Who is afraid to answer questions on jobs in India?

Ideally, the job estimates should be derived using record-level data and appropriate weights for each observation

In a week from the publication of the paper titled “All you wanted to know about jobs in India — but were afraid to ask” in early July 2018, I had raised ten observations to members of the EAC-PM pertaining to serious mistakes in the paper. The report was prepared for the Economic Advisory Council to the Prime Minister (EAC-PM). My observations were limited only to the incorrect use of CMIE’s Consumer Pyramids Household Survey in the paper. These and some audacious assumptions had led to a “conclusion” that 12.8 million jobs were added in 2017.

I have followed up on my observations and even offered the entire record-level data of CPHS along with the services of a research assistant to help in the analysis, but neither my observations nor my offer met with any response. I was given to understand that a revised paper would be submitted to the EAC-PM by mid-September 2018, but I do not know if that was submitted or what happened if it was. Since then the aforementioned erroneous report remains on the EAC-PM website and since there is no correction to the errors and since the concerned EAC-PM member has left the EAC-PM, it is appropriate for me to now make public what, according to me, is wrong with the report.

The first mistake made by the report is in its understanding of how CMIE’s CPHS defines the employment/unemployment status of a respondent. In the CPHS, we take a respondent’s status as of the date of the survey. If the survey is conducted in the early hours of the day, it is possible that a daily wage worker may not know his/her status of employment. In such cases, we take the status of the preceding day. So, the status of a respondent is either on the day of the survey or the preceding day. This ensures easy and correct recall. The report submitted to the EAC-PM has misunderstood this and has assumed that if the respondent did not work on any of these two days but usually worked over the past year, CMIE also classifies them as employed. This is wrong and is at the root of all calculations in the report which renders much of the reports findings dubious.

CPHS is a complex survey with a stratified sampling design. Population estimates derived from such a sample necessarily require the use of appropriate weights inherent in the survey design. Ideally, the estimates should be derived using record-level data and appropriate weights for each observation.

The authors of the report had on offer from me of access to all the record-level data with the weights well before the report was submitted to the EAC-PM, but it was chosen to not use this data or the appropriate weights. The report uses population estimates of 11 age-sex groups published by CMIE using appropriate weights for “demographic extensions” apparently by using UN population projections. The elementary question is why was the available survey design, the available record-level data and the available weights corresponding to the survey design not deployed?

Then, the report cherry picks the labour participation rate for men from CPHS but not for women. The problem is that government-economists are unable to accept the CPHS finding that female labour participation rate fell sharply after demonetisation. The rejection of this finding is arbitrary and the simultaneous acceptance of male labour participation rate is evidence of cherry picking of only “suitable” data.

Nobody is afraid to ask questions on jobs in India. The question is — who is afraid to answer them. 

The EAC-PM official website distinguishes between EAC reports and reports of individual members. The aforementioned report does not appear as an EAC report but as an individual members report. Implicitly, the report has not been accepted by the EAC-PM. But, it is used by commentators and such use often causes confusion. Authors juxtapose CMIE’s estimates that show no job increases with the said paper that shows an increase of 12.8 million. This leaves the reader with a misleading impression that the truth could be somewhere in between. If the truth is always in between two opposing points of view, then it can be easily gamed. But, truth cannot be gamed. Readers must indulge and discern.

If we do not discern then we must be prepared for more confusion when NSSO releases its estimates in the coming few weeks. An informed debate using all available data transparently is imperative.