Welcome to the Congressional Representation Project!

Name: TCRP U.S. House Elections Dataset (1976–2024)
Creator: The Congressional Representation Project (TCRP)
License: https://creativecommons.org/publicdomain/zero/1.0/

This site is an educational hub for understanding U.S. House representation from 1976 to 2024, covering how votes translate into seats, where elections are competitive, and how those patterns change over time and across states.

Explore the data through an interactive map, trend charts, and downloadable project files. Metrics such as fairness, competition, contestation, and the Electoral Health Index (EHI) summarize statewide patterns in plain numbers, with optional raw and adjusted views when uncontested races would otherwise distort the picture.

Start here:

About This Project — why representation is measured this way and what fair representation means in the data.
Key Findings — six national patterns drawn from the dataset before you explore the tools.
Interactive Map — pick a year and metric, click a state for district-level detail, and double-click to return to the national view.
Visualizing Trends — chart one or more metrics across years for the nation or a single state; compare raw vs. adjusted results.
Case Studies — in-depth looks at North Carolina (1992), Massachusetts (2024), the 2010 redistricting cycle, and Texas (2003–2004).
Static Visualizations — five pre-built national charts with CSV downloads.
Raw Data Browser — preview the electoral dataset, source CSVs, and the metric calculation script.
Details & Sources — metric glossary, data downloads, limitations, changelog, and credits.

Important disclaimer: This project is an independent, non-partisan data science initiative designed to visualize structural election data. It does not take advocacy positions, nor is it funded by any political organization.

About This Project

Most people assume their vote automatically translates into representation. In practice, district lines, uncontested races, and election competitiveness all help determine whether that translation is fair. This project exists to measure those structural factors with data, not opinion.

In U.S. House elections, each district uses winner-take-all rules: a candidate who wins 51% of the vote and one who wins 99% both receive exactly one seat. Votes beyond what is needed to win, and all votes cast for losing candidates, produce no additional representation. When races are uncontested or effectively safe for one party, accountability weakens further. Voters have fewer meaningful choices, challengers face steep hurdles, and statewide scores for competition and contestation fall even when a state's overall politics look stable.

Fair representation, as used on this site, means vote totals and seat outcomes line up reasonably well, elections stay competitive, and results are not dominated by uncontested or one-sided districts. The Congressional Representation Project (TCRP) is an independent, non-partisan data science resource covering 1976 to 2024. Every metric, dataset, and script is open for download, inspection, and reuse so researchers, journalists, and citizens can judge for themselves whether they are fairly represented.

Key Findings

Six national patterns from the district-weighted metrics in TCRP’s electoral dataset (1976–2024). Each point summarizes the full country, not any single state or district.

National fairness is down nearly 10 points since 1976. District-weighted fairness averaged 48/100 in 1976 and 38/100 in 2024. That reflects a long-run slide in how evenly House votes translate into seats. Explore year-by-year values on the interactive map (Fairness) or Visualizing Trends chart.
Electoral Health Index peaked in 1992 and has not fully returned. National EHI reached 77/100 that year, the high-water mark in the dataset, and even the relatively strong 2024 cycle landed at 72/100. See the long-run arc in Visualizing Trends and the 2024 state ranking in Static Charts.
Every post-1980 redistricting cycle coincided with a national fairness drop. Fairness fell in the first election under each new Census map from 1982 through 2022; the steepest single hit was −10 points from 2000 to 2002. Compare before-and-after bars in Static Charts (Redistricting Cycle Impact).
Metric confidence usually dips when a new map takes effect. Confidence fell after four of five post-1980 redistricting cycles (1982, 2002, 2012, and 2022), often when contestation data are still settling. Track the pattern in Visualizing Trends (Confidence) or on the interactive map.
The 2010 redistricting cycle sharply rewired the national seat–vote balance. In 2008, Democrats held 3.5 points more seat share than vote share nationwide; by 2012, Republicans held 4.4 points more seats than votes even though Democrats still won a slim nationwide House vote majority (~51%). That is a swing of nearly 8 points in who the map advantaged. Watch the two lines in Static Charts (Seat Share vs. Vote Share).
Uncontested races still weaken competition in many cycles. As many as 22% of districts lacked major-party opposition in 1998 (the dataset high); recent elections have run closer to 6–10%, but safe seats and lopsided wins still pile up wasted votes in both parties. Follow the trend in Static Charts (Uncontested Races Over Time).

