Redistricting: When Participative Geography meets Politics

Prepared for
Spatial Analysis Seminar
Institute for Social Research, University of Michigan
June 2012

Redistricting: When Participative
Geography meets Politics
Micah Altman <Micah_Altman@alumni.brown.edu>
Director of Research,MIT Libraries
Non-Resident Senior Fellow, Brookings Institution

This Talk
 Why is redistricting a difficult problem?
 How is technology changing redistricting?
 Application: DistrictBuilder & PublicMapping
 What kind of inferences can we draw from maps?
 Application: BARD

Redistricting: When Participative Geography
meets Politics

Collaborators*
Michael P. McDonald
Associate Professor
Department of Public and International Affairs
George Mason University
Web: http://elections.gmu.edu
[ A Principle Investigator on the Public Mapping project, regular co-
author since 1999]

Karin Mac Donald
Director Statewide Database
U.C. Berkeley
[ Co-author on studies of computer use in redistricting1980-2000 ]

Research Support

Thanks to the Sloan Foundation, the Joyce Foundation, Amazon,
Inc.
* And co-conspirators

meets Politics

Warning: this presentation is for educational purposes only and may contain
oversimplifications, errors, and/or preliminary conclusions. Caveat Lector.
Related Work For citation and reference please use the related published work below.

Reprints available from: micahaltman.com

 Micah Altman, 1997. "Is Automation the Answer? The Computational Complexity of
Automated Redistricting", Rutgers Computer and Technology Law Journal 23 (1), 81-142
 Altman, M. (1998b). Districting principles and democratic representation. California Institute of
Technology. Ph.D. Thesis.,
Altman, Micah (1999). "Modeling the Effect of Mandatory District Compactness on Partisan
Gerrymanders", Political Geography 17 (8): 989-1012.
 Altman , Micah . 1998. "Traditional districting principles - Judicial myths vs. reality Social
Science History22 (2): 159-200
 Altman, Micah 1998. "Modeling the Effect of Mandatory District Compactness on Partisan
Gerrymanders", Political Geography 17 (8): 989-1012.
 Altman, Micah, 2002. "A Bayesian Approach to Detecting Electoral Manipulation" Political
Geography 22(1):39-48
 Micah Altman, Karin Mac Donald, and Michael P. McDonald, 2005. "From Crayons to
Computers: The Evolution of Computer Use in Redistricting" Social Science Computer Review
23(3).
 Micah Altman, Karin Mac Donald, and Michael P. McDonald, 2005. "Pushbutton
Gerrymanders", in Party Lines: Competition, Partisanship, and Congressional Redistricting
Thomas E. Mann and Bruce E. Cain (eds), Brookings Press.
 McDonald, M.P. (2007) “Regulating Redistricting”, PS: Political Science & Politics, 40 : pp
675-679
 Levitt, J. & M.P. McDonald (2007), ‘Taking the “Re” out of Redistricting’, Georgetown Law
Journal 95(4) 1247-1285.
 Altman, M. and M.P. McDonald. (2010) “The Promises and Perils of Computer Use in
Redistricting”, Duke Constitutional Law and Policy Journal.
 Altman, M. and M.P. McDonald. (2011). "BARD: Better Automated Redistricting." Journal of
Statistical Software..
 Altman, M., & McDonald, M. P. (2012). Technology for Public Participation in Redistricting.
In G. Moncrief (Ed.), Redistricting and Reapportionment in the West . Lexington Books
 Altman, M., and M.P. McDonald (Forthcoming), Redistricting Principles for the 21rst Century,
Case-Western Law Review.

meets Politics

Why is redistricting a difficult
problem?

meets Politics

Definitions?
Redistricting. The aim of redistricting
is to assign voters to equipopulous
geographical districts from which they
will elect representatives, in order to
reflect communities of interest and to
improve representation.

Gerrymandering. Gerrymandering is
a form of political boundary
delimitation, or redistricting, in which
the boundaries are selected to produce
an outcome that is improperly favorable
to some group. The name
“gerrymander” was first used by the
Boston Gazette in 1812 to describe the
shape of Massachusetts Governor
Elbridge Ger- ry’s redistricting plan, in
which one district was said to have
resembled a salamander.

meets Politics

Redistricting Often Fails to Capture the Public Imagination

meets Politics

S i m p le s o lu ti o n – Ve rs i o n 1
“F i rs t p ri n c i p le s ”

Choose the redistricting plan that provides the “best” representation for the state.

Choose district plan X > plan Y, iff.
Representativeness(X)> Representativeness(Y)

meets Politics

Problem with Version 1…
There is a story about a very senior political scientist and a world- renowned
scholar in the field of representation who traveled to Russia shortly after the fall
of communism to lecture to the newly formed Duma.

