It's time for a new QES

The Quarterly Employment Survey is based on an outdated frame and will perpetuate incorrect estimates

Recently, the Labour Bureau, under the Ministry of Labour and Employment, released the Quarterly Employment Survey (QES) results for April-June 2021 for non-farm industries in the organised sector, which includes establishments with 10 or more workers. The organised sector accounts for less than 2 per cent of total establishments and less than 10 per cent of the economy’s total employment. The survey covered about 11,000 units and was based on telephone interviews.

The QES reported a 29 per cent rise in employment, or 7.1 million new workers. There were 30.8 million workers as of April 1, 2021, compared with 23.7 million, according to data from the Economic Census, in 2013-14. The findings met with muted scepticism on the reliability and validity of the quarterly estimates and their comparability with the dated annual census estimates.

Claiming that experts misinterpreted the survey results, Professor S P Mukherjee, chairman of the advisory group for the QES survey, in an interview with this newspaper on October 21 said: “There is no question of job generation. The figures reported were total employment that existed in a particular sector and not generated during a particular period. These were the figures of employment in the manufacturing sector as of April 1. That figure was higher than reported in the Economic Census of 2013-14”. In other words, Professor Mukherjee endorses the comparability of the quarterly QES estimates with the annual Economic Census 2013-14 estimates. He notes that a well-designed sample survey should yield a population estimate close to the census figure. 

However, Professor Mukherjee’s contention merits further attention. Indeed, it is the case that one could compare estimates based on a sample survey with census figures provided the sample is representative of the frame it is drawn from and appropriate data multipliers are applied to sample survey data.

In comparing the QES with the Economic Census, however, this does not appear to be the case, as the sample selection is not proportional to the units in the frame. For instance, establishments in the education sector accounted for the largest number of units in the frame, but the education sector stands third in terms of the number of units in the sample. Additionally, there are problems in comparing a year-long census with quarterly estimates from a sample survey that is designed to create a fixed panel drawn using stratified simple random sampling (without replacement) technique. It is also important to remember that the Economic Census was designed to create sample frames for conducting surveys in the non-farm sector and not for producing employment estimates. Shortcomings of the Economic Census’s employment numbers are widely acknowledged and hence sparsely used. 

Illustration: Binay Sinha

Beyond the comparability question, another critical issue seems to have missed Professor Mukherjee’s attention. Economic Census 2013-14 is the frame (the universe of establishments) used for drawing the QES sample. Further, the Economic Census is based on the “enumeration blocks” of the Population Census, 2011, as the primary geographical units. Hence the universe of establishment for drawing up QES samples is nearly a decade old! One can reasonably expect large scale entry and exits of establishments during the seven years when the economy has grown at an average annual rate of 5 per cent. Estimating quarterly employment using the QES based on such an outdated sample frame and comparing it with the annual Economic Census estimates do not seem statistically appropriate. 

The QES was initiated in 2009 to measure the employment impact of the global financial crisis in 2008 on select industries in the organised manufacturing sector. In April 2016, the Labour Bureau enlarged the survey’s scope by increasing sample size and enhancing sectoral coverage using the Economic Census-2013-14 frame. However, in 2017, the revised QES was abandoned as the employment estimates were unsatisfactory.

The government appointed a task force on employment statistics in 2017 (headed by vice-chairman Niti Aayog) to review the entire gamut of national employment statistics. The task force suggested scrapping the QES, as the outdated Economic Census-2013-14 sample frame was responsible for the poor QES estimates. To quote the report, “This problem is reinforced by the fact that sample frame of the survey is not updated until a new Economic Census is conducted. This means that the implications of the expansion of the relevant universe of enterprises are not taken into account in arriving at the population-wide estimates.” (Report of the Task Force on Improving Employment Data, p. 10)

The task force further recommended: “There is an acute need to strengthen sources of enterprise and establishment-level data … [with] wider coverage across enterprises … bringing both industry and services into the fold… to carry out enterprise surveys at a greater frequency …[and to] conduct the Economic Census at regular intervals.” (p. 14). In 2020 a new Economic Census was carried out in response; results are awaited.

The QES report is aware of the problems of the outdated sample frame. It said, “QES does not capture employment data from units which emerged after the 6th Economic Census in 2013-14” (p. 8). Yet surprisingly, the Labour Bureau chose to embark on a new QES. In 2016-17, when the National Sample Survey Office (NSSO) used a list frame of Economic Census and the Ministry of Corporate Affairs in its 74th round survey on the services sector (2016-17), the entire survey had to be scrapped as these lists were found to be faulty and a sizeable share of units could not be located.

It is also important to highlight that the new QES is a panel of establishments. While panel datasets are desirable as they allow us to track the same units over time, there are problems with their representativeness when the universe of establishments expands. While a panel may be representative when the sample was selected, it becomes less representative over time.  Therefore, describing employment trends in the organised sector based on panels is fraught with challenges. The QES lacks representativeness and the Economic Census-2013’s dated frame only exacerbates the problems. 

Severe defects plague the present QES. It will perpetuate discredited employment estimates that will be reported quarter after quarter. So what should be done to correct for the failed effort? It is, perhaps, best to abandon the current QES, expedite the Economic Census-2020 frame, and re-launch the QES with a statistically sound sample frame.

Kapoor is a visiting Fellow with the Indian Council for Research on International Economic Relations, New Delhi, and Nagaraj is a visiting Professor at the Centre for Development Studies, Thiruvananthapuram