The Dim-Post

May 19, 2012

Tracking poll page

Filed under: polls — danylmc @ 7:00 am

I’ve created a permanent page for the tracking poll graphic. I’ll try to update it as new poll results come out.

I always knew in an intellectual sense that individual poll results were largely meaningless and that the long term trends were what counts, but – to me – this chart really hammers that home. The data is as noisy as hell. Take a look at some of the results just prior to the 2008 election: they showed a really close race, which wasn’t actually close at all.

So every time there’s a new poll out and the political editor publishing it breathlessly speculates on the reasons behind the ‘dramatic gains’ or whatever, it might help to remember this chart and the semi-random distribution of most of these individual results.


  1. I’m curious – is the variance from reality that turns up in the polls within what we’d expect from their confidence intervals, and plausible changes in opinions over the time elapsed between polls (and between polls and the elections) or is it higher? Higher would be interesting – reflecting, presumably, some sort of flaw in the polling process(es).

    Comment by terence — May 19, 2012 @ 8:34 am

  2. Cool idea. But it might need a bit of tidying up as at the moment it is overlying the right-hand column (on my screen at least).

    Comment by bmk12 — May 19, 2012 @ 9:26 am

  3. Would it be possible to make a version with the election results marked in? I’m interested to see how they compare.

    Comment by Merrin — May 19, 2012 @ 12:15 pm

  4. It would be nice to see the actual election results as well.

    Comment by passerby — May 19, 2012 @ 2:43 pm

  5. @bmk12

    That’s your screen resolution at fault. Dude it’s 2012; get with the 1080p.

    erm in windows: control panel | display (settings) | adjust resolution (pref beyond 640×480)

    or just post your OS and I’m sure a nerd round here will help..

    Comment by Luke. — May 19, 2012 @ 7:24 pm

  6. Actual election results are quite interesting.

    # Highlight the actual election results.
    elections <- grep('election', polls $ Company)
    matpoints(polls $ date.num[elections], polls[elections, selected.parties], pch=1, cex=2, lwd=2, col=selected.cols)

    Comment by pete — May 19, 2012 @ 9:40 pm

  7. Kiwi Poll Guy is already running these here:

    Comment by Eric Crampton (@EricCrampton) — May 19, 2012 @ 10:41 pm

  8. Worth comparing to Rob Salmond’s as well:

    Seems like nobody forces the best fit line through the election results.

    Comment by bradluen — May 20, 2012 @ 9:47 am

  9. @Luke

    My screen resolution is at it’s highest possible 1440 * 900. Which isn’t particularly small for a laptop.

    Comment by bmk12 — May 20, 2012 @ 9:59 am

  10. Take a look at some of the results just prior to the 2008 election: they showed a really close race, which wasn’t actually close at all.

    It wasn’t? Banks loses Epsom or Dunne loses Ohariu, and there aren’t the votes to pass the government’s signature economic policy. That seems pretty close.

    Comment by Graeme Edgeler — May 20, 2012 @ 11:38 pm

  11. “Thanks John. Today’s poll shows no statistically significant shifts for any political party, though we can’t rule out massive underlying shifts in support that just didn’t happen to show up. If a government was to be formed based on these results alone, National would almost certainly get somewhere between 51 and 61 seats, we just don’t know.

    “But if we arbitrarily weight this with other recent polls, it’s most likely National’s support hasn’t shifted more than one or two percent either way, and there’s still about a 60% chance they’ll win, and 40% they’ll lose, unless anything changes, or these polls are biased, which they always have been in the past.

    “Really, this is all tea leaves at twenty paces, and it just depends what happens on the day, voters sometimes do the weirdest things at the last moment, like saving Winston. Back to you, John.”

    Comment by tussock — May 21, 2012 @ 4:01 am

  12. @tussock: I think that would be preferable to the status quo, where political editors try to mislead the audience by pretending to read the tea leaves.

    Comment by pete — May 21, 2012 @ 4:05 am

  13. @bmk – found the same thinng, but the window wasn’t set to fulls creen. When I widened it (dragging the right hand edge) that right hand mnenu went with it.

    Comment by MGK — May 21, 2012 @ 9:47 am

  14. Something that jumped out when I replotted this with the election results highlighted: Every single poll in 2009, 2010, and 2011 overestimated National’s election result.

    Comment by pete — May 21, 2012 @ 3:17 pm

  15. @11 – damn right, it makes me so angry when the media tell me the truth.

    Comment by BeShakey — May 22, 2012 @ 1:20 pm

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Blog at

%d bloggers like this: