Polls that consider a broader picture are more reliable in predicting ballot outcomes

On July 20, a tweet that prompted an Oregonian debate looked like an election horse race shocker: a projected Republican win in the state’s 4th Congressional District. 

Said one tweeter, “These Oregon ratings are uh interesting.”

The tweet was a map showing Oregon’s 4th Congressional District (in southwest Oregon, including Lane and Benton counties) as a toss-up, but with Republican Alex Skarlatos favored (57.3% to 42.7%) over Democrat Val Hoyle. The map and statistics struck a nerve because Hoyle has been widely regarded as the front-runner in the race, in a district redrawn for this year to favor Democrats. Skarlatos has run and lost in the district before; Hoyle has been elected to statewide office.

“This is a garbage model,” one tweeter wrote. 

It is one model among many, captured at a particular moment. A broader picture with other models and times tells a different story, about that race and others in Oregon. 

The tweeters did at least consider a poll aggregator rather than a single poll. Individual polls, even when well conducted, should be only lightly relied on. An aggregator incorporating a number of polls and other data has stronger predictive value. 

The arresting 4th District result came from Decision Desk HQ, which as of early August still was listing the race as a toss-up, and placed the 5th and 6th districts (in the northern Willamette Valley, on the east and west side respectively) in the same category. Oregon and Michigan were the only two states listed with as many as three toss-up House contests. Every major aggregator lists Oregon’s 1st and 2rd districts as strongly Democratic and the 2nd as strongly Republican.

But the national aggregator take on districts 4, 5 and 6 do vary. 

Over on the site 270 to Win, the picture shifts. There, the Oregon 5th District contest is the only Oregon House seat listed as one of the nation’s “most competitive,” one of 25 with that ranking. Both of the other two relatively competitive districts here were listed as “likely Democratic.” The catch here is that 270 doesn’t list polling numbers and may be relying more on historical data. (Users are allowed to play with the maps interactively, however, which can be entertaining).

One of the most often-cited aggregators, RealClearPolitics, also focuses on the same three Oregon congressional districts, but ranks them differently. There, the 5th Ddistrict is listed as a toss-up, but the 4th is classed as “leans Democratic” and the 6th as “likely Democratic.” Again, not much actual polling data is listed by the site to back up the estimates. 

For more data breakdown, the best source probably is the aggregator that gets more attention than any other: FiveThirtyEight. This site, founded by Nate Silver, lists few tossups and places all six Oregon districts in some kind of leaning category. As elsewhere, the 1st and 3rd are listed as solidly Democratic and the 2nd as solidly Republican. Among the more competitive districts, both the 4th and 6th are considered likely Democratic, and the 5th as leaning Republican.

Do the differences between likely and leaning matter? Statistically, they do; the chances of a Republican win in the 5th are estimated at close to two in three, but a Democratic win in the 6th better than four in five. The editors at FiveThirtyEight would be quick to note, though, that these numbers can change drastically over time. 

They also would note a lack of current polling information. Historical voting patterns are factored in, and as voting registration numbers, and sometimes estimates by political observers. FiveThirtyEight lets you sift through these data sources, however; it offers prediction estimates in variations like “lite” (poll numbers only), classic (some other hard data added in) and deluxe (the kitchen sink, with pundits estimates and more factored in.) And in many House districts, reliable polling information is scarce. FiveThirtyEight is rigorous in scrounging polling information, but it lists only one recommended poll in the 5th District race (which shows the candidates just a point apart, within the margin of error), and no poll information is listed for District 6 at all. 

So, fun as they can be to track, rely on even the aggregators with caution.

On the other hand, all of the aggregators are in agreement that Sen. Ron Wyden is safe for re-election.

And governor? FiveThirtyEight, which cited three polls, gives Democrat Tina Kotek 70 chances out of 100 to win, and Republican Christine Drazan 30; non-aligned Betsy Johnson didn’t register with them. (That is, Kotek gets a rating of 70 in the deluxe aggregate model; in the classic model she gets 67, and in the lite version 73.) RealClearPolitics didn’t weigh in at all. 

The bottom line? As we head into late summer, little about this year’s Oregon elections is hard wired.