Three weeks after statisticians from the National Sample Survey Office (NSSO) published a report exposing holes in one of the key databases used in India’s gross domestic product (GDP) calculations, economists and statisticians continue to raise questions on the database, MCA-21, and the manner in which it has been plugged into the national accounts.
The data storm erupted after Mint first reported the findings of the NSSO report on 8 May. The report based on a field survey of services firms showed that 16.4% of companies in the MCA-21 database were either non-traceable or closed, and another 21.4% were ‘out of coverage’ or misclassified. These results meant that two detailed reports based on the first-of-its-kind survey had to be junked.
The blow-up questions
In a statement issued on 10 May, the finance ministry said that the overall GDP estimates are unlikely to be mis-calculated despite the problems found in the MCA-21 database.
“From 2012-13 to 2016-17, the number of enterprises whose annual returns were not available for GDP estimation accounted for only 12-15 percent of paid-up capital of all the enterprises in the MCA database," the statement said. “As such the GVA estimated for the responsive enterprises was increased by a blow-up factor of only 1.13-1.17... even when there is a small over- or under-estimation, the blowing up affects the level of GDP and not the year-to-year annual growth rates materially."
The finance ministry’s contention that the GDP estimation methodology may not be overestimating the levels of GDP finds some support from research by G.C. Manna, a senior adviser at the National Council of Applied Economic Research (NCAER) and a former director general at the Central Statistics Office (CSO).
In a 2017 research paper published by the journal of the Pune-based Indian School of Political Economy, Manna showed that the bulk of the non-reporting companies were those in the smallest paid-up capital (PUC) buckets. Further, the GVA to PUC ratio for these PUC size classes was relatively higher. Hence, Manna concluded that the GDP estimates may be underestimating the true size of the corporate sector.
However, Manna noted (in a footnote) that this conclusion was based on the assumption that all non-reporting companies were in full operation. If the reason for non-reporting was that these companies were not in operation, the blow-up methodology may be overestimating GDP, he added.
Manna’s paper does not tell us about the impact of the blow-up factor on growth but another research paper by the economist Amey Sapre of the National Institute of Public Finance and Policy (NIPFP) --- a Delhi-based think tank funded by the finance ministry --- attempted to answer this question.
Sapre used data from the Prowess database of the Centre for Monitoring Indian Economy (CMIE) to simulate how the lack of data on missing or non-reporting companies could impact growth rates. Sapre’s research published by NIPFP in 2016 suggests that the use of the PUC-based blow-up factor can lead to erroneous estimates which can vary based on the extent of coverage in any year, and on average, can overestimate GVA growth in the manufacturing sector significantly. Even if companies accounting for 10-15 percent of paid-up capital in the database have not filed returns, and the growth rates are blown up using a corresponding blow-up factor, this can alter growth rates significantly, Sapre’s research showed. Sapre showed that the use of industry-wise growth rates is a better alternative to estimate growth, with lower chances of errors.
The former chief statistician, Pronab Sen believes that blowing-up is justified for the first set of provisional GDP estimates when data for all companies is not available. But in the computation of the final estimates (third revised estimates, which are published nearly three years later), the use of a blow-up factor is wrong since the final estimates are based on the full universe of reporting firms. Sen also said that this was the prescribed methodology when the new series was introduced.
The methodology notes issued by Mospi --- one in 2015 after the new series was introduced and another in 2017 after revisions were made --- do not offer much clarity on the exact blow-up methodology, the number of companies considered for each revision, and the corresponding blow-up factor used. A detailed set of questions on these issues sent to the current chief statistician and senior CSO officials on 17 May, followed by reminders to Mospi’s spokesperson, has not elicited any response so far.
According to N R Bhanumurthy, a professor at NIPFP and a member of the Advisory Committee on National Accounts Statistics (ACNAS), a blow-up factor is used for the final estimates since the entire set of active companies are considered to be the universe of companies at each stage of revision and even if 10 percent of such companies have not provided data, a blow-up factor has to be applied to account for the entire universe.
The judgement on the methodology thus relies on an assumption regarding non-reporting ‘active’ companies. Active companies are those that have filed returns at least once in the past three years. If the non-reporting active companies are either not producing at all, or contributing negatively to GVA, then GDP is being overestimated because of the blowing-up methodology. If on the other hand, these companies are producing that year, blowing-up may be legitimate, even if PUC is not the best metric to use for that.
The NSSO report --- which exposed the holes in the MCA-21 database --- has shown that the assumption that all active companies produce goods and services is a ‘bureaucratic fiction’, said R Nagaraj, a professor at the Indira Gandhi Institute of Development Research.
The data quality questions
The debate on the use of MCA-21 also hinges on the consistency of the balance-sheet information in the database. Economists and statisticians argue that the database cannot be directly plugged into the national accounts as it is being done now.
“It is very difficult to construct a consistent profit-and-loss or cash flow statement and other disclosures from the (MCA-21) filings," said Mahesh Vyas of CMIE. “Partly, this is because a large part of the essential information is tucked into notes... We read the filings in detail before we add data of a company into our database."
It would be hazardous to assume that the database is usable directly, Vyas added.
The MCA-21 database should be made public so that researchers can check for the reliability and robustness of the data, said Ila Patnaik, a professor at NIPFP. The new database was introduced without adequate validation, she added.
Some of these issues also figured in the meetings of a National Statistical Commission (NSC) committee on real sector statistics headed by Sudipto Mundle. The committee which had representation from both CSO and MCA noted ‘serious inconsistencies’ in the MCA-21 database and suggested that a “procedure of regular meeting of the NAD (National Accounts Divison, CSO) and MCA officials may be put in place to (a) discuss the requirements of the NAD from the MCA data, (b) review the current list of items & equations and (c) devise scrutiny checks on MCA data."
It is not known if any such joint meeting has taken place, nearly a year after the report was published. The Mundle committee report also suggested that MCA should make data on companies ‘more accessible to the public by devising a query based system’. No progress on that front is visible yet.
A detailed set of questions on these issues sent to senior MCA officials on 11 May, followed by repeated reminders to MCA officials and spokespersons, has not elicited any response so far.