This dissertation examines lessons in demand management through consumer behavior in the water sector. The first chapter identifies conservation behavior among households who were not targeted by a local drought policy and explains the mechanisms by which the policy appears to act on these households. Using billing records from two neighboring water districts, I implement a difference-in-differences approach among accounts near the arbitrary border between these districts and identify not only aggregate average treatment effects, but also effects across the distribution. I examine the marginal response to a fine-based policy instituted in only one of the two districts, controlling for contemporaneous common shocks across the two districts. While the policy was targeted only at the high end of the distribution, I establish a response even among households for which the fines were unlikely to bind. Using a simple structural estimation, I rule out a variety of price-based arguments and argue that the most consistent theory for the observed behavior is that the policy announcement temporarily shifted demand, reducing the marginal utility of consumption at all levels. Using both a randomized experiment and an arbitrary threshold in a concurrent natural policy experiment, the second chapter examine the effectiveness of a social comparison intervention across two states of the world with higher and lower pecuniary incentives. Contrary to some of the existing literature, we do not observe evidence of pecuniary incentives crowding out social pressure in this context, despite the fact that both policies were effective. The third chapter focuses on the external validity of drought policies and the population characteristics which might influence their effectiveness. I investigate whether renters are differentially sensitive to price-based and normative instruments. In answering this question, I find that responsiveness is primarily driven by the account characteristics that determine how likely a policy's effective mechanism is to bind. In other words, the main source of heterogeneity in each policy's effectiveness appears to be driven by exposure to the policies themselves. Because this exposure is correlated with my proxy for renting, there are differential responses among homeowners and renters, but these are not due to landlord-tenant relationship.