I have a similar reservation regarding experiments. Of course, it would be possible to set up a survey or lab experiment that analyzes the research question at hand, for example by triggering people to see their government as incompetent, etc.
But it always remains unclear to what extent the results from the survey/lab actually translate into real-life politics. Experiments offer an artificial environment with often unclear generalizability. For me, that is the unique strength of observational data.
In my opinion, important topics should be analyzed with a mix of different methods due to their unique strengths and weaknesses—and cross-sections have their place in this methods mix, despite their weaknesses.
I am fully aware of the weaknesses relating to causal identification of the present paper, but it can show that real-world data is consistent with the presented theory. Optimally, further people will do connected research and use a different approach with
more credible identification to lend credibility to the theory. Even in this optimal world, the cross-sectional analysis still has value—it offers evidence supporting the claim that the credible causal effect estimates from, say, an experiment are “replicated” in observational
Comments