Election 2014

Greying population and youth tremor sees middle age voter squeeze

It wasn't a youthquake, but a youth tremor and a greying population have seen New Zealanders aged 35 to 54 squeezed as a percentage of our voting population.

With youth enrolments down as a percentage of the overall potential voter population this election, a youthquake was always going to be difficult. However, better youth turnout numbers than in 2014 helped counteract the significant greying of voters in those three years.

The good news is that turnout was up across the board, ranging from 6.54 percentage points for those aged 18-24 down to a 0.6pp for those aged 70+. The bad news is that those aged 18-24 made up a lower percentage of overall enrolments than they did in 2014, meaning that their big increase in turnout didn't create a significant change from 2014.

Where the big change did occur was with voters aged 25-29, who increased their representation in the actual voter pool by 0.709pp versus 2014, the largest increase of any age group. Their increase of turnout of 5.45pp, combined with their larger slice of the enrolment pie,  and to a lesser extent those aged 30-34, delivered nearly all of the increase in youth voters this election.

The increase though was largely offset by an increase in voters aged 55 and over. With voters aged 70 and over representing 0.704pp more of voters than in 2014, and those aged 55-59 representing 0.240pp more, it largely balanced out the increases in the youth voter.

All of this means that those aged 35 to 54 saw their influence as voters decrease relative to 2014, which was partially driven by their contribution to the pool of enrolled voters also dropping relative to 2014. The biggest drop was in the 40-44 bracket where they contributed 1.026pp less to the final voter pool than they did in 2014.

You can find the Electoral Commission's numbers here.

Overall there was a marginal greying of voters, going from a median age of 47 in 2014, to 48 in 2017.

And in case you're curious, here's the raw numbers.

Bill and Winston almost owe Colin Craig a beer... almost

For snippet.png

When they're sitting down for coalition negotiations over the coming days, Bill English and Winston Peters should raise a glass in honour of Colin Craig, the man who saved both their parties this election.

Well, I'm exaggerating a bit there. But an analysis comparing preliminary results from both the 2014 and 2017 elections suggests a correlation between electorates where the Conservatives won a high share of the party vote and the electorates in 2017 which comparatively contributed more to the final overall party vote totals of National and New Zealand First this election.

The idea is simple. In 2014 approximately 95,000 New Zealanders voted for the Conservative Party led by Colin Craig. In the preliminary results on 20 September 2014, that figure was 86,616. In 2017's preliminary results that figure had dropped to just 5,318. On election night there were some 81,298 voters who had left the HMNZS Colin Craig in those three years, and I wanted to find them.

There were a few assumptions that were easy to make. I thought it was pretty unlikely many of them would have switched their votes to Labour, the Greens, or even The Opportunities Party given TOP's liberal stance on marijuana.

The issues that motivated voters to cast their ballots for the Conservatives were predominantly social issues that they had a conservative take on. That meant there were only three likely parties for those votes to go to.

Given ACT got so few votes, it was clear they hadn't all rushed there. Plus ACT under David Seymour emphasises a more libertarian ideology than say the ACT of John Banks or Richard Prebble which had a healthy does of social conservatism mixed in with its Rogernomics bent.

That left only National and New Zealand First as the two contenders for those votes. National under John Key was a fairly socially liberal party, but Bill English, as a Catholic, does have a socially conservative streak in him. Likewise New Zealand First and in particular Winston Peters, have always appealed to the protectionist, anti-outsider, harking back to the good-old-days mentality that is commonly associated with voters of a conservative nature. And, if overseas experience teaches us anything, those of a conservative nature aren't shy about voting, so I didn't expect them to stay away in large numbers.

What I wanted to see specifically though was whether this translated into my previous examination of the changing party vote share in each electorate between the two elections. Maybe I'd missed something. Could National's stronger than expected (and Labour's weaker than expected) results in Auckland be explained purely off Conservative voters bulking up the ranks of National and New Zealand First.

