Did India Change GDP to Inflate Growth?

TL;DR

The claim that the Modi government changed one GDP formula in 2015 to make growth look better is false. India’s 2011–12 series was a bundle of changes—new data, new sector boundaries, new prices, and international accounting standards. Work on its most disputed corporate database began under the UPA in 2013; the NDA released the series in 2015.

But the reassuring counterclaim—that this was merely routine statistical maintenance—is also too comfortable. The series relied on an imperfect corporate database, used formal-sector indicators to move parts of the informal economy, and often removed inflation with a single price index. Those choices could overstate real growth in particular periods. A major independent reconstruction puts the post-2011 overstatement at 1.5–2 percentage points a year, but its substitutes and assumptions are too contestable to call it India’s “true” GDP.

The clearest institutional failure came in 2018. A new historical series sharply lowered growth during the UPA years; NITI Aayog helped present it, while the National Statistical Commission says it saw it only post facto. That establishes compromised oversight. It does not establish fabricated data.

Editorial illustration generated for Thenkai; it represents revision and audit, not a government document.

In January 2015, India published a GDP revision with politically convenient and statistically confusing results.

It lowered the estimated size of the economy in 2011–12 by 2%. Yet it raised real growth for 2012–13 from 4.7% to 5.1%, and for 2013–14 from 5.0% to 6.9%.¹

One number said the economy had been smaller. Another said it had grown faster.

That apparent contradiction is the doorway into the whole dispute. “GDP” is not one reading taken from one national meter. It is a constructed estimate: millions of transactions grouped into sectors, valued, adjusted for prices, and extrapolated where direct data do not exist. Change the benchmark levels, coverage, weights, proxies, or deflators, and nominal GDP and real growth can move in different directions.

So the serious question is not whether the method changed. It did, as it has under governments of several parties. The question is whether the changes improved measurement, introduced predictable biases, or were bent toward a political result.

The evidence supports an awkward answer: India modernized its national accounts through technically defensible changes, embedded serious measurement weaknesses in that modernization, and later damaged trust through a politically compromised back-series process. None of that is proof that officials fabricated GDP.

There Was Never One Formula

India has changed its GDP base repeatedly: from 1948–49 to 1960–61, 1970–71, 1980–81, 1993–94, 1999–2000, 2004–05, 2011–12, and now 2022–23. Congress, Janata, and BJP-led governments have all presided over revisions.

This is normal. An economy changes faster than a statistical system. New industries appear, old price weights become absurd, and better tax, survey, or company data become available. A base-year revision is supposed to bring the map closer to the territory.

The 2015 change was unusually large because it was not merely a new price base. It moved toward the UN’s 2008 System of National Accounts; emphasized institutional sectors; reported GVA at basic prices and GDP at market prices; capitalized research and development; expanded the coverage of financial institutions and local bodies; consolidated multi-establishment companies; and replaced a small RBI company sample with filings from the MCA21 corporate database.¹

That last change became the political shorthand for the entire revision. But its timeline matters. An expert subcommittee was commissioned on 11 September 2013—under the UPA government—to examine the private corporate sector and MCA21. Its report and the new series arrived under the NDA in 2015.²

That does not absolve the government that released, defended, and extended the series. It does block the simplest conspiracy story. The architecture crossed governments.

The Revision Moved More Than Growth

The immediate overlap figures show why slogans mislead:

2015 release	Old series	New series	Change
2011–12 nominal GDP	₹90.10 lakh crore	₹88.32 lakh crore	−2.0%
2012–13 real GDP growth	4.7%	5.1%	+0.4 pp
2013–14 real GDP growth	5.0%	6.9%	+1.9 pp

If the sole instruction had been “make the economy look larger,” lowering the benchmark level would be an odd way to comply. But if the political prize were a faster growth narrative, the upward real-growth revisions clearly helped.

Neither observation settles the matter. The level was rebuilt using changed coverage and concepts. Growth depended on how that rebuilt value was moved through time and stripped of inflation. The politically salient result can emerge from a technically chosen method without that method being invented for the result.

The right place to test manipulation is therefore not the direction of one headline revision. It is the mechanism underneath it.

Where the New Series Was Fragile

Three weaknesses matter more than the rhetoric around them.

First, MCA21 was both an improvement and a statistical risk. Replacing a sample of roughly 2,500 companies with hundreds of thousands of filings broadened coverage dramatically. But legal registration does not prove economic activity. Filing companies are not a random sample of all eligible companies; industry codes can be wrong; and non-filers must be estimated. Nagaraj and Srinivasan argued that the direction and size of the resulting selection bias were unknown. B. N. Goldar countered that reporting firms covered most corporate GVA and that a stable scaling factor changes levels more readily than growth.³

Both claims can be true. More data is not automatically cleaner data.

Second, India’s informal economy cannot be measured directly every quarter or even every year. Benchmark surveys establish a level; between them, statisticians move parts of unincorporated activity with organized-sector indicators. That is most dangerous when the two worlds diverge. Demonetisation, GST transition, and COVID could hurt cash-dependent small firms while larger formal firms gained share. A proxy can then mistake formalization for economy-wide growth.

Third, the 2011–12 series frequently used single deflation: one price index to turn nominal value added into “real” value added, effectively assuming input and output prices move together. If commodity input prices fall faster than output prices, measured real value added can grow too quickly. The IMF continued to flag the breadth of this practice years after the new series began.⁴

These are not allegations about ministerial intent. They are mechanisms through which a sincere statistical system can produce biased estimates.

The Strongest Case That Growth Was Overstated

In 2026, Abhishek Anand, Josh Felman, and Arvind Subramanian published the most ambitious reconstruction yet. They combined macroeconomic indicators with sector-level corrections and estimated average GVA growth from 2011–12 to 2022–23 at 4.0–4.4%, against the official 5.9%. Their attribution was roughly one percentage point to deflators, 0.4–0.8 points to informal-sector measurement, and about 0.1 point to formal-sector adjustment.⁵

That is too large to dismiss as statistical trivia. If broadly right, it changes how we interpret job creation, productivity, fiscal space, and the performance of two governments.

But it is not an audited replacement series. The authors substitute consumer-price measures where producer prices are unavailable, bridge informal-enterprise surveys that are not identical, bound sectors with missing data, and exclude part of informal construction. Together these choices create a model, not an observed economy.

Government-linked rebuttals also land real blows. The 2019–20 Economic Survey showed that earlier cross-country claims were sensitive to specifications and weak pre-change trends.⁶ A 2026 EAC-PM replication found that none of five annual correlation changes was statistically significant, that autocorrelation reduced the effective sample, that including COVID years cut the estimated India effect, and that 2011–12 was not uniquely selected as the strongest break.⁷

Yet the rebuttal does not close the case. The EAC-PM note itself identifies the relationship between WPI, CPI, and GVA-to-sales as the critique’s strongest robust result. It does not build a superior series from better producer-price and informal-sector data.

So a claim survives, but in narrower form: material overstatement in parts of the post-2011 period is plausible; the exact all-economy correction is unknown. “GDP was overstated by precisely 2 percentage points every year” exceeds the evidence.

The Year the Past Slowed Down

The most politically damaging episode came three years after the original rebase.

Because reliable MCA21 history did not extend before 2011–12, the new series initially lacked a comparable past. A National Statistical Commission-appointed committee led by Sudipto Mundle released a draft in 2018. Its illustrative production-shift calculation left the old high-growth era intact. Later that year, MoSPI released the official back series—and the past changed sharply.⁸⁹

Published construction	Average real GDP growth, 2005–06 to 2011–12
Old 2004–05 series	8.23%
Mundle committee illustration	8.61%
Official November 2018 back series	6.91%

The official version used direct recalculation where data permitted and ratios or splicing where they did not. MoSPI gave intelligible reasons for lower services growth: a later unorganized-enterprise survey came in below earlier projections; sales-tax data replaced gross-trading-income estimates; telecom minutes replaced subscriber counts; and RBI treatment changed.

There is no pristine like-for-like database that can tell us which path is “the” truth. The Mundle number was a draft illustration, not suppressed official GDP. But the process used to choose and release the final path crossed an institutional line.

NITI Aayog, a policy body chaired by the prime minister, reviewed and co-presented the release.¹⁰ The National Statistical Commission’s own annual report says the National Accounts Division presented the back series to the Commission post facto.¹¹ Two independent commission members later resigned amid wider complaints that the body had been sidelined.¹²

That is the clearest exposure in this story. The official statistical watchdog did not exercise meaningful prior oversight over a politically explosive historical revision, while a government policy body entered the release process.

It is evidence of weakened statistical separation. It is not a document ordering statisticians to hit 6.91%.

Political Interference Is Not the Same as Fabrication

This distinction is easy to caricature, but essential.

Political interference can mean controlling release timing, sidelining independent review, choosing among defensible methods, presenting uncertainty as settled, or using policy bodies to validate official statistics. Those acts can bias institutions and destroy trust even when every spreadsheet cell was calculated as described.

Fabrication is a stronger allegation: data or methods knowingly altered to reach a predetermined number. The reviewed record contains no instruction, leaked target, audit trail, or replicated forensic result establishing that.

The absence of proof is not proof of institutional innocence. It changes the honest verb. The 2018 process was compromised. Deliberate numerical fabrication remains unproven.

This is more serious than splitting the difference. If every methodological dispute becomes “fraud,” authorities can dismiss legitimate critics as partisan. If every internationally recognizable method is treated as neutral, governments can politicize oversight while pointing to technical paperwork. Both moves protect power from inspection.

The 2026 Revision Is an Uncomfortable Witness

India rebased GDP again in February 2026, this time to 2022–23. The new framework uses annual informal-enterprise and labour-force surveys, double deflation for agriculture and manufacturing, more granular prices, activity-wise allocation for multi-activity firms, supply-use balancing, and new GST, public-finance, and vehicle-registration data.¹³

Those are substantial repairs to known weaknesses. They are also evidence that the old method was not good enough.

But the revisions did not move in one politically useful direction:

2026 release	Earlier estimate	New estimate	Change
2022–23 nominal GDP	₹268.90 lakh crore	₹261.18 lakh crore	−2.9%
2023–24 real GDP growth	9.2%	7.2%	−2.0 pp
2024–25 real GDP growth	6.5%	7.1%	+0.6 pp

The same exercise lowered one year’s growth and raised the next. It also cut private-consumption levels by roughly 10–12% and produced large, opposing sector revisions. That weakens the theory of a permanent one-way growth machine. It strengthens the case for publishing far better revision decompositions.

The decisive test is still pending. As of this publication, the 2022–23-base historical back series is scheduled for December 2026. Until it arrives, confident comparisons across earlier governments remain hostage to incompatible statistical vintages.

Why the Latest 7.8% “Exposure” Fails

The Quint article that prompted this investigation makes a valid demand for transparency, then tries to prove too much.¹⁴

It compares ₹88.27 lakh crore for Q1 2026–27 under the new 2022–23-base series with ₹86.05 lakh crore for Q1 2025–26 as published under the old 2011–12-base series. That produces 2.6% nominal growth. Subtract an assumed 2–2.5% deflator, and the article arrives near zero real growth instead of the official 7.8%.

But those two rupee values belong to different accounting systems. The ₹86.05 lakh crore old-base estimate became ₹80.32 lakh crore when the new base was introduced in February 2026, ₹80.44 lakh crore in June, and ₹80.00 lakh crore after new price, industrial-production, and administrative inputs in August. Comparing the current ₹88.27 lakh crore with the already published February same-base ₹80.32 lakh crore gives about 9.9% nominal growth, close to the updated 10.3%—not 2.6%.¹⁵¹⁶

The official constant-price pair, ₹81.36 lakh crore over ₹75.46 lakh crore, yields 7.8%. That does not make 7.8% metaphysically true. Quarterly GDP is provisional and the new series has already moved dramatically. But a cross-base ratio followed by an assumed economy-wide deflator does not expose retrofitting. It exposes a comparison error.

The article’s strongest point survives its failed arithmetic: when the benchmark quarter changes by trillions of rupees, the compiler should show the bridge clearly enough that a reader does not need to reconstruct it across four releases.

What These Revisions Change in Real Life

GDP controversy can sound like a fight among spreadsheet custodians. It is not.

If growth is overstated, the government and central bank may infer that the economy is closer to capacity than it is, tolerate tighter policy, or underreact to weak employment and small-enterprise distress. If it is understated, the opposite mistakes become possible.

The denominator also governs political facts. Debt-to-GDP, deficit-to-GDP, tax-to-GDP, health spending, defence spending, and credit intensity all change when GDP changes—even if the rupee numerator does not. A 3.8% downward revision to nominal GDP mechanically makes every unchanged liability-to-GDP ratio about 4% larger.

Historical revisions reshape accountability. The 2018 back series transformed a story of rapid pre-2011 growth followed by slowdown into one of much flatter performance. The method gap—1.3 to 1.7 percentage points across competing historical constructions—is larger than many partisan claims about one regime outperforming another.

That is why any “UPA versus NDA growth” table without a named series, vintage, and treatment of transition years is not a neutral comparison. It is a choice of statistical past disguised as arithmetic.

What Survived the Audit

The government did not invent GDP rebasing in 2015. The most disputed data reform began before it took office. The revision included internationally orthodox improvements, and later changes have moved headline growth both up and down.

The critics are nevertheless right about something important. India tried to measure a vast, fast-changing, partly informal economy with corporate records that were not fully auditable, sparse household-enterprise benchmarks, formal-sector proxies, and weak price tools. The 2026 repair confirms that several concerns were not imaginary.

And the institutional criticism is stronger still. In 2018, a lower historical growth path with obvious partisan consequences was co-presented by NITI Aayog before the statistical commission received its post-facto briefing. A statistical system does not preserve independence merely by producing a technical note after political boundaries have blurred.

The truth is not “trust the number” or “the number is fake.” It is that GDP is an estimate whose legitimacy depends on methods, revision trails, independent oversight, and the humility to state what the data cannot identify. India’s largest credibility failure was not changing the estimate. It was making the chain of judgment harder to trust.

Where I’d Hold This Loosely

The forthcoming 2022–23-base historical back series could materially alter the comparison. It may show that annual informal-sector data and better deflation close much of the old gap, widen it, or move different periods in opposite directions. Any verdict on long-run regime performance should remain provisional until that bridge is published and independently reproduced.

The 1.5–2 percentage-point overstatement estimate is the strongest quantified challenge, not established ground truth. Better producer-price indices, harmonized informal-enterprise surveys, or alternative treatment of missing sectors could change its magnitude. The official rebuttals weaken its indicator-based identification without solving the underlying measurement problem.

Finally, the reviewed public record cannot reveal every informal conversation inside government. “No proof of fabrication” is a statement about evidence, not a certification of motive. If internal instructions, auditable source-data anomalies, or a reproducible alternative series emerge, the conclusion should change.