Perhaps the most surprising fallout from our conversation is that my opinion of economics declines with nearly every exchange. In my world the sort of graph you cite as evidence that regulations to improve auto safety produced a marginal effect would be roundly rejected as violating the fundamental principles of epidemiology. Correlation does not prove causation. And yet, this conflation seems to be the norm, on both sides. In my world, one begins with a hypothesis inferred from available evidence, and then proceeds to test that hypothesis in the most rigorous fashion one can. In the economist's world one simply skews the evidence however one likes to support one's ideology. And if you throw in a lot of complicated math all the better! (See Krugman, Paul as your favorite example)
Economists don't have the tool of randomized controlled clinical trials that we do. But gheesh, can't they do better than this?