The purpose of the statistical system is to generate data. Official data can be of three types—census, survey and administrative. Census means complete enumeration, surveys are based on samples of the population and administrative data are used for government administration and regulation. MCA-21, GST or information on dashboards are instances of administrative data. To gauge cross-country performances, the World Bank (since 2016) has a Statistical Performance Indicator based on five heads: data use, data services, data products, data sources and data infrastructure. India doesn’t do that badly on the cross-country rankings, though it should do better. The World Bank only includes civil registration and vital statistics for data sources, not any other form of administrative data. Administrative data are no substitute for traditional statistics sources such as censuses and surveys. Incentive structures are such that, depending on the indicator, administrative data tend to overestimate. There is a tendency to overplay success.
For example, dashboards will provide a figure on toilets constructed under Swachh Bharat Mission. That figure, on the percentage of households covered, needs to be validated by a survey like NFHS (National Family Health Survey). Criticisms of India’s official statistics are directed at censuses/surveys. Administrative data are fairly robust (apart from the incentive mechanism) and almost real-time. Because of sole reliance on civil registration and vital statistics, the World Bank doesn’t give due credit to India. Conventionally, administrative data have indeed been that—vital statistics and so on. But just as the world is still coming to grips with what Digital India has achieved, the world and the World Bank are yet to come to grips with what administrative data can accomplish today. By the way, we do use administrative data for policy decisions. MCA-21 and GST are both examples.
In contrast, our performance on censuses and surveys is riddled with holes. There are inordinate delays. It isn’t merely a question of the Population Census 2021. The last Agriculture Census was in 2015–16, SECC in 2011–12, MSME Census in 2006–07 and Livestock Census in 2019. The results of the Economic Census of 2019 are still not ready. In this day and age, why does processing take such a long time? Delays also characterise surveys. The last Annual Survey on Unincorporated Sector Enterprises (ASUSE) is from 2015–16. ASI (Annual Survey of Industries) data are for 2019–20. The 2020–21 results haven’t been released. For any estimation of poverty and inequality, we need data on consumption expenditure distributions. The last NSSO (National Sample Survey Organisation) data are for 2011–12. PLFS (Periodic Labour Force Survey) is not quite a substitute. It shouldn’t be a surprise that we don’t quite have a handle on the informal/unorganised sector and its contribution to GDP. Or on services since there is no survey on services. With PPI (Producer Price Index) awaited, the GDP deflator is a problem, and we don’t quite know what is happening to capital formation. Current prices (as a share of GDP) give one kind of answer, constant prices another. Add to that unrealistic samples in surveys, a point raised by my colleague in EAC-PM, Shamika Ravi.
Samples over-represent rural and under-represent urban populations. With urbanisation throughout India and better provisioning of basic necessities in urban areas, faulty samples overstate deprivation.
It isn’t merely a question of sampling frame and design. MOSPI’s Survey Design and Research Division spends 75% of its time writing reports. With a lack of survey statisticians, only 25% of the time is spent on survey design. There are issues with revamping ISS (Indian Statistical Service). Add irrelevant and inconsistent questions with a mechanical implant of questions from abroad. My colleague in EAC-PM, Sanjeev Sanyal, has given examples of stunting, wasting and anaemia. Sanjeev and I have highlighted how questions on mosquito nets have been donor-driven. There is a broader problem. Since there isn’t sufficient domain knowledge within NSSO, it functions through a system of Working Groups. A fresh Working Group is set up every time there is a survey. That recommends a fresh set of questions, often disregarding what was asked before. This not only renders questions non-comparable across surveys, but it also renders them non-comparable for the same survey over a period of time. NFHS is one example of this. MOSPI should read (or watch) Humphrey Appleby’s educational lecture to Bernard Woolley about sample surveys. It might help frame questions better. It shouldn’t be surprising that questions are raised about the timeliness and credibility of India’s official statistics. Nor should it be surprising that these internal data problems are used to shape a false narrative about India globally.
I am not talking about cross-country global rankings based on perceptions but those based on hard objective data. (Think of discussions on poverty, inequality and unemployment.) Janus was a Roman god who possessed two faces. (January is named after him.) India’s official statistics are like that—excellent on the administrative side, mediocre on censuses/surveys.
From January 1, 2024, India will become a member of the UN Statistical Commission (UNSC), one of 24 members. The UN Statistical Division feeds into UNSC, and UNSC feeds into the UN Economic and Social Council. India was chairman of UNSC in 1954, 1956 and 1976—P C Mahalanobis in the first two and V R Rao in 1976. The UNSC meeting in 1976 was held in Delhi. The official Indian statistical system had a lot of credibility then but progressively eroded over the years. India doesn’t lack great statisticians. It is just that once upon a time, they were interested in working within the government. Now, they prefer to remain outside, underscoring an inability to attract and retain talent within the government.
The last time India was vice chairman was in 2003, but by then, India’s voice no longer led to changes in global statistical systems. (National accounts are one example.) That membership, for four years from 2024, shouldn’t become a mere token. It should be a trigger to rehabilitate the official statistical machinery. For that to happen, there must be much reflection and internal churn. Otherwise, among other things, external indicators will suddenly be thrust upon us, and we will end up reacting, as we have repeatedly done in recent times. That is happening for ESG (Environmental, Social and Governance) norms. Since the days of the Rangarajan report in 2001, a lot has been written on deficiencies in the system. Dr Rangarajan’s report was 22 years ago. Some issues mentioned then have still not been resolved.
Bibek Debroy
Chairman, Economic Advisory Council to the PM