A year of data concealment

Yoginder K Alagh | Updated on January 01, 2020 Published on December 28, 2019

NSSO and NAC data may be subject to sampling error, but that is no reason to suppress it

This has been a year of unconvincing statistical practices. I was upset at the Statistical Office in Sardar Patel Bhavan (where I have spent much productive time through the decades) putting the latest Consumer Expenditure Survey of the NSSO under wraps.

On one hand, it is encouraging to see that independent experts have been appointed to the Statistical Commission. The authorities, however, have still to show that they don't put finished statistical reports in cold storage. Not recognising a problem is not a solution.


The National Statistical Office has released the latest quarterly Periodic Labour Force Survey based on standard practices adopted earlier. A senior overseas economist has criticised the NSS estimates for giving lower estimates of aggregate consumption than National Accounts (NAC) estimates. That’s not all.

Subject to errors

The discrepancy between aggregate consumption estimates from the NAC and the NSS, and the statistical controversies around them, go back almost half a century. Then, the Indian Statistical Institute’s respected statistician, Moni Mukherji, in a terminal paper with GC Chaudhari went into it in some detail. They said that the NSS reports give estimates of errors. That holds true even now. PC Mahalanobis, the founder of the NSS method, had in his famous Royal Statistical Society lecture designed experiments with Interpenetrated Networks of Samples, so that sub-rounds of a Round give error estimates.

The government is right. The NSS estimates have errors. In fact, the report itself reports errors. That is no ground to ban it. Incidentally, it is possible that the distribution of estimates by the class of expenditure may not change much between rounds — but that’s only a gut feeling. Since the report is unavailable, there is no way to comment on it.

Mukherji pointed out that the NAC estimates have errors as well. The great statistician Oscar Morgenstern had argued that all statistics have errors. NAC estimates have big errors in income of the unregistered sector and output estimates of items like vegetables, fruits, animal husbandry products and so on. Mukherji said that we have two sets of estimates. Both have errors. We need to use them with that knowledge.

Selective disclosure

It can be argued that the government feels that it is difficult to take the view that errors have increased. But that doesn’t seem to be the case, when the statistics are favourable to it. The government has been quick to release estimates showing a slight improvement in quarterly estimates of employment.

These estimates incidentally use the ‘Usual Status, end of the week’ NSS workforce norm. This is certainly subject to sampling errors. So, how are the employment surveys kosher despite sampling errors, when the consumption surveys are not?

A more sensible position would be to argue that statistics are subject to error, but assessments have to be made with the available data.

Policymaking in a democratic society is a complex business. Economic policy is not the stuff of cloak and dagger secrets.

In that case, policies are declared and knowledgeable discussion is to be avoided. If a critique comes along, it has to be shot down. If statistics are unflattering, they can be put under cold storage. But this is a country in which information will be released from many sources. How many can be choked out?

It has not been the best of years, in terms of data transparency from the government’s side.

The writer is a former union minister

Published on December 28, 2019
This article is closed for comments.
Please Email the Editor