Groundwater level dynamics and agricultural land-use change : econometric issues and strategies

Abstract

A significant body of literature has pointed to a causal relationship between agricultural irrigation and groundwater depletion in India. Despite these allusions, I know of no rig- orous estimation of the causal impact of cropland water demand on groundwater level change. This gap in research could be due to data availability/quality issues as well as the methodological challenge of identifying the underlying mechanisms that drive change in groundwater dynamics. In order to reconcile these challenges, I construct a unique dataset integrating satellite data products with administrative data including variables that account for climatic, hydrologic, geologic and socio-economic factors. I study three specific issues related to the identification of groundwater depletion mech- anisms. First, I detect systematic or non-random missingness in administrative ground- water data due to the occurrence of “dry wells”. Dry wells signify extensive depletion such that groundwater falls below the maximum depth of monitoring wells. Naive omission of dry wells can lead to severe false optimism about regional groundwater situations. I employ a set of ‘observable’ covariates of groundwater to predict the in- cidence of dry wells in an unlabelled dataset. I then utilize the prediction probabilities to quantify the statistical bias due to non-random missingness in conditional ground- water estimation models. Second, I consider the obstacles in statistical inference that arise from the fact that groundwater aquifers represent a non-exclusive common pool resource whereby the costs and benefits of resource use are shared by spatially proxi- mate users. I employ a statistical tool known as the semivariogram to estimate spatial autocorrelation in groundwater levels. Such estimation provides empirical evidence for delineating the spatial boundaries for resource sharing within a groundwater aquifer. I then assess the impact of the spatial aquifer structure for economic policy and ground- water management science. Finally, I develop a framework for assessing the causal impact of agricultural land-use intensification on groundwater depletion founded on a structural model that is derived from a groundwater balance equation. The identification strategy relies on a 2-stage least squares approach instrumented with spatially varying, crop-specific minimum support price (MSP) which lead to differential incentives for al- locating farm acres across multiple crops and hence groundwater extraction outcomes. This work advances the study of the causal relationship between groundwater irrigation and depletion by addressing three oft-ignored econometric issues that arise in such a study. Overall, my essays bear relevance for groundwater management and policy mak-ing as well as academic research where accurate and efficient estimation of statistical moments of groundwater levels is of paramount importance.