Talk:Opinion polling for the New Zealand general election, 2017

From Wikipedia, the free encyclopedia
Jump to navigation Jump to search
WikiProject Elections and Referendums (Rated List-class)
WikiProject iconThis article is within the scope of WikiProject Elections and Referendums, an ongoing effort to improve the quality of, expand upon and create new articles relating to elections, electoral reform and other aspects of democratic decision-making. For more information, visit our project page.
 List  This article has been rated as List-Class on the project's quality scale.
WikiProject New Zealand / Politics  (Rated List-class, High-importance)
WikiProject iconThis article is within the scope of WikiProject New Zealand, a collaborative effort to improve the coverage of New Zealand and New Zealand-related topics on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
 List  This article has been rated as List-Class on the project's quality scale.
 High  This article has been rated as High-importance on the project's importance scale.
Taskforce icon
This article is supported by the New Zealand politics task force (marked as High-importance).

Graph improvements[edit]

Hi, Here we go again. Thanks to everyone who got this up and rolling again. I will update the graphs accordingly, but before I do so, I'm open to suggestions for improvements etc etc. I can't guarantee everything, but it doesn't hurt to ask. Please add them below--Trevva (talk) 08:26, 30 November 2014 (UTC)

  • Publication of the code in a GitHub repository --Trevva (talk) 08:26, 30 November 2014 (UTC)
  • Suggest adding a margin of error column. Elpaulonz (talk) 20:09, 15 October 2015 (UTC)
  • Is it time to add the graph again? Schwede66 04:32, 22 January 2016 (UTC)

Please bring in the graph ASAP ! It offers the trends hidden in the mass of data, and deals to the weird claims that spring from the polls. — Preceding unsigned comment added by Detxamy (talkcontribs) 23:14, 5 February 2016 (UTC)

It appears Trevva has not edited anything on Wikipedia since the comments in 2014. Is there someone else who is interested in creating the graph? Incredtent (talk) 22:28, 9 May 2016 (UTC)
So I have no idea why (having not really done anything on wikipedia forever), I've had a shot. I'm not 100% sure that I have the GAM smoother working properly (some of them are quite straight lines at the moment), but this should at least get the ball rolling. --Limegreen (talk) 12:06, 13 May 2016 (UTC)
Hi, I wonder if it would be more beneficial to graph polls of parties registered in the 2017 election rather than parties from the 2014 election? Thanks. LePeche (talk) 20:35, 9 March 2017 (UTC)
Anyone able to correct the graph to where other poll averages are? Labour are currently averaging 37-40% based on recent numbers. Shouldn't this be shown on the graph? FiliKiwi (talk) 10:42, 7 September 2017 (UTC)
Fair point. Limegreen, could you please swing into action? Are there any coding issues that prevent you from doing so swiftly? If so, please state what the issue is and I'm sure somebody else will have time to clean up the table so that you can run R. Schwede66 21:15, 7 September 2017 (UTC)

UMR Polling[edit]

@Limegreen: @Schwede66: @Trevva: Sorry had a query about this Mood of the Nation 2016 report released by UMR research back in February 2016. On page 29 of the pdf there is a graph displaying polling results conducted between October 2014 to December 2015, do these polls merit inclusion in the Individual Polls table? - Sleepingstar (talk) 19:59, 6 December 2016 (UTC)

My view is that they are released to far after the fact to be of much use. NealeFamily (talk) 03:02, 7 December 2016 (UTC)
@NealeFamily: I disagree; we should write in all polls regardless of the date they were released. J947 05:20, 7 December 2016 (UTC)
Yes, but the document does not say when the surveys were taken. It only has the month which they fall. The document seems to imply that they publish the results each monthly, so if you have access to them and they state the survey period then they might be relevant. NealeFamily (talk) 07:28, 9 December 2016 (UTC)
@NealeFamily: I understand, but there could still be a separate table for the UMR polls. J947 21:44, 9 December 2016 (UTC)
There are two unpublished polls, UMR (for Labour), and Curia (for National), both of which intermittently become public. Colin James, posting for Radio NZ, has access to both for his poll summary[1]. From the perspective of producing the graphs, it will be quite strange to be adding much more data very much later. It might be better to mention that these unpublished polls exist, and linking to them, and keep this table for public polls. Limegreen (talk) 02:34, 13 December 2016 (UTC)
Good idea. J947 03:31, 13 December 2016 (UTC)
@Limegreen: Hi so I am new in this project and wanted to help. So just recently, there are 2 sources containing/mentioning about a UMR poll. I added it in, though an anon user reverted me stating that it was a poll from a political party. So should we add it in? Typhoon2013 (talk) 03:04, 20 August 2017 (UTC)
The consensus so far has been not to include either UMR or Curia polls, since these are not regularly published and available to the public. They sometimes get leaked, but this means we do not get to see the results of all their polls, and the ones which are leaked may not be representative.-gadfium 04:18, 20 August 2017 (UTC)
@Typhoon2013: The other salient point about these other two polls is that are usually leaked strategically, ie when they send a clear message relating to the party that commissions them. To this end, I think it's good to keep it clean with only the publicly released polls. Quite a few people use this page to scrape the data as the publicly available polls. Limegreen (talk) 15:57, 24 August 2017 (UTC)
Because UMR and Curia conduct research on behalf of political parties I don't think they should be included. The source of the material will clearly have issues around Verifiability and Neutral point of view. Kiwichris (talk) 08:17, 30 August 2017 (UTC)
I agree with the concerns raised by Kiwichris and suggest that UMR and Curia be removed. Schwede66 10:19, 30 August 2017 (UTC)
Surely no harm done if kept in a separate category with a disclaimer? (i.e. like currently). They're good for cross-referencing with actual polls (and they are from a reputable company, and it seems that news agencies generally see the original source material before reporting on it). — Preceding unsigned comment added by (talk) 11:14, 30 August 2017 (UTC)
I disagree, disclaimers aren't sufficient to subvert Wikipedia policies. Kiwichris (talk) 05:23, 31 August 2017 (UTC)

Revised graphs[edit]

I've just change the smoothing line from k=10 to k=5, which means we can add the TOP party for the first time. Also, although these gamm smooths remove most of the 'operator' character, I wonder if they are mostly defaulting to quite straight line fits, and missing some genuine variability. Compare the current graph to the 2011 version,_2011 --Limegreen (talk) 23:04, 30 July 2017 (UTC)

The UK and Australian opinion polling articles use a moving average as a trend line. Perhaps that may improve showing variability. Lcmortensen (mailbox) 02:11, 31 July 2017 (UTC)

Collapsible tables divided by years[edit]

This really doesn't seem necessary considering how short New Zealand election cycles are. It may be more reasonable for the UK where their election cycles can be up to 5 years long. If we could at least connect the tables together so it looks more organized - but otherwise it feels a bit unnecessary.— Preceding unsigned comment added by Clesam11 (talkcontribs) 6 August 2017 (UTC)

I assume you're referring to the recent (last hour or so) breakup of the poll results into sections: 2014-15, 2016, 2017. I object to the breakup because it breaks the continuity of the table, and makes it appear as though poll results in one section were not influenced by events in the preceding sections. That is, the polls are divorced from what previously happened, as though things start afresh with each new year. I see no need for the breakup; it's not like the chart is going to grow to huge proportions; the election is only 7 weeks away. Akld guy (talk) 05:46, 6 August 2017 (UTC)
I agree. It's also fairly difficult to contact the person who made it seeing as he doesn't have a account. I'll reverse the changes for now, if he stops by at some point he can argue for it. Clesam11 (talk) —Preceding undated comment added 06:59, 6 August 2017 (UTC)
Some of these changes also broke the script that produces the figures, and it has taken me an age to debug it; so I'd definitely vote for less screwing with the format of the tables! Limegreen (talk) 11:46, 11 August 2017 (UTC)


I think the source given for the poll that lists Julie Anne Genter as the Green candidate for Epsom is just an oversight by the author. Barry Coates was listed as candidate for Epsom on the Green Party website by 2 May while the Curia poll was conducted 21-28 May, by which time he had already become the candidate. As Genter had contested Epsom in 2014 the article was probably erroneously assuming she would again in 2017, however she has since been confirmed in Mt Albert. Therefore is seems inappropriate to me for Genter to be referred to here as the Green candidate at the time. Kiwichris (talk) 06:49, 14 August 2017 (UTC)

You have provided no evidence for your claims. Do not under any circumstances revert me again until you have provided compelling evidence of your claims here. Akld guy (talk) 07:00, 14 August 2017 (UTC)
An archive snapshot of Coates' profile page (can be accessed here via Waybackmachine) from 2 May lists his electorate as Epsom as early as that date, though potentially any time between then an the previous capture of the page on 11 April. Kiwichris (talk) 07:15, 14 August 2017 (UTC)
That is NOT compelling evidence. You are trying my patience. Do not revert again until you have provided compelling evidence. Akld guy (talk) 07:21, 14 August 2017 (UTC)
Actually, I think the archive link found by Kiwichris is compelling. The error appears to be on the Curia page.-gadfium 08:50, 14 August 2017 (UTC)
You both are kidding, right? All I see is a 2017 calendar with 2 May and 21 May highlighted. That's it in its entirety. No explanation of what it means. What the hell is going on here? Akld guy (talk) 09:29, 14 August 2017 (UTC)
Click on the 2 May and you'll see what the page looked like on that date, which specifies the Epsom electorate. Kiwichris (talk) 09:44, 14 August 2017 (UTC)
When I click on 2 May, I get taken to a bio of Barry Coates, which doesn't specify any dates that he was appointed. You haven't yet provided a reference for Coates being List MP for Epsom at the time the poll was taken. The reference that I'm relying on is shown in the Poll column and is here. It clearly states that Julie Anne Genter was the Green's representative at the time the poll was taken and also states that Coates was nominated as the candidate later. Akld guy (talk) 10:15, 14 August 2017 (UTC)
If the website on that date says he was the candidate that proves he was candidate long before the poll was conducted. You are right that the page for the Curia poll says Genter, what I'm trying to say is that the poll made a mistake by listing the wrong candidate and we should correct it here. Kiwichris (talk) 10:22, 14 August 2017 (UTC)
Nope, that is WP:SYNTHESIS. You are putting together facts from two sources to derive a conclusion that is not stated explicitly by either. Not permitted. You are also drawing a conclusion that the Curia poll is mistaken. Akld guy (talk) 10:33, 14 August 2017 (UTC)
Hardly, I'm not saying A+B=C, I'm saying A has factual inaccuracies and when evidence was requested to support my claim I produced B. That is not SYNTH. Kiwichris (talk) 05:13, 15 August 2017 (UTC)
You are saying that A+B=C. You're putting together the calendar page's date and the Barry Coates Green page to say that Coates was the candidate on 2 May. You couldn't do that to provide a reference. Neither page explicitly says what you want. On the other hand, the Curia page that I cited unequivocally says that Genter was the candidate when the poll was taken in late May. You may be right, but you can't just dismiss what a poll site says. Presumably pollsters are big on accuracy. Akld guy (talk) 06:21, 15 August 2017 (UTC)
So you are seriously suggesting we continue knowingly citing a blog which contains incorrect information just because the citation says so? That seems like a blatant violation of WP:V to me. Kiwichris (talk) 07:12, 15 August 2017 (UTC)

The only reference I can find to the Epson opinion poll is on the Your NZ blog site. That doesn't seem to qualify as a reliable source. Should the poll be cited at all? If it should then you need a better source. NealeFamily (talk) 08:22, 15 August 2017 (UTC)

Actually, it's a verbatim copy of this ACT web page. I was not the person who put the poll result for Epsom into the article. Someone else did it. Akld guy (talk) 10:08, 15 August 2017 (UTC)
That the only source is the ACT party is a concern. If, hypothetically, Curia did another poll a month later which showed a different result (some other candidate in the lead), would you expect ACT to publish it? I think the best solution would be to remove this poll entirely.-gadfium 19:58, 15 August 2017 (UTC)
I think it would be best to remove it too. It has become too contentious. I'm also mindful of the fact that Curia is an entity run by David Farrar, who although qualified as a pollster, is a declared National Party supporter. Akld guy (talk) 20:57, 15 August 2017 (UTC)

I have moved the Epson table here for interests sake and until a better source is found. NealeFamily (talk) 23:18, 15 August 2017 (UTC)


Poll Date [nb 1] ACT

David Seymour


Paul Goldsmith


David Parker


Barry Coates

Curia[1] 21–31 May 2017 46 30 11 112

2At the time this poll was taken, Julie Anne Genter was the Green Party list MP in this seat. Barry Coates was later nominated as the Epsom candidate.

