Via MR:
A rigorous statistical examination has found that smoking bans increase drunken-driving fatalities. One might expect that a ban on smoking in bars would deter some people from showing up, thereby reducing the number of people driving home drunk. But jurisdictions with smoking bans often border jurisdictions without bans, and some bars may skirt the ban, so that smokers can bypass the ban with extra driving. There is also a large overlap between the smoker and alcoholic populations, which would exacerbate the danger from extra driving. The authors estimate that smoking bans increase fatal drunken-driving accidents by about 13 percent, or about 2.5 such accidents per year for a typical county.
That's coming out in the Journal of Public Economics, so it might even be true. Here is the short source article, which surveys other interesting results as well; worth a read.
If other studies bear this out, I can see the ban advocates saying "See? We need to have smoking banned everywhere!"