Carter, C., Pets. v. Chapman, L. ( 2022 )

[J-20-2022] [MO: Baer, C.J.]
IN THE SUPREME COURT OF PENNSYLVANIA
MIDDLE DISTRICT
CAROL ANN CARTER, MONICA : No. 7 MM 2022
PARRILLA, REBECCA POYOUROW, :
WILLIAM TUNG, ROSEANNE MILAZZO, :
BURT SIEGEL, SUSAN CASSANELLI, LEE :
CASSANELLI, LYNN WACHMAN, : ARGUED: February 18, 2022
MICHAEL GUTTMAN, MAYA FONKEU, :
BRADY HILL, MARY ELLEN BALCHUNIS, :
TOM DEWALL, STEPHANIE MCNULTY :
AND JANET TEMIN, :
:
Petitioners :
:
:
v. :
:
:
LEIGH M. CHAPMAN, IN HER OFFICIAL :
CAPACITY AS THE ACTING SECRETARY :
OF THE COMMONWEALTH OF :
PENNSYLVANIA; JESSICA MATHIS, IN :
HER OFFICIAL CAPACITY AS DIRECTOR :
FOR THE PENNSYLVANIA BUREAU OF :
ELECTION SERVICES AND NOTARIES, :
:
Respondents :
:
---------------------------------------------------------- :
PHILIP T. GRESSMAN; RON Y. DONAGI; :
KRISTOPHER R. TAPP; PAMELA GORKIN; :
DAVID P. MARSH; JAMES L. :
ROSENBERGER; AMY MYERS; EUGENE :
BOMAN; GARY GORDON; LIZ MCMAHON; :
TIMOTHY G. FEEMAN; AND GARTH :
ISAAK, :
:
Petitioners :
:
:
v. :
:
:
LEIGH M. CHAPMAN, IN HER OFFICIAL :
CAPACITY AS THE ACTING SECRETARY :
OF THE COMMONWEALTH OF :
PENNSYLVANIA; JESSICA MATHIS, IN :
HER OFFICIAL CAPACITY AS DIRECTOR :
FOR THE PENNSYLVANIA BUREAU OF :
ELECTION SERVICES AND NOTARIES, :
:
Respondents :
DISSENTING OPINION
OPINION FILED: March 9, 2022
JUSTICE MUNDY DECIDED: February 23, 2022
When the political branches approve a redistricting plan, the map will ordinarily
have gone through a public-comment stage, been sent to committee for amendment,
garnered majority support from both Houses of the General Assembly, and been
approved by the Governor. It will subsume a myriad of political choices and tradeoffs
which have been weighed, debated, and voted on by the public’s elected representatives.
These considerations may include how closely the districts should match those of the
previous plan, which non-retiring incumbents should be paired against each other in the
upcoming election cycle, which counties and other political subdivisions should or should
not be divided, which adjacent counties and townships should be grouped together, and
which communities of interest should be kept intact within a single district.
Items such as these are generally viewed as valid districting factors so long as
they do not subordinate the traditional, neutral criteria appearing in the state and federal
charters. See League of Women Voters v. Commonwealth, 178 A.3d 737, 817 (Pa. 2018)
(“LWV-II”) (citing Holt v. 2011 Legis. Reapportionment Comm’n, 67 A.3d 1211, 1235 (Pa.
2012)). As long as the plan that results from the political process does not “clearly, plainly,
and palpably” violate the constitution, League of Women Voters v. Commonwealth, 175
A.3d 282, 289 (Pa. 2018) (per curiam) (“LWV-I”), it will survive a court challenge.
[J-20-2022] [MO: Baer, C.J.] - 2
The present controversy is different. This is an impasse case in which the political
branches have failed to agree on a plan, and we have little choice but to wade into the
“political thicket” of redistricting. Evenwell v. Abbott, 578 U.S. 54, 58 (2016) (internal
quotation marks and citation omitted). Not only that, we are placed in an unfamiliar role:
we must make a selection rather than issue an adjudication. Stated differently, we are
not merely required to judge the legality of a plan, we are put to the task of choosing the
best among a number of competing plans that have been submitted for our consideration
by a variety of parties and amici. To the extent an adjudication is reached in this matter,
it is minimal and undisputed: the current map cannot be used because of population
shifts in the last ten years and, most notably, because Pennsylvania now has only 17
representatives in Congress.
In undertaking our selection task, it is vital that this Court act in a politically neutral
manner – and maintain the appearance of neutrality – to the greatest extent possible in
order that the public may have confidence our decision is reached via compliance with
neutral legal principles alone. In this respect, the Supreme Court has characterized the
need for objectively demonstrable standards in judging redistricting plans as being
necessary to enable the state legislatures to discern the limits of their
districting discretion, to meaningfully constrain the discretion of the courts,
and to win public acceptance for the courts’ intrusion into a process that is
the very foundation of democratic decisionmaking.
Rucho v. Common Cause, ___ U.S. ___, ___, 139 S. Ct. 2484, 2499-2500 (2019) (quoting
Vieth v. Jubelirer, 541 U.S. 267, 291 (2004) (plurality)). It is my position, then, that our
mission should be carried out solely in reference to the politically neutral criteria appearing
in the text of the state charter, namely: contiguity, compactness, population equality, and
respect for political boundaries. See PA. CONST. art. II, §16 (requiring districts which are
“composed of compact and contiguous territory as nearly equal in population as
[J-20-2022] [MO: Baer, C.J.] - 3
practicable,” and specifying further that, “[u]nless absolutely necessary no county, city,
incorporated town, borough, township or ward shall be divided in forming” such districts).1
Limiting our consideration to these express constitutional criteria has multiple
benefits. In addition to maintaining the appearance of neutrality, it helps avoid any subtle,
unconscious influence that political considerations might otherwise bring to bear upon our
decision-making. Relatedly, the map we select will be known by all involved to be that
which is most compliant with the Constitution’s commands as judged by an objective,
neutral standard open to public view.2 Such an approach also appears likely to reduce
any incentive the political branches might otherwise have to view an impasse as desirable
in its own right – in the sense that they would rather “take their chances” with this Court
than seek political compromise – and thereby, to reduce the incentive for those branches
to act strategically. And while I do not discount the theoretical possibility that
gerrymandering might occur within the confines of an effort to comply scrupulously with
1Article II, Section 16 only facially applies to state legislative districts. In the LWV-II,
however, a majority of this Court held that it applies, as well, to Pennsylvania’s
congressional districts through Article I, Section 5, the Free and Equal Elections Clause.
See LWV-II, 178 A.3d at 816.
2 In this regard, I agree with many of the sentiments expressed by Justice Brobson to the
effect that it is the Article II, Section 16 criteria, and not some concept of partisan fairness,
that should control any redistricting exercise; whereas, the experts’ fairness metrics may
be used in proving that a challenged map embodies illegal gerrymandering. See
Dissenting Op. at 8-9 (Brobson, J.). In my view, the neutral criteria appearing in the
Constitution’s text are insufficiently ambiguous to support the consideration of policy goals
that are claimed to have motivated their adoption. As Judge McCullough suggested,
moreover, the use of such policy goals as quality metrics in a map-selection endeavor
can lead to reverse gerrymandering aimed at altering the partisan performance which
arises naturally from the political geography of this state, which in turn stems from the
decisions of many individual voters concerning where they wish to live. See Special
Master Report at 197. Most importantly, the partisan-fairness metrics are not well suited
to an objective scoring methodology because political judgments must be made about
how to rank the maps in relation to such metrics.
[J-20-2022] [MO: Baer, C.J.] - 4
the state charter’s neutral directives, it seems evident that the closer a map adheres to
those directives, the less likely it will be that district boundaries have been manipulated
to give any political or partisan group an artificial advantage. As this Court recently
explained in LWV-II:
Because the character of these [constitutional] factors is fundamentally
impartial in nature, their utilization reduces the likelihood of the creation of
congressional districts which confer on any voter an unequal advantage by
giving his or her vote greater weight in the selection of a congressional
representative as prohibited by Article I, Section 5. Thus, use of these
objective factors substantially reduces the risk that a voter in a particular
congressional district will unfairly suffer the dilution of the power of his or
her vote.
LWV-II, 178 A.3d at 816; see also id. (noting these standards also comport with the United
States Constitution’s requirements for congressional districts).
All of this leads to the question of how to determine which of the proffered maps
best complies with the Constitution’s neutral factors after eliminating any maps that fail to
meet the constitutional floor. See generally LWV-II, 178 A.3d at 817 (“These neutral
criteria provide a ‘floor’ of protection for an individual against the dilution of his or her vote
in the creation of such districts.”).3 To answer this question, two observations may be
made. First, the maps can be analogized to candidates in an election where each criterion
by which they are judged is the equivalent of an individual voter taking part in a ranked-
choice voting exercise:
3A map might fail to meet the floor by, for example, containing districts which are not
contiguous, or by having an unjustified population variance between districts. Such maps
should be eliminated from consideration.
A given map must also comply with federal statutory law such as the Voting Rights Act or
it, too, will not be considered. Here, however, there has been no suggestion that any of
the proposed maps violates federal statutory law.
[J-20-2022] [MO: Baer, C.J.] - 5
When a court or agency purports to select one of many possible outcomes
by ranking the outcomes under a set of criteria, the situation parallels the
democratic process. In place of the preferences of individual citizens,
rankings under criteria determine judicial or administrative choices.
Matthew L. Spitzer, Multicriteria Choice Processes: An Application of Public Choice
Theory to Bakke, the FCC, and the Courts, 88 YALE L.J. 717, 717-18 (1979). This type
of decisional process – having multiple voters rank the contenders in an effort to select
the best one – has been applied in such diverse contexts as selecting the most valuable
player in sports, see Saul Levmore, More than Mere Majorities, 2000 UTAH L. REV. 759,
763, choosing an Academy Award winning film, see National Conference of State
Legislatures, Ranked-Choice Voting, Vol. 25, No. 24 (2017), available at
https://www.ncsl.org/research/elections-and-campaigns/ranked-choice-voting.aspx (last
viewed Feb. 23, 2022), nominating political candidates, see Maine Senate v. Sec’y of
State, 183 A.3d 749, 751-52 (Me. 2018), and electing political leaders, see id.The second observation is that ranked-choice voting can be accomplished through
pairwise comparisons of the candidates, in this case, the candidate maps. As long as
this Court has adequate data concerning how well the maps score for a given quality
metric at the most granular level (for example, the Polsby-Popper compactness metric),
any two maps can be compared to see which one is better, or if they are tied. These
pairwise comparisons can then be used to rank and score the maps for each quality metric
using the “Borda count” system.4 Under this system, for each quality metric, each map
receives one point for every other map it is superior to, plus one-half point for every other
4 The Borda count method is named after Jean-Charles de Borda, an eighteenth-century
French mathematician. See Edward B. Foley, Tournament Elections with Round-Robin
Primaries: A Sports Analogy for Electoral Reform, 2021 W IS. L. REV. 1187, 1200 n.39
(indicating Borda count is viewed as the best method to rank three or more candidates).
[J-20-2022] [MO: Baer, C.J.] - 6
map it ties with.5 In this way, the pairwise comparisons yield a “raw” Borda count score
for each map, for each quality metric at the most detailed level.
The method is simple and transparent. It is also flexible enough to accommodate
virtually any type of quality metric, including continuous metrics such as a map’s score on
a particular measure of compactness; integer-based metrics such as the number of
county splits or county pieces reflected in a given map; binary metrics such as whether a
map splits Pittsburgh (if this were indeed to be considered a valid quality metric); or criteria
with a few discrete points, such as how many non-retiring incumbents are paired and
whether they are from the same or opposite parties.6 These examples are given by way
5 See Bernard Grofman, Public Choice, Civil Republicanism, and American Politics:
Perspectives of a “Reasonable Choice” Modeler, 71 TEX. L. REV. 1541, 1565 n.110
(1993); Jean-Pierre Benoit & Lewis A. Kornhauser, Assembly-Based Preferences,
Candidate-Based Procedures, and the Voting Rights Act, 68 S. CAL. L. REV. 1503, 1522
& n.44 (1995).
With human voters, Borda count can be subject to distortion based on insincere (strategic)
voting, see Cheryl D. Block, Truth and Probability – Ironies in the Evolution of Social
Choice Theory, 76 W ASH. U.L.Q. 975, 987-88 (1998) (providing an example of insincere
ranked-choice voting and its underlying motivation), and it has been shown to sometimes
miss a majority winner, see Saul Levmore, Voting Paradoxes and Interest Groups, 28 J.
LEGAL STUD. 259, 266 n.9 (1999). These problems are absent here, as objective pairwise
comparisons cannot be insincere, and our goal is not to pick the map that comes in first
in most of the quality metrics, but to pick the best map overall.
6For example, the maps before the Court reflect the following non-retiring incumbent
pairings: one (R-D), one (R-R), two (R-D and R-D), two (R-R and R-D), two (D-D and R-
D), and none.
These can be ranked in order from best to worst as follows. Best: none; second-best:
one (R-D); third-best: two (R-D and R-D); fourth-best: one (R-R); worst: two (R-R and
R-D) or two (D-D and R-D).
Returning to the handling of Pittsburgh: the method can accommodate a three-point
quality measure where keeping Pittsburgh whole is best, keeping it whole via a “claw”
[J-20-2022] [MO: Baer, C.J.] - 7
of illustration, but, as explained, I will only be using the neutral constitutional criteria for
the present discussion – albeit in the Appendix, I also fold in the maps’ handling of
Pittsburgh which, for reasons delineated below, is sui generis.
