The role of farm subsidies in changing India's water footprint

The study investigates how India’s output subsidy policy—implemented through government procurement of rice and wheat at guaranteed floor prices (MSP)—has contributed to groundwater depletion. India’s groundwater consumption has increased by about 500% over the past half-century, with average water tables dropping more than 8 meters since the 1980s and up to 30 meters in some regions, especially in the alluvial aquifers of northwest India. Output subsidies, introduced during the Green Revolution to boost yields and ensure food security, have been implemented primarily for rice and wheat, incentivizing their cultivation over other crops. Because these cereals are water-intensive and irrigation needs have increasingly been met by private wells rather than surface systems, the policy may be accelerating groundwater depletion. The research quantifies the link between output subsidies, crop production patterns, and groundwater stress, using national cross-district associations and two state case studies (Punjab and Madhya Pradesh) that span distinct aquifer types and policy timelines.

Prior research documents India’s severe groundwater depletion and its implications for agriculture and climate resilience, including satellite-based assessments (e.g., GRACE) and studies on aquifer dynamics in alluvial versus hard-rock settings. Policy analyses have debated India’s MSP and public procurement regime, noting its historical role in stabilizing prices and ensuring cereal availability but highlighting inefficiencies and environmental externalities. Related work on electricity and irrigation subsidies shows how input price distortions increase groundwater extraction and can exacerbate inequities. Studies also link groundwater scarcity to declines in yields, cropping intensity, and adaptation costs (e.g., well deepening). Nutrition literature indicates that the Green Revolution and coupled consumption subsidies shifted production and diets toward staples, crowding out more nutritious and often less water-intensive crops (coarse cereals, pulses). Global work on virtual water trade shows India exports substantial groundwater embedded in food, contributing materially to global depletion. This paper adds causal evidence on the role of output subsidies in groundwater stress across contrasting aquifer systems and policy contexts.

Data: The authors compile district-level data on crop production/area/irrigation (ICRISAT, 1966–2015), government procurement of rice and wheat (Food Corporation of India and state agencies; Punjab 1981–2018; Madhya Pradesh 2002–2016), groundwater levels and well status (CGWB national monitoring wells, 1996–2016; Punjab state groundwater board, 1973–2003), and tubewell construction (Minor Irrigation Census rounds 2–5). Weather variables (precipitation, temperature) come from IMD gridded datasets aggregated to district growing seasons. Districts are harmonized to 1966 boundaries. Measures of groundwater stress: (a) Groundwater depth (pre- and post-monsoon) at district level; (b) Percentage of defunct wells (active at the start of the sample but ceased recording permanently, with at least four years of no data at the end); (c) Percentage of dry wells (wells missing data in given seasons/years); (d) Number of deep tubewells (>70 m). Different metrics are emphasized by region due to aquifer differences: alluvial Punjab supports long-horizon depth analysis; hard-rock Madhya Pradesh requires short-horizon depth changes, dry wells, and deep tubewells. All-India cross-section: Regress the percentage of defunct wells (1996–2015) on districts’ average annual rice area growth, controlling for growth in other crops, initial rice area, initial gross cropped area, land area, precipitation, and population. Also analyze association with tubewell construction. Wheat is excluded in this national analysis due to limited post-1996 variation and concentration in few districts. Punjab case study (1981–2015): Use panel regressions with district fixed effects, year fixed effects, and district-specific time trends to estimate: (1) Output response: log crop area/production on lagged log procurement (controls for precipitation and temperature). (2) Groundwater response: proportional change in groundwater depth between year t and horizon T on log rice area or log rice procurement in year t, with controls (net cropped area, population, cumulative precipitation to T, temperature). To address slow aquifer adjustment, estimate separate models for multiple horizons (1–7+ years). Inference uses Newey–West standard errors with 10-year lags due to short N (11 districts). Placebo tests regress past groundwater changes on current rice area/procurement. Madhya Pradesh case study (2002–2016): Exploit the sharp expansion of wheat procurement starting in 2008 (state bonus plus strengthened logistics) as a policy shock. Estimate district-year panel models: GWS_dt = β0 + β1 log(wheat proc_dt) + β2 log(wheat proc_dt) × 1{t ≥ 2008} + controls (precipitation, temperature, temperature squared), with district fixed effects and district-specific linear trends. Outcomes: (a) proportional change in groundwater depth between November (t−1) and May (t) (pre-sowing to post-harvest), (b) fraction of dry wells post-harvest (June–September), (c) log number of deep tubewells. Robust standard errors are clustered by district. Mechanism models regress wheat area, irrigated wheat area, and pulses area on lagged procurement and its post-2008 interaction, with year fixed effects and controls (rainfall, temperature, irrigation, other crops). District fixed effects are omitted in mechanism models due to short post-policy panel (addressing Nickell bias concerns). Assumptions and controls: Extensive controls for climate, population, and cropping patterns; fixed effects absorb time-invariant heterogeneity (e.g., hydrogeology), and year effects capture aggregate shocks. Aquifer adjustment dynamics are explicitly modeled in Punjab. Data constraints: District-level procurement data are unavailable for most states; hence the all-India analysis is correlational and focused on rice. Punjab wheat area has little variation post-1981, limiting wheat-specific estimates there. CGWB wells are mostly shallow, implying estimated groundwater effects are conservative relative to deeper aquifer depletion seen in GRACE-based studies.