I also agree that the poll should be removed. If it contains inaccuracies and from a source which has conflict of interest concerns then it is definitely not verifiable. Kiwichris (talk) 05:02, 16 August 2017 (UTC)
  1. ^ "Poll good for ACT in Epsom". Your NZ. 2017-06-19. Retrieved 2017-08-13.

Radio NZ Poll of Polls in Forecasts section?[edit]

Should the Radio NZ poll of polls be included in the Forecasts table? It's equivalent to the Stuff poll of polls (not equivalent to the 3 News and 1 News ones which seem to just be that individual poll translated into seats, if i'm not wrong). Elcalebo (talk) 21:50, 17 August 2017 (UTC)

@Elcalebo: Thanks for pointing it out, have inserted the RNZ data into the data table. Sleepingstar (talk) 06:04, 19 August 2017 (UTC)

Seat projection margins[edit]

I've just altered the "margin" part of the table, and just left the seats of the coalition arrangements of the parties stated. While support partners such as United Future, ACT, Maori would not be necessary for forming a governing coalition/ arrangement, I'm not too sure whether lumping them into the opposition is appropriate (as per the assumptions made when quoting figures such as (66-55) for the One News poll for National-NZ First coalition), since there is still a possibility of supply and confidence to be negotiated with the minor support parties. Likewise the case for the Maori party should Labour/Green/NZ First form a governing coalition...Sleepingstar (talk) 06:01, 19 August 2017 (UTC)

Apocolypse may occur between 21-23 Sept[edit]

This may affect the conservative party vote. Should we include it?

Christian numerologist David Meade has again warned that the mysterious planet 'Nibiru' is about to smash into Earth. According to his theory, the apocalypse will take place between September 20-23, and the clues are written on the pyramids and in the Bible.

Scientists are calling it a hoax with no evidence to back it up (unlike the bible and the pyramids which are a fountain of truth) - the scientists simply state "obviously" as their reason for calling it a hoax:

The scientific community does not agree Nibiru exists. "Nibiru and other stories about wayward planets are an internet hoax," Nasa has said previously. "Obviously, it does not exist."

Source: — Preceding unsigned comment added by (talk) 20:42, 29 August 2017 (UTC)

Wikipedia doesn't engage in conspiracy theories. So the answer to your query is no. Schwede66 20:56, 29 August 2017 (UTC)
But it does include an article about it Nibiru cataclysm NealeFamily (talk) 02:55, 30 August 2017 (UTC)
We don't engage in speculation about whether the outcome of an election will be affected by a disaster based on a theory. If the disaster happens and it affects the election, we will report it. Akld guy (talk) 04:26, 30 August 2017 (UTC)
Well, if we survive the disaster, that is. :-) Schwede66 10:20, 30 August 2017 (UTC)
BUT what if a speculated theory (even one that doesn't come true) is currently impacting the election? All those conservative party voters might hide in their bunkers on election day, fearing Planetary Annihilation! And those conservative party voters are currently polling such strong turnouts too!
(feel free to see this entire section as satire/fun) — Preceding unsigned comment added by (talk) 11:14, 30 August 2017 (UTC)
It comes down to a reliable source. If you can find a reliable source stating that believers in the Nibiru disaster theory are going to avoid venturing out on election day in sufficient numbers to affect a party's result, we may be able to use it. Akld guy (talk) 13:11, 30 August 2017 (UTC)

Scandals and events[edit]

The inclusion of scandals is starting to become a bit over the top. We currently have 10 events over the span of August and only 4 polls. If we start to include every tweet and every possible controversy it's going to take up the entire page. This article is primarily meant to show poll results not every single event that happens during the election cycle. Scandals should be documented on the general election page, not on a page for poll results. To be clear, i'm not saying events like leadership changes and budgets should be removed, but we definitely need to cut back on the amount of events shown. CleSam11 (talk) 12:16, 1 September 2017 (UTC)

I agree with the sentiments above. Kiwichris (talk) 08:26, 1 September 2017 (UTC)
For tidiness I suggest we put all events that are sequential in one cell, similar to the 2014 page. I believe the events should (mostly) stay there, because if somebody reads this page in future and sees a shift in the polls, it's good for them to have a quick idea as to what may have influenced that change. For example, TOP were polling 2.0-3.5% since late July, but the 30th August poll brings them back down to 1.0 and it's very likely that the misogyny perceived in his ill-thought tweet influenced this (though it's only 1 poll at this point, future polls will see if it sticks). Although the amount of events are getting a bit over the top, it's unfortunately reflective of the election. So I would suggest it's best to keep events that majorly effect polls, but remove the rest - for example the appointment of Damien Light can be removed. Arguably things like writ-day could also be removed - however I'm in two minds about this one being that it gives a good feel as to where in the political cycle we are. — Preceding unsigned comment added by (talkcontribs) 1 September 2017, 21:35 (UTC)
I concur to merge adjacent cells into one as per the 2014 page. I agree that the Damien Light appointment is irrelevant; then again, once adjacent items are merged into one cell, Dunne's resignation and Light's appointment could possibly be mentioned in the same sentence even. I'm ambivalent about writ day. The rest would all seem to be quite relevant. It's fair to say that this election is somewhat more interesting than the last one; there's heaps going on! Schwede66 18:27, 1 September 2017 (UTC)
I also agree to merging cells. Kiwichris (talk) 00:24, 2 September 2017 (UTC)

I've tried to condense a bit further - there was already one case where "x days later" was written rather than have two separate rows with two separate dates, so I used that a few more times, in situations where there were no polls in between. I also got rid of a few details, operating under the principle of "What would future election observers need to know when trying to explain polling changes?" Hopefully these work? Thanks, Chuborno (talk) 16:45, 2 September 2017 (UTC)

As a test I've merged news about green/labour MoU + 2016 budget into one cell. Please let me know what you think... (talk) 21:32, 2 September 2017 (UTC)

Table Formatting[edit]

Not exactly every time the table format is tweaked, but at least sometimes, it breaks the R script that scrapes the the numbers to create the graphs. I'm not against the improvements, but getting frustrated at having to trouble shoot the errors and try to understand the new format and then adapt. It has changed a 2 minute job to make a new graph into a 30+ minute job. For example, even just the bolding of the leading party will now require an update. And I think that time would be better invested in adjusting the fit line. There are three solutions I can see

  1. Revert the table to a previously working format.
  2. Work on a new table format in a sandbox, and then develop a new R code to scrape that format.
  3. Remove the graphs.

Sorry for the tone, it's just a bit frustrating. --Limegreen (talk) 16:56, 1 September 2017 (UTC)

I can see why that's rather frustrating. So here's a big thank you to you for producing the graph in the first place; I reckon it's one of the most useful pieces of information for New Zealanders with political interest that comes from Wikipedia. To that end, your third option isn't a real option. With regards to the first two options, I don't really mind. The bolding of the leading party was, until rather recently, somewhat superfluous. Recent events (i.e. last opinion poll) changed that, of course. To that end, I don't mind how the table is formatted. But whatever we decide, it has to work for you, Limegreen, as the graph is much more meaningful than the matrix of numbers. Schwede66 18:12, 1 September 2017 (UTC)
I concur with Schwede, and a very big thank you indeed - while I'm new here I see that you've been doing the graphs a while (i.e. 2014 page) and it is really neat to get a quick idea of polls, etc. May I ask if it is just the ' that causes a problem? Because we could just use the < b > tag :-) (talk) 21:04, 1 September 2017 (UTC)
It's not my code originally, but I have been running it (and adapting it) for about 3 years now (I think). It downloads a copy of the html, so it's is a matter of stripping off the < b > tags, and any other tags that are there. It also originally sourced the party colours dynamically, but with changes to the column headers, it was easier to manually assign them rather than pull them off the table. I've been travelling, so I just haven't had time to update it. I'll try and get the graphs updated shortly, but if we can just then keep the formatting the same up until the election would be good from my perspective! Limegreen (talk) 05:39, 4 September 2017 (UTC)
Sounds good - can you give examples of any changes we've made that have caused issues? Can we still use ' ' ' or change to < b > ? (talk) 09:19, 4 September 2017 (UTC)

OK, I know it's antipodean and many things are reversed down there...[edit]

...but wouldn't it be more sensible for the latest poll results seen to be those at the top of the table and not the bottom? It would save an awful lot of mindless scrolling. Like it is for Opinion polling for the United Kingdom general election, 2017 Boscaswell talk 20:32, 5 September 2017 (UTC) Apologies for the dumb error message. No problem as I see it with a wikilink and no other reference.Boscaswell talk 20:41, 5 September 2017 (UTC)

I'm ok with that in principle but only if Limegreen concurs. Schwede66 22:58, 5 September 2017 (UTC)
I'm happy for that to be the case in the followup article, Opinion polling for the next New Zealand general election, as there will be no urgency for any rework required for the scripts to produce the graphs. Similarly, this article and all earlier versions can be reversed once the final graphs are produced, shortly after the final results of the 2017 election are posted. If Limegreen is happy for this article to be reversed earlier because the script changes are trivial, or they have copious free time to make the changes, then I'm good with that.-gadfium 02:16, 6 September 2017 (UTC)
The UK page is much bigger, with the chart for 2017 alone being bigger than the entire NZ chart to date since the 2014 election. Because it's so much bigger, the UK page is virtually obliged to have the 2017 chart at the top to avoid a lot of downscrolling, which mandates most-recent poll at the top to preserve the backward timeline sense. The NZ article is a lot smaller, unlikely to get much bigger with only 2​12 weeks to go, and I don't think it's any great hardship to scroll down for the latest polls. Makes more sense for the timeline to run down the page and no hardship. Opposed. Akld guy (talk) 04:12, 6 September 2017 (UTC)
Agree with Gadfium - there won't be sufficient polls to require a graph until mid-2018, so we can have a play around and agree on a format first without too much impact. Lcmortensen (mailbox) 06:44, 7 September 2017 (UTC)
The graph actually uses the dates, so in theory, there should be no issue with changing the order. Limegreen (talk) 00:17, 12 September 2017 (UTC)

Newsroom-SSI poll[edit]

But it is a random sample from SSI's nationwide panel and reflects the make-up of the New Zealand population on age, gender, area and ethnicity.

From the Newsroom article. It's cited as a poll that includes all demographics, see Guardian article here. Larger discussion should whether be this article focuses solely on the polling from the main five or includes polling from Newsroom-SSI, Listener-Bauer, etc. Justinhu12 (talk) 05:37, 8 September 2017 (UTC)

Polling Graphical Summary[edit]

As noted in the edit log, I've just inserted temporary graphs (polynomial/ linear regression), awaiting to be replaced by User:Limegreen when File:NZ opinion polls 2014-2017-majorparties.png and File:NZ opinion polls 2014-2017-minorparties.png are updated. - Sleepingstar (talk) 23:18, 8 September 2017 (UTC)

