Long-run study of residential water consumption

The estimation of dynamic models and the measure of long-run effects are rare in residential water demand studies. We show in this paper that a dynamic model of water consumption can be derived from a structural optimisation program solved by local communities. This nonlinear model is estimated on a sample of French municipalities and is found asymptotically equivalent to a dynamic panel data model that is linear in the parameters. The latter includes an original error-component structure that allows for a flexible heterogeneity pattern, including both the usual idiosyncratic effect, and an additional individual effect affected by a multiplicative time-varying parameter. As usual GMM estimators for panel data are not consistent in this case, we propose a new GMM procedure that yields consistent and efficient estimates of short- and long-run price elasticities (respectively -0.26 and -0.40).