A while back I read a study that agreed with your assertion. It examined intersections that had traffic lights removed. They found that in the short term, accident rates went down but after some time passed the accident rates shot back up. The authors theorized that the temporary reduction was due to the novelty of the change.
Edit: typo