Wednesday, July 21, 2010

The value (or lack of) of online polls

On the Daily Telegraph website today a poll asks who will win the upcoming election, Julia Gillard or Tony Abbott.  The result as at 9.35am is:

Untitled

Which is funny considering the latest Newspoll shows:

newspoll 

Gillard: red; Abbott: blue

So how can we marry up these two conflicting polls?  What value should be placed on online (or even phone in) polls? 

I would suggest none. 

Online or phone in polls are based on people voluntarily using the website or calling the poll number.  This means they must be motivated to enter the poll and it may be biased for a number of other reasons. 

It may be that one type of party voter is more interested in taking part in the poll on a particular issue.  Voters who vote ALP may be more interested in reading about IR policy stories which have a linked question conversely Coalition voters may be more interested in a story linked to business.  This would therefore lead to a skewing in the sample pool.  The sample pool is hardly random. 

Other than this you just need to look at the above contrasting polls to know that something is up!

 

UPDATE:

This is further evidence of the lack of cogency of online polls.  On Monday 2/8/10 the Couriermail had a poll asking “Which party would get your first preference”  And the results:

1

So who reads and contributes to the couriermail polls?  Overwhelmingly Liberal supporters!

2 comments:

  1. I'm never in the mainstream with these online polls. I get the impression they're a bit like the old cricket scoreboards - open to interference. It seems for every Labor vote there's 2-plus Liberal. Perhaps its to keep advertisers happy - profit explains all sorts of skewed outcomes.

    ReplyDelete
  2. I agree Cathy. Is generally skewed towards the Liberal Party. May have something to do with readership. More Lib readers?

    ReplyDelete