Lets Get Statistical!

I had half of this in the works when the latest results came out so it was a simple case of plugging them into what I was already working on. Apologies for the wonky layout on the stats, I tried, I really did.

Is anyone else slightly amazed at the astounding 10% in Nationals polling from 43% to 53% in one month via Roy Morgan?

Personally I am calling BS on this one right now.

I admit that I might be slightly bias in my opinion of National (see my previous posts where I have referred to them as criminal scumbags et al) but I don’t think even my bias would blind me to the fact that in our current clime of political ineptitude (housing crisis, diary failure, immigration concerns, housing crisis, possible trade wars, water concerns, housing crisis and the repellent and nauseating image of either Clinton or Trump in the Whitehouse) the mood of the nation would suddenly shift 250,000 people to the political right in the space of one month!

And the idea that 10% of the electorate just suddenly jumping to the right seems even more dubious when you look at these numbers:

NAT              53%             (+10%)

LAB              25%             (-2.5%)

GRN              11.5%            (-3.0%)

NZF              7.0%             (-2.0%)

MAR              0.5%              (1.5%)

UNF              0.0%             (Nc)

  ACT              1.0%              (+0.5%)

MAN              0.5%              (-0.5%)

CON              0.5%              (-0.5%)

What stands out is not just the stupendous surge in popularity with National but the large losses to the Greens, Labour and New Zealand First. I could imagine that some NZ First voters might jump ship (being relative neighbors on the political spectrum) but -3% from the Greens?

But Roy Morgan is just one of the three main thermometers (the other two being Colmar Brunton (TV1) and Reid Research (TV3)) taking the rectal temperature of our nation (sorry Fairfax and Digipol you don’t count).

So let’s compare the polling figures for all three from the last month before quarter of a million kiwis decided that National is the way to go (thanks to Curia Market Research for their handy blog which provides and updated blog on all three).

 

Colmar(TV1)         Roy                     Reid(TV3)

NAT       48% (-2%)            43% (-2.5)           47% (+0.3%)

LAB        29% (+1%)           28% (-1.5%)       31.3% (-1.0%)

GRN       12% (+2%)           14.5% (+2.5%)    11.1% (+0.9%)

NZF        9% (nc)                9.0% (+1%)          7.8% (+0.3%)

MAR        0.7% (-0.4%)        1.0% (+1%)          1% (-0.3%)

UNF        0.0% (nc)              0.0% (nc)              0.0 (nc)

ACT         0.3% (-0.4%)        0.5% (+0.5%)       0.4% (-0.4%)

MAN       0.0% (nc)              1% (+1%)               0.0% (nc)

CON        0.7% (+0.4%)        1.0 (+0.5%)           0.0% (-0.7%)

 

METHOD            Rnd Phone      Rnd Phone     Rnd Phone

SAMPLE              1509/1245     868/820        1000

UNDECIDED       15%           5.5%           Unknown

SUBSCRIBE         Yes           No             Yes

MARGIN              +/- 2.5%      n/a            +/- 1.9%

 

What one gets from last month’s polling is that while National was polling higher for the Colmar Brunton and Reid polls (%48 and 47% respectively) the Roy Morgan poll was down at 43% just 30 days ago the movement of National from previous months was down on two out of the three and the third (+0.3%) was well within the margin of error.

Further the average for National from all three polls was just 45%.

And Roy Morgan’s own data from previous months shows National in gentle decline from its previous high of 50% in April last year

This shows that National was either holding steady or declining under the ongoing pressure of current events and its own limp reactions.

David Farrer, who runs Curia, posts about this on KiwiBlog where he breaks the numbers down a bit more and concludes that while probably not a 10% jump the rise, is probably genuine.

Now I know better than to argue stats with anyone interested in stats, as being a stat freak myself (military stats rather than the more usual Kiwi field of sports stats), but they are only as good as their method of collection and the method of processing, and with all due respect to David (and his statistical probabilities of the data being correct), I am just not convinced that National has increased at all.

As David himself notes on Kiwiblog, it’s been a month of “relentless negativity” for the government after previous months of doom and gloom also. So where is the positive direction coming from? Where is the love for Key and his scaly minions hailing? It can’t be the media, the Reserve Bank or the general public.

