If you are familiar with the various crystal forms of carbamazepine in the CSD, you will know that one particular form of carbamazepine (CSD refcode: CBMZPN03) is in the trigonal space group R-3 and exhibits noticeable channels running down the unique axis. These channels result in void space of roughly 6.6% of the unit cell volume (or 375 Å per unit cell) and appear to either contain trace amounts of solvent or at least are templated by solvent.
I suspect that the space group R-3 is predisposed to forming these kinds of channel crystal structures containing void space. I think this is true in particular for CSD structures where few systems will have the inherent molecular or polymeric symmetry to reside on the actual symmetry element and instead will arrange themselves around the -3 axis.
To test this theory, I've written the attached script which loops over the entries in the CSD in space group R-3 and then determines an average void percentage. The outcome is 5.81%, but how does that match up to the average void percentage across the whole CSD?