the institute of No Shiat Sherlock.
It isn't really that obvious. There was an overall 5% increase in injury accidents at the intersections with cameras. But they did not mention the severity of the injuries. T-bone crashes (which were reduced) are likely to result in more severe injuries than rear-end collisions (which were increased). There were other complications: Most of the additional accidents occurred at intersections that were poorly chosen because they previously had few accidents. So it is possible that cameras improved safety at intersections with a history of accidents, and could improve safety overall if they are only installed at those intersections. Another issue is the yellow light duration. Longer yellows leads to fewer accidents, and some cities installing cameras also shorten the yellow light duration to increase revenue. It isn't clear if yellow light duration was decreased in the intersections studied.
The study shows that cameras can increase accidents, but it doesn't show they always increase accidents. If they are used more intelligently, they could be a net benefit.