Figure 1: Gerry-Mander, Harbinger of Inequality

Elections in Reverse

Leonardo Matone

May 4, 2026

1 Introduction

Gerrymandering in the United States usually involves “packing” opposition voters into few

districts or “cracking” them across many to dilute their inﬂuence. It is a strategic act to

maximize the power of parties or individuals by minimizing the voting power of their opposi-

tion. In 2013, the Supreme Court removed some restrictions on this tactic with the landmark

Shelby County v. Holder decision. The decision renders Section 5 of the Voting Rights Act

(VRA) unenforceable, eﬀectively eliminating it (Brennan Center for Justice 2023). Section

5 established the preclearance requirement, a safeguard which required speciﬁc jurisdictions

with histories of racial discrimination to obtain federal preclearance before changing voting

laws; the goal was to end discriminatory voting changes before they went into eﬀect rather

than relying on lawsuits after the fact. Without preclearance, states and counties with a

history of racial discrimination can implement voting laws and maps with little oversight,

enabling these states to draw maps which limit minority voices and discriminate against

marginalized communities where they could not before. This paper aims to examine the

impacts of the Shelby decision on minority representation in the U.S. as it pertains to elec-

toral fairness. We argue that the decision negatively impacted minority voting power by

demonstrably increasing cracking/packing across the Shelby decision, and that states are in-

centivized to gerrymander with respect to race even when they do not directly gerrymander

with respect to party.

The eﬀects of the Shelby decision are part of a larger trend in U.S. politics (Bernini et al.

2024). On April 29th 2026 the Shelby decision was reinforced by the Supreme Court in

Louisiana v. Callais, which decided that the protections of the VRA no longer apply due to

white and Black parity in “two of the ﬁve most recent Presedential elections.” This is a direct

continuation of the logic in Shelby, characterized by Justice Ruth Bader Ginsburg’s dissent

from 2013: “Throwing out preclearance when it has worked and is continuing to work to stop

discriminatory changes is like throwing away your umbrella in a rainstorm because you are

not getting wet.” The decision speciﬁcally aﬀects Louisiana’s revised district map, which as

a result of Callais v. Landry (2024) included two majority-Black districts out of Louisiana’s

six. Callais v. Landry ruled the original 2020 congressional map an unconstitutional racial

gerrymander under the VRA, ﬁnding that race was improperly prioritized in the creation of

the plan. With the overturn of Callais v. Landry in 2026, racial gerrymandering has been

eﬀectively ruled constitutional.

This trend against the VRA is harmful in two principal ways. Firstly, removing safeguards

to protect minority voting power will inevitably have negative impacts on minority voting

power. Ending preclearance over covered states gives leeway to individuals and institutions

with the capacity to discriminate, and enables them to strategically restrict the voting power

of minorities towards partisan gain or racial prejudice. Secondly, no federal law prohibits

states from undertaking partisan gerrymandering; it is entirely up to states and Congress

to regulate redistricting. One of the only impacts on partisan gerrymandering has been the

VRA, which can indirectly prevent some partisan gerrymandering due to the fact that racial

and ethnic minorities largely identify with the Democratic party (Protect Democracy 2023).

Race is the best predictor of partisanship over any other demographic factor; eﬀorts to

gerrymander on the basis of partisanship often overlap with demographic gerrymandering,

and imply that the Shelby decision would have signiﬁcant impacts not only on minority

representation, but also partisan gerrymandering.

2 Background

The Shelby decision has dangerous implications for minority representation in the U.S. The

literature on the Shelby decision is rich, and usually fall into two major analytical categories.

First is the analysis of voting laws, speciﬁcally changes to voting procedure post-Shelby. Since

2013, covered states have surfaced new laws restricting voting, including voting ID laws, lim-

its on early voting, and same-day registration. Many of these changes have been proven to be

discriminatory in intent; North Carolina notoriously instituted new requirements which were

struck down in 2016 with The North Carolina State Conference of the NAACP v. McCrory,

having proved that the new requirements “target[ed] African Americans with almost surgical

precision.” Despite the wave of controversial new voting laws post-Shelby, studies on the ef-

fects of new voting laws have had mixed results. Some have documented a negative eﬀect on

voter turnout (Hood and Buchanan 2020). Conversely, Cantoni and Pons 2021 ﬁnd that ID

laws had a very limited eﬀect on Black registration and turnout, ﬁnding that mobilization

campaigns targeting minorities were successful in oﬀsetting eﬀorts to disenfranchise them.

This result is crucial in our analysis. From a strategic perspective, it makes sense under the

issue attention cycle (Downs 1972) that these campaigns might lose inertia over time as the

laws remain entrenched in public policy. Even if a change in partisan representation is not

immediately observed, these changes have long-reaching eﬀects that both disproportionally

hurt minorities and prop up the voices which would suﬀer from their representation.

The second category of analysis focuses on redistricting fairness and representation, which we

will focus on. A vital detail in any analysis of gerrymandering is some deﬁnition of fairness.