@Sleepingstar: The image you made looks pretty decent and much more clear. I would prefer to use this graph in the coming days (or years) imo than the previous one. Typhoon2013 (talk) 23:53, 8 September 2017 (UTC)
Yes, good work. Great to have the graphs updated. Akld guy (talk) 00:15, 9 September 2017 (UTC)
Thank you both for the kind words. Unfortunately I am travelling over the next two weeks and won't be returning home until election day... so would probably prefer User:Limegreen to update the graphs in the mean time. Am happy to step in if the graph desperately needs updating within the next two weeks, but cannot make any promises or guarentees as this will all be dependent on my internet availability while travelling. - Sleepingstar (talk) 00:42, 9 September 2017 (UTC)
@Sleepingstar: Oh that sucks. I just want to know what app do you use for the graph? I'll see because I am happy to help and start working within the project! Typhoon2013 (talk) 05:25, 9 September 2017 (UTC)
@Typhoon2013: GraphPad Prism. Although user-friendly for regression, it comes with the cost of having to manually enter data points. - Sleepingstar (talk) 06:02, 9 September 2017 (UTC)
@Sleepingstar: I may be dumb, but what graph did you use, or what is it called? Any how did you get the colours? Data is fine for me though. Typhoon2013 (talk) 08:38, 9 September 2017 (UTC)
@Typhoon2013: There's a wide variety of methods that different people use for polling data, e.g. with LOWESS, GAM, moving average, and non-linear regression. GraphPad doesn't appear to have GAM. The LOWESS is availble but its predetermined settings are hard to alter and the ouput is too grainy for my liking , and moving average ceates too much fluctuation too e.g. File:Opinion polling UK 2020 election short axis.png. So I ended up going with a mixture of linear/polyonmial regression (under XY graphs), trying to keep it simple, but opting out of linear regression when the fits were awful (i.e. R squared values beyond rubbish). Sleepingstar (talk) 10:02, 9 September 2017 (UTC)
@Sleepingstar: Ok so my graph is now similar to yours. Only thing is that I don't have the numbers on the right-hand axis. But anyhow, I am now ready to update the graph with our coming polls and now happy to help. Tbh I would rather use this graph than Limegreen's graph because this one shows the curve more, if that's ok for you and for anyone. Typhoon2013 (talk) 10:55, 9 September 2017 (UTC)
I was planning to change the fit line to a curvy one, but all my time has been spent adjusting my code to the changes in the table format. Just to be clear, older versions of the graph used a LOESS smoother, which was much curvier (e.g., [[2]]), but there was debate around that time that minor adjustments to the smoothing function were leading to changed interpretations, sometimes fitting noise rather than signal, which is why we ended up with the very flat GAM smooth. Though it is clear in the current situation that the GAM is missing a very obvious signal. Limegreen (talk) 23:11, 10 September 2017 (UTC)
OK, I managed to get the R code working again (because when it's working, its incredibly quick). I've switched back to Loess, and picked a coefficient of 0.35. The 0.5 from previous elections really wasn't capturing the present radical change in poll scores. The R code to make the graph is on the description page here [3], if you'd like to run it yourself. I'm not precious about my ownership of this; I've just been doing it because I know R, and the previous editors were not available. Limegreen (talk) 01:36, 11 September 2017 (UTC)
Are the grey shaded areas in the minor party graph meant to look like that? Looks to me like it got confused. User:clesam11 (talk) 14:23, 11 September 2017 (UTC)
@Limegreen: Yay at least you're back updating. I want to ask two questions: 1) are you using the same app (which is GraphPad Prism)? and 2) is the UMR polls added in the graphs? Thanks and again, I just love to help update. Typhoon2013 (talk) 05:37, 11 September 2017 (UTC)
Ok so no reply from @Limegreen: which is concerning. But @Clesam11: and @Sleepingstar:, should we add the UMR polls in the graph? Like the (temporary) graph we discussed and updated when Limegreen was not here. Imo we should because 1) it is a poll and 2) it is better with more dots as we see a curve within the graph. Typhoon2013 (talk) 10:56, 11 September 2017 (UTC)
@Typhoon2013: Adding the UMR polls to the graph is fine with me. Gotta say it's a bit of a strain on the eyes seeing the minor party graph looking messy on account of the grey areas that represent the margin of error not working properly (may have resulted in a conflict from changing from GAM to LOESS @Limegreen:?). I was actually quite fond of the quality of the GraphPad Prism graphs, especially @Sleepingstar:'s one which included a figure for margin of error and would happily welcome them back. Anyway hopefully that conflict with the minor party graph can be resolved. User:clesam11 (talk) 23:30, 11 September 2017 (UTC)
@Clesam11: Yeah I know. I just need to ask @Sleepingstar: again how he did the margin of error, which he hasn't replied yet.Typhoon2013 (talk) 23:08, 11 September 2017 (UTC)
@Typhoon2013: I think Sleepingstar said he was going to be away on holiday until the election, and Limegreen isn't always active which makes it a bit difficult. User:clesam11 (talk) 12:02, 12 September 2017 (UTC)
The graph I make is in R. The full code to make it yourself is on the commons page for the graph. [4]. If there are any particular aesthetic changes (eg the Prism colours are more intense), things are possible.Limegreen (talk) 00:43, 12 September 2017 (UTC)
@Typhoon2013: I didn't include the UMR poll, as the consensus has been that this poll, along with the Curia one is biased. I'm also a little weary of the two new panel-based polls, especially because one of them has an *extremely* surprising result.
Also, I've just been trying to work out what is going on with the minor parties graph. It's a problem with the margin of error for the conservatives, where the polygon is drawing in the wrong shape, but not immediately clear why. Limegreen (talk) 00:45, 12 September 2017 (UTC)
@Limegreen: Yay! Glad to see that the minor party graph got fixed, thanks for that. I'm weary of the Newsroom-SSI Poll as well, their methodology seems like it would be unique to say the least. User:clesam11 (talk) 14:21, 12 September 2017 (UTC)
The Listener - Bauer Media Insight poll uses a similar strategy. They're not actually as weird as they seem, but in this case, they have not had a chance to calibrate their polling method on an election. In the run up to the 2011 election, there was a regular panel poll (some discussion here). Some pollsters even use a single panel over time, which means you can actually detect people changing their mind, as opposed to guessing whether people are (eg) switching from National to Labour or National to NZ First. Limegreen (talk) 02:37, 12 September 2017 (UTC)