The answers surprised me. As you might imagine, there seems to be a pretty clear correlation that generally shows that in electorates where the Conservatives did well in 2014, then those electorates managed to contribute more to the overall party vote totals of National and New Zealand First than they did in 2014.

And bare in mind that this is off preliminary results from 2014 and 2017, so I'm trying to compare apples with apples as much is possible across two elections.

In case you can't expand the above image, the light blue line is the share of the party vote that the Conservatives received in each electorate in 2014. The dark blue and grey lines is the percentage change of how those electorates contributed to National and New Zealand First's overall party vote in 2017 compared to 2014. If the lines jut out to the right it means that electorate has contributed a greater share to the party's overall party vote than it did in 2014. If it's to the left, it contributed less overall. Where I'd earlier just looked at individual electorates in isolation, this allows us to compare electorates within the context of the overall party vote. So I'll dig more into this later.

As you can see in the above image (if you need a bigger version feel free to contact me) where the Conservatives got over 4.47% of the party vote in an electorate in 2014 (that's Waitaki in the above) there appears to have been enough support transferred to National and New Zealand First in 2017 to shore up their support.

As New Zealand First needed fewer of those Conservative votes in each electorate to prop up their party vote numbers, I have to wonder if there was a bit of a churn cycle going on as National sought to replace the support it was losing to Labour. This cycle would see Conservative voters largely going to New Zealand First and some to National, but New Zealand First shedding voters towards National, especially in the large rural or provincial city electorates where National made a big push during the campaign.

Where the Conservatives got 4.47% they generally don't seem to have had enough votes to protect National or New Zealand First. Notable at the bottom of the list are Wellington Central, Auckland Central, and Mt Albert, all places that were in the top 10 locations that Labour grew its party vote relative to 2014.

Where things start to get interesting too is that this highlights the importance of National's performance in West and South Auckland outside of just collecting former Conservative and New Zealand First voters. Looking at the top 10 electorates where National grew it's share of the party vote relative to 2014, only Botany and Pakuranga sit above that tipping point where the number of Conservative voters was high enough to protect National and New Zealand First.

What's also worth noting about all of those electorates is that they have some of the worst voter turnout rates - if not the worst - in the country. It means that not only did National manage to capitalise on Labour's weaknesses with ethnic communities and potentially the mortgage belt in Auckland, but also Labour simply was not able to turn out voters with nearly all of those electorates down in total votes in the 2017 preliminary results compared to 2014 (only Botany and Te Atatu recorded more preliminary votes) which seems incredible when you consider that there should be more people living in those electorates relative to 2014!

There's a few outliers in the data. New Zealand First's Clutha-Southland performance for one, which I suspect is driven by other, more recent events in that electorate, while Northland and Whangarei were likely driven by the presence of Winston Peters and Shane Jones respectively.

That's enough for now! Will churn through some more insights from this dataset soon hopefully. And obviously I'll try to re-run this analysis to to compare the final results from 2014 with the final results from 2017.

No youthquake, youth tremor needs phenomenal turnout

The youthquake looks to be dead in the water ahead of tomorrow's election with the latest enrolment numbers from the Electoral Commission showing  that while there's been improvements in the 25 to 39 segments, the 18-24 segment is well down on 2014, while those over 55 are also well up. Looking at the total makeup of enrolled voters, it shows a clear shift towards grey segments.

Relative change.png

While the gap has eased somewhat from the 19 September numbers I wrote about earlier today there's still a clear shift in the overall makeup of the electorate towards voters aged 55 and over. The scale of the problem facing any possible youthquake tomorrow evening becomes clear when you put the enrolment numbers through the likelihood to turnout from 2014.

Change in makeup of actual voters 22 September.png

Assuming similar rates of turnout across all age groups as 2014, there's still around a 4 percentage point shift towards voters aged 55 an over. 

