rate 11% living below the

Bhalla’s “corruption index” for MGNREGS

Dashed line: Smoothed scatter plot for Bhalla’s data points Unbroken line: Bhalla index using instead the all-India parameters for MGNREGS Kerala Karnataka AP TN Rajasthan Gujarat Mah WB UP JhK Orissa MP Bihar Chhattisgarh

80

60

40

20

0

-20

-40

-60 5 10 15 20 25 30 35 40 45 50 55

Headcount index of rural poverty (% below national poverty line) The figure is based on Bhalla’s (2010) estimates for all data, based on the NSS

Employment-Unemployment Survey for 2009-10. (Bhalla’s estimates differ slightly from those in Dutta et al (2012) due to a difference in how the Tendulkar poverty lines were participation in a applied.) AP = Andhra Pradesh; Jhk = Jharkhand; MP = Madhya Pradesh;

scheme such as the

Mah = Maharashtra; UP = Uttar Pradesh; TN = Tamil Nadu; WB = West Bengal.

The “self-targeting” mechanism of a scheme such as MGNREGS tends to mean that families with a relatively high consumption will be less likely to want to do this kind of work at low wages. But some people in families above the poverty line may still want the work. For example, they may have been hit by a shock that will lower their incomes, but this is not yet evident in their consumption (possibly thanks to the scheme). Or the family as a whole may have a consumption-expenditure per person above the poverty line, but one individual in the household needs help from the scheme.

There is unmet demand for work on MGnREGS, as shown in Dutta et al (2012). This creates scope for corruption through the power of local officials to decide who gets work and who does not. However, using the same NSS round as Bhalla, Dutta et al show that, on balance, the rationing process on MGNREGS generally favours the poor, not the non-poor. Of course, there are some local exceptions to this generalisation. But overall it is the non-poor who are more likely to have unmet demand for work on MGNREGS.

There are undoubtedly important relative-poverty effects relevant to all these calculations. The Tendulkar lines were designed for making consistent interstate comparisons nationally. So they try to adjust for cost-of-living differences between states, but not differences in relative poverty. What it means to be “poor” in a state such as Kerala (with only Tendulkar line) may well be understated by the Tendulkar lines. The nationally “nonpoor” in Kerala may well be considered poor in Kerala. And it should not be forgotten that MGNREGS is implemented at the state level.

The upshot of these

observations is that

many factors influence

60

MGNREGS, besides the average consumption of the family relative to the Planning Commission’s national poverty line. That does not mean the scheme is “corrupt” in any meaningful sense. Nor should Bhalla’s calculations convince us that there is a very large amount of leakage to the “nonpoor” when we allow for defensible, broader, concepts of what it means to be “poor”.

Maybe Bhalla’s numbers are just reminding us of the limitations of measuring poverty by a household’s current consumption per person. There is no doubt that consumption is hugely important to economic welfare in India, but it can never claim to capture everything that matters to welfare, and that matters to participation in a scheme such as MGNREGS.

What Then Is Driving the Bhalla Index?

We have seen that there are good reasons to question the relevance to corruption of Component (i) in Bhalla’s index. However, it turns out that this component is not what is driving his index. Indeed, it is easily verified that if there were no differences across states in the TD then one would get pretty much the same values for his index. The correlation coefficient between CiBhalla and the Bhalla index one would obtain if the TD was identical across all states is 0.98.

Observant readers of Bhalla (2012) may have already the main clue to what is really driving the interstate differences

february 25, 2012

in his index, namely, its high (negative) correlation with the poverty rate (r =–.92). Figure 1 plots the Bhalla index against the rural poverty rates across states (using Bhalla’s estimates). Judged by Bhalla’s index, “corruption” on MGNREGS is pretty much a measure of “lack of poverty!”

In fact this correlation is not surprising when we look more closely at the index. Consider now the second component, which is simply 100 – 2Sipoor. By definition we have:

PoorPoor

Pi WiSiPoor = Hi. .

( )( )

Pi Wi

Here Hi is the headcount index of poverty in state i, Pi is the participation rate in the programme for the population as a whole, Wipoor is the average of the wage earnings from the scheme received by poor participants, and Wi is the overall average for all participants. As we have seen, the participation rate for the poor is greater than that for the population as a whole. This is true in every state. While Bhalla does not give the wage ratio (the last term in parentheses in the above equation), it is possible to back it out from the numbers he does give. The wage ratio for India as a whole is 0.90, and it does not vary much across the states either. And the wage ratio turns out to be negatively correlated with

poor/Pi

pi (r = –0.64). So the two effects in parentheses are partially offsetting each other. The main thing driving the differences between states in the share of wage expenditure going to the poor is thus the poverty rate; the correlation co

poorand Hi

efficient between Si is 0.89. As one would expect, the states with a low share of wage expenditure going to the poor when judged by a common national poverty line are by and large the states with low poverty rates.

Figure 1 also gives the value of the index for each state if MGNREGS had the same performance attributes as the all-India parameter values reported by Bhalla. Then the only reason for differences in the index is the poverty rate, and the index declines smoothly with the latter. By comparing this version of the Bhalla index with his original we see something new: the scheme is

vol xlvii no 8

EPW
Economic & Political Weekly

COMMENTARY

actually working to bring down his index in poorer states, relative to what one would expect if the scheme worked exactly the same way everywhere. While this is not a message Bhalla found in his data, it is there.

It is not the fact that AP and Rajasthan are led by the Congress that leads to a high value of Bhalla’s “corruption index”, but their lack of poverty relative to other states. As is clear from Figure 1, his index is not in fact any higher, or lower, for AP and Rajasthan than one would expect, once one controls for the poverty rate.

There is clearly corruption in MGN-REGS, as in many public programmes, and in countries at all stages of development. But let us not pretend that Bhalla’s index has taught us anything credible about that problem.

Note

1 I write the index the way Bhalla describes it. In his table he multiplies it by minus one, but this is potentially confusing so I will not do so here.

-

-” “

-

----

References

Bhalla, Surjit (2012): “No Proof Required: Corruption by Any Other Name”, Financial Express, 4 February.

Dutta, Puja, Rinku Murgai, Martin Ravallion and Dominique van de Walle (2012): “Does India’s Employment Guarantee Scheme Guarantee Employment?”, World Bank.

Imbert, Clement and John Papp (2011): “Estimating Leakages in India’s Employment Guarantee” in Reetika Khera (ed.), The Battle for Employment Guarantee (New Delhi: Oxford University Press).

Ravallion, Martin (2009): “How Relevant Is Targeting to the Success of the Antipoverty Programme?”, World Bank Research Observer, 24(3): 205-31.

“”

-

--

-

-

-

-

Economic & Political Weekly

EPW
february 25, 2012 vol xlvii no 8

To read the full text Login


To know more about our subscription offers Click Here.

Comments

(-) Hide

EPW looks forward to your comments. Please note that comments are moderated as per our comments policy. They may take some time to appear. A comment, if suitable, may be selected for publication in the Letters pages of EPW.

Back to Top