Why is this Information Not Provided on All Polls? Let us make our own minds up on what they mean

pn415

Showing just the tip of the iceberg

 The results of two major polls were released last week but on almost all of the questions asked no information was provided on the relationship between respondent characteristics and their answers, and this detracted from their usefulness.



Labour or National goes up or down but we need to know what proportion of respondents were Labour or National supporters?  What was their sex, their age, rural/urban place of residence, their household incomes and, in some questions, their religious beliefs?  In statistical language, we need to know the relationship between independent and dependent variables; in common speech,  the relationship between respondent characteristics and their answers, or  who they were and how they answered.

True, people could get buried by the additional information,  but in the printed media, at least, a mention or a table or two would help, and a hyperlink should be provided to the pollsters' reports. The pollsters  had all this information, and presumably gave it to the media. It just was not used in the media releases.  This is what I think we should see in the hyperlink.

Part of the Colmar Brunton poll, published separately, gives this information. If it could be used for this one, it could be givern for all.

The question asked was "Should government continue with the KiwiBuild scheme?" 

 Of the 65% who said "yes", many were Labour supporters, Pacific Islanders, and aged `18-29 years.  Of the 30% who said "no" many were National supporters, more than expected lived in Otago or Southland, and many had household incomes over $150,000. 5% "did not know."

There's probably nothing surprising in this particular poll, but there could well be in others and even here it would be interesting to know the proportion of National and Labour supporters who answered in an unexpected way. For example, Labour supporters who said "no" and wealthy National supporters who said "yes" to this particular question. 

If  we were given access to information of the relationship between respondent characteristics and their answers,  many or us  would be able to make up our own minds on the significance of poll answers.  We would be in a much better position to understand  not just what respondents answered but why they answered the way they did.

We would no longer have to put up with an opinionated columnist or two stuffing  their limited and too often one-eyed interpretations down our throats. 

We could make up our own minds.

-- ACW