GDP revisions undergo revisions, around the world, of 10-50 basis points, but a revision of 250 bp, annually, is unheard of
Former Chief Economic Adviser to the Ministry of Finance, Arvind Subramanian (hereafter AS) contends in a working paper at the prestigious Harvard University that India’s GDP growth between 2012 and 2016 (period II) likely averaged somewhere between 3.5 and 5.5%, yielding an average growth of 4.5%. Given that the official GDP growth for this period was 7%, there is an average overstatement in the official GDP statistics of around 2.5 percentage points (ppt) a year. You can search far and wide, in journals and in newspapers, but no one, absolutely no one, has made such a calculation and assertion as AS has done for any non hyper-inflation economy for such a long period of time (except for one China estimate—see below).
Co-incidentally, AS’s assertion is similar to the contention of two Hong Kong based economists that for the nine year period 2008-2016, official statistics reported GDP growth for China was higher by an average 1.7 ppt a year. However, there is one paper by History and Economics Professor at the University of Pittsburgh, Thomas Rawski, that asserted that cumulative GDP growth in China averaged just 3% over the four years 1998-2001 compared to the official estimates of 8%+. Rawski based his calculations on electricity consumption and his estimates were soon discarded by experts who used trade data to demolish his thesis. The question remains whether the same fate awaits AS’s estimates.
We examine AS’s computational method below, but first, a few general comments. One had thought that post Election 2019, politics would not encroach upon the substantive mis-measurement conclusions reached by AS. Unfortunately, there are still journalists and “experts” congratulating themselves that their election-oriented assertion that the Modi government had presented lies about the economy has now been proven right by AS’s paper.
Two points need emphasis. For the period studied, 2001-2016, is there any additional data that AS now has that he did not have when he was CEA from Oct. 2014 to June 2018? The answer is no and while we all have doubts about all forms of data (in India, China, the US and the world), the fact remains that there are no known revisions of the data post 2018 (and post 2017) for the data used by AS. Given this fact, the question obviously arises that if no new information is available, then what is the point of AS’s contention regarding mis-measurement of GDP growth for the period 2012-2016?
The second point relates to GDP revisions involved with the base-year change from 2004-5 to 2011-12. This is a routine exercise, undertaken for most economies (especially developing economies) of the world from time-to-time. The normal occasion for this revision is the availability of new household survey data on expenditure (or components of income). The consumer expenditure survey data for 2017-18 is expected to be released shortly; so brace yourself for another round of base-year revisions over the next year or two.
However, it is relevant to point out, and AS is well aware of this fact, that 2014-15 base year revisions had important components of “structural change” in data collection and interpretation. In particular, balance-sheet data from the Ministry of Corporate Affairs (MCA) was used to estimate value added in manufacturing and service industries. Previously, CSO used the index of industrial production and the Annual Survey of Industries (ASI) data (available with a two-year lag) to estimate value added in industry. For an important part of the service sector (wholesale and retail trade (WRT) accounting for 10-15% of GDP) the CSO used the previous growth rate in employment as observed via the Employment and Unemployment Surveys. For example, CSO used employment growth trends for 1999-00 to 2004-5 to assume WRT employment (and hence value added) trend between 2004-5 and 2011-12. It turns out that this estimation led to gross overestimate of the WRT sector since employment growth was less than 1% between 2004-5 and 2011-12, a big decline from the near 3% annual employment growth observed between 1999/00-2004/5.
All of this was well known by all statistical experts when they sat down, in multiple committees, to design a new method for estimation of GDP for industry and services. It is important to emphasise that international experts from the UN, World Bank and IMF were involved in the review exercise and most of which was completed before Modi became PM in May 2014. The CSO report and method was finalised in January 2015, when AS was already 4 months in office. Since all major economists were privy to this revision method (along with the National Statistical Commission—again, all appointees of the Manmohan Singh government!) it is a bit unexpected for AS to now claim that he had doubts about the new method of estimating GDP. For the period 2011-2016, there is the whole statistical world (and UPA) approving of the new method—and AS has now positioned himself against these experts and himself (pre-2019 AS). To finish this line of argument, what would have been most useful is if AS had documented the source of the new data (information) and how this new data has changed his interpretation of reality. Note that GDP revisions undergo revisions, around the world, of 10-50 basis points, but a revision of 250 bp, annually, is unheard of, at least since Rawski made a parallel attempt two decades ago.
What about the statistical method that AS uses to reach his conclusions about gross-estimation? Perhaps mindful of Rawski’s fate (AS has written a laudatory book on China’s GDP growth) AS’s estimation is based on four major variables—export and import growth, real credit growth and electricity consumption. He chooses two time-periods—2001-2011 and 2012-2016. I have absolutely no disagreement with his choice of time-periods. The High Level Advisory Group (HLAG) to the Commerce Minister, which I had the privilege and honour of heading, used precisely these break points to measure trade performance in the world, the reason being that world trade literally fell off a cliff in 2012. Between 2001 and 2011, world exports grew at double digit levels (in nominal dollars) and collapsed to -1.5% pa in the 2012-16 period. In real terms, the fall in world exports was from 4.6 to -1%.
The world has changed in many ways since 2011; world trade is down, drastically, and world inflation is now at 3% levels (distributed roughly at 1-2% levels for the developed world and around 2-4% levels for the developing world). In a detailed research note, I will be examining the AS methodology regarding GDP over-estimation in some detail. For the moment, let me point out a few salient facts which suggest that AS may not have got it correct in his over-estimation prediction of 250 bp of annual growth.
Manufacturing—Value-added vs IIP growth: In period I, the two growth rates are virtually identical: 7.7 vs 7.3% pa. In period II, value added in manufacturing increased at 7.2% pa and IIP increased at only a 3.1% rate. If the MCA data for period II are considered problematic, and IIP correct, then there is an over-estimation of 4.1% per year. With a weight of 15%, this will lead to an over-estimation of GDP of 60 bp per year in period II.
Wages in manufacturing (ASI data): Real wage per day growth was only 0.4% an annum in period I, which accelerated to 2.2% pa in period II.
Agricultural wage growth: For (unskilled) ploughmen, this accelerated from 1.9% in period I to 3.3% in period II; for (semi-skilled) rural carpenters, the acceleration was much sharper—from 0.9% to 4.5%.
Real wage growth is an important component of GDP growth, and both ASI and rural wage growth indicators suggest that the second 2012-16 period (spanning both UPA and NDA) contributed a higher portion to aggregate GDP growth. Trend in wages conflicts with AS’s derivation of over-estimation of GDP growth. A meaningful analysis of AS’s assumptions, and cross-country estimation methods, deserves serious investigation—especially a comparison of how other countries performed for the two growth periods.
The author is a Contributing editor, Financial Express Twitter: @surjitbhalla Views are personal