Joint outage identification and state estimation is studied in power systems in which cascading outages dynamically develop and network states dynamically evolve. A recursive algorithm is developed that computes in closed form the joint posterior of cascades and network states at every time step. A beam search technique is employed that prevents the number of cascades to compute from growing exponentially. Because the joint posterior is a sufficient statistic for jointly identifying the cascades and estimating the states, the derived closed forms can be applied to develop the optimal dynamic joint detector and estimator under any performance criterion. We simulate cascading line outages with uncertain network states in the IEEE 14-bus and 57-bus systems, and the proposed algorithm is evaluated for dynamically identifying outages and estimating states at every time step. It is observed that retaining just a few cascades in the beam search can achieve a joint identification and estimation performance close to that with all cascades retained.