Questionable Data, Inaccurate Assumptions: Examining India’s GDP Figures

In the absence of reliable figures, the extent of India’s current economic crisis remains unknown.

In the April–June quarter of the current financial year 2019–20, India’s gross domestic product (GDP) grew at 5%, the lowest in the last six years. Current Finance Minister Nirmala Sitharam called the figure a “dip” in the growth process, but experts fear that even 5% may be an overestimated figure, and actual GDP is between 3% and 4.5%. Former Chief Economic Advisor Arvind Subramanian also questioned India’s GDP numbers, saying that figures between 2011–12 and 2016–17 were overestimated by about 2.5%. Subramanian stated that actual growth during this period was therefore closer to 4.5%, rather than the 6.9% average.

In 2015 the Central Statistics Office (CSO) changed the base year for calculating GDP from 2004–05 to 2011–12. While a change in base year is routine, in this instance it led to above-normal changes in estimates, with GDP increasing from 4.2% to 6.8% for FY 2013–14, and growth in the manufacturing sector moving from -0.7% to 5.3%. Even after demonetisation, official GDP figures for FY 2016–17 were 8.2%, the highest in a decade.

A number of articles in the EPW also questioned the current GDP estimates, and called for an independent inquiry into methodology used. This reading list looks at how, since 2015, GDP in India has been calculated. This list examines the new concepts introduced, and the data sources used in GDP calculation.

1) Flawed Quarter Figures

In the financial quarter post-demonetisation, when 86% of cash in circulation was rendered invalid, the CSO claimed that GDP for that fiscal quarter grew at 7%, a figure that allowed the BJP government to vindicate its policy decision. However, R Nagaraj argues against the validity of these figures, arguing that quarterly GDP estimates lack suitable primary data, and inaccurate estimates of the informal sector were used to predict GDP figures.

The indirect method of estimation is, in simple terms, the product of benchmark estimates of value added per worker (based often on dated sample surveys) and an estimate of the number of workers employed in a year (or a quarter) ... Some of the “fresh” data that go into Q3 GDP estimation could have other shortcomings as well. For instance, in the new NAS series with 2011–12 as the base year, (i) Index of Industrial Production (IIP) figures are used for the noncorporate manufacturing sector and quasi-corporations in the non-financial private corporate sector, and (ii) the Ministry of Corporate Affairs’ (MCA) quarterly corporate financial results for estimating corporate sector GDP.

2) A Flawed Base Year

J Dennis Rajakumar and S L Shetty write that there is a vast difference between advance estimates (AEs) and revised estimates (REs) of growth figures. By changing the base year to 2011–12, along with other methodological changes in calculating growth, the authors question the validity of AEs released by the CSO.

As per these AEs of CSO, the nominal GVA and GDP are expected to grow in 2015–16 at 6.8% and 8.6%, respectively, and in real terms, at 7.3% and 7.6%, respectively … FRE [first revised estimate] of GVA at basic price in 2014–15 was lower than its corresponding AEs by 1.9%, and of manufacturing by 7.3%. For AEs and FRE, different data sources have been relied on which have vastly divergent coverage and sample-size implications.

To better forecast AEs, Rajakumar and Shetty recommend changing the base year from 2011–12 to 2014–15, which would allow estimates to be based on the same data set.

Following this logic, we have reworked the expected growth rate for 2015–16 over AEs of 2014–15. According to our suggestion, GDP is expected to grow at 6.5% over 2014–15, compared to 7.6% based on the FRE for 2014–15. The GVA is also expected to grow slower at 5.9%, against 7.3% using FRE for 2014–15. A similar tendency is noticed with respect to GVA across industries … In particular, manufacturing GVA, expected to contribute 17.5% to total in 2015–16, would grow at 3.1% if we consider its corresponding AEs of GVA in 2014–15, as opposed to 9.5% on FRE of 2014–15. These results point to all not being well with the Indian economy, quite unlike what CSO’s most recent AEs seem to suggest.

3) Misrepresenting Manufacturing Output
The CSO’s new series of National Account Statistics (NAS), which takes 2011–12 as the base year, showed a two percentage point increase in GDP share for the manufacturing sector, at the same prices. Ravindra H Dholakia, R Nagaraj, and Manish Pandya write that the new series also does away with the Annual Survey of Industries (ASI) and instead uses corporate financial data obtained by the Ministry of Corporate Affairs (MCA) to measure manufacturing value added to the GDP. By using MCA data, the authors argue that the methodology used to blow up sample estimates are suspect, and they are not available to independent verification.

The MCA database is large, consisting of 3–5 lakh companies (out of the universe of about 10 lakh companies) compared to 4,500 large companies under the Reserve Bank of India’s database used for estimating private corporate saving and investment … The problem with the new series, therefore, boils down to the following question: Does the new NAS series represent a fuller description of the manufacturing value added (MVA), or is it an overestimation? In other words, how true is it that the ASI omitted non-factory value addition of an enterprise (as averred by the foregoing statements)? To our knowledge, the CSO (or its officials) has not offered evidence to support its stated views.

The CSO justified the shift to MCA data, arguing that the ASI does not cover activity occurring outside the factory—R&D, sales and services, and so on. However, the authors dismiss this argument, asserting that the ASI covers employment, investment and other activities conducted outside the factory. Rather, they caution that the MCA data will overestimate the manufacturing sector.

The very basis of the change in the approach to data collection for estimating manufacturing GDP seems questionable. Hence the higher share and faster growth rate of manufacturing sector reported in the new GDP series seems to have little justification based on mere coverage of ASI. There may, however, be other reasons for expecting the size of the sector and its growth rates to be higher, but the arguments put forth against the ASI as under-reporting value added in manufacturing do not seem to be convincing.

4) Suspect Private Sector Data
R Nagaraj questions the CSO’s estimates on the private corporate sector. After changing the base year and replacing Reserve Bank of India (RBI) data with the MCA database, Nagaraj writes that estimates in the CSO's initial and final reports significantly varied, despite the same methodology and sample size used.

For 2012– 13, in the 2014 version of the subcommittee report, the PCS’s savings were lower than the corresponding NAS 2014 estimate by 70%. But in the 2015 (final) version of the sub-committee report, PCS’s savings were higher than the corresponding estimate in NAS 2014 by 8.5%. In other words, between the two versions of the sub-committee report, PCS savings shot up by an incredible figure of 257%! Can such a revision be deemed reliable without prior careful verification?

Read More: