Can Gerrymanders Be Detected? An Examination of Wisconsin’s State Assembly
| Published date | 01 September 2019 |
| DOI | 10.1177/1532673X18767890 |
| Date | 01 September 2019 |
| Subject Matter | Articles |
research-article2018
Article
American Politics Research
2019, Vol. 47(5) 1162 –1201
Can Gerrymanders
© The Author(s) 2018
Article reuse guidelines:
Be Detected? An
sagepub.com/journals-permissions
https://doi.org/10.1177/1532673X18767890
DOI: 10.1177/1532673X18767890
journals.sagepub.com/home/apr
Examination of
Wisconsin’s State
Assembly
Jonathan Krasno1, Daniel B. Magleby1,
Michael D. McDonald1, Shawn Donahue1,
and Robin E. Best1
Abstract
In October 2017, the Supreme Court heard an appeal of a November 2016
ruling striking down Wisconsin’s State Assembly districts as a Republican
gerrymander that illegally dilutes the weight of Democratic votes. We take
the opportunity to revisit this litigation to evaluate three proposed methods
of detecting gerrymanders: the “efficiency gap,” a count of Assembly
districts carried by statewide candidates, and the difference between the
district-level partisan median and mean. The first two measures figure
either centrally or peripherally in the plaintiffs’ case in Wisconsin, while
the third is the approach we favor. We expand on the analysis offered at
trial by evaluating how these measures fare across a variety of elections in
Wisconsin and with the aid of 10,000 alternative Assembly maps drawn by
computer. The alternative maps provide the appropriate baseline with which
to gauge the level of vote dilution in Wisconsin and distinguish between
the effect of residential geography and the Legislature’s actions. The
results show that Wisconsin’s Assembly map is a substantial gerrymander
according the median–mean comparison across all elections, while the two
tests relied upon by the plaintiffs provide mixed results. We examine the
measurement qualities of each test and show that the efficiency gap and
1Binghamton University, Binghamton, NY, USA
Corresponding Author:
Jonathan Krasno, Binghamton University, P.O. Box 6000, Binghamton, NY 13902-4600, USA.
Email: jkrasno@binghamton.edu
Krasno et al.
1163
districts-carried count both capture elements beyond partisan bias. We find
no similar ambiguity with the median–mean comparison and conclude that
the plaintiffs’ claim that Wisconsin’s Assembly map systematically dilutes the
weight of Democratic votes is correct.
Keywords
gerrymandering, efficiency gap, Gill v. Whitford, neutral maps, partisan
symmetry, median–mean comparison
Introduction
Partisan gerrymandering has been written off by many observers as an inher-
ently subjective phenomenon (Schuck, 1987). When Democrats like one set
of districts, Republicans are bound to object. When Republicans approve of
another, it becomes Democrats’ turn to complain. Yet, it is clear that district
lines do affect who is elected, and obvious that parties often try to press their
control of the line-drawing process to create systematic advantages for them-
selves in legislative elections. The questions are whether (a) these “system-
atic advantages” produce bias that is detectable using objective tests, and (b)
whether that bias is linked to a constitutional violation. In short, can gerry-
manders be measured and might they be unlawful? The Supreme Court, first
in Davis v. Bandemer (1986) and later in Vieth v. Jubelirer (2004) and League
of United Latin American Citizens [LULAC] v. Perry(2006), has held that
partisan gerrymandering is a justiciable issue, effectively answering the sec-
ond question. Some gerrymanders might indeed be illegal provided that
plaintiffs produce objective evidence demonstrating their effects. The plain-
tiffs in Gill v. Whitford, a group of Democratic voters in Wisconsin, invoke
the 14th Amendment by arguing that their state’s Assembly map illegally
dilutes the weight of Democrats’ votes.1 To support that claim, they present
empirical results using one proposed method of detecting gerrymanders, the
“efficiency gap” (EG), and by offering an affidavit about a second measure
focusing on the count of Assembly districts carried (DC) by each party.2
In November 2016, a panel of federal judges ruled in favor of the plaintiffs
by a 2-1 margin, and the Supreme Court agreed to hear the State’s appeal with
the oral argument taking place on October 3, 2017. During that argument,
many of the Justices’ questions were directed toward whether the EG could
serve as a reliable and effective standard for detecting a gerrymander with
several expressing skepticism bordering on derision. Chief Justice Roberts
went so far as to suggest that any empirical assessment of gerrymandering
might be nothing more than “sociological gobbedlygook.”
