The following is a guest post from Jon Eguia:

I propose a measure of partisan advantage in redistricting. A distinctive feature of this measure is that it captures only the partisan advantage that is due to the redistricting map in use, and not the advantage that is due to the geographic sorting of voters. For this reason, I call it the “artificial partisan advantage.”

I propose that we use county lines to identify a benchmark number of seats for each party, and that we then compare the actual seat outcome to this benchmark.

For a given election result, I define this benchmark seat share for Party A to be the share of the population that lives in counties won by Party A. Because county lines are exogenously fixed, this benchmark is not subject to manipulation by redistricting.

The artificial partisan advantage is the difference between the number of districts that Party A wins given the redistricting map in use, and the number of seats corresponding to the share of the population in counties won by Party A.

If Party A does better using the map’s district lines to assign seats rather than using (population-weighted) county lines, then the map confers an artificial advantage to the party.

For instance: in the 2018 election to the US House of Representatives in Ohio, the Republican Party obtained a 52.0% vote share, enough to be the most voted party in 81 out of 92 counties. These counties account for 54.2% of the population of Ohio, so given that Ohio has 16 House seats, the benchmark for the Republican Party was 54.2%*16 = 8.67 seats. Since the Republican Party won 12 House seats, the artificial partisan advantage for the Republican Party in this election was 12 – 8.67 = 3.33 seats.

 The Supreme Court has sought to find an acceptable measure of partisan advantage in redistricting since it ruled that partisan gerrymanders are justiciable back in 1986. Academics have proposed several such measures, but so far the Court has not accepted any of them. Over the years, the Court has expressed a growing wish list for a measure:

1. It must not rely on counterfactual voting outcomes.
2.  It cannot rely only on evidence of an asymmetry in outcomes across parties; it needs to show not just that a party has an advantage, but also that this advantage cannot be explained by the heterogenous geographic sorting of the population of voters.
3. It must be simple: “easily administrable”, “discernible”, “manageable.” It cannot be based on experts’ statistical models that are too sophisticated.
4. It can rely on jurisdictional lines (municipalities or counties), which the Court regards as “natural” limits.

The measure of artificial partisan advantage satisfies these desiderata.

Following other Supreme Court precedents, I propose the following 10% rule to determine if a redistricting map should be presumed to be a partisan gerrymander:

If the artificial partisan advantage, averaged over all years in which the map has been in use, is greater than 0.5 seats plus 10% of the size of the state’s delegation, then I suggest we presume the map is a partisan gerrymander.

If the artificial partisan advantage is below this threshold, then I suggest we presume the map is not a partisan gerrymander. The addition of 0.5 seats is justified because the benchmark allows for fractional seats, and actual maps do not, so a 0.5 margin allows rounding to the nearest interval.