Paper Series

Unstable evidence, stable conclusion: A critique of the GDP overestimation thesis of Anand, Felman, and Subramanian

April 28, 2026

Vidhu Shekhar and Surabhi Mukhopadhyay

Abstract

The Peterson Institute working paper by Anand et al. (2026) claims that Indian GDP growth has been overestimated by 1.5 to 1.9 percentage points per year from 2011 to 2023, with an implied level of overstatement of 22 percent as of 2025. The paper is a direct continuation of the thesis advanced in Subramanian (2019), a Harvard working paper in which Subramanian (a co-author on the 2026 paper) argued the same position as sole author seven years earlier. This paper establishes that the 2026 magnitude claim is not supported by the evidence presented. Six findings support this conclusion. First, of the seventeen macro indicators the 2019 paper used to establish post-2011 correlation breakdown, thirteen are absent from the 2026 paper; on the extended 2012 to 2024 window, ten of the eleven indicators flagged by the 2019 paper as negatively correlated with GDP show positive correlation, including IIP Manufacturing, whose divergence from formal manufacturing GVA the 2019 paper used to argue for a substantial portion of the overall overestimation (on the order of 1 percentage point), and which now correlates with GDP at 0.92. Second, the informal sector contribution of 0.4 to 0.8 percentage points is computed as a compound annual growth rate between the NSS 73rd Round (2015 to 2016) and ASUSE 2023 to 2024, separated by a six-year gap and by a change in survey instrument, sampling frame, and coverage. Third, the sales-to-GVA wedge presented as independent evidence is partly mechanical: it correlates with the CPI-to-WPI gap at 0.81 by construction, because corporate sales are deflated by CPI while GVA is deflated substantially by WPI. Fourth, none of the five correlation changes in the paper’s Figure 2 reaches statistical significance under the standard Fisher r-to-z test, and the reported India dummy in the cross-country regression falls by 24 percent when the two excluded observations are added back. Fifth, the 2026 paper reverses five specific positions from the 2019 paper without acknowledgment: on double deflation (from core mechanism to “relatively minor problem”), on the scope and weight of the informal sector channel (from narrowly-defined informal manufacturing ruled out for the baseline sample period to the full informal economy as a primary mechanism), on tax indicators (from excluded as “unreliable” to headline indicator), on magnitude (from 2.5 percentage points to 1.5 to 1.9, entirely within the lower half of the 2019 paper’s reported 95 percent confidence interval of 1.5 to 3.5), and on the indicator set itself. Sixth, post-2020 tax performance contradicts the overestimation thesis: nonoil tax growth of 14.9 percent against nominal GDP growth of 10.5 percent yields a buoyancy of 1.4 that the paper’s Appendix C attempts to explain through composition effects that, if true, imply aggregate nominal growth above official figures rather than below. The headline claim has remained stable across seven years and two papers; the indicators, mechanisms, decomposition, and magnitude supporting it have not. The pattern is conclusion-anchoring, not evidence-following.

Keywords: GDP mismeasurement; national accounts; informal sector; price deflators; statistical inference; India

Suggested Citation:

Shekhar, V. and Mukhopadhyay, S. (2026). Unstable evidence, stable conclusion: A critique of the GDP overestimation thesis of Anand, Felman, and Subramanian. SPJIMR Paper Series, WP-2026-06. https://doi.org/10.5281/zenodo.19736522

Download PDF Preview PDF