Usage

Quick example: Select 2024 and Electoral Health Index. Then click a state.

If that state’s EHI is high, it usually means fairness, competition, contestation, and stability are all relatively strong for that year.
Switch the metric to Fairness and Competition to see why the overall score looks the way it does.
The Metric Information section below the map updates automatically with the formula and any academic source for the selected metric.
Turn on Display Table to inspect district-level results and spot patterns (many close races vs. many lopsided races).
If lots of races are uncontested, enable Adjusted Metrics and compare: big shifts suggest raw results are being dominated by uncontested districts.
Click the Download table as CSV button to download the current table, whether it be state election results or national metrics.

Takeaway: This is best used to determine why a metric is high or low for a given state or year by comparing that metric to each district's congressional election results.

Taking a Closer Look

The metrics and election results shown in the map are derived from our Electoral Dataset, which is available in the Sources section. The map adds geographic context, allowing you to compare metrics nationally and then drill down to individual states and congressional districts.

Some districts have shapes that are widely cited as unusual or controversial, but shape alone is not proof of gerrymandering or political intent. One way to use this view is to explore well-known redistricting controversies as case studies. For example, Louisiana’s 4th district (1992) is often discussed for its "Zorro"-like shape, while North Carolina’s 12th district (1992) was central to the Supreme Court case Shaw v. Reno. The map can help visualize these districts while the metrics provide additional context for comparison.

Some Louisiana (LA) elections may appear unusual because of the state's use of nonpartisan blanket primaries, often called "jungle primaries." Instead of separate party primaries, all candidates compete on the same ballot, and the top two may advance to a runoff regardless of party. In some historical elections, comparable Democratic-versus-Republican vote totals are not available, so the dataset records only the winning party. On the map and in the data, these elections may appear as 1s and 0s rather than conventional vote totals.

District boundaries are also shaped by legal and political constraints. Common factors include the Voting Rights Act of 1965 (VRA), equal-population requirements, communities of interest, compactness criteria, and whether maps are drawn by legislatures, commissions, or courts. Because these factors often interact, the map and metrics should be viewed as tools for exploration and comparison rather than a definitive judgment of any single cause.

Apportionment can also have a major impact on representation. After each Census, congressional seats are redistributed among the states based on population changes. States may gain or lose seats, requiring district boundaries to be redrawn even when political goals remain unchanged. Population growth, decline, and demographic shifts can all influence how districts are configured and how competitive elections become.

Massachusetts (MA) in 2024 illustrates how limited major-party competition can affect headline metrics. When many races are effectively uncontested, measures of competition and contestation can decline, lowering the overall EHI. The example also highlights a dataset choice: independent and third-party votes are kept separate rather than reassigned to either major party, even when an independent is a significant challenger.

Visualizing Trends

Select one or more metrics to chart across years for the nation or a state. Use State Comparison to overlay a second state, or compare a state to national scores.

Usage

The chart can be used to analyze trends and correlations across years for the nation or a state.

Check each metric you want to chart by selecting it in the Metrics panel. You can select multiple metrics to find correlations and determine how they influence each other.
Select a State/National to compare state or national metrics. Using the Interactive Map, you can compare a state's metrics to it's rank and previous ranks in the table.
Select a Metric Type to compare raw, adjusted, or both metric types. Adjusted metrics are adjusted for uncontested or missing-opposition races.
Select a National Weighting to compare district-weighted or state-weighted metrics. District-weighted metrics give states with more districts more weight.
Select a Start Year and End Year to limit the range of years to chart. This can be used to find and compare short-term trends.

Takeaway: This is best used for comparison and pattern-spotting across years/states. If a metric looks extreme, check Confidence and compare raw vs adjusted before drawing conclusions.

Taking a Closer Look

The metrics shown in the chart are derived from our Electoral Dataset, which is available in the Sources section. The chart lets you compare national or state trends across years. The patterns below use district-weighted national averages of raw metrics as a starting point for exploration. They are not proof of any single cause.

Nationally, the Electoral Health Index (EHI) moves more in cycles than in a straight line. It peaks near 77 around 1992, when contestation and competitiveness were both relatively strong, then falls to a low near 67 in 2002. That year was the first House election after the 2000 Census redistricting cycle, and several states (including Nebraska, Massachusetts, and Louisiana) show unusually weak competition in the data. EHI rebounds again in high-turnout years such as 2020 (near 76) before settling near 72 in 2024. Over the full 1976 to 2024 span, the national EHI ends only slightly above where it began, which is a reminder that “healthier” and “less healthy” election years can alternate even when the long-run average looks stable.

Fairness tells a different story. The national fairness score drifts down from about 48 in 1976 to about 39 in 2024, with sharp dips that often follow redistricting. Fairness drops noticeably after the 2000 and 2010 to 2012 map cycles, when many states were still running new district lines from the prior Census. It rises in 2020 (to about 49) alongside broader two-party competition in a presidential year, then falls again by 2024 as maps from the 2020 Census cycle take full effect. States such as Illinois, Colorado, and Arizona show some of the largest fairness declines between 2018 and 2024 in the dataset. That timing lines up with real-world redistricting fights, court challenges, and partisan mapmaking debates. The metric still measures vote-to-seat balance in the data, not legal intent.

Confidence is usually high nationally (often above 85), but it is worth watching when it dips. The national average falls to about 82 in 2014, when many states are flagged as post-redistricting and several have low contestation, including Massachusetts, Georgia, and Louisiana. Confidence recovers in wave years like 2020 (near 96) when more districts have meaningful two-party vote totals, then eases again by 2024 (near 86). That pattern fits how the score is built: confidence falls when uncontested races, heavy reliance on adjusted values, or fresh map changes make statewide metrics harder to read reliably.

These national lines are averages, so they can hide dramatic state stories. Use the chart’s State/National control to zoom in on a single state, compare raw and adjusted lines, and check Confidence in the same year range before tying a trend to a specific election or redistricting event. For deeper dives (North Carolina’s 12th in 1992, Massachusetts in 2024, the 2010 national cycle, and Texas mid-decade redistricting), see Case Studies below.

Case Studies

Four episodes where real-world redistricting fights, legal milestones, or uncompetitive delegations show up in TCRP’s metrics. Each follows the same structure: what happened, what the numbers say, how to explore it on the site, and what the data cannot prove on its own.

North Carolina’s 12th District (1992)

1. Context

After the 1990 Census, North Carolina added a twelfth House seat and drew one of the most debated maps in modern redistricting. The new 12th district linked Black communities across a narrow, interstate corridor so the state could elect a second Black member of Congress under the Voting Rights Act era rules then in force. Mel Watt won the seat in 1992, the first election under that map.

The district’s shape became a national flashpoint. In Shaw v. Reno (1993), the Supreme Court held that race could not be the predominant factor in drawing lines, even when the goal was minority representation. The case did not erase the 1992 results, but it opened years of litigation and map revisions that reshaped North Carolina politics for decades.

2. What the data shows

Statewide, North Carolina’s metrics moved sharply when the 1990-cycle map first took effect:

Raw fairness fell from 52.3/100 (1990) to 38.3/100 (1992), a 14-point drop in vote-to-seat balance.
The efficiency gap score fell from 62.7 to 19.5, signaling much more asymmetric wasted votes statewide.
Contestation stayed at 100% in both years; every district had two-party vote totals on the ballot.
State EHI eased from 82.5 to 78.0; competitiveness remained high (93.1 → 92.3).

In the 12th itself (1992): Mel Watt took 72.0% of the vote in a contested race (176,664 ballots cast). District competitiveness scored 55.9/100: competitive by formula, but still a clear Democratic hold. The controversy was geographic and legal, not a lack of opposition on the ballot that year.

By 1994, the state was flagged post-redistricting in the dataset; fairness recovered somewhat to 50.9 while confidence dipped to 85.8 as maps continued to move through courts.

3. What the map looks like

On the interactive map, select 1992, choose Fairness or Electoral Health Index, and click North Carolina to open the district view. District 12 appears as a strong Democratic seat with full two-party contestation in that year.

In Visualizing Trends, chart Fairness or Efficiency Gap for North Carolina from 1988 to 1996 to see the 1992 dip against neighboring even years. Compare raw and adjusted lines if you extend into the 2000s, when court-ordered changes stack on top of the original 1990-cycle story.

4. What the data can and cannot tell you

Can: Show that the first election under the 1990-cycle map coincided with a large statewide fairness shift and a highly Democratic, yet contested, 12th district. The metrics summarize vote totals and seat outcomes; they align with the timing of a map that drew national legal scrutiny.

Cannot: Prove why the map was drawn, whether the 12th was “fair” in a legal sense, or how much of the 1992 fairness drop belongs to the 12th alone versus the rest of the North Carolina plan. Shape, intent, and VRA compliance are courtroom and historical questions. TCRP describes election structure in the data, not the motives of mapmakers or judges.

Massachusetts (2024)

1. Context

Massachusetts is one of the most Democratic states in presidential and statewide elections, but its House delegation is not guaranteed to be unanimous in every cycle. In 2024, all nine incumbents won re-election, and the Republican Party did not field a candidate in seven of nine districts. Voters in those seats had no major-party choice on the ballot even though the state as a whole is competitive at the top of the ticket.

That pattern reflects recruitment, safe-seat strategy, and one-party dominance at the district level more than a single controversial map line. It still matters for representation metrics: when opposition is missing, raw vote shares and competition scores can look distorted compared to what a contested two-party race would produce.

2. What the data shows

Massachusetts posted the lowest raw EHI in the country in 2024 (22.9/100). Other statewide raw scores that year:

Fairness: 18.3/100
Competitiveness: 20.2/100
Contestation: 22.2/100 (two of nine districts contested)
Metric confidence: 51.1/100, among the lowest nationally because uncontested races dominate the delegation

Seven districts show 100% raw Democratic vote share with no Republican opponent. TCRP’s adjusted pipeline imputes a plausible two-party split (roughly 73.4% Democratic in those seats using statewide and presidential context). With adjusted metrics enabled:

Adjusted EHI rises to 37.0/100 (still weak, but less extreme).
Adjusted competitiveness rises to 57.6/100.
Adjusted fairness moves to 27.8/100.

Only districts 8 and 9 were contested in 2024 (70.6% and 56.5% Democratic raw vote share, respectively). The gap between raw and adjusted lines is the signature of this case: headline metrics punish missing opposition, while adjusted views estimate counterfactual competition.

3. What the map looks like

On the interactive map, select 2024 and Electoral Health Index. Massachusetts colors among the weakest nationally. Click the state, turn on Display Table, and scan district rows: most show uncontested races and lopsided raw totals.

Toggle Use Adjusted Metrics in the map sidebar and watch statewide scores shift. Mirror the same setup in Visualizing Trends: chart EHI, Contestation, and Confidence for Massachusetts with Metric Type: both to compare raw versus adjusted paths across years.

4. What the data can and cannot tell you

Can: Quantify how a delegation with almost no Republican ballot presence produces rock-bottom contestation and confidence, and show how adjusted estimates change the picture. That is a concrete illustration of why TCRP publishes both raw and adjusted views.

Cannot: Show whether Massachusetts “should” run more competitive races, whether Democrats or Republicans are responsible for missing opponents, or whether district lines (rather than party strategy) caused the 2024 pattern. Adjusted values are model-based estimates, not recorded votes. Low EHI here is a structural signal about competition on the ballot, not a verdict on gerrymandering or voter preferences.

The 2010 Census Redistricting Cycle (National)

1. Context

The 2010 Census triggered a new round of congressional maps nationwide, with state legislatures, commissions, and courts fighting over lines through the early 2010s. The first House election under most new plans was 2012, a presidential year with high turnout but also fresh gerrymanders in several large states.

Real-world fights in Wisconsin, Michigan, Pennsylvania, North Carolina, and elsewhere made this cycle a reference point for partisan mapmaking debates. The dataset captures the national footprint of that transition: not any one state’s courtroom story, but the combined effect of dozens of new maps landing in the same election year.

2. What the data shows

National district-weighted averages (raw metrics):

Fairness fell from 47.4/100 (2010) to 42.1/100 (2012), a 5.3-point drop in the first election under new Census maps.
24 of 43 states with comparable data saw fairness decline from 2010 to 2012; the median state lost about 7.2 points.
Electoral Health Index dipped from 75.6 (2010) to 71.4 (2012) before sliding to 67.2 by 2014.
Metric confidence was still high in 2012 (94.5) but fell to 81.5 by 2014 as post-redistricting flags and uneven contestation accumulated.

Some of the largest state-level fairness drops from 2010 → 2012 include Wisconsin (−64.8 points, 80.5 → 15.7), Michigan (−44.3), Missouri (−41.0), and Indiana (−39.8). Those states anchor much of the national average’s movement even when other states moved modestly or ticked up.

The seat–vote picture flipped nationally. Democrats held 3.5 points more seat share than vote share in 2008; by 2012 Republicans held 4.4 points more seats than votes despite a slim Democratic nationwide House popular-vote majority (~51%). That ~8-point swing in map advantage appears in the Static Charts seat-share versus vote-share graphic.

3. What the map looks like

On the interactive map, select Fairness and toggle between 2010, 2012, and 2014 to watch the national color scale shift. Click individual states (especially Wisconsin, Michigan, or Indiana) to compare district-level results tables before and after.

In Visualizing Trends, chart national Fairness and Confidence from 2008 to 2016. The Redistricting Cycle Impact chart plots national fairness in the election before versus after each Census cycle, including the 2010 → 2012 pair.

4. What the data can and cannot tell you

Can: Document that the 2012 election coincided with a broad national fairness decline and a sharp seat–vote swing toward Republicans in the aggregate data, matching the timing when 2010-cycle maps took hold. State-level drops help explain which map fights moved the national average most.

Cannot: Attribute any single state’s shift to gerrymandering alone, separate presidential coattails from line-drawing in 2012, or replace legal findings from state courts. Commissions, incumbency, and uncontested races also move fairness. This case study is a national benchmark for the cycle, not a substitute for state-by-state investigations (including those planned for other case studies on this page).

Texas Mid-Decade Redistricting (2003 → 2004)

1. Context

Most states redraw congressional lines once per decade after each Census. Texas did so twice in the 2000s. After Republicans won the state legislature in 2002, U.S. House Majority Leader Tom DeLay pushed a mid-decade replan that replaced the court-drawn 2002 map before the next Census. The new lines were in place for the 2004 election.

The move was unusually aggressive and heavily litigated (League of United Latin American Citizens v. Perry, 2006, among others). It remains a textbook example of a party using a second redraw to consolidate seat share without waiting for 2010.

2. What the data shows

Texas delegation outcomes shifted dramatically between the two even-year elections under different maps:

2002 (prior map): 17 Democratic seats, 15 Republican seats (32 total). Democratic candidates won 45.1% of statewide House ballots.
2004 (DeLay map): 11 Democratic seats, 21 Republican seats. Democratic candidates won 40.3% of statewide House ballots.

Republicans gained six seats while their statewide vote share rose roughly 5 points in a year that also featured George W. Bush atop the Texas ticket. Seat share moved much faster than vote share: Democratic seat share fell from 53% to 34% of the delegation.

Statewide raw metrics (both years flagged post-redistricting in the dataset because each map was relatively new):

Raw fairness: 56.9/100 (2002) → 45.9/100 (2004).
Raw EHI: 63.8 → 65.3 (competition and contestation ticked up even as fairness fell).
Contestation: 71.9% → 78.1%; competitiveness: 62.6 → 70.0.
Structural stability rose from 68.0 to 74.6 as incumbency insulation eased slightly (32.0 → 25.4 on the insulation index).

Adjusted metrics tell a parallel story: adjusted fairness was 79.4/100 in 2002 but 40.9/100 in 2004, suggesting the raw and adjusted pipelines disagree more once the new map settles and district-level patterns shift. Always compare both when interpreting Texas in this window.

3. What the map looks like

On the interactive map, select Texas, choose Fairness or Efficiency Gap, and flip between 2002 and 2004 in the year control. Open the district table to see how many seats flipped party and which districts became safer.

In Visualizing Trends, plot Fairness, Structural Stability, and Electoral Health Index for Texas from 2000 to 2008 with Metric Type: both to bracket the mid-decade redraw against the surrounding Census-cycle elections.

4. What the data can and cannot tell you

Can: Show a large, fast seat swing that aligns with a rare mid-decade map change, plus a clear raw fairness decline from 2002 to 2004. Stability and efficiency-gap components move in ways that help you ask whether the map became more structurally locked in for one party even when individual races still saw two-party ballots.

Cannot: Isolate the DeLay redraw from 2004 presidential coattails in Texas, prove illegal gerrymandering (courts later addressed specific districts and claims on their own terms), or capture the full Latino voting-rights story in a single fairness score. Mid-decade replans are rare; Texas in 2004 is a powerful example, but not a template for every state in every decade.

Static Visualizations

Pre-built views of nationwide patterns from the electoral dataset (1976–2024). Each chart highlights one structural story; use the interactive chart above to drill into states or metrics.

Seat Share vs. Vote Share

When Democrats win a smaller share of House seats than their share of total votes, the gap between the two lines is the most direct picture of vote-to-seat imbalance. This chart aggregates every congressional district nationally and plots Democratic vote share and Democratic seat share across even-year elections from 1976 to 2024.

The shaded area between the lines is that gap in percentage points. A persistent seat line below the vote line means Democratic votes are translating into fewer seats than a proportional system would produce; the reverse would indicate a Republican vote-seat shortfall.

Wasted Votes Over Time

In winner-take-all districts, votes that do not help elect a candidate are wasted: all votes for losing candidates, plus any winning votes beyond what was needed to win (50% + 1 of the district total). This is the same intuition behind the Efficiency Gap metric on the map.

The stacked areas show estimated Democratic and Republican wasted votes nationwide each election year. When one party’s wasted total consistently dominates, statewide seat outcomes can diverge from overall vote share even without any single “smoking gun” district. Safe seats and blowout wins inflate wasted votes on both sides but often unevenly.

Uncontested Races Over Time

A district counts as uncontested when one major party is effectively missing from the race, with no meaningful two-party opposition on the ballot. The line shows what percentage of all House districts nationally fit that description in each even year.

Rising uncontested rates weaken competition and contestation scores and are a major reason the site offers adjusted metrics. When many races lack opposition, raw vote totals alone can make a state look more competitive than voters actually experienced. Watch this line alongside Confidence and EHI in the interactive chart.

State EHI Rankings (2024)

The Electoral Health Index (EHI) combines fairness, competitiveness, contestation, and structural stability into one 0–100 score. This lollipop chart ranks all states by raw EHI in 2024, giving an immediate snapshot of outliers without clicking the map.

Longer bars mean higher overall electoral health for that year’s House elections. States at the bottom often combine low contestation, weak competition, or fairness gaps; Massachusetts and Louisiana frequently appear in that tier in recent cycles. Compare this snapshot to the interactive map or trend chart to see whether a low score is a one-year blip or a longer pattern.

Redistricting Cycle Impact

After each Census, states draw new congressional maps. The grouped bars compare the national district-weighted average fairness score in the election before a new map (e.g., 2010) with the first election after it (e.g., 2012) for five post-1980 cycles: 1982, 1992, 2002, 2012, and 2022.

Drops after a cycle suggest the new maps coincided with less balanced vote-to-seat translation nationwide. That is not proof of gerrymandering in any one state, but it is a concrete before/after benchmark. Individual states can move in the opposite direction; use the map and state comparison tools to see which states drove each national shift.

Raw Data Browser

Preview the processed electoral dataset, source CSVs, and metric calculation script. All files are available to download under Data sources and downloads in the Details & Sources section.

Select a tab to preview a file.

Details & Sources

This section is for readers who want the data sources, technical caveats, and a quick site index.

Quick index

About This Project — why representation is measured this way and what fair representation means in the data.
Key Findings — six national patterns drawn from the dataset.
Interactive Map — statewide and district-level metrics by year.
Visualizing Trends — chart metrics over time for the nation or a state.
Case Studies — four in-depth redistricting and representation episodes.
Static Visualizations — five pre-built national charts with CSV downloads.
Raw Data Browser — preview project data files and scripts.
Details & Sources — overview of reference material on this page.
Metric glossary — formulas and sources for every map metric.
Data sources and downloads — datasets, GeoJSON, and direct file links.
Exceptions and limitations — what these statistics can and cannot show.
Changelog — site updates and release notes.
Credits & disclosure — AI use, contact, and corrections.

Metric glossary

Definitions, formulas, notes, and academic sources for every metric available on the interactive map. The Metric Information panel below the map shows the same content for whichever metric you have selected.

Data sources and downloads

All project data lives under data/. You can preview most files in the Raw Data Browser or download them directly below. Site code is licensed under MIT; dataset licensing is listed per source.

electoral-dataset.json

Processed dataset powering the interactive map, trend charts, and tables. Includes state and district vote totals, raw and adjusted metrics, and presidential context by state/year.

Download:

License: CC0 (public domain)

Attribution: electoral-dataset.attribution
mit-1976-2024-house.csv

Raw U.S. House candidate-level results, 1976-2024.

Download:

Source: MIT Election Data and Science Lab

License: CC0 (public domain)

Attribution: mit-1976-2024-house.attribution
princeton_CE_results.csv

Congressional election results by state, year, and district (Dem/GOP votes, vote share, incumbent). Used as input to metric calculations.

Download:

Source: Princeton Gerrymandertests

License: CC0 (public domain)

Attribution: princeton_CE_results.attribution
dailykos_GE_results.csv

State-level presidential election results (Dem%, Rep%, swing) used when estimating vote shares in uncontested House races.

Download:

Source: Stephen Wolf and David Nir, Daily Kos Elections (2021) — spreadsheet

Attribution: dailykos_GE_results.attribution
compute_metrics.py

Reads the Princeton and Daily Kos CSV inputs, cleans and imputes uncontested races, calculates all statewide metrics (EHI, fairness, competition, and others), and writes the electoral dataset. Comments in the file label each metric calculation; preview it in the Raw Data Browser.

Download:

License: MIT (site code)
manifest.json

Index mapping each state (by FIPS code) and election year to the correct boundary file for that redistricting cycle.

Download:
District GeoJSON files

One file per state and map era (325 files), named {FIPS}_{startYear}_{endYear}.geojson (for example, 37_2022_2022.geojson for North Carolina in 2022). These files are loaded on demand by the map; browse the data/geojson/ folder to download individual files.
LICENSE

MIT License for boundary data.

Download:

Source: UCLA Congressional District GeoJSON (Jeffrey B. Lewis et al.)
LICENSE

CC0 license text for corresponding datasets.

Download:
LICENSE

MIT License for this website, front-end code, and project scripts (including compute_metrics.py).

Download:

License: MIT

Citation: When using the processed electoral dataset in research, journalism, or derivative work, please cite The Congressional Representation Project (TCRP) (crproject.org).

Exceptions and limitations

This site uses statistical summaries of election results to help compare states and years. Those numbers are useful for exploration and pattern-spotting, but they are not a complete picture of representation, intent, or legality. Below are important limits of the approach and real-world factors the metrics do not fully capture.

What statistical analysis can and cannot show

Metrics are not proof of gerrymandering or bad intent. A low fairness score or unusual district shape can reflect many causes, including geography, the Voting Rights Act, population constraints, court orders, or partisan mapmaking. The site highlights patterns in vote-to-seat balance and competition; it does not determine why a map looks the way it does.
Scores compress complex elections into single numbers. Efficiency gap, mean-median, proportionality, and EHI summarize statewide patterns. They can hide local stories, such as one competitive urban district inside an otherwise safe delegation.
Adjusted metrics rely on estimates. When races are uncontested, the script imputes plausible two-party vote shares using statewide House results, presidential results, and neighboring districts. That improves comparability but introduces modeling choices raw vote totals alone do not contain.
Confidence scores flag uncertainty, not guilt. Lower confidence often means many uncontested races, fresh post-redistricting maps, or heavy use of adjusted values; it does not mean the underlying election was irregular.
Cross-year comparisons are approximate. District boundaries change after each Census. A state's metrics in 2010 and 2020 may refer to different geographies, so long-run trends should be read cautiously around redistricting cycles.

Real-world factors this analysis does not fully model

Geography and communities. Mountain ranges, coastlines, and urban/rural splits naturally cluster voters. A compact, geographically coherent map can still produce lopsided seat outcomes, and a bizarre-looking shape is not, by itself, evidence of manipulation. Metrics here do not measure compactness, communities of interest, or how well boundaries follow local geography.
Legal and political mapmaking constraints. Maps must satisfy equal population, comply with the Voting Rights Act, respond to court rulings, and follow state rules about commissions vs. legislatures. Those forces can simultaneously improve representation for some groups and reduce headline competitiveness, which a single fairness number cannot disentangle.
Apportionment and seat counts. When a state gains or loses seats after a Census, districts are redrawn even if partisan goals stay fixed. Shifts in competitiveness or fairness may reflect seat reallocation and population change, not a new gerrymander.
Unusual election systems. Louisiana's nonpartisan blanket ("jungle") primaries sometimes yield races without comparable Democratic-Republican vote totals. In those cases the dataset may record only a winning party, and district totals can appear as 1s and 0s rather than ordinary vote counts, which can distort raw competition and contestation metrics.
Third-party and independent candidates. Votes for independents and minor parties are not folded into either major party. A strong independent challenger can depress major-party competitiveness scores even when the race was genuinely competitive (Massachusetts in 2024 is one example discussed on this page).
Incumbency, fundraising, and local politics. Safe seats can reflect long-standing incumbents, weak recruitment, or regional one-party dominance, not only district lines. Incumbency insulation measures structural protection in the data but not campaign spending, media markets, or candidate quality.
Presidential vs. House behavior. Adjusted estimates lean on presidential vote shares when House races are missing opposition. States that split tickets or shift sharply between federal elections may not be well represented by that shortcut in every year.