Horizon are actually back polling [5] Limegreen (talk) 02:39, 12 September 2017 (UTC)

@Limegreen: Thank you so much for all your hard work maintaining the graphs, and also having to re-create the code and adjust it to fit with all the highlighting and bolding (which I am so sincerely sorry for putting in -- at the time I thought it was a good idea because it felt like Labour could surge into the highest polling party, and the formatting was to correspond with the US polling pages... )
@Typhoon2013 and Clesam11: Thanks for your interest in the Prism graphs, but to be honest, I feel that Limegreen's graphs are superior in a few respects. Firstly I never figured out how to plot the margin of confidence on the line graph in Graphpad Prism, which the R script handles rather well. On a side note, as per Typhoon2013's question, you can go back to the coefficient output in pRism and look at their CIs and then back-calculate the polynomial equation to figure out the margin of confidence. Nevertheless, this is tedious, and much of Graphpad Prism requires manual input and manual calculations, it's probably best to stick with the "R" script which Limegreen has already done all the donkey work for, which does all the calculations automatically. You can always feel free to modify the "R" script so the output is .SVG instead of .PNG, and the colour scheme etc, but in my own opinion I think the R script LOESS regression is a much more elegant approach.
- Sleepingstar (talk) 06:25, 13 September 2017 (UTC)
I can't take too much credit for this. Most of the R code is unchanged from the 2005-2008 electoral cycle!!! And the reason why it takes me a bit of time to troubleshoot is that I didn't write the code, so it is always harder to understand other people's code. It's a remarkably clever piece of work, because it pulls quite a lot of material directly from the table. Just at some point, there wasn't another editor around who used R, so I stepped up. Limegreen (talk) 09:13, 13 September 2017 (UTC)

Stuff polls[edit]

I don't know why we don't include the Stuff polls]. It's still a poll and from a site lot of people use, right? And I don't see a problem with it, so I'm going to add it in. Typhoon2013 (talk) 11:09, 9 September 2017 (UTC)

The reason we don't include them is because Stuff isn't conducting polls, they're just taking other polls (like Reid Research, Colmar Brunton, etc.) to show the average poll results at a given date. It's similar to the graphical summaries we have on this article. User:Clesam11 (talk) 11:42, 9 September 2017 (UTC)
Ok thanks for clarifying. Typhoon2013 (talk) 11:51, 9 September 2017 (UTC)
Please take more care when finding polls to load, as I've also just deleted content from you where the source doesn't state what year it was... google news didn't find it ("IPSOS" poll -> sort by date). Also, I've just now looked at Opinion polling for the New Zealand general election, 2014 which has an identical poll dated 2012. (talk) 19:39, 9 September 2017 (UTC)

Preferred Prime Minister[edit]

So I figured it would be good to see data from: (a) Leaders of minor parties (i.e. Maori); (b) Every deputy leader (i.e. Paula Bennett) to see if hardcore followers (who won't vote for other parties) think the leader is doing a bad job (e.g. like Jacinda/Bill pre-little/key standing down). However, I can see the PPM table is getting quite large and hence someone decided to delete that - fair enough... but maybe we should also be deleting every leader that polls 3% and below (i.e. James Shaw, Hone Hawawira, Gareth Morgan, Peter Dunne, Meteria Turei)? Also note that the Opinion polling for the New Zealand general election, 2014 has a similar setup of minor parties being deleted. (talk) 19:55, 9 September 2017 (UTC)

I'm getting a little annoyed at the presumption that because we know that a candidate or party is polling very low, we should deny the reader the right to know that. Suppose a reader wants to see whether there's a trend developing for a minor candidate or party, that might result in the creation of a powerful coalition partner. Should we deny them that info? You haven't made clear whether minor results in the 2014 polling were weeded out during the campaign or after that election. Akld guy (talk) 21:01, 9 September 2017 (UTC)
The concept of only including high polling leaders isn't exactly controversial, we're actually quite lenient in the sense of including people that are polling -1% (Gareth Morgan, Peter Dunne, etc.) Take a look at the UK opinion polling or the Australian polls, they only include the incumbent prime minister and the opposition leader. I understand the idea that a minor party leaders can become popular, which is why I think a threshold of about 3% is completely reasonable. If a note above the table was included that specifies that only leaders that have polled above 3% or so are included then there can be no confusion over other minor leaders that may be polling (below 3%) as well. User:clesam11 (talk) 11:33, 10 September 2017 (UTC)
The UK uses the First past the post (FPP) system and Australia uses the Single transferable vote (STV) system, neither of which guarantees the minor parties much power by way of coalition, because the winning party can usually govern alone. In New Zealand's Mixed member proportional (MMP) system, coalition arrangements are virtually guaranteed, and have indeed been the case in every election since MMP was introduced in 1996. Therefore NZ's MMP system makes the minor parties potentially far more powerful. Akld guy (talk) 00:09, 10 September 2017 (UTC)
That is technically true, but if less that 3% of New Zealanders supported someone becoming Prime Minister there'd be virtually no circumstance that person would be made Prime Minister, even with some sort of deal. This is mainly because it would likely destroy any chance of re-election. This page is primarily about public opinion anyway, not a deal that could be struck in a coalition. I still think it would be fair to put a 3% threshold rule or something similar in place, if for any reason, to keep the preferred pm table manageable and tidy. User:clesam11 (talk) 12:48, 10 September 2017 (UTC)
To be clear I think it's good to put all party results. But putting all PPM results gets a bit ridiculous (e.g. Hone Hawawira doesn't even get polled on non-Colmar Brunton polls) and I agree with all of clesam's points. If we put all the options, it would look similar to this Colmar Brunton poll which has 27 options listed). I put this to the talk page simply to bring forth a discussion (not a presumption of rights, etc.) to ask about what results we should be putting there before things start looking silly (talk) 04:45, 10 September 2017 (UTC)

Dashes and zeros[edit]

