The old GDP series overstated sectoral growth; Here’s how

Updated: December 12, 2018 3:37:11 AM

The close relationship between the old and new series in current prices hides the fact that there are sharp divergences at the sectoral level

When the base revision exercise was done in 2015, the revised series showed improved growth performance from 2011, as compared to the earlier series.

By TCA Anant

CSO recently released a press note on the backcasting of the 2011-12 series of national accounts. Like with many other statistical exercises, this, too, got immediately sucked into a slanging match, much of it driven from political narratives. The confusion arises principally on two accounts. Firstly, a simple reading of the back series reveals that the y-o-y growth in constant 2011-12 prices between 2004-05 and 2011-12 is now lower than what had earlier been computed by the 2004-05 series. What added fuel to this fire was a existing, though now somewhat dormant, debate on the new series.

It is useful to connect some of the lessons from the earlier discussion in order to throw light on and dispel confusion in the current debate. When the base revision exercise was done in 2015, the revised series showed improved growth performance from 2011, as compared to the earlier series. This, unfortunately, happened at a time when there had been a major political change, leading all political commentators to include this discussion as a part of their argumentation. This discussion tended to focus on issues of growth, and for the most part, ignored the technical and methodological factors underlying the changes.

The differences between the old (2004-05) and new (2011-12) series can be summarised in a few stylised facts. Firstly, the new series significantly reduces the value added in the household or unincorporated segment of the economy. This was on account of the employment estimates from the 2011-12 NSS survey as compared to those in the earlier surveys of 1999-2000 and 2004-05. Secondly, the new series significantly reduces the contribution of the services segment of the economy. This was primarily due to the reduction in value added in trade. Countering these two reductions was the increase in the contribution of the corporate segment. This was due to the fact that the MCA data gave a more comprehensive picture of value added in this segment than the earlier piecemeal compilation from different sources. The overall effect of these changes was an approximately 3% reduction in gross value added in current prices in 2011-12. These changes in levels aside, there were changes in the way estimates beyond the benchmark year were compiled, leading to changes in the growth profile of the economy. These included the use of sales tax/service tax collections to project growth in retail trade and some service sectors; the use of MCA data to estimate corporate value added in all compilation categories.

For computing the back series, ideally we should have used the same methods which were used after 2011-12 and then compiled or projected backwards. This, however, is not always possible, since similar data is available in earlier years in all cases. Thus, what CSO has done is to use a medley of approaches depending on the availability of data in different segments of the economy. So, what can we make of this exercise?

The first point to note is that the simplistic comparisons of aggregate growth rates between the old and new series hides many interesting features of the backcast. The close relationship between the two series in current prices hides the fact that there are sharp divergences at the sectoral level. Further, in both current and constant prices, the new series has higher growth rates in many years in sectors like mining, manufacturing, electricity, construction and public administration. However, one sector which has consistently had lower growth rates and levels is trade, repair, hotels and restaurants. This divergence is almost completely explained by the fact that the 2004-05 series had significantly over-estimated the contribution of this sector to GVA/GDP. The growth divergence is also linked to the difference between the new sales tax-based indicator and the older indicator of gross trading income. The over-estimation of GVA in the informal or unincorporated sector is a second factor behind the lower growth estimates in the back series.

However, in our review of the new series, we had pointed out that the use of the MCA database in 2011-12 showed that the 2004-05 series had underestimated GVA from the corporate sector. For the computation of the back series, MCA data was not available. This then leads to the question of whether this biases the backcast towards underestimating growth rates. Since MCA data was not available, for corporate sector estimates in the back-series, CSO has used a melange of approaches using ASI data in manufacturing and RBI corporate studies’ data in other cases. This approach has one element in its favour: This was the approach used in earlier series when MCA data was not available. However, in using this approach, CSO appears to have made some significant adjustments. Firstly, the computation of the back series has been done separately for different institutional categories. Secondly, the ASI and RBI data has been used to generate growth indicators which were used to back cast the 2011-12 estimates. In other words, the level differences of 2011-12 have been incorporated into the back series. These adjustments imply that the back series estimates are an improvement over the 2004-05 estimates.
There is, however, still an open question that the growth dynamics of these indicators may be different from those of MCA. Some partial results about the comparison of ASI and RBI datasets and MCA is available from 2011-12. These seem to suggest that the growth in the MCA-based GVA has been more than that of these two sources. But before we jump to any conclusion, we should note that this may vary over the business cycle due to the fact that the coverage of both ASI and the RBI studies are more limited than the MCA frame and that these elements may differ in their growth dynamics. To sum up, the back series does meet the requirements of researchers for generating a consistent time series but it will remain an open question as to whether we have been able to completely account for the contribution of the corporate sector, an area which should continue to engage researchers.

Before concluding, it may be useful to point out some other dimensions of confusion. The first relates to the increased share of gross fixed capital formation (GFCF). There are two reasons for this. Firstly, the denominator, i.e., GDP, is lower in the new series, thus increasing the share of GFCF. The second reason is that the output of manufacturing is higher in the new series and the output of services lower. Since, in India, expenditure side aggregates are, to some extent, derived from production estimates, this has also contributed to raising the implied estimates of capital formation. A second source of confusion relates to differences in the GDP deflator. The explanation here is, in part, on account of the varied sectoral composition of GDP, and, in part, to the index number effect of changing the base year.

To conclude, in assessing the back series, it is best we discard simplistic assessments based on our political prejudices and, rather, focus on the estimates and its structural characteristics. This may help us improve our understanding of the past and help us avoid making errors in assessing our present.

-The author is former chief statistician of India

Get live Stock Prices from BSE and NSE and latest NAV, portfolio of Mutual Funds, calculate your tax by Income Tax Calculator, know market’s Top Gainers, Top Losers & Best Equity Funds. Like us on Facebook and follow us on Twitter.