ONE News Colmar Brunton Poll: July 2008 - TVNZ response

> LAWS179: "ONE News Colmar Brunton Poll: July 2008" A prompt and self-explanatory response from TVNZ and its pollsters below. It's a little trap to be aware of if one is re-calculating the poll numbers based on other assumptions. For example, I know that David Farrar's Curiablog has been re-calculating the seats in the House based on the published figures and has similarly ended up with different figures to TVNZ. I've asked if TVNZ's pollsters could make the more precise figures available - but we'll see. ------------------------- Subject: RE: ONE News Colmar Brunton Poll: July 2008 The reason Dean gets a different result is due to rounding. As you know within the report, for Party Support, we round figures up or down that fall above 5% and for anything below this, we present the percentages to one decimal place. In the rubric at at the front of the report we do mention this: "...For Party Support, percentages have been rounded up or down to whole numbers, except those less than 5% which are reported to 1 decimal place...". So in the last report the figures were as follows: - National 51.8% (which we rounded up to 52%) - Labour 35.2% (which we rounded down to 35%) - Green Party 5.9% (which we rounded up to 6%) - New Zealand First 2.4% - Etc When we calculate seat allocation, we in fact use the actual number of respondents stating a preference for a party, rather than using percentages (either rounded or unrounded) so we can be as precise as possible. See tab 'CB Calc'. In the following tab, in the attached Excel document, 'On line calc', you will see three sets of computations taken from the Elections New Zealand on-line seat calculator. These are: - Result A - Seat allocation using the actual numbers - how we normally do this calculation. This gives us the same result as on tab 'CB Calc'. - Result B - Seat allocation using unrounded percentages. On this occasion it gave the same result as A) - Result C - Seat allocation using rounded percentages. This is how, I think, Dean did it - and it is the use of these rounded percentages (that fall above 5%) that gives us this variation, on this occasion. In short, the issue is 'rounding'. ... -------------------------