Cap-and-Trade for Water under Seasonal Forecasts: MDB and Po Valley Evidence and a Theoretical Model

Farmers worldwide face increasing hydrologic variability and water scarcity due to climate change. Two policy tools often considered to enhance adaptation are water markets (to reallocate scarce supplies) and seasonal forecasts (to inform ex-ante decisions). Prior studies typically examine these instruments in isolation, overlooking how expectations about water availability may interact with anticipated trading opportunities in land-use choices. This thesis studies them jointly. Empirically, using HAC-robust inference, we examine the Murray–Darling Basin—one of the most developed water markets—and show that higher pre-sowing allocations over the prior 12 months are systematically associated with larger acreage of water-intensive opportunistic crops (rice, cotton), conditional on pre-sowing field rainfall. Timing and robustness checks are consistent with a causal interpretation, though we do not separately identify price versus quantity channels of the water-scarcity signal. Theoretically, we develop a competitive-equilibrium model that links water availability and its forecasts in a joint probabilistic framework with heterogeneous producers and technology-driven price thresholds. Simulations indicate that (i) under scarcity, cap-and-trade delivers the largest efficiency gains by reallocating water to higher-value uses; (ii) forecast information adds benefits when forecast skill is high and variability is pronounced; and (iii) forecasts and trading can be synergistic, with combined welfare gains exceeding the sum of stand-alone effects, particularly for risk-averse producers. Overall, the results motivate evaluating cap-and-trade and seasonal forecasts together to capture both anticipatory (extensive-margin) and within-season gains. We discuss policy implications for basins with small, risk-averse farms (e.g., the Po Valley) and note key limitations, including the omission of forecast products in the empirical design and stylized assumptions in the model.