To the extent that school shooting are, statistically, a "tail" event, you would expect that to change more sensitively than the "average" (guns per capita). So a tripling of the average gun measure (going from Serbia to US), might correspond to a much larger change in these extreme events. I doubt it's a simple statistical consequence, but it also doesn't follow that amount of guns would be (linearly) proportional to extreme events.