To put this in context - a 10 point lift in turnout on those reduced enrolment numbers for the 18 to 39 segments is the bare minimum required just to negate the growth in the grey segments since 2014. That also assumes there's no though it'll be those aged between 40 and 54 who actually get squeezed the most due to the growth of the age segments above them over the past three years, and their higher probability to vote.

While there's no official data yet - I believe we might get figures in October - for the scale of Jeremy Corbyn's youthquake, current estimates put it at around 10 points. It's worth remember that Corbyn's youthquake only worked for him because of their FPP system, as overall the Conservatives did drive up their share of the vote to 42% (from 37% in 2015). So while the youthquake was valuable in individual seats, in an MMP system it would have done very little to impact things other than mitigating the likely damage caused to UK Labour.

That means that Jacinda Ardern and Labour are hoping on not only a bigger youthquake than Jeremy Corbyn managed, but also lower turnout in those aged 55 and over, or for those aged 40 to 54 to turnout out in favour of Labour more so than they have in their past. Of those, given that our youth turnout has been higher than the UK's, I suspect hoping on a bigger youthquake than what we saw in June seems very unlikely, so they'll now be hoping on those final two things to go their way enough.

With all this in mind, I think what we'll see at most is a youth tremor. The good news for Labour is that in such a close MMP election, even a tremor be just enough to tip the numbers in their favour for a Labour, Greens, and Maori Party government.

Enrolment data doesn't yet support youthquake

Change in enrolment numbers 2014 to 19 September 2017.png

Wishing something is happening doesn't make it so, such appears to be the case with claims that a youthquake is in progress. While youth enrolments might be increasing at greater numbers since writ day than they did in the 2014 election, the overall number of those aged 18 - 24 is well down, and there's only a small increase if you extend that up to those aged 39, on the overall numbers who were enrolled in the 2014 election.

What's more, the biggest increases in enrolments since 2014 aren't in the youth segment but in what I call the grey segment - those aged 55 and above. Even if you take likely voter turnout out of the equation, there's still a clear swing in the voting population towards grey voters.

Relative change.png

You can see that while there's promising signs in the 25 to 34 brackets, the only other growth has been the over 55 segments, meaning grey voters are poised to dominate the election, especially as they turnout in much greater numbers. When you take into account the turnout figures for each age group from 2014, and put that against the 2017 figures so far, you can see what I mean a lot more clearly.

Change in makeup of actual voters.png

Once again there's an overall swing towards grey voters, with their greater numbers and greater propensity to vote, meaning they'll have a bigger impact.

If a youthquake was happening it should be showing up in the enrolment data, with more young people enrolling to vote and thus lifting their overall share of the voting population. While there have been signs of that taking place over the past few days which I've written about the problem is that it might be too little, too late, to reverse the drift to older voters.

None of this is to say that a youthquake isn't possible, but it's not likely to happen due to surging youth enrolments, which means it's down to those who are enrolled to turnout and vote.

As much as I've written about the data not showing a youthquake, I do hope I'm proven wrong, because what matters most to the future of our civic institutions and democracy is that young people participate in our elections to ensure that our government is more representative of New Zealand as a whole, and that political parties are forced to adapt their policies to reflect that reality.

Don't bother waiting up. Advance votes will tell you who won

Party vote change 2002 to 2014 from advanced to final.png

If you were planning on sitting around the house on Saturday night for a good old fashioned election party, don't. Well, maybe you could stick around if you're supporting a minor party and they're hovering perilously close to the 5 per cent threshold or near to getting another MP.

Thanks to advanced voting we'll know the outcome as soon as advance voting results are released, which will be anywhere between 7:01pm and 8pm judging by past elections.

As you can see above, with ever increasing numbers of people advance voting, the difference between the party vote from advance voting results and the final result has been quickly decreasing. Based on the current numbers of people advance voting relative to 2014, I've estimated we'll get around 1,560,000 advance votes this year (versus 717,579 in 2014), around a 118% lift. I'll keep updating this each day as new statistics come through from the Electoral Commission.