After speaking, a newly-minted member of the Duma approached him and asked
him a question with great earnestness.

“I have been elected as a representative,” the
Duma member asked, “so when I vote, should
I vote the way I think the electors want me
to, or should I vote the way I think is right?”

“That’s a good question… Scholars have
been studying this for two thousand
years. And, let me just say, there are
many opinions.” Participative Geography
Redistricting: When
meets Politics

Simple solution – Version 2
“Let’s Randomize”
 Pure random redistricting equivalent to at-
large elections
[Grofman 1982]
 Compact districts on randomly clustered
population disadvantage parties with
geographically clustered support [Altman 1999,
Jerit and Barabas 2004; Rodden and Chen 2010]

In Vieth vs. Jubelirer 2004 , Justice Kennedy Agreed:

Second, even those criteria that might seem promising at the outset
(e.g., contiguity and compactness) are not altogether sound as
independent judicial standards for measuring a burden on
representational rights. They cannot promise political neutrality when
used as the basis for relief. Instead, it seems, a decision under these
standards would unavoidably have significant political effect, whether
intended or not. For example, if we were to demand that congressional
districts take a particular shape, we could not assure the parties that
this criterion, neutral enough on its face, would not in fact benefit one
political party over another.
meets Politics

Simple Solution Version 3
“Neutral Criteria”
 Eliminating judgment leads to calcification:
Electoral District-based systems are unique in
incorporating expert judgment into this process
converting voter preferences to candidate selection

 Weak empirical links between process and
outcomes
 Little empirical support for restrictions other
than population
 Population restriction, etc. has not prevented
gerrymanders
 Unintended consequences
 Baker & Karcher lead to widescale
abandonment of other traditional principles
(Altman 1998a)
 Intended (second order) consequences
 Choice of combination of neutral rules to
disadvantage minorities (Parker 1990
 Compactness rules have partisan
consequences (Altman 1999; Barabas 2005;
Rodden & Chen 2010) (Parker 1990)

meets Politics

Solution Version 4
“Consensus Criteria”
U.S. State Criteria
Scholarly criteria
Neutrality’ (unbiasedness) [Niemi & Coincidence with “major roads, streams, or
Deegan 1978] other natural boundaries”.
symmetry of seats-votes curve Coincidence with census tract boundaries.
‘Range of responsiveness’ [Niemi & Being “square, rectangular or hexagonal in
Deegan 1978]
range of vote shares across which shape to the extent permitted by natural or
electoral results change political boundaries.”
Constant Swing [Niemi & Deegan 1978] Being “easily identifiable and understandable by
increase of seat share is constant in voters”.
increase in seat share
‘Competitiveness’ [Niemi & Deegan 1978] Facilitating “communication between a
maximize number of districts with representative and his constituents”.
competitive margins Preserving “media markets”.
Compactness – perception of district
appearance [see Altman 1998b] Enhancing “opportunity for voters to know
Minimize voting for a loser their representative and the other voters he
(anticompetitiveness) [Brunell 2008] represents.”
‘Cognizability’ [Grofman 1985] Aligning with “prior legislative boundaries”.
‘Communities of Interest’ [See Forrest
2004] Consistency with “political subdivisions”.
Clustering [Fryer & Holden 2007] Utilizing “vernacularly insular regions so as to
Conformance with natural/administrative allow for the representation of common
boundaries interest”.
Media market preservation
Moderate majoritarianism
Continuity of representative relationship
(incumbency protection) [ see Persily
2003]
Graphical symmetry around expected
partisan vote share [Kousser 1996]
meets Politics

Problems with Version 4 - Satisfiability
 Logically exclusive:
 Competitiveness and anticompetitiveness

 Mathematically bounded:
 Can’t maximize competitiveness & guarantee constant
swing [Niemi & Deegan 1978]
 Can’t maximize competitiveness & symmetry [Niemi &
Deegan 1978]

 Empirically bounded:
 Compactness, communities of interest, competitiveness,
symmetry, etc.

meets Politics

Solution Version 5
“Let a Computer Do it”
 First suggested: 1961.
[Vickrey]
 Regularly proposed
 Advantages:
 Could increase transparency
 Could reveal range of
alternatives not otherwise
generated by political
process Results from “redistricter”
software. [Olson 2008]

meets Politics

Problems with Solution 5
 Too many solutions to enumerate:

1  r
r! 
S ( n,r) = ∑ ( −1) ( r − i)
r n
=
r! i= 0  ( r − i) !i!

