4 min read.Updated: 28 Nov 2019, 11:15 PM ISTSurjit S. Bhalla,Karan Bhasin
Big inconsistencies in NSO’s Consumer Expenditure Survey suggest it’s beyond redemption but its data must still be released
Some results of the Consumer Expenditure Survey (2017-18), CES1718, were leaked in a newspaper report. Soon after, a press release issued by the government stated that the draft report based on CES1718 was not considered to be of acceptable quality and would not be part of an official release. No sooner than you can say “censored", media was abuzz with dire forecasts of the end of freedom as we know it.
It is important to note two points about the controversy. First, there has been no mention by any government official that the unit-level data of the survey, on which the draft report was based, would not be released. If unit-level data is released, as we hope it will be (and the government has already released parts pertaining to health expenditures), then all chest-beating about censorship should end. Whether the data is good or bad, informative or useless, can be determined by an army of researchers worldwide.
The reason given by the government for dumping the report was that there was “a significant increase in the divergence in not only the levels in consumption pattern but also the direction of change when compared to other administrative data sources like the actual production of goods and services" [emphasis added]. Translated: the CES data appeared to be junk.
There is still an ostensibly legitimate concern that the draft report must be published because the government has spent millions of rupees collecting the data and hence is obliged to publish it. We show that other high-frequency and administrative data published by the government (and on which far more money was spent) sharply contradicts the information contained in CES1718. The money spent on CES will not be wasted if the report is not published, but it will be a waste if we, and our statistical authorities, do not learn anything from the mistakes made. The sad reality is that we have seen the writing on the wall of the “badness" of National Statistical Office (NSO) consumption surveys for more than two decades now, and spanning governments of all colours and dispositions. In other words, assessment of NSO credibility, or lack thereof, is not a political matter—it is just non-political “numbers". So please judge for yourself.
What is the problem? In the left corner is the CES data, which shows that almost for the first time ever, per capita real consumption has declined in India. Not a decline in the growth rate, but in absolute levels. Stated simply, the average Indian was consuming 4% less in 2017-18 than she did six years ago (in 2011-12). In the right corner is the summary data as published by the national accounts division of the same ministry as the NSO. This data shows that rather than declining, average consumption increased by 37% between the two NSO years. Both cannot be right: a decline of 4% versus an increase of 37%. The figures are so divergent that they cannot be of the same country at the same time.
Sadly, many Indian “experts" judge any data via a political lens. If the data speaks poorly of India, and if your prior assumption is that the government is “bad", you defend the data at the cost of your credibility. For example, even before the ink had dried on the leaked draft report, some experts said that malnutrition and poverty, and worse phenomena, were on the rise in India in 2017-18. Again, only two years of comparison are involved, 2011-12 being the other. It turns out that the marketable surplus of food brought to the market by farmers in 2017-18 was 21% higher than in 2011-12. All the data is in quantities, and therefore in real terms. The farmer was better off and the average poor person had more food to eat. The fraction of income devoted to food declines with an increase in per capita income.
Some other consumption indicators are that car sales rose by 31%, two-wheeler sales by 46%, internet usage by 241%, mobile purchases by 44% over the six years, and smartphone usage rose by 786%. Do you still want to believe average per capita consumption declined between 2011-12 and 2017-18? Keep dreaming, but do recognize that the number of airline passengers went up by 96% over the period when the NSO CES surveys believe average consumption declined.
Also note that the number of people engaged in education above the age of 15 rose from 103 million in 2011-12 to 122 million in 2017-18. You may choose to believe that education expansion is for the privileged few, but you would be wrong. The gross enrolment ratio today is close to 25%, and the gender-parity ratio in college—the proportion of women in college relative to men—was close to unity in 2017-18. As every lower-income tiger mother knows, education of children involves blood, sweat and consumption expenditure etc.—the third item, in particular, has been rising significantly.
Given the bizarre nature of the CES data, we find that the government decision to not certify the report with a “Good Housekeeping" seal is correct. At the same time, and with equal emphasis, it is important to release the unit-level data, if for no other reason than to let the world know just how bad CES data is. Releasing the unit-level data will allow independent academics to identify the problems and perhaps constructive suggestions will emerge so that such bad data is not collected again.
Surjit S. Bhalla & Karan Bhasin are, respectively, executive director, IMF, representing India, Sri Lanka, Bangladesh and Bhutan, and an independent economist. These are the authors’ personal views