I would like to know why ip changed dashes to zeros in this edit of Roy Morgan data and then changed them back in this edit nine or ten days later. The zeros are correct and appear in the published Roy Morgan poll results. Akld guy (talk) 21:17, 9 September 2017 (UTC)

Really a dash or a zero implies that they didn't get any support in the poll, so i'm not really bothered either way. If you'd like to change them to zeros I doubt anyone would complain. User:clesam11 (talk) 11:37, 10 September 2017 (UTC)
They mean quite different things. A dash means that the pollster (Roy Morgan) didn't include the party or candidate in the question, whereas zero means they were included and scored 0. I don't want to change them to zeros because I have already restored them after the ip's edit above. What I want him to answer is why he changed them on 30 August and then changed them back again today. It looks like that editor may be having a little fun. Akld guy (talk) 23:47, 9 September 2017 (UTC)
Problem I see is that zero means the pollster definitely asked about the party and reported it in the results. Where-as dash means it wasn't reported in the poll which means one of two likelihoods (1) The pollster never asked (or allowed for) that party in their poll, or (b) the party polled <0.1% - and this is where putting them all as dashes (rather than some as zeros) would allow for consistency in the numbers. Feel free to argue the point if I'm wrong - I don't really care, I just thought it was cleaner (talk) 04:36, 10 September 2017 (UTC)
We are in agreement about what the dashes and zeros ought to mean, up to where you talk about <0.1%. In the case of Roy Morgan poll results, the cited references don't omit any of the minor parties we report here, and only ever report them as 0's, so I don't know where this <0.1% comes from. I don't see that in the references. So we should use what Roy Morgan reports, which is 0, not dash. In my opinion:
  • dash should be used where the pollster does not report a figure
  • 0 should be used when reported as such
  • N/A should be used only where the party or candidate was not eligible or not in existence at the time the poll was taken; ie. it should mean Not Applicable, not Not Available, which should be reported as a dash. Akld guy (talk) 06:32, 10 September 2017 (UTC)

@ Your latest edit, which I reverted and which you promptly restored, changed Roy Morgan poll results that were 0 to 0.0 and also changed several other whole numbers, eg 5 to 5.0. That is a WP:OR violation since the references do not state the figures to the first decimal point. It's likely that Roy Morgan has rounded those figures to the nearest whole number, so you have no right to attempt to make them more precise than what the company reported. Akld guy (talk) 21:11, 10 September 2017 (UTC)

For say Roy Morgan, they appear to be rounding to the nearest 0.5 rather than whole number and hence it makes sense to me to add .0's to the whole numbers for consistency. If you disagree, I'll simply restore the zeros without the .0's - again, I don't care (talk) 21:25, 10 September 2017 (UTC)
They "appear to be"? It's not your call. We go by what the references say. If Roy Morgan reports 0 or 5, we report it that way. Akld guy (talk) 21:31, 10 September 2017 (UTC)
"appear to be", as in roy morgan reports EVERYTHING as "X.XX" (talk) 21:33, 10 September 2017 (UTC)
Citation for that please. That's not the way they are cited in the references. Akld guy (talk) 21:37, 10 September 2017 (UTC)
See graphs (talk) 22:10, 10 September 2017 (UTC)

@ If we get the impression that your contributions are more disruptive than useful, you may face a block. Hence the suggestion to get agreement on the talk page, rather than simply changing details in the article that are not supported in citations. Schwede66 21:55, 10 September 2017 (UTC)

Not to mention this insult. Akld guy (talk) 22:15, 10 September 2017 (UTC)
The careless way my contributions are undone I would say is disruptive and I don't have to put up with it (e.g. Mana party polled 1.42 in the cited source, not 1.40 - and this was reversed... other parties were polled in the cited source so I put zero since it was agreed above this is how they should display, there was an argument about decimals - but still fully reversed). I don't need to put up your threats either. So whatever... I cared enough about what is going on in my political world to look at the data, I felt an obligation to contribute to the wiki as an enjoyer of the data... but I don't need to deal with this bullsh%t so no need to block me, I won't be back. (talk) 22:10, 10 September 2017 (UTC)


All editors - please watch carefully the edits by User:FiliKiwi. This user lied in this edit. Akld guy (talk) 23:36, 15 September 2017 (UTC)

@Akld guy: I actually think that edit was good faith; he was rounding and was confused by the previous edits done. J947(c) (m) 00:08, 16 September 2017 (UTC)
@Akld guy: Calling a contributor a liar is a bit harsh. I've given a caution about introducing factual errors ({{uw-error2}}) on their talk page . Lcmortensen (mailbox) 00:24, 16 September 2017 (UTC)
If it was a mistake, why didn't they pick it up before hitting the "Save changes" button? Does this user use the "Show preview" button? Quite a few of their edits have been reverted. Akld guy (talk) 02:48, 16 September 2017 (UTC)
@Akld guy:, I do not appreciate or condone your comment in any way. I had done the edit based on previous edits done so on the page; Roy Morgan poll's and others had been previously rounded up to reflect the statements in the article that the results were 'virtual ties'. They were done so in good faith and to state that I am a liar is rather rude. — Preceding unsigned comment added by FiliKiwi (talkcontribs) 03:23, 16 September 2017 (UTC)
@FiliKiwi: So you admit your changes were not a mistake. You changed one party's result from 39.5 to 40 (sic). You changed another set of results from 38.5 and 38.2 to both being 38 (sic). All of those changes were contrary to what the pollster's published results showed and favoured one party over another. You had no right to change published poll results on the basis that you were rounding up and I'm calling you a liar and sticking by it. There's a little bit of politicking going on over these results (see section immediately above) and I'm determined to stop it. Akld guy (talk) 04:14, 16 September 2017 (UTC)
If we are having issues with multiple unregistered (IP) and new users, then maybe we should consider protection until 24 September. Lcmortensen (mailbox) 07:07, 16 September 2017 (UTC)
I have semi-protected the article until 24 September, requiring extended autoconfirmed status. And after successfully protecting the talk page, I have now also protected the article itself. D'oh! Schwede66 03:43, 19 September 2017 (UTC)
@Schwede66: Was that really necessary? There's only been about 4 edits in the past 3 days. This seems like a bit of an overreaction to a handful of disruptive edits. User:Clesam11 (talk) 16:02, 19 September 2017 (UTC)
Oh hold on there's been more than that. My mistake. Maybe i'm wrong, but it still seems like an overreaction. User:Clesam11 (talk) 16:04, 19 September 2017 (UTC)
No, I've been concerned about this article for the last six weeks or so. There's been a lot of toing and froing, and it's getting too disruptive in my opinion. The protection expires on Sunday. So should be all good. Is that ok with you? Schwede66 07:01, 19 September 2017 (UTC)

