Good quality or bad, data always has plenty to reveal: Let data speak

But this is what one expects from politicians. Strangely, though, one has witnessed a parallel development among esteemed economists and policy analysts. I will just summarize the data arguments to show that some of the conclusions drawn by my tribe are even more bizarre than those drawn by politicians.

GDP and consumption growth are important factors that determine voting choices. Ten of the last 14 Lok Sabha elections can be explained by observing whether per capita GDP growth during the incumbent’s term was above or below 3.25% (close to 5% GDP growth). This result has not been lost on the economic critics of the Narendra Modi government, including several renowned foreign experts and foreign publications. The narrative is the same: Don’t believe the data, because data that is favourable to the government’s view of the economy is likely to be exaggerated and probably even false.

S. Subramanian (The India Forum, ‘The Household Consumption Expenditure Survey 2022-23’) accurately reflects this scepticism when he bluntly states that worries about the HCES 2022-23 results are justified because of the government’s track record: “A part of that experience is reflected in the delay in the release of the 2017-18 Periodic Labour Force Survey report which carried information on a record level of unemployment in India… in the suppression of the NSSO’s 2017-18 survey on consumption expenditure."

But this narrative has no time for logical objections. Note its emphasis on the delay in the release of PLFS 2017-18. That survey showed record high unemployment rates. Also, leaked results of the “suppressed" 2017-18 HCES showed a record steep decline in real consumption of 5.5% between 2011-12 and 2017-18. A 5.5% decline in a consecutive year would be bizarre, but somewhat plausible (it did happen in covid year 2020-21). But a decline of this magnitude after five years of positive economic growth (2011-12 to 2016-17; two years of United Progressive Alliance rule and three years of National Democratic Alliance rule) would imply an approximate decline of 25-30% in one single year, 2017-18. A decline of such dimensions in such a short period would have led to widespread starvation or even famine across the countryside. And, given its all-India nature, the worst catastrophe to befall India ever—worse than World War II and also the Bengal famine.

Over the last five years, what scholarly questions have been raised over the results of the 2017-18 HCES survey? And isn’t it peculiar that many of the scholars who staunchly believed the authenticity of the 2017-18 HCES survey now equally strongly believe that the HCES 2022-23 survey is not comparable with the 2011-12 survey? Because identical questions were spread over three visits rather than one? If you believe that to be the real reason, then, as they say in New York, I have a bridge in Brooklyn to sell.

It now turns out that the International Labour Organization (ILO), which conducts labour-force surveys that get the same respect and authority as the World Bank’s surveys and studies on poverty around the world, has now deemed the 2017-18 and 2018-19 PLFS results to be, well, worthless.

Here goes the ILO’s explanation: “In the model of labour force participation, the PLFS observations for 2018 and 2019 have been excluded as they appear to present limited comparability with both the previous NSS results and the newer PLFS results. Given the country’s size, this has a sizeable impact [on] the global aggregates," (shorturl.at/elst5)

We have this rather unusual reality of what happened with the first national surveys conducted in India after the 2014 general election. The 2017-18 HCES survey was not released by the Indian government on grounds of bad-quality data. The twin PLFS editions of 2017-18 and 2018-19 (‘twin’ because these surveys are conducted over eight quarters) are now deemed to be of uniquely bad quality by not the Indian government, but the United Nations’ authority on labour markets, the ILO. The probability of these two statistical events happening by chance is close to the probability of my being able to find a buyer for the Brooklyn bridge.

Two important lessons are to be drawn from this analysis. First and foremost, we should all let the data speak. This is relevant for all of us—government officials and scholars included. Only by the release of data, specifically the PLFS results for 2017-18 and 2018-19, was the ILO able to determine that for these years (and only these years), the survey data was not usable. In my long experience with international and national data, this is a most unusual condemnation and should lead to introspection among all. Second, we shouldn’t let our prejudices determine our conclusions. Sooner or later, the data will speak, and the ‘truth’ will be out. So let no one have an opportunity to sell a Brooklyn bridge.