Where India's Productivity Growth Came From, 1991–2025: An Exact Three-Term Shift-Share Decomposition, the Agricultural Productivity Gap, and Four Asian Comparators

Output per worker in India grew by 3.77% a year between 1991 and 2025. I split that growth exactly into productivity gains inside agriculture, industry and services (the within term), the movement of workers towards sectors with high initial productivity (the static term) and the movement of workers towards sectors whose productivity was rising (the dynamic term). The data are World Bank WDI sectoral value added at constant prices and ILO-modelled employment shares, vintage 8 October 2026. Over 1991–2025 the within term explains 71.6% of the gain, the static term 8.7% and the dynamic term 19.7%. Structural change contributed about a quarter of growth in each sub-period from 1991 to 2019, then turned negative in 2019–2025 (−9.8%), when agriculture's share of employment rose from 40.7% in 2019 to 44.7% in 2020 and was still 41.6% in 2025. Non-agricultural output per worker was 3.32 times agricultural output per worker in 2025, up from 2.59 in 1991. I prove that the decomposition is exact, that the static term equals the sum of share changes times the gap between each sector's initial productivity and the aggregate, and that moving 5% of the workforce out of farming would raise output per worker by 3.0–4.9% to first order, depending on the labour elasticity. Had the 1991 employment structure been frozen, output per worker in 2025 would have been 79.6% of its actual level. Among China, Indonesia, Vietnam and Bangladesh, India is the only economy where the gap widened and the farm employment share fell by less than 22 points.