How is GDP actually measured, and why does the number keep changing?

The National Statistical Office (NSO), part of the Ministry of Statistics and Programme Implementation (MoSPI), is responsible. It publishes the National Accounts Statistics (NAS), a thick document released quarterly and annually. The process is not a single count; it is a system. Every stage of the GDP release has a name and a purpose. The first estimate for a year, the Advance Estimate, arrives in January, before the financial year even ends. Then the Provisional Estimate in May, followed by several Revised Estimates as more data trickles in, and a Final Estimate years later. The calendar is public, and the revisions are built in.

The NSO does not just add up each shop's sales. For large swathes of the economy, it must estimate, using indicators like the Index of Industrial Production, GST collections, corporate financial statements, crop output data, and vehicle sales. A steel plant in Jamshedpur sends monthly output data; the IIP captures that. But the chaiwala outside the plant gate does not. The process is designed to deliver a timely number, not a perfect one, because an economy the size of India's cannot be measured in real time. The fact that the NSO revises its own numbers repeatedly is not a sign of error. It is how an honest statistical office works when better data arrives months, or even years, after the fact.

The Advance Estimates are released in January, two months before the financial year ends. At that point, the NSO has data only for the first half of the year, or at most nine months. For the remaining months, it must project. The core method is the benchmark-indicator approach: the NSO takes a detailed benchmark year, say the last year for which it has full corporate filings and household survey results, and assumes that the relationship between a sector's output and a set of high-frequency indicators (IIP, GST, bank credit, tractor sales) holds constant. Then it tracks those indicators for the current year and scales the benchmark accordingly.

For agriculture, it uses the first advance estimates of crop production. For manufacturing, it relies on the IIP and corporate results for listed companies. For services, it cobbles together railway freight, cargo traffic, credit growth, and telecom subscriber data. Every one of these indicators is a stand-in, not a direct measure of value added. The first estimate, then, is an educated extrapolation built on a model, not a tally of actual output. It gets the direction right more often than not, but the magnitude can be off by a percentage point or more. That is why later revisions matter.

Because the data keeps coming. The Advance Estimates of a financial year are built on partial high-frequency indicators. The Provisional Estimates, released in May after the year ends, incorporate more actual data: the first full-year corporate filings for listed companies, updated agricultural output, and more complete tax collections. Then the First Revised Estimate, a year later, folds in the full results of the Annual Survey of Industries and the audited accounts of non-financial corporations. The Second Revised Estimate incorporates finalised government accounts and updated rural-urban consumption surveys. The Third Revised Estimate, typically three years after the year in question, is the last major revision before the number becomes "final."

Each stage brings in harder, slower-to-collect data, replacing earlier educated guesses with observed outcomes. The GDP growth rate for a given quarter can shift by 0.3 to 0.5 percentage points between the advance and final numbers. That is normal in every country that produces quarterly GDP. The Indian process is more reliant on high-frequency indicators than some, because India still lacks a comprehensive, timely system of quarterly corporate reporting for the unlisted sector. The revision chain is not a conspiracy. It is what an honest statistical office does when better data arrives. Calling it manipulation misunderstands how any large-economy national accounts are built.

A base year is the reference year whose prices are used to compute real GDP, the measure that strips out inflation to reveal actual volume growth. Real GDP in India is currently expressed in 2011-12 prices. That means every rupee of output in any year is converted into what it would have cost in 2011-12, using a set of price deflators. The choice of base year fixes two things: the relative prices used to weight different goods and services, and the basket of products that represents the economy. An old base year distorts the picture because relative prices shift over time. A mobile phone in 2011-12 was expensive and simple; a modern smartphone is cheaper and does vastly more, but a 2011-12 price index may understate its real contribution. Similarly, new services like app-based aggregators barely existed in 2011-12.

When India moved the base year from 2004-05 to 2011-12 in 2015, it also switched the headline measure from 'GDP at factor cost' to 'gross value added at basic prices' and 'GDP at market prices', aligning with the United Nations System of National Accounts. The new series showed faster growth for some overlapping years than the old one, surprising economists and fueling a debate about whether post-2011 growth was overstated. A further move to a more recent base year, something like 2022-23, has been discussed for years but not yet implemented. The delay means the current real GDP numbers are slowly growing less accurate in capturing the changing structure of the economy.

Below we show the current GVA breakdown by broad sector, which the base-year weights attempt to capture. An out-of-date base year would misweight these.

When India switched to the 2011-12 base in 2015, the new methodology raised a question: what would growth look like for years before 2011-12 if the new method were applied backward? In 2018, the NSO released a "back-series" that attempted to answer that. Two versions emerged. The National Statistical Commission's committee produced one set of numbers in July 2018. Then, in November, the NSO, working with the government think-tank NITI Aayog, published a different official version that lowered the growth rates for the years 2004-2014, the period when the previous Congress-led UPA government was in power.

The contrast was stark: the committee's version showed faster growth for those years, the official version slower growth. The former finance minister, who had served under UPA, called the official series a "hatchet job." Some former chief statisticians questioned the process. The government defended the new methodology, but the damage to public trust was done. The episode taught the Indian public that GDP numbers are not just technical exercises; they are political objects, and the choices made in how to construct them can rewrite a government's economic legacy. The back-series remains a touchstone for anyone who suspects that official statistics in India serve the ruling party of the day.

A large chunk. The informal sector, the kirana shops, street vendors, small workshops, and home-based workers who account for perhaps 80-90% of employment, is not captured by corporate tax filings or the Annual Survey of Industries. For these activities, the NSO relies on proxies and surveys that are themselves infrequent. For unorganised manufacturing, for instance, it uses the results of a quinquennial survey and then projects forward using IIP growth for related sectors. For unorganised services, it often uses labour force survey estimates of employment and assumptions about output per worker.

The method essentially takes a benchmark ratio, the relationship between formal and informal activity in a base year, and assumes it holds until the next survey. If the informal sector has shrunk relative to the formal sector (as it likely did during the demonetisation-GST-COVID sequence), the method can overstate informal output. This is the core of the criticism levelled by economists like Arvind Subramanian and colleagues at the Peterson Institute for International Economics. They argue that India's post-2011 growth rate has been overestimated because the methodology implicitly assumes the informal sector is growing in step with the corporate sector, while in reality it was hit much harder. The NSO has defended its approach, and in February 2026 it revised the methodology in part to address such concerns. The debate is not settled. The proportion of GDP that is directly measured versus estimated is not a statistic the NSO publishes, but the critics' point is that the estimated component is large enough to matter for the growth rate.

Trust the direction. Distrust the decimal. The broad trajectory, whether the economy is expanding at 7.4% or contracting, is likely correct. The precise rate, particularly the change from one quarter to the next, is subject to revision as better data replaces earlier projections. This is not unique to India. The United States revises its GDP quarterly and annually, and the revisions are often comparable in magnitude. What is different about India is the thickness of the informal sector and the reliance on dated surveys to estimate it, which makes the initial estimates more model-driven.

The back-series controversy and the ongoing methodological debate remind us that GDP is a constructed number, built on a set of choices about what to count, how to price it, and how to bridge data gaps. Every choice is contestable. That does not make the number worthless; it makes it a best-effort estimate that must be read with the knowledge of how it is built. When the finance minister announces that the economy grew at 7.4%, an informed reader should understand that the "7" is firmer than the ".4," and that the "7" itself rests on assumptions that serious scholars have questioned. Use the number for what it is: a directional signal, not a precise meter reading.