India's Ministry of Statistics and Programme Implementation is currently publishing two accounts of the same financial year, and the arithmetic separating them is four times larger than the difference that has been reported. Under the old 2011-12 base year, the ministry's First Advance Estimates put 2025-26 nominal GDP at ₹357.14 lakh crore; under the 2022-23 base introduced on 27 February 2026, the Provisional Estimates put it at ₹346.36 lakh crore — ₹10.78 lakh crore lower, or 3.02 per cent. Gross value added, the measure before net indirect taxes, falls by ₹8.61 lakh crore, from ₹323.48 lakh crore to ₹314.87 lakh crore. But the eight industry lines beneath that total moved ₹34.48 lakh crore between them to produce it, and two of the four largest moves point up, not down. A lakh crore is a trillion rupees, and MoSPI's data API reports in crore — ten million rupees — so the new series' ₹34,635,638 crore is the ₹346.36 lakh crore of the press note.
Both headline levels come from the ministry's own press notes rather than from any reconstruction. The First Advance Estimates note of January 2026 gives ₹357.14 lakh crore for FY 2025-26 against ₹330.68 lakh crore the year before; the Provisional Estimates note of May 2026 gives ₹346.36 lakh crore against ₹318.07 lakh crore. Those are different vintages, and the difference matters: the old-base figure is a January forecast, the new-base figure a May estimate carrying four more months of actual data, so not all of the ₹10.78 lakh crore is a base-change effect. The one strictly like-for-like comparison the overlap allows is 2022-23, where both series publish Final estimates — ₹268.90 lakh crore on the old base against ₹261.18 lakh crore on the new, a gap of ₹7.73 lakh crore, or 2.87 per cent. The four-year pattern is consistent, with nominal GDP reductions of 2.87, 3.78 and 3.81 per cent for 2022-23, 2023-24 and 2024-25.
The industry detail sits in the ministry's National Accounts API, which serves both bases from the same endpoint at current prices. Three lines fall. Trade, hotels, transport, communication and broadcasting-related services drops ₹11.49 lakh crore, from ₹56.41 lakh crore to ₹44.92 lakh crore, a fifth of its old-series level; public administration, defence and other services drops ₹8.61 lakh crore, or 17.4 per cent; construction gives up ₹1.45 lakh crore. Five lines rise. Financial, real estate and professional services gains ₹8.55 lakh crore, up 11.2 per cent, agriculture ₹2.37 lakh crore, manufacturing ₹1.08 lakh crore, mining ₹0.90 lakh crore on a 17.7 per cent increase, and utilities ₹0.04 lakh crore. Gross reductions of ₹21.55 lakh crore against gross increases of ₹12.94 lakh crore leave the ₹8.61 lakh crore net.
That structure is the reason a tempting statistic should not be written. The trade line's ₹11.49 lakh crore fall is 133.4 per cent of the ₹8.61 lakh crore national reduction, which reads as concentration until the next line is checked: public administration, defence and other services is 100.0 per cent of the same reduction on its own, financial, real estate and professional services is minus 99.3 per cent of it, and agriculture minus 27.6 per cent. When three separate industry lines each account for roughly the whole of a total, the total is a residual left over from much bigger opposing moves, and no single line explains it. The share-of-the-total framing works on four different lines here and therefore establishes nothing about any of them.
The ministry has already published why the lines move. Its 27 February explainer, issued through the Press Information Bureau, describes the segregation of multi-activity corporations: value added by diversified enterprises was previously assigned wholly to their principal activity, and with company filings under MGT-7 and MGT-7A now available it is “distributed across activities more accurately”. The same document says household-sector levels no longer rely on growth rates carried forward between benchmark surveys or on proxy indicators, but on the annual ASUSE enterprise survey and the PLFS labour survey. A reallocation of output between industries, plus a fresh level for the informal-heavy sectors, is precisely the pattern the numbers show — a large transfer out of trade and public administration, a large transfer into financial and real estate services, and a small net residual at the bottom.
The direction of these revisions was identified in March, five months ahead of this decomposition, and by two separate readers. A DMI Finance review of the new GDP and GVA series named public administration and trade, hotel and transport as the downward-revised services lines and financial services, real estate and professional services as upward, attributed the shift to direct annual surveys replacing indicator-based extrapolation, and reported the same 2.9 to 3.8 per cent band of nominal reductions across FY23 to FY25. An Ashoka University policy discussion paper by Dani and Shrimal put it in share terms, finding that trade and public administration “witness significant declines of 3 ppts and 2 ppts respectively” in their contribution to FY26 nominal GVA; the API reproduces those shares, with trade falling from 17.44 to 14.27 per cent of GVA and public administration from 15.28 to 12.96 per cent. What neither published is the rupee size of the moves or the ratio between them: ₹34.48 lakh crore of gross reallocation for ₹8.61 lakh crore of net change, four to one.
The rebasing produced one juxtaposition worth stating carefully. The 27 February release reported the trade, hotels and transport sector growing 10.1 per cent at constant prices in 2025-26, against 7.6 per cent real growth for the economy as a whole, and the API's constant-price rows reproduce it at 10.13 per cent. In that same release the sector's nominal level under the new base sits about a fifth below the old series' level for the same year. Neither figure contradicts the other, since a sector can grow quickly from a lower base, but the two travel together in one document and only the growth rate has been widely quoted.
The double account has a closing date. MoSPI publishes Q1 estimates for 2026-27 on 31 August, six days out on its advance release calendar, and that print will be new-series only, with no old-base quarterly running alongside it for comparison. The back series is the reason the two accounts coexist at all: the ministry's explainer says back-series data “is expected to be released by December 2026”, recalculated on the revised methodology up to the previous base year and then linked at a disaggregated level to extend the series to 1950-51. Until it arrives the old base stays on the API because nothing else covers the earlier years, and the overlap that creates is the only window in which the ₹34.48 lakh crore of reallocation can be measured from published data at all. After December there will be one account, and the reshuffle will be inside it.