National Council of Applied Economic Research
Opinion: Poonam Gupta.
Each sequential release of GDP estimates by MoSPI (Ministry of Statistics and Programme Implementation) generates considerable excitement. Yet, perhaps many of us are unaware that each year’s estimates are revised six times over a stretch of three years before they are finalized.
For example, the first estimate of growth for the current year 2023-24, the First Advance Estimate (FAE), would be released in the first week of January 2024. A Second Advance Estimate (SAE) would be released after about a month and a half in end-February 2024. A Provisional Estimate (PE) would be released three months later in end-May 2024.
These three “early estimates” would then be followed by annual revisions spanning nearly a three-year period. The First Revised Estimate (FRE) would be released in end-February 2025; the Second Revised Estimate (SRE) and Third Revised Estimate (TRE) would be released exactly one and two years thereafter, that is, in February 2026 and February 2027, respectively.
Thus, as per the current practice, the final estimates for GDP and GDP growth are available only three years after the end of the respective fiscal year.
This practice raises certain inevitable questions. Does it really need to be such a long-drawn process? Does the release of six different estimates over a three-year period ensure better accuracy? Is there a case to consolidate some of these iterations and move to only three estimates: an early estimate; a provisional estimate; and a revised estimate, spread roughly over a period of a year-and-a-half?
In order to answer these questions, we analysed the data for four recent years for which all the six estimates are available (Table 1).
Three features of the various estimates stand out from this analysis, which then lead to three corresponding recommendations.
First, given the limited time in the interregnum between the FAE and the SAE, the two estimates are very similar. In fact, in two of the four years, the two estimates were identical. The SAE was upgraded by 0.1 percentage points in one year, and downgraded by 0.2 percentage points in another year.
This pattern lends itself to the obvious recommendation that the FAE and SAE should be merged into one estimate. One may simply release these numbers as the Advance Estimate in February, around the time when, as per the current process, the SAE is released.
Second, once all the data available up to two years after the end of the fiscal year has been utilized to prepare the SRE, very little new information is available in the subsequent year. Therefore, unsurprisingly, the SRE and the TRE have also been quite similar to each other.
As can be gleaned from Table 1, the SRE and TRE were identical in one of the four years; the TRE was upgraded by 0.1 percentage point in one year and by 0.2 percentage points in another year; and was downgraded by 0.2 percent in one year. These revisions have not only been rather minor, but in the medium term, they mostly even out, as a result of which the size of the economy is not impacted in any specific direction.
Therefore, doing away with TREs would be largely inconsequential for the accuracy of the estimates, but it would instead make the process shorter, nimbler, and more streamlined.
A final recommendation would be to combine the FRE and SRE into one estimate. The GDP estimates are revised significantly between these two estimates. The SRE, underpinned by more detailed information, is more accurate than the FRE. It would thus be prudent to release the combined estimate at an appropriate time between the two currently designated periods. If the data collection and its compilation can be accelerated with some effort, one could perhaps achieve a reliable revised estimate about 14-15 months after the fiscal year has ended.
A transition step may be go to four estimates first by eliminating the FAE and TRE; and subsequently go to three estimates by merging the FRE and SRE within two years.
Figure 1 graphically depicts the existing and recommended processes of arriving at the GDP estimates.
The final slate would then look as follows: an Advance Estimate, a Provisional Estimate, and a Revised Estimate, all three estimates would be available within a span of 16 months. Simple, precise, and agile.
The writer is director general of NCAER, and a member of the Economic Advisory Council to the Prime Minister of India. Shubhashree Jha contributed to this piece.