GDP revisions and jugglery

The easiest way for the economy to achieve the Prime Minister's target of a 9 per cent growth rate for the economy must be for the Central Statistical Organisation to keep revising its data. This seems to be what the GDP statistics presented in the Economic Survey for 2000-01 suggest.

It is a part of the data generation process that the CSO first makes an Advance Estimate of GDP, to be followed by the Revised, Quick, Provisional and ultimately the Final Estimates. The revisions can be substantial as indicated for 1999-2000 in Table I. These revisions cannot be criticised though one can only hope that over time the extent of the modifications will be reduced.

But what is also happening is that even the Final Estimates of GDP are being revised and that too years later. This year's Economic Survey reports CSO data which have revised GDP estimates for as far back as 1994-95 and 1996-97! And the revision is not small. As Table I shows, the increases in GDP growth rates for these two years in the latest estimates have been fairly large. If this process continues, 1996-97 may end up as the first year in a decade in which growth crossed 8 per cent.

These revisions are not deliberate and meant to show one or the other government in a poor/good light. (If it were so the NDA Government would not have wanted to show that one of the years of the United Front Government showed the highest growth rate in the past decade - 7.8 per cent in 1996-97.) Knowledgeable economists say that as the CSO keeps receiving revised information on industrial production, agricultural output and even of services it is forced to revise the GDP estimates. That is why even after the ``Final'' GDP estimates have been made for the early and mid- 1990s they are revised once more. Until recently all revisions were upwards. However, as the latest figures for 1997-98 (the second year of the short-lived UF Government) show the odd downward revision too takes place.

GDP estimates may not be revised to prove a point but the analysis of the same data can be carried out in a certain fashion to prove one's point of view. And this is what the Economic Survey 2000-01 does to show that growth in the post-reform era has been higher than before. The Survey compares the average GDP growth in the period 1980-81 to 1991-92 with that in the period 1992-93 to 2000-01. As Table II shows, this comparison makes GDP growth in the post-reform era a good percentage point higher. But there is a problem with the periodisation. When were the reforms introduced? In June-July 1991. So strictly speaking, 1991-92 should belong to the post-reform era. And if it is, the average pace of growth in the reform era is no higher than before (Table II).

One can argue that as 1991-92 was a year that straddled both eras and since it was one of stabilisation, GDP growth in that year should not be included in either period. In that case, the difference between the two periods narrows down quite a bit. So going by the average rates of GDP growth one cannot really come to a definite conclusion either way. (And if one looks at the sectoral composition the fact that it is the service sector rather than industry or agriculture that has boosted economic growth in the 1990s gives things a completely different complexion.)

Actually, the authors of the Economic Survey need not have resorted to all these tricks. Instead of taking averages for short periods, they could well have used the median (middle) GDP growth rate in each period. Indeed, the more careful analysts of statistics suggest using median values measure rather than averages when making comparisons over relatively short periods (See ``Voodoo of statistics and woes of growth rates", The Hindu, July 27, 2000).

Indeed, internationally median values are used more often than averages. If the authors of the Economic Survey had in fact compared median growth rates in the pre and post reform eras, they could have comfortably included 1991-92 in the better period. The median growth rate in the period 1980-81 to 1990-91 was 5.5 per cent and that in 1991-92 to 2000-01 6.4 per cent. This is not a major acceleration but still close to a percentage point higher.

CRR

Send this article to Friends by E-Mail

Section  : Business
Previous : Glaxo, SmithKline boards approve merger
Next     : Monitor