Friday, October 25, 2013

When a poll is and isn't an outlier

As I'm a little under the weather today, a short post to highlight my latest article for The Globe and Mail, on the topic of outlier polls.

With a series of polls out in the last week giving the Conservatives everything from 26% to 32%, and the Liberals from 32% all the way to 40% support, the piece is somewhat timely. EKOS was the first out of the gate and had some unusual results, making it look like an outlier poll. But the new numbers from Nanos and Forum go a long way to back them, as the updated federal polling averages can attest to.

The Globe article is available to Globe Unlimited subscribers only. If you like reading this sort of deep polling analysis (which, if you'll allow me to say, you'll have a hard time finding anywhere else except right here), please consider getting a subscription. In the end, the work I'm lucky enough to do for places like the Globe help subsidize my work on this site.

Speaking of which, a few days ago was the fifth anniversary of the launch of ThreeHundredEight.com, which began all the way back in 2008. I just wanted to give a big thank you to all of my readers, who have made this project possible. And thanks also to those of you that I have interacted with in the comments section over the years, on Twitter, and in real life, who help make this strange job of mine so fulfilling. I'll do my best to keep your interest for the next five years. It helps that those coming years are setting up to be pretty darn interesting.

13 comments:

  1. Éric,

    Félicitations à l'occasion du 5e anniversaire de 308. À mon avis, une des meilleurs sources de données sur la scène politique canadienne. Bravo et bon courage

    ReplyDelete
  2. Congratulations on hitting your five year mark, I thoroughly enjoy reading your sight. Yours is one of the few that make it onto the RSS subscription. Here's to another five successful years for you Eric.

    God Bless. And keep up the great work.

    ReplyDelete
  3. How does one find the regional Nanos numbers? Are they no longer publishing these? The less disclosure there is about numbers from any pollster, the less likely I am to trust them.

    ReplyDelete
    Replies
    1. Nope, no regional numbers. I agree it is a bad thing.

      Delete
  4. Hello Eric,

    Thank you for your work. I read your blog regularly and appreciate your professionalism. Congratulations on the past 5 years. I hope you keep going strong.

    Sincerely,

    Derek.

    ReplyDelete
  5. Many thanks from a Sharbot Lake political enthusiast. I appreciate the local and provincial polling as well as the national. Ken Fisher

    ReplyDelete
  6. Many thanks for your dedication to reporting polls not only on the federal level, but also provincial and municipal. I check in at least once a week.

    ReplyDelete
  7. Thanks Eric for all your hard work. This place has become a daily necessity !

    Again many thanks.

    ReplyDelete
  8. Two questions Eric. I always use polls to give myself a picture of what is happening generally in a certain area. For example according to recent polls in eastern Canada the Tories are well under the layer of support they received in 2011. This would suggest they could lose half their eastern seats come election time. My question for you is practically speaking how much extra credit should be given to an incumbent party or historical voting history providing they haven't completely tanked in the polls? For example in some of your projections (which I understand are based purely on polling data) you have the liberals and NDP winning multiple seats in alberta. Practically speaking I just don't see that happening. I look at the west the same way as a whole in that I really dont see the conservatives dipping below 70 seats unless they really tank in the polls. In Ontario again barring a catastrophe for the conservatives and with 2 progressive parties competing on the left i don't see how the Tories will win less than 70 seats under the new boundaries. I guess what im trying to say is I like to see polls for a general trend but practically speaking the seats have to come from somewhere. I understand one has to be scientific but do you think injecting your own opinion into your projection might make them more accurate?

    ReplyDelete
    Replies
    1. The projection model does take into account the effects of incumbency, giving incumbents a bonus when their party is dropping in the polls.

      As to an injection of my own opinion, absolutely not! The model has to be objective and applied uniformly, otherwise it is just a guess.

      Delete
  9. Eric, I appreciate the work that you do; and that of other experts in the polling field (e.g., Frank Graves, Dan Leger). Unfortunately, some people on low, fixed incomes can't fit the price of subscriptions to newspapers and magazines into their budgets. If we're mobile enough to get to a well-equipped library, we might catch up there. Otherwise, we're out of luck.

    There are several news media to which I'd subscribe if I could manage it. The Globe would be one of them. Maclean's Magazine and the National Post also come immediately to mind.

    I wanted to drop off this note because it's an issue that comes up time and time again, and it bothers me. Not all non-subscribers think print and other media should be free. We're not all foolish enough to suppose that DOING quality journalism, investigative reporting and analysis comes cheap. Some of us know that the people who write such material and the organizations that support them must be paid, and paid well.

    ReplyDelete
  10. Thank you so much for your hard work and congratulations on 5 years.

    ReplyDelete
  11. Congrats on 5 years Eric, you're doing a great job!

    ReplyDelete

COMMENT MODERATION POLICY - Please be respectful when commenting. If choosing to remain anonymous, please sign your comment with some sort of pseudonym to avoid confusion. Please do not use any derogatory terms for fellow commenters, parties, or politicians. Inflammatory and overly partisan comments will not be posted. PLEASE KEEP DISCUSSION ON TOPIC.