Visualising peaks and troughs in the election cycle

I’m not just an evaluation researcher. I’m also an election geek.

I’d like to put the blame on the lyrical descriptions of election results woven by my politics lecturer, Professor Bob Chapman, but the truth is it goes well back to my teenage years. I vividly recall tallying every electorate result on the night of the desperately close 1981 election; those numbers just really mattered to me. I find election night parties incredibly frustrating if people just want to chat – for goodness sake people, there are votes being counted!

So when I came back from the data-viz workshop this month, there was really only one thing on my mind. What could I do with my database of poll results from the last three years?

One of the challenges of visualising poll results, particularly in New Zealand where we have so many minor parties, all with a fighting chance of gaining representation in parliament, is that change over time graphs of polls results inevitably turn into the fairly distinct blue, a red and green lines of National, Labour and the Greens respectively, followed by an inelegant spaghetti of the other parties where discerning performance is at best challenging, like this one.

With this in mind, I thought a visualisation, at this stage of the election cycle, of how close each party is from their high and low points of the past three years, might be worth exploring. For anyone following electoral fortunes, it’s a pretty common question, one that’s often sought for glimmers of hope.

And using a fairly simple poll of polls, here’s how they look: at the time of writing, the governing National Party had peaked only a month earlier, at 53%. By the end of July they were still holding strong at 50%.

Labour hit a trough of 26% in early July and had clawed back a little by the end of the month, but still well off their peak of a year earlier. The Greens performance remains strong enough to return them to parliament. But the sobering reality for Labour and the Greens is that they need to regain close to their support of nearly a year earlier to have a chance of governing, whereas National simply needs to hold on to much of its June peak of support, and look to see if any potential coalition partners make it back into parliament.

Among the smaller parties, the recently formed Mana-Internet alliance has hit a peak just out from the formal campaign. This is not entirely surprising for a fledgling alliance, but given earlier in 2014 the Mana Party was registering less than 0.5% in many polls, the whole does appear at this stage to be more than the sum of its parts.

New Zealand First are well off their peak, but still in with a good chance of crossing the 5% threshold. Other small parties are struggling to retain support. The Conservatives star has waned from February, and with little chance of securing a crucial electorate seat (unlike many other smaller parties), looks unlikely to be in parliament post-September.


All this said (and yes, this is the equivalent of “cricket’s a funny old game and anything can happen”), election campaigns can change fortunes rapidly – look at how Helen Clark nearly snatched victory in 1996, or how Peter Dunne became the ‘common sense’ touchstone in 2002, or how Don Brash imploded in 2005. Suffice to say, the heat of the campaign and the head to head tussles of party leaders may yet see some shifts.

A note on the data: they’re based on a rolling average of the past four political polls (Roy Morgan, One News Colmar Brunton, 3 News Reid Research, and Fairfax-Ipsos), with a very simple weighting that gives greater emphasis to the most recent polls, and less emphasis to the least recent. It’s intended to smooth out some of the variation between polls, but still capture the fluctuating fortunes of parties. The underlying assumption is that voting intentions both have an element of inertia, and at the same time are responsive to recent events.

My thanks to Stephanie Evergreen for feedback on this visualisation. She rightly pointed out that this isn’t change over time, it’s just peaks and troughs. If you want better ways of showing change over time, you need to do a change over time graph. So, ahead of the election, I plan to look at the shifting fortunes of potential coalition partners.

In the meantime, what do you think of the approach I’ve used here? Feedback welcome!