Fairness to the people is the cornerstone of the U.S. Constitution, and voting is arguably

the most important mechanism behind that fairness. The “fairest” possible system might

be one in which the vote of any person is worth just as much as any other person. Here

is where the trouble begins: this restricts the power of a “state” as an entity that should

have independent and comparable power to other states. Under this deﬁnition, a state with

more people is arguably more inﬂuential in federal decision-making than a state with fewer

people. Famously, this dilemma is what the U.S. Congress and Senate aim to alleviate, but

it also creates complexity around the seemingly simple idea of one person, one vote.

Gerrymandering, while termed in 1812, is an unfortunate tradition nearly as old as the United

States itself and has been studied for at least as long. Steering state districts in favorable

directions for a political party is a natural tendancy for political institutions wishing to

maintain power and inﬂuence. A benevolent institution which aims to protect single votes

still operates in a representative government, requiring a degree of information loss from

single vote counts. The tools for district control in the U.S. have always rested in the

laps of incumbent interests. Gerrymandering may be profoundly undemocratic, but it is

also unsurprisingly widespread in its usage across party and ideology. In the context of

millions of people, fairness, especially in redistricting, has historically proven diﬃcult due

to these incentives. Diﬀerent groups and political parties aim to limit the power of their

opposition, taking whatever deﬁnitions of fairness might be most convenient to maintaining

or implementing a power dynamic (Becker and Gold 2022).

Another straightforward idea of fairness is political representation. Presently U.S. politics is

dominated by two political parties, so with relatively little information loss, ensuring partisan

symmetry across population and representation is an ideal metric (Gelman 2002). In this

formulation, fairness of a particular redistricting plan can be quantiﬁed by the relationship

between a proportion of votes a party receives and the proportion of seats it wins. If a

state has ten seats and a perfect partisan split, one might expect that both parties have

ﬁve seats each. There are two versions of this formulation, one comparing the seat share

each party would win if support were shifted to 50% per party, and the second comparing

with a speciﬁc symmetry parameter ¯p. This enables us to evaluate and compare plans with

a mathematically rigorous deﬁnition of electoral responsiveness and bias. In addition to

partisan bias, political scientists have developed a quantiﬁer termed the ”eﬃciency gap”

(Warrington 2018). Partisan bias quantiﬁes the degree to which parties would win diﬀerent

fractions of the total seats with the same original votes. The eﬃciency gap relies on the

concept of a wasted vote, these being votes that are cast over the majority required to win

an election. We can see this in a simple toy example detailed in Figure 2. Two districts 1

and 2 have ten voters each. One district is won by party A with 7 votes and the other by

party B with 6 votes. In district 1, every vote over 50% of the population is ’wasted.’ An

optimal outcome for party A would be to redistrict 3 of its voters into District 2, where,

despite holding 50% of the partisan vote across districts, party A would hold 100% of seats.

District Party A Party B

District 1 7 3

District 2 4 6

Figure 2: Two-district vote distribution (10

voters each)

Another requirement for our analysis is that we

have a frame of reference to compare diﬀerent re-

districting plans. While we can quantify fairness

with regard to the voting populous or partisan

representation atomically using partisan bias or

the eﬃciency gap, these plans are often drawn to

maximize a power dynamic with respect to a set

of rules which vary by state. There is no baseline

for comparison. The ALARM Project (McCar-

tan, Kenny, Simko, Garcia, et al. 2022) proposes

a simulation-based approach to contextualize congressional district plans within the rules of

a state, enabling state-speciﬁc district simulations. The authors simulate possible district

plans based on the rules of a state, considering geography, demographics, institutions and

redistricting requirements that constrain the space of possible maps. Drawing possible sim-

ulated plans thousands of times, they are able to evaluate diﬀerent quantiﬁers of fairness

over the generated set. This formulation allows for a counterfactual ”ground truth” distribu-

tion of possible plans which can be used to contextualize existing plans, and has been used

extensively in further literature.

The simulated plans take into account existing redistricting rules per-state, but lack more

complicated rules from the Voting Rights Act (VRA). These ensembles are still eﬀective

baselines, but are ineﬀective for contextualizing proposed plans with regard to prohibitions

against racial gerrymandering and minority vote dilution, speciﬁcally the “Gingles factors”

which mandate the creation of minority-opportunity districts. This is ameliorated with

“VRA-conscious” ensembles that are designed with these constraints (Becker, Duchin, et al.

2021), advancing the ALARM Project’s framework by integrating the legal requirements of

the VRA into simulation algorithms.

The ALARM Project oﬀers a comprehensive simulation-based framework to contextualize

redistricting plans, and oﬀers excellent analysis of district fairness in the context of neutral

counterfactual alternative districts. While district demographics are a facet of their Se-

quential Monte Carlo (SMC) algorithm (and therefore existing plans that deviate from the

neutral counterfactuals can be said to possibly deviate from VRA requirements), they do not

oﬀer speciﬁc analysis of the eﬀects of current district plans on minority group representation.

There exist speciﬁc case studies Gay 2007 and general improvements on simulating districts

Becker, Duchin, et al. 2021, but there is limited material assessing the statistical impact on

marginalized groups from these district plans on a federal level, as well as targeting covered

states from the VRA post-Shelby decision.

McCartan et al. (McCartan, Kenny, Simko, Ebowe, et al. 2025) utilize the ALARM frame-

work to evaluate the pre and post-Shelby cycles. They perform a DiD experiment with a

treatment consisting of states that implemented new redistricting reforms between cycles,

with the control consisting of states that maintained their existing map-drawing processes.

They compare these to the ALARM Project’s neutral baseline. Our proposed framework

performs a DiD with the control as the non-covered states (which presumably would see little

change) and the treatment group as the covered states, which presumably would change more

in response to the Shelby decision. While these authors evaluate redistricting outcomes as a

function of institutional reform type, they do not study the eﬀect of the removal of federal

preclearance itself on minority representation.

Komisarchik and White (Komisarchik and White 2020) provide an empirical audit of the

Shelby decision’s eﬀects on minority voter participation. Utilizing a DiD design comparing

previously covered jurisdictions with similar non-covered areas, they study a possibly widen-

ing gap between Black and White voter turnout following the ruling, but ﬁnd limited eﬀects.

Crucially, this work’s thesis demonstrates that the eﬀects of Shelby extend beyond the physi-

cal lines of redistricting maps; the removal of federal preclearance enabled a suite of electoral

changes that target minority groups and can systematically depress minority participation,

establishing that structural legal protections are directly linked to the exercise of electoral

power. These authors establish that Shelby depressed minority voter participation through

procedural changes, but do not examine whether the the degree of racial packing in enacted

maps shifted in response to the removal of federal oversight.

A critical institutional backdrop to post-Shelby redistricting is the Department of Justice’s

“Max-Black” strategy, pursued aggressively during the 1990s redistricting cycle under VRA

Section 5 preclearance authority. Under this interpretation of the non-retrogression princi-

ple, the DOJ pressured covered jurisdictions to maximize the number of majority-minority

districts, often producing “super-majority” districts in which Black populations were con-

centrated well beyond what was necessary to elect a minority-preferred candidate (Lublin

1997). Scholars have further noted that some covered state legislatures exploited VRA com-

pliance as a partisan pretext, using the mandate to pack reliable Democratic voters into

fewer districts while engineering safer Republican seats in the surrounding areas.

Majority-minority districts are further explored by Trautman and Smith 2025, analyzing how

the 2018 and 2022 midterm redistricting cycles inﬂuenced the turnout and mobilization of

Black voters. The study ﬁnds that the structural conﬁguration of a district, speciﬁcally the

degree to which minority voters are packed or cracked, impacts grassroots mobilization and

engagement with the electoral process. This suggests that the harm of unfair redistricting

is not limited to a loss of legislative representation, but extends to a structural dampening

of democratic participation itself. This is further studied in majority-minority districts by

Gay 2007, who examines the trade-oﬀs of packing minority voters into dedicated districts

with respect to the responsiveness of legislators in neighboring ones. She argues that while

such districts ensure descriptive representation, they simultaneously “liberate” surrounding

white legislators from having to appeal to minority interests in order to win. This dynamic,

which she terms legislating without constraints, creates a scenario in which the substantive

policy inﬂuence of a minority group may decrease even as their nominal seat count remains

stable.

The downstream consequences of this packing strategy on minority groups is dangerously

substantial. Lublin 1997 argues that while majority-minority districts reliably increase the

number of Black oﬃceholders, they produce a perverse eﬀect on the surrounding area: by

drawing Black voters who lean heavily Democratic out of adjacent districts, those districts

become more likely to elect Republicans whose policy agendas may be misaligned with

minority interests. Because districts must be equally populous under federal law, the creation

of high-concentration majority-minority districts in covered states operates as a zero-sum

reallocation of minority inﬂuence across the broader political map.

A vital consideration in our analysis should also be how these maps are designed. In practice,

district maps are drawn by whichever partisan body controls the state legislature, typically

with the assistance of professional mapmakers and Geographical Information System soft-

ware. One such mapmaker was Thomas Hofeller, a renowned gerrymanderer for the Repub-

lican party. While his work in states like North Carolina is more famous, his gerrymandering

strategy was adopted professionally by the Republican party across the south through his

company, ”Geographic Strategies.” From some points of view, Hoefeller could be described

a proponent for the VRA: he was a loud advocate for Black majority districts across the

south. This strategy, while initially helping to elect more Black representatives, also served

to pack Democratic-leaning Black voters into fewer districts and ultimately constrict their

voting power by making the surrounding white districts more reliably Republican (Bouton

et al. 2023). After his death, a trove of his personal ﬁles were released which detailed some

of the data and analysis used to develop these maps, exposing the extensive data on race

and other demographics which work behind the scenes to restrict minority voting power and

give Republicans a partisan edge (Common Cause v. Lewis 2019). The Hoefeller example

demonstrates the ways racial and partisan gerrymandering are often linked, and that even

when not speciﬁcally targeting partisan outcomes, the process of drawing electoral maps

often speciﬁcally target marginalized groups using very speciﬁc data analytics.

3 Case Study: Louisiana

A motivating example for this study is the state of Louisiana as the focus of Louisiana v.

Callais (2026). The decennial 2020 district plan exempliﬁes what we aim to study, that

minority packing may be more likely in covered states, and that these states might increase

across the Shelby decision. The state also serves as an example of the mechanics of racial

and partisan packing, and will motivate our chosen features of importance in our analysis.

