PENNER STRATEGYResearch
SILPHIUM · INVESTIGATION

Primary-to-general changes in 78 Washington races, 2014–2024

Do the 2022 and 2024 races behave like the longer historical sample?

The recent races had a narrower spread.

In the 19 races from 2022 and 2024, the middle 90% of changes in Republican vote share ran from −1.47 to +1.30 percentage points. For all 78 races from 2014–2024, the range was −4.82 to +3.05.

“Middle 90%” describes the span from the 5th to the 95th percentile in these past races. It is not a 90% probability range for this November. A one-point change means a share moving, for example, from 54% to 55%.

Charts from this investigation

Comparison of historical vote-share changes in 78 and 19 races
A comparison of past changes, with the two samples built the same way. These ranges are not a calibrated forecast for November. Open full-size figure

Recent and earlier election changes

Change in the Republican candidate’s two-candidate share, percentage points
Reference periodRacesMean change5th percentile95th percentile
2014–202478-0.36-4.82+3.05
2022 and 202419-0.02-1.47+1.30

The recent sample contains 19 races across two election years. These are historical percentiles, not probability bounds for November 2026.

Download all 78 races and the R reproduction script · Reference races (CSV) · Mean Republican-share change by year (CSV)

31st District examples in the broader historical panel
31st District races included in the broader two-candidate panel; candidate orientation follows that panel
YearOfficeCandidate followedAugust shareNovember shareChange (points)
2014House 2Christopher Hurst54.64%51.60%-3.04
2016House 1Drew Stokesbary (R)73.98%71.59%-2.39
2018House 1Drew Stokesbary (R)56.33%58.92%+2.59
2024House 1Drew Stokesbary (R)63.86%63.69%-0.17

Changes also varied by election year.

The broader 302-race Republican-versus-Democrat panel; this is not the 78-race incumbent subset
YearRacesMean change in Republican share
201044-1.54 points
201453+0.42 points
201634+0.80 points
201851+0.83 points
202045-0.91 points
202232+0.59 points
202443+0.68 points

Yearly values (CSV) · The underlying 393-race panel (CSV)

How the 78 races were selected

The code begins with 14 official primary and general-election files, totaling 4,369 rows. It matches the same two named candidates across August and November, producing 393 races. Of these, 302 pair a Republican and a Democrat. Exact name matching to earlier general-election winners identifies 213 races with an inferred incumbent, including 78 with a Republican incumbent.

The sample’s appointment history and boundary changes have not been fully checked. The recent subset shares only two election years. A different sample construction produced 96 races in the earlier investigation; that unresolved difference affects how representative these ranges may be.

Download source files, 78 races and reproduction code

Calculation and record checks · Return to the district results

Sources and production

Silphium is Penner Strategy’s research system. These studies use AI-assisted analysis and code-generated calculations and charts. The linked review records describe the calculation checks and remaining limitations.