You might have noticed that the Oregon primary polls were a bit... off.
That’s the question the Oregonian’s Betsy Hammond asked Wednesday. As it turns out, when it comes to DHM Research, they can be extravagantly wrong.
DHM predicted Hillary Clinton would defeat Bernie Sanders in Oregon, 48 percent to 33 percent. But with nearly all the votes counted, Sanders is winning handily, 55 percent to 43 percent. That’s a 27 point miss — a colossal margin noticed by political reporters across the state and country. Willamette Week called it “lousy polling.” Polimedia noted that even though DHM gave their results a huge margin of error, they “still missed big time.” And Benchmark Politics called it their “biggest miss in history.”
What was DHM’s response when asked why they missed so badly? They blamed Oregon’s new automatic registrants as the reason. But that excuse doesn’t quite add up (and communicates a pretty large lack of understanding of how our new system works), since, according to the publicly available records from the Secretary of State, almost every voter registered through Oregon’s new Motor Voter law is a non-affiliated voter, unable to cast a ballot in a Democratic primary.
So while DHM might not understand why they got the answers so wrong, it’s pretty easy for the rest of us to see why they were so far off the mark: They simply surveyed the wrong people. Their survey of “likely voters” was skewed towards older, lifelong Democrats (who tend to support Clinton) and undercounted younger voters (who overwhelmingly support Sanders).
The impacts of this big miss weren’t just limited to Oregon. Because state polls are combined into national models, DHM’s airball threw off predictions all across the country. Benchmark Politics lamented, “without the DHM +15 Clinton poll we would have been dead on.”
A miss this big calls into question DHM Research’s leaves us with more questions than it answers. Surely a whiff this big will have pollsters and journalists alike thinking twice about trusting any of DHM’s future findings.