What the Score Means
Every Texas county on this site has a Party Lean score like R+13 or D+7. This is the margin of victory: R+13 means Republican candidates win by about 13 points (roughly 56.5% to 43.5%). D+7 means Democrats win by about 7 points. Scores within 1 point of zero are labeled Even.
The score is based on contested federal and statewide races in recent general elections: President, U.S. Senator, U.S. Representative, Governor, Lieutenant Governor, Attorney General, Comptroller, Land Commissioner, Agriculture Commissioner, and Railroad Commissioner. These are the races where both parties field candidates and voters make a direct partisan choice.
How We Calculate It
The lean score is a recency-weighted average of the Republican two-party vote share across the last five general elections (roughly 10 years), converted to a margin. More recent elections count more heavily:
The weighting follows an exponential decay: each election cycle is worth half as much as the one after it. This means the most recent general election accounts for about half of the score by itself, and the two most recent generals together account for over 75%.
This approach is similar to the Cook Partisan Voter Index, which moved from equal-weighting to a 75/25 split in 2022 to better capture rapid partisan shifts. Our steeper decay curve is appropriate for Texas, where suburban counties like Collin, Tarrant, and Fort Bend have shifted significantly in recent cycles.
Worked Example: Harris County
Harris County illustrates why this methodology matters. Under our previous method (unweighted average of all election types including primaries), Harris showed a lean of Even. But its recent general elections tell a different story:
The weighted average gives Harris County a Republican two-party share of about 46.7%, producing a lean of D+7. That reflects the county's consistent Democratic lean in recent cycles while accounting for tighter results in 2024 and 2016.
Why These Races
Our previous calculation included votes from all election types and all races. This created two significant distortions.
First, Texas primaries are single-party elections. In a Republican primary, 100% of votes are cast for Republican candidates by definition. Including these votes inflated the Republican share in counties with high primary turnout and deflated it in counties with active Democratic primaries.
Second, aggregating all races in a general election (including downballot judicial and local races) exaggerates lean in both directions. In Republican-leaning counties, many downballot races are unopposed Republicans. In Democratic-leaning counties, the reverse. Tarrant County, for example, showed R+30 using all races but R+8 using only federal and statewide contested races, which better reflects its competitive reality.
We now use only contested federal and statewide races: President, U.S. Senator, U.S. Representative, Governor, Lieutenant Governor, Attorney General, Comptroller, Land Commissioner, Agriculture Commissioner, and Railroad Commissioner. These races reliably feature candidates from both parties and reflect the kind of partisan choice voters actually make.
Why Weight Recent Elections More Heavily
Texas counties are not static. Fort Bend flipped from R+7 to D+3 in just four cycles. Tarrant went from solidly Republican to a battleground. An equal-weight average over 20 years of data would obscure these shifts.
Recency weighting ensures the score reflects where a county is, not where it was. The Cook Political Report made the same change in 2022, shifting from a 50/50 split to 75/25 because "the dramatic increase in straight-ticket voting since 1997 made the most recent election a better predictor of future behavior."
Our decay curve is steeper (roughly 50/25/13/6/3 rather than 75/25) because we have more data points (every 2 years rather than every 4) and because Texas's partisan shifts have been particularly rapid at the county level.
What We Tested and Chose Not to Use
Primary Turnout as a Predictor
We tested whether primary turnout ratios (R primary voters vs. D primary voters) could improve the lean estimate, particularly for years where primary data is available but the general hasn't happened yet (like 2026 at the time of this writing).
The analysis covered 2,019 county-year observations from 2010 to 2024. We excluded 2006 and 2008 because of a well-documented crossover voting effect during the 2008 Democratic presidential primary.
Finding: Primary turnout is not reliably predictive of general election lean beyond what prior general election results already tell us.
- Mean absolute error of predicting the general using only the prior general: 4.18 points. Blending in 10% primary data: 4.06 points, a marginal 3% improvement that was inconsistent across cycles.
- Directional accuracy: When primary data suggested a county was shifting R or D relative to the prior general, it was correct only 58.7% of the time, barely better than a coin flip. In gubernatorial years (2014, 2018, 2022), it was worse than a coin flip (31-48% correct).
- Bias instability: The gap between a county's primary R% and its general R% varied wildly across cycles. Harris County's primary-to-general bias ranged from -17 to +18 points (std dev 10.0). Even the most stable counties had 3-5 point swings.
The core problem is that primary turnout reflects enthusiasm within a party (driven by contested intra-party races), not relative party strength. A hotly contested Republican governor's primary can spike R primary turnout statewide without meaning anything for November.
Equal Weighting Across All Years
An unweighted average of all available general elections (2006-2024) produces scores that lag behind reality. Harris County, which has voted Democratic in four of the last five generals, would show as nearly even under equal weighting because its pre-2016 Republican results pull the average up.
National Baseline Normalization
Methods like Cook PVI express partisan lean relative to the national average. We considered this but chose to use absolute two-party share instead. For a Texas-specific site, users want to know "does this county lean R or D?", not "does this county lean more R than the national average?" Since Texas is a red state, normalizing to the national baseline would shift nearly every county to the right, which would be confusing for users comparing Texas counties to each other.
How This Compares to Other Methods
Limitations
- Third-party and independent candidates are excluded from the two-party calculation. In races where a Libertarian or Green candidate drew significant votes, the two-party share may overstate the winner's support.
- U.S. Representative races are included alongside statewide races. Because congressional districts don't align perfectly with county boundaries, a county may only see results from one or two districts, which can differ from the county's statewide voting pattern.
- The score reflects historical voting behavior, not a prediction of future results. County demographics, candidates, and issues can cause significant shifts between cycles.
Source Data
All vote totals are sourced from the Texas Secretary of State. Our dataset covers general elections from 2006 through 2024 across all 254 Texas counties. Lean scores are recalculated whenever new election data is imported.