Australian robopolls

Tim Colebatch at the Morning Herald reports:

If you believe the opinion polls, there must be two elections going on. One is the Federal election, where the established pollsters agree that the Coalition has established a small but growing lead, averaging 52-48 on their latest polls. That's a swing of 2 per cent.

The second election seems to be taking place in individual seats. It's brought to us by the new kids on the block, the robo-pollsters, who use automated phone calls to bombard us with surveys that report huge swings to the Coalition.

On Thursday, a Guardian Lonergan poll reported that Kevin Rudd stands to lose his own seat of Griffith in a 10.5 per cent swing against Labor. If you haven't heard of Lonergan, it's because they are brand new; their accuracy is untested.

But Lonergan is on a roll. Last week it stunned us by reporting that Labor's other big name in Queensland, Peter Beattie, would lose Forde in a swing of 8.5 per cent to the Coalition. And it told us Assistant Treasurer David Bradbury stands to lose his western Sydney seat of Lindsay in a swing of 11 per cent.

Uh-huh. Lonergan's own national poll reports only a 2 per cent swing against Labor. Yet in the three seats it polled individually, it found an average swing of 10 per cent. That's huge, far bigger than we have seen in any Federal election since 1943.

Its Rudd poll came out as The Australian's Newspoll reported a swing of 2 per cent to Labor in Queensland. Is there a swing to Labor in the other 28 seats in Queensland, but a landslide against it where Rudd and Beattie are standing? Maybe not.

It's not just Lonergan. Another of the new kids using robo-polling, ReachTEL. shocked us last week with polls in four Sydney seats reporting an average swing against Labor of 10 per cent, with Treasurer Chris Bowen another big name heading for defeat.

Last month ReachTEL went polling in Tasmania, and reported swings against Labor of 11 per cent in Bass and Franklin, 14 per cent in Braddon and 17 per cent in Lyons. There's a bit of a pattern here.

Last Saturday JWS Research polling in eight seats found an average 6 per cent against Labor, relatively modest by robopollster standards.

Yet the established polls with a strong track record such as Nielsen, Morgan, Newspoll and Galaxy on average report a swing of 2 per cent. All of them came within 2 per cent of the actual result last time.

Well, they and the robopollsters can't both be right. Someone will be left looking pretty silly on election night.

I don't think any pollster in NZ uses robopolls. Their advantage is they are cheap (no staff to pay!) and data is entered directly by the respondents (pushing buttons on phones) so analysis can be instant.

However the concern is that they become self-selecting, and hence not representative.

Traditional phone polling relies on the fact that because there is an actual human on the phone nicely asking you for a couple of minutes of your time, you will agree – even if not greatly interested in the topic.

I suspect with robopolls (and to a degree some Internet panel polls) that those with a higher degree of interest in an issue will stay on the line, and those less motivated will hang up as you are hanging up on a machine, not a person.

Hence it is not surprising that robo polls may be showing bigger swings against the Labor , as people are often more passionate about booting a Government out than keeping it in.

We'll see of course come the election, but like the author I am sceptical of the massive swings being shown in some Labor held seats.