Here are two basic facts about politics:
- Figuring out how to win elections and be popular once in office is very important.
- Doing the political science research necessary to demonstrate the best way to do that is really hard.
There are a few reasons for this second point. First, data quality in politics is often bad. Polling is hard (and getting harder) so figuring out what voters actually want is tricky. Second, you can’t really run randomized control trials of different political strategies. Parties want to win, and they’re not inclined to assign each of their 435 congressional candidates a slightly different strategy to produce the kind of data that would make it past peer review. Third, the effect size of any given political maneuver is generally pretty small. So without randomization and large sample sizes, it’s hard to know whether a politician’s victory should be attributed to their policy ideas, the national environment, or local events.
In this situation, I think it’s useful to zoom out for a second, and to look to the states to try to find some broad patterns of what seems to work for politicians and what doesn’t.