Accidents Happen: Where to improve safety

Drunk Driving Lindsay Lohan in jail

In your city, the intersection with the most accidents is probably,
per car, one of the safest. And that is a good thing, as it is
probably a useful intersection on the way to somewhere you want to be.
But if you are on city council and you are presented with engineering
recommendations to improve the intersection with the most accidents,
would you ignore the recommendations because, per car, the
intersection is already one of the safest?

Depending on the recommended work, it might be a cost effective way of
reducing the number of car accidents in your city. But is that fair?
Should some drivers encounter significantly less risk in order to
maximise the reduction of accidents? Or should the risk be spread out
as equally as possible, even if that way more cars are damaged and
more people injured.

You can apply this dilemma to food safety, healthcare, crime,
unemployment, economics, etc.

This keeps me up at night.

Update (2012-05-30): Here is an interesting bicycle accident report of 4 years of consolidated data. I say interesting because it is not. Accidents are remarkably spread out. They also go up in June. But if you read that data and conclude January is the safest month to drive your bike in Montreal, then you are a fool.