Radley Balko says that the true purpose of red light cameras is to raise money, not stop accidents:
[O]verall collisions are up at the intersections where Clarksville has installed red light cameras (a result we've seen nearly everywhere they've been installed). The city just chooses to ignore rear-impact collisions when evaluating the cameras. Those collisions increased from 138 in 2008, to 173 in 2009, to 169 through October of this year. It's true that side-impact collisions are generally more dangerous than rear-impact collisions. But even taking that into consideration, it's a bit of a stretch to say that a decrease of eight side-impact collisions coupled with an increase of 25 rear-impact collisions shows that the cameras are preventing accidents.
The best approach doesn't bring in any revenue, for camera makers or city governments: Lengthening the duration yellow lights has proven to be much more effective at preventing accidents than cameras. Which of course is why several cities have been caught making intersections more dangerous by shortening yellow lights in order to generate more tickets.