Estimating Water Demand Using Price Differences of Wastewater Services

Many homes in Hawaii use cesspools and other on-site disposal systems (OSDS) instead of the municipal sewer system. Because bills combine water and waste-water services, and homes with OSDS do not pay for sewer service, OSDS residences have lower monthly bills compared to those with sewer-connected systems. We use this price difference in conjunction with selection on observables and matching methods to estimate the price elasticity of residential water demand. Matching methods indicate that OSDS residences have systematically different characteristics than those with sewer-connected systems, suggesting an imperfect natural experiment. We show traditional methods lead to biased elasticity estimates, even though they are robust when selecting on observables using OLS with or without census tract fixed effects, census block fixed effects, and non- parametric controls trained using cross-validation and a lasso. We then estimate demand using a limited sample of OSDS homes that have sewer-connected neighbors, which gives estimates from -0.06 to -0.08. The neighbors have no systematic differences in other characteristics and estimates are robust to further selection on observables, but the sample differs slightly from population means in their physical characteristics. These more defensible demand elasticity estimates, however, are much more inelastic than estimates not based on comparison of neighbors and are generally more inelastic than previous studies. Taken collectively, the results highlight the susceptibility of demand estimates to omitted variable bias. Highly inelastic water demand suggests that considerably higher prices may be needed for sustainable water management, creating some practical challenges under current regulatory guidance. We also use our results to estimate willingness to accept a tax credit for upgrading an OSDS system, a targeted policy that aims to improve water quality. Regardless of whether consumers respond to average or marginal prices, our estimates imply that the tax credit is far too small to induce voluntary participation in the program. Additional consumer welfare topics are also considered.