Monday, July 21, 2014

POLLS DON'T TELL THE REAL STORY


“Public opinion polls,” said J B Priestley, “are rather like children in a garden, digging things up all the time to see how they’re growing.” 

Love or hate them, the odds are that you will remember having heard about at least one, if not all, of these polls in recent times; One News Colmar Brunton, 3 News Reid Research, Herald-Digipoll, Fairfax Media Ipsos, Roy Morgan Research.  Ring a bell with you?  That’s because in New Zealand these are the big five, and they’ve been polling weekly since the 2011 election on everything from Party Vote to Preferred Prime Minister. 
But how accurate are they, and how much faith should Maori put in them? 
When done right, political polling is a social science with strict rules about representative sample size, random selection of participants and margins of error. 
It isn't always so scientific of course.  A straw poll alludes to the ancient farming practice of tossing straw in the air to see which way the wind is blowing.  Such polls are run regularly by different media, including this newspaper.  They give a very rough idea of which way the wind of public opinion is blowing on an issue. 
However the big five polling companies in New Zealand use mathematical methods and computer analysis to get the most representative sample of the New Zealand voting public they can get, in order to gauge the political opinion of the entire country.  That means the sample group has to represent the larger population and has to be selected as randomly as possible.
The most popular method for doing that is through random digit dialing (RDD) using a continually updated database of all listed landline numbers in the country.  Engari, if pollsters only called the numbers in the database, then they'd exclude all unlisted numbers, which would muck up the randomness of the sample. So they also programme their computers to randomly dial every possible number combination of all area codes, exchanges and numbers in active use.
To randomise the sample even further, pollsters not only dial random numbers, they also try to choose random respondents. To do that they may ask to speak to the voting-aged member in the whare with the most recent birthday. 
Next week we’ll cut through some of the mystery around ‘margins of error’ in polls.
In the meantime, if you are one of those (un)fortunate enough to be called by a pollster, you no longer have to wonder how they got your number.  But you might want to keep in mind that there is always a ‘but’ to these polls. 

For example, they include landline numbers, but exclude mobiles; so much for representativeness.  And as Helen Clark and the last Labour Government can attest, the polls can be devastatingly accurate.  But as Iwi Insiders, Ipurangi MANA and Winitana know, the ‘polls don’t tell the real story.”

No comments: