A couple of months ago, a CMIE report on unemployment in India led to a huge cacophony in the political sphere. With the unemployment rate standing at 7.2%, the report said that in 2018 alone, as many as 11 million (1.1 crores) jobs were lost. Another report, by the National Sample Survey Office, reported an even more alarming statistic – that the unemployment rate was at a 45-year high.
The opposition parties in the country smelled blood and unitedly decided to draw their daggers of criticism out against the Modi government. If the government’s policies led to a loss of 1.1 crore jobs in one year alone, one had to wonder how many jobs would have been lost over its entire nearly 5-year tenure! But then, what would be the ground reality? Let’s put it in perspective.
For various definitions as CMIE uses them, refer to Point 1 in the comments section. Now let us see how CMIE computes the unemployment rate. The CMIE has formed what it calls a Consumer Pyramids panel of around 173,181 households comprising of around 522,000 such members who are in the age category of 15 years or more i.e the age where a person is considered to be a part of the labor force.
Out of the 173,181 households mentioned, 110,615 households surveyed were from urban areas, and 62,566 samples were surveyed from rural areas. The urban households are from 320 different towns and cities of different sizes, and the rural households are from 3911 different villages. These towns, cities, and villages are from almost all states (the exceptions being certain North East Indian states and certain Union Territories). Next, the recorded data is run on a Survey package for the R statistical environment, where the Horvitz Thompson Estimator is used to generate the estimation of unemployment. You can download the file ‘Unemployment in India: A Statistical Profile (Sep-Dec 2018)’ on the link for a deeper understanding of the process.
Now, there are certain things to consider:
- 4000 towns and cities in India. The CMIE survey covers 320 towns and cities i.e 8%
- 640,867 villages in India. The CMIE survey covers 3911 villages i.e 0.62% of the villages
- 248 million in the whole of India, 73% in rural India, therefore 27% in urban India, and 0.27*248 = 66.96) in urban India. The CMIE survey covers 110,615 urban households i.e 0.17%
- 460 million. The CMIE survey covered 522,000 people who can be classified as members of the labor force i.e 0.11%
Shitavarun bhatachi pariksha’ which translates to ‘when testing the quality of rice, one would need to examine a few grains’. Statistical estimation is a similar exercise. However, rice grains are expected to be similar to each other. Human beings are not, and hence, statistical estimates can be highly error-prone.
Another hypothetical scenario where statistical estimating can go wrong is mentioned in point three in the comments. I have made the particular choice of numbers in the hypothetical example deliberately. The CMIE survey sampled 0.11% of the population i.e 11 out of every 10,000. The NSSO’s sample size is even smaller, at 101,724 households which involve 326,810 individuals of age 15 or 15+ years (which as we know is considered Labour Force).
Of course, over large populations and large samples, the sampling error rate is expected to even out. Also, on a practical level, one cannot choose a sample size beyond a point. A 5 lakh+ sample-size would be normally considered good enough. But yet, can we truly ignore aspects such as the survey covering only 0.03% of the total rural households? Also, could there be a deliberate bias in choosing samples themselves, which could end up generating an adverse report? We will look into this aspect later in the article so as to not digress from the statistical aspect of the discussion. Now we look at some of the obvious inconsistencies and rather outrageous conclusions in various claims on unemployment in both the CMIE and NSSO reports:
- As per NSSO, the unemployment rate in 2017-18 at 6.1% was the highest in 45 years. Which means, since 1973, there was not a single year when the unemployment rate was equal to or more than 6.1%. Now when we look at the graph showcased on the CMIE website; we see the following figure:
The graph starts in January 2016 and ends in March 2019. Readers can easily verify this by clicking on the link and hovering the mouse pointer over the graph. What the graph shows is, in January 2016, the unemployment rate was 8.72, and it reached highs of 9.65% and 9.59% in May and August of 2016 respectively. So now, do we trust the NSSO report which says that the unemployment in 2017-18 at 6.1% was the highest since 1973 when the CMIE graph suggests that in 2016 itself, the unemployment rate was much higher than that in the two months mentioned above?
- As per this news item on the basis of the CMIE report, India lost 11 million jobs in 2018 with rural areas worst hit. Now, the survey has clearly included a lot more household samples from urban areas (110,615) which form 27% of the total number of households, as against the household samples from rural areas (62,566) which form 73% of the total number of households. In all, they have sampled 0.03% of the rural households (3 in every 10,000). How accurate could be expect the estimations drawn to be?
