This is Predict Oregon's first scientific statewide poll — a live-telephone survey of 656 verified likely Oregon Republican primary voters drawn from the Secretary of State voter file. Fieldwork ran May 7–13, 2026, with data completion on May 14th, just five days before the May 19 primary election.

That means our poll took place inside the primary with votes already cast. Unlike other polls which measured solely voter intent — who someone is likely voting for — ours also measured actions. Who they did vote for; not a possible future action.

The headline finding is a statistical dead heat: Christine Drazan at 40.5%, Ed Diehl at 37.6% — a 2.9-point gap that falls inside the ±3.83% margin of error, meaning Diehl's actual support could statistically exceed Drazan's. Chris Dudley trails both at a distant 17.2%. Among the 295 respondents who had already returned their ballots at the time of the interview, Drazan leads by less than one point — essentially tied.

🔬 Stratified probability-based sample · n = 656 · Live telephone interviews · ±3.83% MoE at 95% confidence
656
Completed Interviews
Verified likely voters
±3.83%
Margin of Error
95% confidence
34/36
Counties Represented
Statewide coverage
295
Already Voted
Ballots already returned
Primary Preference — Decided Voters + Leaners

Respondents who named a candidate (Q2/Q3) plus undecided respondents who leaned toward a candidate when pressed (Q2A). Combined n=538. Percentages exclude those who refused to answer or had no preference after both questions.

Christine Drazan 40.5%
218 respondents · Already-voted share: 39.7% · Net favorability: +42.8
Ed Diehl 37.6%
202 respondents · Already-voted share: 39.0% · Net favorability: +44.0
Chris Dudley 17.2%
93 respondents · Already-voted share: 16.8% · Net favorability: +32.5
David Medina 2.8%
15 respondents · Net favorability: +12.5
Danielle Bethell 1.7%
9 respondents · Net favorability: +9.6
What the Topline Tells Us

The 2.9-point gap between Drazan and Diehl falls well within the ±3.83% margin of error, which means Diehl's true support could statistically exceed Drazan's — this is a genuine statistical tie. Either candidate could be ahead. The race is being decided in the margin. Both candidates are in the high 30s to low 40s; everything above 37.6% is contested terrain between two candidates running neck-and-neck statewide.

Dudley's 17.2% is the most stable number in the dataset. It appears in every sub-sample — decided voters only (16.5%), already-cast ballots (16.8%), not-yet-voted respondents (17.7%) — varying by less than a percentage point across each cut. When a number is that stable, it's not noise. It's where his support actually is.

Already-Cast Ballots — 295 Locked-In Votes

At the time of interview, 295 of 656 respondents (45%) had already returned their mail ballot. These are locked-in, immovable votes. This is the most direct pre-election read available for this race.

Candidate Already Voted (n=295) Not Yet Voted (n=243) Difference
Drazan39.7% (117)41.5% (101)+1.8
Diehl39.0% (115)35.9% (87)−3.1
Dudley16.8% (50)17.7% (43)+0.9
Medina2.4% (7)3.2% (8)+0.8
Bethell2.0% (6)1.4% (3)−0.6
The Signature Finding of This Poll

Drazan's early-voting advantage is real but thin: among ballots already returned, Drazan leads by 0.7 points — two votes. Drazan's not-yet-voted number is actually her stronger side. Among those who haven't voted yet, she leads by 5.6 points, suggesting she may benefit from final-days turnout.

Diehl's strongest numbers are already in the pile — he runs 3 points higher among returned ballots than among those who haven't voted yet. Given the gap, Diehl will need his supporters who already cast their votes to convince others to do the same.

Dudley's numbers are consistent and stable — but at 17.2%, consistent and stable may not enough.

The actual outcome of this race will depend on which candidate's not-yet-voted supporters return their ballots in the final days before May 19. And whether supporters of the trailing candidates decide to use their vote strategically — backing their second choice rather than a candidate unlikely to win.

The Dudley Conviction Gap — Votes vs. Expectations

Q5 asked all 656 respondents who they think will win the primary — independent of who they personally support. This "wisdom-of-the-crowd" question measures perceived momentum, not personal preference.

Candidate Preference Share (Q2+Q2A) Perceived Winner (Q5) Gap
Drazan40.5%43.6%+3.1
Diehl37.6%21.3%−16.3
Dudley17.2%8.5%−8.7
Medina2.8%1.3%−1.5
Bethell1.7%0.5%−1.2
The Conviction-Voting Gap

Dudley's supporters are voting for a candidate they themselves don't expect to win. His 17.2% preference share drops to 8.5% when you ask who will win — a 9-point gap. If his voters were making a strategic bet on electability, you'd expect that gap to run the other direction, with more people expecting him to win than personally supporting him. It runs the opposite way. They're voting conviction, not calculation.

Diehl's gap runs the other direction and is even larger: 37.6% of respondents support him, but only 21.3% expect him to win. His voters are confident in their choice but uncertain the rest of the electorate will follow. In a primary with early low-turnout numbers, the candidate whose supporters feel most certain they're backing a winner often benefits from that confidence in the final days.

Convergence — Numbers That Don't Move

One test of polling data quality is whether headline numbers hold up across independent sub-samples. When a number is consistent across multiple cuts of the data, it's signal, not noise.

