Or maybe a professor could divide his undergrads randomly into two groups and send them one at a time for a 10pm stroll through either a high-class neighborhood or a housing project, keeping careful statistics of how many in each group have their money taken from them by violent means. Could differences in willingness to violently rob others be construed as saying something about differences in ethics or greed?
Of course the second study would never pass the research ethics board, but that's okay because, as it turns out, the police keep mountains of such statistics from the natural laboratory social scientists are supposed to be studying. Data on per capita "theft from vehicle" by neighborhood, per capita robbery, and so on, are easily available and could be incorporated into the study, except that...they wouldn't support the desired conclusion, making the study less helpful for promoting other agendas.