- On November 8th, 2016, at 20:00 hours, Prime Minister Narendra Modi made an announcement that had the nation buzzing within minutes – Demonetization of old Rs. 500 and Rs. 1000 currency notes. Now according to a report by CMIE in 2017, a year after the Demonetization day, 1.5 million people lost their jobs post demonetization between January-April of 2017.
Now let us look at the graph published on the CMIE link again. The unemployment rate in January 2017 was 5.93%, and in April, it was 3.87%. If we trust both the CMIE graph and the above report, then what we would conclude is despite the GDP growth rate for the Jan-March 2017 quarter clocking at 6.1% , an expected substantial fall from 7.4% in the previous quarter there was a net addition of 9.48 million jobs in the three months just after the 50-day demonetization exercise concluded!
In fact, as per the CMIE graph, the unemployment rate touched a low of 3.37% in July 2017, suggesting that the six months post demonetization, despite being those of slowest GDP growth rate in the past 5 years, actually generated several million jobs. What do we then make of the reports of thousands of SMEs having to shut down because of a shortage of cash?
- The other issue with blindly taking the CMIE estimates on the unemployment rate is the incredible volatility it suggests in the net jobs created and lost. For example, in April 2016, the graph shows an unemployment rate of 8.90% which shoots up to 9.65% in May 2016 and again back to 8.91% in June 2016. That means a net loss of 3.45 million jobs from April 2016 to May 2016, and immediate replenishment of this loss by a nearly equivalent net job addition from May 2016 to June 2016. While this is the most extreme example of the kind of volatility under consideration, several such instances can be found throughout the chart. For example, from an unemployment rate of 9.59% in August 2016, it shows a reduction to 6.71% in October 2016, suggesting the net addition of 13.25 million jobs in 2 months, which for all practical purposes is an impossibility.
Now let us briefly touch upon another lesser quoted statistic of Labour Participation Rate, the definition of which I have already included earlier. Now according to this write-up on the CMIE website, the Labour Participation Rate fell from 43.2% in January 2019 to 42.7% in February 2019.
Firstly, a decrease in Labour Participation Rate suggests that people are leaving the labour force (i.e they are either losing employment or are unemployed and yet no longer willing to work or actively seeking a job). Now, there are logical reasons why this is bound to happen. Firstly, at 15 years of age, a person has just cleared his secondary school.
A decreasing number of people today are actively seeking jobs immediately after passing their 10th grade. With an increasing number of students opting to pursue at least an undergraduate degree, the earliest age at which they could enter the workforce can be 20-21.
However, as we have seen earlier, the CMIE statistical estimates tend to throw up highly volatile computations, and even the fall of 0.5% in the Labour Participation Percentage which would correspond to an approximate 4.5 million persons quitting the labour force in the span of one month is improbable.
When Rahul Gandhi today speaks on stage about how ‘Unemployment is the highest in 45 years’ and how ‘Millions of youth have lost their jobs’ , he just has to engage in shallow rhetoric for which the CMIE and NSSO (by the way the NSSO (National Sample Survey Office) is named so and not NPSO (National Population Survey Office) for a reason) reports have armed him with. And Modi cannot in his rallies afford to get into the nitty-gritty of statistical estimation as described above.
What he can, and does quote, however, is the fact that the MUDRA Bank has managed to disburse around 5.4 crore loans amounting to a total of Rs. 2.73 lakh crore to various micro and small entrepreneurs. And remember, this is not an estimation – these are real numbers. For an example of what this means in terms of job creation, refer to Point four in comments.
The Dr. Manmohan Singh-led UPA government had once upon a time had made a very silent admission that seems to be buried in the pages of recent history – 500,000 jobs could be lost in 2013. But guess what? If you read the article, it will mention specifically which companies have fired how many employees at which locations along with the predictions for what could be the number of job losses in a particular year.
Not one Opposition leader today would give any details of a similar nature, where they could say that here – this is the sector in which so and so companies have fired an X, Y, and Z number of employees respectively. Which is why their superfluous quoting of the CMIE and NSSO data cannot be trusted.
Now let me end with a slight bit of trivia regarding CMIE. On the CMIE Board of Directors, among five names, two are interesting — Mahesh Vyas – Managing Director, and Ajay Shah. Now focus on the latter for now. Ajay Shah is currently being investigated for an algorithmic trading scam. Mahesh Vyas is the brother-in-law of Ajay Shah. Ajay Shah is said to be close to P. Chidambaram, and it is this proximity that is said to have been of aid to him while going about his algorithmic trading ‘business’. Any conclusions that could be drawn from this bit of information are entirely left to the reader’s judgment.