Craig M. Burnett, Ph.D. Vladimir Kogan, Ph.D. 26 Riggs Place Locust Valley, NY 11560

August 29, 2017

RE: Village of Port Chester, NY, RFQ #2017-02

Village of Port Chester 222 Grace Church Street Port Chester, NY 105753 Attn: Janusz Richards, Village Clerk

Dear Mr. Richards:

We write to you in response to Port Chester's request for qualifications(RFQ #2017-02). As a research team, we have analyzed and ultimately published research on questions relating to local voting issues and representations in some of the most prominent peer-reviewed journals that examine voting in local elections.

Our work begins with the premise that all voting systems entail tradeoffs. Indeed, the selection, implementation, and continued use of any electoral method comes with its own set of pros and cons. What makes us unique, however, is that we can bring an academic perspective to the question of which system might provide the best mix of benefits and costs for the Village of Port Chester, informed by empirical research and our own knowledge of the relevant caselaw.

As you will see from both of our included curriculum vitaes, we are experts in the field of electoral systems, urban, state, and local politics and representation. Our work relies heavily on the sophisticated use of modern statistical analysis and software to examine large-scale voting datasets. This work includes analysis of Hispanic/Latino voters, minority representation, and compliance with relevant federal laws. We have included a few sample publications with our response that are representative of the nature, scope, and sophistication of the work we can provide the village.

As academics, a part of our normal duties is to summarize our analysis in ways that are understandable to a broad audience in a variety of formats (conferences, media inquiries, legal consultation), including elected officials and the legal community. As well-published experts in this field, we are confident we can meet the needs of the village under tight time constraints.

Please find included with this letter our statement of qualifications and references. We look forward to working with the Village of Port Chester.

Best regards.

Craig M. Burnett, Ph.D. Vladimir Kogan, Ph.D. Assistant Professor Assistant Professor Hofstra University The Ohio State University Craig M. Burnett, Ph.D. and Vladimir Kogan, Ph.D. Statement of Qualifications

Resume of Craig M. Burnett, Ph.D,

Craig M. Burnett Department of Political Science Hofstra University Hempstead, NY 11549 805-252-7540 (Cellular) Email: [email protected]

Academic Appointments

Assistant Professor of Political Science, Hofstra University, 2016-present.

Assistant Professor of Political Science, University of North Carolina at Wilmington, 2013-2015

Assistant Professor of Political Science, Appalachian State University, 2010-2013

Education

Ph.D., Political Science, University of California, San Diego, 2010

M.A., Political Science, University of California, San Diego, 2006

B.A., Political Science and History, honors. University of California, Santa Barbara, 2003

Peer-Reviewed Publications

[17]"Parties as an Organizational Force on Nonpartisan City Councils." Forthcoming at Party Politics.

[16]"The Politics of Potholes: Service Quality and Retrospective Voting in Local Elections," with Vladimir K.ogdin. Journal ofPolitics, Volume 79, No. I (2017): 302-314.

[15]"Reconsidering the Construct Validity of Political Knowledge." Critical Review, Volume 28, No. 3-4 (2016): 265-286.

[14]"Exploring the Difference in Participants' Factual Knowledge between Online and In- person Survey Modes." Research and Politics, Volume 3, No. 2(2016): 1-7.

[13]"The Personal Politics of Same-Sex Marriage," with Aaron King. Politics and Policy, Volume 43, No. 4(2015): 586-610.

[12]"Voter Knowledge and Candidates' Judicial Philosophies," with Lydia Tiede. Justice System Journal, Volume 36, No. I (2015): 49-62. [11]"Party Labels and Vote Choice in Judicial Elections," with Lydia Tiede. American Politics Research, Volume 42, No. 2(2015): 232-254.

[10]"When Does Ballot Language Influence Voter Choices? Evidence from a Survey Experiment" with Vladimir Kogan. Cow/;7z/w/co//oA7, Volume 32, No. 1 (2015): 109- 126.

[9]"Ballot (and Voter)'Exhaustion' Under Instant Runoff Voting: An Examination of Four Ranked-Choice Elections," with Vladimir Kogan. Electoral Studies, Volume 37, No. 1 (2015): 41-49.

[8]"Local Logrolling? Examining the Impact of District Elections on Distributive Politics," with Vladimir Kogan. Urban Affairs Review, Volume 50, No. 5 (2014): 648-671.

[7]"Gubernatorial Endorsements and Ballot Measure Approval," with Janine A. Parry. State Politics and Policy Quarterly, Volume 14, No. 2(2014) 178-195.

[6]"Sex and the Ballot Box: Perception of Ballot Measures Regarding Same-Sex Marriage and Abortion in California," with Mathew D. McCubbins. Journal ofPublic Policy, Volume 34, No. 1 (2014), 3-33.

[5]"Gaming Direct Democracy: How Voters' Views of Job Performance Interact with Elite Endorsements of Ballot Measures," with Mathew D. McCubbins. California Journal ofPolitics and Policy, Volume 5, No. 4 (2013), 627-643.

[4]"Does Campaign Spending Help Voters Learn About Ballot Measures?" Electoral Studies, Volume 31, No. 1 (2013), 78-89.

[3]"Familiar Choices: Reconsidering the Institutional Effects of the Direct Initiative," with Vladimir Kogan. State Politics and Policy Quarterly, Volume 12, No. 2(2012), 204-224.

[2]"Do Blacks and Whites See Obama through Race-Tinted Glasses? A Comparison of Obama's and Clinton's Approval Ratings," with Marisa A. Abrajano. Presidential Studies Quarterly, Volume 42, No. 2(2012), 363-75.

[1]"The Dilemma of Direct Democracy," with Elizabeth Garrett and Mathew D. McCubbins. Election Law Journal: Rules, Politics, and Policy, Volume 9, No.4 (2010), 305-24.

Other Publications

[3]"Government and Politics in New York State" Companion Chapter for CQ Press/Sage Publications. Forthcoming Fall 2017. [2]"Marriage on the Ballot: An Analysis of Same-Sex Marriage Referendums in North Carolina, Minnesota, and Washington During the 2012 Elections," with Mathew D. McCubbins. Chapman Law Review, Vo\umt 16, No. 1 (2016): 1-34.

[1]"When Common Wisdom is Neither Common Nor Wisdom: Exploring Voters' Limited Use of Endorsements on three Ballot Measures," with Mathew D. McCubbins. Minnesota Law Review, Volume 97, No. 5 (2013), 1557-1595.

Working Papers

"County Managers as Political Agents," with Christopher Prentice. Revise and Resubmit at Politics and Policy.

"Is Political Knowledge Unique?," with Mathew D. McCubbins. Revise and Resubmit at Political Science Research and Methods.

"DoNonpartisan Ballots Racialize Candidates Evaluations in Low-Information Elections?" with Vladimir Kogan. Under review

Conferences

Annual Meeting of the American Political Science Association (2008, 2010, 2012, 2013, 2014, 2015, 2016, 2017) - Presenter (8x), Discussant(5x), Chair(4x)

Annual Conference on State Politics and Policy (2009, 2011, 2012, 2014, 2015, 2016, 2017) - Presenter (7x), Discussant (5x), Chair(2x)

Annual Meeting of the North Carolina Political Science Association (2014, 2015) - Presenter (2x), Discussant, Chair(2x)

Annual Meeting of the Midwest Political Science Association (2009, 2010, 2011, 2012, 2013) - Presenter (5x), Discussant(2x)

Initiatives and Referendums in the Elections of 2012, University of Southern California, Los Angeles, CA, November 16, 2012 - Presenter

Annual Meeting of the Society for Political Methodology (2012) - Presenter

World Congress of Political Science (2012) - Presenter

Annual Conference on Empirical Legal Studies (2009, 2011) - Presenter (1 x). Discussant(2x) The Past, Present, and Future of Election Law: A Symposium Honoring the Work of Daniel Hays Lowenstein, UCLA School of Law, Los Angeles, CA, January 29, 2010 Presenter

Resume of Vladimir Kogan,Ph.D.

Please see appended curriculum vitae for Vladimir Kogan.

Work Samples

See included copies of relevant peer-reviewed publications. VLADIMIR KOGAN

The Ohio State University Phone (cell); (510) 415-4074 Department of Political Science Phone (office): (614) 292-9498 2140 Derby Hall Fax: (614) 292-1146 154 N. Oval Mall E-mail: kogan.180osu.edu Columbus, Ohio 43210-1373

EDUCATION

Ph.D. in Political Science, University of California, San Diego. 2012 M.A. in Political Science, University of California, San Diego, 2009 B.A. in Political Science, summn ami laude, University of California, San Diego, 2006

ACADEMIC APPOINTMENTS

Department of Political Science. Ohio State University, Columbus, OH

Assistant Professor: August 2012 - Pre-sent

AFFILIATIONS

Faculty Affiliate, Democracy Studies Program, Ohio State University Faculty Affiliate, IVanslational Data Analytics Institute, Ohio State University

RESEARCH

Books Paradise Plundered: Fiscal Crisis and Governance Failures in San Diego. 2011. Palo Alto, CA: Stanford University Press. (With Steven P. Erie and Scott A. MacKenzie)

Journal Articles ".Means, Motives, and Opportunities in the New Preemjition Wars." Fortlicoming. PS: Political Science and Politics.

"Administrative Centralization and Bureaucratic Resfwusiveness: Evidence from the Food Stamp Program." Forthcoming. Journal of Public Administration Research and Theory.

"I^irect Democracy and Administrative Disruption." 2017. Journal of Public Administration Re- .Hcarch and Theory Vol. 27, No. 3: pp. 381-399. (With Stephane Lavertu and Zachary Peskowitz)

"Parties Without Brands? Evidence from California's 1878-79 Constitutional CJonvention." 2017. Studies in American Political Development Vol. 31. No. 1: pp. 68-87. (With Michael Binder)

"Do Anti-Union Policies Increase Inequality? Evidence from State Adoption of Right-to-Work Laws." 2017. State Politics and Policy Quarterly Vol. 17. No. 2: pp. 180-200.

"The Politics of Potholes: Service Quality and Retrospective Voting in Local Elections." 2017. Journal of Politics Vol. 79, No. 1: pp. 302-314. (With Craig M. Burnett)

"Do School Report Cards Produce Accountability Through the Ballot Box?" 2016. Journal of Policy Analysis and Management Vol. 35, No. 3: pp. 639-661. (With Stephane Lavertu and Zachary Peskowitz) "When Voters Pull the Trigger: Can Direct, Democracy Restrain Legislative Excesses?" 2016. Legislative Studies Quarterly Vol. GO, .N'o. 2: pp. 207-325.

"Performance Federalism and Local Democracy: Theory and Evidence from School Tax Referenda." 2016. American Journal of Political Science Vol. 60, No. 2: pp. 418-435. (With Stephane Lavertu and Zachary Peskowitz)

"Pushing the City Limits: Policy Responsiveness in Municipal Government." 2016. Urban Affairs Review Vol. 52, No. 1: pp. 3-32. (With Katherinc Levine Einstein)

"Ballot (and Voter) 'Exhaustion' Under Instant Runoff Voting: An Examination of Four Ranked- Choice Elections." 2015. Electoral Studies Vol. 37: pp. 41-49. (With Craig M. Burnett)

"When Does Ballot Language Influence Voter Choices? Evidence from a Survey Experiment." 2015. Political Communication Vol. 32, .No. 1: pp. 109-126. (With Craig M. Burnett)

"Local Logrolling? As.sessing the Impact of Legislative Districting in Los Angeles." 2014. Urban AJJairs Review Vol. 50, No. 5: pp. 648-671. (With Craig M. Burnett)

"Mobilizing Latino Voters: The Impact of Language and Co-Ethnic Policy Leadership." 2014. Amer ican Politics Research Vol. 42, No. 4: pp. 677-699. (With Mike Binder, Thad Kousser, and Costas PanagopoLilos)

"Familiar Choices: Reconsidering the Institutional Effects of the Direct Initiative." 2012. State Politics and Policy Quarterly Vol. 12, No. 2: pj). 204-224. (With Craig M. Burnett)

"Redistricting California: An Evaluation of the Citizens Commi.ssion Final Plans," 2012. California Journal of Politics and Policy Vol. 4, No. 1. (With Eric McGhee)

"The Irony of Comprehensive State Const itntional Reform." 2010. Rutgers Law Journal Vol. 41, No. 4: pp. 881-906.

"Changing Tracks? The Prospect for Clalifornia Pension Reform." 2010. California Journal of Politics and Policy Vol. 2, No. 3. (With .Mathew D. McCubbins)

"Redevelopment, San Diego Style: The Limits of Public-Private Partnerships." 2010. Urban Affairs Review Vol. 45, No. 5: pp. 644-678. (With Steven P. Erie and Scott A. MacKenzie)

"Lessons from Recent State Constitutional Clonventions." 2010. California Journal of Politics and Policy Vol. 2, No. 2.

"The Problem with Being Special: Democratic Values and Special Assessments." 2009. Public Woi'ks Management & Policy Vol. 14, No. 1: pp. 4-36. (With Mathew D. McCubbins)

Book Chapters "Redistricting and Representation: Searching for 'Fairness' Between the Lines." 2017. In Changing Ilow America Votes, edited by Todd Donovan. Lanham, MD: Rowman k. Littlefield Publishers. (With Eric McGhee)

"Machine Bosses, Reformers, and the Politics of Ethnic and Minority Incorporation." 2016. In Oxford Handbook of the Histouy of Arneiicxm Immigration and Ethnicity, edited by Ron Bayor. New York. NY: Oxford University Press. (With Steven P. Eric)

"Paradise Regained? Nonparti.san Appeals and Special Election Rules in San Diego's 2013-14 May oral Race." 2015. In Local Politics and Mayoral Elections in 21si Century America: Keys to City Hall, edited by Sean D. Foreman and Marcia L. Godwin. New York, NY: Routledge. (With Steven p. Erie, Nazita Lajevardi, and Scott A. MacKenzie)

"Causes of Fiscal Crises in State and Local Government." 2015. In Emerging Trends in the Social and Behavioral Sciences, edited by Robert Scott and Stephen Kosslyn. Thousand Oaks, CA: SAGE Publications.

"From Machines to Service Centers: The Evolution of State and Local Political Parties." 2014. In CQ Guide to Political Parties, edited by Marjorie Randon Hershey, Barry C. Burden, and Christina Wolbrecht. Washington, DC: CQ Press.

"Redistricting: Did Radical Reform Produce Different Results?" 2013. In Governing California: Politics, Government, and Public Policy in the Golden State, 3rd Edition, edited by Ethan Rarick. Berkeley, CA: Berkeley Public Policy Press. (With Eric McGhee)

"Planning L.A.: The New Politics of Neighborhood Development and Downtown Revitalization." 2013. In New York and Los Angeles: The Uncertain Future, edited by David Halle and Andrew Beveridge. New York, NY: Oxford University Press. (With Andrew Deener, Steven P. Erie, and Forrest Stuart)

"Great Expectations and the California Citizens Redistricting Commission." 2011. In Reappor- tionment and Redistricting in the West, edited by Gary Moncrief. Lanharn, MD: Lexington Books. (With Thad Kousser)

"How G.A.V.E.L. Changed Party Politics in Colorado's General Assembly." 2011. In State of Change: Colorado Politics in the Twenty-First Century, edited by John Straayer, Robert Duffy, and Courtney Daum. Boulder, CO: University Press of Colorado. (With Mike Binder and Thad Kousser)

Other Publications Book review of John J. Betancur and Janet L. Smith, Claiming Neighborhood: New Ways of Under standing Urban Change (Urbana, IL: University of Illinois Press, 2016); Susan J. Popkin, No Simple Solutions: Transforming Public Housing in Chicago (Lanham, MD: Rowman & Littlefield, 2016); and Costas Spirou and Dennis R. Judd, B^iilding the City of Spectacle: Mayor Richard M. Daley and the Remaking of Chicago (Ithaca, NY: Cornell University Press, 2016). Forthcoming. Perspectives on Politics.

Book review of David J. Karjanen, The Servant Class City: Urban Revitalization Versus the Working Poor in San Diego (Minneapolis, MN: University of Minnesota Press, 2016). Forthcoming. Journal of Urban Affairs.

Book review of Parke TVoutman, Cone Awry: The Smart Growth Quest to Stop Urban Sprawl (San Diego, CA: Three Acres Press, 2015). Forthcoming. Journal of Urban Affairs.

Book review of Sean Nicholson-Crotty, Governors, Grants, and Elections: Fiscal Federalism in the American States (Baltimore, MD, Johns Hopkins University Press, 2015). 2017. Political Science Quarterly Vol. 132, No. 1: pp. 171-173.

Work in Progress "Election Timing and the Composition of the Electorate." Invited to revise and resubmit at Amer ican Journal of Political Science. (With Stephane Lavertu and Zachary Peskowitz)

"Do Nonpartisan Ballots Racialize Candidate Evaluations? Evidence from Who Said What? Exper iments." Under review. (With Craig M. Burnett)

"Electoral Consequences of Welfare State Expansion: The Case of the Food Stamp Program." Under GRANTS AND FELLOWSHIPS

Marnette and Paul Perry Fellow, 2014-2016 Co-Principal Investigator (with Stephane Lavertu and Zachary Peskowitz),"Educational Governance and Accountability Project," Lyle Spencer Research Award, Spencer Foundation, August 2015 ($632,778.90) Principal Investigator,"The Consequences of Nonpartisan Elections," Democracy Studies Program, The Ohio State University, December 2014 ($5,484) Research Grant, College of Arts and Sciences Small Grant Program, The Ohio State University, October 2014 ($2,000) Co-Principal Investigator (with Stephane Lavertu and Zachary Peskowitz),"Ohio School District Democracy Project," Democracy Studies Program, The Ohio State University, April 2013($17,584.65) Time-Sharing Experiments for the Social Sciences Research Grant, 2010 Pre-Doctoral Research Fellowship, Bill Lane Center for the American West, Stanford University, 2009-2010

AWARDS

Political Research Quarterly Outstanding Reviewer Award, 2014

APSA Urban Politics Section Best Book Award for Paradise Plundered: Fiscal Crisis and Gover nance Failures in San Diego (with Steven P. Erie and Scott A. MacKenzie), 2012 APSA Urban Politics Section Best Paper Award for "Who Benefits from Jurisdictional Competi tion?", 2011

APSA Urban Politics Section Stone Scholar Award, 2010 Urban Affairs Association Alma H. Young Emerging Scholar Award, 2010

SERVICE

Editorial Board Member, Urban Affairs Review: 2015 - Present Section Co-Chair, Conference Program Committee (Impact of Politics on the Policy Process Section, Association for Public Policy Analysis and Management); 2016, 2017 Second Year Transformation Program Faculty Mentor (Ohio State University): 2016, 2017 Proposal Reviewer, Translational Data Analytics Institute Seed Grant (Ohio State University): Spring 2017 Member, Best Paper Award Committee (Urban Politics Section, American Political Science Associ ation): 2016 Democracy Studies Director Search Committee (Department of Political Science, Ohio State Uni versity): 2015-2016 Member, State Politics and Policy Quarterly Best Paper Award Committee (State Politics and Pol icy Section, American Political Science Association): 2015 Chair, Best Journal of Urban Affairs Article Award Committee (Urban Affairs Association): 2015 Undergraduate Studies Committee (Department of Political Science, Ohio State University): 2012 - Present American Politics Search Committee (Department of Political Science, Ohio State University): 2014 Undergraduate Awards Committee (Department of Political Science, Ohio State University): 2013, 2015, 2016, 2017 Alma H. Young Emerging Scholar Award Committee (Urban Affairs Association): 2011 Dissertation Committees: William Massengill (current) Faculty Advisor, Multi-Partisan Coalition, Ohio State University: 2015 - Present Discussant: Midwest Political Science Association (2012, 2013, 2014, 2017), Southern Political Sci ence Association (2016), American Political Science Association (2014), State Politics and Policy Conference (2016, 2017), Association for Public Policy Analysis and Management (2016) Referee: American Political Science Review, American Journal of Political Science, Journal of Pol itics, British Journal of Political Science, Political Research Quarterly, American Politics Research, Governance, Journal of Policy History, Journal of Public Policy, Legislative Studies Quarterly, MIT Press, Political Behavior, Political Communication, Public Administration Review, Public Manage ment Review, Public Works Management and Policy, PS: Political Science and Politics, Publius, Rowman & Littlefield, State Politics and Policy Quarterly, Urban Affairs Review, Urban Studies

Teaching Experience

Instructor Political Science 7125 (OSU), Federalism (Graduate Seminar)

Political Science 3115 (OSU), Introduction to Policy Processes

Political Science 4125 (OSU), American State Politics

Political Science 4127 (OSU), Governing Urban America

Political Science 4130 (OSU), Law & Politics

Political Science 4920 (OSU), Politics in Film and Television

Arts and Sciences 1137 (OSU), Political Courage (First-Year Seminar)

Political Science 162 (UCSD), Environmental Policy

Teaching Assistant

Political Science 10 (UCSD), Introduction to American Politics

Political Science 13 (UCSD), Power and Justice

Political Science 30 (UCSD), Political Inquiry and Analysis

Political Science 225E (Stanford), The Law of Politics

Grader

Political Science lOOM (UCSD), Political Psychology

Political Science 104M (UCSD), Law and Sex

Political Science 160AA (UCSD), Introduction to Policy Analysis References for Craig M. Burnett, Ph.D. and Vladimir Kogan,Ph.D.

Kevin J. Minnick, Skadden, Arps, Slate, Meagher & Flom LLP [email protected] 213-687-5272 300 South Grand Avenue, Los Angeles, CA 90071

Project: Legal Challenge of California's Top-Two Primary (Burnett) Dates: October 1, 2015 to January 31, 2016

Eric McGhee, Public Policy Institute of California [email protected] 415-291-4439 500 Washington Street, Suite 600, San Francisco, California 941

Project: Redistricting California(Kogan) Dates: July 1, 2012 to Dec. 31, 2013

Stephane Lavertu, Glenn College of Public Affairs, Ohio State University [email protected] 614-247-4060 1810 College Road, Columbus, Ohio 43210

Project: Education Governance and Accountability Project(Kogan) Dates: Oct. 1, 2015 to Present

THE CALIFORNIA Journal of Politics & Policy

Volume 4, issue 1 2012

Redistricting California: An Evaluation of the Citizens Commission Final Plans

Vladimir Kogan University of California, San Diego Eric McGhee Public Policy Institute of California

Abstract

For the first time in California history, a carefully vetted commission of citizens has overseen the delicate task of redrawing the state's political boundaries. By ana lyzing the maps produced by the commission and comparing these plans to the redistricting overseen by the legislature a decade earlier, we show that the new pro cess has produced important improvements in terms of both the criteria voters said they cared about and the representational implications of interest to academics and political observers. In many respects, however, the magnitude of these gains has fallen short of what many political reformers may have hoped for. Perhaps the most important lesson from the 2011 round of redistricting is that a fair process, no mat ter how nonpartisan and participatory, cannot avoid the reality that any redistricting scheme produces both political winners and losers.

Keywords: redistricting, redistricting reform, California politics, political polar ization Copyright £) 2012 Dc Gruyter. All rights rcscr\ ed. w\v\v.degruy tc r.com/v i e\v/J/cj pp Kogan and McGhee: Redislricting California: PA'aiuating the Citizens Commission Final Plans

THE CALIFORNIA Redistricting California: Journal of An Evaluation of the Citizens Politics & Policy Commission Final Plans

Vladimir Kogan University of California, San Diego Eric McGhee Public Policy Institute of California

Introduction

California's recently completed redistricting has been a sharp break with the past, both in style and substance. For the first time in state history, the decision about the location of district lines has been made by a commission of citizens from outside of government, rather than elected officials or court-appointed special mas ters, in a process that has been significantly more participatory and transparent than in the past.' This new commission has not wanted for critics. Many observers, including members of the Republican Party and some minority groups, have loudly complained about the process and the commission that has overseen it. Indeed, Re publicans have sought to overturn the congressional and state Senate maps in court, and have submitted signatures for a referendum to overturn the state Senate maps and throw that process to the courts as well. In this paper, we help place these criticisms in historical context and offer a range of important information about the maps the commission produced. We dis cuss the political forces that led to the creation of the Citizens Redistricting Com mission (CRC) and show how the process overseen by the new commission has differed from redistricting of the past four decades. We then evaluate the newly adopted plans, both on the specific criteria the CRC was mandated to consider and on the political implications the CRC was required to ignore. On most of the mandated criteria—such as maximizing compactness and avoiding city and county splits—the commission's plans represent an improvement over the maps drawn by the legislature in 2001. The maps are also somewhat more likely than the current plans to produce competitive races and to elect Democrats to office, though the lat ter effect is only notable for the congressional plan. In short, although the plans are far from perfect on any single dimension, they look like the sort of result one could expect given the mandates and constraints faced by the commission.

DOI: 10.1515/1944-4370.1197 I California Journal of Politics and Policy. Vol. 4[2012 j, Iss. I, Art. 2

California's Redistricting Wars and the Creation of the Citizens Commission

The creation of California's Citizens Redistricting Commission through the pas sage of Proposition 11 in November of 2008 represented an important shift in the redistricting battles that have been waged between the state's two largest political parties for decades (see Kousser 1997, 1998; Quinn 1981). Although it is common to say that the commission has taken redistricting power away from legislators, California's redistricting process has not historically been the exclusive purview of the legislature. Instead, the process has involved a series of actors who have used the courts and direct democracy to influence the outcome, making each redistricting a protracted battle with significant input from actors outside the elected branches of government. The courts have played a direct role in drawing lines in almost halfthe California redistricting battles since the U.S. Supreme Court's one-person, one-vote decisions began the modem era of redistricting in the 1960s. In both 197land 1991, vetoes of Democratic redistricting plans by Republican govemors handed the matter to the state Supreme Court, which appointed "Special Masters" to craft the new boundar ies. These masters then created plans that, as demonstrated by both qualitative and quantitative analysis, were more favorable for the party of the governor who had appointed a majority of state Supreme Court justices (Kousser 1997, 2006). Even when members of one party have controlled both the legislature and the governorship in California, they have not exerted unchecked power over how to draw the lines. In the early 1980s, the Democrats controlled all levers of govern ment but saw their redistricting plan defeated by a set of referendums—one each for the congressional, state Senate, and state Assembly lines—placed on the bal lot by Republicans. Republicans then allied with good govemment reformers on a follow-on proposition to establish a redistricting commission. When this measure lost, Democratic legislators rushed to redraw the lines, passing a compromise plan before lame duck Democratic Gov. Jerry Brown had left office and his Republican successor George Deukmejian could been sworn in. The plan was generous enough to Republicans to garner a two-thirds vote in each house, allowing it to go into ef fect swiftly and avoid another referendum challenge (Kousser 1997). Those lines stayed in place for the remainder of the decade,' but a clear lesson from the episode was that direct democracy—whether real or threatened—would constrain the dis tricts legislators could draw. This contentious history—and the successful referendums in particular—pro vided the backdrop as California prepared for the 2001 round of redistricting. The Democrats agreed to a bipartisan gerrymander that protected incumbents from both parties with safe districts that minimized partisan turnover, and they cemented the deal with a bipartisan two-thirds majority that exempted the plans from any refer- Kogan and McGhee: Rcdistricting California: Evaluating the Citizens Commission Final Plans endum challenges. The resulting plans angered many outside the legislature. Some minority group leaders argued the maps missed an opportunity to increase rep resentation for historically excluded groups (Ingram 2001b). Political reformers and newspaper columnists considered the plans a corrupt bargain drafted behind closed doors that put the interests of elected officials ahead of the voters they were supposed to represent. Some political observers also argued that, by creating safe political seats in which one party controlled an overwhelming majority of the votes, the maps exacerbated the partisanship and polarization that had created perennial delays in the adoption of the state budget(see, e.g., Skelton 2009; although see Mc- Carty, Poole, and Rosenthal 2009, and McGhee 2008 for evidence to the contrary). This popular displeasure with the 2001 maps provided fuel for the campaign behind Proposition 11. As with previous failed reform initiatives. Proposition 11 was funded primarily by Republican interests—nonetheless, its advocates made a concerted effort to broaden their coalition. They attracted endorsements from sev eral high-profile Democrats, and attempted to incorporate concerns from minority voting rights groups. These efforts paid off on Election Day 2008, when Proposi tion 11 squeaked through by the barest of margins. A follow-on measure. Proposi tion 20, then passed in November 2010, extending the commission's authority to congressional lines as well.

The Citizens Redistricting Commission

Propositions 11 and 20 created a 14-member Citizens Redistricting Commis sion to oversee the process. CRC members were chosen through a complicated process designed to purge the applicant pool of any connection with the elected of ficials whose districts would be redrawn. The commission also included represen tation from every major ethnic group, and the law mandated five Democratic, five Republican, and four decline-to-state or third-party commissioners for partisan bal ance. The rules laid out in Proposition 11 required at least some members from each partisan delegation to approve the final maps, ensuring that the commission would either act in a bipartisan manner or deadlock trying. In the event of a deadlock, the new law required the state Supreme Court to appoint a panel of special masters to draw the districts as in the 1970 and 1990 redistricting cycles. Thus, there was no scenario under the new system for the legislature to be directly involved in crafting the maps. Proposition 11 also added new language to the state constitution listing specific criteria, in order of importance, that the CRC was to use to craft new political dis tricts. In many ways, these criteria represented a backlash against perceived abuses of the 2001 process, including districts that split cities and created jagged edges and long protrusions that, to outside observers, appeared to serve few purposes

Published by Dc Gruyter. 2012 3 California Journal of Politics and Policy. Vol. 4[2012}, Iss. 1.Art. 2

Other than creating constituencies that advanced the political interests of elected incumbents.^ Table 1 below lists the formal redistricting criteria adopted by voters as part of Proposition 11. Aside from the equal population, district contiguity, and Voting Rights Act requirements already explicitly or implicitly a part of California law, the measure added several other constraints that had historically been less central to de bates about redistricting in California. First, Proposition 11 explicitly required the commission to protect the "geographic integrity" of cities, counties, neighborhoods, and other very broadly defined "communities of interest," cautioning line-drawers against splitting communities into multiple districts.'' Second, the new language mandated that districts be drawn to maximize geographic compactness—although only to the extent that this did not conflict with other higher criteria—to avoid the creation of sprawling and irregularly shaped districts. Despite the existence of many different measures used to assess compactness in the context of redistricting (see, e.g., Niemi et al. 1990), the law left it to the commission to define and measure the concept. Finally, the new law adopted the principle of"nesting," calling for the incorporation of two state Assembly districts within a single Senate district. Although all six criteria included in Proposition 11 draw on widely accepted redistricting considerations, there are inherent tensions between them. Maximizing either geographic compactness or nesting often hurts minority representation, and nesting can split city and county boundaries as well (Barabas and Jerit 2004; Cain and Mac Donald 2007). Yet aside from ranking the criteria in order of importance, Proposition 11 provided little guidance on how to resolve these conflicts, leaving the difficult task of reconciling them to the commission. Equally important are the criteria specifically excluded from the measure, such as increasing political competition. Although establishing boundaries for the pur pose of creating close elections was not one of the goals included in Proposition 11—indeed, the language of Proposition 11 actually prohibited the maps from be ing drawn "for the purpose of favoring or discriminating against an incumbent"— the promise of political competition was one of the central arguments made by the measure's proponents. The growing geographic segregation of California's elector ate has made it more difficult to draw such districts by accident(Cain, Mac Donald, and Hui 2008), and the Voting Rights Act requirements help ensure that a certain number of uncompetitive districts have to be created. Once again, the commission faced a difficult balancing act: reconciling voter expectations that redistricting re form would lead to dramatically more competitive elections with the actual criteria and constraints written into the measure that made drawing such districts more difficult. Kogan and McGhee: Redistricling California: Evaluating the Citizens Commission Final Plans

Table 1. Official California Redistricting Criteria

Priority Criteria 1 Districts shall have reasonably equal population. 2 Districts shall comply with federal Voting Rights Act. 3 Districts shall be geographically contiguous. 4 To the extent possible, geographic integrity of any city, county, city and county, neighborhood, or community of interest shall be respected. 5 To the extent possible, districts shall be drawn to encourage geographical compactness. 6 To the extent possible. Senate districts shall be comprised of two whole, complete, and adjacent Assembly districts, and each Board of Equaliza tion district shall be comprised of 10 whole, complete, and adjacent Sen ate districts.

Source: California State Constitution Article XXI. Section 2

Evaluating the Commission Plans: Mandated Criteria

In this section, we assess how well the CRC's maps adhere to the formal redis tricting criteria adopted by voters. The inherent tensions among these criteria make it extremely difficult to determine whether the maps drawn by the commission represent the "best" possible set of plans that could have been created. However, by comparing the commission's work to the maps drawn by the legislature a decade earlier, we can quantify the extent to which the new district boundaries represent an improvement on the 2001 plans. We also contrast the final maps adopted by the commission with the draff boundaries released in June 2011. This comparison is important for two reasons. First, the June maps generally attracted significant praise from political observers, including some notable Republicans.^ This reception contrasted sharply with the polarized partisan response to the final maps adopted in August, which were almost universally assailed by Republican leaders and strategists, as well as some minority rights advocates. By identifying the differences between the June and August maps, we can assess whether criticisms of the latter were motivated by legitimate concern over the redistricting criteria put into place by Proposition 11 rather than narrow partisan interests that may have emerged as the political implications of the new maps became clear. Second, by tracking changes made to the June maps, we can identify the tradeoffs made by the commission in response to public input.

Published by De Gruyter, 2012 California Journal of Politics and Policy. Vol. 412012], Iss. I, Art. 2

Minority Representation

Proposition 11 differed from earlier redistricting reform initiatives in the extent to which the measure sought to reassure advocates of minority rights that their in terests would be respected (Kogan and Kousser 20II). One such reassurance was language calling for compliance with the federal Voting Rights Act. In Thornburg V. Gingles(1986), the Supreme Court interpreted Section 2 of the law to require the creation of majority-minority districts when minority populations are sufficiently large and geographically concentrated and when there is evidence of "racially po larized" voting. Even where the Gingles preconditions may not be satisfied, other parts of Proposition 11 that task the commission with preserving the geographic integrity of local "communities of interest" provide additional protections, since minority groups undoubtedly represent populations that share "common social and economic interests"(California Constitution, Article XXI,§ 2, subdivision (d)(4)). To gauge the implications of the various redistricting schemes for minority rep resentation, we calculated the proportion of each district's citizen voting-age popu lation (CVAP) that is Asian, black, and Latino, the three groups identified by the CRC as potentially worthy of VRA protections under the Gingles criteria (Citizens Redistricting Commission 2001, 20). We focus on CVAP because this number best captures the universe of eligible—rather than actual or registered—voters in each district.^ For 20II, we rely on census block-level CVAP figures from the Statewide Database; for 2001, we use district-level measures calculated using a special tabu lation from the 2000 census data. Although our discussion focuses on point esti mates, it is important to recognize that the CVAP figures are inexact because they are constructed, at least in part, from survey data. Unfortunately, due to a variety of methodological challenges discussed at length by McCue (2011; see, in particular, pp. 16-17), it is not possible to construct confidence intervals or margins of error that adequately capture this uncertainty. These data limitations should be kept in mind when interpreting the results below. Table 2 lists the estimated number of majority-minority districts drawn under each set of redistricting plans. Overall, the 2011 maps adopted by the commission resulted in the creation ofeight new majority Latino districts, with most of the gains coming from the Assembly plan. One notable detail is that the bulk of the gains appeared only in the final commission plans, with the draft maps released in June 2011 creating just one net increase in majority Latino districts, a fact that helps explain the initial criticism offered by some civil and voting rights groups after the release of the June maps. The jump in the number of majority Latino districts be tween June and August suggests that the gains for Latinos were produced primarily through a deliberate effort on the part of the commission to increase representation for this group, rather than as a result of natural growth in the relative size of the La- Kogan and McGhee: Redistricting California: Evaluating the Citizens Commission Final Plans

Table 2. Number of Majority-Minority Districts (by CVAP)

2001 Plan 2011 Draft Plan 2011 Final Plan Assembly Black 0 0 0 Latino 8 10 14 Asian 0 0 1 Senate Black 0 0 0 Latino 4 4 5 Asian 0 0 0 Congress Black 0 0 0 Latino 6 5 7 Asian 0 0 0 tine population in California between 2000 and 2010. In addition, the commission created one new majority-Asian district in the San Gabriel Valley—an area that had been split up among several districts in 2001 in the face of strong opposition from Asian voting rights groups (Ingram 2001a). Given the sizeable increase in the number of Latino Califomians over the past decade—from 32.4 percent of state residents in 2000 to 37.6 percent in the most recent census—it may be surprising that Latinos did not see greater gains in repre sentation, particularly in the state Senate and congressional maps. This puzzle may be explained in part by the smaller share of the Latino population that is eligible to vote compared to whites—due both to lower rates of citizenship and a higher proportion of Latinos falling below the legal voting age (see Arvizu and Garcia 1996 for an overview)—as well as lower turnout rates even among Latinos who are eligible to vote. As a percent of CVAP, the Latino population increased from 22.5 percent in 2000 to 25.6 percent in 2010.^ Moreover, Latino growth has not been limited to heavily Latino areas, leading in many parts of the state to large Latino gains that still fall short of producing a Latino voting majority (California Depart ment of Finance 2011). Consistent with this idea, the number of districts where La tinos make up 30 to 40 percent of the voter-eligible population has increased even as the number of majority-Latino districts has shown much smaller change (see Figure I, Panel b). Latino candidates might succeed in these "influence" districts by attracting votes from other minorities or cross-over support among white voters.

Published by De Gruyter. 2012 Figure 1. Minority CVAP by District:(a) Black Representation

Black Representation ^ack Representation ^ack Representation Assembly Senate Congress

a S.

Kj o

T 1 1 1 1 1 r

10 20 30 40 50 60 10 20 30 40 50 60 0 10 20 30 40 50 60

Blac k CVAP percentage Black CVAP Percentage Blac k CVAP Ftercentage

2001 LegislaCve Plan '— 2011 Draft Qjrrmlsslon Han 2011 Rnal Ccmrissbn Ran Figure 1. Minority CVAP by District:(b) Latino Representation 2^ o (TO 03 3 Latino Representation Latino Representation Latino Representation Assembly Senate Congress o o 3-

cro n

n

n o T 1 1 1 1 1 r 3 3 0 10 20 30 40 50 60 10 20 30 20 30 40 50

Latino CVAP Percentage Latho CVAP Ftercentage Latno CVAP percentage

2001 Legislatlve Plan — 2011 CfraJt Commission Ran — 2011 Rnal Commission Ran 3. ■3 Figure 1. Minority CVAP by District:(c) Asian Representation

Asian Representation Asian Representation Asian Representation Assembly Senate Congress

& S.

-| 1 1 1 r

0 10 20 30 40 50 60 10 20 30 40 50 60 10 20 30 40 50 60

Asian CVAP Percentage Asian CVAP Ftercentage Asian CVAP Percentage

2001 Legislative Plan 2011 Draft (Donmission Ran 2011 Rnal Convnissbn Ran KJ

Note: The Y-axis in these and other figures presented below display the "probability density"—or roughly the proportion of districts in the data corresponding to each value on the X-axis. The density does not correspond to any absolute measure (such as number of voters) because, by construction, the total area under each density curve must add up to one. As a result, the actual value ofon the Y-axis will depend on the units used for the X-axis. However, it is useful to compare such density curves because they provide an intuitive visual representation of the full distribution, which is often more informative than a comparison of district means or medians. Kogan and McGhee: Rcdistricting California: Evaluating the Citizens Commission Einai Plans

City, County, and Community Splits

In addition to recognizing the representational aspirations of minority groups, Proposition 11 also directed the CRC to minimize division of cities, counties, neighborhoods, and other "communities of interest." Indeed, one of the reasons the commission held dozens of hearings across the state was to solicit input from local residents to help commissioners identify such communities. As an empirical matter, it is difficult for outside observers to identify the geographic boundaries of established of communities of interest, and this task is made even more challenging by the rather vague and open-ended definition for the term added to the constitution by Proposition 20. Although legal scholars have proposed geospatial methods for identifying communities of interest by comparing the "homogeneity" of districts on various socioeconomic dimensions (Stephanopoulos 2012), this method is not designed for making comparisons among redistricting plans over time. Due in large part to such measurement and definitional challenges, we do not attempt to evaluate the plans on this dimension. Our focus instead is on two types of political com munities with clearly identifiable geographical boundaries: incorporated cities and counties. For each type of geography, we calculated how many jurisdictions were left intact as part of the redistricting process—that is, the proportion of communi ties that were drawn into a single political district. In Table 3, we report the percentage of intact incorporated cities produced by each set of plans. This calculation focuses only on communities small enough to fit into a single district and thus excludes major cities that would have been split un der any redistricting plan.^ The table also tallies up the total number of community splits, including geographies too large to fit in a legislative district. Table 4 presents a similar analysis for county splits, except instead of percentages, this table reports the raw counts of intact counties.'^ The tables suggest that the 2011 commission plans represented a modest im provement on the 2001 legislative redistricting in preserving the integrity of exist ing communities. Despite criticism from one political observer that the commis sion's work "whacks and hacks cities and counties for no apparent reason"(Quinn 201 lb), between 91 and 96 percent of all incorporated cities were kept intact under the 2011 plans. For both the Assembly and Senate maps, the number of intact cities increased slightly compared to 2001, although the 2011 congressional plan actually increased the number of city splits compared to the maps adopted by the legisla ture a decade earlier. The primary difference between the two sets of maps is what happened to split cities. Compared to a decade earlier, more of these divided cities were contained in only two districts in 2011, thus reducing the overall number of city splits. If we expand the analysis to include unincorporated census-designated places, which are well-defined areas with clear community character, the improve-

Published by De Gruyter. 2012 II California Journal of Politics and Policy. Vol. 412012}, Iss. I.Art. 2

Table 3. Comparison of City Splits

2001 Plan 2011 Draft Plan 2011 Final Plan

No, of No. of No. of Community % Community % Community % Plan Type Splits Intact* Splits Intacta Splits Intacta Assembly 97 98% 62 92% 62 92% Senate 41 96% 37 94% 33 96% Congress 53 94% 73 88% 59 91%

Note: We use a list ofincorporated cities available from Census CIS shapefiles. The number of incorporated cities increased from 452 in 2000 to 459 in 2010. "Percent intact calculated only for communities with a small enough population to fit in a single legislative district.

Table 4. Comparison in County Splits

2001 Plan 2011 Draft Plan 2011 Final Plan

No. of No. of No. of Community Intact Community Intact Community Intact Plan Type Splits Counties Splits Counties Splits Counties Assembly 99 31 93 31 92 30 Senate 51 33 59 32 54 38 Congress 62 34 69 31 67 33 ment between the 2001 and 2011 maps is somewhat more pronounced.'" However, aside from the Senate plans, where the 2011 maps increased the number of intact counties, the commission did not notably increase or reduce county splits.

Maximizing Geographic Compactness

Compactness—the spatial "spread" ofa district—is a long-established criterion for evaluating political boundaries, and a legally mandated redistricting consider ation in many states (Niemi et al.l996). Indeed, sprawling districts are often used by critics to make the case for redistricting reform. In California, for example, proponents of Proposition 11 often pointed to California's 23''' Congressional Dis trict drawn by the legislature in 2001. Known as the "Ribbon of Shame," the costal district stretched nearly 200 miles from Monterey County in the north to Oxnard in the south.

12 Kogan and McGhee: Redislricting California: Evaluating the Citizens Commission Final Plans

Despite the straightforward intuition for what it means to draw compact dis tricts, consensus on a formal mathematical definition has remained elusive, and scholars have proposed a variety of measures. Niemi et al. (1996) showed that ag gregate statistics produced by these measures generally lead to similar conclusions about specific redistricting plans, so the precise measure we choose is not likely to affect our conclusions. We use the measure proposed by Polsby and Popper(1991): It ranges from zero to one, with higher values indicating greater compactness." To compare the degree of compactness achieved by the 2001 legislative redis tricting and the two sets of commission-drawn plans. Table 5 presents the mean district Polsby-Popper score produced by each plan. Figure 2 also plots the dis tribution of district compactness scores for each set of maps. Overall, the larger Polsby-Popper scores for the CRC-drawn districts provide strong evidence that the commission drew notably more compact boundaries than the legislature. This was the case for both sets of commission plans, although the August maps achieved somewhat greater compactness than the initial draft maps released in June.

Nesting Assembly Districts

The final criterion we evaluate is "nesting": the process of fitting two Assembly districts within a single Senate district. One impetus for nesting is the belief that doing so makes it easier for voters to correctly identify their elected representatives and accurately apportion credit and blame for policy outcomes. Some research has also shown that nesting can improve legislative productivity in bicameral systems by aligning electoral incentives of lawmakers and thus making it easier to build legislative coalitions across chambers (Chen 2010). In fact, the Special Masters adopted perfectly nested plans in both the 1970s and the 1990s (Special Master 1991,40). Although nesting of lower house districts was approved by voters as part of Proposition 11, this goal was given the lowest priority among the criteria listed in the constitution. Table 6 compares the degree of nesting across the commission and legislative plans. For each Senate district, we calculated the percentage of its population that was contained in the two Assembly districts with the largest number of overlapping residents. In a perfectly nested plan, the percentage would be 100—because the two Assembly districts would contain all of the residents of the larger Senate district. The first row of Table 6 reports the average amount of population overlap between each Senate district and its two most nested Assembly districts. The second row reports the average number of Assembly districts covered, at least partially, by each Senate district. A perfectly nested plan would have a value of2 on this measure, and higher values indicate weaker nesting.

Published by De Gruyler, 2012 13 California Journal ofPolitics and Policy, Vol. 412012], Iss. I, Art. 2

Table 5. Average Polsby-Popper Compactness Scores Plan Type 2001 Plan Draft 2011 Plan Final 2011 Plan Assembly 0.20 0.22 0.22 Senate 0.12 0.17 0.19 Congress 0.13 0.19 0.19

As the table makes clear, the draft plan released in June 2011 achieved the greatest amount of nesting, with nearly perfect population overlap between the dis tricts of the upper and lower houses of the legislature. The amount of nesting was reduced in the final plan adopted by the commission, in large part due to the cre ation of new majority-minority districts in response to public input and in an effort to reduce the number of community splits in the draft plan (Citizens Redistricting Commission 2011). However, the final commission maps still produced more over lap between the Senate and Assembly districts than the redistricting plan adopted by the legislature in 2001.

Summary

Overall, the results reported here make clear that the final commission maps adopted in August represent a notable improvement over the 2001 legislative redis tricting on nearly all of the criteria voters said were important to them when they passed Proposition 11. This is true despite the fact that the commission did not produce lines that fully achieve any of the goals written into the state constitution. Though slightly more communities have been left intact in 2011 than in 2001, a small number of cities are still split into multiple districts under the CRC-approved plan. Likewise, while the new districts are more compact than the old ones, none of them look like perfect rectangles or circles. The fact that the commission did not achieve perfect compliance with all of the criteria reveals the inherent tensions between these important considerations — tensions that limit the potential gains from the reform and which few voters likely understood when they voted for the new system. Equally surprising, however, is that the apparent conflicts between the redis tricting criteria were less pronounced than many redistricting scholars may have expected. This is best demonstrated by comparing the draft maps released by the commission in June to the final maps adopted in August. The changes made to the plans in the meantime resulted in improvements on nearly all of the redistricting criteria. Only nesting, the lowest-ranked priority under Proposition 11, declined in the final maps — and this likely occurred because the commissioners worked to

14 Kogan and McGhee: Redislricting California: Evaluating the Citizens Commission Final Plans

Figure 2. Comparison of Compactness Across Plans

Assembly

0.0 0.1 0.2 0.3 0.4 0.5 0.6

Pdsby-Popper ODmpactness

State Senate

0.0 01 02 03 04 0 5 0 6

Pdsby-Popper (Dompactness

Congress

to c a

0.0 0.1 0.2 0 3 0 4 0.5 0.6

Polsby-Popper Conpactness

2001 Ran * 2011 Draft Ran — 2011 Rnal Ran

Published by De Gruyter. 2012 15 California Journal of Politics and Policy, Vol. 412012], Iss. I, Art. 2

Table 6."Nesting" of Assembly Districts Inside Senate Districts

2001 2011 2011 Plan Draft Plan Final Plan Senate District Population in 74% 96% 83% Two Largest Assembly Districts (Average)

Average Number of Assembly 6.35 2.95 4.95 Districts per Senate District maximize other higher-ranked criteria, as directed by the state constitution. The criteria were in tension with each other, but they were not mutually exclusive.

Evaluating the Commission Plans: Political Effects While the law explicitly required the CRC to consider the impact of its maps on geographic and minority representation, it also specifically directed the CRC to ignore the partisan and political effects of the plans it drew. All the same, many ac tors hoped for a particular political outcome: either more competitive seats, more victories for their own party, or both. In fact, the commissioners might have found ways to produce a partisan advantage using some combination of the demographic information available to them and their own knowledge of particular communities or incumbents. Even if the commission pursued its task with strict neutrality, the product of its efforts was certain to produce partisan winners and losers—and a partisan impact that emerges from a neutral process still has real consequences. In this section we evaluate the plans for two possible political effects: the num ber of competitive seats, and the share of seats each major party is likely to win un der the new maps. Although we show that the plans will likely impact the nature of political competition and the balance of partisan power, these effects are not always as large as some observers have suggested, nor have they necessarily emerged for the reasons that many critics claimed.

Competitiveness

Although the CRC was not allowed to consider partisanship, advocates of re form clearly hoped the commission would draw a larger number of competitive seats (Common Cause Education Fund 2005, Johnson, et al. 2005, Kogan and Kousser 2011), and the campaign on behalf of Proposition 11 strongly hinted that its passage would lead to greater competition.'- Did the commission live up to those expectations? We consider a seat "competitive" if we predicted that it would lead to

16 Kogan and McGhee: Rcdistricting California: Evaluating the Citizens Commission Final Plans

elections in which no candidate would win by more than 10 percent. This definition of"marginal" districts is commonly used by scholars of legislative elections, and a ID-point advantage is certainly close enough to keep a representative or legislator concerned about reelection without forcing us to adopt an overly restrictive defi nition.'^ We use party registration and incumbency to produce our estimates, first by fitting a statistical model that used these two variables to predict the results of previous elections with data from those elections, and then by combining the model coefficients with the relevant data from the new maps to generate predictions under the new lines.''' Using our definition, the share of seats predicted to be competitive under the new maps can be found in Table 7, with separate estimates for the 2001 districts, the CRC's draft plan from June, and the final plan adopted in August. All the com mission's maps are more competitive than the existing maps, with increases rang ing from three percentage points (i.e., between one and two additional seats) for the final Assembly plan to 13 points (i.e., between six and seven additional seats) for the final House plan. The draft plan in June offered slightly greater competi tive gains in both the Senate (6%, or between two and three seats) and Assembly (6%, or between four and five seats), but not in the House (11%, or between five and six seats), although none of the differences are statistically significant. In fact, while there is at least an 80 percent chance of greater competition in each of the final plans, the competitive gains only clear the 95 percent statistical significance standard for the House. It is worth noting the large role incumbency plays in these estimates of compe tition. If all incumbents were suddenly to retire and leave every district open, our model predicts that the new lines by themselves would make between 19 percent (state Senate) and 25 percent (Congress) of the races competitive—higher than the 14 percent to 18 percent competitive when incumbency is factored in. Of course, the same is true of the old lines as well: between 13 percent and 18 percent would have been competitive if all incumbents had retired, far higher than the 5 percent to 11 percent we observed in practice. Thus, district lines are only one of several factors that determine the competitiveness of races."' If we assume no incumbent effects—that is, if we take every district on its face and remove the effect of the "incumbency advantage"—^which seats would be the most competitive? Table 8 lists the most competitive districts for each plan, along with the general location in the state, the balance of party registration in each district (i.e., the difference between the Democratic and Republican shares of voter registration), and the probability that its outcome will be competitive by our estimates. Since our methodology explicitly incorporates uncertainty, none of these districts is certain to see a competitive race, but each of them has a competitive outcome in at least half our simulated elections. Because these particular estimates

Published by De Gruyter. 2012 17 California Journal of Politics and Policy, Vol. 4[2012], Iss. I, Art. 2

Table 7. Predicted Share of Seats with Competitive Outcomes: 2001 Plan vs. CRC Plans Significant Difference ? 2001 2011 2011 (2001 Plan to 2011 Final Plan Draft Plan Final Plan Plan) Senate 11% 17% 15% No (p=0.20) Assembly 11% 17% 14% No (p=0.16) Congress 5% 16% 18% Yes (p=0.02) treat every race as open, they include many districts political analysts would typi cally omit from a list of this type. But the list gives a sense of which districts could be competitive should they come open, based on hundreds of previous elections results. Indeed, if incumbency is factored back in, the list shrinks from 30 seats to 19, further emphasizing the important role incumbents play in determining the competitiveness of races. Overall, the new maps will produce a modest increase in the number of competitive elections — though perhaps fewer than many reformers hoped for.

Partisan Balance

We conducted a similar analysis to gauge the impact of the new districts on the partisan balance of power, as measured by the Democratic share of seats in each legislative house and the California congressional delegation. Predictions of this kind can be sensitive to assumptions about broader partisan tides, especially when several competitive seats are in play: A good year for Democrats might push a num ber of these seats into the Democratic column, while a good year for Republicans could produce the opposite effect. To address this challenge, we produce two sets of estimates: one based on 2008, which we call a "Good Democratic Year," and one based on 2010, which we call a "Good Republican Year.'"' Basing our analysis on those two years also ensures we are working with a relatively stable period for partisan registration. A large registration shift away from Republicans occurred in the 2008 election, leading to a new, more Democratic equilibrium that has persisted to the present day.'® State Senate races create special complications that are worth a short discus sion. The new even-numbered seats will not be used until the 2014 election cycle, and in the interim, incumbents currently representing even-numbered districts will be allowed to remain in office. If they step down, the resulting special election will be held under the old district lines. To capture this dynamic, we have gener- Kogan and McGhee: Redislricting California: Evaluating the Citizens Commission Final Plans

Table 8. Seats most likely to be competitive in CRC Final Plan (no incumbent effects) Registration Probability of Location Balance(%) Competitive Race Senate 5 San Joaquin +4D 0.74 21 Palmdale +4R 0.53 27 Ventura County +6D 0.64 29 Fullerton +6R 0.55 31 Riverside 4-2D 0.76 34 Huntington Beach Even 0.75 Assembly 8 East Sacramento County +2D 0.79 12 Modesto +3R 0.50 16 Tri-Valley +6D 0.60 35 San Luis Obispo +5R 0.59 36 Palmdale +4R 0.63 40 Redlands +2R 0.69 44 Ventura County +2D 0.81 60 Corona +4R 0.64 61 Riverside +7D 0.68 65 Fullerton +2R 0.78 66 Torrance +3D 0.73 77 North San Diego +9R 0.60 Congress 7 East Sacramento County +ID 0.66 9 San Joaquin +9D 0.60 10 Stanislaus +5D 0.66 21 Kings + I ID 0.62 24 Santa Barbara +4D 0.64 25 Palmdale +6R 0.54 26 Ventura County +6D 0.62 31 San Bernardino +4D 0.69 36 Coachella Valley +3R 0.52 39 Fullerton +9R 0.54 41 Riverside +5D 0.66 52 San Diego +3R 0.65

Note: Location copies the description given by Redistricting Partners, a Democratic consulting firm. Party registration balance is the difference between the share oftotal registered voters who are Republicans and the share who are Democrats. The probability of a competitive race is the share of 1.000 simulated elections where the district's predicted outcome fell between 45 and 55 percent Democratic, based on the statistical model described in Footnote 16. Only districts with probability above 0.5 have been listed here.

Published by De Gruyter. 2012 19 California Journal of Politics and Policy, Vol. 4 12012], Iss. I, Art. 2 ated separate Senate predictions for 2012 and 2014. For 2012, we generate model predictions only for the odd-numbered seats and assume that the rest will continue to be held by the party that currently controls them. We then generate estimates for 2014 based on all 40 seats, which assumes that our model predictions for those odd-numbered seats represent the best guess as to party control moving forward. We split the seats in the same way for our predictions using the old maps, to make the numbers as comparable as possible.'' Table 9 summarizes the results of our analysis, including predicted Democratic seat share for the existing and new maps, as well as the probability of Democrats claiming a two-thirds majority under each plan.-'(Because the estimates from the June draft plan were virtually identical, for the sake of space we have reported only a comparison of the final CRC map and the existing 2001 map.) Perhaps the most interesting result is that our model predicts a strong Democratic performance in the legislature under the existing maps put in place in 2001—far stronger, in fact, than the Democrats have actually achieved.^' Under the 2001 maps, a good Democratic year in 2012 would give Democrats better than even odds (p=0.60) of claiming a two-thirds majority in the Senate, and only a good Republican year predicts a seat share(63%) close to what Democrats currently hold. Likewise, the existing Assem bly districts already seem to give the Democrats good odds (p=0.53) of claiming a two-thirds majority, at least in a good year for their party. Nonetheless, the final CRC map still improves the odds of a two-thirds majority for the Democrats in 2012. The increase is modest for the Assembly (0.53 to 0.66), but much larger for the Senate (0.60 to 0.94). Taken together, if 2012 turns out to be a good Democratic year, we predict a six in ten chance that Democrats will win two-thirds in both chambers under the new maps, compared to a one in three chance under the old. The results are much different for 2014 (recall that estimates for this year incor porate model predictions for all seats). Under the 2001 maps, the odds of a Demo cratic two-thirds in the Senate are a little higher in 2014 than in 2012, while under the new maps the odds are actually a little lower. In fact, when the Senate plan is considered as a whole in this way (rather than as odd and even districts separately), the maximum chances of a Democratic supermajority in the Senate are, if anything, marginally smaller under the new map than the old one (0.79 vs. 0.89), while the maximum chance of a supermajority in both chambers is about the same (0.52 vs. 0.47). Thus, the potential Democratic gains are less a function of the district lines themselves and are more closely related to the numbering of the Senate districts, which "front-loads" seats in 2012 where the Democrats can expect to do better. When it comes to the U.S. House, we predict more substantial gains for Demo crats under the new plan, with as many as four additional seats (7%) in a good Democratic year. Even a bad year for Democrats will likely net them the same share

20 Kogan and McGhee: Redistricting California: Evaluating the Citizens Commission Final Plans

Table 9. Predicted Democratic Seat Share: 2001 Plan vs. Final 2011 Plan

Predicted Seat Share Prob. Democratic 2/3 2001 Plan 2011 Plan Significant 2001 Plan 2011 Plan Difference? Senate: 2012 Good Democratic 67% 70% No (p=0.30) 0.60 0.94 Year Good Republican 63 66 No (p=0.32) 0.04 0.43 Year

Senate: 2014 Good Democratic 69 69 No (p=0.56) 0.89 0.79 Year Good Republican 66 63 No (p=0.32) 0.35 0.11 Year

Assembly Good Democratic 67 68 No (p=0.52) 0.53 0.66 Year Good Republican 63 62 No (p=0.43) 0.00 0.01 Year

Congress Good Democratic 65 72 No (p=0.14) Year Good Republican 63 63 No (p=0.53) Year Note: "Good Democratic Year" estimates are based on an OLS regression on 2008 election data; "Good Republican Year" estimates are based from the same model using 2010 election data. Model coefficients are available from the authors upon request. Significance estimates indicate the proportion of simulations where the difference between the old and new districts held the opposite sign from the reported average difference. of seats they currently hold. In contrast to the Assembly and Senate plans, Repub licans have performed about as expected by our model in recent elections, with the predictions using the 2001 seats closely matching what actually occurred.^^ Thus, it seems fair to attribute any estimated gains to the redistricting plan itself. Nonethe less, the uncertainty inherent to competitive seats places even this large seat gain within the margin of error, because slight vote share deviations in competitive dis tricts can alter the predicted seat share dramatically.-^

Published by De Gruyter, 2012 21 California Journal of Politics and Policy, Vol. 4[20! 2}, Iss. I. Art. 2

Figure 3. Seats-votes curves: existing maps vs. final CRC maps

Assembly Plans: 2001 vs. CRC

ts e

o a

E 2001 Pisn o a C PST;

—T"

Ci C4 1 C Democratc Vote S hare

Why does the congressional plan produce so many more seats for the Demo crats than either the Assembly or the Senate plans? Figure 3 offers some explana tion by tracing out the seats-votes curves for the both the old districts and the new ones. A seats-votes curve displays the relationship between the average share of votes received by a party in each district and the seats won as a result across a broad range of potential election outcomes (Niemi and Fett 1986). The curves therefore allow us to identify the partisan seat share we should expect for any average vote share in that plan.^"* The black lines in Figure 3 identify the curve for the 2001 plan

22 Kogan and McGhee: Redislricting California; Evaluating the Citizens Commission Final Plans

Figure 3. cont.

Assembly Plans: 2001 vs, CRC

CO ,w o a Id b o E 2001 pai o a

o _

—r -r- —r-

c; c-t C6 c I 1 0 Democratc Vote S hate in each case, while the dotted red lines identify the curve for the final CRC plan. We also label the 2008 and 2010 outcomes on each graph for reference. Each ofthe 2001 plans has a distinctive bowing in its seats-votes curve between a vote share of about 55 and 65 percent Democratic—exactly the range where elec tions in California tend to occur. This bowing is a graphical representation of the lack of competitive seats, because at the point where the curve bows, a shift in vote share (on the horizontal axis) produces a smaller change in seat share (on the verti cal axis). By contrast, all three CRC plans come close to eliminating the bowing.

Published by De Gruyter. 2012 23 California Journal of Politics and Policy, Vol. 4[2012], Iss. I, Art. 2

Figures, cont.

U.S. House Plans: 2001 vs. CRC

® □ o .c. CO

m ® CO o a ■s

2O01 Pen Q

"T"

02 10 Democratic Vote Share leaving a straight line where additional votes translate into new seats at a much faster pace. In the Assembly and Senate plans, the original bowing was modest so the consequences of straightening it are less pronounced. But the bowing was quite severe in the 2001 congressional plan. Given the range of vote shares across which elections in California have historically occurred, this has limited the gains the Democrats might have achieved under a more competitive map. The predicted seat shares for 2008 make this clear: The same vote share produces a much higher seat

24 Kogan and McGhee: Redistricting California: Evaluating the Citizens Commission Final Plans share for the Democrats under the new map than under the old one, because the new map allows the additional Democratic support to translate into greater seat gains. These results do not necessarily suggest that the CRC maps avoid any and all attempts at gerrymandering. But they do show that the new plans all resemble each other in the seat share that is predicted for similar shares of the vote, so the differ ences in partisan gains for Democrats between the new plans have at least as much to do with the 2001 plans to which they are compared as with the lines the commis sion has drawn. Indeed, given the strongly uncompetitive nature of the 2001 con gressional plan, it seems unlikely that it is possible to draw any plan that increases competition among congressional seats without also advantaging the Democrats.

Summary

Our analysis of political effects points to several conclusions. First, as many observers had hoped, the districts are somewhat more competitive than the districts drawn by the legislature in 2001. Apart from the congressional plan, however, the gains in competition are fairly modest. Indeed, for the Senate and Assembly, the level of competition predicted by our model is higher for the 2001 districts than ac tually occurred. This suggests that factors besides either incumbency or the district lines have conspired to keep competition low in recent years and may continue to do so moving forward. Second, the partisan change we predict for the new plans is generally modest, and where it is larger, the new lines are not necessarily responsible. Based on party registration and incumbency. Republicans have been outperforming expectations in the current set of districts, especially in the Senate, where the Democrats probably ought to have claimed a two thirds majority already. This fact is visible without any statistical model. Under the 2001 Senate map, every Democrat represents a safe seat, while three Republican senators either represent Democratic-leaning districts or won their last election by a small margin (Sen. Blakeslee in District 15, Sen. Cannella in District 12, and Sen. Strickland in District 19). Had two of these three races turned out differently, the Democrats would already have a two-thirds major ity. The handwriting for Republicans has been on the wall for several years. Because the map in place before the redistricting already predicted better Dem ocratic performance than actually occurred, only a small fraction of the partisan effects are properly attributed to the new lines drawn by the commission. In the Senate, improved Democratic performance is a product of district numbers more than any other cause. It is conceivable that Democrats could claim more than two- thirds of the seats in 2012 but lose the super-majority in 2014, based largely on the districts at stake in each year. Of course, one might accuse the commission of deliberately producing this result. We offer no opinion on this idea, except to note

Published by De Gruyler. 2012 25 California Journal of Politics and Policy. Vol. 4(2012]. Iss. I. Art. 2 that the commission adopted a systematic method of numbering Senate districts (Citizens Redistricting Commission 2011, 25), so any argument about the number ing would presumably need to take issue with that process itself.-^ We find the largest partisan effects in the House map, where the Democrats may pick up four or more seats. These gains stem from the old congressional plan as much as the new one. The 2001 congressional plan was the least competitive of the three considered here. Such low levels of competition ensured that a strong Democratic performance would not lead to gains in seat share. By contrast, the new congressional plan is the most competitive of the three, so a good year for Demo crats will probably allow them to pick up the seats they did not win before. Yet because these gains come from competitive seats, they could easily be lost again in a countervailing partisan tide. Greater competition brings opportunity, but also risk. The implications of these results depend greatly on one's primary concern. If the question is whether the new maps are better for the Democrats, the answer is clearly yes. But the gains for Democrats are either uncertain, because they depend on competitive seats, or conditional, because they depend on the esoterica of Senate seat numbering. Moreover, there is nothing about the new lines that guarantees a good Democratic performance. If past Republican overperformance is due to sys tematic factors we have not accounted for in our model (e.g., strong candidates or fundraising). Republicans might continue to perform well and hold at least one- third of the seats into the foreseeable future.-''

Discussion

There is little doubt that the maps produced by the CRC,and the process through which these plans came about, represented an important improvement on the legis lature-led redistricting of 2001. The new district boundaries kept more communi ties together and created more compact districts while at the same time increasing opportunities for minority representation. If these maps survive the coming referen dum and legal challenges, they have the potential to modestly increase competition in California elections and the responsiveness of the legislative branch to changing voter preferences. These are important accomplishments, and many supporters of Propositions 11 and 20 will rightly point to them as Californians and political observers elsewhere evaluate our state's innovative approach to redistricting. Yet any political reformer who believed that simply giving the job of drawing districts to regular citizens and attempting to take politics out of the process would produce maximum improve ment on every desired dimension will surely be disappointed. As the 2011 expe rience has showed, many other social and political factors—such as the partisan self-sorting and segregation of the electorate—greatly limit the extent to which the

26 Kogan and McGhee: Redistricting California: Evaluating the Citizens Commission Final Plans rediStrict!ng process, no matter how fair and nonpartisan, can change the product of redistricting. Maximizing certain criteria, such as keeping communities intact and protecting the voting rights of historically underrepresented groups, makes achiev ing other goals, such as increasing competitiveness, more difficult. Perhaps most importantly, the experience of the CRC has shown that a nonpartisan, participatory, and transparent process does not eliminate the zero sum nature of electoral com petition. The commission's maps have both winners and losers and may result in lasting impacts on the balance of political power. If Califomians decide that they like their new system of drawing lines, they will need to jealously guard it in future iterations, because the organized interests who care most about the process will surely learn from this maiden voyage. They will attempt to either steer future com missions in their direction or sabotage the commission's efforts entirely.

References

Abramowitz, Alan 1., Brad Alexander, and Matthew Gunning. 2006. "Incumbency, Redistricting, and the Decline of Competition in U.S. House Elections." Jour nal ofPolitics 68(1): 75-88. Arvizu, John R., and F. Chris Garcia. 1996. "Latino Voting Participation: Explain ing and Differentiating Latino Voting Turnout." Hispanic Journal ofBehavioral Sciences 18(2): 104-28. Barabas, Jason, and Jennifer Jerit. 2004. "Redistricting Principles and Racial Rep resentation." State Politics and Policy Quarterly 4(4): 415-35. Brown, Louis. 2000. "Reapportionment in California; Where We've Been, Where We Go from Here." Manuscript. University of the Pacific McGeorge School of Law. Cain, Bruce E., and Karin Mac Donald. 2007. The Implications ofNesting in Cali fornia Redistricting. Berkeley, Calif.: Institute of Governmental Studies. Cain, Bruce E., Karin Mac Donald, and Iris Hui. 2008. "Sorting or Self-Sorting: Competition and Redistricting in California." In New Political Geography of California, ed. Frederick Douzet, Thad Kousser, and Kenneth P. Miller. Berke ley, Calif.: Institute of Governmental Studies Press. California Department of Finance, Demographic Research Unit. 2011. "Map 3: Nonwhite Population by Percent in County: 2000 & 2010." . California Secretary of State. 2008. California General Election, Tuesday, Novem ber 4, 2008: Official Voter Information Guide. Chen, Jowei. 2010. "The Effect of Electoral Geography on Pork Barreling in Bi cameral LtgxsXo.iures.''' American Journal ofPolitical Science 54(2): 301-22.

Published by De Gruyler, 2012 27 California Journal of Politics and Policy. Vol. 4[2012}. Iss. I. Art. 2

Citizens Redistricting Commission. 2011. "Final Report on 2011 Redistrict- ing." . Common Cause Education Fund. 2005. Designer Districts: Safe Seats Tailor Made for Incumbents, Washington, D.C. Hoflfenblum, Allan. 2011. "Memo to Redistricting Commissioners: Beware of Those Who Protest Too Much!" Fox and Hounds Daily. . Jacobson, Gary C. 1990. The Electoral Origins ofDivided Government: Competi tion in U.S. House Elections, 1946-1988. Boulder: Westview Press. Johnson, Douglas, Elise Lampe, Justin Levitt, and Andrew Lee. 2005. "Restor ing the Competitive Edge: California's Need for Redistricting Reform and the Likely Impact of Proposition 77." Claremont, Calif.: The Rose Institute of State and Local Government. Kogan, Vladimir, and Thad Kousser. 2011. "Great Expectations and the California Citizens Redistricting Commission." In Reapportionment and Redistricting in the We.st, ed. Gary Moncrief. Lanham, Md.: Lexington Books. Kousser, J. Morgan. 1997. "Redistricting: California, 1971-2001." In Governing California: Politics, Government and Public Policy in the Golden State, ed. Gerald C. Lubenow and Bruce E. Cain, Berkeley, Calif: Institute of Govern mental Studies Press, 137-55. . 1998. "Reapportionment Wars: Party, Race, and Redistricting in Califor nia, 1971-1992." In Race and Redistricting in the 1990s, ed. Bernard Grofif- man. New York, N.Y.: Agathon Press, 134-90. 2006. "Estimating the Partisan Consequences of Redistricting Plans— Simply." Legislative Studies Quarterly 21(4): 521-41. Ingram, Carl. 2001a."Revised Districts Hinder Minorities, Critics Say." Los Ange les Times, Sept. 6, B8. . 2001b. "Davis Oks Redistricting that Keeps Status Quo." Los Angeles Times. Sept. 28, B2. McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2009. "Does Gerryman dering Cause ?o\2iv'\z2i\.\or\T' American Journal ofPolitical Science 53(3): 666- 80. McCue, Kenneth F. 2011. Creating California's Official Redistricting Database. Available at . McGhee, Eric. 2008. Redistricting and Legislative Partisanship. San Francisco: Public Policy Institute of California.

28 Kogan and McGhee: Redistricling California: Evaluating the Citizens Commission Final Plans

. 2011. California's 2011 Redistricling: The Commission's Final Plans. San Francisco: Public Policy Institute of California. Niemi, Richard G., and Patrick Fett. 1986. "The Swing Ratio: An Exploration and an Assessment." Legislative Studies Quarterly 11: 80-83. Niemi, Richard G., Bernard Grofman, Carl Carlucci, and Thomas Hofeller. 1990. "Measuring Compactness and the Role of a Compactness Standard in a Test for Partisan and Racial Gerrymandering." Journal ofPolitics 52(4): 1155-81. Quinn, T. Anthony. 1981. "California." In Reapportionment Politics: The History ofRedistricling in the 50 States, ed. Leory Hardy, Alan Heslop, and Stuart An derson, Beverly Hills, Calif: Sage Publications, 53-57. . 201 la. "Redistricting Commission: The Blind Squirrel Finds a Nut." Fox and Hounds Daily, . -. 201 lb."The Commission's First Draft Maps: Doing Worse than the Legis lature Did." Fox and Hounds Daily, . Polsby, Daniel D., and Robert D. Popper. 1991. "The Third Criterion: Compact ness as a Procedural Safeguard against Partisan Gerrymandering." Yale Law & Policy Review 9(2): 301-53. Skelton, George. 2009. "Polarization is Paralyzing California. Los Angeles Times. Aug. 10. Special Masters on Reapportionment. 1991. "Report and Recommendations." Available at . Stephanopolous, Nicholas. 2012. "Spatial Diversity." Harvard Law Review 125.

Notes ' California voters first created a redistricting commission in 1926. However, this body was dilTcrent in important respects from the commission that carried out redistricting in 2010. First, the 1926 commission was made up of public officials — including the lieutenant governor, who was to serve as the chairman. Second, the commission was given power over drawing political boundaries only if the legislature first failed to agree on a plan. The earlier commission never actually oversaw the redistricting process in the state. See Brown (2000) for additional details. 'Even after this late-1982 maneuver, California's redistricting wars of the 1980s were not over. Republican assemblyman and winery owner Don Sebastiani funded an initiative to draw new plans in 1983, but the measure was thrown out by the state Supreme Court on the grounds that it violated the state constitution's requirement that redistricting occur once a decade. In the November 1984 election, Gov. Deukmejian pushed another initiative to create an independent commission, which was defeated, as was Proposition 1 19 in 1990 (Kousser 1997). Moreover, on the same ballot with

Published by De Gruyter, 2012 29 California Journal of Politics and Policy, Vol. 4[2012 J, Iss. l.Art. 2

Proposition 119 was Proposition II8. a measure to require a two-thirds vote in each chamber of the legislature to pass a redistricting plan, and it too was defeated. ^ In fact, there are many legitimate reasons, including minority empowerment, for the drawing of irregularly shaped districts. A different section ofthe state constitution prior to the passage of Proposition 11 had included nearly identical language protecting the boundaries of cities and counties, though not necessarily of neighborhoods and other "communities of interest." Proposition 20 added a specific definition of the term to the constitution:"A community of interest is a contiguous population which shares com mon social and economic interests that should be included within a single district for purposes of its effective and fair representation. Examples of such shared interests are those common to an urban area, a rural area, an industrial area, or an agricultural area, and those common to areas in which the people share similar living standards, use the same transportation facilities, have similar work opportunities, or have access to the same media of communication relevant to the election process" (California Constitution, Article 21, Section 2(d)(4). 'Influential Republican analyst and redistricting expert Tony Quinn praised early visualizations of the maps, commending the commission for listening to community input and concluding that "[t] he maps are balanced in partisan terms"(Quinn 2011a). This is not to suggest that the June plans were universally liked. Some Republican leaders criticized the draft maps soon after they were re leased (Hoffenblum 2011). Moreover, Quinn himself quickly had a change of heart, writing several weeks later that the maps "have bombed" and arguing that the commission "managed to gerryman der the state even more than the legislature did in 2001"(Quinn 201 lb). In Romero v. City ofPomona {\9^9\ the Ninth Circuit Court of Appeals endorsed the use of CVAP to assess compliance with the Voting Rights Act. ^ These numbers are calculated by the authors using data on citizenship rates from the 2000 Census and the 2010 American Community Sur\'ey estimates. " This follows the methodology used by the commission in its final report, except the commis sion did not count splits that produced areas with zero population; such splits are included in our count. Despite overall population growth, the number of counties too big to fit in a single legislative district did not change between 2000 and 2010. Thus, the comparison of the raw counts provides a useful metric for the differences between the 2001 legislative redistricting and the 2011 commission plans. Results available from the authors on request. "The Polsby-Popper score for each political district is equal to the ratio of the area of the dis trict to the area of a circle that has the same perimeter length as the district. For instance, supporters of Proposition 11 argued in the November 2008 voter information guide that "[t]here is a serious conflict of interest when legislators arc allowed to draw their own district boundaries. They divide up neighborhoods and communities to create districts where they are virtually guaranteed reelection. Once elected, these politicians aren't accountable to voters be cause they don't have to earn our votes."(California Secretary of State 2008, 72) Later in the same guide, supporters also argue, "[Proposition 11] means fair election districts drawn by citizens, not politicians, so we can hold[the politicians] accountable and throw them out ofoffice ifthey aren't doing theirjobs." (California Secretary of State 2008, 73, emphasis in original.) One could measure competitive seats in a number of ways. Past efforts have used the number of Democratic and Republican voters (Abramowitz, et al. 2006, McGhee 2011), turnover (i.e., the seats that actually change hands)(Johnson, et al. 2005), or close outcomes (Jacobson 1990). We assume that supporters of greater competition value it for its ability to keep incumbents concerned about public opinion outside their own party base. A seat need not actually change partisan hands

30 Kogan and McGhee; Redistricting California: Evaluating the Citizens Commission Final Plans to frighten an incumbent in this way. Likewise, the number of Democratic and Republican voters in a district might be a decent proxy for election results, but it is the results themselves that matter. So while both turnover and party registration are defensible, close outcomes are a more satisfying approach than either one. We regress the Democratic share of the two-party vote on the Democratic and Republican shares of total registration and separate dummies for Democratic and Republican incumbents. Be cause we conceptualize competitiveness as a characteristic of each district independent of specific elections, we pool together all the elections from 2002 through 2010 into a single multilevel model, with random intercepts for election years, and then generate predictions for an average year. We also omit uncontested seats for the purposes ofestimation but generate predictions for all seats. Thus, our predictions implicitly assume that all districts would be contested under the new maps. For the sake of coding incumbency, we assume that all incumbents will run in their district of primary residence, and that any seat occupied by a termed-out incumbent is open. After running these models, we sample 1,000 vectors ofcoefficients and errors to generate our estimates of probabilities. All models were run in R. Coefficient and goodness-of-fit estimates from these models are available from the authors upon request. " Our predictions overstate the number of competitive races that actually occurred in the As sembly (11% predicted vs. 8% actual) and Senate(11% predicted vs. 7% actual) while perfectly pre dicting competitive races in the Mouse(5% each). Nonetheless, there is little reason to think that the predicted changes will be consistently over- or underreported for the new plan. The same model also overpredicted competitive races in the 1990s, but there was no bias in the estimated change in com petitiveness from the 1990s to the 2000s. These results are available from the authors upon request. One might also use the vote for the seven partisan statewide offices in 2010 and for president in 2008 as the basis for a measure of competitiveness—for instance, the districts whose average vote across all those races falls within some range around 50 percent, or the number of districts won by each party at least once. We tend to prefer the regression-based approach because it ties the estimates explicitly to votes, Nonetheless, these measures produce broadly similar estimates of competitive races, with the new Assembly plan always the least competitive and the House and Senate plans the most competitive depending on the measure. For the Assembly and House, we ran separate OLS models for 2008 and 2010 and generated predictions off each model separately. We omitted uncontested seats and used the same predictors as for the estimates of competition: Democratic and Republican registration, and separate dummies for Democratic and Republican incumbency. Since the Senate offers so little data(a maximum of20 races per cycle), we ran one regression for both 2008 and 2010 and included a dummy to capture the intercept shift between the two cycles. There is a common misconception that the national Republican tide of 2010 did not touch California. Although Republican candidates did perform poorly statewide, they gained substantial ground in legislative and congressional races. Controlling for party registration and incumbency and excluding uncontested seats, the average Republican candidate in 2010 earned an additional 4 percent of the vote in the Assembly, 5 percent in the Senate, and 6 percent in the House compared to 2008. Shifts of these magnitudes would qualify as substantial partisan tides in the broader sweep of American history. However, given the lopsided registration advantage for one of the two large parties in most districts, these swings did not produce any extra victories for Republican candidates. "Because both the odd- and even-numbered district estimates are based on the same model for 2014 (i.e., either a "Good Democratic Year" or a "Good Republican Year") it assumes that both 2012 (when the odd seats are up) and 2014(when the even seats are up) would be good years for the same party. In reality, a good year for one party is rarely followed by another good year for the

