Getting the GDP numbers right

Getting the GDP numbers right

Estimates are not perfect, but the process is revised and fine-tuned.

Arvind Subramanian, Arvind Subramanian research paper, Arvind Subramanian GDP claims, India GDP, Indian economy, indian GDP overestimation, Economic Advisory Council to Prime minister, EAC-PM, indian express
Former Chief Economic Adviser Arvind Subramanian. (Express photo by Prem Nath Pandey)

Former Chief Economic Adviser Arvind Subramanian’s recent paper claims that the Indian GDP growth may have been overestimated by 2.5 per cent per annum between the period 2011-12 and 2016-17. A note by Prime Minister’s Economic Advisory Council (PMEAC) rejects the methodology, arguments and conclusions of Subramanian’s paper. A study done at our institute by Ashima Goyal and Abhishek Kumar show that after removing various flaws in Subramanian’s data and procedures used, these indicators suggest official growth rates are overestimated before 2011 too. This study also shows that the GDP growth in a large number of countries is either overestimated or underestimated using these indicators. Therefore, the study rightly says that “these regressions cannot be used for predicting growth or for concluding Indian growth is overestimated or for pointing to problems in the estimation methodology”. It is better now to concentrate on the next base revision by focusing on methodology, coverage and data in estimating GDP. There is no point in saying that GDP estimation has problems because ground level realities are different using some indicators.

I was a member of the National Statistical Commission (NSC) during 2013 to 2016 (Pronab Sen was the chairman) and was part of the process in base revision. In 2015, CSO introduced a new series of National Accounts Statistics (NAS) with 2011-12 as the base year. The guiding principles for change in base are: One, revision of base year to a more recent year; two, complete review of existing data base and methodology employed in the estimation of various macro-economic aggregates and alternative data bases; and three, implementation of the international guidelines based on SNA, the System of National Accounts 2008 to the extent possible. Therefore, it was not just base revision but there were significant changes in methodologies, coverage and data.

The base revision for 2011-12 was undertaken by CSO under the guidance of the Advisory Committee on National Accounts Statistics, which constituted five sub-committees for the purpose. The reports of these committees provide a comprehensive assessment of the changes and their rationale. Improvements in GDP estimation is a continuous process. In fact, Rangarajan’s commission on NAS was appointed 20 years back to give recommendations on significant deficiencies and improving credibility of official statistics. There have been several changes in recent GDP estimation. We focus here on two issues: Ministry of Corporate Affairs (MCA) 21 data base; estimation of household sector.

First, it is known that for the corporate sector, the old series used the RBI study on company finances from a sample of around 2,500 companies. There has been a long-standing demand to change this data. In 2011-12 series, corporate sector, both in manufacturing and services, has been comprehensively covered by MCA 21 data. For the “manufacturing” enterprises, MCA 21 data base has been used to supplement the information available in the Annual Survey of Industries. In the new series, the CSO used the MCA 21 data set which had about five lakh non-financial private companies. The new series also adopted the concept of enterprise in place of establishment. This led to a big change in manufacturing sector value added. One major issue is on the blow-up method for non-responding companies. Critics say this method is leading to overestimation of GDP. This may impact the level of GVA, but it is not clear about the impact on growth rates. However, there are some suggestions on the alternative indicators for blow-up method and also on having sample surveys for companies not submitting accounts.


Second, there are changes in the estimation of value added in household sector. NSS establishment surveys and the NSS employment surveys are used to get value added in household sector. Earlier, workers were treated as a single homogeneous group. In the base revision, a concept called “effective labour input” was developed. Another change in tune with SNA is that quasi-corporations have been separated from the household sector and added to the corporate sector in the new series.

The use of MCA data and other changes in household sector led to increase in the share of value added in corporate sector and decline in the share of household sector. But, it has not been established that overall GDP growth rates are overestimated.

Another issue is double deflation method — that is, deflating output and input separately while arriving GDP at constant prices. Single deflation has been in practice due to absence of separate deflators for inputs. An exercise done by G C Manna on manufacturing GVA with ASI data using double deflation method showed that it resulted in lower growth for the year 2012-13, and higher growth for the year 2013-14. Therefore, one cannot conclude that it is overestimation or underestimation if we use single deflation.

On 2011-12 base revision, it was a conscious and consensus (may not be unanimous) decision of CSO, NSC and various expert committees to have methodological revisions on the lines of UN SNA 2008. One cannot say GDP estimates are perfect as one has to operate on the data available in different domains. We have improved with each round of revision and with each improvement in data sources. We should now focus on methodology and data of CSO rather than saying ground-level realities are different using few indicators.

The writer is director, IGIDR, Mumbai and was member of the National Statistical Commission