Course 3 of the Demographic Architecture series. Moving from theoretical law into applied quantitative analysis, computational redistricting, and Ecological Inference.

At a Glance

Course

3 of 3 in the series

Length

12 modules

Intended Audience

Analysts, geospatial data scientists, academic researchers, and legal practitioners requiring technical fluency. This quantitative framework aligns with the data science needs of the Brennan Center for Justice, FairVote, and the Democracy Funders Network.

Capstone Artifact

Methodology Portfolio: A structured, multi-scale analytic framework capable of being handed directly to litigation counsel as the technical backbone of a civil rights lawsuit.

Course Purpose

Course 1 defined who is allowed in the building. Course 2 mapped the physical architecture of the building. Course 3 turns on the lights and runs the algorithms.

Modern voting rights litigation is won and lost through data analysis. State legislatures defending gerrymanders rely on the complexity of spatial data to hide their intent behind claims of "natural geographic clustering" or "race-neutral" compactness. To defeat them, practitioners cannot simply argue fairness; they must deploy ecological inference (EI) to prove racially polarized voting, and computational ensembles (MCMC) to prove the mathematical deviation of a map. This course exists to cross the bridge from legal framework to hard data generation, surveying the complex technical methodologies used by expert witnesses to model American democracy.

Key Concepts

Data Degradation

Census data is not static. We track how differential privacy mechanisms and natural demographic drift degrade the fidelity of the public data used to draw maps, forcing analysts to reconstruct reality.

Computational Scale (MAUP)

We take the Modifiable Areal Unit Problem introduced in prior courses and fully operationalize it. We examine the Markov Chain Monte Carlo (MCMC) algorithmic ensembles used to generate 100,000 neutral maps to definitively detect partisan rigging.

Ecological Inference

Because ballots are secret, we only possess aggregate demographic data and aggregate vote counts. We study the statistical methods used to cross-reference these sets to explicitly prove "Racially Polarized Voting" in federal court.

How to Use This Course

Unlike previous courses designed for organizers and policy generalists, this course leans heavily into methodology. While no coding is required, readers are expected to engage with statistical and spatial reasoning. The reading tiers have been adjusted accordingly:

Conceptual Readings: Texts explaining the intuition behind the analysis. Why do we need this algorithm, and what does it prove in court?
Methods Readings: Applied examples of the algorithms running in the wild against real demographic data.
Technical Reference: Direct access to the codebases, mathematical notation, or deep-dive federal expert reports required for practitioners actively building their Methodology Portfolios.

View Full Course Bibliography

Module List

01

Architect and Builder
The framework and ethical constraints guiding the translation of human rights into pure data.
02

Scale 1: Census Infrastructure
Auditing the bedrock data source: differential privacy, PL 94-171, and block-level geography.
03

Scale 1: Data Degradation and Reconstruction
Handling the decay of data between Census years and the spatial interpolation methods required to fix it.
04

Scale 2: Geography of Participation
Transitioning from demographic data to behavioral data (turnout deserts, precinct consolidation).
05

Scale 2: Measuring Suppression
A technical post-mortem of Maricopa County 2016. How to quantitatively measure administrative friction.
06

Scale 3: MAUP and Redistricting Fundamentals
Operationalizing the Modifiable Areal Unit Problem through mathematical limits (compactness scores).
07

Scale 3: Computational Redistricting
The core technical block on algorithmic ensembles and Markov Chain Monte Carlo (MCMC) / ReCom methodologies.
08

Scale 3: Running Case Block I
Applying ensemble analysis directly to the North Carolina baseline to isolate systemic bias.
09

Scale 4: VRA Framework
Preparing data explicitly to satisfy the rigorous three-part Gingles test required under Section 2.
10

Scale 4: Ecological Inference and RPV
The statistical mechanics of Ecological Inference used to prove Racially Polarized Voting.
11

Scale 4: Running Case Block II
The Alabama (Allen v. Milligan) tracking sessions, modeling how EI and VRA math survived the Supreme Court.
12

Synthesis: Multi-Scale Analysis
Combining all scales into the Methodology Portfolio, and acknowledging the limits of mathematical truth in hostile courts.

Core Concepts & Inquiries

What is Ecological Inference (EI) in voting rights analysis?

Ecological Inference is a statistical method used to estimate individual-level behavior (such as how different racial groups voted) from aggregate-level data (such as precinct-level vote totals and demographic counts). It is the primary tool used to demonstrate Racially Polarized Voting in court.

How are MCMC ensembles used in redistricting litigation?

Markov Chain Monte Carlo (MCMC) ensembles are used to generate thousands or millions of alternative, algorithmically "neutral" maps that follow specific legal constraints. By comparing a challenged map to this ensemble, analysts can determine if the challenged map is a statistical outlier, indicating partisan or racial gerrymandering.

What is Census 'Data Degradation'?

Data degradation refers to the loss of accuracy in Census data over time due to population drift, as well as the intentional introduction of "noise" through differential privacy mechanisms (like the TopDown Algorithm) to protect individual confidentiality.

What is PL 94-171 data?

PL 94-171 is the specific public law that requires the Census Bureau to provide states with the small-area population counts (down to the block level) necessary for legislative redistricting.

What is the role of spatial interpolation in democratic analysis?

Spatial interpolation is used to estimate data for geographic units where information is missing or where boundaries (such as precinct lines and Census blocks) do not align, allowing for the reconstruction of a consistent dataset for analysis.

Capstone: The Methodology Portfolio

Through this course, your engagement actions will transition away from diagnostic audits and toward building a structured Methodology Portfolio.

This living document serves as the architectural blueprint for any civil rights lawsuit. By the end of Module 12, you will have a comprehensive manual stating exactly which spatial data is required, the precise algorithmic processes (Ensemble/EI) that must be run, and the explicit legal frameworks under which those results will be submitted as expert testimony.

Are you ready to run the numbers? Start Module 1.

Course 3: Democratic Analysis