This paper is about the proportional voting system with open lists and many votes. Such voting systems differ in the number of votes per voter and in the number how often they can cumulate their votes on one candidate. In Europe's Nordic countries voters have usually one vote. In Switzerland, they have as many votes as representatives are elected, but can give only two votes to one candidate at maximum. Voting systems for local elections in Germany are split between these two versions. The southern communities follow the Swiss model, while the northern and eastern communities and city-states are different. Voters have three to five votes which they can all cumulate. In some sense, voters have the most options to distribute their votes. They can cross-vote for a team of five even from different parties similar to Switzerland or they can join all their voting weight on just their favorite candidate as in the Nordic countries. However, this freedom is undermined by strategic incentives in favor of full cumulation in two ways. First, cross-voters might fear a loss for all their preferred candidates and thus have an incentive to concentrate their votes. Second, candidates who can mobilize voters who fully cumulate have an advantage against candidates trying to mobilize cross-voters because they need a much lower number of voters for the same amount of votes. The paper shows that these strategic advantages seem to lead to empirical learning of candidates and the electorate in the two-city state Bremen from the 2011 to the 2015 election. We can observe a strong increase of ballots with fully cumulated votes. Further on, there is also an increase in the average number of votes per voter which candidates receive, especially for those candidates who finally get elected. BY the time of the ECPR general conference also the results of the 2019 election should be available for further analysis of these effects. The paper finally argues that among the open-list voting systems only the Swiss model (as many votes as representatives with limited or no cumulation) and the Nordic model (one vote per voter) are preferable. The Nordic model does fit to relatively small voting districts, while the Swiss model fits for both smaller and larger districts.
