Political poll results may exclude 30%?

The Electoral Commission might be well advised to see if some research companies are excluding up to 30% of respondents before they publish party vote poll results.

The commission in 2008 allocated $3.2 million to political parties to fund election broadcasting.

One of its criteria to allocate the money includes considering “public support for a political party such as the results of opinion polls”.

However, Horizon Research says an investigation may show some polling firms are excluding up to 30 out of every 100 people who respond to them from published results, providing a distorted picture of  party support levels in the electorate as a whole.

This could over represent support for larger parties and under represent support for minor parties.

If this is happening it could result in poor information going to the Electoral Commission when it decides on the allocation of public money.

Horizon says if 30% of respondents have been excluded from published results for the latest post-Budget DigiPoll this would be an example.

In that event, it would be rating National at 16.3% to 19.6% higher than its actual support.

Labour support would also be over stated, while minor parties’ support could be understated by up to about 3%.

Horizon says the differences arise because some of the polls are publishing results only for those who say they have decided which party they will vote for. This may include people not eligible to vote, not registered to vote and not intending to vote.

However, respondents who do not know, choose not to vote, and prefer not to say may be being excluded. This precludes providing a total picture of the electorate.

In its post-Budget poll of a representative national population sample of 2,254 Horizon found 15.6% undecided, 6.3% choosing not to vote, 3.9% wouldn’t say and 4.2% voting for minor parties (other than NZ First, United Future, Mana and ACT), a total of 30%.

This would mean a survey like the DigiPoll one would have had 1071 total respondents, 750 decided.

Recalculating the DigiPoll on this basis would bring the latest HorizonPoll and DigiPoll results within each other’s maximum margins of error.

National’s committed support would be 38.2%, not 54.5%, Labour 23.6%, not 33.7%.

At the 2008 election, National won 32.9% support among the population aged 18+, Labour 25%.

Horizon says results are also affected by sample size, and the factors used to weight responses.  For example, responses can be affected by about 2% to 3% for the main parties if they are not weighted by respondents’ 2008 party vote choices.

Horizon weights by age, gender, ethnicity, personal income region and party vote 2008 and seeks sample of sizes of about or more than 2000, in order to reduce margins of error of the total sample to about 2.1%.

Factors the Electoral Commission must consider when allocating party political broadcast time are detailed here.

The Horizon Research analysis, comparing its post-Budget HorizonPoll with a DigiPoll one which had 30% added to the respondents is here.

For further information please contact:

Grant McInman, Manager, Horizon Research Ltd

Telephone: 021 076 2040

E-mail: gmcinman@horizonresearch.co.nz