Exploring Racial Gerrymandering Using Moment of Inertia Measures Levi John Wolf Chao Fan Wenwen Li Alan Murray SCHOOL OF GEOGRAPHICAL SCIENCES AND URBAN PLANNING November 13, 2014 MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 1 / 26
Contents 1 Introduction Terms & Conditions Laws & Regulations 2 Theory Detecting Gerrymandering Compactness Theory 3 Method Compactness Measures A New MMI Testing Claims 4 Results California Overall Race Group Comparison 5 Conclusion Takeaways Extensions MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 2 / 26
Specifics Gerrymandering: Deliberate manipulation of spatial boundaries to provide a political advantage to a particular group MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 3 / 26
Gerrymandering As Seen In practice, gerrymandering is: often unintentional Objectives abound: gerrymandering to protect social justice (Weber 2013) retrogression and preservation of Majority-Minority Districts protecting incumbents against challenges protecting ideological segments of parties against each other protecting media markets & political capital investments enforcing compact districts manipulation gets fuzzy MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 4 / 26
In the courts Thornberg v. Gingles (1986): Majority Minority districts should only protect sufficiently large and geographically compact communities with RPV Shaw v. Reno (1993): Appearances matter Miller v. Johnson (1995): Compactness problems are not a necessary condition for gerrymandering LULAC v. Perry (2006): Partisan Symmetry is a desirable standard, but not talismanic Barrett v. Strickland (2009): Maj-Min districts must have a numerical majority Shelby County v. Holder (2013): Preclearance is over unless updated (hint: won t get updated!) MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 5 / 26
Main Theory Gerrymandering is a move in the broader game of political control It involves many different strategies at the state level These strategies are the same everywhere They are not always employed every time Each strategy has an observable effect: Partisan bias partisan gerrymandering Racial bias racial gerrymandering MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 6 / 26
Why? Why Use Compactness Measures? Efficient to calculate Strong construct validity (Neimi et al 1990) Decent face validity Legal credibility Why Not to Use Compactness Measures? Not a necessary condition for bias! Poor content validity Too complex? (ten properties of Angel et al 2010) MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 7 / 26
Warrants Compactness arguments rely on an unspoken hypothesis (Chen & Rodden 2009) We need: Districts to be drawn over smooth population distributions CoIs that are each contiguous and compact, but dispersed CoIs that are not large enough to warrant their own districts Thus, compactness measures work when gerrymanders group together dispersed communities that are contiguous, and compact MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 8 / 26
Warrants Compactness arguments rely on an unspoken hypothesis (Chen & Rodden 2009) We need: Districts to be drawn over smooth population distributions CoIs that are each contiguous and compact, but dispersed CoIs that are not large enough to warrant their own districts Thus, compactness measures work when gerrymanders group together dispersed communities that are contiguous, and compact i.e they pack MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 8 / 26
Warrants Compactness arguments rely on an unspoken hypothesis (Chen & Rodden 2009) We need: Districts to be drawn over smooth population distributions CoIs that are each contiguous and compact, but dispersed CoIs that are not large enough to warrant their own districts Thus, compactness measures work when gerrymanders group together dispersed communities that are contiguous, and compact i.e they pack So, Gingles provides a good description If districts are compact & MMDs, they probably are protected If not, they may be packing minorities But, simple counterexamples exist for cracking and stacking (Humphreys 2009) MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 8 / 26
Elections-tested, Court-approved Polsby-Popper (1991): Schwartzberg (1966): Reock (1961): Equiperimeter Circle Area District Area Adjusted District Perimeter Equal Area Circle Perimeter Min Bounding Circle Area District Area MMI (Weaver & Hess 1963): m i d 2 i atomic polygons And on and on and on... MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 9 / 26
A New MMI Given: NMMI = I 0 2π ( ) ρ i I i + m i dir 2 r : a region in the problem frame i : atomic units from which the region r is composed m i d ir ρ i I i I 0 : some mass of atomic unit i : distance from centroid of atom i to center of mass type m of region r : density of mass type m in atom i : Moment of area for atom i : Effective mass moment for region R MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 10 / 26
