This is the second out of four papers devoted to the 2021 German federal elections continuing our anal-ysis of the 2009, 2013 and 2017 Bundestag elections. This paper arranges the contesting parties into a ‘spectrum’ that reflects the spatial proximity of their policy profiles. The latter are 38-dimensional vectors of the parties’ answers to 38 policy questions from the 2021 Wahl-O-Mat, the German voting advice application (VAA). Applying Principal component analysis (PCA), we construct a contiguous party ordering where the neighboring parties have close policy profiles. The ordering fits to the left-right ideological axis rolled up in a circumference, which can be unfolded by splitting it at one of its largest gaps. Rigorously speaking, we obtain a horseshoe-shaped left-right axis with the far-left and far-right ends approaching each other. For comparisons, alternative party orderings are constructed using four other models. Finally, the 2013, 2017 and 2021 German political spectra are compared.