And what statements or announcements from the Lizard King himself or any announced policy (Nationals weak willed attempt at dealing with the housing crisis?) could be driving this? Where is the momentum for 10% of those being polled to shift to National at the expense of all other parties?

Farrer again has his own take on this calling Roy Morgan a “yo-yo poll” which sounds like a polite way for statisticians to put each other’s work down.

But before we dive into that lets have a bit more of a look at political polling in Godzone.

Firstly two of the big three polls have to confirm to the New Zealand Political Polling Code by being members of the Research Association of New Zealand (RANZ). Guess which poll is not a member? Hint it’s not Colmar or Reid. This is probably because Roy Morgan is based in Australia.

The code is reasonably robust with prescriptions for conducting, reporting and publishing the data covering the sampling, the collection method, the weighting, the margin of error and results. It does have a few grey areas like excluding unlikely voters from the sample but in general is sound and if followed should lead to consistent and accurate results and transparent reporting.

Second have a look at the sample size, Morgans is by far the smallest at 820. Now I know that for polling you don’t need to poll all people to get a representative sample (usually above 1000 is considered acceptable) but I do know that polling at such small levels can magnify small shifts in the data (my own undergrad study in Pol Sci was relentlessly American in technique, which as anyone taught under that system knows is very heavy on data collection and analysis over theory or analysis).

Then there is the margin of error (MOE) and the undecided portion of the polling. Roy Morgan does not have a margin of error that I could find but did have a 5.5% rate of undecided voters. This is not as high as Colmar’s 15% for +/-2.5% margin from a sample of over 1500 people but in a poll of just over 860 people a nearly 50 person hole in the data is problematic to say the least. Also Reids data does not even include the number of people undecided so we only know that it was less than 1000 listed.

Now I’m not linking these two inextricably but in such polls the MOE and percentage of those undecided are key measures for how reliable your data is and not being bound by the Polling Code or having a MOE leaves me concerned at this result, yo-yo poll or not.

Of course Colmar’s 15% hole in the data and Reid’s undeclared undecided are also problems but at least there is a margin of error to give some guarantee and I will be surprised if their new data shows such shifts.

With all three polls there are deeper issues with the data, one of which is the method of polling (calls to households with landlines).  Current data from Stats NZ has landlines in NZ at 85% which means that any house without a landline is automatically excluded.

The standard “wisdom” for this is that any household without a landline would be extremely low income and not likely to vote anyway. The issue is that I myself have a cell phone and an internet connection in my house but not a landline and many people I know don’t have one either being that mobile and internet can cover all the bases in modern life better than a landline can AND we are all politically active (ie we vote!). But that does not appear to register for the pollsters.

So the assumption that no landline equals no political participation is dubious at best and flawed at worst. I do acknowledge that the high rate of non participation in politics in NZ, which is reflected in only 76% of eligible voters voting in the last election, may have some correlation with economic well being and possibly not having a landline but as far as I know there is nothing to show exactly what those numbers are. Buts that’s an issue for another day.

So back to the question, where is National supposedly getting this 10% surge in votes from? Probably, as Farrer noted, there is likely some statistical error or readjustment (ie they were too low so the previous results were out of whack so this month’s result is more a readjustment than a surge in votes. That theory I can accept but I remain dubious of any increase in popularity for National at all given the current pressure they are under.

But apart from my grumbles about Key and Co there is a lot of other interesting data that can be taken from all three polls.

The first is that United Future is 0.0% across the board and with no change from previous polls. Add to this that Dunne currently holds less than a 2% majority in his one and only electorate and it’s not hard to imagine what’s going on in both Peter Dunne’s and the other parties minds as they consider the coming election and any electorates which might be up for Grabs.

Another is that the Maori Party, United Future and ACT are all one seat parties and all living well within the margin of error. Loose that seat and its goodbye baby. In the case of Act and the Maori Party both hold comfortable majorities (12% and over 20% respectively) but as noted above United Future is the straggling calf in the herd of political Wildebeest and the predators are circling.