I use the term “raw scores” because the Borda count methodology must be
modified slightly to be of use here. A map’s overall raw score is not ultimately what
matters, but its overall weighted score, as explained infra.7 As for terminology, I will refer
to high-level measures such as compactness and respect for political subdivision
boundaries as the neutral constitutional criteria, and the different ways of measuring those
criteria as individual quality metrics. This distinction is needed because there are multiple
ways to measure compliance with each criterion. For example, there are several
individual quality metrics associated with compactness, each capturing a different aspect
of mathematical compactness, and some accounting for such features as jagged state
borders or peninsulas which necessarily make districts less compact. See N.T., Jan. 27,
2022, at 214 (reflecting expert testimony stressing the importance of considering multiple
compactness metrics); Holt, 67 A.3d at 1242 (recognizing “an apparent variety” of
compactness models). Likewise, there are various different quality metrics relating to
subdivision splits, such as county splits, ward splits, county pieces, and so on.
shape which grabs it, as in the House Democratic Caucus’s proposed map, is second-
best, and splitting it is worst. The attached Appendix illustrates this scenario.
7 The weighting of criteria has been used in a variety of multi-criteria decision making
(“MCDM”) tasks involving selection. See Thiel v. W. Mifflin Borough, 2007 WL 1087773,
at *3 (W.D. Pa. Apr. 9, 2007) (hiring and promotion); Transactive Corp. v. N.Y. State Dep’t
of Soc. Servs., 665 N.Y.S.2d 701, 704 (N.Y. App. Div. 1997) (public procurement); Pickus
v. U.S. Bd. of Parole, 507 F.2d 1107 (D.C. Cir. 1974) (parole selection); Doe v. Alternative
Med. Md., LLC, 168 A.3d 21 (Md. 2017) (licensure selection); Lohn v. Morgan Stanley
DW, Inc., 652 F. Supp. 2d 812 (S. D. Tex. 2009) (assignment of client accounts to
financial advisors); Universal Grading Svc. v. eBay, Inc., 2009 WL 2029796 (E.D.N.Y.
June 10, 2009) (assessment of rare-coin grading services).
[J-20-2022] [MO: Baer, C.J.] - 8
Thus, for example, if compactness and respect for political boundaries are
considered equally important and each is given a total weight of 10, there may be X ways
to measure the former and Y ways to measure the latter. It follows that each
compactness-related individual quality metric should have a weight of 10/X, and each
boundary-related individual quality metric should have a weight of 10/Y. A map’s score
for a given individual quality metric, then, is its Borda count raw score multiplied by the
weight of that quality metric.8
Consistent with my remarks at the beginning of this opinion, I would hold that this
Court should rank and score all proposed maps according to each of the individual quality
metrics and select the map with the highest total weighted score. The process entails
five steps: (1) eliminate any map which fails to meet the constitutional “floor” or which
violates federal law; then as to each of the remaining maps: (2) compute raw scores for
each map for each individual quality metric using pairwise comparisons and Borda count;
(3) compute weighted scores for each map for each individual quality metric by multiplying
the raw scores by the weight for that individual quality metric; (4) compute the total
weighted score for each map by summing all weighted scores for that map; and (5) select
the map with the highest overall weighted score.
8 This type of weighting might also be useful in situations where secondary factors such
as preserving communities of interest are included in the analysis. This is because not
all such metrics are equally important, nor are they as important as the constitutional
criteria. See Majority Op. at 15 (noting such factors are “wholly subordinate to the
traditional core criteria”). Assigning different weights can reflect those realities. Similarly,
weighting can be useful if this Court ultimately reads the “unless absolutely necessary”
language in Article II, Section 16 as signifying that the Constitution places a higher value
on avoiding subdivision splits than on compactness. See generally Holt, 67 A.3d at 1242
(indicating that achieving population equality and avoiding subdivision splits may
“necessitate[] a certain degree of unavoidable non-compactness in any reapportionment
scheme.” (internal quotation marks and citation omitted)). For example, a total weight of
10 could be assigned to compactness, 7 or 8 to avoiding subdivision splits, and 3, 4, or 5
to the subordinate historical considerations.
[J-20-2022] [MO: Baer, C.J.] - 9
The maps presented to us, and the data contained in the expert reports concerning
those maps, reveal that all meet the contiguity and population-equality criteria, which are
essentially binary in nature.9 As noted, moreover, none are alleged to violate federal law.
See supra note 3. This leaves only the compactness and adherence-to-political-
boundaries criteria on which to form a judgment concerning which is the best of the maps
under review.
Twelve maps have been submitted for this Court’s consideration: the Carter
Petitioners’ map (“CARTER”) , the Gressman Petitioners’ map (“GRESSMAN”), Governor
Wolf’s map (“GOV”), the map approved by the General Assembly (“HB-2146”), the first
map by the Senate Democratic Caucus (“SEN-DEM-1”), the second map by the Senate
Democratic Caucus (“SEN-DEM-2”), the House Democratic Caucus’s map (“HOUSE-
DEM”), the first map by the Reschenthaler group (“RESCH-1”), the second map by the
Reschenthaler group (“RESCH-2”), the map submitted by the “Voters of the
Commonwealth of Pennsylvania” group (“VOTERS-PA”), the map submitted by the “Draw
9 Pursuant to the 2020 census, Pennsylvania’s population was 13,002,700, resulting in
17 districts with an average population of 764,864.7 per district. See Special Master
Report at 3 n.6. Because the population is not a multiple of 17, there must be a population
deviation, that is, the population of the most-populous district minus the population of the
least-populous district must be at least one person.
