Two things worth noting.
First, Kip Viscusi and others have found that when the costs are tallied correctly, smokers save the government money: they die earlier of cheaper diseases.
Based on surveys of smokers in the United States and Spain, for instance, he demonstrates that smokers actually overestimate the dangers of smoking, indicating that they are well aware of the risks involved in their choice to smoke. And while smoking does increase medical costs to the states, Viscusi finds that these costs are more than financially balanced by the premature mortality of smokers, which reduces their demands on state pension and health programs, so that, on average, smoking either pays for itself or generates revenues for the states.
Second, it's unclear whether these "cost the health system money" effects really are inefficiencies. They're called fiscal externalities, and are of pretty dubious status regardless of the direction of the sign. If we take Buchanan and Stubblebine's (1962) categorization of pecuniary and technological externalities, fiscal externalities don't get to count as the kind that cause inefficiencies. In the Buchanan-Stubblebine framework, pecuniary externalities affect others via the budget constraint while technological externalities enter directly via the utility function: only the latter has efficiency consequences. In short, under that view, we should be no more worried about inefficiency caused by somebody smoking more and costing the health system money than we should be worried about inefficiency caused by somebody bidding against you at a house auction.
Even if you want to count fiscal externalities as technological, it's well-neigh impossible to tell where to stop. If the existence of a public health system induces me to smoke more than I otherwise would, imposing costs on you, the taxpayer, it will also have effects on all kinds of other things I might choose to do. Folks will take on more risks on many margins and will invest less in avoiding health costs. At what point does the State get to tell us we're forbidden from risky sexual practices because of the costs of STDs on the health care system? As Edgar Browning so nicely showed us, the full set of taxes and subsidies that would eliminate all of the incentive effects inherent in the public health system would have an incidence similar to an actuarily-fair private health insurance premium, but with massive transactions costs.
Finally, Reason points me to an FDA campaign against electric cigarettes, which deliver nicotine vapor to smokers. While these electronic delivery mechanisms are far safer than regular cigarettes, the anti-tobbac folks worry that they'll prove the gateway into regular tobacco for non-smokers. Or, at least that would be the rational version of the argument. Arguments against harm-minimization devices of this sort often seem more akin to the Pope's reaction to a campaign promoting masturbation over pre-marital sex. Consider the position of the anti-tobbac lobby here in New Zealand with respect to chewing tobacco, or snus, which here is banned.
Snus is an addictive tobacco product that contains nicotine. The risk of developing cancer is far less among snus users than it is among smokers. However, it cannot be said that snus is an entirely safe product nor can it be claimed that it has been proven to be an effective cessation aid for smokers who want to quit. ...Rather than deregulating the use of oral tobacco in New Zealand more substantial health gains are to be had from the regulation of all tobacco products.In other words, even though it could well improve health outcomes for smokers, it's still tobacco, and tobacco is evil, and suffer not tobacco to live among you.
It's an empirical question around elasticities on the extensive and intensive margins, but I'd be extremely surprised if either electronic cigarettes or chewing tobacco proved net harm increasing. It's a shame that the onus isn't on the pro-ban folks to prove harm.