(a) 2020 Louisiana districts, colored

by BVAP share

(b) 2024 Louisiana districts, post-VRA enforce-

ment

Figure 3: Louisiana district maps pre-Callais (2026)

Following the 2020 decennial census, the Louisiana State Legislature enacted a map (Fig-

ure 3a) that maintained only a single majority-Black district (the 2nd District), despite

Black residents comprising approximately one-third of the state’s population and registered

Democrats making up 36.50% of the electorate. This 2020 plan faced intense legal scrutiny

in Ardoin v. Robinson (2023) and Callais v. Landry (2024), characterized by the extreme

concentration of minority voters into a single district to prevent the formation of a second

competitive seat. With Callais v. Landry, the plan was redrawn in order to comply with Sec-

tion 2 of the VRA. This revised plan, seen in Figure 3b, introduced a second majority-Black

district (the 6th District), stretching from Shreveport to Baton Rouge and signiﬁcantly al-

tering the state’s max BVAP and eﬃciency gap metrics. With Louisiana v. Callais (2026),

the state is now free to return to the original plan.

As a covered state, Louisiana illustrates the dangers of removing federal oversight from

historically discriminatory institutions. The state has a documented and long history of

racially motivated redistricting, with enacted maps that have been successfully challenged

for packing and cracking Black voters in violation of the VRA. Its redistricting record also

elucidates an important question for our analysis: what features of a district map, constitute

a violation of the VRA or a meaningful departure from fairness?

4 Analysis

We focus the scope in our analysis to estimations of district packing, which we extract

from simulated and real districts by taking the maximum over all districts. This gives a

simple, eﬀective metric across our statistics. We speciﬁcally focus on two primary metrics

computed for each enacted plan. The ﬁrst is max BVAP, the maximum Black voting-

age population (BVAP) share across all districts in a plan, which serves as our measure

of racial packing: the degree to which Black voters have been concentrated into a small

number of districts. The second is enacted EGAP, the eﬃciency gap, which indexes partisan

packing and cracking jointly. By focusing on the maximum BVAP and eﬃciency gap, we

isolate the ‘packing’ mechanism of gerrymandering, where minority or partisan groups are

concentrated to a degree that exceeds what would be expected under neutral geographic and

demographic constraints with respect to the Black population and partisan packing. We do

not separately assess cracking in this analysis as these statistics are not as naive to track as

simple maximums, though we acknowledge that this would be an interesting area for future

research; this is particularly true considering how Louisiana’s 2020 plan cracks the BVAP

(Callais v. Landry), which is harder to visually identify or computationally analyze when

comparing the revised 2024 plan to the original.

The demographic and electoral data underlying our analysis come from the ALARM Project’s

redistricting data ﬁles (McCartan, Kenny, Simko, Garcia, et al. 2022), which merge and

tidy two primary sources. Racial and ethnic breakdowns of both the total population and

the voting-age population are drawn from the 2020 decennial Census and 2010 decennial

Census, tabulated at the precinct level. These are joined to precinct-level election returns

from the Voting and Election Science Team (VEST), which provide average vote counts

for Democratic and Republican candidates across statewide races. Where 2020 precinct

boundaries are unavailable, census block-level data are provided instead.

Our analysis is two-pronged. We ﬁrst evaluate the diﬀerences between covered and non-

covered states within the 2020 district plans build oﬀ of our Louisaiana case study and

provide a basis for our argument: covered states are more likely to pack one max district than

non-covered states, in both minority and partisan packing. This analysis mainly focuses on

the 2020 district data and contextualizes it with 2010 ﬁgures, which we focus on to highlight

existing disparities as context for our test. Second, we detail several statistical experiments

to evaluate the eﬀect of the Shelby decision. We evaluate the diﬀerence between covered

and non-covered states across the Shelby decision to determine whether or not minority

packing increased for covered states, and theorize what eﬀects it may have. The states we

will examine include:

• Fully covered (n=8): AL, AZ, GA, LA, MS, SC, TX, VA

• Partially covered (n=5): CA, FL, NH, NY, NC

• Not covered (31 states with ALARM data)

Because redistricting outcomes are heavily contingent on a state’s unique geography and de-

mographic distribution, raw metrics like the Eﬃciency Gap or BVAP percentage cannot be

directly pooled in a regression. To ensure comparability across states, we assess departures

from the simulated distribution using two complementary simulation-relative statistics: a z-

score and a percentile rank. Within simulation metrics, we deﬁne the former as the enacted

plan’s deviation from the simulation mean in units of standard deviation. In demographic

metrics on a per-state basis, we deﬁne it as the enacted plan’s BVAP with regard to the pop-

ulation of the state atomically. We deﬁne the latter within simulation metrics alone, as the

fraction of simulated plans falling below the enacted value. These metrics enables compar-

isons across states with diﬀerent populations, standardizing the magnitude of “extremeness”

across jurisdictions. This approach transforms state-speciﬁc results into a common metric

of “departure from neutrality,” allowing for a statistically sound comparison between states

with vastly diﬀerent baseline populations and partisan leans.

4.1 2020 Plan Baselines

Figure 4: Histogram of covered states’ simulated max BVAP with enacted plan as a red

vertical line.