 Even redistricting using common criteria is NP-complete
[Altman 1997]
 Optimal partisan gerrymander and optimal unbiased districts
also NP-complete [Puppe & Tasnadi 2008,2009]
 Not mathematically possible to find optimal solutions to
general redistricting criteria!
meets Politics

Are Redistricting Criteria
more Transparent than Plans?
 Compactness
 30+ different base measures to choose from, e.g.
Then variations …
 Map orientation can change results for Length-Width/Bounding box measures
 Map measurement scale can matter for perimeter based measures
 Map projection can matter for any measure
 Treatment of water?
 Ignore it
 Assign to closest land area C = .1 5 8
C= .1 9 4
 Include it
 Transform it away
 Treatment of bounding region?
 Ideal area – all area in bounding region
 Practical area – all area available for redistricting
 Treatment of population
 Ignore it
 Drop zero population blocks A Square is more
A: B: Rotating a district makes it
 Weight it compact than a circle? less compact?
 Type of population: any, voting age, citizen, eligible voter
 Transform map
 Even ‘contiguity’ involves many operational decisions:
 theoretical contiguity vs. census block contiguity vs. transportation feasibility
 single point vs. multi-point vs. line segment
 crossover permissibility
 nesting permissibility

meets Politics

No unbiased algorithm
 Dozens of ad-hoc heuristics for particular criteria
 Lots of general heuristics applied:
tabu-search, hill-climbing, evolutionary optimization, GRASP, TSP ,
recursive partitioning, k-means
[See Altman 1997, de Cortona et al 1999, Duque 2007]
 No heuristic does best on all problems – “no free lunch”
[Wolpert & Macready 1997]
 General heuristics still require adaptation/tuning to particular
problem in practice
 Thus potential for interaction between geography, criteria
and algorithms (“third order” bias)
 Not possible to completely disconnect algorithm, criteria,
and political geography

meets Politics

Solution Version 5
“Institutional Design”
 Redistricting/boundary “commissions” appear to be
reasonably well-insulated from partisan politics in most
other developed democracies using geographical districts:
United Kingdom, Canada, Australia, New Zealand
[See Altman-McDonald 2012 for a summary]
 Redistricting commissions in many states

meets Politics

The Problem with Institutions –
Participation Theater?
 Many U.S. independent non-partisan redistricting
commissions aren’t
 Ex-ante bi-partisanship  incumbent gerrymanders
 Ex-ante power to veto/modify by legislature
 Ex-post partisan litigation against commission?
 Many public hearings on redistricting are theater
 Do not change the plans actually adopted
 Do not produce evidence that can be used by the courts later
 Much public information isn’t
 Much electoral data is not shared
 Proposed plans not available for analysis
 Closed meetings and back-room deals are common

meets Politics

How is technology changing
redistricting?

meets Politics

Media Coverage is Oversimplified

 “In “Until is advances inkeepingisdistricts equalthesimple andduringcomputer allcan play
summary, this Article reapportionment — feed into the would to redraw
recently only political parties had
“The rapid only while computer…
“There
manpower and the tools
eliminationto describe in education politically feasible
of gerrymandering the last two
“The purpose of way to do technology and population. Now anybody
boundaries one a
seemfactorsprogrampoliticalcan reapportion a little asdistrictsautomatic
the to require as a registration.” as legislaturean other body
this make at least the establishment of or the geographic
decadesgame,it relatively simple to draw contiguous $3,500 of equal of
computer except which kibitzer. For
population ReaganCaliper Corp. will letdistricts. …The software and census data you
- Ronald [and] at the same time to further have the secondary goals the State
people who represent 1973]
analysis firm [Goff geographical you whatever redistricting program
andneedcomputer donovel geometries onvalue judgmentsHarvardaresearcher Micah
has.”impersonal procedure for PC screen. of those responsible
“Let a to is designed to implement the a carrying out
proposed try out it”
redistricting.in Karcher v. Daggett (1983)not at all difficult to districts.
for reapportionment”– [Nagel 1965] be
Altman and It appears to
- -Justice Brennan,others have put together a program that draws compact
Washington Post, 2003
( And software is free.
His many, many blogs)
devise rules for doing this which will produce results
notDemocratic redistricting could work like which would becommission in
markedly inferior to those this. After a census, a
each state entertains proposals from the political parties and any do-gooder
arrived at individual willing to compete. The commission picks the most--
group or by a genuinely disinterested commission.”
[Vickrey 1961] according to some simple criterion. (Say, add up the miles
compact solution,
of boundary lines, giving any segments that track municipal borders a 50%
discount, and go for the shortest total.) The mathematical challenge might
inspire some gifted amateurs to weigh in.” – William Baldwin, Forbes 2008

How can computers improve redistricting?

