Robert Aumann is a famous economist, usually described as a mathematical economist. He won the Nobel Memorial Prize in Economics in 2005. Many years ago, I attended one of Bob Aumann’s seminars. He remarked, “Life as a mathematical economist is difficult. If mathematics comes up with a counter-intuitive result, people feel the maths must be wrong. If mathematics comes up with an intuitive result, people ask, why did you need fancy mathematics for what is obvious?” Particularly on the first part of the statement, I find a similarity in the way people react to data and statistics. Data is the plural of datum. It means something that is “given”. In Sanskrit and many Indian languages, we would say “datta”. I once heard a speaker seriously suggest that data is the plural of anecdote. That’s not true. Anecdote has a different etymological root and means something that is not to be published. Perhaps one might say—not fit to be published. Yet, when a census or a survey throws up numbers, and it doesn’t tally with our intuitive anecdotal perceptions, we instantly reject it. “I spoke to a cab driver and he doesn’t believe this,” is sufficient to immediately reject a proposition. Any sample survey, which isn’t complete enumeration, unlike a census, can certainly be subjected to a critique on grounds of sample size and design. But any sample size will typically be preferable to a cab driver, with a sample size of one.
The immediate trigger for this comment is National Family Health Survey (NFHS)-5. The survey has been going on since 1992–93, when NFHS-1 was held. NFHS-5 was held in two phases, June 2019 to January 2020 and January 2020 to April 2021. (Some states/Union territories were covered in Phase-I, others in Phase-II.) NFHS-5 covered 6,36,699 households. This sample size is large. Therefore, even though a sample, by definition, is not the same as the population, the size is large enough to eliminate biases caused by size. NFHS isn’t the equivalent of a cab driver response. Why do we have something like NFHS? To obtain data on health and nutrition, disaggregated to the level of districts. We want to take stock of developmental targets at a single point in time and wish to track improvements (or deterioration) over time. To gauge improvements over time, ideally, we should have what statisticians and economists call a panel. In a panel, across time, questions are asked to the same individuals/households. For something like NFHS, that’s not possible. In addition, for NFHS-5, compared to NFHS-4 (2015–16), additional questions have been asked. For those questions, gauging improvements over time is naturally impossible. NFHS-5 occurred in the midst of the Covid pandemic. What does such an exogenous shock do? Does it make matters worse? We don’t know.
India is a large and heterogeneous country and there will be differences between states and differences between districts within them. Subject to this, at an all-India level, it is impossible to dispute improvements between 2015–16 and 2019–21. School attendance, literacy, internet usage, sex ratio, birth and death registration, households with improved drinking water, improved sanitation and clean fuel, insurance, age at marriage, infant and child mortality rates, maternal and child health, institutional delivery, vaccination—you take your pick. Data, as opposed to anecdotes, show a clear improvement in these social sector indicators and some (Ayushman Bharat, maternal health through Pradhan Mantri Surakshit Matritva Abhiyan) were yet to fully kick in when the NFHS-5 questions were asked. Articulated in a different way, this is what the Economic Survey sought to capture by constructing and measuring improvements in a Bare Necessities Index. In passing, the total fertility rate has declined to 2, a shade below the replacement level of 2.1. In simple terms, the demographic dividend will soon be over.
Everything is not unmitigated good news. There are under-5 children who are stunted (measured as height), wasted (measured as weight), severely wasted and under-weight. Except for severely wasted (where the decimal point difference is marginal), it’s improving over time, but not fast enough. And we now have an additional problem of over-weight under-5 children and obese men and women, all relatively urban phenomena. If health indicators are improving, it is not obvious why anaemia should increase, both among children and adults. There won’t be satisfactory answers if one looks only at all-India figures. One has to take it down to the level of states and districts. Take Gujarat as an example. Children aged 6–59 months who were anaemic were 62.6% of the child population in 2015–16, but 79.7% in 2019–21. For women aged 15–19 years, the deterioration was from 56.5% to 69.0%. It’s not just the deterioration, but its scale. In West Bengal, the deterioration for children aged 6–59 months was from 54.2% to 69.0% and for women aged 15-19 years, it was from 62.2% to 70.8%, both alarming drops again. I have chosen two different states for purposes of contrast and both exhibit this sharp deterioration in anaemia. One shouldn’t accept the good news from NFHS-5 and reject the bad on grounds of sample design or methodology. Logically, that’s not acceptable. We know reasons for the prevalence of anaemia—not enough iron, folic acid, Vitamin B-12 and so on. But over time, why has it become worse in the last four years? Could subsidised rice and wheat, or penetration of packaged foods have something to do with it? I have no idea what the answer is. But we need one.
(Views are personal)
Bibek Debroy
Chairman, Economic Advisory Council to the PM
(Tweets @bibekdebroy)