We investigate statistics of dynamical exchange events in coarse-grained models of supercooled liquids in spatial dimensions d=1, 2, and 3. The models, based upon the concept of dynamical facilitation, capture generic features of statistics of exchange times and persistence times. Here, distributions for both times are related and calculated for cases of strong and fragile glass formers over a range of temperatures. Exchange-time distributions are shown to be particularly sensitive to the model parameters and dimensions, and exhibit more structured and richer behavior than persistence-time distributions. Mean exchange times are shown to be Arrhenius, regardless of models and spatial dimensions. Specifically, < t(x)>similar to c(-2), with c being the excitation concentration. Different dynamical exchange processes are identified and characterized from the underlying trajectories. We discuss experimental possibilities to test some of our theoretical findings. (c) 2005 American Institute of Physics.