Recent GDP revision gives rise to doubts

According to a recent press release of the Central Statistics Office (CSO), the growth rate of gross domestic product (GDP) at market and at constant (2011-12) prices, a measure of the country’s economic growth, has been accelerating five years in a row — 5.5 per cent in 2012-13, 6.5 per cent in 2013-14, 7.4 per cent in 2014-15, 8 per cent in 2015-16 and 8.2 per cent in 2016-17 — and it is expected to fall to 7.2 per cent in 2017-18.

In particular, the growth rate of the last two years is based on the second and first revised estimates, respectively, and they deviate considerably from their respective previous estimates.

The GDP estimates go through several rounds of revisions. The first advance estimates of GDP, released at the end of first week of January of a given financial year, get revised as second advance estimates at the end of February of that year.

The provisional estimates (PE) are released two months after completion of the year (end May). Although the benchmark-indicator method is employed for both advance estimates (AE) and PE, the AE covers indicators for 7-9 months and the PE for the whole financial year.

Key indicators

The CSO considers indicators such as Index of Industrial Production (IIP), quarterly results of listed companies, advance estimates of crop production, deposits and credits, performance of railways, cargo handled by civil aviation and ports, government accounts and sale of commercial vehicles among others.

The PE are subsequently revised; first revised estimates (FRE) are released 10 months after completion of the year (end January), second revised estimates (SRE) after one year and 10 months and, finally, third revised estimates (TRE) after two years and 10 months. For the purpose of these revisions, the CSO essentially moves away from benchmark-indicator method and relies on actual values.

For FRE, it uses revised estimates of Budgets of both Central and State governments, available financial statements of public and private corporations and local bodies, and production, prices and inputs related information of over 42 crops and horticulture.

For SRE, the CSO uses actual figures from government budgets and improved coverage of accounts of public and private corporations and local bodies.

At this stage, the IIP is replaced with Annual Survey of Industries, where required.

The TRE considers improved coverage of both State and Central government accounts, and of corporations and local bodies. Thus, one would expect better and reliable estimates, as we move from AE to final estimates.

The reported growth rate of 8.2 per cent for 2016-17, which is based on SRE of that year, is more reliable as they are based on firmer and actual values, to a large extent. However, compared to its corresponding FRE, it is a huge jump by 1.1 percentage points.

Since the methodological considerations and data sources remain nearly the same for each of these revisions, one may expect such a jump to mirror the past as well.

We have, therefore, mapped the GDP growth rates of various years in 2011-12 series under different estimates (Table 1).

We do not have AE and PE from 2011-12 to 2013-14, because the new series (2011-12 base) of GDP was released on January 31, 2015. For the remaining years, the data show hardly any differences in the growth rates of GDP between PE and AE, probably because both estimates follow benchmark–indicator method with only the difference being AE covers 7-9 months, whereas PE one year of a given indicator.

Moreover, the growth rate difference as per FRE and PE in most of the years also appears to be meagre, except in 2017-18 when it was 0.5 percentage points. That is, the real GDP in 2017-18 grew at 6.7 per cent as per PE and 7.2 per cent as per FRE. There are only marginal differences in the GDP growth rate of a given year under SRE and FRE till 2015-16, but it rose by 1.1 percentage points in 2016-17. Such observed scale of differences in the growth rate between two successive revisions is unprecedented.

To buttress this argument, we have further worked out percentage differences of a given year’s GDP under various revisions. As we have noticed that the FRE of 2017-18 and SRE of 2016-17 witnessing major changes compared with their corresponding previous estimates, we examined the percentage differences of FRE over its corresponding PE, and SRE over its corresponding FRE for all years for which data are available in the 2011-12 series (Table 2).

The SRE of GDP of 2013-14 was placed lower at (-) 0.82 per cent over its previous FRE. Similarly, the FRE of GDP of 2014-15 was reduced by (-) 0.86 per cent compared to its PE and the SRE of GDP of that year was still lesser at (-) 0.28 per cent than its FRE.

On the contrary, since 2015-16, the size of GDP has moved upward in successive revisions in both FRE and SRE.

Magnitude of rise

The magnitude of such rise is phenomenal in 2016-17 and 2017-18. For instance, the SRE of GDP in 2016-17 is 0.84 per cent higher than its corresponding FRE, and more significantly, the FRE of GDP of 2017-18 is 1.3 per cent higher than its PE. Such a trend has never been the characteristic of previous GDP revision in the new series.

This raises a question as to why such major differences have been noticed in the latest CSO’s revision of GDP of these years. Is it because the source data for PE, that is, benchmark indicators, are becoming outdated such that actual data of the recent periods are at variance with their movements?

Even then, the observed differences of SRE over FRE in 2016-17 remains a concern as these estimates have relied on nearly the same data sources.

The commentaries over the reliability of GDP estimation, with respect to methods and data sources, are thus yet to complete their course.

The writer is Director, EPW Research Foundation.

COMMENT NOW