The failing polls..

An article in The Economist looks at the failing basis of polling techniques in the USA. It isn’t that much different to the circumstances here.

THEY are used to determine spending on political campaigns, to guide tactical voting and to enthuse or cow voters. News outlets, bloggers and activists cannot get enough of them. Candidates live and die by them. But even pollsters are worried that many of the political polls in individual races unveiled in the run-up to the mid-term elections in November may be inaccurate. “We’re in a dark age for polling,” says Jay Leve, boss of SurveyUSA, a polling firm. “It’s a very uncertain time.”

The immediate problem is the rapid growth in the number of people who have only a mobile phone, and are thus excluded from surveys conducted by landline. About a quarter of Americans are now “cellphone-onlys” (CPOs) in the industry jargon, and this poses both practical and statistical difficulties. They are less likely to answer their phones, and less likely to participate in a survey when they do, says Frank Newport of Gallup, another polling firm. They often retain their telephone numbers, including the area code, when they move from state to state, so it is hard to know where they are. And it costs more to call a mobile phone in the first place.

This is exactly the same problem that is steadily undermining the statistical validity of polls in NZ.  I’ve been involved in doing the numbers and canvassing in Labour for decades, and we see exactly the same issues. The significant difference here is that we are probably a lot further along. As well as the CPO’s we also have an awfully large number of urban centre people whose land-line is unlisted. This is largely the inevitable result of telemarketing – which was why I unlisted my phone number almost 20 years ago.

Pollsters can always “rinse” their data to try to correct such flaws. As Mr Newport of Gallup points out, most firms produced reasonably accurate findings ahead of this year’s primaries. It helps that the sort of people who do not have landlines are also the sort who tend to turn out in relatively low numbers in mid-term elections, notes Scott Keeter of the Pew Centre.

Which isn’t going to last. The CPO’s tend to vote more as they get older. In NZ the percentage of people with listed land-lines is currently pretty static in the over 45’s and starts falling away dramatically in people who are younger to very small percentages in the 20-25 age group. The actual profile is different in different areas. But when I first noticed it over 15 years ago the tipping point was in people in the 30-35 age group.

In the central Auckland electorates the number of households with listed phone numbers is now typically less than 60% and falling rapidly, whereas those in rural or more technically conservative areas can be over 80% but falling slowly.

Furthermore with the advent of caller id onto most phones, you get technophiles like me using it to screen calls. I don’t pick up if I don’t know who the caller is. They can leave a message because they’re probably not important to me.

If I’m doing anything like coding, moderating, or writing a post – I won’t answer even if I know who it is (unless it is family). Anyone that actually knows me also knows that they should e-mail or txt – it is far less disruptive to my work flow and limited leisure time. I hate having to disrupt my concentration. Of course I’m probably somewhat extreme and living around the bleeding edge. But other people are starting to head the same way…

But even if fears of a systematic bias prove unfounded, the fuss about CPOs points to a broader problem for pollsters: the ever-increasing difficulty of persuading Americans to take part in political polls. The proportion of those called who end up taking part in a survey has fallen steadily, from 35% or so in the 1990s to 15% or less now, according to Mr Keeter. Reaching young people is especially difficult. Only old ladies answer the phone, complains Bill McInturff of Public Opinion Strategies, a Republican polling firm.

Of course this isn’t just a problem for the pollsters.

It is also an issue for political parties canvassing to get a feel for voters in particular areas. Canvassing is a good tool for both getting to know the details of your electorate, and for convincing people to get out to vote. Typically you not only activate the person you talk to, but also those to whom they talk about it.

When campaigners of the left are doing their planning in the next few months they need to think about the limits for different canvassing strategies. Phone canvassing is a relatively cheap and effective tool because it allows wider coverage across our rather large electorates of fifty thousand potential voters. But we also need to seek out the pockets of people who are not on a listed land-line.