>>Then it becomes equivalent to a 50% treshold locally. You are as represented as many places you can muster where you are a majority. You can have 49% everywhere and end up with zero seats, if you don't have 51% anywhere.
>
>That would be true in a 2 party system, but we have far more than that. We have 4 major parties and an up and comer which will mean 5 major parties before long. There is also a smattering of other parties that are lucky to get a few votes here and there. A party could conceivably win a riding with 21% of the vote if the rest of the votes were almost evenly split over the other 4 parties. (not that it's likely, of course). In that case, the party with 21% would get a seat and the others splitting 79% would get 'thanks for playing'.
Not exactly the same as gerrymandering, but basically a majority system. Miloševic pulled a similar thing a few times, when he began losing votes. He started splitting Serbia into smaller and smaller electoral units, until there were so few of the seats per unit as to make no room for fourth parties - with 20+ units in a 250 seat parliament, you get about 10 seats per unit; usually 7-8 go to two major parties or coalitions (the system being proportional but favoring stronger parties), and the remaining six parties, which may have carried 30% all together, would fight for the remaining two or three seats, which would go to the strongest one of them, and the others would blow.
You could have a fourth or a fifth party if you were regionally strong, but then he'd redesign the regions and split your region so that your few strongholds get spread across several neighboring districts, so even if you had, say, 20% in one region, more than a half of that would be with the region across the river where you haven't even started, and on election day you'd get 12% in one, 5% in the other (because your 8% meant less there, it's a larger one). It's not without reason that the opposition demanded to have the whole country's votes tallied together - so that not so much would be lost to rounding errors.