Washington State’s right wing movement is once again circulating skewed survey data to make the argument that one of its top candidates this cycle has far better prospects than they really have, in what appears to be a reprise of their failed attempt in 2022 to put the Evergreen State in play for Democratic Senator Patty Murray’s Republican challenger Tiffany Smiley, who Murray dispatched with relative ease in the midterms.
This time around, the pollster deploying bad data to try to shape the narrative is Echelon Insights, rather than Trafalgar, and the intended beneficiary is Republican gubernatorial hopeful Dave Reichert, rather than Smiley.
Echelon’s March 2024 survey for “Concerned Taxpayers of Washington State” has Reichert up nine points over Democratic Attorney General Bob Ferguson in a head-to-head matchup. Echelon also claims that a plurality of voters favor Republicans in a generic legislative ballot question. Those findings are totally at odds with our own recent public opinion research and other credible private polling that we’ve seen.
Echelon obtained these findings primarily by constructing a skewed statewide sample that doesn’t have enough Democratic and progressive voters in it to properly resemble the likely Washington State November electorate, just as it did last year when it polled for the right wing group Future42 and Brandi Kruse’s podcast.
This is a gambit that our team has seen before.
Publishing skewed data in an effort to create a self-fulfilling prophecy didn’t work last cycle for Tiffany Smiley and we don’t think it’ll work this time around, either.
But Republicans — who are tired of being on the losing end of the longest current streak of consecutive gubernatorial election victories by one political party anywhere in the country — are desperate.
They crave relevance and dream of flipping the Evergreen State despite Donald Trump’s stranglehold on their party, which nowadays functions more like a cult. And so they’re paying Echelon Insights to periodically publish information that suggests a favorable electoral landscape for Republicans that simply doesn’t exist in reality.
If last cycle’s pattern holds, then we can expect a few other Republican-aligned firms to eventually materialize and drop surveys that show Reichert ahead or tied with Democratic frontrunner Bob Ferguson. That’s what we saw in the midterms — it was the final act in their data pollution gambit. First, Trafalgar produced a series of periodic surveys that were favorable to NRSC recruit Smiley, and then in the home stretch, firms like Moore Information, co/efficient, and KAConsulting joined the party, contributing their own flawed data to make it look like Smiley was competitive when she really wasn’t.
Credible preelection polling from nonpartisan and Democratic-aligned pollsters had found Murray well ahead, but Republicans were able to “flood the zone” and obtain a lot of favorable media coverage with their bad polling. NPI worked before, during, and after the election to hold them accountable, and we are doing so again now.
In a moment, I’ll share the deck that Echelon released and invite you to scrutinize it. Then we’ll discuss why the findings lack credibility and should not be taken seriously.
How we evaluate polls
First, a few words about how we evaluate polls.
We believe that subjective organizations are perfectly capable of conducting objective research, so we won’t reflexively dismiss data from a survey merely because it comes from right wing sponsor. However, we are sticklers for the scientific method, so we will look to see if that sponsor and their pollster adhered to best practices for obtaining sound data, or whether they cut corners.
The key to accurate, credible polling is neutral questions asked of a representative sample. You’ve got to have both, or your data is going to be skewed. You can’t find out what people think if you tell them what to think first, which is why neutral questions are important. And your sample needs to properly resemble the electorate or the population of the area you’re surveying, or the findings simply aren’t going to accurately reflect what public opinion is.
It’s extremely difficult to write neutral questions and it can also be tough to build representative samples. But it can be done, as we’ve demonstrated over the past decade in our work with Public Policy Polling, a well-regarded pollster, and more recently also with Change Research and Civiqs, two other trusted firms we regularly work with.
Survey details
According to Echelon Insights, this was a survey of 600 registered voters in Washington State “aged 18 or older”, conducted for Concerned Taxpayers of Washington State. They were interviewed “from March 18–21, 2024 using a mix of live telephone calls (30% landline) and text to online (70%), matched to the L2 Washington voter file.”
The brief methodology statement goes on to say: “The sample was weighted to reflect population benchmarks for November 2024 registered voters for gender, age, race/ethnicity, education, region, party, turnout probability, and 2020 vote. The margin of error for a random sample of 600 voters is ± 4.7%.”
Echelon is based in Virginia, and is Republican-aligned.
(As you can probably guess, a Republican-aligned firm is partisan and works mainly with right wing clients. A Democratic-aligned firm like one of our pollsters is the equivalent, also partisan and works with progressive clients.)
Ironically, according to BuzzFeed, the firm was created to “fix the chronic Republican problem of bad polling.” Co-founders Patrick Ruffini and pollster Kristen Soltis Anderson were frustrated that many Republican campaigns were relying on internal polls with skewed samples that weren’t accurately gauging current electoral dynamics, which was leading to Election Night surprises in a number of high-profile Republican campaigns.
And yet, here they are, generating the sort of bad data that they were calling on Republicans to stop wasting money creating around a decade ago.
A copy of their deck summarizing selected results is below.
