Arvind Subramanian's recent observation that India's GDP growth has been overstated over the years would raise some eyebrows as he was part of the system which was involved in the so-called new methodology. While his observation holds for both NDA and UPA regimes and hence is not specific to just the ruling party when he worked as CEA, it does come more as an afterthought considering that he was known to defend demonetisation when it was introduced but became critical of the same after his tenure.
Similarly, he had made the term 'soft spot' quite popular in his authored Economic Survey as he took pains to show that India was the best-performing country and hence in a sweet spot which must have based on certain views one gets when in the government. Now a sudden turnaround based on an academic exercise does raise some questions on which judgment is right.
When the CSO calculates GDP it is wholesome and covers all the sectors. When data is not available proxies are used so that it becomes inclusive and does not miss out on any activity. This is one reason as to why it is revised regularly and the exercise is arduous because it takes a lot of time to come up with a refinement by which time a new base year may be warranted. India gets almost 40-45 percent of its GDP from the unorganised sector which can be captured only with legacy data and large machinery as several proxies need to be found here.
There are definitely shortcomings in the data that is collected or proxies which are used but can be accepted given the nature of the exercise. The methodology has been clearly defined with the proxies being stated upfront and hence cannot be contested. The approach is the same as is pursued in developed countries and also used by multilateral agencies. Therefore there can be no better way of having such coverage. Interestingly the problem is there in all countries as all agencies have to rely on proxies as not all activity is captured on the formal sense. The predominance of an informal sector in India makes it susceptible to errors which however would even out over the years.
For example, one will never know the value addition from Kirana shops as these are not defined. Yet there are millions of such entities of which a large part may not even be registered. Similarly for agricultural production one can never arrive at the actual amount produced because the entire crop does not enter the mandi, where the coverage is not more than 30-60 percent. There is a lot of grains used for seed or self-consumption or is sold in the local village and hence is known only through imputations. The CSO has machinery which has been tracking these changes and hence is more likely to be right when using proxies.
Now the former CEA has come up with his calculation that shows that there has been an overstatement of GDP based on limited variables that he has used. His argument from what has been made out from his write-up is that his proxy variables had a very good linkage with the GDP growth rates prior to 2011-12 after which it deviated which led to this overstatement of 2.5 percent or so over a longer time period. Economists have the advantage of using models where the fundamental tenet is that once the assumptions are stated and the model satisfies the basic tests, then the results should be acceptable. We cannot question the assumptions. Therefore if the author sues variables used like cement, steel, power etc. as proxies for the entire country, it should be accepted if the results are robust. However, models tend to exaggerate the fit of data as they are statistical exercises. The data does not touch on the unorganised sector which has a different story to tell. Similarly using bank credit is quite limited in scope as the financial sector is much larger.
So do we debunk this new number? IT cannot be rejected because the model shows that it worked well. But it should be treated as an academic exercise. There are other economists who have run models and provide that there has actually been deceleration or rather negative growth in GDP post demonetisation which is hard to believe but has been validated by models. Therefore these exercises should be looked at as academic ones and the CSO numbers should be the ones to be used.
This leads to two conclusions. First, the CSO needs to ensure that the credibility of data is retained and should hence be seen as being free from political nudging or else the public would tend to rely more on these academic exercises which are not comprehensive in scope as practically speaking no individual has access to data across the country.
Second, whenever we look at data the number of say 7 percent should be seen in the context of the time series being used to understand whether there is a progress of low growth. This is important. Also creating back series must be eschewed because they become theoretical exercises and just as the case has been led to unnecessary controversy as everyone tries to make political capital. Ideally using common years to extrapolate backwards would be a better way out of finding out how the old series looks with the new methodology. The debate should end at this stage and the data must be used purely for statistical analysis rather than comparative purposes.
We need to stick to the CSO numbers and treat the '2.5 percent overstatement' as a view which is theoretically right but not valid. The CSO on its part must work towards enhancing its credibility.
(The writer is chief economist, CARE ratings)