> I've never seen a half-decent study in social sciences that draws rigorous conclusions.
Perhaps you should look harder. The dominant approach in economics for 20 years has been to reject correlational studies and try very hard to get at casuality, by:
* Running randomised controlled trials, often at scale (see eg Esther Duflo);
* Laboratory experiments, which have provided a body of robust paradigms and results;
* Seeking natural experiments;
* Statistical techniques like regression discontinuity and instrumental variables.
There's plenty of bad work in the social sciences. So is there elsewhere in the natural sciences (cough Lancet). There's plenty of good work too.
FWIW, I realise I didn't phrase it correctly. What I wanted to say was that any non-terrible ("half decent") study does not attempt to draw a final, black or white (I wrongly used "rigorous") conclusion, but that mainstream media will do that instead by choosing a particular interpretation of the study results.
I did not want to imply that social science studies are non-rigorous, I was actually trying to defend their scientific nature, but with an incorrect phrasing.
Economics is not really something people think of when they talk about humanities or social sciences.
Indeed, in Econ grad school I learned a lot about control theory, statistics, dynamic programming, etc. But I was never told to read Foucault, Levi-Strauss or even Marx - something that sociologists and other people in humanities usually have at least a basic understanding of.
If we judge what is science by level of quantitative rig our then economics is the only social science.
There are definitely areas of overlap with social sciences--especially these days. Behavioral economics (for which Richard Thaler won a Nobel Prize a couple years back) grew directly out of behavioral psychology for example.
Perhaps you should look harder. The dominant approach in economics for 20 years has been to reject correlational studies and try very hard to get at casuality, by:
* Running randomised controlled trials, often at scale (see eg Esther Duflo);
* Laboratory experiments, which have provided a body of robust paradigms and results;
* Seeking natural experiments;
* Statistical techniques like regression discontinuity and instrumental variables.
There's plenty of bad work in the social sciences. So is there elsewhere in the natural sciences (cough Lancet). There's plenty of good work too.