[shkkmo]

To me, this sounds like the rate of unreported crimes would be lower in dense areas, making them even safer than they appear in the statistics.

You are correct that if the average number of observers per crime is higher, that could indeed make a place feel comparatively more dangerous since the while the liklihood of being a victim is lower, the liklihood of seeing a victim is higher. However, I would posit that a lot more goes into feelings of safety and we haven't seen any data on how large this effect size might be.