Introduction

Humans are rapidly depleting many of Earth’s natural resources, a problem that resonated with economists a century ago and persists today. This research focuses on groundwater, which supplies 40% of the water used for global agriculture. In response to declining water tables in numerous aquifers, governments in regions such as California, Arizona, Texas, Australia, India, and Iran are debating or implementing regulations. The study systematically tests whether prevailing extraction is socially optimal, as many observers believe most aquifers are common pools where users act myopically, leading to extraction levels that exceed the social optimum.

Background on water resources

Groundwater is a vital component of total terrestrial water storage, accounting for 20% of global freshwater. Agricultural irrigation accounts for 70% of global groundwater extraction. Most of the Earth’s land surface stores groundwater, with hydrological agencies identifying 3,424 distinct global aquifers. This research highlights four key case study aquifer systems: the Central Valley in California, the Ogallala in the U.S. Great Plains, the North China Aquifer System, and the Indus Basin in South Asia. These systems are selected for their scale, agricultural importance, and significance in public policy debates.

Model

The study utilises a dynamic model of renewable resource extraction where a single agent manages an aquifer to maximise the present discounted value of profits. The agent faces a trade-off: consuming water today increases current profits but lowers the water table, which in turn raises future extraction costs. This model is used to recover a “decision” discount factor that rationalises observed extraction and compares it against benchmarks, specifically a Ramsey discount factor and a market-based benchmark of 0.98. This comparison assesses whether institutional or market failures prevent users from internalising the long-term costs of depletion.

Data

Measurement of the model’s parameters involves three main groups of objects: the physical state and flow of groundwater, the marginal return to groundwater in production, and the marginal cost of extraction. Remote sensing data from the Gravity Recovery and Climate Experiment (GRACE) satellites is used to measure annual changes in lift height. Productivity and output are proxied using gridded nighttime luminosity data and gridded GDP data. The cost function is calibrated using data on electricity prices, pump efficiency, and lift heights from global hydrological models such as WaterGAP and data from Fan et al. (2013).

Parameter estimation and model calibration

Production function parameters are estimated through instrumental variables regressions to address potential biases from measurement error and omitted variables. The preferred specification implies a groundwater elasticity of 0.10 and a resource share of 0.51. The marginal returns to groundwater are calculated as a weighted sum of crop-specific returns, varying significantly by aquifer. For instance, high-value crops in California’s Central Valley yield high marginal returns, while areas like the Sahara have very low returns. Calibration shows that median and mean net extraction as a share of capacity are near zero, suggesting many aquifers are in a steady state, though not necessarily an optimal one.

Results: discount factors

The research finds that 61% of global aquifers have decision discount factors below the Ramsey benchmark and 82% below the market benchmark, implying that extraction in these aquifers exceeds optimal rates. The median decision discount factor is 0.924, while the mean is 0.860. While excessive extraction is common, many aquifers have discount factors within a few percentage points of benchmarks, suggesting that welfare costs might not be extreme in all cases. In the Central Valley, the decision discount factor is 0.89, where users act as if 41% of extraction costs depend on the choices of others.

Results: resource management and welfare

Global aquifers are estimated to have a total present value of at least $103 trillion, a figure comparable to global GDP or the value of all publicly traded firms. The welfare loss from non-optimal resource management is estimated at 2.6% of aquifer present value, equating to between $2.6 trillion and $8.4 trillion in aggregate losses. These losses are driven by significantly higher extraction costs; the present value of additional costs incurred on current extraction paths exceeds optimal paths by trillions of dollars. Three counterfactual scenarios—myopic extraction, constant extraction, and sustainability—are analysed, with the first two generating substantially higher welfare costs (3.4% to 12.9% of value) and leading to the complete drainage of 10% to 31% of aquifers respectively.

Conclusions

The study concludes that while many aquifers are in a nominal steady state, they often have inefficiently low water stocks. An optimal management plan would involve cutting short-run extraction to refill aquifers, eventually reaching a steady state with higher water levels and lower extraction costs. This transition would reduce short-term profits but provide a long-term benefit that outweighs the cost. Improving management in aquifers with decision discount factors near the benchmark could recover a significant portion of current welfare losses. Future work is suggested in areas such as spatial analysis and the integration of stochastic recharge data.