How often do you review?
In the latest early releases from the American Journaol of Epidemiology, the editors discuss the concern of authors who submit heavily but do not accept requests to review articles. This situation is a classic case of the "Tragedy of the Commons"; the credit for reviewing is small (and typically anoymous). If nobody reviewed then the peer-review system would break down. However, the cost of any one person removing themselves from the reviewer pool is small compared to the benefit (for that person) in being able to focus on their own research projects.
Even worse, it is hard to know if the person who refuses at journal X might be a common reviewer at journal Y. Is the person who refuses refusing because they already do a lot of reviewing for another journal? Heck, I have reviewed for the American Journal of Epidemiology and I've never had a paper accepted there!
Female Science Professor once assembled a list of why people might see reviewing as a rewarding experience. It is an interesting list but maybe not completely convincing that the benefits outweigh the costs.
On the other hand, it is somewhat insirational that this system works despite everything that is working against it. Maybe science still has a critical mass of idealists left?
It's rather a nice thought.