APR24-Echelon-Survey-DeckEvaluating the sample
As was the case last year, the sample suffers from a fatal flaw: it doesn’t have enough Democratic and progressive voters in it. On the “Survey Demographics” slide, you can see that Echelon identifies 28% of the respondents as Republican, 34% as Independent, 34% as Democratic, and 3% as “something else.”
With respect to ideology, 35% are identified as “conservative”, 32% as “moderate” (which is a problematic label, not an ideology), and 29% as “liberal”. Echelon claims that a total of 42% of its sample are Republican or lean Republican, 10% are pure independent, and 38% are Democratic or lean Democratic.
Washington State doesn’t have voter registration by party. Nevertheless, we can see from election results that a majority of Washingtonians are Democratic or lean Democratic — not a plurality. A properly representative sample needs to match Washington’s electoral profile. Data going back many cycles shows that most Washington voters reliably vote for Democrats at the top of the ticket.
For example, in the 2020 presidential election, 57.97% of Washington voters supported Joe Biden, and in the 2022 U.S. Senate election, 57.15% of Washington voters supported Patty Murray. Republicans haven’t won a gubernatorial election since 1980, the state’s presidential electors since 1984, or a U.S. Senate race since 1994.
A properly representative statewide sample for Washington should consist of at least 54% or so of Biden voters, and ideally closer to 57%. Our February 2024 statewide survey, which was conducted by Public Policy Polling, consisted of 56% Biden voters, 37% Trump voters, and 7% who reported voting for someone else or didn’t vote.
41% of the respondents in that poll identified as Democrats, 23% as Republicans, and 36% as independents.
Contrast that with Echelon’s sample. They have too few Democratic voters, which means their sample resembles a Washington electorate that doesn’t exist.
Though Echelon omitted the 2020 vote breakdown from its slide deck, Echelon’s Patrick Ruffini did provide, in response to a tweet we posted, the following screenshot, which shows the 2020 vote breakdown in the survey… 55% Biden, 37% Trump.
So that at least clears that up. But who are those Biden voters? That’s important: merely having alignment with presidential vote results doesn’t mean a sample is representative.
Across the entirety of the Echelon survey, Democratic candidates look much weaker than they do in our polling, which utilizes properly representative sampling. Failing to get the partisan balance of the electorate right is public opinion research malpractice, and given that Echelon has continued to undersample Democrats even after we called them out for doing so last summer and urged them to do better next time, we can only conclude that they’re doing this deliberately to try to drive a pro-Reichert narrative.
Comparing Echelon’s findings to NPI’s most recent findings
NPI’s most recent survey fielded in February. Each link below goes to the finding for that contest, with the full question text and toplines and accompanying analysis.
Governor (Head to head)
- NPI/PPP: Bob Ferguson 46%, Dave Reichert 42%, 11% not sure
- Echelon: Dave Reichert 39%, Bob Ferguson 30%, 31% not sure
Governor (Top Two field)
- NPI/PPP: Bob Ferguson 35%, Dave Reichert 27%, Semi Bird 9%, Mark Mullet 4%, 25% not sure
- Echelon: Dave Reichert 28%, 23% Bob Ferguson, 7% Semi Bird, 5% Mark Mullet, 37% not sure
President
- NPI/PPP: Joe Biden 54%, Donald Trump 38%, 8% not sure
- Echelon: Joe Biden 48%, Donald Trump 37%, 14% not sure
U.S. Senate
- NPI/PPP: Maria Cantwell 53%, Raul Garcia 37%, 10% not sure
- Echelon: Maria Cantwell 44%, Raul Garcia 36%, 19% not sure
Generic legislative ballot
- NPI/PPP: Democrats 51%, Republicans 41%, 8% not sure
- Echelon: Republicans 42%, Democrats 40%, 18% not sure
You can see that in each contest, Echelon found the Democratic candidate / Democratic ticket in a weaker position than we did. The deficit ranges from six to sixteen points compared to our February 2024 statewide poll. And even though the number of “not sure” voters is also higher in Echelon’s poll, Republican candidates and the generic Republican legislative ticket nevertheless performed at levels comparable to our survey.
Echelon also asked about Brian Heywood and Jim Walsh’s slate of initiatives, but curiously, did not test the actual ballot titles. Instead, respondents were asked somewhat biased questions which don’t meet our criteria for neutrality. Considering that the sample is also skewed, those findings aren’t useful. Amusingly, despite the skew of the sample and the somewhat biased questions, the responses still weren’t all that great for the slate.
Credible polling versus bad data
Credible polling like NPI’s adheres to the scientific method. While it cannot predict election results, it can suggest what is most likely to happen with a reasonably strong degree of confidence. It’s very important to understand that distinction.
Our public opinion research has a track record of excellence that dates back to when we began commissioning surveys to gauge voter sentiment in upcoming Washington State elections. At every level that we’ve polled — statewide, congressional, legislative, local — the candidates who’ve led in our surveys have gone on to win.
The most recent example is our survey of the Washington Republican presidential primary. Working with Civiqs, we found Donald Trump at 77% and Nikki Haley at 8%, with x% undecided. In the primary, which was recently certified, Trump received about 76% and Haley got 19%. That’s about as spot-on as electoral polling can be!
Observers from across the ideological spectrum know that NPI’s polling has consistently been on the mark; some have called it the gold standard. There’s a simple reason for this success: we are sticklers for following the scientific method. We commission research because we want to know what people think, not because we want to generate favorable numbers for progressive causes or candidates.
Those who put their faith in bad data, meanwhile, end up looking pretty out of touch.
Take Jason Rantz, a local right wing talk radio host employed by Bonneville. He went all in on Trafalgar’s 2022 WA-Sen polling and was hyping it excitedly on the air and on posts for MyNorthwest.com right up until the election demonstrated it was total bunk. Here’s one from October 30th, titled WA Senate race statistically tied as Tiffany Smiley surges:
The Washington Senate race is now statistically tied, according to a new Trafalgar Poll.
Republican newcomer Tiffany Smiley earned 48.2% support to 30-year incumbent Patty Murray’s 49.4%. The margin of error is 2.9%, making this a statistical dead heat. 2.4% of voters are undecided.
It appears the surge is coming from independent voters, in line with the significant swing expressed in a recent Seattle Times poll. Trafalgar oversampled Democrats at 44.2% versus 33.4% of Republican respondents. Independents accounted for 22.4% of the respondents. And there were more female (53.1%) respondents than male (46.9%).
Smiley is on a statewide bus tour and her stops have been standing-room only. Contrast that with Murray who, even with guests like Senator Elizabeth Warren, has sparse participation.
Rantz kept the drumbeat going a few days later with this piece — Big names step up support for Tiffany Smiley as polls show ‘dead heat’ in senate race.
Of course, credible polling did not show a dead heat in Washington’s U.S. Senate race. Rantz ignored the credible polling and put his faith — at least publicly — in bad data. A few weeks later, he was forced to lament Tiffany Smiley’s concession and bemoan that “too few people chose to take this seriously and let her [Smiley] down.”
He did not apologize for having played a leading role in promoting garbage right wing polling that was feeding a fabricated narrative about Washington’s U.S. Senate race.
An opportunity squandered
Washington State and the Pacific Northwest could benefit from having a Republican-aligned pollster that contributes to our region’s body of credible public opinion research, complementing the work that NPI and its pollsters do from the progressive side and the work that media organizations like The Seattle Times and Crosscut do with their pollsters (usually SurveyUSA and Elway Research, respectively.)
Echelon Insights could have been that firm.
Sadly, instead, it has chosen to behave like Trafalgar or Moore and use its infrastructure and reputation to peddle bad data. That’s extremely disappointing. And unfortunately, this is the sort of behavior that goes totally unpunished nowadays in Donald Trump’s Republican Party, which pretends to care a lot about the truth, but in practice has become a firehose of misinformation and disinformation.
Andrew, as always an interesting post. I’ve been around the intersection of media and politics a while now, and when this poll appeared I thought (being right of center) that it was too good to be true. But after examining some of the same data that caught your attention, I’m now thinking the poll might be catching a trend: that as Seattle voters pulled away from an out of touch city council and swung toward the center in last year’s city council elections, the state’s voters seem poised to do the same thing. I initially thought the poll undercounted Democrats, but I noticed that when asked who they voted for in 2020, the respondents favored Biden by 18 percentage points. Biden actually won by.….19 points. Pretty close. So what’s happening? My guess is that just as many self described independents moved into the Democratic camp during the Trump presidency, they are now migrating back to the independent camp. We’ll see if this is indeed a trend or a blip, but the fact that the Dems in Olympia voted to repeal their own law restricting police chases suggests that they realize they overplayed their hand, and are trying to contain the damage, lest they lose the three “money” initiatives in November. Thanks for the chance to offer feedback.
Hi John. We aren’t seeing any evidence across our research to support your theory that “self described independents” are “migrating” towards Republican candidates and causes. When we ask Biden voters if there is anything the Republicans can do to earn their vote in the next election, most say no. Those who say yes are asked to comment. It turns out that the vast majority of these yes folks are also nos, however. The sentiments we hear from them go like this: Yeah, if the Republican Party repudiated Trump and everything he stands for, and found a good candidate who wasn’t an extremist, I’d consider voting Republican in a future election, but not this next one. No way, no how.
The Republican Party’s transformation into a political cult centered around Donald Trump has been catastrophic for several consecutive election cycles. Even in the midterms, when Trump wasn’t on the ballot, Trumpism was… and Washington State Republicans did very poorly.
The Legislature’s decision to vote I‑2113 into law was a strategic choice to keep it off the ballot to deny Republicans a messaging tool and guarantee the next Legislature’s ability to modify the law again (which they first did last session, before the I‑2113 signature drive). It was not an admission of overreach. And the outcomes of last year’s Seattle city council races are not a foreshadowing of anything. By that same logic, Adam Fortney’s loss in Snohomish County and Nadine Woodward’s loss in Spokane are a very bad omen for Republicans.
Candidate elections turn on identity and trust, not ideology. Way too much political analysis is ideology-driven or ideology-centric. What research shows us, though, is that ideology is only one facet of people’s political identity. It is important, to be sure, but there are other facets, other scales, to be considered, like means and ends or the desired rate and speed of change.