Published by De Gruyter, 2012 31 California Journal of Politics and Policy, Vol. 4[2012], Iss. I, Art. 2

same party. Thus, it is more than likely that the actual 2014 results will fall somewhere between the estimates we present. The results are similar if we use an average of the vote for statewide offices in 2010 and the presidential vote in 2008 as our measure of the likely outcome, assigning those districts with an av erage vote share greater than 50% to the Democrats and all others to the Republicans. This approach does not allow us to distinguish between good years for each party, so the specific estimates vary somewhat. However, they all show a modest gain for Democrats in the Assembly and Senate plans and a larger gain in the House plan. Our regression model predicts vote share, which we then translate into seat share. In an oth- eru'ise competitive race, even slight deviations from our model's prediction will produce the wrong winner. The Democratic underperformance is therefore a consequence of better than expected Re publican performance in key competitive districts. We cannot say whether this performance is a function of something predictable and systematic that we have not included in our model or simple random chance. Our model predicts a 65% Democratic seat share in a good Democratic year and a 63% seat share in a good Republican year, compared to the 64% Democrats actually won in both 2008 and 2010. "Consistent with this idea, the predicted seat shares under the old plan are much more precisely estimated than they are under the new plan, while there is no such difference in precision for vote shares. Specifically, the seats-votes curve graphs the implications of a hypothetical partisan tide moving the vote share in every seat by a uniform amount. For example, if the Democrats received an average of60 percent of the vote and then suddenly lost five percent in every district, the seats- votes curve would indicate how many seats the Democrats could expect to lose for that five percent, assuming that nothing else about the districts changed. The commission's process intended to minimize the number of voters moving from an odd to an even numbered district or vice-versa. To our knowledge, neither the 2001 plan drawn by the legislature nor even the 1991 plan drawn the Special Masters followed a similar systematic process. The legislature made no attempt to explain any of its decisions, while the Special Masters simply noted that they "tried to assign the numbers rationally" (Special Masters 1991, 74)and that any er rors they made could be corrected by the California Supreme Court (Special Masters 1991, fn 60). There are two ways that we tested our Senate results to ensure that the real outcome for Republicans was not worse than we estimate. First, it is possible tht party registration is not the best gauge of a district's political .sensibilities, so we tried estimating our models with each dis trict's 2008 presidential vote and 2010 gubernatorial vote. We obtained these vote results from the web site of Redistricting Partners, a Democratic consulting firm (http://redistrictingpartners.com/). Predictions based on these numbers actually suggested better outcomes for Republicans than the analysis using party registration. Second, we allowed for certain key incumbent decisions affecting competitive districts that have been announced or that might soon be announced: Republican Tom Benyhill running in the safe Republican District 8 in 2014 instead of the competitive District 5 in 2012; Republican Sam Blakeslee retiring, leaving solidly Democratic District 17 open; and Juan Vargas running for Congress instead of challenging fellow Democrat Christine Kehoe for District 39. The results were the same.

32

Contents lists available at ScienceDirect Elecforcl Electoral Studies

Ei^V^ER journal homepage: www.elsevier.com/locate/electstud

Ballot (and voter) "exhaustion" under Instant Runoff Voting: CrossMark An examination of four ranked-choice elections'^ Craig M. Burnett \ Vladimir Kogan ^ Untversi'O' of North Carolina Wilmington, 601 S. College Road, Wilmington, NC 28403-5607. L/Si^ '' Oli/o State University, USA

ARTICLE INFO ABSTRACT

Article liisfoo'; Some proponents of municipal election reform advocate for the adoption of Instant Runoff Received 31 July 2014 Voting (IRV), a method that allows voters to rank multiple candidates according to their Received in revised form 5 November 2014 preferences. Although supporters claim that IRV is superior to the traditional primary- Accepted 10 November 2014 runoff election system, research on IRV is limited. We analyze data taken from images of Available online 18 November 2014 more than 600,000 ballots cast by voters in four recent local elections. We document a problem known as ballot "exhaustion." which results in a substantial number of votes Keywords: being discarded in each election. As a result of ballot exhaustion, the winner in all four of Instant Runoff Voting our cases receives less Chan a majority of the total votes cast, a finding that raises serious Ranked choice voting Alternative vote concerns about IRV and challenges a key argument made by the system's proponents. Ballot exhaustion ® 2014 Published by Elsevier Ltd.

1. Introduction election system. Under the primary-runoff format, voters participate in two separate elections. In the first round,voters Instant mnoff voting (IRV) — also known as ranked- cast a vote for one candidate from among the entire field. If a choice voting and. outside of the United States, the alter candidate receives a majority, no runoff election occurs. If no native vote — promises to guarantee majority winners in candidate receives a majority of votes, the top two vote- single-member district elections. Under IRV. voters rank getters compete in a runoff election. IRV. by contrast, only the candidates in accordance with their preferences. If no requires a single election where voters rank the candidates. candidate receives a majority after the initial count of first- Proponents of IRV argue that a single election is less choice votes, the candidate with the fewest number of first- demanding on voters' time, cheaper for taxpayers, and limits choice votes is eliminated; the ballots supporting the the influence of moneyed interests in politics by reducing eliminated candidate are then redistributed according to fundraising among candidates (for a longer discussion, see the voters' ranked preferences indicated on the ballots. This Richie, 2UU3). Furthermore, IRV advocates assert that the process continues until a candidate receives a majority of instant runoff ensures that no "spoiler candidates" can the votes. emerge to deprive the winner of a majority — for example, In the United States, a number of local jurisdictions use Ralph Nader in the 2000 United States presidential election IRV as a replacement for the traditional primary-runoff — which remains a possibility in a traditional runoff election. How widespread is the use of IRV? According to Fair- Vote.org,^ eighteen municipalities and four states in the ^ We thank David Cary. Harold Clarke. Kristin Kanthak. and Rob Richie United States use some variant of IRV. In some cases, the for comments. All errors remain our own. A replication dataset is avail able at https://u.osu.edu/kogan.l8/research/. • Corresponding author. Tel.: -t-1 910 962 2161; fax: -f 1 910 962 3286. ^ A list of municipalities, countries, and organizations that use IRV is E-mail addresses: [email protected] (C.M. Burnett). Kogan.lS@osu. available at: http://www.lai1vote.0rg/1eforms/:nstant-runofT-voting/ edu (V. Kogan). where-instant-runoff-is-u.sed/. http://dx.doi.0rg/lO.lOI6/J.electstiid.2OH.ll.OO6 0261-3794/® 2014 Published by Elsevier Ltd. 42 CM. Burnett. V. Kogan / Elecloral Sludies 37(2015) 41-49

NONPARTISAN NONPARTISAN NONPARTISAN CtTY CITY CITY

FOR MAYOR FOR MAYOR FOR MAYOR

FIRST CHOICE SECOND CHOICE THIRD CHOICE riMiiaaftedM fUMiiBirt A*dllMNl Ian ><■« fKlcOufJ M awwwtf caecMi vaih(0« WMtarOn* euANORRoosevar ELEANOR ROOSEVar ELEANOR Roosevar

BOOKER T.WASMMOTON BOOKER T.WASKINQTON BOOKER T.WASKMGTON

OSOO RIVERA OKCORtVERA OtEGORiyQU

ARTHUR lOUXR ARTHUR ICUER ARTHUR NBLLER

SMIRIEYKORN SHRIEYKORN SKRLEYHORN

BRUCE LEE BRUCE LEE

JtWOA*!

Rg-1. Sample IRV ballot.

method is used for the election of all major city officials, third choice. All candidates are listed in all three columns, while in others. IRV is only available for overseas voters and voters are asked to select only one candidate from each who would almost certainly be unable to complete and column. It also states that each choice should be different mail in two ballots in the short window between the pri from the others. Almost every implementation of IRV in the mary and runoff elections under the traditional primary- U.S. limits the number of rankings that a voter can make, as runoff format. Additionally, a number of governments in this example, because allowing voters to rank all possible outside of the United States use IRV to elect a variety of candidates is too technically taxing to implement in prac officials, as does the Academy Awards (Oscars) and a tice given the available voting and tabulation technology. number of organizations and corporations. Australia is Under most iterations of IRV. if no candidate receives a perhaps one of the best-known examples of IRV use: voters majority of first-choice votes, the candidate with the have used this method to elect members of the Australian smallest number of first-choice votes is eliminated. The House of Representatives for over 90 years. ballots that ranked the eliminated candidate as the first Despite its supposed advantages, IRV also has the po choice are then redistributed to the second listed choice. tential to suffer from a number of democratic shortcom The process is then repeated in the second round and so on. ings, three of which we consider here. First, ranking If at any point the voter did not rank a next choice candidates — up to three candidates in the cases we (assuming her most favored choice or choices are elimi consider — is more difficult for voters when compared with nated). or all of the choices on the voter's ballot have been a traditional election where they must choose only one in eliminated, the ballot is "exhausted" — meaning that it is each race. Put another way, ranking preferences beyond the excluded from future vote redistributions, and it does not most favored alternative can be a cognitively laborious task affect the final outcome of the election. The ballot, in for voters who often seek to minimize the time and effort essence, is discarded. The process ends once a candidate needed to make political decisions (Downs, 1957; Popkin. receives a majority of the remaining valid votes. 1994). Second, IRV does not ensure that the winning IRV is very similar to the single transferrable vote (STV)^ in candidate will have received a majority of all votes cast, only that — at least theoretically — both electoral systems have the a majority of all valid votes In the final round of tallying. Thus, potential to provide better representation for the electorate it is possible that the winning candidate will fall short of an compared to First Past the Post (FPTP) systems, with pro actual majority when a substantial number of ballots are ponents defining "better" to mean the election of candidates eliminated, or "exhausted." during the vote redistribution supported by a greater percentage of voters. Indeed, unlike process. Third, and related to the previous point, there is some probability that a voter's ballot will become exhaus ted. eliminating their influence over the final outcome. We ^ STV is. in essence. IRV in muitimember distrias. Under STV. however, return to this point in our concluding discussion. it is difficult for both parties and voters to be strategic because there is the possibility of wasting votes on one candidate when the extra votes would be more impactful had they been cast for a different candidate from the 2. Instant Runoff Voting: benefits and clialienges same party (B.iitholdi and Orlin. 1991). Parties, recognizing this problem, often encourage their party identifiers to "spread the preferences" among all candidates from the party to ensure that as many of the party's can Instant runoff voting (IRV) is an electoral system that didates will be elected as possible (Bowler and Fari ell. I99S). Unlike IRV, provides voters the opportunity to rank-order candidates STV introduces an element of randomness to the process: After a according to their preferences. A voter under IRV ranks her candidate receives the requisite number of votes (called the Droop quota), most favored candidate as her first choice, her second most which votes should be transferred to the next-ranked candidates? In favored candidate as her second choice, and so on. See Fig. I most iterations of STV. the votes that are transferred are chosen through a random draw (Fairell and McAllister 2003). For a longer description of for a sample IRV ballot. In this example, the ballot has three how STV works, see Doron and h'ronick (1977). Richie (2003). and columns corresponding to the voter's first, second, and Tideman i lOOS i. CM Burnett, V. Kogan / E/ertora/ Snidics 37(2015) 41-49 43

