Political Spaces as Conceptual Spaces

The idea of a political space is intricately connected to the ubiquitous use of spatial metaphors when talking about politics. In particular the idea of a “distance” between “political positions” would suggest that politics could be situated in a metric space.

Using the electoral manifestos from the Manifesto project database we investigate to which extent the spatial metaphors so common in political talk can be brought to mathematical rigor and to which extent they can be considered as conceptual spaces in the sense of Gärdenfors.

In the manifesto data set each document is split into quasi-sentences that are assigned to one 56 categories. Each manifesto can be represented by a 56-dimensional vector of counts. These counts are not uncorrelated and therefore a set of manifestos can be approximately represented in a low dimensional space which is an example of a “political space”. We compare these political spaces withsimilar spaces based on topic modeling. There each party program is represented by a vector of topic probabilities, but in contrast to the manifesto categories these topics are learned fully unsupervised and thus more flexible than pre-defined categories.

When comparing different elections parties do not only change their positions in these political spaces, but one can also argue, that the meaning of the dimensions that span these spaces, change. We explore this "reconfiguration of the political space" in the context of the recent rise of populist parties, for instance in Germany.