A Response to NITI Aayog’s Rajiv Kumar on Seasonality and Job Losses After Demonetisation

Mahesh Vyas of the Centre for Monitoring Indian Economy (CMIE) responds to the Niti Aayog vice-chairman's dismissal of CMIE's research on unemployment in India.

In an interview with Karan Thapar for The Wire, NITI Aayog’s vice chairman, Rajiv Kumar, has been dismissive about CMIE-BSE unemployment statistics. He says that I had no answer to chief statistician T.C.A. Anant’s question on seasonal adjustment of unemployment data. This is an incorrect description of a brief interaction, which I relate below.

A couple of weeks ago, I was in a meeting in Delhi with Kumar and Anant, where the subject of discussion was about measuring employment and unemployment in India. After the meeting, as we were exchanging pleasantries and moving out, Anant effectively pulled me up on CMIE’s recent efforts at measuring unemployment in India.

What he said, essentially, was that it is not fair for me to state in what is no more than a footnote that the fall in jobs during January-April 2017 compared to September-December 2016 could reflect a seasonal effect, but then not take seasonality into consideration.

Anant was referring to my column of July 11 in Business Standard. The article can also be accessed from CMIE’s unemployment website.

This was the first time that I put out an estimate of the fall in jobs post demonetisation. I believe I was careful in not being alarmist and in not jumping to conclusion. The header did not scream that demonetisation had caused job losses. Most of the article focusses rather labouriously on statistics and even concepts. I then wonder and conjecture in the article on what could have caused this fall in jobs. I quote:

“Note that the 9.6 million fall in the unemployed count is close to the addition to the workforce. This is like saying that almost the entire new workforce of January-April 2017 did not offer themselves for employment. This is odd. Is this a seasonal phenomenon?

A longer time-series could help us de-seasonalise this to understand this phenomenon better. For now, we can make some intelligent guesses of what might be at work.

September-December is a busy season as the kharif crop is harvested during this period and most festivals fall during these months. 2016 was a good kharif crop and this could have kept employment levels high. January-April is a relatively lean season. Further, demonetisation could have had its full impact during these months while its impact during September-December was partial.”

Note that the CMIE’s effort at estimating unemployment began only in January 2016. With less than two years of monthly series, it was (and is still) not possible to make adjustments for seasonality.

I respect the chief statistician’s observations and explained that my time-series is just too short to make any seasonal adjustment. He, of course, knew this. So I initially thought he was just engaging in friendly banter. But then he said that I could refer to the detailed data of 2011-12 where they have given some estimates of seasonality. It is not in the main volume. I promised to look into this and see what adjustments are possible. It is not possible to simply implant the seasonality of one series onto another series, but that can be investigated.

CMIE's data also showed a drop in the labour participation after demonetisation, which was in line with new investments falling. Credit: CMIE

CMIE’s data also showed a drop in the labour participation after demonetisation, which was in line with new investments falling. Credit: CMIE

Anant then went on to say that once I adjust for seasonality, all the job losses I claim and attribute to demonetisation would vanish. I did not ask him if he had done the maths to make such a claim. That would be rude. I took his observations as those from a venerable professor as leads to do more work on the data.

Kumar made no observations and was not even a part of this casual conversation. The official meeting on measuring unemployment did not mention seasonality at all. It is apparent that Kumar did not know that I had acknowledged seasonality in my article even before Anant raised the issue.

So it is totally incorrect for him to say that I had no answer to Anant’s observation. I was the first to flag seasonality because I also lead India’s first effort to estimate unemployment on a monthly and even weekly basis in the country. Without such fast-frequency estimates, there is no question of observing or adjusting for seasonality.

While I respect the observations of the chief statistician on seasonal adjustment, I feel that the summary dismissal by Kumar of professional work does not build confidence in the government’s objectivity. Will Kumar similarly dismiss the Index of Industrial Production (IIP) statistics produced by the CSO because it is not seasonally adjusted? Or does the NITI Aayog and Kumar always seasonally adjust the non-seasonally adjusted series produced by the government before interpreting it? Furthermore, one may ask why does the Indian official statistical machinery not produce a seasonally adjusted IIP series like most developed countries do?

Looking at the larger picture, India’s official statistical machinery does not produce a monthly series of unemployment. A private Indian initiative by CMIE and BSE has filled this gap. Kumar says in his interview with Thapar that he is setting up a system to build a fast-frequency data gathering system at NITI Aayog. I sincerely hope that he does not ignore the CMIE-BSE databases in this regard just because they are not seasonally adjusted yet.

Independently, CMIE will continue its efforts to make fast-frequency estimations based on large surveys and other ways of crunching statistics. When the series is long enough, it will also make seasonal adjusted series. In the meanwhile, it is still useful to use and argue with the available data.

The CMIE-BSE partnership has made available, for free public consumption, five statistical profiles that provide very detailed data on employment and unemployment. These volumes collectively provide a time-series of age-wise, gender-wise, education-level wise, rural-urban-wise and state-wise data for labour force, labour participation, unemployment and several other related indicators. These volumes cover five waves of household surveys from the first one during January-April 2016 through the most recent one during May-August 2017.

Kumar is an accomplished economist and it would be an honor for me and for the team at CMIE to engage with him to dispel his doubts (or that at the NITI Aayog), if any, regarding the data and its utility.

Mahesh Vyas is managing director & CEO of the Centre for Monitoring Indian Economy.

Join The Discussion