Another interesting statistic is that if you add Mana and The Conservatives to the above three then five out of the nine parties in the poll are functioning non entities politically (ie no real representation in NZ). All live within the margin of error and all are political equivalent of the living dead if not actually dead (Mana and the Conservatives). The fact that all these zombie political parties, barring Mana, are or would assemble under the banner of National cements John Keys status as the Necromancer king of New Zealand politics.

But the most telling statistics of them all is that no matter how much National is up or down in the polls no combination of Labour and the Greens has enough to beat National at this time. They come close, ironically, in the Roy Morgan poll of last month, but nowhere can they actually get enough in the numbers to beat any poll result national has.

And if you’re thinking like I am thinking then you already know what those results are really saying, which has been said before by myself and many others, which is that the balance of power in all of this remains the MP for Northland, Winston Peters!

But with the current Roy Morgan results not even Winston can help the Greens and Labour but as I have been saying I do not believe the results to be that high and such polling always gets closer come election time as minds are made up and campaigning has an effect.

And why did Labour go down in the polls after announcing its own policy on housing which is streets ahead of Nationals own tepid response? There had been cautious indicators that the Labour/Greens MOU had helped build both brands and raise both in the polls but the current stats would have us believe that both have suffered for it and for actually proposing a solution to the housing circus.

So what has happened here? Did National pay someone at Roy Morgan to fudge the results? I would not put it past them but let’s assume no for the moment.

But the message, if echoed by the other two polls results (soon to be out), could have a chilling effect on any momentum the two parties have been building up the last months as they keep the pressure on the government through an ongoing barrage of criticism AND alternate solutions.

Political polling is the barometer of modern politics with its desperate reliance on unstable voter bases and shiftless ideologies. But as I use the barometer in my kitchen to give me an idea of the what the next 24 hours weather will be I also take those results with a grain of salt as the local and immediate reality can and does differ.

Most of the time we take political polls as gospel and never question their results, they are the life giving air that inflates or deflates party fortunes in western democracies far more than anywhere else in the world but they are, at the end of the day, just statistics and while a useful tool and just measure of past performance are not, in the quantum storm that is politics, always a good indicator of things to come.

I remain dubious of the 10% jump in preference as well as National having any uptake in the current round of political polling but I will have to respect the data if they all come in with the same conclusion. The question I would then ask is why? This is something that the stats and statistical data can never really answer.

5 thoughts on “Lets Get Statistical!

  1. Plenty of affluent, well educated voters don’t have land lines. This is the gap in the polling. Many of those polled, who ‘pick up’ may be the older Nat leaning demograph. Do they poll from the same list all the time or is it really randomized? Just sayin!! Barbara Matthews.

  2. Barbra: My thoughts exactly and I think it is a bit of a blind spot in the polling.

    That said I do get the mechanics of polling and the possible issues of getting a persons cell number to call but thats not an insurmountable problem.

    As for how random it is, the sites of the various pollsters do have info on how they actual do their polling and the Polling Code guidelines spell out clearly how reporting is done so while there might be some issues there I feel its really the sample selection (ie the phone issues) which may be the weak link.

    I am tempted to see if there is any data on people who don’t have landlines and who DO actually vote and on what possible political affiliation those who do have landlines is. Would be good fodder for this area.

  3. Good initiative, pretty hard to do. Then there’s caller ID where people won’t pick up if unknown caller unless they are terminally bored. Sorry for the negatives! Given all the pox on John Key and all his houses in various media I really can’t give much credit to polls of any description. Are they really indicative or just cheer leading the status quo?

  4. I feel that it can be a very fine line. The worm in 2002 which thrust Peter Dunne into the center (no pun intended) of NZ politics was possibly a good example of polling leading opinion by the nose.

    On the other hand the more prosaic ones which come out after the fact and tabulate monthly results are probably better measures.

    I, like you, do distrust the polls as media have shown themselves to be less honest arbiters or fact or truth and more cynical and manipulative media masseurs who present news as entertainment and entertainment as news.

    At least in the age of the internet we can get alternate sources of info.

    While caller ID would be an issue a more robust polling of the electorate is less an impossibility and more a technical challenge.

  5. I agree with you about the ‘massaging’ of info! Thanks for the feedback, helps to enlarge perspective.

Leave a Reply

Your email address will not be published. Required fields are marked *