The strongest deterrent for the general populace.
Generally speaking, crime rates tend to be pretty low already. So the sample shifts from general populace to those who already commit crimes, or in such an emotional fervor that they gain the capacity for crime.
Among that population, I don't think surveillance cameras are stopping much.
Also, how can we know how much crime isn't happening due to cameras? If it's like "we installed a camera at location X and crimes there dropped 72%", that's not taking into account that the criminal just found an easier target, leaving the same amount of net crime.
I.e. they are much more likely to rob a grocery store if they think they won’t get caught, but the penalty for robbing a grocery store being 1 year or 10 years doesn’t have a strong effect on deterrence.
To your second point - I don’t think it is helpful to find hypothetical holes in their methodology, without reading the individual studies.
The control group. Aka, the current crime rates right now with current infrastructure. Not a blank slate
In a lawless anarchy, you're probably right that "will I be held accountable for my actions?" Is the nost important question to ask. But we don't live in that society. The question we're asking instead is
1) how much does surveillance augment law enforcement?
2) how much does surveillance deter would be criminals compared to current deterrents that is law enforcement patrolling and reporting?