Examining how votes from 1st round of elections shift in the 2nd round

In my country, the presidential elections are set in two rounds. The two most popular candidates in the first round advance to the second round, where the president is elected. I have a dataset of the election results on municipality level (rougly 6.5k observations) - the % of votes in 1st and 2nd round for each candidate. I also have various demographic and socioeconomic variables for each of these municipalities.

I would like to model how the voting of municipalities …

advance country datascience dataset election elections popular president results set shift