In 2015, the Central Statistical Office (CSO) revised the way GDP is calculated in India. According to the new series, India is the fastest growing large economy in the world, with a 7.1 percent real growth rate. Other trusted measures of the state of the economy convey a discordant picture. This discrepancy has led to an active debate comprising two parts. One part of the debate has been arguments about the extent to which the official GDP data are accurate. The second part of the debate is based on criticising CSO's methods.
This article summarises the literature that examines CSO's methods. There are two main areas of concern: the way manufacturing Gross Value Added (GVA) has been estimated and the methodology for calculating deflators.
The manufacturing sector has been at the centre of the GDP debate. The methodological changes for this sector and consequently the data revisions have been substantial. Manufacturing growth for 2012-13 was revised up from 1.1 percent to 6.2 percent, while that for 2013-14 was increased from -0.7 percent to 5.3 percent. Various authors (Nagaraj, 2015a, 2015b, 2015c; Rajakumar, 2015; Nagaraj and Srinivasan, 2016; Sapre and Sinha, 2016) have questioned the reliability of the new estimates, on several grounds.
Enterprise vs. Establishment approach: In a major innovation, the new GVA methodology shifted data collection from establishments (or factories) to enterprises (or firms). Sapre and Sinha (2016) point out that lack of clarity on measures of output and costs at the enterprise level can lead to imprecise estimates of GVA. The activities of firms can be much more diverse than those of factories, and not all of these functions would qualify as manufacturing. Yet all the value added of enterprises classified as "manufacturing firms" has gone into the calculation of manufacturing GVA. This will inflate the level of output and possibly also the growth rate, if the ancillary activities are growing faster than the manufacturing ones.
Blowing up of GVA: Extrapolating from samples ("blowing up") is not a new feature of the current GDP series. What has changed is the database used. Previously, manufacturing GVA was based on the RBI's fixed sample of large private companies. Under the new series, the MCA21 database is used to compile a set of "active" companies, which have filed their annual financial returns at least once in the past three years. The problem is that for any given year, information from several active companies remains unavailable till a cut-off date of data extraction. In such a case, the GVA of available companies needs to be blown-up to account for the unavailable companies. There are multiple issues in this blowing-up method.
The year wise number of available and active companies in manufacturing is not publicly available. Hence, year on year, the exact number of companies for which the GVA is blown up is unknown.
While the Ministry of Corporate Affairs has made filing of annual financial returns mandatory for all registered companies, it is not known how many of these companies produce any output on a regular basis.
The blowing-up factor is the inverse of the ratio between the paid-up capital (PUC) for the available companies and that for the active set as a whole. Nagaraj (2015a, 2015b) argues that this is inappropriate since a large fraction of the MCA21 active set are "fictitious, shell companies" that exist only on paper. In that case the blowing-up method is likely to overestimate GVA.
Sapre and Sinha (2016) argue that blowing-up using the PUC is an inappropriate method because PUC and GVA do not have any one-to-one relation. Also, it is possible that the actual GVA of some "active but unavailable" companies is negative for a particular year. In those cases, blowing up of GVA using the PUC factor method can lead to overestimation.
The actual computation of the blowing-up factor applied by the CSO in the new series has not been described in detail in the official documents. This makes it difficult to replicate the process and analyse it.
A single blowing-up factor has been used for private as well as public limited companies. Rajakumar (2015) points out this is not appropriate as the two groups are widely divergent in their patterns.
The number of "available" companies reporting their annual financial returns with MCA varies across the years. As a result, the blowing-up factor that accounts for the non-reporting companies will also vary from year to year. As highlighted by Nagaraj (2015a, 2015b) and Sapre and Sinha (2016), this variation will result in wide fluctuations in the final GVA estimates.
Identification of manufacturing companies: Sapre and Sinha (2016) find that within the manufacturing sector several companies operate as wholesale traders or service providers. These companies may have changed their line of business since they were originally registered. These changes do not get reflected in the Company Identification (CIN) code assigned to the companies. Such misclassification of companies will distort the manufacturing estimates, although not the overall GVA.
MCA 21 vs. IIP : There are other problems with the manufacturing GVA calculation that have not been written about much. For the manufacturing sector, the GVA is derived from a combination of MCA 21 numbers, Index of Industrial Production (IIP) estimates and estimates of the unorganized sector from the Annual Survey of Industries (ASI). While the MCA21 is a new database, the base year for the IIP data is still 2004-05. Also the data obtained from MCA 21 follows an "enterprise" approach as mentioned earlier, but the data obtained from ASI follows the old "establishment" approach. The full implications of these discrepancies are yet to be fully understood.
