Is the National Sample Survey (NSS) Organization, the statistical arm of the Government of India, seeking to play a “statistical fraud” by saying that overall Muslim poverty levels in rural Gujarat have gone down from 31 per cent in 2004-05 to 7.7 per cent in 2011-12? It would seem so if one looks at the NSS’ 2011-12 extremely small sample size for Gujarat’s OBC Muslims – just five households! Based on this sample size, the NSS’ unreleased report on monthly per capita expenditure (MPCE) for socio-religious groups suggests that the OBC Muslims’ poverty rate went down from 40.5 to 1.9 per cent during the seven years. Rajiv Shah analyses:
In the second week of March 2014, a few newspapers carried a report that Gujarat had emerged as one of the top states in India where poverty among Muslims in both rural and urban areas has declined drastically in the last seven years. This was reportedly brought to light by a high-level committee formed under the Jawaharlal Nehru University professor, Amitabh Kundu. The committee was appointed by the Government of India in August 2013 to evaluate the socio-economic condition of the Muslim community during the period following the implementation of the Sachar Committee report on the condition of minorities in India, released in 2006.
Basing its findings on an unreleased National Sample Survey (NSS) Organization report on monthly per capita expenditure of socio-religious groups in 2011-12, the media reports said, the committee has highlighted that that, in rural Gujarat, the number of Muslims below poverty line (BPL) fell from 31 per cent in 2004-05 to a “measly” 7.7 per cent in 2011-12. Comparison of the same period reveals that those in the BPL population in urban areas also declined, from 42 per cent to 14.6 per cent. This happened at a time when the maximum MPCE (an indicator of economic well-being) of Muslims in rural Gujarat went up from Rs 209 to Rs 291. In urban areas, it shot up from Rs 259 to Rs 328 in the last seven years.
“I am neither a BJP supporter nor a Narendra Modi fan, but as far as the welfare of the Muslim community is concerned, the facts say that Gujarat did much better in the past seven years compared to other states,” Prof Kundu was quoted as saying. A right-wing website went gaga over the findings, and it reported how National Sample Survey (NSS) statistics have now confirmed that poverty level among Muslims in Gujarat has declined significantly in comparison to other states, “presently ruled by the parties which fight polls with agenda to lift the condition of people of the community.” The website also highlighted that in rural Gujarat, Muslim poverty has come down to 7.7 per cent – less the number of poor Muslims in rural Kerala, eight per cent!
There was nothing new in what the reports were been saying. A working paper by well-known Columbia University professor Arvind Panagariya and Vishal More, “Poverty by Social, Religious & Economic Groups in India and Its Largest States: 1993-94 to 2011-12” (October 2013), referring the unreleased MPCE data (quoted by the committee headed by Prof Kundu), had said, “In as many as seven out of the sixteen states for which we can credibly estimate poverty rates for both communities, the poverty rates for the Muslims have dropped below those for the Hindus. The seven states are: Andhra Pradesh, Gujarat, Jammu and Kashmir, Kerala, Madhya Pradesh, Rajasthan, and Tamil Nadu.”
The working paper adds, “In four of these seven states, Andhra Pradesh, Jammu and Kashmir, Kerala and Tamil Nadu, the poverty ratio for the Muslims is below ten per cent. In the fifth, Gujarat, at 11.4 percent, it is only marginally above the ten percent mark.” The paper doesn’t stop here. It says, the “perhaps the single most striking feature” found, on the basis of the 2011-12 NSS data, is, “In the rural areas, Gujarat leads with the lowest poverty ratio of 7.7 per cent for the Muslims. Other states with lower poverty ratio for the Muslims than Hindus are Andhra Pradesh, Bihar, Jammu and Kashmir, Kerala, Madhya Pradesh and Rajasthan.”
Well-known economic commentator Swaminathan S. Anklesaria Aiyar, while analyzing the Pahagariya-More paper, was the first to see that there was something wrong with the NSS data, based on which the Kundu committee appeared to have reached its conclusions about a steep decline in the poverty levels in Gujarat. Writing in his blog swaminomics.org, he agrees that there may be good news for Muslims, as their poverty ratio declined in this seven-year period by 18.2 per cent, faster than the 15.6 per cent for Hindus, even though “the absolute level of Muslim poverty remains higher than of Hindus”.
Now coming to the “whopping surprise” of Panagariya and More – that “in as many as seven states, Muslims are less poor than Hindus” – Aiyar says, “Astonishingly, the rural Muslim poverty ratio is lowest of all in Gujarat, at 7.7 per cent. This is much lower than the urban ratio, and so raises the question whether it is a statistical quirk.” He adds, “Muslims in Gujarat were victims of terrible violence in 2002, and many still feel unsafe.” He sounds a note of caution with the data used by Panagariya-More to reach their conclusions: “The year-wise variations in some states look very steep, and seem to need a pinch of salt”. In Aiyar’s view, Panagariya and More conclusions were guided by the “rapid GDP growth in the seven years between 2004-05 and 2011-12”.
A deeper look into the data, which are part of the still unreleased Kundu committee report, suggests a much deeper malaise than what Aiyar has noted — “a statistical quirk”, which “need a pinch of salt.” A breakup of the Muslim poverty ratio suggests that in rural Gujarat (where the poverty ratios in are down to 7.7 per cent), other backward classes (OBC) Muslims’ poverty levels were 40.5 per cent in 2004-05. And these went to just 1.9 per cent in 2011-12! As against to this, the Other Muslims’ poverty levels in rural Gujarat have gone down at a much lower pace – during the same period, these have gone down from 23.2 per cent in 2004-05 to 18.7 per cent in 2011-12.
These data form part of an unpublished paper, prepared by Prof Kundu, who has clarified, through an e-mail, that there is no “typographical error” here, but are a result of “standard error”. To quote him: “The standard error in the estimates for different socio-economic groups in rural and urban areas would be high due to small sample size in the NSS.” He has said, “I looked at the disaggregated data for different socio-religious categories at the state level. The sample size for OBC Muslims in Gujarat, for example, is very small, even smaller than the size in 2004-05. I would therefore not be very confident in drawing conclusion for these sub-categories at the state level.”
And what is the sample size of OBC Muslims in Gujarat, based on which their poverty has allegedly gone down from 40.5 per cent to 1.9 per cent in the seven years in question? Prof Kundu says, “The sample size for OBC Muslims is five in 2011-12 and 25 in 2004-5. For total Muslims also the size is much below the acceptable level and with high standard error.” Hence, he points out, “The committee did not want to comment on the state-level estimates for socio-religious groups.” He adds, “Also, the growth rate in consumption expenditure and poverty reduction figures can be misleading due to their problem of uneven base.”
Prof Kundu further says, this is not just true of Gujarat but several states. To quote him, “The data for estimation of Muslim OBC poverty is below ten for a few states.” Overall, too, the Muslim sample size of OBC Muslims and Other Muslims taken together for several states is extremely small, hence it is impossible to reach a conclusion. This is one reason, he explains, that the committee under him has “not used the state-level NSS data for drawing inferences regarding the conditions of the Muslims.” Muslims’ small sample size is very small many states – it is “less than 100.” He underlines, “It is important that people discussing the results know about this.” Gujarat-based activists, working among poor Muslims, who suffered during the 2002 riots, believe, the data suggest nothing but a “statistical fraud.”