So I knew Brazilian parties were confusing and not ideological in the way we normally imagine parties to be. I looked at all of the coalitions made for governor’s races in 2018 and tried to see if there were patterns. The expectation was that left-wing parties would ally with other left-wing parties more often than with right-wing parties, etc.
I’m sure a clustering algorithm would have been a better way to map this, but at this point I don’t understand that, so instead I just assigned each party a number from 1-33, then found the frequency-weighted average score of a party’s allied partners, then sorted the parties by these alliance scores, then reassigned numbers according to this new order and recalculated alliance scores. By repeating this process, eventually the parties shuffled themselves into the right order.
The conclusion: there seem to be distinctly far-left parties (PCB and PSOL), distinctly left-wing parties (PT and PCdoB), a clump in the middle (shaded in green) that ally with everyone, and a clump on the right (shaded in blue) that ally with each other as well as with the middle.
Why each clump has so many parties within it, though, is a mystery.
I also graphed the final “alliance scores.” This is rather less interesting; the only finding here is that this graph is asymmetric: left-wing parties are more choosy about who to ally with (these parties have low numbers and they primarily ally with other parties with low numbers, giving them low alliance scores) than are right-wing parties.
Koji's Posts 