We ﬁrst utilize the ALARM Project’s simulated district data and 2020 precinct-level demo-

graphic data to construct district-level minority voting power metrics, utilizing a Gingles

packing excess measure deﬁned as the surplus Black Voting Age Population (BVAP) above

the 55% safe-harbor threshold that is typically suﬃcient to elect a minority-preferred can-

didate. We use this to compare covered and non-covered states and establish a basis for

our claim: covered states are more likely to pack than non-covered states in single years.

In Figure 4, we can see that the max BVAP share across all districts in covered states is

far from the neutral simulation distribution, greater than 95th percentile across all states.

This means that the observed levels of racial concentration in these jurisdictions are not the

product of non-partisan geographic constraints. Instead, the enacted maps are statistical

outliers that concentrate Black voters into “super-majority” districts at rates signiﬁcantly

higher than what would occur under a race blind, neutral redistricting process.

This trend can also be compared across states. Figure 5 details a national choropleth of

racial packing in 2020 congressional redistricting plans, measured against neutral simulation

Figure 5: 2020 z-scored max BVAP packing across the US

baselines from the ALARM redistricting project. For each state, approximately 5,000 theo-

retically race-blind redistricting plans were simulated under their neutral algorithmic process.

Per state, we record the max BVAP share across the congressional districts of each enacted

plan, and compare it to the distribution of that same statistic across the simulated plans.

States are colored only where the enacted plan constitutes a statistically signiﬁcant outlier

(with regard to the state’s simulated plans alone), which we deﬁne to be if fewer than 5% of

simulated plans produced a maximum Black district BVAP share equal to or greater than

the enacted plan’s (p ¡ 0.05). Color intensity reﬂects the BVAP z-score, with purple indicat-

ing greater packing than neutral plans would predict and blue indicating less. States shown

in gray are not statistically distinguishable from neutral redistricting with regard to BVAP.

States with thick black borders were fully covered. This plot is central to our argument-

when corrected for population, covered states pack Black voters into super-majority districts

where non-covered states deviate less.

Figure 6 tracks the trajectory per-state of the max BVAP congressional district between the

2010 and 2020 redistricting cycles. For each state, we identify the single enacted district with

the highest Black voting-age population (BVAP) share - the ”rank-1” district - and plot an

arrow from its position in 2010 to its position in 2020. The gray cloud shows a sample of

districts drawn from neutral, race-blind ALARM simulations for the 2020 cycle, providing a

baseline for what the rank-1 district might look like under non-partisan redistricting. The

left panel shows the eight fully-covered states under the VRA prior to the Shelby decision; the

right panel shows all other states with data available in both cycles. Arrows moving rightward

indicate increased racial concentration in the most-Black district across cycles; arrows moving

downward indicate reduced Democratic presidential performance in that district, consistent

with a dilution strategy. While the max BVAP district across 2020 and 2010 may not be

the same, it is nonetheless interesting to observe that over the Shelby decision max BVAP

districts covered states do not increase BVAP share, and that many decrease their share.

4.2 Diﬀerences-in-Diﬀerences Across Shelby

Our central inquiry examines whether the trajectory of district packing from 2010 to 2020

diﬀered signiﬁcantly between fully VRA-covered states and non-covered states. Speciﬁcally,

we ask whether congressional districts in Section 5 covered states are more racially packed

than would naturally arise from race-neutral redistricting, and whether the Shelby decision

exacerbated this diﬀerence between covered/non-covered states. We evaluate this for two

targets: demographic packing (deﬁned by the creation of majority-minority districts that

hyper-concentrate minority voting power) and political packing (measured via the eﬃciency

gap). Here we again utilize the ALARM Project’s 50-state simulations.

To evaluate these shifts, we deﬁne our temporal and legal variables to precisely isolate the

impact of the Shelby decision. We deﬁne our P ost-treatment period as the 2020 redistricting

cycle. By comparing the 2010 and 2020 decennial cycles, we capture the atomic redistrict-

ing events immediately preceding and following the ruling, minimizing noise from oﬀ-cycle

or court-mandated mid-decade redraws and ensuring maps are compared under the same

fundamental constitutional requirements. Regarding legal status, our T reated variable fo-

cuses on the VRA Section 5 coverage formula active prior to Shelby. We opt to categorize

partially-covered states (e.g., CA, FL, NY) as non-covered due to their ambiguous signal at

the congressional level. Sensitivity analyses conﬁrm our results are robust to this speciﬁca-

tion, as well as to the inclusion of state ﬁxed eﬀects and raw eﬃciency gap outcomes; we

ﬁnd no signiﬁcant diﬀerence in coeﬃcients whether partially covered states are included in

the control group or excluded entirely.

We use a Diﬀerence-in-Diﬀerences (DiD) experiment to estimate packing changes in two

groups across the pre and post-Shelby era, where P ost is 1 for the 2020 group and 0 for the

Figure 6: Max BVAP districts plotted by dem presidential vote share and BVAP share.

Left covered, right not covered. Colored dots are the maximum BVAP districts per state,

annotated with their change from 2010 in the same space.

