There has been considerable discussion in these pages on the accuracy and/or credibility of the GDP data for March-June 2026. My guess is that internationally also this debate has assumed importance. Given this, I hope I can be excused for penning another op-ed on the subject.
It is important to separate the commentary on Indian data into two broad categories — national accounts data, and other data. I have had the privilege of being a member of several government committees since 1998. In 2006, I was appointed a member of India’s first statistical commission (2006-2009) under the leadership of the late Suresh Tendulkar. Interacting frequently with staff members of the NSS and national accounts division, I emerged with the following observation on keepers of the statistical frame (and flame): “Don’t build a road because you may have an accident”. Stated more favourably, official Indian statisticians would err on the side of “accuracy” or “precision”. If a road is not built, if knowledge is not advanced, so be it — we purists cannot be wrong.
It is for this reason that India has not undertaken the task of conducting a household income distribution survey. I have begged the authorities to do an income distribution survey since 1998.
Note that in conducting consumption surveys India has been a pioneer in world statistical surveys; and we have not worried about survey consumption ranging between 55 and 60 per cent of consumption estimated via national accounts. India was the first country to outline poverty measurement. The World Bank’s famous dollar-a-day poverty line was the Indian poverty line! The first household panel survey was conducted by NCAER, the Additional Rural Incomes Survey, 1968-69 to 1970-71; an honour shared with the University of Michigan Panel Study of Income Dynamics, which conducted the first panel in 1968.
South Asian countries like Sri Lanka, Pakistan and Bangladesh have been conducting income distribution surveys for years, and India has finally joined the world in this endeavour. Given the dissection and healthy criticism of Indian data, it is important to point out the overwhelmingly unhealthy analysis, let alone criticism, of one of the most successful development transitions in history — China. I myself have published, and discussed, the “fact” that China has shown the fastest-ever poverty reduction over the last 25 years. But how many of us have asked how China obtains this result? And how do the World Bank and IMF treat and analyse Chinese data on poverty?
They don’t. They accept at face value what China provides via deciles of income and consumption, for rural and urban China. What analysis? China conducts household surveys like India does. But no unit-level data has ever been released to the public. My best understanding is that China provides four excel files, one each for consumption and income, rural and urban. There are 12 observations in each file — the 10 deciles and possibly the 5 per cent and 95 per cent mean levels. On that basis, the world continues to applaud the Chinese poverty eradication miracle.
But nations living behind glass should not throw stones at those with steel veneers. One criticism of the Indian statistical apparatus that refuses to go away, precisely because it is unanimous, is this: Something happened to India’s statistics since 2014 – Indian data started being officially censored. That is the primary reason why, despite world-class improvements (not perfection, but we are not a developed country yet), the efforts on the GDP data are, well, questioned.
Examine the most widely known, and acknowledged, instance of censorship. After 2011-12, India undertook a consumption (and employment-unemployment) survey in 2017-18. There was no official report or data release of the 2017-18 consumption survey. After much reluctance, and internal debate, the 2017-18 employment survey was released (PLFS 2017-18). No such luck with the consumption survey. Why? Because the data showed (via leaked reports) that real mean consumption levels had declined by 3.7 per cent over the six-year period between 2011-12 and 2017-18; and rural consumption, where most of the poor are, declined by 8.8 per cent; urban per capita consumption rose by 2 per cent.
There is a lesson from what the Indian government had previously done when survey “findings” were unpalatable. NSS consumption surveys had been on a five-year schedule since 1983. The 2009-10 consumption survey results were released in mid-July 2011. Because of the fear that there was a severe drought in 2009-10 (rainfall data did not confirm a drought) and the dampening effects of the Great Financial Crisis, the government had commissioned an unprecedented additional consumption survey for the agricultural year 2011-12 (July 2011 to June 2012). In other words, a new survey was already in the field when the results of the “old” survey were announced.
The Indian government had acted in the best, pioneering nature of the Indian statistical system. Be conservative (no income distribution survey!) but do not ever censor data.
The tragedy, and mistake, is deeper. I have participated in several analyses of Indian poverty data, including an analysis of what likely happened to consumption and poverty in 2017-18. In a 2019 article, co-authored with Karan Bhasin, we wrote: “It is important to release the unit-level data, if for no other reason than to let the world know just how bad the CES data is”.
I have reiterated that view several times, privately and in print. No need to endorse the data — just release it. The “unintended” consequences of that censorship have become evident since then: No matter how good the data produced in India is, there will always be doubt, one that is exploited by political opponents. There is an easy no-cost solution to this data and credibility crisis — in the name of god, release the unit-level 2017-18 consumption data.
The writer is chairperson of the Technical Expert Group for the first official Household Income Survey of India. Views are personal
Note that in conducting consumption surveys India has been a pioneer in world statistical surveys; and we have not worried about survey consumption ranging between 55 and 60 per cent of consumption estimated via national accounts. 
Share