When computers became ubiquitous
 Computing systems
used in only a handful NO Voting Registration Block

of states in the SOFTWARE Data Data Data

1980’s 1992 Congressional 5.3%
(2)
71.4
(30)
64.3%
(27)
66.7%
(23)

 In 1990 redistrictings Legislative .9.3%
(4)
64.6%
(31)
58.3%
(28)
46.2%
(24)
2002 Congressional 0% 72.7% 75% 71 %
computer use was Legislative
(0)
0%
(32)
66%
(33)
68%
(27)
53.9%
nearly universal (0) (33) (34) (28)

 In 2000, computer
use was universal,
data use had
increased

meets Politics

What could redistricting system do in the last
round?
All systems used 2000 Congressional redistricting
could perform:
 Thematic mapping

Software % of Thematic Numeric Automated
Most could perform: Package States Mapping Tabulation Redistricting
&
 Tabulations Geographic
 demographic/voting variables Reporting
Autobound 45% Yes. Yes Yes.
 Geographic reports (Digital .
Engineering
 (compactness, holes, plan comparisons, Corp.)
contiguity) Maptitude 14% Yes. No. No.
(Caliper Corp.)

Only one could perform practical automated Maptitude for 12% Yes. Yes No.
redistricting. Redistricting
(Caliper Corp.)
Plan 2000 5% Yes. Yes No.
(Public Systems
With this exception, the available software tools Associates)
were qualitatively the same, although much Custom 14% Yes. Yes. No.
cheaper and faster, as in the previous 1990’s systems (Except
round of redistricting Texas)

meets Politics

GIS – Ubiquitous
 Goal: Aid in the efficient creation of maps associated
with data.
 First invented: 1962. [Tomlinson]
 Maturity: The 2000 round of redistricting.
[Altman, Mac Donald, McDonald 2005]

meets Politics

Computer use and compactness,
compared

 Measured Plan Compactness
Voting Data Automated Redistricting
 ‘Reock’ and Perimeter-Area 0.11
0.1
0.11
0.1

Compactness scores 0.09
0.08
0.09
0.08

 congressional districting plans (>3 0.07 0.07

PA
PA
0.06 0.06

districts) 0.05
0.04
0.05
0.04
0.03 0.03

 No significant difference by 0.02
0 1
0.02
0 1

software capability or voting data Block Data
0.11
USE_VOTING_DATA Tabulations and Geographical Reports
0.11
soft_automated_redis

use 0.1
0.09
0.1
0.09
0.08 0.08
 Districts using block data are 0.07 0.07

PA
PA
0.06 0.06

slightly more compact 0.05
0.04
0.05
0.04
0.03 0.03
0.02 0.02
0 1 0 1

USE_BLOCK_DATA soft_tabulation

Redistricting: When Participative

Expert Support Systems Unjustly Feared
 Fears
Mappers were able to specify a desired outcome or outcomes — the
number of people in a district, say, or the percentage of Democrats in it
— and have the program design a potential new district instantly. These
systems allow redistricters to create hundreds of rough drafts easily and
quickly, and to choose from among them maps that are both politically and
aesthetically appealing. [Peck and Caitlin, 2003]
 Evidence
 Widespread adoption of computers in the 1990’s post-dates precipitous
changes in district shape and composition
 Redistricting software prices dropped in 2000, but features remained essentially
the same.
 Competitiveness declined in 2000, after computers and election data already
ubiquitous.
 No statistical correlation between computer use/data and bad outcomes

meets Politics

Optimal Automated Redistricting
– State of the Art
 Enumeration
 Explicit enumeration intractable even for small #’s of units
 Early work with implicit enumeration (branch and cut) yielded
solutions for 30-50 units [E.g. Mehohtra, et. al 1998]
 Practice: Integer Programming
 Similar, but not equivalent, school districting problem solved
for < 500 units.
[Caro et. al 2004]
 Related multi-site land-use allocation problem solved for 900
units [Aerts, et al 2003]
 Integer programming applied to < 400 units, but used early
termination, rendering solution non-exact. [Shirabe 2009]

meets Politics

Heuristic Automated Redistricting
General Limitations
 Limited to population, compactness, contiguity
 Ad-hoc definition of compactness
 Often implicitly include a specific geographic model for districts
Recent Work
 “Redistricter” [Olson 2008]
 Not peer reviewed, but open source
 uses kmeans with ad-hoc refinements (including annealing) to solve
 Using 500000 census blocks can find solutions within 1% of population
 Weighted Voronoi Diagrams [Ricca, et. al 2008]
 Applied to up to <1300 population units
 Yielded large population variances
 Q State Pott’s Model [Chou and Li 2007]
 Applied to <450 population units
 Shortest Split-line [Kai et al 2007]
 Population variance of 5%
 Appears to require continuous data, results of discretizing solution not clear
 Ad Hoc Greedy Heuristics [Sakguchi and Wado 2008]
 <1000 units
 Yielded large population variance

