Candidate Margin Overperformance

1. What CMO measures

CMO is a state-legislative analogue to Split Ticket's WAR. It begins with the observed two-party margin gap between a legislative race and the same district's ticket baseline, predicts the portion normally associated with structural conditions, and treats the remainder as candidate margin overperformance.

Republican values reverse the Democratic race residual, so every race is zero-sum. The residual is retrospective and is not a causal estimate of either candidate's personal contribution.

2. Data and eligibility

The model covers contested Democratic-versus-Republican Alabama House and Senate races from 1994 through 2022. Official and reconciled returns provide legislative, federal, statewide, and presidential results. Census and ACS sources supply era-appropriate demographics; finance enters only through a capped campaign-effort term.

3. Ticket baseline

The preferred comparison is the same-cycle U.S. House and U.S. Senate result measured within the legislative district. It captures the national political environment facing the legislative candidates. Eighty-one races without a usable federal comparison use a documented same-cycle state-ticket fallback; 428 use the federal-primary baseline.

4. Structural expectation

A regularized ridge model predicts the normal legislative-ticket gap. Major terms are symmetric party-oriented incumbency and era-specific downballot lag—the difference between same-cycle federal voting and preceding presidential voting. Prior presidential margin and available presidential swing provide additional political context.

Nonwhite share and white-college share are minor demographic terms capped at ±3 points. Campaign effort uses spending, fundraising, or resource ratios according to availability and is capped at ±2 points. Ideology and cycle fixed effects are excluded: ideology is an outcome to test afterward, while cycle indicators would absorb statewide Conservadem overperformance.

5. Model selection and checks

Ridge strength is selected with leave-one-cycle-out testing. Published checks cover vote arithmetic, ticket-source selection, component reconciliation, caps, party symmetry, candidate orientation, and exact zero-sum scores.

6. Candidate and career summaries

Election-level CMO is the primary product. Candidate histories use normalized full-name identities; unresolved surname-only records remain race-specific. A separately labeled partial-pooled career summary shrinks repeat candidates toward zero and never replaces the candidate-cycle result.

7. Limitations

• Eight cycles provide many races but few independent statewide environments.
• The effort cap reaches its ±2-point bound often, signaling that finance measurement remains coarse.
• Repeat-candidate WAR persistence is near zero in the current sample, so CMO should not be treated as a durable talent rating or forecast signal.
• Fallback ticket sources and historical geographic allocation add uncertainty that is documented row by row.
• Zero-sum residuals cannot separately identify the contributions of opposing candidates.

8. Reproducibility

The versioned build publishes race rows, candidate rows, component decomposition, coefficients, cycle diagnostics, construct checks, and a provenance manifest.