Re-Run DiD with Multiperiod data about alderman_machine HOT 3 CLOSED

This is proving trickier than expected.

1. The election cleaning R code is badly written. Super confusing. Need to refactor
2. I need to collect winners and losers in data. Focusing only on incumbent vote shares loses the incumbents who lost in the general election.
3. Furthermore, I need to create better lists of experience, retirements, and other removals from office.

from alderman_machine.

Currently have obtained results for "close-runoff" filtered results. That proved easier to produce due to less hard-coding required for the treatment periods. The results are roughly in line with previous results, but are much noisier and don't die off. Need to determine if this is due to different categorical definition or what.

from alderman_machine.

I think abandoning the category-based DiD is the best step here. These results are not really all that interesting and are extremely underpowered. All effort needs to be going to the geography-based DiD.

from alderman_machine.