We applied the self-consistent polarization density functional theory (SCP-DFT) to water. SCP-DFT requires only minimal parametrization, self-consistently includes the dispersion interaction neglected by standard DFT functionals, and has a cost similar to standard DFT despite its improved performance. Compared to the DFT functionals BLYP and BLYP-D (where the latter contains a simple dispersion correction), SCP-DFT yields interaction energies per molecule and harmonic frequencies of clusters in better agreement with experiment, with errors in the former of only a few tenths of a kcal/mol. BLYP and BLYP-D underbind and overbind the clusters, respectively, by up to about 1 kcal/mol. For liquid water, both BLYP and SCP-DFT predict radial distribution functions that are similar and overstructured compared to experiment. However, SCP-DFT improves over BLYP in predicting the experimental enthalpy of vaporization. A decomposition of the dimer interaction energy attempts to rationalize the performance of SCP-DFT. The SCP-DFT approach holds promise as an efficient and accurate method for describing large hydrogen-bonded systems, and has the potential to model complex systems with minimal parametrization.