On the dual of Lyapunov's second theorem