What this means is that the advance voting result is increasingly similar to the final result. So much so that in 2014 National only dropped 0.81 points, Labour gained 0.64, the Greens gained 0.72, NZ First lost 0.47, the Māori Party lost 0.01, ACT gained 0.09, and United Future (who I've included purely because they had an MP in the 51st Parliament) gained 0.02.

With that in mind, I've attempted to calculate what impact more than double the number of advance votes cast will have on the change between the two results by looking at the shift between 2011 and 2014, as well as trying to calculate in the impact of what is currently an 11 point difference in growth between 2011 and 2014 and 2014 and what 2017 is tracking at. (2014 had 129% growth on 2011, whereas we're currently looking at 118% for 2017).

The net result of those calculations are below:

Projected Vote Change.png

What you can see is that at most it might rob National of a seat, and grant Labour and the Greens a seat each. NZ First looks set to lose at least one seat, if not two depending on how close they sit to a factor of 0.6%, while for the Māori Party, ACT, and United Future, it's unlikely to make much difference, though it could cost ACT a seat if they're sitting just over 1.2%.

Admittedly there might still be a bit of interest around certain marginal seats where the Māori Party is concerned that could create some fascinating outcomes, but it is largely the party vote that determines who will form the next government.

Update 7pm 21 September: With the latest advance voting numbers in, my projection for advance voting turnout has dropped by 45,000 to 1,515,000. The reason for this is largely driven by a big slow down in advance voting yesterday relative to 2014. For the same day in 2014, advance voting increased 37% on the day prior. This year it only increased 9%. Across the final four days of advance voting in 2014 numbers really surged. This might be a one day aberration and we could see normal service resume tomorrow, but I can't see it increasing enough to even meet my ambitious projection.

Assuming that the final two days also deliver 133,000 advance votes, we're looking at advance voting reaching around 1,070,000, so well short of the 1,560,000 I originally projected. 

The most important graphs you'll see today

Three Years Total Fans.png

As of 20 August 2017 the Labour Party overtook the National Party in terms of Facebook likes, a position they hadn't been in since 22 September 2014. At it's height, the gap between the two parties was around 18,000 page likes, as at the time of writing Labour is now approximately 400 likes ahead. Having been 200 ahead overnight.

National has enjoyed two significant boosts in the past year, in September of 2016 and from December through to January 2017 following Sir John Key's resignation and Bill English becoming leader and Prime Minister.

Labour's two boosts are from Parliamentary paid advertising just before the three month Regulated Period kicked in, and then from Jacinda Ardern becoming Labour leader at the start of this month.

Three Years Party Leaders.png

When you look at party leaders over a similar period (unfortunately my analytics tool can only pull data from September 2015 for most of them) you can clearly see the moments Bill English became Prime Minister and Jacinda Ardern became leader of the Labour Party.

Jacinda's rate of growth is now trailing off from the initial boost she received, but she's still growing by about 3,200 page likes more than Bill English each week. Assuming the rate of growth over the past week is more indicative of the campaign period itself, Jacinda will be just shy of 100,000 page likes by the election, while Bill will be around 107,000. Looking back at 2014's data, neither John Key or David Cunliffe seemed to enjoy a bounce from debates, but then again those debates didn't set the world alight, so a clear debate win for Jacinda could make a big difference for her.

As I've mentioned before, Facebook page likes aren't going to win you an election, but insofar that the more you have, the more people you can reach, and the morale boosting effect of knowing you are more popular than your opposition will have some impact on how you perform.

If you're curious where we sourced our data and graphs from from check out Quintly. It's an easy to use social media analytics tool that takes a lot of the hassle out of visualising your social media data.

The social media political revolution wasn't televised - you missed it back in 2014

A recurring theme I see in commentary on the 2017 New Zealand election is how this year will be known as the "social media campaign". It really isn't. 2014 was the campaign that social media revolutionised, what'd you think all those kids were doing with those selfies with John Key? They were posting them to Facebook, Twitter, and Instagram while at the same time being served up with a barrage of specially designed, targeted content from all of the major campaigns, political leaders, and third party campaigns.

