The Wall Street Journal did a pretty superficial article on the effectiveness of red light cameras. The most important thing in the article was Kimberly Eccles statement,

She added that red-light cameras generally are effective when deployed correctly. However, "because of the controversial nature of red-light cameras, I do believe an agency should consider everything else" before installing them, she said.

However the article was marred by bistro statistical nonsense like

Simple before-and-after comparisons also won't do, researchers say. For one thing, intersections typically are chosen for camera installation because they have had a spate of accidents. That makes them due for a fall just by statistical chance.

One of the problems towns have faced is doling out the cameras in a political way. The tech needs to be deployed by the traffic pros.
One thing that happened at a light in Washington DC is that, while citations went down, virtually all the people later cited were from out of town. Safety was achieved as locals adjusted but out-of-towners (DC is a huge tourist town) kept screwing up. This indicates to me that there was a problem with the intersection. Monitoring the intersection, however, could lead designers to find a better solution.
Of course we can't do this in Cincinnati since voters, after a fear campaign spearheaded by COA T & NA CP, threw the baby out with the bathwater. Now every red light related crippling injury & death in Cincinnati can be laid at their blood drenched feet.

Read the whole article here

The truth about Cincinnati's plan that counters the FUD perpetrated about it here