FPTP,IRV ensures that the winner of the election receives the make broader appeals to multiple ethnic groups for their majority — rather than plurality — of the eligible votes. Re lower-ranked votes, which will eventually be distributed to formers who advocate for the adoption of IRV make the determine the actual election winner. Others, however, normative claim that plurality winners are less representative dispute this argument (Fraenkel. 2001; Fraenkel and of the electorate than are majority winners. As Richie argues, Grofman. 2004, Fraenkel and Florowitz, 2006; see also under FPTP, "It Is quite possible that most voters dislike the Horowitz 2004). winner who 'represents' them"(2003, p. 503). Proponents ofIRV also contend that the system Is cheaper FalrVote.org. the leading advocacy group for election to administer compared to the traditional runoff system reform In the U.S., emphasizes this argument In Its case for currently In use by many local governments In the United IRV adoption. On a page titled "Comparing IRV With Plu States. Under the existing system, candidates compete In a rality Voting," a section spells out "The Problems With primary election, and the top two vote-getters move on to a Plurality Voting" and begins by noting: "Plurality voting, runoff that determines the final winner. While this system whereby the candidate with the greatest number of votes almost guarantees that the winner receives support from a wins. Is the norm In most American elections. As a result, majority of voters. It requires two separate and costly elec time and again we witness some of our most powerful tions and allows for the possibility of much lower turnout In elected offices filled with candidates who were not sup the runoff stage, especially If the runoff Is not held concur ported by the majority of voters. ... In fact, the prospect rently with other elections. As Richie, Bouriclus,and Macklln becomes very real that the winner of an election may even argue, "IRV duplicates a series of traditional runoffs, but have been disliked by a majority of the population. This Is without the need for additional elections that cost taxpayers the first and most basic problem with the plurality system." and candidates more money and often lead to falloffs In voter The subsequent section, titled "How IRV Addresses These participation" (2001, p. 303). Indeed, anecdotal evidence Problems," begins with the assertion that "IRV Protects suggests that the potential for cost savings Is one of the main Majority Rule." It states: "A raw mathematical aspect of IRV reasons for why local governments that have adopted the Is that whoever wins will have done so with more than 50% IRV method chose to do so. of the votes. ... This winner will be the candidate that Is In making the case for IRV, Richie (2003) generally ar considered at least acceptable to a true majority. Plurality gues that FPTP (1) does not require a majority,(2) allows rules are such that a candidate who Is opposed by the "spoiler candidates," who can alter the outcome of the majority can win."(Emphasis added.) election, and (3) creates Incentives for negative cam IRV and FPTP sometimes produce different winners. As paigning." Richie also finds fault In the traditional two- Doron and Kronick (1977) note, the IRV runoff process can round runoff method In that It (1) requires candidates to produce a majority winner who did not obtain a plurality of raise more money,(2) asks taxpayers to finance an extra the first-round votes.'' Bean's(1997) research on Australian election, and (3) reduces voter participation by requiring elections and a simulation by Sanders et al.(2011) further voters to go to the polls more than once. IRV, according to demonstrate that FPTP and IRV can lead to divergent Richie, can alleviate these problems — a proposition that Is election outcomes.^ In the case of Sanders et al.(2011), the especially attractive from an administrative perspective. authors find that. If the United Kingdom had adopted IRV, IRV, however.Is not a panacea for the problems associated the Liberal Democratic Party would have won more seats In with local elections. First, some research suggests that, by 2010 and both the Labour and Conservative parties would requiring voters to rank multiple candidates rather than have won fewer. Some of the differences are due to stronger simply Identifying the one they most prefer, IRV can become Incentives for strategic, or tactical, voting In FPTP elections more difficult and confusing for voters (e.g.. Bowler and among those who wish to avoid "wasting their vote" by Farrcll. 1995; Diinieavy et al., 1997).' While the system supporting a candidate with a low probability of winning. seems to work well In some places (e.g., Australia), research Some advocates of IRV argue that the method encour on decision-making has shown that as the number of choices ages the election of more moderate candidates and dis Increases, so does the Individuals' difficulty In making de courages negative campaigning by creating Incentives for cisions (e.g., Schwartz. 2003).^ While most of the localities In candidates to appeal to a broader section of the electorate. The veracity of this claim has been especially contentious In research on divided (plural) societies. Horowitz (1991, ^ By contrast. Richie argues that IRV discourages negative campaigns 1993) posits that IRV can moderate ethnic cleavages In because the winning candidate will likely need to receive at least some these fraught political contexts. While ethnic solidarity second-choice votes. If. however, a candidate feels she can win a majority may result In voters picking their co-ethnlcs as their first- in the first round, the incentive structure for negative campaign remains choice candidate, Horowitz argues that IRV rules identical under IRV and FPTP. encourage voters to choose more Inclusive candidates who ^ Ne.'lv and Cook ;2008] analyze the results from San Francisco's voters' experiences with IRV. They conclude that, while there is definitely a learning curve, voters seem to adapt to the new system reasonably well. " Evidence from Great Britain, where a sizable majority of voters " This also happens in traditional runoff formats (e.g.. Antonio Villar- rejected a referendum to shift from a FPTP system and instead use the aigosa won the 2005 Los Angeles mayoral election after finishing the alternative vote for parliamentary elections, provide evidence on this primary in second). point. Surveys conducted in the month before the election showed that a ® The primary differences in outcomes are produced by strategic voting substantial number of voters reported that the ranked-choice voting — and how incentives for strategic voting differ across the two systems. alternative was hard to understand and support for the reform was In other words, voters do not necessarily support the same candidates in strongly and positively correlated with voters' level of formal education FPTP elections as they do in the first round of an IRV election. and political knowledge (see Whiteley ci al., 2012). 44 CM BumeU. K Kbgan / Electoral Studies 37(2015) 4I~49 the U.S.(and the ones we study here)implement a modified Table 1 version of IRV that asks voters to rank only their three top Selected demographic characteristics of cases. candidates, making three choices is more difficult and im Oakland Pierce San San poses substantially higher information costs than a single county Francisco Leandro choice. Population 389,301 792,504 799,512 84.013 Second, as we document in our analysis, IRV does not Median Household Income 550,094 557,214 571.779 558.978 guarantee that the winner in the final round actually se Median House Value 5475,700 5268,700 5773,600 5413,500 Poverty Rate (Families) 16.a 8.1% 7.4% 7.0% cures the majority of all votes cast. This occurs because, in Unemployment 9.6% 10.6% 8.2% 12.3% practice, a large number of ballots are eliminated during White 26.3% 70.7% 42.0% 26.8% earlier rounds of redistribution due to exhaustion, and are Black 27.2% 6.8% 5.9% 12.3% thus excluded from the final vote tally. Latino 25.1% 8.9% 14.9% 28.0% Asian 16.4% 5.7% 33.4% 27.6% Third, as noted earlier, there is a substantial probability that a voter's ballot will become exhausted in the process of Source: American Community Survey, 2008-2010 3-Year Estimates. counting votes, and therefore will not be considered in the final round.This is especially true in the American cases we We focus on these cases because all four use the same study, all of which limit the number of candidates each ballot design and election technology and all have made voter can rank to three. That is. if each of a voter's top three digital images of ballots cast in the elections available to the candidates is eliminated, his or her ballot becomes public. By using these images, we can analyze the behavior of individual voters, and examine how these voters ranked exhausted and, as a result, is excluded from the final total. the candidates. The same will be true for voters who rank fewer than three Table 1 provides some basic demographic information candidates, and whose preferred candidates are eliminated in early rounds. This reality may undermine the democratic about each of the four cases in our sample. Overall, the legitimacy of IRV in the eyes of voters whose ballots tables suggest that the cases are diverse in size, socioeco- become exhausted prior to the final round. nomics, and ethnic makeup." The four cases provide sub stantial variation among which to study the dynamics of elections and voter behavior. It is important to note that all 3. Case selection four cases use the same variant of the IRV method,allowing voters to rank up to three candidates in each contest A small but growing number of local Jurisdictions in the regardless of the total number of candidates running.'" We United States have adopted IRV or some variant thereof as consider the extent to which the three-candidate limit their method for electing public officials. Here, we examine contributes to ballot exhaustion. In this section, we provide four recent elections run under IRV rules,representing a fairly a brief overview of the four contests we examine. large proportion of governments that use this electoral method. Two, San Leandro and Oakland, are cities in Cali 3.1. Oakland. 20J0 fornia, and both used IRV for their mayoral contests in 2010. One, Pierce County, is a county in Washington that used the In late 2010,Oakland Mayor Ron Dellums announced that method to elect the county executive^ in 2008.The final case he would not seek a second term. Dellums, a retired long is San Francisco, a consolidated city-county that elected its time congressman, was first elected mayor in 2006. mayor using the method in 2011. With the exception of San Throughout his term as mayor, he was widely criticized as Francisco, which first implemented IRV in 2004,the elections being ineffectual and largely inactive. He also endured a we examine are the first in these Jurisdictions to use the IRV series of controversies, including a settlement with the IRS method instead of the more traditional primary-runoff elec that resulted in the mayor owing more than $200,000 in tion format. It is possible, therefore, that the patterns we back taxes and penalties. Growing concern over the increase document may become less pronounced in future elections as in crime throughout the city plagued his administration. voters become more familiar with this system.'" Although ten candidates qualified to run to replace Dellums in November 2010, leaked polls identified only four credible contenders for the Job. Don Perata, who had ® The county executive is similar to the county manager or adminis recently ended his stint as the president of California's state trator and serves as the chief executive for the agency. Given that the experience of Oakland. Pierce County, and San Leandro Senate due to term limits and had previously served on the largely match the findings from San Francisco, where voters have utilized Alameda County Board of Supervisors, appeared to enjoy IRV during a number of earlier election cycles, we should not simply support from the largest number of voters.'^ While Perata assume that this would be the case, however. Shortly after this paper was won backing from leading interest groups, including many accepted for publication in November 2014, both Oakland and San Leandro held their second mayoral elections using the IRV system. In San Leandro, preliminary results indicated that the winning candidate captured 50.6 percent of total valid votes cast, with an exhaustion rate of " With the exception of San Leandro, these cases are substantially 5.2 percent. In Oakland, however, the winning candidate appeared to larger than the typical municipal and county government in the United secure only 47.2 percent of valid votes, with an exhaustion rate of 24.7 States; they arc also more ethnically and racially mixed than the average percent. Oakland's experience provides further evidence that our results city and county. are not limited only to first elections held under this method. Results While the three-choice limit is one of the most common imple accessed at luip://vvww.,icgov.org/rnv/cuirent .ek'ction/226/index.htm mentations of IRV in the United States, some jurisdictions allow voters to and htrp;//www.acgnv.org/rov/icv/iesults/226/rcvrcsults_y292.htm on rank a greater number of candidates. Nov. 5, 2014. Oakland is the largest city in Alameda County. CM. Bumeit. V! Kogan / Electoral Studies 37(2015) 41-49 45

unions and developers, he also suffered from a checkered The candidates and independent committees spent a political past, including a five-year FBI investigation into total of$l million on the fall 2008 election (Wickeit, 2008). questionable payments made to Perata by consulting firms Bunney won the most first-round votes(35 percent), while owned by his friends and relatives that had received money the two Democrats split most of the remaining votes (26 from his campaign accounts. Additionally, Perata was the percent for McCarthy and 23 percent for Goings). In the chief architect of the financing agreement that brought the third round, however, McCarthy eked out a narrow victory Raiders football team back to Oakland from Los Angeles in over Bunney, beating him by just over 4000 votes. She 1994; the deal proved to be quite costly for both the city became Pierce County's first female executive. and the county. Second in the polls after Perata was Jean Quan, a city 3.3. San Francisco, 2011 councilwoman who previously served as a member of the Oakland Unified school board and was the city's most After Mayor Gavin Newsom was sworn in as California's recent vice-mayor, a largely symbolic post on the city lieutenant governor in January 2011, San Francisco's Board council.The remaining two major candidates were Rebecca of Supervisors appointed City Administrator Ed Lee to fin Kaplan, an at-large city councilwoman first elected to office ish the final year of Newsom's term. Lee became the first two years earlier, and joe Tuman, a San Francisco State Asian-American mayor in the city's history and pledged University political science professor and local television that he would not run for a full term. In August, however, political analyst. Notably, the Oakland Tribune, Oakland's after a highly publicized "Run, Ed, Run" campaign bank largest newspaper, endorsed Kaplan and selected Tuman rolled by his supporters, including leaders of the local and Quan as second choices. business community. Lee announced that would file his Overall, pro-Perata forces spent nearly $1 million on his candidacy for mayor. campaign, which included more than $600,000 spent by Some opinion polls put Lee in the lead, at around 35 his official campaign committee and remaining funds percent of the vote, but he fell well short of a majority coming from independent groups that backed his bid (Burt, needed to win the November election outright in the first 2010a). Many ads, especially those purchased by the inde round: a bevy of other well-known candidates trailed Lee pendent committee, targeted Quan by attempting to link closely. Taking advantage of the city's generous public her to the unpopular Dellums, the outgoing mayor. Quan, financing system, 16 candidates qualified for the ballot, who signed a voluntary pledge to stay below a $379,000 most of whom had previous experience in elected office. spending cap, countered by running ads criticizing the Other noteworthy leading candidates included City Attor Raiders financing deal that Perata brokered (Burr. 2010b). ney Dennis Herrera and state Senator Leland Yee (Mailer On Election Day, Perata won a plurality of first-choice and Ross, 2011). Lee's opponents criticized his flip-flop in votes cast. He led with 34 percent of first-round votes, running for a full term and accused him of being a closet compared to 25 percent for Quan, 21 percent for Kaplan, Republican, a stinging criticism in the heavily Democratic and 12 percent for Tuman. After nine rounds of elimination city. Eschewing the front-runners, the Democratic central and redistribution, however, Quan won the majority of committee endorsed the progressive Supervisor John Ava- remaining ballots, beating Perata by fewer than 2000 votes. los, although it picked Herrera as its second choice. Lee, who refused to accept public financing, spent more than $1.3 million on the campaign on top of hun 3.2. Pierce County, 2008 dreds of thousands spent by independent groups. Other candidates also raised and spent substantial sums, The retirement of County Executive John Ladenburgh in including several million dollars in public funds. Herrera 2008 — who was forced out by term limits — created an received $712,000 in public financing, followed by opportunity for candidates seeking a high profile executive $682,000 for former Supervisor Bevan Dufty,$563,000 for position that could serve as a potential springboard for Board of Supervisors President David Chiu, $516,000 for higher office. A total of four candidates qualified to run for Yee: and $451,000 for Avalos(Gordon and Knight, 2011).'"' the position of overseeing the state's second most populous Overall, Lee led the field in first-choice votes, winning 31 county, which includes the city of Tacoma and its suburbs. percent in the first round. He was followed by Avalos, with Shawn Bunney, a county councilman and former head of 19 percent, Herrera with 11 percent, Chiu with 9 percent, the regional transportation planning agency, built his and Yee with 8 percent, with the remainder split among platform around economic development and job creation the other candidates. In the twelfth round. Lee beat Avalos — arguing primarily for additional investment in local by almost 30,000 votes, winning 60 percent of the final transportation infrastructure. The only Republican in the votes. race, Bunney faced off against two Democrats, incumbent county auditor Pat McCarthy, who ran on a platform of 3.4. San Leandro, 2010 increasing investment in public safety, and county coun cilman Calvin Goings, who outlined an ambitious agenda Most local political observers expected MayorTony Santos focusing on increasing economic development and to win re-election in 2010 with ease. Santos enjoyed support improving public safety. The final candidate was indepen dent Mike Lonergan, who had twice been elected to the Tacoma city council and championed squeezing efficiencies Other candidates not listed here also received substantial public from the county's criminal justice functions. financing. 46 CM. Bumeff, V. Kogan / Electoral Studies 37(2015) 41-49 from major public employee unions,won the endorsement of Table 2 the Democratic Party, and faced relatively weak opponents. Percent of votes cast for election winners. The most visible of these was second-term city council- Oakland Pierce San San woman Joyce Starosciak. The other was former school board county Francisco Leandro member Stephen Cassidy, who raised substantially less Total Votes for Winner 53.897 136.346 84,457 10.277 money and did not win endorsements from any prominent Total Valid Votes Cast 119.607 299.132 194.418 22.484 stakeholder group. Cassidy, Santos, and Starosciak were all Winner's Vote Share 45.1% 45.6% 43.4% 45.7% Democrats, though Cassidy did not actively seek out the party's nomination. His campaign emphasized prudent budgeting and he antagonized organized labor by advocating process, resulting in the exclusion of these ballots from the cost-saving reforms and cuts in employee retirement bene final round. In each of the four elections, the rate of ballot fits. an issue that attracted significant attention due to rising exhaustion was substantial. Exhaustion was least common pension costs that squeezed out spending on basic city ser in San Leandro, where 9.6 percent of ballots with valid first- vices (San LeandroTalk, 2011). Two other largely unknown round ballots were discarded prior to the final round. By candidates also qualified for the ballot. contrast, the exhaustion rate was highest in San Francisco, Santos received a tiny plurality in the first round, winning a where 27.1 percent of valid first-round ballots did not make total of 36 percent of first-choice votes, compared to 35 it to the final round. Voters who cast these discarded ballots percent cast for Cassidy. Starosciak trailed both with 23 had no say in the final round of vote redistribution, which percent. Cassidy, however, prevailed in the fifth round of decided the election outcome. elimination,edging out Santos by slightly more than 200 votes. There are two possible causes of ballot exhaustion. First, voters may have ranked three different candidates on their 4. Assessing the benefits and challenges of IRV ballots, but each of these three may have been eliminated prior to the last round of counting. Although this was un We begin our analysis by presenting aggregate statistics likely to occur in elections with only a few candidates on voter behavior compiled from more than 600,000 in running, such as the mayoral race in San Leandro (five dividual ballot images from these four elections. Our pri candidates) and the county executive contest in Pierce mary focus is on the electoral dynamics in these contests. County (four candidates), it may have occurred with some We use the aggregate numbers to consider the extent to frequency in both San Francisco and Oakland, where a larger which the implementation of IRV produced the types of number of high-quality candidates appeared on the ballot. salutary effects its proponents have argued. We also use Second, voters may have chosen to rank fewer than three these data to identify problems with the IRV method, candidates, only to have their marked choices eliminated which have attracted relatively little attention among re prior to the final round. The first problem is primarily one of formers, scholars, and election officials. technology — at least theoretically, adopting a different First, we examine whether victorious candidates in each ballot format that allows voters to rank a greater number of race did, indeed, win the majority of the total votes cast — candidates can reduce the number of exhausted ballots. By one of the primary arguments for using IRV over the contrast, changing the ballot format does not address the traditional FPTP method. In Table 2, we calculate the share second problem, which is due to voters simply failing to fill of vote won by each winning candidate by taking the out a ballot completely by ranking three distinct candidates. number of ballots allocated to them after the final round of Table 4, which reports the percent of ballots that con redistribution and dividing this figure by the total number tained three unique choices and the breakdown of incom of valid ballots cast in each election. Somewhat surpris plete ballots, shows that a large number of voters failed to ingly, given the arguments made by IRV proponents, none rank all three of their top candidates.This was least likely to of the four elections we examine resulted in the winner occur in San Francisco, where 73 percent of voters identi capturing the majority of all votes cast in the election. fied different candidates for their first, second, and third Although the victors won the majority of ballots that made slot. By contrast, it was much more common in Pierce it to the final round of vote counting in each election, a County, where nearly half of all voters failed to cast a substantial number of ballots in each case became complete ballot. A substantial number of voters — between exhausted during the redistribution process, reducing the 5 percent in San Francisco and 11.9 percent in San Leandro number of ballots needed to prevail in the last round. — listed the same candidate in more than one spot, sug Using the ballot images, we are able to identify each gesting that at least some members of this subset of voters exhausted ballot. Table 3 reports the number of complete did not understand how the IRV system works.'® The fact first-round votes cast in each election'^ and the share of that this problem occurred with regular frequency even in these ballots that were exhausted during the redistribution

It is also possible that this phenomenon, sometimes described as The number of valid first-round votes in Table 3 is lower than the "bullet voting." was the result of voters deliberately listing the same number of total valid ballots reported in Table 2. although the differences candidate in multiple slots to signal strong support for him or her. are very slight. This is due to the fact that a very small number of voters Another potential explanation for incomplete ballots is that voters are did not indicate a first choice on their ballot, but did mark candidates in simply indifferent between the candidates who remain in the final round. higher-ranked spots. Their votes were counted using specific policies in In each of the cases we examine, however, the ideological differences place for dealing with these types of under-votes in each jurisdictions. We between candidates in the final runoff were stark, making it highly exclude these unusual ballots from our analysis. implausible that indifference explains the high rate of incomplete ballots. CM. BumeH, V. /(ogan / Electoral Studies 37(2015) 41-49 47

Table 3 Table 5 Rate of ballot exhaustion. Rate of ballot exhaustion by category.

