I'm sure there's causal effects in there, but I also wonder if it's not similar to how riding an unregistered motorcycle puts you in a much higher risk cohort.

I'm sure there's causal effects in there, but I also wonder if it's not similar to how riding an unregistered motorcycle puts you in a much higher risk cohort. Of course, riding an unregistered bike is not going to cause an accident, but it's a behavior of people with very high risk profiles. As a motorcyclist, if you have a license, and your bike is registered, you become part of a behavior cohort that's barely represented in crash statistics when compared to the other cohort. Someone who's likely to smoke, is potentially the type of person who doesn't take care of themselves in a number of dimensions, putting them in a higher risk behavior cohort.