More rambling on polls & bias (UPDATED)

I should clarify that my post yesterday was not intended to produce another poll of polls, but to explore differences between polling companies. As one commenter suggested, making a line more sensitive would probably be more interesting/useful for a poll of polls.

And as I eventually realised, with Roy Morgan contributing over half the poll results, it does weigh heavily on the poll of polls. With those two thoughts in mind, I’ve removed Roy Morgan from this chart, and also added a Loess (span = 0.5), which was the sensitivity used in the Wikipedia charts for the 2011 election.

As you can see, the more sensitive line does certainly vary more, and certainly counters Stuff’s headline yesterday suggesting that National were bouncing back. It’s also really obvious here just how high the Fairfax Ipsos (pink dots) is compared to most of the rest of the polls at the moment. Interestingly, at this point in the last election, it was the Roy Morgan poll that seemed to be tracking much higher for National.

This is perhaps the attraction of producing some sort of poll of polls, but trying to correct for the polling frequency, so that the average isn’t over influenced by polling frequency makes sense. I didn’t mention David Farrar’s rolling average of polls yesterday (probably because it isn’t presented as a graph), but this close to an election, he only uses the last poll, so those figures would not be influenced by that. Some discussion yesterday on this, and the suggestion that Roy Morgan ought to perhaps be overweight, as it does add more information, because it is more regular. However, perhaps an alternative would be average each polls trend. Pictured below are loess(0.5) by polling company. Interestingly, they don’t as a rule particularly follow each other, though (again in contrast to the Fairfax headline yesterday), do suggest that National’s support is softening.

Finally, back to bias, Russel Brown, on twitter, asked about how my figures would look for the smaller parties.

Fairfax have been low on NZ First (corresponding to their being high on National?). Roy Morgan again driving the trend, and with the highest estimates for NZ First.

3 News Reid Research have the highest estimates for the Conservatives overall. This is one where Roy Morgan don’t dominate, and they don’t seem to have detected the uptick in support that all the other polls have seen recently.

Not enough data points really

Definitely trending down.

Not much to say here, other than that this make’s Roy Morgan’s policy to round to 0.5 percent (and 1 percent for larger parties) really obvious, relative to the other pollsters.

So final *bias* summary

  1. Fairfax are high on National, seemingly at the expense of NZ First
  2. Roy Morgan are higher on Labour and the Greens, seemingly at the expense of National.

UPDATE: The point I meant to make this morning (but forgot)

It seems reasonable to me that there will be systematic quirks in each firms polling methodology (handily outlined here). Combined with changes over time in the population, and perhaps polling companies changing their methods (eg to try to increas response rates), there ought to be some variance attributable to the polling company. Trying to work out whose is “right” is a fool’s errand, but attempting to account and model for it seems like a good plan.


About James

I'm a scientist. Sometimes I get distracted.
This entry was posted in Uncategorized. Bookmark the permalink.

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 )

Google photo

You are commenting using your Google 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 )

Connecting to %s