The recent embarrassment on the growth in the Index of Industrial Production or IIP for January 2012 — which turned out to be at one-sixth of its provisional value — has stirred up a hornet's nest. But a study of this data over a two-year period reveals that such revisions are common.
A look at the IIP numbers for the 26-month period between September 2010 and January 2012 reveals that significant revisions are quite frequent. For instance, in 15 out the 26 months — about 58 per cent of the time — the final/revised number differed from the provisional one by 1 percentage point or more.
Understating numbers
On an average, the provisional IIP growth rate in this period was off by 2.2 percentage points compared to the final number. If you thought that the Ministry of Statistics and Planning is trying to paint a rosy growth picture, you would be wrong.
The Ministry, in fact, underestimates IIP growth more often than it overstates it. Over the 26-month period, the provisional number turned out to be lower than the final one two out of three times. There is another saving grace too. The average difference in the estimated and actual growth rates that was 2.7 percentage points in 2010 “improved” to 1.15 percentage points in 2011. So, to its credit, the Ministry actually reduced the errors by more than two times from 2010 to 2011.
On an average, the actual IIP growth during the period diverged from the provisional number by a factor of 1.7 (or 1.7 times). But as mentioned earlier, we are improving! This average actually decreased from 1.6 times in 2010 to 1.3 times in 2011.
A probability game
Let's play a guessing game based on fundamental probability theory! In the ensuing months, if the average IIP growth rate is projected to be 4 per cent, what is likely to be the actual growth rate?
The trend between December 2009 and January 2012 suggests that there is one in three chances that this value could be 2.4 and two in three chances that it could be 7.4. Then, the actual growth rate is the probabilistic sum of the two, that is, 5.6 per cent, which is definitely way off from the 4 per cent that was projected. In either case, the take-away is that what is supposed to be the barometer of the industrial performance in India is flawed most of the times, and perhaps not serving its very purpose.
Why the revisions?
About 15 source agencies supply data on output from three sectors — electricity, manufacturing and mining and quarrying — based on which the IIP figures are constructed.
However, not all data on production may be available at the time of constructing the index. In addition, the methods used to estimate data points from units that don't respond are not uniform.
So a provisional index is created which is typically updated twice in the subsequent months to incorporate the necessary changes.
Is there light at the end of this tunnel? The government could trim down and consolidate the data source agencies to reduce inconsistency on data collection. It could also standardise the method of estimating data points for non-responding units across source agencies. Such a regulation could at least reduce the difference between the estimated and actual figures.