The merits of legalization as harm reduction?

[bcubed]

Now, I know this might be a touchy issue, especially with the nature of this site and its mission, but I'd like to know if an answer is even capable of being determined.

One very true statement pro-legalization and -decriminalization proponents use is that, "The two legal drugs (alcohol, tobacco) kill more than all illegal drugs combined."

However, I can see the corollary that prohibitionists could use: "Well, that just goes to show what happens to the death rate when you make a drug legal!"

One obvious effect (to anyone who's thought about it) of criminalization is that use becomes more dangerous (outside the law, questionable quality, etc); the other effect is that TOTAL usage is reduced (both from increased price and legal consequences). As someone who's always identified more strongly with those who use anyways, I care more about lowering risk among users...plus the added ideological beliefs about choosing what one will/won't put into one's body.

But, from a "public policy" sense, you'd have to weigh the two, opposing effects: lowered usage, but what usage remains is more harmful.

What I'd like to know: is it possible to determine the risk of negative outcome (which could be any number of things, but let's just keep it simple and deal with fatalities) PER USE?

I mean, more people use alcohol and tobacco as drugs (in the US,at least) than any other, and they may well use it more often with more opportunity. Since legality makes alcohol and tobacco use look "safer" than they otherwise would, if the probablilty of death PER USE is higher than certain other drugs, then they are uniquely dangerous, because the risk is already skewed in their favor.

If, on the other hand, the high death toll is due chiefly to widespread use, then it's possible that the increase in use outwieghs the decrease in per-use risk--meaning that criminalization is not necessarily counter-productive (in this narrow sense).

How would one go about determining per-use risk? The gov't certainly wouldn't want this data known, even if it tended to support prohibition, because it'd make the trade-offs public knowledge.