FiveThirtyEight

Welcome to Pollapalooza, our weekly polling roundup.

Poll of the week

A new Texas Senate poll came out this week showing Democratic Rep. Beto O’Rourke within 1 point of Republican Sen. Ted Cruz. It gave new life to a familiar question: Could Texas finally flip blue? Normally, this would be the point where we lecture everyone about the dangers of putting too much stock in one poll. But this poll wasn’t even much of an outlier — it came on the heels of a few others that also show O’Rourke within a few points of Cruz. That’s an awfully close race for a state that President Trump won by 9 points in 2016. Is it time to start betting on a Democrat in Texas?

We still say you should hold onto your chips.

OK, first the new survey: Emerson College’s poll of registered voters, conducted online and through robo-calls, reported 38 percent support for Cruz and 37 percent for O’Rourke, well within the poll’s 4-point margin of error. About a fifth of respondents remained undecided.

Emerson was the sixth consecutive pollster in our database to find the Texas Senate race within single digits:

Summer polls show a close Senate race in Texas

But how much significance should we really give to a few polls this far out from Election Day? FiveThirtyEight will publish a Senate forecast soon that will give you a more comprehensive answer, but in the meantime, let’s just look at how closely Senate polls conducted in the late summer have matched the eventual election results.

We searched our database (which goes back to 1990) and collected all the Senate polls that were conducted in August of an election year, giving us 594 polls in all. We took the margin separating the candidates in each August poll and then subtracted the margin that separated them in the final election results to find out how much each poll missed by. We then averaged all the misses in each year to figure out how far off that year’s August polls were from the final results:

Average error of Senate polls taken in August

Average difference, in percentage points, between August polling margins and the final election margin

The August polling averages were off by between 5 points and 14 points in any given year. On average since 1990, they were off by about 8 points.

Of course, that error could cut either way — it doesn’t necessarily favor Cruz. The senator leads by 3 points, on average, in polls taken this month. If this year’s polls are off by that average amount — about 8 points — then we really shouldn’t be surprised by anything from Cruz winning by 11 points to O’Rourke winning by 5. And even that understates how wide the range of possible outcomes still is — that’s just what could happen if the polls are off by the average amount. Who’s to say polls won’t be off by 14 points like they were in 1990 and 1994?

None of this is to say O’Rourke has no chance. Expert ratings, such as those provided by the Cook Political Report, Inside Elections and Sabato’s Crystal Ball, have a track record of accuracy in forecasting elections, and all three have pinned the race as either “lean Republican” or “likely Republican,” meaning Cruz is favored but they believe O’Rourke has a shot to at least make the race competitive. Moreover, O’Rourke, a three-term congressman, has pulled far ahead of Cruz in the fundraising race. According to the Texas Tribune’s analysis of Federal Election Commission data, Cruz raised $15.6 million from the start of 2017 through July; O’Rourke pulled in a whopping $23.6 million. That sum makes O’Rourke’s campaign the third-biggest fundraiser among all Senate candidates this election cycle, according to OpenSecrets, a nonpartisan research group that traces money in politics.

O’Rourke also has the potential to gain ground by improving his name recognition in the next two months. The Emerson poll found both candidates with similar favorability ratings, but 38 percent of Texans either felt neutral toward O’Rourke or said they had not heard of him, compared to only 19 percent who said the same about Cruz. In a national environment where the generic ballot favors Democrats and Congress’s approval ratings remain low, incumbency may not be as much of an advantage for Republicans as it has been in some other years. This suggests that as O’Rourke becomes more familiar to voters in the next couple of months, he may have more potential to win over undecided voters than Cruz does.

Still, not a single poll so far has shown O’Rourke ahead of Cruz. So don’t make any serious bets just yet — it’s still too early to get carried away with speculating about big changes in Texas based on August polling alone.

UPDATE (Aug. 31, 4:00 p.m.): We noticed that some FiveThirtyEight readers were discussing in the comments section that our approach of showing the average error of polls tells you only that the average August poll is about 8 points off historically. But how about the average of all August polls in a particular race — has that more accurately predicted the election? Since you asked, we calculated the average of polls for each individual race, then filtered out any race that didn’t have at least three polls (as this Texas race does). We found that the average of polls in each race was also off by approximately 8 points. If you’d like to dive in further, we’ve also published the data behind this article, you can check it out on our data page or grab it from GitHub.

Other polling nuggets

Trump approval

Trump’s net approval rating currently sits at -12 points, according to our tracker. (That’s a 41.5 percent approval rating and a 53.5 disapproval rating.) One week ago, his net approval was -10.9 points; 42.1 percent of Americans approved of Trump’s job performance and 53 percent disapproved. At this time last month, that net approval was -11.7 points — 41.3 percent approval, 53 percent disapproval.

Generic ballot

Per our tracker of generic ballot polls, Americans currently opt for a hypothetical Democratic House candidate over a hypothetical Republican by a 8.2-point margin (48.2 percent to 40 percent). One week ago, their lead was a similar 7.8 points (47.8 percent to 40 percent). At this time last month, our tracker sat at Democrats 46.8 percent and Republicans 39.6 percent, or a 7.2-point Democratic advantage.

Check out our 2018 House forecast and all the polls we’ve been collecting ahead of the midterms.

Filed under