• Overproduction: India’s rice and wheat production exceeded domestic consumption and buffer needs by about 30% in 2020; government stocks were ~2.5x required norms (excess ~36 million tons, ~20% of annual consumption), indicating policy-induced overproduction of water-intensive cereals.
• All-India associations: A 1 standard deviation increase in district rice area growth (1996–2015; SD ≈ 4.22%/yr) is associated with at least a 5.44 percentage point (pp) increase in defunct wells, controlling for confounders. Considering rice area growth since 1966 (SD ≈ 2.58%/yr), effect sizes roughly double. A 1 SD increase in rice area growth is also associated with a ~50% increase in tubewell construction (Supplementary Table S1).
• Punjab causal dynamics: A doubling (100% increase) in rice area is associated with groundwater depth declines that accumulate over time: ~8 pp after 1 year and ~28 pp after 3 years (post-monsoon), stabilizing around ~72 pp (post-monsoon) and ~40 pp (pre-monsoon) by ~6–7 years. Similarly, a doubling of rice procurement leads to a ~40 pp fall in groundwater depth by year 7. Given rice procurement rose ~3%/yr from 4.4 to 13 million tons (1981–2015), model-based estimates imply 1.2–2 pp/year declines attributable to procurement-induced cultivation, totaling roughly 50–65% of the observed groundwater depth fall over 34 years. Placebo tests (lag-lead) show no spurious pre-trends.
• Madhya Pradesh causal effects (post-2008 procurement shock): Table 2 estimates indicate that, relative to pre-2008, a doubling of wheat procurement causes: +3.9 pp increase in groundwater depth change (AGWL) between Nov(t−1) and May(t) (col 1; p≈0.097), +7.6 pp increase in dry wells post-harvest (col 2; p<0.01), and +4.8% increase in deep tubewell construction (col 3; p<0.01). From 2007 to 2016, procurement rose by ~70% (0.057 mt to 4 mt), implying an overall +5.3 pp rise in dry wells and +3.4% increase in deep tubewells.
• Mechanism in MP: Post-2008, the marginal effect of procurement on next-year wheat area increased by ~13.5%; a doubling of procurement increased irrigated wheat area by ~22%, with some area shifting from pulses to wheat, intensifying groundwater demand.
• Cross-region insight: Effects manifest differently due to aquifer properties—gradual, cumulative declines in deep alluvial Punjab versus short-horizon stress signals (dry wells, deep tubewells) in hard-rock Madhya Pradesh—underscoring the need for region-specific metrics.
• Overall implication: Output subsidies focused on rice and wheat likely induced roughly 30% overproduction of water-intensive crops and are a significant driver of groundwater stress across diverse hydrogeologies.

Findings indicate that India’s output subsidies, implemented primarily through MSP-backed procurement of rice and wheat, have materially distorted crop choices toward highly water-intensive cereals, increasing irrigation demand and depleting groundwater. In alluvial Punjab, the long adjustment of aquifers masks contemporaneous effects; when modeled over appropriate horizons, procurement-induced rice expansion accounts for at least half of the multi-decade groundwater depth decline. In hard-rock Madhya Pradesh, procurement causally increases dry wells and deep tubewell installations within a few years, revealing rising stress even when long-run depth trends are muted. These results reconcile regional differences by aligning metrics with aquifer behavior and caution against naive contemporaneous analyses that would understate policy impacts. The environmental externality conflicts with policy goals of food security and farmer welfare: groundwater scarcity reduces yields and buffers against climate variability, threatening long-term production stability. The coupling of output subsidies with consumption subsidies entrenches cereal overproduction and undermines dietary diversity, with potential nutrition deficits. Internationally, the policy mix contributes to virtual water exports despite domestic scarcity. The study underscores the broader global lesson that distortionary agricultural supports can impose significant natural resource costs, necessitating careful policy design that aligns income stabilization with environmental sustainability.

The paper demonstrates that India’s output subsidy regime has been a significant driver of groundwater stress by inducing overproduction of water-intensive rice and wheat and shifting irrigation toward groundwater. Causally, procurement explains a large share of Punjab’s long-run groundwater decline and has already increased well failure and deep well demand in Madhya Pradesh within a short period. Policymakers should reconsider subsidy design: shift from output-linked procurement toward less distortionary instruments (e.g., income transfers such as PM-KISAN, price deficiency payments), rationalize procurement to reflect nutrition and water objectives, and invest in agricultural R&D and water-saving technologies. Transitioning will require building farmer trust and institutional capacity. Future work should integrate deeper well observations across regions, refine procurement criteria to minimize water footprints while safeguarding nutrition and food security, and evaluate policy reforms’ impacts on SDG trade-offs under climate variability.

• Data availability: District-level procurement data are available only for select states (Punjab, Madhya Pradesh), limiting nationwide causal identification; the all-India analysis is correlational and uses rice only due to limited variation in wheat post-1996.
• Measurement constraints: CGWB monitoring wells are predominantly shallow; deeper pumping common in rice-growing regions may not be fully captured, making estimated effects conservative relative to GRACE-based storage declines. Monitoring well levels are indicators and may not match farmers’ irrigation wells exactly.
• Model identification: Although models include extensive controls and fixed effects, unobserved confounding cannot be entirely ruled out. Punjab regressions rely on Newey–West SE due to small N; MP mechanism models omit district fixed effects to mitigate Nickell bias given short panels, implying potential bias though estimates are consistent with the proposed mechanism.
• Generalizability: Effects vary with aquifer type and local implementation intensity; translating magnitudes to other states requires careful consideration of hydrogeology and policy execution.
• Time horizons: Punjab results hinge on long adjustment periods; short-window analyses could understate true impacts. MP results reflect a relatively recent policy window (post-2008); longer-term impacts may be larger.