The main focus of the Quantifying Gerrymandering group is the development of what has been dubbed the ensemble method for outlier analysis. This perspective is relatively recent to the discussion of gerrymandering but rests on a rich scientific tradition in computational sciences and statistics.

  • A brief overview of our perspective is given below.
  • A more detailed account can be found in our blog posts.
  • A guide to our post is contained in the Reader’s Guide.
Quantifying Gerrymandering

The basic idea is to create a representative sample of non-partisan maps from a distribution on  redistricting which captures traditional redistricting criteria as historically expressed in the state. This collection, or ensemble, of maps can then used to create baselines to different questions of interest. Typical example questions include:

      1.  The expected number of districts won by the Democrats or Republicans  using a collection of precinct level votes.
      2. Quantify structural biases to one party or another due to natural packing or other geopolitical structures in the state.
      3. The typical margins of victory either in specific geographic locals or across the entire region.
A baseline and the Definition of Gerrymandering

The basic dictionary definition of gerrymandering is


      1. Manipulate the boundaries of (an electoral constituency) so as to favour one party or class. (Oxford)

      2. to divide or arrange (an area) into political units to give special advantages to one group.  (Merriam-Webster)

Hence the very definition of Gerrymandering rests on  having a baseline against which to detect “manipulat[ion]” or “special advantages.” The ensemble method gives a way to reveal what would have typically happened if only particular (usually non-partisan ) criteria were considered.

In many discussion, results are compared to proportional representation or other abstract ideals such as partisan-symmetry. The ensemble method does not make any assumptions about proportional representation. Rather, it  determines the effect of preferring particular redistricting criteria (usually non-partisan) when filtered through our district-based political system. Hence, it naturally accounts for effects such as natural packing or geographic features of the state. It only registers as atypical those manipulations or special advantages which go beyond what would be caused by natural packing.

A Quantifying Gerrymandering through outlier analysis

One strength of the ensemble method is that it can provide quantitative estimates of how likely it would be to find a map with a particular character if one was not looking particularly for a map with that property. This can speak to intent by showing that finding such a map by chance would be highly unlikely if one were only to consider the specified set of, often non-partisan, redistricting criteria listed.