Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

The effect of gerrymandering in four states exceeds that of population clustering in all 50 states

December 8th, 2015, 7:00am by Sam Wang

My New York Times piece was specifically focused on the legal question of what a party-neutral standard should look like. It didn’t address the political question: where do such deviations come from? As it turns out, the answer is: post-2010, partisan redistricting accounts for more than half of the total asymmetry in the House. The top four gerrymandered states have an effect that exceeds the effect of population clustering in all 50 states combined. The net effect of all gerrymanders combined is similarly large.

It is a commonly believed that the predominant force in partisan asymmetry is population clustering: groups that tilt Democratic are clustered into cities, generating a natural packing effect. A clustering effect certainly exists. However, as of 2012-2014, this effect has become secondary to gerrymandering in a handful of states.

Population clustering and partisan actions are not mutually exclusive. In fact, partisan gerrymandering relies on the fact that voters are not distributed perfectly uniformly. Using this fact, redistricters lasso voters into districts to suit political ends. For this reason, it is easy to mix up the two processes.

My statistical measures provide a way to separate these effects. As an obvious example of how redistricting can cause asymmetry, see the Michigan example at left. As you can see, the mean-median difference took a sudden jump from 2010 to 2012. The intervening event? Redistricting.

The same pattern appeared across many states in 2012. Here is a table of results calculated using my simulated delegations analysis, which I first proposed here at PEC, published in the New York Times in 2013, and detail further in my new paper on gerrymandering standards. The key tell is the difference is between two numbers, “Dem. seats” and “Simulated average.” Pink and blue shading indicate discrepancies that deviate from neutrality – too many or too few seats for a given vote share. In 2012, the shading blossomed: mostly favoring Republicans and in a few states favoring Democrats. In virtually all cases, the benefit went to the party that controlled redistricting.

This pattern of partisanship arose too suddenly to be caused by population migrations. Americans have been migrating to live in places where they find politically like-minded people. This has been called the Big Sort, and there is plenty of recent evidence documenting the phenomenon. One thing we know about the Big Sort is that it’s a gradual process. So these jumps, coincident with post-Census redistricting, are likely to arise from redistricting.

It is possible to quantify the effects of population clustering and partisan redistricting separately. In my SSRN paper, I estimate that population clustering was responsible for a net shift of 9-10 House seats towards Republicans. The total effect of post-2010 redistricting added a net gain of 11 additional seats for Republicans. (this combines 14 seats in seven GOP-controlled states with 3 seats in two Democrat-controlled states). In other words, partisan redistricting in just seven states created a distortion that exceeded the effects of population clustering in all 50 states combined.

One consequence of this is that a legal standard based on district shapes alone would not address the problem of asymmetry. In this respect, the Supreme Court was wise to avoid establishing such a standard. They could, however, adopt a symmetric standard like what I have proposed.

Tags: Redistricting

2 Comments so far ↓

  • Joseph

    As I understand it, Dr. Wang, what you’ve done is to create a statistical means for defining fair representation. As such, it is nothing less than a new way to approach building democracy. It’s hard to think of anything that would be more threatening to existing powers.

    It should also give you an advantage in predicting the results of elections, since you can now quantify both the population shift aspect and the purely political manipulation aspect. It may be that, over time, your improved accuracy will garner the necessary attention to the process whereby you’ve improved your predictability. That may be your best chance for eventual adoption of your work as a counter to blatant gerrymanderng, since your work is so clearly going to be anathema to those existing powers.

Leave a Comment