The City Paper article suggests the absolute number of crashes should track people per square mile, and is surprised that we have the most crashes but not the highest population density. That's silly. By that logic, a one-block county with everyone packed into massive apartment buildings should have huge numbers of crashes, while a vast county with nothing in it but a busy interstate and one house should have virtually none.
In reality, we should expect crashes to rise with both the size of a county and its population. Add people to a county and naturally you'd expect more crashes. But imagine you hold the population constant and spread them out over a bigger county. That typically means more roads and more driving. (Of course, this ignores lots of stuff. Counting the number of miles of road in a county might be a better predictor of absolute crash numbers than its area, but the article focuses on square miles, not road mileage, so let's stick with that.)
It's perfectly plausible that Allegheny would have the highest absolute number of crashes, regardless of how our road design compares to other counties. After all, we have the second-highest population in the state, but the one county with more people (Philadelphia county has 25% more) is much smaller (it's 19% of our size).
I found 2014 crash stats by county here. Here are the top counties for crashes, in order, from page 59:
Now, let's approximate the expected number of crashes. If you take each county's population and multiply it by its area^0.3, to account for population being a more significant factor than area in crashes, you get this top ten list:
|County||Pop||Area||Expected relative crashes|
Notice that the top few entries are identical. I just guessed at the 0.3 figure, but already it's a pretty close approximation of the actual relative number of crashes.