I think part of the reason for this is that our traditional media companies themselves still don't quite get social media. Reading or watching much of their use and commentary on social media it's often dismissively served up as some sort of light entertainment medium that only millennials care about (the Paul Henry and AM Show's use of it's social media bunker is a good example of this), or as something that should have the country on the verge of a moral panic* (Stuff's recent "The Takeover: How Facebook is everywhere" was a good example of this).

The reality is that political social media is fundamentally the same as it was when it made such a big impact for Barack Obama first in the 2008 election, an approach that was ruthlessly refined by the 2012 U.S. elections. It's useful to sometimes think of political campaigns as an arms race - where each side continually looks for an edge over its rivals in terms of both getting its message to its audience, but then also converting that audience to voters. The development of this arms race from a data perspective is nicely documented up to 2012 in Sasha Issenberg's "The Victory Lab: The secret science of winning campaigns".

Essentially, Issenberg argues that George W Bush's team in 2000 and 2004 came up with a much better targeting model and method getting their messages out and converting voters to their side. In 2008 it was the Democrats and Obama's team, in large part thanks to a lot of earlier leg work by Howard Dean, who first clued onto the immense power of the web, email, and social media, to create two amazing campaigns from a content perspective, but underpinned by one of the best and most in-depth data operations ever seen. Dan Balz's "Collision 2012: The Future of Election Politics in a Divided America" picks up where Issenberg's work leaves off and provides a good platform for understanding how Obama's campaign leveraged social media to devastating effect against Mitt Romney and the GOP.

Which brings us to 2014 where National, Labour, the Greens, and Internet Mana all had significant social media and digital operations. It was just that traditional media was geared so heavily towards following the traditional on-the-ground elements of campaigning - campaign launches, policy announcements, debates and the like - that they simply didn't notice the importance or volume or work going on in the digital space.

The social media revolution happened to our political campaigns back in 2014, and it went entirely unnoticed except by those who were at the coalface of those functions. The tools and methods have been continuously refined since then, but the game- changing moment happened three years ago.

Without going into specifics, the numbers of unique New Zealanders who would have seen a piece of social media content from either John Key or the National Party on a weekly basis during that campaign were such that it was a massively more cost effective channel to reach a greater audience than other mediums could ever hope to deliver, and it had the added bonus of being an amazingly easy platform for people to engage with content in a way that traditional communication channels - face-to-face, TV, newspaper, radio, or physical mail/leaflets - simply couldn't (and still can't) match.

The political social media revolution well and truly happened back in 2014. The only difference between 2014 and 2017 is that traditional media companies have finally realised what people like Matthew Beveridge had already clicked onto in 2014 - that social media is a game-changer for the scale and nature of political campaigns in much the same way that radio, direct mail, and television have been in the past.


*Further to this point about inciting a moral panic about social media - the reaction to social media by traditional media across the world parallels, in many respects, the backlash to all new methods of mass communication as they're appeared by the previously dominant sources of information. Pamphleteers were routinely censored and harassed throughout Renaissance Europe, newspapers heavily controlled by governments in their early days.

More recently serial novelisations, when they first appeared in the 19th century, had established book publishers lashing out at these new shorter forms of writing with their appeal to the working class who were rapidly becoming more literate. Charles Dickens was one of those serial novelists who was slammed by the established content producers of the day for his work - especially its moral character, the social themes addressed in it, and its mass appeal - but is now considered to be one of our greatest literary figures.

There was a similar reaction from newspapers when radio stations first became commonplace, with many forbidding radio stations from reading their newspaper stories on air until after the physical papers had been delivered. Radio stations responded by employing their own news departments. The reaction to cinema and television in particular was similar too, with established players seeing the new technology as a threat to not only their own survival, but common decency.