Illustration: Ajay Mohanty
Cryptologist Bimal Kumar Roy
, former director of Kolkata-based Indian Statistical Institute, took charge as chairman of the National Statistical Commission
(NSC) at a time when questions were raised over the autonomy of statistical institutions. In his first interview after being appointed NSC chief, Roy speaks to Somesh Jha
on the recent controversy over the government’s decision to scrap the consumer expenditure survey. Edited excerpts:
The government recently decided to junk the consumer expenditure survey of 2017-18. What do you have to say?
We are not in favour of scrapping the survey report. We are exploring possibilities in terms of adding certain caveats to the report and releasing it in the public domain within two months. This is our target. The government’s concern is that the estimate does not look realistic. We are trying to examine the issues. We have decided to set up two-three committees to look into it.
Do you think it was right for the government not to consult the NSC and scrap the report on its own?
I won’t be able to comment on it. I can take responsibility for my actions. I won’t like to comment on what others did. The government might have thought that the estimates did not look good.
But then it boils down to the autonomy of the NSC…
The government is trying to bring a Bill for giving more autonomy and powers to the NSC. The question of the NSC’s autonomy is not clear at this stage. The government has taken note of this and it is the process of framing a new law.
The convention has been to place the surveys for approval of the NSC before taking any decision. This has not happened in the past few months…
We are a newly constituted body and the survey took place long back. Perhaps, the government would have thought that the NSC is not aware of the background of the surveys.
Don’t you think the NSC should be the final authority for approving the surveys?
It will be difficult for me to comment on this. The government has to take a call on it and we will follow that.
Do you think the unit-level data of the consumer expenditure survey should be put in the public domain immediately, as has been argued by many economists?
As an academician, I am in full support of releasing the unit-level data of all surveys and we will be happy to disseminate it.
Do you think the surveys produced by the NSO have data quality issues as have been talked about recently?
There are some issues observed by the ministry (of statistics and programme implementation). We are working out a way forward.
Can you throw light on some of the issues?
For instance, when you conduct the household consumer expenditure survey, certain expenses do not factor in the social security benefits received through government schemes. If consumers get more social security support, their expenditure on those items on which they have received the benefits will be low. If a below poverty line person gets rice at highly subsidised prices, naturally her expenditure will look low. Though this is not exactly a data quality issue, all these factors need to be explored.
There are a lot of government programmes related to health run by states. The survey shows that health expenditure has probably gone down. Though this looks bad on the face of it, we are trying to find explanations for such causes. Then, other surveys are now being conducted through CAPI (computer-assisted personal interviewing) and one has to be
careful as to how investigators are coping up with this technology.
Further, there is an issue related to the use of contract workers to do fieldwork for surveys. Once the contract tenure comes to an end after the survey, it is hard to reach out to such surveyors in case errors are detected. We are trying to examine the impact of all these factors on all the surveys. It’s an eye-opener for us and we are looking at the previous surveys, too. One important factor is to see how the coefficient of variation differs over time. If it is high, then the estimates are not dependable.
How do you plan to address some of these issues?
We can take the help of technology. For instance, if households report that they have brought a certain amount of flour for consumption, instead of asking at what price they purchased it, we can take the price data from other sources, such as mandi (local market). If we can do that, it will be redundant to ask for the price. The present questionnaire is too long and respondents become impatient in 10-15 minutes. So, the question arises whether we can borrow certain sets of data on expenses from other sources, instead of asking these to the respondents.
But it may still be feasible for food items. Can this be possible for non-food expenses?
We are exploring all options. The average time taken by investigators to complete the schedule (or questionnaire) in a single household is 45 minutes. We want to reduce it by half or to at least bring it down to less than 30 minutes by clubbing a few questions. We will do a pilot survey and see how people are responding to it.
One argument is that if one tinkers too much with the schedule or the set of questions, the data may not be comparable with the past surveys. How do you see that?
In the pilot survey, we will ask half the households questions based on the shorter schedule and the other half based on the longer format. If there is not much difference in the output, it will be comparable. Before July, when the next set of surveys is commissioned, we want to be prepared with a set of recommendations based on the field tests.