How to Read the Polls in This Year’s Midterms

It will be trickier than usual, because of changes in polling and the specific states involved.

Comments: 26

  1. Why was this sentence stuck mid article but never developed or explored: "These polls wouldn’t be so problematic if we could simply ignore them and focus on nonpartisan polls. "
    You can unless someone else is writing this column.
    Now that you dudes have noticed that while GOP was wailing about "skewed polls" during 2012 election for President Mittens, they were actually skewing the polls intentionally. What the math geeks fail to discern is this is intentional double speak on the part of gray-faced rich men to manipulate elections, kind of an anemic geriatric coup d'etat. Overwhelming support for one of the ranters is intended to discourage participation from busy working sane people who may weigh the need to show up at the polls, ergo allowing the aggressive white male supremacists to win again by cheating is some way or another. Add it to the list of voter discrimination measures by the holy-rollers. The likely reason the nerd squad misses this partisan manipulation is they were probably doing math problems or building computers while the rest of us were watching Wyle E. Coyote battle Roadrunner with all of his new-fangled Acme stunts. I am glad to see our hours of dedication to cartoons paid off, as we recognize the behavior as the old men temper tantrums. They aren't stupid anymore than the wildly crazy mob bosses are stupid when they act that way in court. Thanks for catching on that many red polls do not use science; remember, they hate science and smart people.

  2. I have not trusted polls since I had to compose and conduct a poll for a college class. The results depend on the question and how it is asked and who is asked. Polls are manipulated like everything else.

  3. There is a lot of mumbo-jumbo about election polls. The media talk about plus or minus X amount, but that depends on a random sample, which does not exist.

    I have done work in earlier politics, and the insiders like FDR's Jim Farley were extremely good at judging the result (they used figures like 50,000 margin rather than %). There are all sorts of signs, often comparing attitudes of individuals from the last election for example. I sometimes wonder if modern pollsters don't talk with insiders to get their "feel" and then weight to get that result.

  4. The Tea Party crowd here in North Carolina is upset at Tillis for beating their guy in the primary and have adopted the idea that Tillis is not conservative enough. Their support will give Kay Hagan the edge and she will be reelected I think.

  5. No Tea Party member will vote for Hagan. She is a reliable Obama stooge.

  6. "Mr Citizen, do you plan on voting for the Democrat Corporate Party that represents big money but still has some concern for the rest of us or the Republican Corporate Party that represents big money but has no concern for the rest of us?"

    That's the reality of our so called "elections" in our country today.

  7. Interesting analysis--thank you!

  8. Your comments about the Arkansas polls would have been more meaningful if you had reported the historical error data about each of the polling firms. For example, in the non-partisan polls where only Rasmussen is calling it a Republican win, doesn't that polling company have a history of erring on the side of the Republicans (and often, by quite a lot)?

  9. Partisan politics that use polls to try and manipulate the masses, old polling techniques that don't fit contemporary communications paradigms, and respondents who flat out lie because they resent the media have all but mortally wounded modern polling, at least for local and regional elections. I think the nation is large enough that you can get a feel for the presidential race, but Senate, local reps, governors, if it's not a blowout you're not going to know until election day. And even then -- just ask Eric Cantor.

  10. Here we go again. Politics as a game for politicians, and psephologists

    For the rest of us it might be the beginning of wisdom to see that the alternation of Republican and Democrat means very little and has little real world impact and that what we are being offered is fake choice. Same here in UK. If this is 'democracy', it's not worth much.

  11. Instead of the title, "How to Read the Polls in this Year's Midterms," how about "Why read the polls in this year's midterms"?

    I simply don't understand the media's obsession with predicting results. I know that pollsters like to get paid, and politicos hope to influence voting patterns by essentially handicapping races. Just like a Superbowl. If political betting were allowed in elections, wow, what a boon for state and federal coffers, assuming the money stayed there instead of going to the pockets of pollsters.

    Enough with the polls now. With undecided voters not deciding until a few days before, such polls are meaningless and a waste of media time and energy. Instead of predicting the midterms, how about more hard reporting on why the country is in such a mess?

  12. While I agree about the lack of hard reporting on issues we are facing and the general laziness of a lot of news people in simply spouting what pops up on Yahoo without further analysis or research, I disagree with the idea that polls are meaningless and a waste of time.

    Good polling tells you the lay of the land, so to speak. It gives you an idea of what is possible, so that when a "surprise" happens you can tell if it the kind of surprise that should be investigated as a federal crime (i.e. election rigging). An example of good polling has to be the 2012 election results (good polling ended up matching the actual election results) and a good example of the opposite situation (bad/no polling, surprising result) is the Cantor/Brat election in Virginia. What good polling would have told you about Virginia was that Brat was not the massive underdog he was made out to be - something that people in that district had a sense of according to after-the-fact reporting, but never really showed up in the polls because of the organizations doing the polling and that there were few polls done.

    There are guidelines in Statistics for getting a good, representative sample. Sometimes these have to be compromised because of costs (like in Alaska) or other problems getting results, but they can also be intentionally compromised by the pollster. It is hard to tell which without looking at the methodology of the poll - which brings us back to lazy reporting.

  13. As accurate as Silver has been in the past, it looks like he is starting to hedge his bets regarding the 2014 senate now. Could it be that people are starting to notice Obamacare's rising success, the continuing lunacy of the tea party, and the GOP still failing to heed their own 2012 post-mortem analysis?

  14. Silver isn't here anymore.

  15. This is Nate Cohn, not Nate Silver, he is no longer with the Times.

  16. Oooops! I stand corrected. Thank you

  17. the partisan polls I discuss in my echo chamber often convince me and my like-minded friends that my candidate will win, but my candidate loses, which proves the election was rigged.

  18. Polls based on responses to poll questions have possibly outlived their true purpose. They’ve become more of a sporting event… with respondents using them to send messages… then following the results on social media and the evening news. Even if, for example, a respondent will ultimately vote for a particular candidate, he/she often feels it’s not good to express unfettered satisfaction. As a way of keeping politicians on their toes, respondents would rather send a message with a negative response.

    Pollsters have an agenda, respondents have an agenda, and the news media reports results and discusses them as if they’re handicapping the World Cup. The entertainment value has crowded out the information value.

  19. The media are obsessed with polls. As Nate Cohn shows, in 2014, voters need to know which bias a partisan polling organization is presenting. With the cost of quality polling rising that includes the large increase in cell phone use, makes it almost impossible to predict outcomes in close states. Some popular polling organizations poll only 30% of cell phone users. Studies (2013) show that more young people and Democrats use cell phones; and only 13% of voters 65 & older used cell phones. This points to a bias against younger voters, and Democrats. Older voters, who mostly use land line phones tend to be more conservative. We only need to look at the polling of Eric Cantor's campaignr to see the flaws in polling, when he was "polled" ahead of his opponent by a large percentage. Excellent article by Mr. Cohn.

  20. In Arkansas it looks like if you throw out the Rasmussen poll -- which Cohn describes as "conservative leaning" rather than truly independent -- the average result for independent polls and Democratic polls is identical. Seems like its really not a problem with partisan polls, but with Republican polls

  21. I think polling data is getting weaker and less reliable because many people will not respond. As I said in a similar article I can't believe pollers are able to get any answers any more. With legal (political, non-profits and other business relationships) but still unwanted telemarketing calls and scam calls that all border on harassment along with the knowledge that big data continues to build databases probably using some polling information shared to identify me, my preferences, illnesses, political and sexual orientation and more I will not answer any calls or any polls, customer service questions or anything else period. None. I will not contribute or help these people. In my opinion, they are in the process of putting themselves out of business with their overly sophisticated data bases and data mining. And in my opinion, it won't be too soon. George Orwelll is way too close.

  22. Nate Cohn, I'm pretty sure Rasmussen, which is listed in the non-partisan polls, should be in the Republican column. While Scott Rasmussen identifies as independent, he is also a frequent guest of Fox News, and used to write a column for World Net Daily. His polling numbers are often used to support Republican talking points, and Nate Silver showed in 2013, here at the Times, that Rasmussen numbers had the largest house bias towards the GOP in the 2012 elections.

  23. Caller ID killed the telephone poll.

    Used to be one always answered the phone because you didn't know if it was a family member in need or something mundane. Today, many people don't answer a number they don't recognize today. I'll guess repose rates for cell phones are lower than that of landlines.

    Polls are weighted towards those without caller ID.

  24. Your mistake is one all-too-typical of American punditry. To further the delusional "both sides are to blame" mentality, you used a blatantly partisan "Rasmussen" to skew your "non-partisan" results.
    Rasmussen is a Republican pollster. Categorize him as he belongs, and the "non-partisan" polls line up with the in-house Democratic polls. Pryor is ahead.
    The Republicans are a fantasy based party, looking for confirmation bias and a feel good result. The Democrats are a reality based party, looking to determine who is actually winning.
    This is the same kind of reporting that left the Republican-friendly punditry with their pants down in Election 2012.

  25. From the article: "Libertarian candidates in North Carolina usually receive a fraction of the vote. "
    Has any candidate, of any party, ever received more than a fraction of the vote?