2010 group, T reated is 1 for fully-covered VRA states (“AL”, “AZ”, “GA”, “LA”, “MS”,

“SC”, “TX”, “VA”), and β

is the DiD estimator:

∆packing = β

+ β

× T reated + β

× (P ost × T reated) + ϵ (1)

We also perform the same experimental setup to predict the eﬃciency gap, utilizing the

following formulation:

∆egap = β

+ β

× T reated + β

× (P ost × T reated) + ϵ (2)

From Table 1, the eﬃciency gap results are statistically insigniﬁcant. The primary ﬁnding

from this Diﬀerence-in-Diﬀerences model is a null result regarding partisan packing/cracking.

The data does not show a statistically signiﬁcant diﬀerence in how the Eﬃciency Gap (z-

score) changed between VRA-covered (T reated) and non-covered states following the 2020

redistricting cycle (P ost-Shelby). The DiD estimator (β

= 0.904, p = 0.408) fails to reach

signiﬁcance, suggesting that any shift in partisan symmetry within VRA-covered states was

not meaningfully diﬀerent from the shifts observed in the rest of the country. This null ﬁnding

is evidence against our hypothesis, that the Shelby decision increased packing. While the

legal landscape for minority representation shifted dramatically, the “partisan eﬃciency”

remains un-impacted across the decennial plans.

Variable Estimate Std. Error t p

Intercept 0.434 0.357 1.215 0.229

Post -0.818 0.505 -1.621 0.110

Treated 0.840 0.768 1.094 0.277

Post × Treated 0.904 1.086 0.833 0.408

Observations 74

0.106

Adjusted R

0.068

Residual Std. Error 1.922 (df = 70)

F -statistic 2.767

∗

(df = 3, 70)

∗

p < 0.05;

∗∗

p < 0.01;

∗∗∗

p < 0.001. Dependent variable: simulation z-scores for the eﬃciency gap.

Table 1: Diﬀerence-in-Diﬀerences: Eﬃciency Gap (z-score)

The Diﬀerence-in-Diﬀerences analysis of max BVAP in Table 2 demonstrates that the pri-

mary distinction in racial packing is not the change over time, but the persistent and extreme

baseline of the treated states. While the Shelby decision removed the federal preclearance

requirement, the “Post × Treated” coeﬃcient suggests that this legal shift did not lead to a

statistically signiﬁcant surge in further packing beyond the already extreme levels established

in the 2010 cycle. Instead, the data suggests that covered states have maintained a “packing

equilibrium.” The z-score of 6.010 for the Treated variable conﬁrms that these jurisdictions

were already outliers in the 2010 cycle, likely a result of the “no-retrogression” standard

of Section 5 which, ironically, often incentivized the maintenance of super-majority districts

pushed by gerrymandering entities like Thomas Hoefeller. The lack of a signiﬁcant DiD eﬀect

(p = 0.154) indicates that the post-2020 maps in these states continue to be extreme outliers

relative to neutral simulations, regardless of the change in their legal oversight status.

Variable Estimate Std. Error t p

Intercept 0.229 0.453 0.507 0.614

Post -0.069 0.635 -0.109 0.914

Treated 6.010 0.987 6.091 <0.001 ∗ ∗ ∗

Post × Treated -2.005 1.393 -1.439 0.154

Observations 77

0.428

Adjusted R

0.404

Residual Std. Error 2.480 (df = 73)

F -statistic 18.17

∗∗∗

(df = 3, 73)

∗

p < 0.05;

∗∗

p < 0.01;

∗∗∗

p < 0.001. Dependent variable: simulation z-scores for maximum BVAP share.

Table 2: Diﬀerence-in-Diﬀerences: Max BVAP (z-score)

4.3 Ideology and Packing

Building oﬀ of our analyses, we examine the eﬀect of packing these districts on ideology of

the candidates which represent them. To accomplish this, we utilize the DW-NOMINATE

(Dynamic Weighted Nominal Three-step Estimation) metric, which acts as a political spa-

tial mapping, placing every member of congress on a spectrum based on their roll-call vot-

ing records (Poole and Rosenthal 2007). The motivation behind this experiment is that

majority-minority districts (often the result of packing) are less competitive, which lead to

idiosyncratic representatives Gay 2007. Beyond restricting the power of minorities, we also

want to understand how these representatives behave ideologically, and whether or not racial

packing impacts ideological scores.

By placing every member of the 117th Congress on a spatial map ranging from -1 (most

liberal) to 1 (most conservative), we can determine if the packing identiﬁed in our simulation-

based z-scores translates into a statistically signiﬁcant shift in ideological extremity. This

allows us to assess whether the Shelby decsion primarily aﬀected the number of minority-

inﬂuenced seats or if it also fundamentally altered the ideological extremity of representation

within those districts. We use the following setup:

NOMINATE = β

+β

× BVAP+β

× Treated+β

× (BVAP× Treated)+β

× Party+ϵ (3)

Our analysis of DW-NOMINATE scores reveals that the impact of the Shelby decision has

limited impact on NOMINATE scores, and that interestingly, T reated districts do not have

statistically signiﬁcantly large impact on NOMINATE. We examine whether the racial com-

position of a district or its covered status under the VRA meaningfully predicts the ideology

of its representative. Black VAP share is not a signiﬁcant predictor of DW-NOMINATE

scores (

β = -0.110, p = 0.172), and the interaction between BVAP and covered state status

is similarly insigniﬁcant (

β = -0.136, p = 0.245). This means that packing more Black voters

into a district does not seem to produce a more ideologically diﬀerent representative, and this

null relationship holds equally in covered and non-covered states. The one notable ﬁnding is

that representatives from covered states carry a conservative baseline independent of party

and district composition (

β = 0.099, p ¡ 0.001), but this reﬂects regional political culture

rather than anything speciﬁc to redistricting. Taken together, the ideology analysis suggests

that while covered states pack minority voters into extreme outlier districts, the primary

consequence is the structural containment of minority voting power, not a measurable shift

in the ideological character of the representatives those districts produce.