meets Politics

Metaheuristics Approaches
 Genetic Algorithms [Xiao 2003]
 <500 Units (?)
 Population variance< 1%
 Genetic Algorithm w/TSP Encoding [Forman and Yu 2003]
 <400 Units
 Some post-processing
 Population variance< 1%
 Annealing [Andrade & Garcia 2009]
 <400 Units
 Tabu Seach [Bozkaya et. al 2003]
 <850 units
 Population variance <25%
 General Metaheuristics [Altman & McDonald 2010]
 Framework for multiple metaheuristics & criteria
 Preliminary results on <1000 units

meets Politics

State of the Practice

meets Politics

Application: DistrictBuilder &
PublicMapping

meets Politics

The Next Wave – Open Access Redistricting

 In last round of redistricting much more data was
available publicly, but public participation lagged.
 Generally, only well-organized political interests –
political parties, incumbents, and minority voting
rights groups – have had the capacity to draw
redistricting plans.
 This can change… “We have the technology.”

Ohio 2009 Competition
 Demonstration project by
Ohio Secretary of State
 Based on 2000 round data
 Complex user interface.
 Spearheaded by Mark
Salling, for state
 100+ participants, 11 final
plans

Major Components
 Identify barriers and principles
 Software
 Data
 Education
 Dissemination

Challenges to Transparency
 Participation matters
 Systems that help people identify their neighborhoods
 Systems that can create plans meeting all measurable legal criteria
 Systems that people can use

 Code matters
 Open Source for verification, replication, and correction
 Documented algorithms for measuring and adjusting plans

 Data matters
 Open data
 Complete information
 Accessible formats
 Known provenance

 Online systems do not guarantee transparency
 Are algorithms, code and data used open and transparent?
 Is sponsorship of the system transparent?
 Can data and plans be transferred in and out of the system freely

Principles for Transparency

 All redistricting plans should include sufficient
information such that the public can verify,
reproduce, and evaluate a plan
 Proposed redistricting plans should be publicly available in non-proprietary formats.
 Public redistricting services should provide the public with the ability to make available all
published redistricting plans and community boundaries in non-proprietary formats.
 Public redistricting services must provide documentation of any organizations providing significant contributions to their
operation.
 All demographic, electoral and geographic data necessary to create legal redistricting plans and define community
boundaries should be publicly available, under a license allowing reuse of these data for non-commercial purposes.

 The criteria used to evaluate plans and
districts should be documented.
 Software used to automatically create or improve redistricting plans should be either open-source or provide
documentation sufficient for the public to replicate the results using independent software.
 Software used to generate reports that analyze redistricting plans should be accompanied by documentation of data,
methods, and procedures sufficient for the reports to be verified by the public.
 Software necessary to replicate the creation or analysis of redistricting plans and community boundaries produced by the
service must be publicly available.

Bi-Partisan Endorsements for Principles
Endorsements for Principles of Transparency

• Americans for Redistricting Reform
• Brennan Center for Justice
• Campaign Legal Center
• Center for Governmental Studies
• Center for Voting and Democracy
• Common Cause
• Demos
• League of Women Voters of the United States.

Project Advisory Board

- Nancy Bekavac, Director, Scientists and Engineers for America
- Derek Cressman, Western Regional Director of State Operations, Common Cause
- Anthony Fairfax, President, Census Channel
- Representative Mike Fortner (R), Illinois General Assembly
- Karin Mac Donald, Director, Statewide Database, U. C. Berkeley
- Thomas E. Mann*, Brookings Institution
- Norman J. Ornstein*, American Enterprise Institute.
- Leah Rush, Executive Director, Midwest Democracy Network
- Mary Wilson, President, League of Women Voters
* Co-Principle Investigators and Editors

Public Mapping Software – Features
Powered by Open Source
 Create
 Create districts and plans
 Identify communities*
 Evaluate
 Visualize
 Summarize
 Population balance
 Geographic compactness
 Completeness and contiguity
 Report in depth
 Share
 Import & export plans
 Publish a plan
 Run a contest
* Coming soon
meets Politics

The Public Mapping Project
Supported by
The Sloan Foundation
Joyce Foundation
Amazon Corporation
Judy Ford Wason Center at Christopher Newport Univ.

Michael McDonald Micah Altman
George Mason University Harvard University
Brookings Institution Brookings Institution

Robert Cheetham
Azavea

The DistrictBuilder software was developed by the Public
Mapping Project with software engineering and implementation
services provided by Azavea

meets Politics

Advisory Board

Mike Fortner Illinois state Representative, 95th District
Carling Dinkler John Tanner's office, Tennessee 8th Congressional District
Mary Wilson Past President, League of Women Voters
Derek Cressman Western Regional Director of State Operations, Common Cause
Anthony Fairfax President, Census Channel
Kimball Brace President, Election Data Services
Gerry Hebert Executive Director, Campaign Legal Center and Americans for Redistricting Reform
Leah Rush Executive Director, Midwest Democracy Network
Nancy Bekavac Director, Scientists and Engineers for America
Karin Mac Donald Director, Statewide Database, Institute for Government Studies, UC Berkeley
Thomas E. Mann Senior Fellow, The Brookings Institution
Norman J. Ornstein Senior Fellow, The American Enterprise Institute

meets Politics

Map a State -- Change the Nation

 Identify communities
 Explore the alternatives
 Understand political consequences
 Establish transparency
 Catalyze participation
 Create alternatives to politics-as-usual

meets Politics

Launch Your Own Competition
www.districtbuilder.org

More Information at
www.publicmapping.org

meets Politics

Now…Public Access Redistricting

Open Data Open Access Open Source
meets Politics

Choose Your Legislature

meets Politics

Look Around – Get the Picture

meets Politics

Drill Down – Get The Facts

meets Politics

Make A Plan

meets Politics

Get the Details

meets Politics

Run The Numbers

meets Politics

Is it legal? How Well Are You Doing?

meets Politics

Spread the Word

 Shareyour plans with others
 Have a contest
 Mash up and plug in

meets Politics

Early Results
 More plans produced in the
Virginia Competition than in
any previous public effort
 Suggests that public plans can
be truly different
 Illuminates tradeoffs among
redistricting goals

meets Politics

Completed Redistricting Competitions
 Eight competitions
in different states
 Hundreds of legal
plans
 Thousands of active
participants
 Millions of viewers

meets Politics

External Impact?
 Politico
“best policy innovation” of 2011
 APSA ITP Award
 Dozens of local and national media articles

meets Politics

What didn’t we do?
 Surveys
 Randomized interventions – group level
 Pre-post evaluation
 Randomized intervention – individual level
 Deploy community mapping effectively

meets Politics

Increasing Public Participation

Draw the Lines

Evaluate maps
Get the data
Watch the
News

"Forthe first time in U.S. history, a court has allowed the
public to submit their own redistricting maps for
consideration."
meets Politics

Where Do We
Go From Here?

meets Politics

What kind of inferences can we
draw from maps?

meets Politics

Redistricting as Graph
Partitioning

meets Politics

Forecasting
 Forecasting, Inference & Optimization in redistricting
 Forecasting election results does not require combinatoric optimization,
may be convincing
 Observational/correlational evidence may be compelling when analyzing
change of institutions
 Revealed preference sometimes revealing
 Statistical causal inference (e.g. that district plan intended as a
gerrymander)
often rests on hidden & unreliable computational assumptions

meets Politics

Using WARP in Redistricting
1. Divide the relevant characteristics of a plan into three categories.
 Criteria that describe the feasible set of districts, K, such as the maximum population deviation between
districts, contiguity, etc.
 A characteristic, I, that best proxies the intent you wish to test.
 Characteristics, R, representing any other politically relevant criteria.
2. Enter the current plan, and any alternative plan that is part of the public record, into a GIS system, along
with the data necessary to evaluate K, I, and R.
3. Use these plans as starting points for metaheuristic optimization (simulated annealing, genetic algorithms,
or GRASP)
4. Use the optimization algorithm to search for a plan p* that such that:
If a feasible p* exists, then the motive proxied by I can not have been overriding (lexically preferred).
5. To further explore the trade-off among criteria. Use the optimization algorithm to search for a plan p**
that such that:
And then maximize I, subject to holding all but one (j) of the other relevant criteria constant
 Use optimization algorithms to find alternatives to given plan
 Attempt to find alternatives that differ in one dimension
 Probe the trade-offs among maj-minority seats, partisan seats, etc.

meets Politics

Sampling Problem - General

meets Politics

Sampling Bias in “Random Districting”
Analytic Approach

3 1

2 4 2 6

5*
5*
1 3

2 2
4 6
1 2 3 1 2 3 1 2 3 5 5

4 5 6 4 5 6 4 5 6 3

4 Event-trees showing the
1 generation of the district plans in
5*
2-District Plans using 6 counties 2 six block case. Each sub-tree is
equally likely (P=1/6), and the
1
3
probability of following any branch
5* at each node is equal to 1/
(number of branchs). Starred
6 nodes indicate illegal plans, which
5*
cause the algorithm to restart with
subtree selection
meets Politics

Legal Standards for Proving Discriminatory Intent
 “Ely”: no other rational basis can explain state action
 “Arlington Heights”: action would not have occurred
but for illicit motives
 “Feeney”: action was caused, in part, because of illicit
motives
 “Miller”: race is the predominant motive of
legislature

meets Politics

Intent: Definitional Problems Collective Intent Predominant Intent
 Collective Intent. Hypotheticals Hypotheticals

 collective choice
fundamentally differs 1. Four legislators split on 1. Legislature wants to
from individual intent plan for non-racial reasons.
Fifth legislator has only racial
create maj-min. districts, but
would trade two maj-min.
(Arrow) motivations votes for
legislation solely on racial
seats in to gain an additional
democratic seat overall.
 Scalia:with respect to 99.99 grounds.
2. Like 1 above. But 5th 2. Control of legislature is
percent of the issues of legislator has lexicographic threshold criteria (lexically
construction reaching the preferences, race is 10th in preferred). Once this
the list of her preferences, threshold is reached,
courts, there is no legislative but plans are identical on first legislature prefers to create
intent, so that any clues 9 criteria. majority minority seats.
3. Only one legislator in a 3. As above, but legislature
provided by the legislative (non-minimal) majority has will only values creating a
history are bound to be any racial motive. That maximum of K maj-min
false.” – Antonin Scalia legislator’s sole motive is seats.
racial.
 Predominant Intent 4. Legislature assigns simple
linear weights of {3,3,4} to
 Must be defined precisely protecting (non-racial)
before a mechanical test can be communities of interests,
applied protecting incumbents, and
creating majority-minority
seats

meets Politics

Intent: Statistical plan p* from the set of all possible plans
The planner chooses some
Problems
conditioned on their intent y.
The scholar must supply plausible hypotheses about the value that y could
take on. E.g.:
(0) intent only to create legal districts
(1) mixed -- intent to maximize probability of controlling the legislature,
protect incumbents, protect communities, and create k maj-min seats
(2) predominant intent to create a racial gerrymander.

We observe characteristics of the plan C, and of elections EV(E)
When is it likely that the observed characteristics indicate a
gerrymander? Using Bayes’ rule, this occurs iff. (Altman 2002):
( {~
prob y ∈ {1 ∪ 0} | C, EV (E)
=
}) ~
({ } )
prob C, EV (E) | y ∈ {1 ∪ 0} prob( y ∈ {1 ∪ 0})
<< 1
( { ~
prob y = 2 | C, EV (E) }) ({
~
} )
prob C, EV (E) | y = 2 prob( y = 2)

* If you prove the existenceRedistricting: When Participative Geography
of a deity…
meets Politics

Current Statistical Methods for Assessing Intent
Ignore Counterfactuals*

 CDR Methodology: if “random” plans differ from actual
plan racial motivation predominates
 Typical Use of Bias/Responsiveness: if bias is high 
assume partisan motivation
 Flaws:
 At most, we can reject ‘null’ hypothesis that
(
~
prob y = 0 | C, EV (E) { })
 No basis for evaluating likelihood of competing hypotheses
regarding intent

* Will the real null hypothesis please stand up… Redistricting: When Participative

Application: Better Automated
ReDistricting (BARD)

meets Politics

BARD – Open Source for Experts

 Create plans
 Fix/Refine plans
automatically
 Generate reports
 Compare plans
 Analyze

… Released in 2006

BARD
 What is BARD
 A tool for modifying and analyzing redistricting plans
 An open model for implementation of redistricting criteria
 A framework for experimentation with redistricting
 An R module
 What BARD is not
 A GIS system
 “the redistricting game”
 A comprehensive battery of tests
 A pushbutton gerrymander
 A solution to optimal redistricting

meets Politics

What can you do with BARD?
 Analyze district plan characteristics
 Create districting plans
 Explore trade-offs among different redistricting goals

meets Politics

Modular, Object-Oriented
 Configuration
 Data
 Goals
 Plan generation
 Randomly generate
 Use existing plan
 Create plan interactively
 Refinement
 Refine starting plan to fit goals
 Create sequence of plan with
reweighted pairs of goals
 Analysis
 Analyze single plan
 Compare pairs of plans
 “Comparative statics” for pair of
goals

meets Politics

Plan Input and
Manipulation

 Load a plan from a GIS shape
file
 Generate plans
 kmeans
 greedy
 random assignment
 Create or edit a plan
interactively

meets Politics

Generating Plans

meets Politics

Plan Comparison
 Generate reports
 contiguity
 holes
 compactness
 population equality
 competitiveness
 partisanship
 communities of interest
 completely extensible

meets Politics

Plan Comparison (Visual)

meets Politics

Metaheuristic Refinement
 Greedy Local Search:

 Metaheuristics combine
 some form of greedy local search with
 a mechanism for making non-improving selections – needed to escape local optima
 a mechanism for increasing search effort in a selected solution space, based on
history of plans evaluated
 As much “art” as science
 Examples:
 Simulated annealing
 Tabu Search
 Genetic optimization
 GRASP
meets Politics

Arbitrary Plan Sampling

meets Politics

Some Informal Observations
 Field experiments are hard …
 Kranzberg’s law
– technology is neither good, nor bad – neither is it
neutral
 Technology matters in politics
 Transparency, data and information technology are
interconnected
 Data transparency can enable participation
 Transparency & Data Involves
 IP law
 Electronic Access / formats
 Timeliness
 Completeness
meets Politics

Additional References
 J. Aerts, C.J.H,. Erwin Eisinger,Gerard B.M. Heuvelink and Theodor J. Stewart, 2003. “Using Linear Integer Programming for Multi-Site Land-Use
Allocation”, Geographical Analysis 35(2) 148-69.
 M. Andrade and E. Garcia 2009, “Redistricting by Square Cells”, A. Hernández Aguirre et al. (Eds.): MICAI 2009, LNAI 5845, pp. 669–679, 2009.
 J. Barabas & J. Jerit, 2004. "Redistricting Principles and Racial Representation," State and Politics Quarterly¸4 (4): 415-435.
 B. Bozkaya, E. Erkut and G. Laporte 2003, A Tabu Search Heuristic and Adaptive Memory Procedure for Political Districting. European Journal of
Operational Re- search 144(1) 12-26.
 F. Caro et al . , School redistricting: embedding GIS tools with integer programming Journal of the Operational Research Society (2004) 55, 836–849
 PG di Cortona, Manzi C, Pennisi A, Ricca F, Simeone B (1999). Evaluation and Optimization of Electoral Systems. SIAM Pres, Philadelphia.
 J.C. Duque, 2007. "Supervised Regionalization Methods: A Survey" International Regional Science Review, Vol. 30, No. 3, 195-220
 S Forman & Y. Yue 2003, Congressional Districting Using a TSP-Based Genetic Algorithm
 P. Kai, Tan Yue, Jiang Sheng, 2007, “The study of a new gerrymandering methodology”, Manuscript. http://arxiv.org/abs/0708.2266
 J. Kalcsics, S. Nickel, M. Schroeder, 2009. A Geometric Approach to Territory Design and Districting, Fraunhofer Insititut techno und
Wirtshaftsmethematik. Dissertation.
 A. Mehrotra, E.L. Johnson, G.L. Nemhauser (1998), An optimization based heuristic for political districting, Management Science 44, 1100-1114.
 Grofman, B. 1982, "For single Member Districts Random is Not Equal", In Representation and Redistricting Issues, ed. B. Grofman, A. Lijphart, R.
McKay, H. Scarrow. Lexington, MA: Lexington Books.
 B. Olson, 2008 Redistricter. Software Package. URL: http://code.google.com/p/redistricter/
 C. Puppe,, Attlia Tasnadi, 2009. "Optimal redistricting under geographical constraints: Why “pack and crack” does not work", Economics Letter
105:93-96
 C. Puppe,, Attlia Tasnadi, 2008. "A computational approach to unbiased districting", Mathematical and Computer Modeling 48(9-10), November 2008,
Pages 1455-1460
 F. Ricca, A. Scozzari and B. Simeone, Weighted Voronoi Region Algorithms for Political Districting. Mathematical Computer Modelling forthcoming
(2008).
 F. Ricci, C, Bruno Simeone, 2008, "Local search algorithms for political districting", European Journal of Operational Research189, Issue 3, 16
September 2008, Pages 1409-1426
 T. Shirabe, 2009. District modeling with exact contiguity constraints, Environment and Planning B (35) 1-14
 S. ,Toshihiro and Junichiro Wado. 2008, "Automating the Districting Process: An Experiment Using a Japanese Case Study" in Lisa Handley and
Bernard Grofman (ed.) Redistricting in Comparative Perspective, Oxford University Press
 D.H. Wolpert, Macready, W.G. (1997), "No Free Lunch Theorems for Optimization," IEEE Transactions on Evolutionary Computation 1, 67
 N. Xiao, 2003. Geographical Optimization using Evolutionay Alogroithms, University of Iowa. Dissertation

meets Politics

Questions & Contact

micahaltman.com

meets Politics

Redistricting: When Participative Geography meets Politics

Recomendados

Recomendados

Mais conteúdo relacionado

Semelhante a Redistricting: When Participative Geography meets Politics

Semelhante a Redistricting: When Participative Geography meets Politics (20)

Mais de Micah Altman

Mais de Micah Altman (20)

Último

Último (20)

Redistricting: When Participative Geography meets Politics

Notas do Editor