All manner of interpretations and conspiracy theories are floating around regarding the CSO release of the new GDP back-series for the period 2004/5 to 2011/12 (or FY05 to FY12). Let me state that I, along with others, also found it inappropriate for Niti Aayog to be directly involved in the presentation of statistical data by the CSO.
The question is whether the back-series makes economic sense, and hence, whether it can be “interpreted” as an accurate reflection of reality. I have written several articles over the last three years on the new GDP data — the basic conclusion has been that the new series (base 2011-12) has been vetted by statistical experts at the international agencies involved in this worldwide exercise like the UN, IMF and World Bank. My conclusion has been that the unprecedented discussion about the “accuracy” of the new GDP data (now in its fourth year) is entirely political and would not have happened if the Congress had not suffered a defeat in 2014.
The new GDP series was presented barely seven months after the 2014 election. GDP changes to a new series are routine, and have occurred at least six times in India, and never before with any discussion, let alone debate. Base-year changes occurred in 1960/61, 1970/71, 1980/81, 1993/94, 1999/00, 2004/5 and now 2011/12. I challenge anyone to show me one instance of any discussion on base-year change in India, or elsewhere. Why did it happen in India, and post 2014?
I will now get into some technical details which strongly indicate that the lowering of GDP growth for 2004/5 to 2011/12 was entirely expected. Primarily because of the surprise low employment growth between 2004/5 and 2011/12.
The story behind the new GDP series, and the just-released CSO back series, has its (major) explanation within the realm of NSSO employment data for the last 20-odd years. There is one large sector of the economy — wholesale and retail trade (WRT) — whose GDP estimation is directly dependent on employment data as revealed by the NSSO employment data. This sector, in the old 2004-5 series, accounted for 16 per cent of GDP; in the new 2011-12 series, it accounts for only 11 per cent. Why this large decline in the share of this sector?
Let us look at the traditional method of estimating WRT GDP. On the release of the NSSO employment data (approximately every five years), the CSO looked at the employment gains of this sector and assuming some productivity growth of the labour, arrived at an estimate of the sector’s GDP. In 1999/00, there were 34.4 million people working in WRT, and this figure increased to 41.7 million in 2004/5 yielding an annual growth rate in employment of 3.9 per cent per annum. Total employment in the economy increased at a CAGR of 2.4 per cent per annum. GDP growth for this period: 5.5 per cent per annum, implying an average productivity growth of 2.1 per cent per annum.
This healthy growth in employment was assumed \by the CSO in making estimates until the next major NSSO survey, in 2011/12, became available. (There was a drought-infected NSSO survey for 2009/10 which was rightly ignored by the CSO). However, the results of the 2011/12 NSSO survey were a shocker for employment gains, and in “polite” circles, the jobless growth encountered during these years is not much discussed. For the period 2004/5 to 2011/12, NSSO data revealed a total job gain of only 9 million (from 419 million to 428 million). Nine million over seven years translates into a CAGR of only 0.3 per cent per annum. GDP growth for this period: 8.1 per cent per annum, implying an average productivity growth of 7.8 per cent. Some acceleration in labour productivity growth was expected, given the large increase in investment — but 7.8 per cent?
This was the first broad hint that there was some overstatement in the GDP series for the UPA period. For WRT, the growth in employment was even lower than the aggregate — only 0.2 per cent per annum. The CSO (and international advisers) rightly got down to the task of changing the method of estimating GDP for the WRT sector. They rightly converged on using growth in real sales tax revenue as an indicator.
How much difference does the new and improved method make? For starters, it decreases the share of WRT in GDP by about 5 ppt (from 16 to 11 per cent) — a natural consequence of the fact that the share was grossly overestimated, according to the incorrect estimate of high growth in WRT employment. This large decline in the share can only mean that growth in the WRT sector was significantly lower than growth in the non-WRT sector. One estimate of the decline in aggregate growth, because of a change in method of WRT computation is about 50 bp a year.
The new CSO back series projects GDP growth to be 6.6 per cent per annum, FY05 to FY12, versus the 8.1 per cent contained in the old GDP series.
Inflation (as measured by the GDP deflator) between FY05 and FY12 has also been corrected. The deflator is a weighted combination of the WPI and CPI inflation indices. The two increased at a CAGR rate of 6.4 and 7.9 per cent respectively between FY05 and FY12. However, the old GDP deflator has an average inflation rate of only 6.7 per cent. A mid-point of the CPI and WPI inflation yields 7.2 per cent, that is, GDP deflator was underestimated by around 50 bp in the old series.
These two simple computations suggest that any back series earlier than 2011/12 should lower GDP growth. Incorporating employment growth estimate for just 10 per cent of the labour force, and an under-estimation of inflation, the GDP back series lowers GDP growth in the 2005/2012 period by around 100 bp per year. The CSO estimate, incorporating all factors, is a very credible estimate. But don’t tell that to the politicians and maybe some economic experts as well!