Oakland Pierce San San Oakland Pierce San San county Francisco Leandro county Francisco Leandro

First-Round Votes 119.408 298.912 194.046 22.421 Three Unique Candidates 7.8% 0.0% 22.5% 2.7% Rate of Exhaustion 11.6% 10.2% 27.1% 9.6% Marked Duplicate Candidates 25.3% 18.4% 48.1% 21.3% Marked San Francisco, where voters have been using the IRV Two Unique Candidates 13.3% 5.3% 29.8% 6.1% method since 2004, suggests that even a substantial Marked One Candidate Marked 27.9% 31.1% 44.8% 29.0% amount of public education about the process is unlikely to eliminate this type of voter confusion. Which type of problem — elimination of all three can and cannot be resolved by simply changing the design of didates chosen by voters or incomplete ballots — was the the ballot or the voting technology used. The first is that a most common cause of ballot exhaustion? In Table 5, we large number of voters may lack sufficient information report the rate of ballot exhaustion by type of votes cast. about the candidates to be able to develop a rank ordering Overall, in each election, voters who identified three of their top choices. Informational barriers are one poten unique candidates faced the lowest probability of having tial reason for why voters fail to participate even in high- their ballots become exhausted during the counting pro profile federal elections (Downs. 1957; Riker and cess. In Pierce County, no voter who selected three unique Ordeshook. 1968), where they need only to identify their candidates saw their ballot become exhausted. We expect most preferred candidate in each race. Ranking multiple this result as there were only four candidates on the ballot, candidates in much lower-profile local contests is thus a and selecting three unique candidates guaranteed the difficult task for a substantial part of the electorate. ballot a spot in the final tally. The highest rate of exhaustion Second, ranked-choice voting reduces the incentives for among voters casting a ballot with three unique candidates strategic voting by making it more difficult for voters to was 22.5 percent in San Francisco. By contrast, in all but one determine which candidates are likely to be eliminated in election, voters who cast a ballot only for their single top- early rounds of vote redistribution (for a discussion, see ranked candidate while leaving the other two choices Farrell and McAllister 2003)and which candidates stand to blank had the highest rate of exhaustion, between 27.9 benefit from redistributed votes. As a result, many more percent in Oakland and 44.8 percent in San Francisco. voters are likely to "waste their votes" by supporting can Table 6 presents the same data in a slightly different way. didates with a low probability of prevailing. Although this Rather than calculating the rate ofexhaustion for each ballot is sometimes seen as an advantage of IRV — by increasing type (as shown in Table 5), Table 6 identifies the total per political opportunities for third-party candidates, such as centage of all exhausted ballots accounted for by each level would be the case in British elections (Sanders et al. 2011) of ballot completeness. In San Francisco, voters who ranked — the lower rates of strategic voting creates a serious three different candidates accounted for more than 60 democratic dilemma among voters who fail to rank mul percent of all exhausted ballots. The San Francisco election tiple candidates. Wasting their votes on losing candidates was the exception, however. In the three other cases, means that these voters see their ballots exhausted, and completed ballots accounted for a minority of exhausted thus have much less impact on determining the final ballots, suggesting that giving voters the opportunity to election outcomes than do voters who choose to rank rank more than three candidates may help reduce, but multiple candidates. The interaction between these two almost certainly will not eliminate ballot exhaustion. phenomena — high informational barriers that prevent Across all four elections, a substantial number of voters from ranking enough candidates to submit exhausted ballots came from voters who ranked only one or completed ballots and weakened incentives for strategic two candidates. This was true even in San Francisco, where voting — appear to explain a substantial share of exhausted voters had previous experience with using ranked-choice ballots in the four cases we study here. voting on Election Day. This fact underscores to two serious weaknesses that are endemic to the IRV method

Table 6 Table 4 Share of exhausted ballots by category. Number of complete ballots cast and reasons for incompletion. Oakland Pierce San San Oakland Pierce San San county Francisco Leandro county Francisco Leandro Number of Exhausted 13.800 30.463 52,665 2148 First-Round Votes 119.408 298.912 194.046 22.421 Ballots Three Unique Candidates 72.3% 51.2% 73.0% 60.8% Three Unique Candidates 48.5% 0.2% 60.4% 17.1% Marked Marked Duplicate Candidates 8.2% 6.4% 5.0% 11.9% Duplicate Candidates 18.0% 11.5% 8.9% 26.4% Marked Marked Two Unique Candidates 10.7% 16.2% 10.3% 10.9% Two Unique Candidates 12.4% 8.4% 11.3% 6.9% Marked Marked One Candidate Marked 8.7% 26.1% 11.7% 16.4% One Candidate Marked 21.1% 79.8% 19.4% 49.6% 48 CM. Burnett, V. Kogan / Electoral Studies 37(2015) 41-49

5. Discussion who led in first-choice votes but lost in the final round of redistribution, refused to concede defeat for several days, Proponents of IRV argue that the electoral method can arguing that he was the rightful winner of the election. San mend many of the flaws of local democracy. Our results Leandro's Tony Santos,too, argued that IRV was the cause of show, however, IRV comes with Its own set of concerns. his defeat. Although initially one of the leading proponents Despite the claim made by its leading proponents that of adopting the method in the city, Santos became a vocal ranked-choice voting can ensure that victors win a majority critic after the election, bashing IRV in press interviews. of the votes cast in the course of a single election, none of Santos even testified in the state capitol against a bill to the elections we examined here resulted in such an make it easier for local governments to adopt ranked- outcome. This occurred because many votes were exhaus choice voting (San LeandroTalk, 2011). ted during the ballot redistribution process. As Table 3 Although ensuring that the winner of elections enjoys indicated, the exhaustion rate ranged from 9.6 percent to support from a majority of the electorate is certainly not as high as 27.1 percent. This rate of exhaustion led to the necessary for an electoral system to be legitimate, it is winners in all four cases earning, on average,45 percent of nevertheless an important consideration. As FairVote.org the total votes. The reality that a substantial number of itself notes:"The democratic standard ought to be majority ballots cast under IRV are discarded raises two important rule: it is a fundamental principle of republican gover questions: Does ballot exhaustion affect specific subgroups nance. So in choosing the method of electing our leaders, of the electorate? And, would the election outcome have we should demand that it holds to the principle of majority been different if the voters who cast exhausted ballots had rule." While FairVote.org claims that IRV produces a winner a chance to voice their preference in the final round? who "is considered at least acceptable to a true majority", One concern is whether the rates of ballot exhaustion our research shows that the "true majority" in these elec varied systematically and affected disadvantaged and low- tions may often be a plurality of all votes cast. socioeconomic groups more than others. Not only would Our data show that the IRV method has important this raise a broader democratic worry about IRV but also shortfalls. As the number of candidates increases, so does might violate federal laws, including the Voting Rights Act. the level of ballot exhaustion. For example, San Francisco, Our data, however,do not allow for precise individual-level which had sixteen candidates listed on the ballot, the rate analysis, and relying on precinct data would invite signifi of ballot exhaustion rate was strikingly high (27.1 percent). cant ecological inference concerns, a particular problem While the precise electoral rules and ballot technology given the relatively low rate of turnout in these elections used — which limited voters to only three choices — may and the fact that only population-level demographic have contributed to the exhaustion rate, we are skeptical covariates are available from the Census. that asking voters to rank more than three candidates will Nevertheless, this possibility deserves close examina dramatically reduce the rate of exhaustion. As we see in tion. As the history of Progressive municipal reforms has Tables 4 and 5, a large number of voters failed to submit a demonstrated, electoral rules supported by well-meaning completed a ballot, which constituted a substantial share of reformers can sometimes systematically reduce participa the exhausted ballots. This likely reflects, at least in part, tion among the most disadvantaged segments of the elec the reality that few voters possess enough information to torate. At the turn of the 20th century. Progressive rank more than a few of the candidates running, regardless reformers adopted at-large nonpartisan elections. They also of how many they are allowed to select. The case of Port delinked the timing of local elections from state and federal land, Maine, presents a useful case in point. In Portland's contests. Although reformers believed that these changes 2011 mayoral election, voters had the option of ranking as would encourage more professional governments within many of the 15 candidates running in the election as they cities, reduce corruption, and eliminate the influence of wanted.This did not appear to substantially reduce the rate party bosses, Bridges(1997) shows that these changes — by of ballot exhaustion: almost 18 percent of valid ballots were dramatically increasing the informational costs of voting — exhausted prior to the final round. systematically reduced participation among low-income Some supporters of IRV may argue that the problem of and minority voters. The result was a shift toward pol exhaustion is not a serious one from the perspective of icies that favored middle-class residents who were democratic theory. Exhausted ballots, they sometimes comparatively more active in these off-year elections. assert, are essentially treated no different than votes for Because IRV rules similarly increase informational costs — losing candidates cast in plurality contests.'^ This argument by requiring voters to rank multiple candidates to reduce is problematic for three reasons, however. First, it compares the probability of having their ballot exhausted — future IRV to an alternative — plurality elections — that is not research should examine whether their adoption adversely representative of what actually occurs in the U.S. context. impacts poor and minority voters. Most local elections use the two-stage runoff method, The available data also do not allow us to assess how which results in far fewer wasted votes. Second, voters ballot exhaustion affected the final election outcomes. Regardless of the answer, however, the possibility that exhaustion might tip the balance in the final round poses a " The Ninth Circuit Court of Appeals made this argument in Dudum v. serious risk to the democratic legitimacy of the method and Armz (2011,1: '"Exhausted" ballots are counted in the election, they are just counted for losing candidates in the tally of total votes. In the terms of the outcomes it produces by allowing defeated candi used by election experts, these are "wasted" votes, not because they aren't dates to attribute their loss to the high rates of ballot counted, but because they were cast for candidates not ultimately elec exhaustion. This occurred in Oakland, where Don Perata, ted"(emphasis in the original). CM. Burnett, V. Kogan / Electoral Studies 37(2015) 4I~49 49 frequently avoid wasting their votes in the plurality context government to have a clear understanding of these tradeoffs because it is relatively easy for them to behave strategically, before they engage in electoral reform. adjusting their choices to avoid having their vote go toward a candidate with no chance of victory. With IRV, strategic voting is more difficult, so the risk of having one's votes References wasted becomes more severe. Finally, it is important to note that this line of argument concedes one of our main B.irrholcli III.J.J., Orlin. J.B.. 1991. Single nansferrable vote rc.slst strategic points: That IRV need not, and frequently does not. produce voting. Soc. Choice Welt". 8, 341--354. a winner who wins the majority — rather than plurality — Bean. C.. 1997. Australia's experience with the alternative vote. Repre sentation 34, 103-110. of all votes cast, one of the alleged advantages of this voting Bowler, S.. Farrell, P.M.. 1995. Vote siiategies under preferential electoral that proponents sometimes highlight when comparing it to systems: a single transterrable vote mock ballot survey of London FPTP elections. If IRV tends to produce plurality winners of voters. In: British Elections and Parries Yearbook, pp. 14-31. Bridges, A.. 1997. Morning Glories: Municipal Reform in the Southwest. all votes cast, it is important for policymakers to consider Princeton University Press. Princeton. NJ. this fact when evaluating whether to adopt IRV. Burt. C.. 2010a. Don Perata Breaks Spending Record in Oakland Mayor's We conclude by noting that the goal of this study is to Race. Tribune. Oakland. Burl, C. 20101). Mud Flying in Oakland Mayoral Race. Oakland Tribune, p. 4A. examine ballot exhaustion, a little-noticed feature of IRV Doron. G., Kronick, R.. 1977. Single transferrable vote: an example of a elections, and explore its causes. We do not take any position perver.se social choice function. Am. J. Polit. Sci. 21. 303-311. about whether local jurisdictions should choose this method Downs. A.. 1957. An Economic Flieory of Democracy. HarperCollins. New over their existing electoral systems.Indeed, a shortcoming of York. Dudum v. Arntz. 10-17198 (9th Clr. 2011)(Seeborg. R.). our analysis is that we do not consider other prominent Dunleavy, P.. Margetts. H.. O'DulTy, B., Weir. S., 1997. Making Votes Count: promises ofIRV reform — that the method increases political Replaying the 1990s General Elections under Alternative Electoral participation among voters and encourages a more inclusive Systems (Democratic Audit Paper No. 11). Democratic Audit of the UK. London. style of campaigning. We do not examine these claims Farrell. D.M., McAllister. 1.. 2003. The 1983 change in surplus vote transfer directly because our four cases do not provide us with suffi procedures for the australian senate and its consequences for the cient data to study the differences in turnout or campaign single transferable vote. Aus. J. Polit. Sci. 38. 479-491. Fraenkel. J.. 2001. The alternative vote system in Fiji: electoral engineering styles while holding constant a variety of other important or ballot-rigging? Commonw. Comp. Polit. 39.1 -31. factors (e.g., presence of incumbent, quality of challengers, Fraenkel. j.. Grofman. B.. 2004. A Neo-Downsian model of the alternative timing of runoffelection, amount of campaign spending)that vote as a mechanism for mitigating ethnic conflict in plural societies. Public Choice 121. 487-506. have a substantial impacton both outcomes in local elections. Fiaenkel. J.. Horowitz. D.L. 2006. Does the alternative vote foster Nevertheless, these are likely to be important considerations moderation in ethnically divided societies? Tlie case of Fiji. Comp. for policymakers who are contemplating adopting IRV. With Polit. Stud. 39. t)23-65h respect to turnout, however, we suspect that a well-designed Gordon. R.. Knight. H.. 2011. Taxpayers' Tab. Breakdown Per Candidate Per Vote. San Francisco Chronicle. runoff system, in which both the first round election and the Hajnal. Z.L. Lewis. P.G.. 2003. Municipal institutions and voter turnout in runoff are scheduled to coincide with state and federal elec local elections. Urban AIT. Rev. 38 (5). 645-668. tions,can achieve many ofthe participatory benefits(see, e.g., Hoiowitz. D.L. 1991. A Democratic South Africa? Constitutional Engineer ing in a Divided Society. University of California Press. Berkeley. CA. Hajnal and Lewis, 2003) with a much lower incidence of Hoiowitz. D.L. 1993. Democracy in divided societies. ]. Democr. 4.18-38. wasted votes."^ Hoiowitz. D.L. 2004. The alternative vote and Interethnic moderation: a Overall,our findings suggest that IRV is not a magic cure- reply to Fraenkel and Grofman. Public Choice 121. 507-516. Mailer. P.. Ross. A.. 2011. Lee Popular, but Falls Short of Vote in Poll. all to popular complaints about the quality of local de Chronicle. San Francisco. mocracy, as is true for every electoral method available. Neely. F.. Cook. C.. 2008. Undervotes. over votes, and ranking in San While the method has the potential to reduce administrative Francisco's instant-runoff elections. Am. Polit. Res. 36 (4). 530-554. Popkin. S.M.. 1994. The Reasoning Voter: Communication and Persuasion costs for local governments, it also increases the difficulty of in Presidential Campaigns, second ed. University of Chicago Press. the task facing voters. As our data make clear, a substantial Chicago. IL. number of voters either cannot or choose not to rank mul Richie. R.. 2003. Instant runolf voting: what Mexico (and others) could learn. Election law]. Rules. Polit. Policy 3. 501-512. tiple candidates, even when they have the ability to do so. Richie. R.. Bouricius. T. Macklin. P.. 2001. Candidate number 1: instant Instead, many opt to cast a vote for their top choice, runoff voting. Science 294. 303 306. neglecting to rank anyone else. Substantial educational ef Riker. W.H.. Ordeshook. P.C., 1968. A theory of the calculus of voting. Am. forts regarding the mechanics of IRV may help alleviate this Polit. Sci. Rev. 62. 25-42. Sanders. D.. Clarke. H.D.. Stewart. M.C.. Whiteley. P.. 2011. Simulating the concern and should be part of any transition plan for mu effects of the alternative vote in the 2010 UK geneial election. Par- nicipalities that adopt this method in the future. As our data ham. Aff. 64. 5-23. show, however, even individuals who mark three distinct San Leandro Talk. 2011. RCV in San Leandro: a Primer to the Nov. 2010 Mayoral Election. choices often face the prospect of exhaustion, so education Schwartz. B.. 2003.The Paradox ofChoice: Why More is Less. Ecco. New York. alone will not fix the problem. The results highlight a key Tideman. N.. 1995. Tlie single transferrable vote.]. Eton. Perspect.9.27-38. political reality: no electoral system is perfect and every Whiteley. P.. Clarke. H.D.. Sanders. D.. Stewart. M.C.. 2012. Britain says NO: voting in the AV ballot leferendum. Parliam. Aff. 65. 301-322. method comes with its own set of democratic tradeoffs. It is Wickert. D.. 2008. Pierce County Political Spending Reaches Million-dollar important for voters and their elected representatives in Mark. News Tribune.

Note that this is not always the case. In San Francisco, for example, the run-off stage usually occurred during an off-cycle eleaion that saw dramatic fall off in participation.