That's the beauty of randomization.

In North America, where, for ethical reasons, randomization requires informed consent, only two such trials have been conducted. The first was in New York in the 1960s, using annual mammograms and breast examinations for the screened group.

But even in microeconomics, where it is sometimes possible to generate precise empirical estimates using randomization techniques, the results must be extrapolated in order to be applied in other settings. New economic evidence serves at best to nudge the views a little here, a little there of those inclined to be open minded.