I am aware that some of the maps have a population deviation of two persons. However,
I do not consider the difference between a one-person and a two-person deviation to be
legally significant, particularly as the census numbers are only approximate due to
imperfections in data gathering combined with subsequent births, deaths, and relocations.
Put differently, discounting two-person-deviation maps as compared to one-person-
deviation maps would, in my view, be an exercise in false precision. Whether or not the
Constitution allows for a de minimis population deviation, I would find a deviation of two
persons to be sub-de minimis. For purposes of this case, then, I consider all maps with
a one- or two-person deviation as satisfying the constitutional equal-population criterion.
[J-20-2022] [MO: Baer, C.J.] - 10
the Lines” citizens’ group (“DRAW-LINES”), and the map submitted by the “Citizen
Voters” group (“CITIZEN-VOTERS”).10
These twelve maps have been given a compactness score for each of six different
mathematical compactness measurements: Polsby-Popper, Schwartzberg, Reock,
Convex Hull, Population-Polygon, and Cut Edges.11 Each map, in fact, has 17 scores for
these metrics because each has 17 districts for which a compactness measure can be
calculated. Helpfully, for each map the record contains average scores for each of these
quality metrics – that is, an average score which comprises the mean value for the 17
districts contained on a particular map. It is these averages that are used in the pairwise
comparisons between maps. Per the above discussion, each of the compactness metrics
is assigned a weight of 1.67 (10 divided by 6, rounded to the nearest hundredth).
The averages for the twelve maps on four of the six compactness metrics were
given by Dr. Daryl DeFord, see Majority Op. at 24, the expert who testified on behalf of
the Gressman Petitioners. The only two compactness metrics missing from Dr. DeFord’s
data are the Schwartzberg and Population-Polygon measures. Fortunately, however,
those are reflected in a table supplied by Dr. Moon Duchin, Governor Wolf’s expert, which
10 A thirteenth map was submitted by the Khalif Ali amici. It has been excluded because,
unlike all of the other maps, its boundaries were drawn based on data which attempted
to assign prisoners to their last known home address without first establishing a legal
basis for doing so. When assessed according to the data used by all the other maps, its
population deviation was too high to meet the constitutional requirement of equi-populous
districts. In any event, the record suggests it would not be the highest-scoring map in
terms of compactness and subdivision splits even if accepted on its own terms.
11 As explained, each such metric captures a different aspect of geometrical
compactness, and each has its strengths and weaknesses. Further elucidation of this
topic from a mathematical point of view is beyond the scope of this dissenting opinion. I
only note at this juncture that, for each metric except “Cut Edges,” a number closer to 1.0
is better. With the Cut Edges metric, a lower number is better.
[J-20-2022] [MO: Baer, C.J.] - 11
was endorsed by the Special Master. See Special Master Report at 141-43.12 All six of
these compactness measures are shown below in the row containing the map name.
From these averages, raw Borda count scores are obtained using pairwise comparisons;
as previously noted, a map’s raw score includes one point for each pairwise win, plus a
half-point for each pairwise tie, and so a higher raw score indicates better performance
on that metric. The raw scores are then multiplied by the weight for that metric to arrive
at the weighted score for each map for each metric:
MAP Polsby- Schwartzberg Reock Convex Population Cut
Popper Hull Polygon Edges
Weight 1.67 1.67 1.67 1.67 1.67 1.67
CARTER .31 1.8103 .41 .78 .7416 5896
Borda raw score 2.5 3 6.5 2.5 1 2
Weighted score 4.175 5.01 10.855 4.175 1.67 3.34
GRESSMAN .33 1.7351 .40 .80 .7582 5546
Borda raw score 5 5 4.5 8.5 5 4
Weighted score 8.35 8.35 7.515 14.195 8.35 6.68
GOV .37 1.6534 .40 .81 .7834 5154
Borda raw score 9.5 10 4.5 10.5 11 8
Weighted score 15.865 16.7 7.515 17.535 18.37 13.36
HB-2146 .31 1.8197 .38 .78 .7524 5882
Borda raw score 2.5 1 1.5 2.5 3 3
Weighted score 4.175 1.67 2.505 4.175 5.01 5.01
SEN-DEM-1 .30 1.8144 .37 .77 .7519 6016
Borda raw score 1 2 0 1 2 1
Weighted score 1.67 3.34 0 1.67 3.34 1.67
SEN-DEM-2 .32 1.7478 .38 .79 .7601 5476
Borda raw score 4 4 1.5 5.5 6 5
Weighted score 6.68 6.68 2.505 9.185 10.02 8.35
HOUSE-DEM .27 1.9693 .39 .75 .7205 6821
Borda raw score 0 0 3 0 0 0
Weighted score 0 0 5.01 0 0 0
RESCH-1 .35 1.6859 .43 .81 .7737 5061
12In Dr. Duchin’s report and table of map statistics, see Special Master Report at 141,
the DRAW-LINES map is referred to as the “CitizensPlan.” See N.T., Jan. 27, 2022. This
should not be confused with the CITIZEN-VOTERS map.
[J-20-2022] [MO: Baer, C.J.] - 12
Borda raw score 8 8 9 10.5 10 11
Weighted score 13.36 13.36 15.03 17.535 16.7 18.37
RESCH-2 .34 1.7127 .41 .80 .7658 5208
Borda raw score 6.5 7 6.5 8.5 7 6
Weighted score 10.855 11.69 10.855 14.195 11.69 10.02
VOTERS-PA .38 1.6069 .44 .79 .7681 5120
Borda raw score 11 11 10.5 5.5 8 10
Weighted score 18.37 18.37 17.535 9.185 13.36 16.7
DRAW-LINES .37 1.6625 .44 .79 .7725 5202
Borda raw score 9.5 9 10.5 5.5 9 7
Weighted score 15.865 15.03 17.535 9.185 15.03 11.69
CITIZEN-VOTERS .34 1.7133 .42 .79 .7575 5144
Borda raw score 6.5 6 8 5.5 4 9
Weighted score 10.855 10.02 13.36 9.185 6.68 15.03
In addition to the compactness metrics, there are five quality metrics relating to
how well a map keeps political subdivisions intact: counties split, county pieces,
municipalities split, municipality pieces, and wards split. Including a score for “ward
pieces” would amount to double-counting, as Dr. DeFord’s data reflect that no ward is
split more than once. The combined weight of these individual metrics will be set to
approximately 10, in accordance with the decision mentioned above to give equal weight
to compactness and respect for subdivision boundaries. Still, it is something of a
judgment call whether to consider these five quality metrics equally important and assign
each a weight of 2.0. In my view, doing so would diminish the importance of ward splits
without constitutional warrant, as all types of subdivisions are listed in Article II, Section
16 on equal terms. See PA. CONST. art. II, § 16 (“Unless absolutely necessary no county,
city, incorporated town, borough, township or ward shall be divided[.]”).
Separately, giving county splits and county pieces each a weight of 2.0 would
involve double-counting as the number of county pieces will depend, to a large extent, on
the number of split counties (and similarly for split municipalities and municipality pieces).
To ameliorate these concerns, I am assigning a weight of 2.00 for county splits, 1.34 for
[J-20-2022] [MO: Baer, C.J.] - 13
county pieces, 2.00 for municipality splits, 1.34 for municipality pieces, and 3.34 for ward
splits.13 The total weight is 10.02, the same as the total weight for the compactness
measures (6 x 1.67).14 The scores are set forth below in a manner similar to that for
compactness:
MAP Counties County Municipali- Municipality Wards
split pieces ties split pieces split
Weight 2.00 1.34 2.00 1.34 3.34
CARTER 14 31 23 44 21
Borda raw score 8 7 2.5 1 5
Weighted score 16 9.38 5 1.34 16.7
GRESSMAN 15 32 19 36 15
Borda raw score 5 5 10.5 10.5 10
Weighted score 10 6.7 21 14.07 33.4
GOV 16 35 22 41 25
Borda raw score 2 1 4.5 4 1.5
Weighted score 4 1.34 9 5.36 5.01
HB-2146 15 33 21 39 18
13 The county and municipal pieces metrics include all pieces, not merely “extra” pieces.
I note this because the data supplied by Dr. DeFord only includes the number for extra
pieces. For example, if a map splits, say, 20 municipalities into two pieces each, Dr.
DeFord’s data shows 20 split counties and 20 split pieces rather than 20 split counties
and 40 split pieces. The Borda counts will not change, however, as the ranking of maps
according to the “pieces” metrics is the same regardless of whether all pieces, or only
“extra” pieces, are counted.
As a separate matter, for consistency with the majority opinion, per Dr. DeFord’s data the
splits and pieces shown in the table include boroughs split by county lines. See Majority
Op. at 32.
14 A reasonable argument could be made that these items should be weighted differently.
One possibility would be to consider each type of municipality – cities, incorporated towns,
boroughs, and townships – on equal terms. But this could be distortive as there are
different numbers of the different types of municipalities. For example, Pennsylvania has
only one incorporated town (Bloomsburg). In the end, since counties are the basic sub-
units of governance, and because splitting wards can be especially problematic, I am
assigning a weight of 3.34 to counties, 3.34 to wards, and 3.34 to all other municipalities
combined.
[J-20-2022] [MO: Baer, C.J.] - 14
Borda raw score 5 4 6.5 5.5 7
Weighted score 10 5.36 13 7.37 23.28
SEN-DEM-1 17 36 25 45 17
Borda raw score 0 0 0 0 8
Weighted score 0 0 0 0 26.72
SEN-DEM-2 16 34 21 38 14
Borda raw score 2 2.5 6.5 7 11
Weighted score 4 3.35 13 9.38 36.74
HOUSE-DEM 16 34 24 43 21
Borda raw score 2 2.5 1 2 5
Weighted score 4 3.35 2 2.68 16.7
RESCH-1 13 29 20 37 25
Borda raw score 10.5 10.5 8.5 8.5 1.5
Weighted score 21 11.39 17 11.39 5.01
RESCH-2 13 29 20 37 24
Borda raw score 10.5 10.5 8.5 8.5 3
Weighted score 21 11.39 17 11.39 10.02
VOTERS-PA 15 31 23 42 41
Borda raw score 5 7 2.5 3 0
Weighted score 10 9.38 5 4.02 0
DRAW-LINES 14 30 22 39 16
Borda raw score 8 9 4.5 5.5 9
Weighted score 16 10.72 9 7.37 30.06
CITIZEN-VOTERS 14 31 19 36 21
Borda raw score 8 7 10.5 10.5 5
Weighted score 16 9.38 21 14.07 16.7
The final two steps are to compute the total weighted score for each map and
select the one with the highest total. Doing so yields the following scores, from highest
to lowest.15 As can be seen, RESCH-1 is the top-scoring map, followed by DRAW-LINES:
15 For the scoring in this opinion and the Appendix attached hereto, I have used a
spreadsheet to facilitate the calculations. The weights, raw data, and raw Borda scores
were entered manually. All other computations were performed by the spreadsheet
program. All total weighted scores are rounded to two decimal places.
[J-20-2022] [MO: Baer, C.J.] - 15
MAP Place Total weighted score
RESCH-1 1 162.83
DRAW-LINES 2 158.83
RESCH-2 3 142.79
CITIZEN-VOTERS 4 142.28
GRESSMAN 5 138.61
VOTERS-PA 6 121.92
GOV 7 114.06
SEN-DEM-2 8 109.89
HB-2146 9 81.66
CARTER 10 77.65
SEN-DEM-1 11 38.41
HOUSE-DEM 12 33.74
I note that I used Dr. DeFord’s data to align my scoring with the data used by the
majority (supplemented where necessary). To guard against possible distortion from the
use of only one data set, I also scored the maps based on Dr. Duchin’s table on page 141
of the Special Master’s Report. While there were slight variations in placement as among
all twelve maps, the top two scoring maps remained the same:
MAP Place Total weighted score
DRAW-LINES 1 166.51
RESCH-1 2 155.98
RESCH-2 3 138.45
CITIZEN-VOTERS 4 134.60
VOTERS-PA 5 131.27
GRESSMAN 6 129.26
SEN-DEM-2 7 116.57
GOV 8 113.89
HB-2146 9 83.15
CARTER 10 68.80
HOUSE-DEM 11 42.42
SEN-DEM-1 12 41.75
[J-20-2022] [MO: Baer, C.J.] - 16
Thus, with Dr. Duchin’s data the DRAW-LINES map was the top scorer, with
RESCH-1 as the runner-up. As between those two maps, however, only RESCH-1 keeps
Pittsburgh whole, whereas DRAW-LINES splits it in two.16 If this factor were to be given
weight as recommended by the Special Master, see Special Master Report at 150-51
(discussing evidence suggesting Pittsburgh should be kept within a single district); see
also id. at 149 (finding that splitting Pittsburgh allows a map to achieve a higher
compactness score), I would conclude that the RESCH-1 map should be chosen
regardless of which data set is used.
In all events, the CARTER map does not come close to rising to the top of the
pack. It seems notable, moreover, that, when compared with the other maps, the majority
does not purport to find that the CARTER map scores particularly well on the neutral
constitutional criteria on which the maps primarily compete, namely, compactness and
respect for county and municipal boundaries. See Majority Op. at 28 n.23 (reflecting that
the CARTER map is only a mid-level scorer in terms the compactness quality metrics
listed); id. at 33 n.26 (same with regard to the split-municipalities quality metrics).
Whichever data set was used, the CARTER map placed tenth out of twelve – thus,
in the bottom quartile. As the majority chooses that map for Pennsylvania, I respectfully
dissent.
16 With a population of approximately 302,000, Pittsburgh is the second-largest city in
Pennsylvania, and it is the largest city that does not need to be split to maintain population
equality among congressional districts. The third-largest city, Allentown, has a far-lower
population – around 125,000 as of the 2020 census. See
https://www.census.gov/quickfacts/allentowncitypennsylvania (last viewed Mar. 4, 2022).
Therefore, and because of the distinctly local emphasis of Pittsburgh’s political culture as
described by the Special Master, there appears to be particular importance attached to
the precept that Pittsburgh should not be split. The Appendix to this opinion reflects the
weighted quality scores of the maps if the handling of Pittsburgh were to be subsumed
as a quality metric. In that scoring, the RESCH-1 map scores highest.
[J-20-2022] [MO: Baer, C.J.] - 17
APPENDIX
As suggested in the attached dissenting opinion, the Borda-count scoring system
is versatile enough to subsume virtually any quality metric. All that is needed is the ability
to perform pairwise comparisons in reference to that metric. The handling of Pittsburgh
can be used to illustrate this concept. Per the Special Master’s report, it can be deemed
best to keep Pittsburgh within a single district. At the same time, keeping that city whole
via a normal-looking district can be viewed as superior to keeping it whole by grabbing it
with what the Special Master termed a “Freddy Krueger-like claw,” which gives the
appearance of gerrymandering. Special Master Report at 152, 203. Thus, one can
construct three quality levels in the following descending order of desirability: “whole,”
“claw,” and “split.” In that event, the seven maps that keep Pittsburgh whole would receive
a raw score of 8 because each is superior to five other maps and tied with six (5 + (0.5 x
6) = 8); the “claw” map would receive a raw score of 4 by being superior to the four maps
that split Pittsburgh; and those last four maps (the ones that split Pittsburgh) would receive
a raw score of 1.5 because each is tied with three other maps. Giving the handling of
Pittsburgh quality metric a weight of 4 (less than half as weighty as either of the neutral
constitutional criteria which each received a weight of 10.02), the maps’ handling of
Pittsburgh can be folded into the scoring system with the following raw and weighted
scores:
[J-20-2022] [MO: Baer, C.J.] - 18
MAP Handling of
Pittsburgh
Weight 4.00
CARTER Whole
Borda raw score 8
Weighted score 32
GRESSMAN Whole
Borda raw score 8
Weighted score 32
GOV Split
Borda raw score 1.5
Weighted score 6
HB-2146 Whole
Borda raw score 8
Weighted score 32
SEN-DEM-1 Split
Borda raw score 1.5
Weighted score 6
SEN-DEM-2 Split
Borda raw score 1.5
Weighted score 6
HOUSE-DEM Claw
Borda raw score 4
Weighted score 16
RESCH-1 Whole
Borda raw score 8
Weighted score 32
RESCH-2 Whole
Borda raw score 8
Weighted score 32
VOTERS-PA Whole
Borda raw score 8
Weighted score 32
DRAW-LINES Split
Borda raw score 1.5
Weighted score 6
CITIZEN-VOTERS Whole
Borda raw score 8
Weighted score 32
[J-20-2022] [MO: Baer, C.J.] - 19
When these weighted scores are added to the previous totals, the following ranking
emerges:
MAP Place Total weighted score
RESCH-1 1 194.83
RESCH-2 2 174.79
CITIZEN-VOTERS 3 174.28
GRESSMAN 4 170.61
DRAW-LINES 5 164.83
VOTERS-PA 6 153.92
GOV 7 120.06
SEN-DEM-2 8 115.89
HB-2146 9 113.66
CARTER 10 109.65
HOUSE-DEM 11 49.74
SEN-DEM-1 12 44.41
A similar ranking is generated when only the Dr. Duchin data are used:
MAP Place Total weighted score
RESCH-1 1 187.98
DRAW-LINES 2 172.51
RESCH-2 3 170.45
CITIZEN-VOTERS 4 166.60
VOTERS-PA 5 163.27
GRESSMAN 6 161.26
SEN-DEM-2 7 122.57
GOV 8 119.89
HB-2146 9 115.15
CARTER 10 100.80
HOUSE-DEM 11 58.42
SEN-DEM-1 12 47.75
[J-20-2022] [MO: Baer, C.J.] - 20
The above tables show that, when the handling of Pittsburgh is taken into account,
the RESCH-1 map scores highest, followed by either the RESCH-2 map (using the Dr.
DeFord data supplemented by the Dr. Duchin data) or the DRAW-LINES map (using only
the Dr. Duchin data). Moreover, the CARTER map is consistently in the bottom three
even though it keeps Pittsburgh whole.
[J-20-2022] [MO: Baer, C.J.] - 21