Variable Estimate Std. Error t p

Intercept -0.37664 0.01526 -24.675 <0.001 ∗ ∗ ∗

Black VAP Share -0.10961 0.08008 -1.369 0.172

Treated 0.09890 0.02407 4.109 <0.001 ∗ ∗ ∗

Party 0.88153 0.01569 56.187 <0.001 ∗ ∗ ∗

Black VAP Share × Treated -0.13579 0.11668 -1.164 0.245

Observations 307

Residual Std. Error 0.149

F -statistic 1353

∗∗∗

∗

p < 0.05;

∗∗

p < 0.01;

∗∗∗

p < 0.001. Dependent variable: DW-NOMINATE ﬁrst dimension score.

Table 3: BVAP Share and Ideological Extremity

5 Conclusion

The empirical results from our Diﬀerence-in-Diﬀerences analyses indicate that the Shelby

County v. Holder (2013) decision did not fundamentally alter the mechanics of district pack-

ing in formerly covered states. The statistical insigniﬁcance of our primary DiD estimator for

max BVAP (β

= −2.005, p = 0.154) suggests that covered states largely maintained their

pre-existing redistricting strategies into the 2020 cycle. Rather than triggering a new wave

of racial concentration, the removal of Section 5 preclearance appears to have had no eﬀect

on the existing packing equilibrium that was already ﬁve standard deviations removed from

neutral simulation baselines in 2010, itself a legacy of the DOJ’s Max-Black strategy and

the no-retrogression standard which, institutionalized super-majority districts across covered

jurisdictions (Lublin 1997). The null eﬃciency gap result (β

= 0.904, p = 0.408) further

suggests that Shelby did not produce a detectable shift in partisan packing either, imply-

ing that the structural geometry of these maps remained largely static across the decennial

plans.

The ideological analysis reinforces the seemingly little inﬂuence of the Shelby decision. While

representatives from covered states are systematically more conservative than peers in non-

covered states (a novel result indeed) controlling for party and district composition (p <

0.001), the interaction between BVAP share and covered status is insigniﬁcant (p = 0.245).

This aligns with Gay 2007’s framework of legislating without constraints: racial packing

in these states functions primarily as a mechanism of containment, cordoning oﬀ minority

voting blocs into super-majority districts and liberating surrounding representatives from

accountability to minority interests, rather than producing a measurable radicalization of

the representatives elected from those packed districts themselves. The ideological signal is

driven by party and regional alignment, not by the precise degree of packing.

The more consequential eﬀects of Shelby likely lie outside district maps, as the literature

does explore. As Komisarchik and White 2020 demonstrate, the removal of preclearance en-

abled a suite of procedural changes including stricter registration requirements, polling place

closures, voter ID laws, all which may widen the gap between Black and White voter turnout

in formerly covered jurisdictions over time. Trautman and Smith 2025 further establishes

that the structural conﬁguration of districts directly dampens grassroots mobilization and

psychological electoral engagement, suggesting that the harm of the existing packing equi-

librium compounds over time even absent new gerrymandering. Under the issue attention

cycle (Downs 1972), the mobilization campaigns that Cantoni and Pons 2021 ﬁnd oﬀset early

disenfranchisement eﬀorts are unlikely to sustain their eﬀectiveness indeﬁnitely as restrictive

laws become further entrenched. The long-run costs of Shelby, in other words, may be less

visible in decennial map comparisons than in the gradual erosion of participatory democracy

in covered jurisdictions.

Ultimately, these results underscore that while the Shelby decision did not drastically worsen

the state of redistricting in covered jurisdictions, the baseline reality revealed by this analysis

remains deeply concerning from a fairness perspective. Covered states consistently exhibit

levels of racial packing that are extreme statistical outliers relative to race-neutral simu-

lations, concentrating minority voting power in ways that dilute political inﬂuence across

a broader set of seats, precisely the dynamic that Lublin 1997 identiﬁes as the tradeoﬀ of

descriptive representation. Whether or not these conﬁgurations survive legal scrutiny in

the future, they represent a structural inheritance of decades of federally incentivized con-

centration that Shelby has now left without meaningful oversight. The persistence of these

outliers suggests not that the decision changed the rules of the game, but that it removed the

referee, leaving in place a system in which minority communities in historically repressive

jurisdictions retain signiﬁcantly less political leverage than a neutral redistricting process

would aﬀord them.

