One might argue the intersection itself is the problem and should be redesigned, as well as adjoining roadways feeding into the offending intersection.
If it's consistently high something needs fixing. But accidents are random, there will always be some intersections that by pure chance have more accidents. Put cameras on those, presto, cameras "work".
Let's make a hypothetical city with 10 intersections, all absolutely identical. Watch until you have seen 20 accidents. Do you really think there will be exactly 2 accidents per intersection??
Select the intersection that gets the most accidents, declare it dangerous.
Case A: stick a camera on it. Case B: do nothing. Watch another 20 accidents.
In both cases you expect to see fewer accidents at the intersection as the original number was just chance. But if you're trying to prove cameras work...
