Wednesday, November 09, 2011

Numbers and reality

One thing I’ve been consistent about when talking on this blog about polling is that to be believable, pollsters must report all their data, not just the headline result. This means not just their margin of error, but also their confidence level. They also must release their polling questions, methodology and sample size. If any of these details are withheld, we cannot be sure whether the poll is accurate or reliable.

This is true for all polls, whether on issues, attitudes or elections. While we can evaluate a polling organisation’s reliability by looking at their track record, full disclosure removes doubt—especially because past performance is not necessarily an indicator of future accuracy.

All of this is especially relevant here in New Zealand. Political polling has consistently shown that the ruling conservative National Party has around 50-53% popular support, the Opposition, the Labour Party, has around 30-35%, the Greens around 10-12% and the other parties dividing the rest.

Will it hold—is the election over? Of course not.

History is relevant here. First, polls always tighten up when the election happens. Part of the reason for that is that most New Zealand polls exclude undecideds, even if they’re strongly leaning toward a party. In 2008, the final result was reasonably close to the average of polls, but there was also low turn out, particularly among Labour voters.

And that’s the danger in this election: With the newsmedia constantly saying that National is polling highly enough that it can govern alone, Labour voters may stay home, or vote for the Greens. This is why the reliability of the polls is so important—if the polls are wrong, they may nevertheless become a self-fulfilling prophecy.

There’s also no way to be sure that something won’t happen in the final two weeks that could change what the polls seem to indicate. In 2005, the polls were showing that National was on track to defeat Labour, but their campaign collapsed in the final stretch when their collusion with a secretive far-right “Christian” cult was revealed. So, reversals of polling fortune telling do happen.

But with the newsmedia pretty much uniformly declaring the election all but over, and many of them clearly favouring and all but openly promoting John Key (the New Zealand Herald and Fairfax being the worst of the lot), Labour has a tough job ahead of it—certainly not impossible, but difficult.

The New Zealand newsmedia have made a switch from reporting the news to leading it (not creating it, which is different), and that’s a serious threat to democracy. I don’t know that anything can be done to impose ethics and balance on the newsmedia, but one thing the newsmedia can and must do is to be completely transparent about their polling. Until and unless that happens, we have no alternative but to assume there is a wide gulf between their numbers and reality.

1 comment:

Roger Owen Green said...

I hate polls. I hate that they determine how it decides elections, or how one votes. I could (and did) vote for Nader in 2000 because I believed, from the pols (which were accurate) that gore would win NY. But if I had done that in, say Florida, I would have been "throwing my vote away."

I also hate the reporting of polls as if they were news. Almost no one mentions margin of error, so Romney beating Cain 24-23 is statistically no different from Cain beating Romney 25-23 the week before, when the MOE is =/-4.