Revisions and the base year

Chapter 4 ended on the Ministry telling you its figures would change. This is how much, and why it is not a failure.

Why a number about the past keeps moving

Three separate things cause a GDP figure to be revised, and they are worth separating because they have different implications.

Better data arrives. The first estimate uses the fast indicators of chapter 4. Later, company filings, tax returns, surveys and audited government accounts come in. The indicator-based guess is replaced by the real thing, sector by sector.

Source agencies revise their own inputs. The press note names this explicitly: "revision in input data made by source agencies". If the Index of Industrial Production is revised, everything built on it moves.

The method or the base changes. The largest and least understood cause, and the subject of the rest of this chapter.

The first two shrink over time and eventually stop. The third does not, because it happens on its own schedule and reaches backwards.

The sequence

An Indian financial year's GDP is published several times before it settles:

Stage	Roughly when
First Advance Estimates	January, before the year has ended
Provisional Estimates	May or June, after it ends
First Revised Estimates	the following February
Second and Third Revised Estimates	in later annual cycles

The press note records the current cycle precisely: the new series was released on 27th February 2026, and "Provisional Estimates of GDP for 2025-26 was released on 5th June 2026."

So a figure quoted in January is a forecast of a year still running, and the figure for the same year quoted three years later has been through at least four vintages. They are not the same number and they were never meant to be.

The practical consequence: when two people disagree about what growth was in some past year, they are often both right and reading different vintages. The first question is "from which release?", not "which of us is wrong?".

The base year change

In February 2026 India moved its national accounts from a 2011-12 base to 2022-23, and this is a larger event than it sounds.

A base year change is not only a change of reference point. It rebuilds:

• the weights — the structure of the economy as it was in the new base year, replacing a picture that had become a decade stale
• the price indices used to deflate — the press note names a new series of output Producer Price Index, a new Banking Services Price Index and a new Index of Industrial Production, all on 2022-23
• the methods — chapter 4's switch to double deflation for manufacturing came with this change
• the sources — updated administrative data across the board

Because real GDP is nominal GDP deflated, changing the deflators changes the real series. Because the weights changed, the sectoral composition changed. And because all of that applies to history as well as to the present, the past got restated.

The press note says exactly how far back:

the Quarterly estimates of GDP from Q1 of 2022-23 to Q4 of 2025-26, along with Provisional Estimates of GDP for the FY 2025-26 has been updated

Four years of quarterly history, republished.

What that means for anything you read

Three consequences, in increasing order of how much trouble they cause.

Growth rates for settled years changed. A quarter that was reported at one rate when it happened now carries a different one. Nothing about the past changed; the lens did.

Old and new series cannot be joined. A chart running from 2015 to 2026 crosses a break. Splicing the two without saying so produces a line that looks continuous and is not, and this is the single most common error in long-run economic charts. The honest options are to show the break, to show only one series, or to use a back-cast if the statistical office has published one.

Comparisons across the break need care. The Q1 2026-27 press note reports growth of 7.8%, against 6.9% for the same quarter a year earlier — and both of those are on the new 2022-23 series, which is why they can be compared. A figure for Q1 2025-26 taken from the press note issued in August 2025 would be on the old series and would not be comparable with either. The exercise at the top of this chapter is to go and find out by how much.

Why rebasing is right anyway

It would be easy to read all this as a reason for suspicion. It is the opposite.

A 2011-12 basket of weights, deflators and methods, used in 2026, would be measuring an economy that no longer exists — one with a different share of services, different manufacturing processes, different payment systems and different prices. The estimate would get quietly worse every year while appearing perfectly stable.

Rebasing is the mechanism by which a statistical system stays accurate at the cost of being comparable. Stability and accuracy are in direct conflict here, and accuracy is the right thing to choose — provided the break is disclosed, which it is, in the release, in the paragraph the exercise sends you to.

The thing to carry forward is narrower and more useful than suspicion: a growth rate has a vintage and a series, as well as a value, and quoting it without them is the same error chapters 2 and 3 found in GVA versus GDP and real versus nominal.

The point

GDP figures are revised because better data arrives, because source agencies revise their inputs, and because the base and the methods change — and only the third reaches backwards. India moved to a 2022-23 base in February 2026, which rebuilt the weights, the deflators and the manufacturing method, and restated every quarter from Q1 of 2022-23 onwards. Old and new series cannot be spliced. A growth rate therefore has a vintage and a series as well as a value, and the first question about any historical figure is which release it came from.