Friday, September 15, 2017

The problem is numbers

Recently, New Zealand political commentators were thrown into a frenzy of commentatoring when a poll seemed to show the National Party had suddenly zoomed way out in front of Labour and could govern alone. Was it right? Commentators breathlessly commentatored about that up until a new poll came out a couple days later showing that Labour and the Greens could form government together, and National couldn’t form one even if all the minor parties (apart from the Greens, of course) backed it. What this actually shows isn’t who will win the election—which is far too close to call—but that New Zealand has a huge problem with numbers.

The first and most obvious problem is the polls themselves. Only three companies are conducting public polls this year, only two of them are doing so for news organisations. This means that we have very little data to work with, and the “poll of polls” averages are actually pretty meaningless. In the USA, presidential elections have dozens of polls, and many major ones, to draw on when averaging them out, giving those averages an accuracy New Zealand can’t ever achieve.

Another problem is with how the polls are conducted. Colmar Brunton, which conducts polls for TVNZ’s One News, phones 1000 landlines only, and not mobiles, in an era when people are giving up their landlines. Even so, they vet their respondents to make their sample match the electorate as closely as possible, so I’m not convinced that fact makes much difference—but inevitably, it will.

Reid Research, which does polling for Newshub (TV 3’s News), only samples 750 landlines, plus 250 “online”, which they don’t further explain. We can evaluate, up to a point, the phoneline surverying, but because we don’t how Reid does its online polling, we have no reason to believe it’s accurate OR inaccurate: We simply cannot even guess. This is totally unacceptable, and any company conducting political polling during an election campaign—especially during the final 2 weeks, when most New Zealand voters make up their minds—has an ethical obligation to be totally transparent with about their methodology so experts in polling can fairly evaluate the data, especially how it was collected.

Which brings us to the second numbers problem: Political journalists. All news organisations have a political editor, some of which are better than others. One who is often good, Newshub’s Patrick Gower, damages his believability with over-the-top histrionics which oversell whatever story he’s hyping, most recently the Reid Research poll. TV One’s Corin Dann, in contrast, has maintained a far more measured demeanour during this year’s campaign—as have most others.

But a few good individuals notwithstanding, NZ doesn’t have a class of journalists who are good at reporting on politics or statistics. They are too quick to accept what they are told, especially by the government of the day, without looking for supporting evidence—or contradicting evidence. They also don’t understand statistics, opinion polls in particular.

Even worse, the quality of our political commentators—pundits—is appalling and awful. Most are chosen only for partisan bias, and the commentary they make is useless and banal regurgitation of partisan talking points. They offer no insight, no unique perspective, just partisan noise. It’s implausible that in a nation as diverse as New Zealand, and despite its small size, those same few people are the ONLY ones who can offer coherent political commentary on the issues of the day. I’m beginning to think that the idea of “term limits” for pundits is a great one.

Where all of this leaves us is—well, no one can say. We have wildly conflicting polls, which we can’t fairly evaluate, we have commentators—both journalists and partisans—who can’t be relied upon to tell us the straight truth, and yet we also have a clear mood for change. This all adds up to this: No one can tell what will happen because the election is too close to call.

Despite that, in the first four days of voting, 229,259 people voted, as compared to 98,063 during the same four days in 2014. But even this number, widely shared on Twitter, is misleading. In fact, 2014 Advance Voting began five days earlier than it did this year, and during the whole 2014 Advance Voting up to and including September 14, 147,560 New Zealanders voted, as compared to 229,259 this year, in fewer days. So, Advance Voting is up pretty dramatically this year, and is on track to be another record year. [complete stats on Advance Voting are posted on the Elections NZ website at 2pm each day]

In 2014, Advance Voting favoured the National Party. I’m not convinced that will be true this year, but we’ll see on Saturday the 23rd. We’ll also find out then whether any opinion poll was even somewhat close to getting it right. What, leave it to the voters?! Imagine that!

No comments: