One combined fundamentals score
As of October 19, 2026 · Senate research
A combined fundamentals score has not shown a statistically reliable improvement beyond the polling forecast.
No new forecast weights adopted.
New: Cook, Sabato and Inside Elections tests
In 130 races across 2016–2024, the candidate-plus-lean index, blended using weights chosen from earlier validation elections, changes average error from 4.613 to 4.638 points: slightly worse. It does not validate an added forecast weight.
Historical comparisons
Lower error is better. Errors are in percentage points. Each row has its own polling benchmark because source coverage differs.
| Composite | Test coverage | Polling error | Learned blend error | Fixed 25% blend error |
|---|---|---|---|---|
| Candidate factors, without partisan lean | 130 races / 5 cycles | 4.613 | 4.613 | 4.946 |
| Candidate factors + partisan lean | 130 races / 5 cycles | 4.613 | 4.638 | 4.348 |
| Candidate factors + lean + generic ballot | 79 races / 3 cycles | 4.726 | 4.871 | 4.282 |
| All candidate factors + individual-dollar share + lean | 13 races / 1 cycle | 3.695 | 3.695 | 4.225 |
| All candidate factors + dollar share + lean + generic ballot | 13 races / 1 cycle | 3.695 | 3.695 | 4.222 |
| All candidate factors + individual-dollar amount + lean | 13 races / 1 cycle | 3.695 | 3.695 | 4.224 |
Does removing variables help?
Across 130 races in 2016–2024, the full five-factor score is the best of all 32 fixed-25%-blend versions, with error 4.348. Removing any one factor makes it worse. Choosing the subset, shrinkage and blend using earlier validation elections gives error 4.642, versus 4.613 for the current forecast. The earlier-selected procedure does not validate a new weight.
Full score: state lean, incumbency/tenure, earlier personal performance, challenger or open-seat experience, and scandal news. Fixed 25% blend; lower error is better.
| Factor removed | Average margin error | Improvement over full score |
|---|---|---|
| State partisan lean | 4.946 | -0.597 |
| Incumbency and Senate tenure | 4.417 | -0.069 |
| Earlier personal election performance | 4.354 | -0.006 |
| Challenger/open-seat experience | 4.365 | -0.016 |
| Scandal news | 4.354 | -0.006 |
In the generic-ballot cohort (79 races, three elections), dropping scandal news is the best fixed-blend subset but helps by only 0.000383 point. In the financial cohort (13 races, one election), the best subset drops fundraising, experience and scandal news; its error is 4.216, still above the current forecast’s 3.695. Financial significance cannot be estimated from one test election.
Removal-test methods
I tested removing each factor family and all smaller combinations: 32 versions without generic-ballot or financial coverage restrictions, 64 in the generic-ballot cohort, and 128 in the matched financial cohort. Every subset within a cohort uses the same training and test races.
The best fixed subset is chosen after inspecting the evaluated elections, so it is a retrospective diagnostic. The earlier-selected procedure chooses variables before each target result, using only earlier validation elections. These are development comparisons on elections already studied; no new weights are activated. Correlated remaining inputs can retain related information, so removing a family does not isolate its causal effect. Missing source indicators are removed with their associated family.
What the promising version tells us
A fixed 25% blend is more encouraging: 4.613 → 4.348 points, a 0.264-point improvement. It helps in 2020 and 2022 and worsens 2016, 2018 and 2024. The two-sided whole-election sign-flip p-value is 0.625; the exploratory 95% cycle-bootstrap interval for the improvement is -0.192 to 0.869 points.
Most of that fixed-blend gain is available from lean and incumbency alone, which reach 4.381 points. The richer candidate score adds only 0.033 points of improvement over that simpler score (p = 0.625).
Fundraising and the generic ballot
The full fundraising combination can be tested in only thirteen 2022 races after training on the matched 2016/2018 reports. Its fixed 25% blend raises error from 3.695 to 4.225 points. The generic-ballot version is similar. One test election cannot establish a general fundraising coefficient or a statistically significant joint effect.
Adding the generic ballot has only three later complete-case test cycles (2020/2022/2024; 79 races). The nested blend raises error from 4.726 to 4.871 points. A fixed 25% blend lowers it to 4.282, but the gain is inconsistent (p = 0.750). This subset is not directly comparable with the five-cycle table.
Full explanation and limits
The score combines incumbency, earlier personal election performance, appointed/elected status, challenger or open-seat experience and dated scandal news. Versions add state partisan lean, the generic ballot, and individual fundraising. Each index is one expected Democrat-minus-Republican margin, learned from earlier elections without current race polling as an input.
This is a set of specified combinations, not every possible weighting or interaction. A linear composite packages existing information; its usefulness depends on out-of-election performance. Historical development results are exploratory because these elections have already been studied. Whole-election tests avoid pretending that races in the same national environment are independent, but five cycles remain few, and overlapping training history also limits formal inference. The five-cycle sign-flip test cannot attain a two-sided p below .0625. Bootstrap intervals are descriptive sensitivity checks, not precise confidence guarantees. More financial history and future elections are needed; no new weights have been activated.
Ridge prediction of actual D-minus-R margin using no current race polling as an input. Coefficients use only earlier elections; prior personal performance also predates the target election.
Incumbency; appointed vs multiple election wins; latest comparable prior personal overperformance for incumbent and nonincumbent; nonincumbent state/House/governor experience hierarchy; dated public-news severity; optionally partisan lean, cutoff-dated generic ballot, and Q3 individual-dollar share or amount.
Source and prior-performance gaps have explicit signed missing indicators; financial and generic-ballot gaps are excluded from their complete-case specifications, never zero-imputed.
Equal-cycle race weights with an eight-year recency half-life. Intercept penalty .10; PVI/10 prior 10, penalty .10; incumbency prior zero, penalty .5; added features prior zero, penalty 2; multiplied by lambda. No current-cycle outcome or race polling enters the pure index.
For each outer election, fit and score inner indices using only still-earlier elections. Select ridge strength and blend on the mean MAE across at least two earlier inner validation elections. With fewer, use lambda 1 and alpha 0. Also show fixed 25% blends as exploratory diagnostics, with lambda 1.
P + clip(alpha * (F - P), -3, +3) percentage points. F is the one-number fundamentals estimate and P is the frozen active forecast.
Keep inherited per-cycle uncertainty/calibration fixed for paired diagnostics. A point-error gain alone does not validate uncertainty or authorize adoption.