You can find proof of anything you want, of course. Some people still believe the world is flat, so anything goes.
When you see a study, you need to know what age groups it covers, whether it applies to all passengers or just properly restrained passengers, does it only consider crashes with a fatality, all crashes or those with some threshold on injury measures, etc.
For example, the Freakonomics people will tell you that in their research, child restraints are no more effective than putting a child into a seatbelt and giving them a DVD player to keep them from squirming around. The problem is, to my knowledge, they still haven't had any of their tests or studies published in a respected, peer-reviewed medical or technical journal. Instead, they went for the hype in a money-making book and website.
It's one thing if you convince a parent to take a relatively calm child who is incorrectly restrained in an expired or obsolete overhead shield seat and put them into a seatbelt with a DVD player. Maybe it's true that their risk of a fatal injury hasn't changed much. It's another thing if you do that to a squirmy child who is properly restrained in a 5-point harness. I wish the authors would be at the emergency room to see what happens when that child puts the shoulder belt behind their back and comes in with septic abdominal trauma and a fractured spine because of lap belt related injuries.
I even referred one of the authors to an excellent trauma surgeon who is an expert on seatbelt syndrome injuries, conveniently at their own university. I wonder if they ever had the intestinal fortitude to see one of this physicians' presentations on the topic:-(
Anyway, some good links for studies and statistics are:
http://stokes.chop.edu/programs/injury/our_research/pcps.php
http://www.chop.edu/consumer/jsp/division/generic.jsp?id=77971
http://www.usa.safekids.org/tier3_cd.cfm?content_item_id=1133&folder_id=540