An mathematical perspective on this problem is if you consider the German system as a transformation from a FPTP to a Popular Vote system: Let's say you have a

An mathematical perspective on this problem is if you consider the German system as a transformation from a FPTP to a Popular Vote system: Let's say you have a FPTP system with a total of D direct seats. To evaluate how distorted the FPTP vote is, you run a PV election in parallel. Let's say a party wins 'd' out of the 'D' FPTP seats. The party is over-represented if d/D > PV_percentage How do you fix this? Step 1: Introduce a scale factor 's' that increases the number of seats. You choose a minimal s > 1, such that for all parties: d/(s*D) <= PV_percentage In particular, for the most over-represented party the less-equals will become an equals. Step 2: Award additional seats 'x' to all other parties, until the invariant becomes equal for all of them: (d+x)/(s*D) == PV_percentage Perfect! Now, in the German system the 's' is additive (s'+D) rather than multiplicative (s*D). But you can always transform: s*D == ((s'/D)+1)*D == s'+D Originally, the system set s' to a fixed s'=D (implying s=2). With the rationale that parliament ends up with half FPTP seats and half additional (list) mandates. But, of course, then you can't scale past s=2 in Step 1 and end up with "Überhangmandanten". Since 2011 the new constraint is s>=2, making the size of parliament variable. This year, the CSU is predicted to get around 43 (out of 299) direct seats, but less then 5% of the popular vote. So: 43/(s*299) == 0.05 // s=2.88 and sizeof(parliament)=s*299=861 --- Imo, the solution would be to reduce D (the number of direct seats) and increase the number of list seats (s'). For example: D=200 and s'=400 (implying s>=3), which leaves more headroom for over-represented parties. Another lever would be to change from FPTP to approval voting as froh said.