Using the quasiparticle random phase approximation, we calculate the nuclear matrix elements governing two-neutrino and neutrinoless double-beta decay. We show that a consistent treatment, including both particle-hole and particle-particle interactions, helps to resolve the longstanding discrepancy between experimental and calculated two-neutrino decay rates. The particle-particle force, which allows us to bring calculated EC/β+ decay rates in semimagic nuclei into closer agreement with experiment, is in large part responsible for suppressing calculated two-neutrino decay rates that are otherwise too fast. We test the validity of our procedure by comparing quasiparticle random phase approximation results with exact solutions for a solvable model, in which the suppression of two-neutrino decay by the particle-particle interaction is confirmed. We then extend our approach to the neutrinoless decay associated with a finite Majorana neutrino mass and, conceivably, with majoron emission, and demonstrate that the nuclear matrix elements governing these processes are also suppressed. We present predicted half-lives for both two-neutrino and neutrinoless double-beta decay in several candidate nuclei, and discuss the difficulties associated with the calculation of such highly suppressed quantities.