<p>  </p> <p>Irrigation water is crucial for agricultural production and farmers’ livelihoods in many regions of the world. Water scarcity, however, due in part to extended droughts and groundwater depletion, is putting ever-increasing pressure on regulatory agencies to implement water conservation policies, possibly hampering farmers’ livelihoods and food supply. It is therefore imperative to implement policies that allocate scarce water resources in the most efficient way, while allowing water buyers and sellers to benefit from trading. In this dissertation, I study the design of water policies that are in use for agricultural water management in many dry regions: markets, quotas, and water pricing. A key distinction among them is that water markets constitute a decentralized allocation mechanism, while the other policies are centralized. </p> <p>In my first two essays, I compare two prominent market designs, namely once-a-year trading (i.e., non-recursive design) and year-round trading (i.e., recursive design). I examine the effect of these competing market designs on efficiency and distributional outcomes. I do so by accounting for two unique characteristics of irrigation water markets: 1) farmers differ in the marginal valuation of water, and 2) water is thinly traded. In my first essay, I theoretically develop a 3-seller and 3-buyer trading model and show that a recursive design can hinder efficiency when traders differ substantially in their water valuations. This is because sellers greatly increase markups, which strengthens trading frictions. </p> <p>Theoretical predictions from essay 1 rest on rather strong behavioral assumptions – sellers and buyers are able to identify and play a rather complicated subgame perfect Nash equilibrium. But human subjects may deviate from these due to many reasons, including bounded rationality and off-equilibrium beliefs, among others. In my second essay, I conduct a laboratory experiment with human subjects to examine the extent to which they follow theoretical predictions, and the effects of their behavior on the relative performance of competing market designs. Experimental evidence shows that subjects do deviate from predicted equilibrium, particularly when agents are highly heterogeneous. As a result, I find that some of my theoretical predictions are reversed in the laboratory. In particular, the recursive design outperforms the non-recursive design because experimental agents behave much more competitively under the recursive design than predicted by my subgame perfect Nash equilibrium.</p> <p>In my third essay, I empirically compare the performance of quantity and price policies, as well as a hybrid to conserve water, using observational data from Mexican farmers. A fixed cap (the policy chosen by Mexican regulators) effectively limits water consumption, but the rigidity of the cap results in substantial welfare losses due to volatile weather patterns that induce drastic fluctuations in water demand. Consequently, I study two counterfactual policies, a price policy and a hybrid that combines quantity and price policies. I find that the hybrid policy performs very well relative to its pure counterparts because it provides an intermediate level of flexibility in water consumption, preventing large losses under anomalous weather. </p>