4 ms·
Political scientists are generally focused on national governments and/or international organizations. To paraphrase a well known economist and political commen
by zhdc1 5y ago
Political scientists are generally focused on national governments and/or international organizations. To paraphrase a well known economist and political commentator, modern governments are essentially insurance companies with their own militaries. Add on a wealth redistribution function, and it's easy to see why competent national governments mostly behave alike, regardless of the policy positions of the ruling party.
My feeling is that this changes dramatically when you look at local politics, which is where most of the interesting action is. There's not much that the French government can do that, when zoomed out, is fundamentally different from Germany or Spain. There's a lot more leeway when you compare Tokyo or Shanghai with Mexico City.
- ContrarianBrit 5y agoGood point. This article uses quasi-experimental methods and finds important partisan differences in Swedish local government: https://www.jstor.org/stable/40283092?seq=1#metadata_info_tab_contents https://www.jstor.org/stable/40283092?seq=1#metadata_info_ta...
- uniqueuid 5y agoThanks! I haven't read the paper yet, but noticed it uses a regression discontinuity design without showing raw data (i.e. a scatterplot). There have been some cases where such discontinuity analyses were extremely misleading [1], so some caution is advised. [1] https://statmodeling.stat.columbia.edu/2021/03/11/regression-discontinuity-analysis-is-often-a-disaster-so-what-should-you-do-instead-do-we-just-give-up-on-the-whole-natural-experiment-idea-heres-my-recommendation/ https://statmodeling.stat.columbia.edu/2021/03/11/regression...
- ContrarianBrit 5y agoThanks for the article! I love methods :-)
- uniqueuid 5y agoAwesome! Maybe you'll like this better method then: Bayesian counterfactual time-series analysis [1]. In essence, it compares a "neutral" time series and another one affected by a natural experiment and produces bayesian point-estimates for the causal impact. Super powerful, there's a great R package, and really easy to use. [1] https://research.google/pubs/pub41854/ https://research.google/pubs/pub41854/
- ContrarianBrit 5y agoThanks! I'll check it out :-)