Author’s Note:

No Artiﬁcial Intelligence (AI) was utilized in the ideation behind this project, the analysis

of our data, or writing. However, I am bad at LaTeX, and some of the formatting was done

with AI tools, speciﬁcally Claude.ai.

References

Becker, Amariah, Moon Duchin, et al. (2021). “Computational Redistricting and the Voting

Rights Act”. In: Election Law Journal: Rules, Politics, and Policy 20.4, pp. 407–441. doi:

10.1089/elj.2020.0704.

Becker, Amariah and Dara Gold (Oct. 2022). “The Gameability of Redistricting Criteria”.

In: Journal of Computational Social Science. doi: 10.1007/s42001-022-00180-w. url:

https://link.springer.com/article/10.1007/s42001-022-00180-w.

Bernini, Andrea et al. (2024). “Sixty Years of the Voting Rights Act: Progress and Pitfalls”.

In: Oxford Review of Economic Policy 40.3, pp. 486–497. doi: 10.1093/oxrep/grae024.

url: https://academic.oup.com/oxrep/article/40/3/486/7907277.

Bouton, Laurent et al. (2023). “Gerrymandering with Diﬀerential Turnout”. In: National

Bureau of Economic Research (NBER) Working Paper w31442. Analyzes tactical models

of ”pack-crack-pack” inspired by data-driven strategies found in modern redistricting ﬁles.

doi: 10.3386/w31442.

Brennan Center for Justice (June 2023). Shelby County v. Holder. https://www.brennancenter.

org/our-work/court-cases/shelby-county-v-holder. Originally published August 4,

2018; last updated June 25, 2023.

Cantoni, Enrico and Vincent Pons (2021). “Strict ID Laws Don’t Stop Voters: Evidence

from a US Nationwide Panel, 2008–2018”. In: The Quarterly Journal of Economics 136.4,

pp. 2615–2660. url: https://www.nber.org/system/ files/working_papers/w25522 /

w25522.pdf.

Common Cause v. Lewis (2019). Order Lifting Conﬁdentiality Designation on Hofeller Files,

No. 18-CVS-014001. Wake County Superior Court, North Carolina. Includes over 100,000

documents and maps belonging to Dr. Thomas Hofeller. url: https://www.brennancenter.

org/our-work/court-cases/common-cause-v-lewis.

Downs, Anthony (1972). “Up and Down with Ecology: The “Issue-Attention Cycle””. In:

The Public Interest 28, pp. 38–50.

Gay, Claudine (2007). “Legislating Without Constraints: The Eﬀect of Minority Districting

on Legislators’ Responsiveness to Constituency Preferences”. In: Journal of Politics 69.2,

pp. 442–456. doi: 10.1111/j.1468-2508.2007.00542.x.

Gelman, Andrew (2002). “Voting, Fairness, and Political Representation”. In: CHANCE

15.3, pp. 22–26. doi: 10.1080/09332480.2002.10554807.

Hood, M. V. and Scott E. Buchanan (2020). “Palmetto Postmortem: Examining the Eﬀects

of the South Carolina Voter Identiﬁcation Statute”. In: Political Research Quarterly 73.2.

Accessed: May 2, 2026, pp. 492–505. url: http://www.jstor.org/stable/45381403.

Komisarchik, Mayya and Ariel White (2020). Throwing Away the Umbrella: Minority Voting

after the Supreme Court’s Shelby Decision. Working Paper. MIT Center for International

Studies. url: https : / / cis . mit . edu / sites / default / files / paper - documents /

attnZIPgHeEwYhvId.pdf.

Lublin, David (1997). The Paradox of Representation: Racial Gerrymandering and Minority

Interests in Congress. Princeton, NJ: Princeton University Press.

McCartan, Cory, Christopher T. Kenny, Tyler Simko, Emma Ebowe, et al. (2025). “Redis-

tricting Reforms Reduce Gerrymandering by Constraining Partisan Actors”. In: American

Political Science Review. Forthcoming. arXiv: 2407.11336.

McCartan, Cory, Christopher T. Kenny, Tyler Simko, George Garcia, et al. (2022). “Simu-

lated Redistricting Plans for the Analysis and Evaluation of Redistricting in the United

States”. In: Scientiﬁc Data 9.1, p. 689. doi: 10 . 1038 / s41597 - 022 - 01808 - 2. url:

https://www.nature.com/articles/s41597-022-01808-2.

Poole, Keith T. and Howard Rosenthal (2007). Ideology and Congress. 2nd revised. New

Brunswick: Transaction Publishers.

Protect Democracy (Oct. 2023). How Did We Get Here: Understanding the Demographic

Sources of America’s Party Divisions. Chapter in More Than Red and Blue: Political Par-

ties and American Democracy. Accessed: May 2, 2026. url: https://protectdemocracy.

org/work/understanding-the-demographic-sources-of-americas-party-divisions/.

Trautman, Linda and Michael A. Smith (2025). “Redistricting and Black Political Mobiliza-

tion: Implications from the 2018 and 2022 Midterm Elections”. In: Social Sciences 14.2,

p. 60. doi: 10.3390/socsci14020060.

Warrington, Gregory S. (2018). “A Comparison of Partisan-Gerrymandering Measures”. In:

Public Choice 195.1–2, pp. 1–23.