Reply #9: It's a complicated problem. [View All]

[skids]

A competant discrete mathemetician might be able to solve it outright, however, the shortcut for those of us without that acuity is to just use a computer to iterate through a large variety of test cases.

You basically take a population of people and give them a few secondary attributes. Then you decide on a correlation factor between those attributes and vote preference. Then you run through the election and count up how many people with certain values for their attributes were disempowered by the results.

The attributes need to be divided into groups of different sizes. For example, one of the attributes might have values of urban/suburban/exurban/rural and out of every 100 voters, for example, you'd assign 50 urban 30 suburban 15 exurban and 5 rural. Then for another run you'd assign 50 urban 15 suburban 15 exurban and 20 rural.

The preference strength (e.g. 60% of "urban" voters strongly/weakly prefer candidate X over candidate Y) would also have to be varied to see what happens in close races versus those where a candidate highly popular with one attribute.

I'm sure there are reams of academic papers on this.