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

Partisan Gerrymandering Across the 50 States

July 16th, 2017, 8:49am by Sam Wang

Note: I’ll pretty this up later. In the meantime, the files are available for you to download and inspect at the end of this post.

Over at the Associated Press, reporter David Lieb has published a new, in-depth analysis of the effects of gerrymandering in the 2016 Congressional and statehouse elections. The analysis found that the same states identified as partisan gerrymanders in 2012 and 2014 in my Stanford Law Review article — North Carolina, Michigan, Pennsylvania, and Maryland — also show clear signs of advantage to the same political paty.

The analysis is important for two reasons: (1) It means that the advantages built into the district maps in 2012 aren’t dissipating, and that these gerrymanders will likely hold up through the 2020 elections, and (2) These advantages aren’t an accident, because they are echoed at the level of state legislatures. In short, the parties that were in power in 2011 are likely to have a strong hand in drawing the maps again in 2021. They will go unfettered unless the opposing party gains the governorship*.

While the AP report primarily relies on the “efficiency gap” analysis, the Princeton Gerrymandering Project provided a separate t-test** analysis of election results for the article. The agreement between the two tests is striking. Virtually*** every state flagged by the t-test at the Congressional level is also flagged by the efficiency gap. Generally, we have found that the efficiency gap does well except for having a higher false-positive rate than the t-test, which is unsurprising since the t-test has such a venerable history. At the state legislative level, 4 of the 6 worst offenders according to the efficiency gap are also captured by the t-test with exceedingly low p-values.

The results of our analysis of Congressional races can be found here, and results for state house elections can be found here.

I thank Brian Remlinger and Naomi Lake for assistance with this post.


*Currently, the opposition party holds the governor’s mansion in Pennsylvania (Tom Wolf, a Democrat) and Maryland (Larry Hogan, a Republican). Note that in North Carolina, the governor has no role in redistricting.

**A note on statistical testing: Our analysis used one- or two-tailed t-tests, depending on redistricting authority. For states with single party control of redistricting, we carried out one-tailed tests for advantage in the direction of that party. For states with bipartisan or nonpartisan redistricting, we carried out two-tailed tests.

***The exception is California, whose top-two primary system complicates analysis.

Tags: 2012 Election · 2014 Election · 2016 Election · governors · Redistricting

2 Comments so far ↓

  • LondonYoung

    In your data file, I agree with your calling out Maryland, but it seems “un-scientific” to do so. And therein lie dangerous waters …
    Add in the exception for California and the waters muddy further.

    • Colin McAuliffe

      What do you think of Georgia? Looks like EG and the T-test have it as a kind of borderline case but our analysis determined it is one of the worst gerrymandered states since the 70’s. There are of course some methodological differences but it seems our analysis is the outlier among yours and the Brennan Center’s.

Leave a Comment