Horizon polling[edit]

Horizon may be back polling, but as they are still polling a self-selected group they're still not meeting the requirements of these electoral polling articles. Because of this Horizon polls have been excluded in previous election articles, and should be excluded here also. I'll hold off editing the entries for a day so Schwede and Limegreen (et al) can comment. FanRed XN | talk | 07:58, 19 September 2017 (UTC)

Yes, Limegreen pointed out in a discussion above that Horizon is back at it. For the 2011 election, we decided to exclude Horizon. The company has the same methodological problem of self-selection, and they encourage this behaviour by giving people prizes when they sign up. Unless I'm missing something (e.g. the big polling firms employing the same recruitment methods), Horizon should thus be removed. I shall do so now, but if anybody can provide further background, please discuss. If we reach consensus that all is good with Horizon, we can add their results back in. Schwede66 19:00, 19 September 2017 (UTC)
I'm reasonably happy for Horizon to be excluded, based on the previous precedent. Having said that, I don't think it is correct to describe the Horizon poll as self-selected. Normally when people say a poll is self-selected, people are opting in for one very specific poll (eg the recent poll on whether landlords would raise rents under Labour, with around 800 responses out of 120,000 landlords). In contrast, Horizon run a panel, where a large number of people (likely greater than 10,000) sign up, and then a random sample of those people (probably with quota to match NZ voter demographics) responds to the specific survey. Though there is some bias introduced by those who sign up to the panel, the sampling from the panel should produce a much better outcome. Colmar Brunton have a 25% panel component in their sample, as an attempt to limit bias from using only landlines.
Further, the Newsroom SSI poll is also an online panel poll, and I'm pretty sure that the Listener Bauer Media Insights is as well, so I think we should be consistent. Limegreen (talk) 21:38, 19 September 2017 (UTC)

"The Newsroom-SSI poll has a margin of error of 4.4 percent. It is from a smaller sample than the television networks' polls and does not involve phone interviews. But it is a random sample from SSI's nationwide panel and reflects the make-up of the New Zealand population on age, gender, area and ethnicity. Those polled do not opt in, but are invited to participate in accordance with SSI's required demographic balance." source=

^ quote on the SSI poll Limegreen (talk) 21:39, 19 September 2017 (UTC)

Forecast section needs regular refreshing - can someone with editing privileges please do it now & regularly[edit]

@Schwede66: @Lcmortensen: @Sleepingstar: Hi - when Template:2017_NZ_election_forecasts is updated, this page doesn't automatically display the updates. (For example right now you can see that the Herald forecast hasn't been updated yet on this page although it's been updated on the template.) It only updates when this page is edited and then saved. What I have been doing is editing that section, changing nothing, and saving, with the edit summary "refreshing template" and "minor edit" ticked. Can someone with editing privileges please do that now? And can you please make a point of checking the template regularly (e.g. there will be a Newshub poll tonight and subsequent edits to the Radio NZ and Herald forecasts) and refreshing it here whenever it needs refreshing? Elcalebo (talk) 20:22, 20 September 2017 (UTC)

Not a real problem. It's a cache issue – you'll have to purge your cache for it to display the most recent version (append ?action=purge to the end of the page's url or, if using Firefox, ctrl-shift-R). Mélencron (talk) 20:30, 20 September 2017 (UTC)
It's also an issue with Wiki's cache. See the "I edited the template, so why didn't the page it is used on change?" question here: @Mélencron: @Schwede66: @Lcmortensen: @Sleepingstar: can one of you please refresh this morning for the final projections? Elcalebo (talk) 19:31, 21 September 2017 (UTC)
Caching is done by a local computer. Schwede66 20:48, 21 September 2017 (UTC)
@Elcalebo: {{ping|Schwede66}: There is a cache on the Wikipedia server end too. To empty it, append ?action=purge to the end of the URL (i.e.,_2017?action=purge), then click OK on the prompt that appears. This will purge the cache at the Wikipedia end. Lcmortensen (mailbox) 22:00, 21 September 2017 (UTC)
Thanks! That's working for me. Elcalebo (talk) 05:48, 22 September 2017 (UTC)

Extended-confirmed-protected edit request on 21 September 2017[edit]

Newshub reid research results 13-20 sept 2017. Results for United 0.0, Conservatives 0.6, and Mana 0.1 should be included. ref Apolonia 2 (talk) 10:32, 21 September 2017 (UTC) Newshub Reid Research[106] 13–20 Sep 2017 45.8 37.3 7.1 7.1 0.4 0.6 – – – 0.9

Newshub Reid Research[106] 13–20 Sep 2017 45.8 37.3 7.1 7.1 0.4 0.6 0.0 0.6 0.1 0.9

The link does not contain any results. This is probably because the data was removed because it's illegal to display material influencing voting in NZ today. NZ editors should be careful updating political articles today.-gadfium 22:38, 22 September 2017 (UTC)

Graph tweak[edit]

There's historically been a lot of debate here about how sensitive to make the smoothing on the graph, but I've just made it much more sensitive, as the trend line has not really captured recent volatility. Can revert if that seems unpopular. (And even within that, it's really hard to tell just how sensitive to make it). Limegreen (talk) 21:16, 21 September 2017 (UTC)
Cite error: There are <ref group=nb> tags on this page, but the references will not show without a {{reflist|group=nb}} template (see the help page).