Despite the Bharatiya Janata Party-led government making job creation one of its focus areas, India’s unemployment rate has increased – from 4.9% of the labour force in 2013-’14 (before the BJP assumed power) to 5% in 2015-’16 – data shows. Or has it?

While much ink has been spilt in decoding official employment numbers and what they mean for the economy, a report by government think tank NITI Aayog, put up on its website on Thursday, suggests it may have all been in vain. According to the report, submitted by a task force set up in May to improve job data, much of India’s official statistics are either outdated or suffer from design problems when it comes to employment numbers, rendering debates based off these numbers quite pointless.

This report comes at a time that many refer to as a period of “jobless growth”, where gross domestic product (the sum of all goods and services produced in the country) continues to rise while labour bureau data shows an erosion of jobs rather than the creation of new employment.

This conclusion can be drawn from other independent sources too. According to the Centre for Monitoring Indian Economy, 15 lakh jobs were lost in the first four months of 2017 as the number of employed people came down from 406.5 million to 405 million. The centre said these numbers were based on the finding of its Consumer Pyramids Household Surveys, which have a sample size of 161,167 households, which include 519,285 adults, across the country. The most recent survey was conducted in the January-April period.

NITI Aayog Vice-Chairman Arvind Panagariya, however, is of the view that the government data itself is flawed. Therefore, the think tank’s report – which is open to public feedback till July 23 – suggests ways to improve the credibility of employment data and examines possible sources of errors in the labour bureau’s statistics. It also points out flaws in current methods of data collection and dissemination, while suggesting that new and more frequent household and individual surveys be conducted.

This is not the first time the credibility of employment data in India has come into question. Economists have for long been demanding more reliable economic numbers for a country as big as India, which presumably has the manpower and technical skills to estimate correctly if it is creating enough jobs for its workforce.

Bad data

While writing about unemployment in India, publications often refer to one statistic – that each year, 12 million people enter the workforce. This should be good news for a country looking to make the most of its demographic dividend, with 65% of its population under the age of 35. However, the 12 million figure comes from the Labour Ministry, which largely relies on the Employment-Unemployment Survey conducted by the National Sample Survey Organisation. And this survey, though seen as the most comprehensive survey on labour force participation in India, suffers from many limitations, as the NITI Aayog report points out.

The Employment-Unemployment Survey collects data typically once every five years, which is too large a time frame for a bustling economy like India, according to the report. Moreover, after the survey, the data takes almost a year to be released to the public, making it even more outdated.

The report goes on to say that enterprise or establishment surveys (which collect data from worksites rather than households) could prove to be a good proxy for estimating India’s working population and their current employment status, if not for the fact that only enterprises registered under the Factories Act are considered for such surveys. Furthermore, the Act only takes into account establishments that employ 20 or more workers if not using power or 10 or more workers if using power.

“This leaves out all service sector establishments and all industrial establishments employing less than 10 workers if using power and less than 20 workers if not using power,” the report stated.

Additionally, the Economic Census covers all non-agricultural segments regardless of size or sector, but the survey has been conducted infrequently and at arbitrary intervals in the past, the report pointed out. For instance, the survey was conducted in 1977, 1980, 1990, 1998, 2005 and then 2013-’14.

The report goes on to say, “… The 2.7 crore workers covered by the Economic Census only represent about 6% of the total workforce of about 47.3 crore workers.”

In addition, the document points to large variations in data released by various surveys.

“According to the sixth and latest round of the Economic Census, the total workforce employed in all establishments in 2013-’14 was 13.1 crore,” the report noted. “Comparing this figure to 24 crore workers employed in industry and services as per 2011-’12 NSSO [National Sample Survey Organisation] household survey, we see that a substantial part of non-agricultural force is not captured by the Economic Census.”

A section that perhaps all such surveys fail to capture is the self-employed population or disguised unemployment. For instance, units employing less than 10 workers together account for 79% of India’s workforce, the report said, adding that only 2.7 crore workers are employed in establishments with 10 or more employees, which represent a “tiny 1.3%” of all establishments in the country.

The task force to improve India's employment data was set up in May under the chairmanship of NITI Aayog Vice-Chairman Arvind Panagariya.

Fixing the data

While pointing out problems with the way data on employment is collected and distributed in the country, the report made several suggestions for the government to improve the credibility and reliability of its statistics.

Household surveys

To correctly identify the state of unemployment in the country, household survyes must be conducted annually rather than erratically, the report said. And to gather additional data, it recommended the use of technology, through mediums such as phones, after one face-to-face interaction has taken place. Another proposal is to introduce a time-use-based survey every three years that deciphers how people allocate their time during their days and weeks, apparently to get a more credible estimate of women’s participation and disguised unemployment.

Enterprise surveys

For surveys of business establishments, the NITI Aayog task force has recommended that these be made wider and deeper, covering more enterprises. At the same time, it added, data generated through the Goods and Services Tax Network can be used as a potential sample to get inputs on a continuous basis in real time.

“Samples drawn from GSTN will have the virtue of covering enterprises of all sizes except those with [annual] turnovers below Rs 20 lakhs and from industry as well as services sectors,” the report stated.

Additionally, the task force proposed a new survey for enterprises outside the ambit of the goods and services tax, such as health and education, and those with a turnover of less than Rs 20 lakhs a year to make sure that the dataset is fully representative of the population.

The task force also pointed out that organisations have multiple identifiers – such as their Goods and Services Network registration, Employees’ Provident Fund Organisation, Employees’ State Insurance and Factories Act among others – and recommended using the GST Network across ministries and departments as the universal establishment number.

Using government databases

Another recommendation of the task force is to use existing government datasets, such as those with the Employees’ Provident Fund Organisation, Employees’ State Insurance and National Pension Scheme, for cross-referencing of employment data.

However, it cautioned that there would be substantial limitations in using government datasets as there would be a large amount of duplication and subsequent aggregation problems for statisticians.

“First, there is very substantial overlap across them,” the report said. “Aggregation across them requires de-duplication. In turn, this requires providing a common identifier for individuals listed in these datasets. Modern statistical techniques allow de-duplication without a common identifier but this is not without error.”