BREAKING: A statistical pet peeve!

WEDNESDAY, JANUARY 24, 2018

Statewide congressional vote versus gerrymandering: Gerrymandering can sometimes be in the eye of the beholder.

That pretty much isn't the case with Pennsylvania's current congressional districts. The 18 districts are so wildly irregular that they would look crazily gerrymandered to pretty much anyone's eye.

For one crazy-quilt example, see this Kevin Drum post. On the down side, Drum advances one of our favorite "pet peeve" statistical howlers at the start of that very same post:

DRUM (1/23/18): Yesterday the Pennsylvania Supreme Court declared the state’s congressional districts unconstitutional because they had been so badly gerrymandered. The result of the 2016 election bears out how effective the gerrymander was: Republicans won 54 percent of the congressional vote but received 72 percent of the congressional seats (13 out of 18).

Doggone it! That familiar construction suggests that a party which gets 54 percent of the statewide House vote should get something resembling 54 percent of the congressional seats. It certainly can work out that way, but it ain't necessarily so.

Imagine a state with 18 non-gerrymandered House districts. The imaginary state's congressional districts are as "regular" as can be.

This state's imaginary electorate is 54 percent Republican, 46 percent Democratic. But also imagine this:

That population is evenly spread across the state's 18 districts. In each of the 18 congressional districts, the electorate is 54 percent Republican.

In a state like that, the GOP would get 54 percent of the statewide vote, and it would win every single congressional seat! In principle, no gerrymandering is required to produce an outcome like that.

In the real world, Pennsylvania's districts plainly have been gerrymandered. No one seems to dispute this. But every time you see a gap between the statewide vote and the statewide allotment of districts, you aren't necessarily seeing evidence of foul play.

As has been widely observed, gerrymandering isn't the only force which can tilt the allotment of House against the Democratic (or Republican) party. As has been widely observed, Democratic voters tend to be heavily concentrated in large urban areas. In the absence of gerrymandering, this tends to produce urban House districts which are heavily Democratic.

In states which feature such concentrations, Democrats may win a smaller number of districts by huge margins. Republicans will thereby win a larger number of districts by smaller margins.

(Our own sprawling campus sits in one such urban district with lots of extra Democratic votes.)

These remarks have nothing to do with the claim that Pennsylvania is gerrymandered. They have to do with a statistical pet peeve, one which drives us wild.

Pictures from the leading authority: How screwy are the shapes of some Pennsylvania districts? The leading authority on the state shows you all the shapes here.