Logic of NMMI NMMI measures how well a population is packed into a frame Relaxes importance of assumptions Population MI can reflect non-smooth distributions Noncompact or large CoIs are standardized by areal & MMI terms Poor population compactness = district/distribution mismatch Good population compactness = district reflects distribution Breaking mass of NMMI down by race captures different racial population distributions MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 11 / 26
Active Hypotheses To ensure racial fairness, MMDs are necessary If racial compactness is particularly low, MMDs should improve racial compactness for their districts If racial compactness is indistinguishable from total population compactness, then this is a moot point. To make MMDs, we have to make noncompact districts to collect minority CoIs Stems from basic assumptions of compactness theories in gerrymandering MMDs would have low compactness MMDs are self-defeating: they disenfranchise minority voters in neighboring districts (Shelby County v. Holder) Negative spatial autocorrelation between Majorities and Minorities in general MMDs may be spatial outliers in their bivariate racial compactness comparisons MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 12 / 26
0.6 Compactness Scores: Hispanic/Not Hispanic 0.4 0.2 Total Corr: 0.983 Corr: 0.967 0 0.2 0.4 0.6 0.6 0.4 0.2 Non.Hispanic Corr: 0.919 0 0.2 0.4 0.6 0.4 0.2 Hispanic 0 0.2 0.4 MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 13 / 26
0.6 Compactness Scores: Black/White 0.4 0.2 Total Corr: 0.972 Corr: 0.879 0 0.2 0.4 0.6 0.6 0.4 0.2 White Corr: 0.77 0 0.2 0.4 0.6 0.6 0.4 Black 0.2 0 0.2 0.4 0.6 MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 14 / 26
Introduction Theory Method Results Conclusion CA 112 112th Congress: California NMMI Total Population [0,.1] [.1,.2] [.2,.3] [.3,.4] [.4, 1] Bay Area Los Angeles MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 15 / 26
CA 113 113th Congress: California NMMI Total Population [0,.1] [.1,.2] [.2,.3] [.3,.4] [.4, 1] Bay Area Los Angeles MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 16 / 26
CA Overall Results Population Range Mean Median SD Mann-Whitney Total (.011,.596).245.213.145 N/A White (.012,.664).242.209.141.405 Black (.008,.666).196.165.145.004 Not-Hispanic (.012,.596).243.214.141.201 Hispanic (.007,.562).205.176.131 2.2e 16 Holds for each Congress Blacks & Hispanics are not as well packed as reference groups Reference groups are indistinguishable from overall Better in the 112th than the 113th MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 17 / 26
CA113 Total 113th Congress: California NMMI Total Population [0,.1] [.1,.2] [.2,.3] [.3,.4] [.4, 1] Bay Area Los Angeles MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 18 / 26
CA113 African Americans 113th Congress: California African American Population [0,.1] [.1,.2] [.2,.3] [.3,.4] [.4, 1] NMMI Bay Area Los Angeles MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 19 / 26
CA113 Hispanics 113th Congress: California NMMI Hispanic Population [0,.1] [.1,.2] [.2,.3] [.3,.4] [.4, 1] Bay Area Los Angeles MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 20 / 26
Groups Between Group tests Hypothesis (no pair) P-value NMMI B less than NMMI W?.004 NMMI H less than NMMI nh?.021 Holds in each Congress as well No global spatial effects: within any score between any two scores MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 21 / 26
* denotes Hispanic MMD CA 112th Congress LISAs District Race NMMI I i p-value 23 Total.05-1.18.04 - Hispanic.04-1.01.02 30 AfAm.20.15.02 35* AfAm.08.68.01 36 Total.38-1.00.04 - Hispanic.29 -.89.03 38* Total.26.58.01 - * Hispanic.19.25.01 - * AfAm.18 -.02.02 41 Total.14 -.42.04 - Hispanic.12 -.45.02 MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 22 / 26
* denotes Hispanic MMD + denotes Asian MMD CA 113th Congress LISAs District Race NMMI I i p-value 2 Total.14 -.70.04 - Hispanic.14 -.48.04 - AfAm.08-1.10.01 10* Hispanic.11 -.60.03 23 Hispanic.05 1.00.03 27+ Total.37 -.40.04 30 Total.23 -.24.04 - AfAm.24.21.02 MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 23 / 26
Main Takeaways Districts are not compact for Hispanics and African Americans in CA Both are worse than total, sometimes even in MMDs Blacks v. Whites and Hispanics v. Non-Hispanics also bad Between Congresses: No big change in MMDs Atleast they didn t get worse! Aligns with small improvement in geometric in CA Some neighboring districts to MMDs have very poor NMMI Some districts clearly deserve stricter scrutiny (23rd, 36th, 2nd) Still, many questions... MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 24 / 26
Going Further What is the reference distribution for this measure? How compact should we expect an NMMI to be? Use p-compact regions simulation (Li et al 2014) Predict outcomes over simulation (Chen & Rodden 2009) Further examine spatial dependence in NMMI Scraps of MMDs getting picked up by other districts? MMDs will result in disenfranchisement in neighboring areas Some interesting behavior in MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 25 / 26
Thank you! ljw2@asu.edu MMI & Racial Gerrymandering ljw2@asu.edu GEOGRAPHICAL SCIENCES & URBAN PLANNING 26 / 26