GDP data must not be released in a hurry if so far, significant changes have had to be made, and consistently so.
The CSO releases on GDP have tended to be controversial given the back series that was brought out, and which changed the course of debate. However, from the point of view of a user of such data, there is growing concern over the constant revisions made to these numbers. Such revisions are inevitable given that a large part of our economy is unorganised, leaving several imputations to be made. Even for the organised segment, data flow is not regular and lacks consistency, making estimating GDP challenging. For example, the IIP used for the unorganised sector, to arrive at GDP growth, is susceptible to revisions that automatically change the GVA. The WPI indices, used as deflators, are also susceptible to change.
In the latest release, the GDP growth forecasts for Q1 and Q2 for FY20 have been revised quite significantly, from 5% and 4.5%, respectively to 5.6% and 5.1%, respectively. Against these two numbers, the Q3 forecast is now at 4.7%, which indicates a slowdown rather than an increase given that 4.5% was the Q2 number as revealed in November. In fact, it was widely believed that Q3 should be better than Q2 due to the festive demand phenomenon, and low base year advantage. However, this did not happen. But, an upward revision cannot be ruled out, as was the case for the first two quarters.
These numbers lead to some confusion because when the first advance estimate is put out in early January, it is based on the extrapolation of Q1 and Q2 numbers. This annual projection would then be used for all the calculations and forecast made in the Economic Survey, the Union Budget, and by RBI. Now, with higher revised estimates for Q1 and Q2 of FY20, the full year forecast should have gone up, but has been taken to be unchanged in the second advance estimates released on February 28.
Interestingly, these revisions for the first half involve an addition of Rs 42,000 crore in a total increase of Rs 3.63 lakh crore in the first half of the year. This revision is around 12% of the total increase. This is quite significant as it indicates a forecast error of just over 10%. In fact, after revising these two numbers, the CSO has an unchanged estimate of 5% for FY20. This is surprising given that there is no reference made to corona virus, and there is a sharp positive base effect expected in Q4, which should ideally prop up the Q4 and full year growth numbers.
Quite clearly, there is need to revisit the process of data collection and dissemination. In our attempt to provide high frequency data, the probability of providing information that is amenable to constant revisions increases. The accompanying graphic shows how different CSO releases have projected these data points from FY16 onwards.
Starting from the bottom—the starting point of the first estimate provided—to the top, the latest number, shows some interesting patterns. Growth for FY16 has moved from 7.6% to 8.2% over two and a half years, after which it stabilised. In FY17, the year of the infamous demonetisation process, the initial forecast was 7.1%—lower than that of FY16, but then increased sharply by 1.2%, and stabilised in the January 2020 release. This changed the narrative from one which said that 0.5% growth in GDP got shaved off due to demonetisation to one which showed that the economy bloomed like never before to reach the highest mark in seven years! But, this was revealed almost two years after the event.
For FY18, the increase was from 6.6% to 7%, and came down from 7.2% in between. Here, too, the narrative changed from initially saying that GST impacted the economy in a negative manner to one that said it was not too bad.
Though GDP growth for FY19 can still be revised, so far, the forecast has stepped downwards from 7.2% to 6.1%. The sharp fall came in January 2020. Hence, in three of the four years, there were upward revisions, and a downward one in the fourth. For FY20, it needs to be seen if the 5%forecast will hold or be revised upwards. Typically, the final number should be out within a span of a year and a half to two years.
While revisions may be a part of the process of forecasting, the CSO numbers become a part of all business planning as the present GDP growth number is always taken by companies to forecast future GDP growth, which is then linked with business targets and goals. While the absolute numbers that are put out, and which enter all planning models, tend to overstate or understate projections, it gets stickier when the direction of growth changes.
During demonetisation, the initial indicator was that growth had slowed down, which would have been used to give a positive thrust to almost all corporates when planning for FY18. However, the revised number for FY17, which was higher, would have shown most plan forecasts to be conservative. Similarly, all macro targets benchmarked to the GDP number (usually, nominal)—fiscal deficit, current account deficit, investment ratio, etc—would get distorted. It can be seen that frequent revisions of such magnitude lead to asymmetrical forecasts for companies, and upsets plans.
The question to be posed is whether it makes any sense to release numbers in what can be called a hurry if there are, consistently, significant revisions to be made. While the prevalent view is that it is preferable to have some information rather than no information, the fact that there are a plethora of decisions being taken based on these numbers calls for introspection. Today, the demand for all industries—steel, coal, metals, automobiles, consumer goods, etc—are based on projections of GDP, which is the standard variable used globally as it tends to give the best results over a longer period of time.
A suggestion would be to stop making extrapolations in January based on H1 numbers, and have only the May figures, which are based on more facts. There would still be major revisions as the data pockets are fragile. Also, the sectors responsible for these divergences need to be reworked. Understandably, this will be an evolving process, but reducing the noise caused by constant revisions is a necessity.
The writer is Chief Economist, CARE Ratings. Views are personal