Previously, estimates of real GDP relied heavily on production indices such as the IIP. Now, most real numbers are derived by taking nominal data and deflating them by price indices. If done well, this approach can give a more accurate measure of value added. But if the deflators used are inappropriate, the estimated real magnitudes will be distorted. And this may well have happened in the past few years, since there have been very large changes in relative prices (especially petroleum and other commodities), which are inherently difficult to capture in aggregate deflators. The issues here are as follows.
Double deflation: In most G20 countries, real manufacturing GVA is computed using a methodology known as double deflation. In this method, nominal outputs are deflated using an output deflator, while inputs are deflated using a separate input deflator. Then, the real inputs are subtracted from real outputs to derive real GVA. But in India things are done differently. Here, we compute the nominal GVA, and then deflate this number using a single deflator.
If input prices move in tandem with output prices, both methodologies will give similar results. But if the two price series diverge- as they have for the past few years in India- single deflation can overstate growth by a big margin.
The reason is not difficult to see. If the price of inputs falls sharply, profits will increase, and nominal value added will go up. Since real GDP is supposed to be measured at "constant prices", this increase needs to be deflated away. Double deflation will do this easily. But single deflation will not work. In fact, if a commodity-weighted deflator like the Wholesale Price Index (WPI) is used, as is the case under the current methodology, nominal growth will be inflated, on the grounds that prices are actually falling! In this case, real growth will be seriously overestimated. A fuller explanation is provided here.
As the gap between input and output inflation starts to close, the problem will diminish. But that could also send a misleading signal, because it might seem that growth is slowing, when only the measurement bias is disappearing.
Service sector deflator: Deflator problems also plague the estimates for the service sector, which accounts for the bulk of GDP. Currently, the deflator used for much of this sector is the WPI. But the weight of services in the WPI is negligible. If instead the services component of the Consumer Price Index (CPI) were used, growth in this sector would be far lower than currently estimated.
WPI vs. CPI : Finally, there are questions about whether the WPI should really be used as a deflator, at all. The weights are now more than a decade old, and India's economic structure has changed radically over this period. In addition, the sample frame (the selection of goods sampled) is also out of date. The CPI is a better price index [link, link].
Based on the foregoing, a number of refinements to the GDP methodology could be considered:
- Releasing disaggregated information on firm output and cost items, to permit more precise estimation of manufacturing GVA given the shift from the establishment to the enterprise approach.
- Altering the definition of the active set of manufacturing companies, to ensure the companies are truly active.
- Releasing the number of active and available companies every year by industry or sector, to get a sense of the companies contributing to GVA.
- Shifting the blowing-up factor from paid-up capital to another indicator, such as replacing growth rates for "active but unavailable" companies by the overall growth rate for the relevant subsector.
- Using separate blowing-up factors for public and private limited companies. Currently the blowing-up factor does not take into account the size, industry or ownership of the unavailable companies.
- Reviewing the classification of companies to ensure they are categorized appropriately.
- Providing greater clarity and transparency about the database and methodology used to estimate the manufacturing sector GVA. Also, documents could be released explaining the precise method used to blow up the GVA estimates.
- Adopting the double deflation method to calculate real manufacturing GVA.
- Using the relevant CPI components to deflate service sector GVA.
- More generally, the WPI could be replaced by the relevant CPI components, in the long period before a Producer Price Index (PPI) is developed which would be an ideal deflator.
Until this methodological debate subsides, official GDP data should be used with caution as it may not accurately reflect conditions in the economy. Other proxies for output are required.
I thank Josh Felman, Deep Mukherjee, R. Nagaraj, Amey Sapre, and Pramod Sinha for useful conversations
Nagaraj, R. (2015a), Seeds of doubt on new GDP numbers Private corporate sector overestimated?, Economic and Political Weekly, Vol. L, No. 13.
Nagaraj, R. (2015b), Seeds of doubt remain: A reply to CSO’s rejoinder, Economic and Political Weekly, Vol-L, No. 18.
Nagaraj, R. (2015c), Growth in GVA of Indian manufacturing, Economic and Political Weekly, Vol-L, No. 24.
Nagaraj, R. and T.N. Srinivasan (2016), Measuring India’s GDP Growth: Unpacking the Analytics & Data Issues behind a Controversy that Refuses to Go Away, India Policy Forum, July 2016.
Rajakumar, J Dennis (2015), Private corporate sector in new NAS series: Need for a fresh look, Economic and Political Weekly, Vol-L, No. 29.
Sapre, Amey, and Pramod Sinha (2016), Some areas of concern about Indian Manufacturing Sector GDP estimation, NIPFP Working Paper 172, August 2016.
The author is a researcher at the Indira Gandhi Institute of Development Research.