Examples worth keeping in mind

North Carolina's 12th district (1992) — Central to Shaw v. Reno and often cited for its shape. The map helps you see the boundary; metrics add vote-to-seat context but do not settle legal or historical debates about the map's purpose.
Louisiana's 4th district (1992) — Frequently discussed for its unusual ("Zorro-like") geometry. Shape draws attention; statistical scores describe election outcomes within that geometry. They are complementary, not interchangeable.
Massachusetts (2024) — Many effectively uncontested House races can push competition and contestation down and drag EHI lower, even when the state's overall political landscape is stable. Comparing raw vs. adjusted metrics helps separate structural map effects from uncompetitive delegations.
Post-Census years (e.g., 2002, 2012, 2022) — National fairness and confidence often move after new maps take effect. A dip may track implementation of new boundaries and incomplete contestation data as much as a permanent structural change.

Bottom line: Use these tools to ask better questions: why a metric is high or low, how raw and adjusted results differ, and how a state compares before and after redistricting. They are not meant to support definitive conclusions about intent, fairness in a legal sense, or the full lived experience of representation in each district.

Changelog

Site updates and notable changes. Dates and version numbers are placeholders until a formal release schedule is in place.

v1.0.0 (2026-07-04)

Official v1.0.0 release of the Congressional Representation Project (TCRP) House representation site (1976–2024).
- Added About This Project, Key Findings, Case Studies, and Static Visualizations sections.
- Added metric glossary under Details & Sources (all map metrics in one place).
- Added five static charts with per-chart CSV download buttons.
- Merged static chart code into the main site script; removed separate static-charts.js.
- Added CSV download for Visualizing Trends and static charts; map table CSV retained.
- Added state comparison overlay to Visualizing Trends.
- Adjusted Interactive Map configuration to always show metric information.
- Adopted TCRP as the project acronym in the page title, footer, citation, and About copy.
- Updated Welcome and Quick index navigation to list all main sections.
- Added mobile compatibility.
- Linked to Twitter/X account for updates (@CRProjectorg).
v0.2.0 (2026-06-11)

New features include:
- Added Visualizing Trends section.
- Added Raw Data Browser section.
- Added Credits & Disclosure sub-section under Details & Sources.
v0.1.0 (2026-06-06)

Initial public release of the Congressional Representation Project site with interactive map.

Credits & disclosure

AI disclosure: Artificial intelligence tools were used in part to help build this site, including code and layout. All data, metrics, and source material were reviewed and verified by a human before publication.

If you spot an error in the data, a metric calculation, or anything else on this site, please let us know at contact@crproject.org.

Want to be informed about new features and updates? Follow us on Twitter / X:

@CRProjectorg

Created by Aydon Fauscett
Email: aydonsoffice@gmail.com
GitHub: github.com/aydon14

Welcome to the Congressional Representation Project!

About This Project

Key Findings

Interactive Map

Usage

Taking a Closer Look

Visualizing Trends

Usage

Taking a Closer Look

Case Studies

North Carolina’s 12th District (1992)

1. Context

2. What the data shows

3. What the map looks like

4. What the data can and cannot tell you

Massachusetts (2024)

1. Context

2. What the data shows

3. What the map looks like

4. What the data can and cannot tell you

The 2010 Census Redistricting Cycle (National)

1. Context

2. What the data shows

3. What the map looks like

4. What the data can and cannot tell you

Texas Mid-Decade Redistricting (2003 → 2004)

1. Context

2. What the data shows

3. What the map looks like

4. What the data can and cannot tell you

Static Visualizations

Seat Share vs. Vote Share

Wasted Votes Over Time

Uncontested Races Over Time

State EHI Rankings (2024)

Redistricting Cycle Impact

Raw Data Browser

Details & Sources

What statistical analysis can and cannot show

Real-world factors this analysis does not fully model

Examples worth keeping in mind

Credits & disclosure