Soumya Kanti Ghosh and Pulak Ghosh in their article for ThePrint find fault with the State of Working India report.
First, Ghosh and Ghosh begin with a puzzling mischaracterisation of the report. They say it relies on the Labour Bureau’s Quarterly Employment Surveys (QES), which have several limitations, the most important of which is that they only cover eight sectors (manufacturing, construction, trade, transport, education, health, accommodation and restaurant, and IT/BPO). We agree with these limitations.
That is why, in addition to the QES, we use the Labour Bureau’s employment-unemployment survey and the NSSO’s household survey, as well as establishment surveys by the Ministry of Statistics and Programme Implementation (Annual Survey of Industries for the organised sector and the National Sample Survey for the unorganised sector). We use the QES only very sparingly to provide estimates of employment in the formal service sector since these are not available from the other two establishment surveys. Therefore to say, as Ghosh and Ghosh do, that our report ‘uses the QES as its basis’ as if this is the only data source used, is simply false. We can only conclude that they either did not read the report or have, for whatever reason, omitted to mention the many other sources.
The second point is regarding the use of payroll data, such as those available from the EPFO, the NPS and the ESIC. The recent attention being given to this data is both welcome and problematic. Ghosh and Ghosh’s intervention earlier this year was pioneering because they explored a new source of information that can be brought in to understand the jobs scenario. But, as the authors themselves note, these databases are yet to stabilise.
Further, several people have pointed out, since the original study, that such data cannot be used to draw conclusions about the aggregate state of employment. Note that the EPFO has around 60 million active members. This is around 12-13 per cent of the entire workforce. So, job gains or losses in this part of the workforce tell us very little about what has happened in the rest of the economy. This does not make the EPFO data useless, it just makes it an unsuitable indicator of aggregate employment. A recent article in the Economic and Political Weekly from labour economist Radhicka Kapoor points out the problems in detail. That Ghosh and Ghosh seem to think that it is sufficient to use EPFO data to make such projections suggest that they are happy to make very heroic assumptions about the rest of the workforce based on this data.
The final point made is regarding the limitations of household surveys. Once again, most researchers would agree that household surveys are not foolproof but need to be used with due caution. However, the example chosen by the authors to make this point makes an elementary mistake. They note that the CMIE (Centre for Monitoring Indian Economy) household survey data shows both a fall in the total workforce (net loss of jobs) and a fall in the unemployment rate between September-December 2016 and January-April 2017. This apparent contradiction has a simple explanation available in the same CMIE data, but overlooked by the authors.
The data shows a fall in the labour force participation rate (LFPR) during this period. Note that the unemployment rate is defined as the fraction of workers without work as a proportion of those who have work or are seeking work. If the fraction of working age people either working or seeking work (the LFPR) falls, i.e. people stop looking for work (say due to demonetisation which occurred during this period), then one gets the result observed: a loss of jobs and a fall in the unemployment rate.
Thus, in the face of low or falling LFPR, a low or falling unemployment rate is not a good indicator of the health of the labour market, a point that has been made in the US’ case also in the context of the lingering effects of the 2008 crisis. So, this is not about survey design, but rather a basic fact of the labour market and it is surprising that Ghosh and Ghosh appear not to understand it.
Finally, for what it’s worth, we completely agree with the need for a better and more robust data architecture and say as much in the report. This needs to be a combination of household and establishment surveys alongside payroll data. In this respect, we also note that such data should not be provided to a select few, but should be made available to all so that researchers and the general public can properly scrutinise and understand the patterns of employment.
The authors teach Economics at Azim Premji University.
A previous version of this article erroneously said that the QES data covers 15 per cent of total employment in the country and is no worse than the EFPO. The authors wish to thank Soumya Kanti Ghosh for pointing it out. The error is regretted.