Sunday, August 14, 2005

Moral issues

The Sunday Star-Times have published the results of their morality survey. The most significant result is the polarisation on party lines: 83% of self-identified National voters thought that sexual infidelity was wrong, versus 65% of Labour voters; the split was 64% vs 44% on whether a politician's adultery should be a factor in voting decisions, 83% to 15% on whether the government had lost touch with family values, and 56% to 9% on whether homosexuality was immoral. Green supporters tended to be more liberal than Labour, NZFirst supporters slightly more conservative than National, and the Christian loon parties far more conservative. This is a fundamental difference in outlook, though demographics are also likely to be a factor; Labour and Green voters tend to be younger (and therefore more liberal on at least some of these issues) than National and NZFirst ones.

There's more summaries here and here, but you probably need to buy the paper for the graphs.

We should be cautious about these results - the 10,000 respondents were self-selected, and not entirely representative:

David Fougere, of independent pollsters Phoenix Research, which conducted the survey and analysis for the Sunday Star-Times, says the findings are in part the result of people writing in who may feel otherwise that their important opinions are not being heard. "The survey may also have attracted those with more extreme opinions, whether liberal or conservative.

There's also a heavy bias towards United Future supporters, and evidence of efforts by more conservative groups to stack the survey. Despite this, they conclude that "the 10,000 survey participants are not too different from the profile of all New Zealanders". I'm not so sure, but I think the results are still interesting, even if of questionable validity.


Nice blog. Unable to resist the similarities in titles, I've added a link to you over at Swerve Left.

Posted by Karlo : 8/14/2005 02:38:00 PM

I can't see how any self-selecting sample can be representative, except by coincidence.

You might find that such a sample mirrored the political makeup of the population - but would still tend to favour the opinionated.

It would be interesting, should the "Dunne decision" be upheld, whether one could seek to legally prevent a TV company from airing inaccurate "polls" as representative of national opinion.

Posted by Rich : 8/15/2005 10:05:00 AM

Hrm. Although I understand that there is a strong bias, I can't help let the n = 10,000 part take over my thinking.

10,000 respondents must be indicitive of some sort of trend.

Posted by T : 8/15/2005 11:39:00 PM

Why? It's less than 1/3 of 1% of the adult population.

That's why media promotion of these "polls" is so insidious - most people don't understand statistics (actually, they don't even have a grasp of the relative size of numbers - largely because school arithmetic consisted of trying to make kids learn what 6x7 is).

Posted by Rich : 8/16/2005 02:44:00 PM

Rich: yeah, but how many responses would a survey on economic attitudes have received?

I'd also note that a survey with 10,000 responses is fairly large for New Zealand (well, outside of anything done properly by the government). Its a shame about the systematic bias - though the subsample data on views by political parties would still be valid.

Posted by Idiot/Savant : 8/16/2005 03:31:00 PM