In December 2022, the Employees’ Provident Fund Organisation (EPFO) released its Annual Report for the fiscal year 2021-22. According to this, 46.34 million members contributed to the EPFO during 2021-22. This was 0.06 million higher than the 46.28 million contributing members in 2020-21. The latter estimate is sourced from the previous year’s Annual Report of the EPFO.

The 46.34 million contributing members in 2021-22 is much lower than the 48.92 million contributing members seen in the pre-pandemic year of 2019-20. Evidently, employment in the formal sector measured through EPFO contributions had not recovered to its pre-pandemic levels till at least 2021-22. It had declined by 5.3 per cent.

The EPFO Annual Reports also reveal that the contributing establishments have declined from 660,204 in 2019-20 to 591,184 in 2021-22. The number of contributing establishments has shrunk by a massive 10.5 per cent during these two years. The scars of the pandemic on employment have endured all claims of a V-shaped recovery as is evidenced even in the most touted official data.

This finding, however, is at odds with the regular monthly flow of EPFO data. We hope to explain this mystery in the following paragraphs.

The regular publication of EPFO payrolls data is a relatively new initiative of the government. It began in early 2018. It provides data from September 2017. The objective of this initiative is to measure employment in the formal sector in India. This payroll data is, in fact, the net addition of EPFO subscribers in a month. This is the gross additions less exits but including re-subscription of exited members every month. The EPFO is, in fact, a savings scheme applicable to a section of salaried employees and not a (non-farms) payrolls database as is usually understood in other countries where the term is used to measure employment. But, that is a different problem.

This monthly data release suggests that net addition to EPFO subscriptions increased by 7.7 million in 2020-21 and by another 12.2 million in 2021-22. This implies a 20 million net increase in EPFO subscribers. This is dramatically different from the Annual Reports placed in Parliament which suggest a fall of 2.6 million in the same period.

The monthly data released by EPFO are always provisional. Data for a month is released with a lag of less-than two months. The latest release is dated December 20, 2022 and the latest data it provides is the provisional estimate for the month of October 2022.

The first release of the data undergoes revisions in subsequent releases for a maximum of 12 months. Or, it may not be revised at all. Whether an initial estimate will be revised or not or, how often it will be revised depends upon the month. We will revert to this phenomenon in a bit. First, let’s examine the revisions.

The revisions are substantial. We examine revisions to the estimates of April 2021. The first estimate of net addition to EPFO subscribers in April 2021 was 1,275,729. This estimate was released in June 2021. It was revised downwards in each subsequent release till the release dated May 2022. By then, the April 2021 estimate fell by a massive 45 per cent to 698,876.

Similarly, the first estimate for May 2021 released in July 2021, which was 919,772, dropped by more than half to 441,760 by May 2022. This was a 52 per cent downward revision in 11 months.

The revisions seem to be systematically and substantially downward. Ergo, reading the initial releases of the EPFO can be significantly misleading.

The twist in these revisions is that they are available for a month till all the monthly data for its financial year are released. So, the first release of April sees 11 revisions in as many subsequent months, May sees 10 revisions, June sees 9, and so on till the March release, which never sees a revision. This is strange, and an unsatisfactory position because it produces a monthly time series of net EPFO subscriptions that has data of varying levels of reliability. Given that the revisions are very substantive and given that the revisions are incomplete, the time-series is rendered almost useless for any analysis or to draw any meaningful inference of the trend.

When the March data are released, the EPFO releases the annual estimate for the year ended in that March. These annual estimates are never revised. This implies that there were no revisions to the data once the last month of the data is released. But, this is implausible. If monthly estimates of April and May can be revised to less than half their original values then March estimates may also be getting revised in subsequent months. But, the EPFO monthly releases do not oblige with these revisions.

The presentations also raise a doubt whether the April 2021 estimate that was last revised in May 2022 was the final estimate or were there more revisions that were never released.

So, what has the new initiative that was much hyped when it was first released achieved? Its first estimates are unreliable by a huge margin as the reality could easily be half of the given estimate. Its time-series is meaningless because it contains different levels of substantial revisions. Its annual estimates are frozen and never revised when it seems these should be revised.

As a result, the monthly EPFO data releases have only misled. The good old EPFO Annual Report tabled in Parliament with a lag of about 9 months is still the best provider of this data. And, it suggests that employment fell during 2020-21 and stagnated in 2021-22.