(Written by Davendra Verma)
An article on the overestimation of GDP in India by Arvind Subramanian (‘New evidence for fresh beginnings’, IE, June 12) has received widespread criticism. Subramanian is a former CEA, an insider. He is not just any individual or economist. He should have been more careful before bringing out a working paper with a questionable methodology. By doing so, he has done great injustice to the Indian statistical system. He has gone to the extent of saying that he is using only statistics produced outside the central statistics office (CSO) in his analysis, though he has used the index of industrial production (IIP), which is produced by the CSO.
The compilation of GDP is a completely apolitical and complex exercise, much beyond the correlation and regression analysis undertaken by Subramanian. Further, a lot of the analysis carried out by him is based on the CSO’s quarterly estimates, data for which is limited and hence no serious scientific study can be based on these estimates. These estimates are good for giving the direction of economic growth in the short term and a broad idea of the quantum of growth.
Subramanian has claimed that “GDP is overestimated by 2 percentage points post 2011-12 with base revision and new methodology”. But, he was consulted when the new methodology was finalised in a meeting conducted by the then National Statistical Commission chairman, where he gave no suggestion. Before releasing the new GDP series, a presentation of the new methodology and estimates was made to select officers in the finance ministry, where too he was present.
The former CEA has admitted that there was a need to use the corporate sector data for GDP compilation. He may have some reservations in the way shell companies have been accounted for. But the estimates arrived at after blowing up subscribed capital using MCA data are far superior to what was being done in the earlier series, when the estimates were obtained by blowing up the results from the RBI’s study of a much smaller sample of companies. How can a series based on data pertaining to a fewer number of companies be considered to be more reliable than that based on a much larger set of companies? This data was adopted in the new GDP series after proper examination and analysis.
Subramanian has argued that the growth rate (2000-2010) in the manufacturing sector was closer to the IIP growth rate. People familiar with the compilation of GDP in India should be aware that the manufacturing sector estimates in the old series were being compiled using IIP. Prior to the base year 2005-06, data from the annual survey of industries (ASI) was being used for the computation of growth estimates of the organised manufacturing sector. Later, as it was observed that the estimates from ASI were diverging from the IIP, and it was becoming difficult to defend the revisions of these estimates during the first revised estimates of GDP, it was consciously decided to continue to use the IIP. Thus, the estimates continued to be based on the IIP and were not revised using ASI estimates, which was the practice in the 2000 base series.
Another issue which keeps coming up: IIP is highly volatile, and does not properly capture the growth rate of the manufacturing sector. Both the IIP and ASI were not found to be useful for estimating manufacturing sector growth during 2000 to 2010. Since there was no other source of data available for estimating gross value added by the manufacturing sector, these estimates/growth rates continued to be used.
It has also been argued that high GDP growth is impossible in the absence of a rise in exports and imports. To establish such a causality, one should look at the share of exports and imports in the overall output and input of the Indian economy, especially in the manufacturing sector. If they do not have any major share, it is difficult to establish any causal relationship.
Electricity generation, airline passenger growth, commercial vehicle production, and petroleum products consumption are closely associated with GDP growth. Any relationship using the quantum of growth in any of these indicators does not have any significant implication. Each of these indicators should be further bifurcated for further understanding their contribution to economic activity.
Subramanian’s paper has assumed that the estimates from 2000 to 2011 were correct, and hence the estimates in the 2011 series are overstated. But, a number of improvements were made at the time the 2011 series was introduced. So, any evaluation of this series with not so well estimated past series is invalid.
In the 2011 series, there was a significant shift in the GDP compilation process — from establishment approach to institutional approach. So, the estimates of the manufacturing sector (or any sector) in the 2011 series were not strictly comparable with the previous series.
In conclusion, researchers must have knowledge of the GDP compilation process before undertaking any econometric exercise which only leads to more confusion. There is a serious gap in the availability of the right kind of data. Concerted efforts should be made to fill this gap.
The writer is former director general, Central Statistical office