Tuesday, March 28, 2006

Polls - not news

I'm glad to see most people ignore the reports of approval ratings that appear in the media periodically. You know what they say about statistics ? When a poll is taken, there are many contentious issues, from the selection and size of a sample group to the specific wording of the questions, right through to the way different media entities report the results.

For example, the NY times will always spin the results of a poll to show that Pres. Bush's approval ratings have taken a nosedive in response to some policy of his, even if there is nothing in the results to show that.

Most newspapers don't even report how the data was gathered, and how big the sample size is. If you are trying to approximate what the general population's opinion is, you need a fairly big sample. If you sample 10 people on a street, the margin of error is very huge. If you sample 100 or even 1000, its still a few percentage points of error. And sometimes, on a rare occasion, the unreliability of these polls is exposed for all to see.

The Age:

KIM Beazley has been dealt a massive personal blow in the latest AgePoll and the shockwaves of Labor disunity are hitting the party’s support. …Labor’s primary vote fell 3 points in the ACNielsen poll to 37 per cent, while its two-party preferred vote was down 2 points in the past month, putting it equal with the Coalition on 50 per cent
The Australian:
The Newspoll in The Australian showed Labor leading the coalition 53 per cent to 47 with a seven-point rise in its primary vote, while the ACNielsen poll in Fairfax newspapers had the government and opposition level-pegging at 50-50.
Often the only reason the media run these polls is to give themselves some news to report on, and so they can write a few editorials about what might be behind the change in approval ratings.