Sometimes a sample is also a population. We might not be able to generalize to all nations, but knowing the effect on the US would be very interesting in and of itself.

Other times it seems reasonable to draw conclusions without a sample, if we expect little variability in the population on the measure in question. For instance, if Obamacare has been in effect in Massachusetts for a few years, you don't say "n=1" and that the results have no bearing on what will happen in other states. You might argue that there are reasons it won't apply due to differing conditions, but few would say that it is as irrelevant as "n=1" would imply.