Whether to enforce or not capital punishment is one of those never-ending debates Europeans and Americans cannot grow tired of. In theory (and the classroom), we economists like to think that imposing a capital punishment on a specific action will make it disappear. Yet, people still do get the capital punishment, despite the fact that it is well know it is exists. Well, maybe the criminals were not quite sure of that, or they were wrongly convicted, or they were framed. But in any case, we should be seeing some evidence that capital offenses should be less prevalent in jurisdictions where capital punishment is enforced.
Steven N. Durlauf, Chao Fu and Salvador Navarro look at the literature and find that there is no consensus whether capital punishment has a deterrence effect. And quite far from it. These diverging results sometimes even come from the same datasets. This is an indication that the model specification matters a lot, and they demonstrate that indeed this matters. A linear specification always leads to a positive deterrence effects, while a non-linear one favors a negative deterrence effect. Of course, we have no idea which specification is the right one, thus we have no idea whether capital punishment is effective or not.