Back Talk has a lengthy, fascinating discussion of the deterrence effects of capital punishment. The article discusses the uses and misuses of data from the U.S. and Canada. His conclusion, which doesn't surprise me, is somewhat ambiguous:
One possibility is that you need about that many [40] executions per year to influence the murder rate in this country [the U.S.].My own view, as I have argued before, is that there are many fates worse than death, and I can't imagine that capital punishment is much, if any, more of a deterrent than really bad prison conditions might be.
... I do not mean to suggest that this evidence proves that executions serve as a deterrent to murder.
... [A] potenital murderer does not think to himself "gee, the state legislature just passed a temporary moratorium on executions, so I think I'll run out and kill Joe Blow before that changes." That kind of cognitive computation is not (in my view) how executions would deter murder.