1164
American Politics Research 47(5)
Given the high stakes involved, we take the opportunity here to revisit the
case to address a question of interest to political scientists, judges, and citizens:
Can gerrymanders be reliably and objectively detected. Specifically, we exam-
ine these two measures used by the Whitford plaintiffs as well as a third we
favor, a comparison of the partisan median and mean at the district level. We
expand on the analysis offered at trial by examining a wider range of elections
in Wisconsin and by comparing them to an appropriate, within-sample counter-
factual that allows us to distinguish the effect of Legislature’s actions from resi-
dential patterns. The results show that the plaintiffs’ two measures do not
reliably identify Wisconsin’s Assembly districts to be a Republican gerryman-
der. The comparison of the partisan median and mean district, however, does
detect a substantial Republican gerrymander achieved by diluting the weight of
Democratic votes in every election and every test. If the plaintiffs’ evidentiary
standards are adopted, the ruling that Wisconsin’s Assembly districts are a ger-
rymander is potentially in jeopardy. If the median–mean (MM) is to be believed,
the plaintiffs’ claims of vote dilution are correct. Sorting out who is right and
why is essential for the current case as well as other lawsuits that may follow.
The circumstances of this lawsuit aside, Wisconsin is an ideal setting for
this inquiry for several reasons. As we describe below, the State Legislature
maintains and makes available an unusually comprehensive collection of
election data. More important, the state is politically competitive. In the 13
statewide elections conducted between 2008 and 2014 (the two cycles before
and after redistricting), Republicans won eight, Democrats won five, and all
but two were fairly close. That makes the stakes of vote dilution particularly
high for Wisconsin is the sort of state where either side could reasonably
expect to win control of its Assembly in a given election. Gerrymandering in
these sorts of circumstances could make it possible for a minority of voters to
consistently win a majority of legislative seats. In an area of law where the
debate rages over the proper translation of votes into seats, the notion of
“majority rule” is perhaps the single clear and agreed-upon principle.3
The 2016 presidential election serves as a reminder that majority rule is
not universally applied to all U.S. elections. Whatever the merits of the
Electoral College, its example provides a useful contrast to partisan gerry-
mandering. Obviously, state boundaries would have unknown effects on elec-
tions taking place more than a century later. Legislative boundaries, by
contrast, are redrawn every decade. They carry none of the historical weight
of state lines, but they also give the party in control an opportunity to entrench
its majority anew each decade. Indeed, the popular vote winner has almost
always carried the Electoral College in U.S. history. An arrangement within a
state or other jurisdiction that is biased so that it consistently awards a major-
ity of seats to a minority of voters is far different.
Krasno et al.
1165
We focus on three related issues in our examination of bias in the form of
vote dilution in Wisconsin’s Assembly districts: its magnitude, persistence,
and source. The first two dimensions are straight forward. Magnitude refers
to the size of the bias produced by a gerrymander, and persistence to its pres-
ence across elections. The latter is essential because we expect vote dilution
would be detectable across a range of elections else it suggests that voters
might do and undo the bias with their ballots.4 Source is particularly impor-
tant given the objections raised at trial and by the dissenting judge in Whitford
that the apparent pro-Republican lean of the Assembly map could be the
result of high concentrations of Democratic voters living in Madison and
Milwaukee. This effect of residents essentially packing themselves is known
as the “natural” or “accidental” gerrymander and has been recognized for
decades (Chen & Rodden, 2013b; Erikson, 1972, p. 1237; Vieth, 2004, pp.
289-290). As a result, it is useful to distinguish between the effect of geogra-
phy and the actions of the mapmakers. We do so here with the aid of 10,000
alternative Assembly maps of Wisconsin drawn by computer without refer-
ence to voting history. We argue these maps provide the appropriate baseline
with which to establish the extent of the natural gerrymander and differenti-
ate it from mapmakers’ actions.
We proceed in this essay to evaluate these three methods of detecting gerry-
manders as applied to Wisconsin’s Assembly map. “Three Measures of
Gerrymandering” section introduces and discusses the methods. “Data” section
moves to the data,...
To continue reading
Request your trial