Measure Drazan Diehl Dudley
Decided + leaners (n=538)40.5%37.6%17.2%
Decided only (n=507)41.0%38.0%16.5%
Already voted (n=295)39.7%39.0%16.8%
Not yet voted (n=243)41.5%35.9%17.7%
What Convergence Means

Dudley's number ranges from 16.5% to 17.7% across every independent slice of the data — a spread of 1.2 points across four different measurement cuts. Drazan and Diehl are similarly stable, each within roughly 2 points of their headline number in every sub-sample. When numbers are this consistent across independent measurements, it is not a quirk of the sample. It is where the electorate actually is.

How Our Numbers Compare

Two other polls were publicly released during this race. Here's how Predict Oregon's results compare.

Pollster Field Dates n / MoE Drazan Diehl Dudley
Predict OregonMay 7–13, 2026656 / ±3.83%40.5%37.6%17.2%
Hoffman Research GroupApril 23–24, 2026620 / ±3.93%35%18%14%
Nelson ResearchApril 14–17, 2026515 / ±4.3%31.1%15.6%14.8%

* Hoffman Research ballot test listed only Drazan, Diehl, Dudley, and "some other candidate" — Bethell and Medina were not offered as named options. Nelson Research included all five candidates. Hoffman's 25% undecided and Nelson's 29.9% undecided mean candidate percentages in both earlier polls are calculated on a smaller base than Predict Oregon's. All three polls are directionally comparable but not identical in question design or methodology.

Reading the Trend

The Crosstabs/Hoffman poll (April 24) explicitly flagged Diehl as having the highest conversion rate in the field and noted that his only constraint was name ID. Three weeks later, our data shows Diehl gained roughly 12 points. That trajectory is consistent with, not in conflict with, the earlier poll — name ID resolved, support consolidated. The Nelson poll (April 14–17) is the outlier; its methodology and sample frame have been questioned separately.

Predict Oregon's poll is the most recent, the closest to Election Day, and conducted via live telephone interviews of verified likely voters from the Secretary of State voter file.

Full Results Packet — Download Any or All

Five documents covering every dimension of this poll — from the one-page headline to full crosstabs

Methodology

How This Poll Was Conducted

Sample source: Oregon Secretary of State voter file with demographic and phone appends. Respondents were stratified across three likely-voter tiers — highly likely (HLV), very likely (VLV), and likely (LV) — with tier shares calibrated to expected primary electorate composition.

Method: Primarily live telephone interviews conducted by Predict Oregon volunteer interviewers and by an independent professional research firm under contract with Predict Oregon. A small portion (fewer than 20) of respondents unable to complete a live interview were offered a secure one-time access link to complete the survey online.

Fieldwork: May 7–13, 2026. n=656 completed interviews.

Margin of error: ±3.83% at 95% confidence.

Weighting: Results are weighted by gender to a 51.5% male / 48.5% female benchmark consistent with peer-pollster electorate composition. Age raking was tested and produced movement under 0.2 percentage points on every candidate; age weighting was not applied.

Geography: 34 of Oregon's 36 counties represented. Gilliam and Wheeler are not represented.

Funding: This poll was privately funded. It was not sponsored by, paid for by, or coordinated with any candidate, campaign, party, or donor. No candidate knew it was happening until the results were publicly announced.

Questions We Keep Getting Asked

Respond to bad-faith attacks with facts. Here are straight answers to the questions circulating about this poll.

Is this just an internet poll?
No. This poll used live telephone interviews of 656 verified likely Republican voters from the Oregon Secretary of State voter file after the April 28th deadline to register. This is a scientific, probability-based sample with a ±3.83% margin of error at 95% confidence. It is the opposite of an internet poll.
Who paid for this poll?
This poll was privately funded. No candidate, campaign, party committee, PAC, or donor commissioned, sponsored, or had advance access to this research. Predict Oregon published the results regardless of outcome and did not share them with any campaign before publication.
Why should I trust a polling operation I've never heard of?
Fair question. Judge the methodology, not the brand name. The sample is drawn from the Secretary of State voter file — the same source every professional pollster uses in Oregon. The interviews were live phone calls. The sample is stratified by likely-voter tier. The margin of error is ±3.83% at 95% confidence. The full methodology is published and documented. Compare that standard to any other poll released in this race.
Didn't a candidate say on the radio that this poll is fake?
Yes. The claim was made without any specific methodological critique — it amounted to not liking the results. Candidates who trail in polls routinely challenge the poll rather than the findings. We document our methodology publicly and transparently so you can evaluate the work yourself. The sample source, fieldwork method, dates, n, and margin of error are all disclosed above and in our published documents.
How is this different from your earlier polls?
This is Predict Oregon's first scientific poll. Our March 2026 and April 2026 polls were self-selected non-probability samples — useful for directional signal, but not scientific in the traditional sense. This poll is different: verified voter file sample, live phone interviews, stratified by likely-voter tier, and a proper statistical margin of error. Same commitment to transparent methodology — different and more rigorous sample design.
Why don't the numbers add up to 100%?
The topline percentages are calculated on the base of respondents who named a candidate or leaned toward one. Undecided respondents who did not lean (n=118) and those who refused to answer are excluded from that denominator. The full topline report, linked above, shows all response categories including undecided and refused.
Can I see the full question wording?
Yes. The full topline report (linked in the downloads above) includes the complete question text for every question in the poll, along with all response categories and counts. Candidate names were presented in randomized order to each respondent to prevent order bias.