India’s GDP debate: Right questions, wrong numbers

If the true overestimation were really 22 per cent of GDP, the official revision would look nothing like it does. That gap is not a detail to be explained away

How accurately does India’s GDP capture what is happening in the economy? It is a question serious economists have been wrestling with for some years. The February 27 revision to the national accounts methodology — developed through what even the authors of a recent paper on the issue describe as “commendable consultations” — is the most substantive effort to improve GDP estimation, addressing deflator choices, the treatment of the informal sector, and the use of administrative data. Into this conversation steps a working paper by Abhishek Anand, Josh Felman, and Arvind Subramanian, published by the Peterson Institute in March 2026, claiming to provide “new evidence” of systematic misestimation.

The paper’s two core methodological complaints — that India used WPI-based deflators tracking commodity and oil prices rather than prices of actual production, and that it used formal-sector corporate data as a proxy for informal-sector activity — are not new. They have been in the academic literature since at least 2016, including in Subramanian’s own earlier work. These choices were driven by structural data limitations rather than analytical oversight. In the absence of regular annual data on informal-sector output and a comprehensive Producer Price Index, the use of WPI-based deflators and formal-sector corporate data was pragmatic and arguably the most viable option available.

The paper’s assertion that the WPI is an inappropriate deflator and the CPI a better alternative is also misplaced. India’s WPI is conceptually close to a Producer Price Index, which is what international recommendations prescribe. The CPI reflects price movements relevant to private consumption; goods and services such as steel, cement, minerals, chemicals, IT services, trade, and professional services are produced primarily for industrial use and rarely appear in the CPI basket. Real estimates for many services also derive from a much broader set of volume indicators than the paper considers.

With annual data from the unincorporated sector survey and labour force surveys now available, the base-year revision to 2022-23 addressed several of these issues. The February 2026 methodology revision — described by the authors as resulting from commendable consultations — was in any case announced before this paper appeared.

The informal sector corrections are shakier still. The paper uses unincorporated enterprise survey data covering only half the informal economy — construction is excluded — and applies the performance of surveyed sectors to the entire informal economy to construct its “lower bound” estimates. That phrase, lower bound, is doing a lot of work. The survey data also exclude housing services, which account for a substantial share of informal-sector GVA in the national accounts. Using raw survey figures without accounting for this omission structurally understates informal-sector performance and overstates the formal-informal divergence that sits at the heart of the mismeasurement argument.

Moreover, the paper’s use of inter-survey growth rates to infer divergence from official GVA growth overlooks the fact that India’s National Accounts rely solely on GVA per worker from the unincorporated sector survey. There is also a circularity problem: The paper uses corporate sales data as its primary benchmark to demonstrate GDP mismeasurement, which assumes corporate sales are the correct benchmark — precisely what is at issue. Worse, corporate sales measure turnover rather than value added. Changes in input cost structures and productivity can drive large wedges between sales growth and GVA growth that have nothing to do with measurement error.

The authors also argue that using MCA data may lead to overestimation. But the MCA database provides comprehensive coverage based on actual reported data, not survey-derived estimates like the ASI; corrections relying solely on ASI data for organised manufacturing are therefore unacceptable. More fundamentally, the paper never confronts the fact that India’s economy changed substantially after 2015: Rapid expansion of the digital economy, strong growth in financial services and insurance, the emergence of India as a major hub for Global Capability Centres, and deliberate policy-driven formalisation. These are precisely the activities that do not show up in energy consumption, trade volumes, or bank credit data — the very indicators the paper treats as reliable proxies for aggregate output.

Some weakening of correlation with the selected variables is therefore not unexpected, given the significant changes in indicator relevance over time. The use of direct tax growth ignores major tax policy rationalisation, most notably the corporate tax rate reduction introduced in 2019. The Index of Industrial Production does not adequately capture structural changes in India’s manufacturing sector since 2011-12. And bank credit overlooks the growing role of alternative financing sources in supporting investment and consumption.

But the most damaging issue is one the paper never addresses. Subramanian, as India’s Chief Economic Adviser, was the principal author of the Economic Survey of India 2017-18. Chapter 2 of that Survey, using early GST data, found that purely informal firms — outside both GST and social security coverage — accounted for only about 7 per cent of total economic turnover, even though they represented 87 per cent of firms by number. Firms within the GST net accounted for nearly 80 per cent of total turnover. The Survey documented rapid, self-reinforcing formalisation, as small firms registered voluntarily to access input tax credits from larger buyers.

The 2026 paper’s mismeasurement argument rests on the claim that the informal sector accounts for roughly 44 per cent of GVA and that formal and informal sectors diverged so sharply after 2015 that using formal data as a proxy led to sustained overestimation. Both claims are substantially weakened by what the 2017-18 Survey found. If 80 per cent of turnover was already in the formal economy by late 2017, and formalisation was accelerating, the informal sector’s weight would have shrunk further still in the years that followed. The 2026 paper makes no attempt to reconcile its baseline assumption with the evidence its own author produced. The omission is not peripheral: The informal sector mismeasurement component accounts for between 0.4 and 0.8 percentage points of the paper’s estimated annual overestimation. Weaken it, and the quantitative edifice wobbles considerably.

The strongest empirical check on the paper’s claims is the official February 2026 revision itself. The authors describe the consultation process behind it as commendable. The revision, produced by statisticians working with full access to administrative data and without the methodological shortcuts the paper employs, produced a substantially more modest adjustment than the paper’s estimates imply. If the true overestimation were really 22 per cent of GDP, the official revision would look nothing like it does. That gap between the paper’s claims and the official outcome is not a detail to be explained away. Knowing where to stop is part of the discipline of empirical work. This paper does not stop at the right place.

Nageswaran and Garg are the Chief Economic Adviser to Government of India and Secretary, MoSPI, respectively. Views are personal