This is a guest post by Alec Ramsay of Dave’s Redistricting about a new class of metrics that measure whether redistricting plans favor one political party or the other.
People tend to use the terms “bias” and “fairness” interchangeably and sometimes use the term “partisan bias” generically even though it has a specific meaning in the literature. People also tend to treat prominent measures of “bias” as though they all measure the same thing, even though some metrics measure partisan gerrymandering via packing & cracking, some measure partisan symmetry, and others measure “fairness” relative to some normative standard.
The measures of partisan fairness allowed in the Senate’s recent Freedom to Vote Act (S2747) measure a quantity I call partisan advantage: the difference between the ideal and actual seat shares. As Nick Stephanopoulos pointed out in his analysis of its Rebuttable Presumption of gerrymandering provision, the language in the Act excludes measures of partisan fairness “that don’t specify an optimal seat share for a party’s given statewide vote share” – such as measures of partisan asymmetry, mean–median difference, and declination – but includes the efficiency gap and disproportionality.
In my paper, Estimating Seats–Votes Partisan Advantage, I extend Jon Eguia’s notion of partisan advantage to formalize this class of metrics which I call seat–votes partisan advantage. I show that most prominent measures of “bias” – including declination, lopsided outcomes, mean–median difference, seats bias, votes bias, geometric seats bias (), global symmetry, disproportionality, and the efficiency gap – don’t measure it specifically or aren’t reliable for states that are unbalanced politically. In contrast, also show that both disproportionality and the efficiency gap do measure partisan advantage and are reliable across a wide range of statewide vote shares.
This approach – using a measure of partisan advantage with an ideal seats–votes benchmark – can be enshrined in new federal law, like the Senate’s Freedom to Vote Act (S2747), and in state constitutions, as it is in Ohio.