Voting with preferences has not been used in New Zealand Parliamentary elections before, so we have no historical data to draw on. We have instead relied on a survey done in 2008 by the New Zealand Election Study, which contained a question asking respondents to rate political parties on a scale of 0 to 10. The survey sampled 2330 respondents from the general electoral roll and 712 from the Maori roll. You can read the survey questions here.
We converted the NZES responses into preference orders. Where there were ties (two parties given the same score out of 10), we included both possible orders, each with half weight. Parties with the lowest score, as well as any parties with no score given, were omitted from the preference order. In preferential voting systems, this effectively ranks them equally among themselves, and below the lowest explicitly ranked party. The resulting preference orders were assembled into two collections: one for voters on the general electoral roll and one for the Maori roll, with each collection categorized according to the first preference. Some rare preference orders were then discarded in order to keep the file size and execution time of our calculator to reasonable levels. In each subcategory, no more than 20% of the total (weighted) number of responses was discarded in this way. The final preference database comprises 1501 different rank orderings of up to seven parties, as follows:
Number of preference orders | ||
---|---|---|
First preference | General Roll | Maori roll |
National | 352 | 85 |
Labour | 205 | 91 |
Green | 127 | 48 |
NZ First | 109 | 43 |
ACT | 92 | 10 |
Maori | 121 | 116 |
United Future | 92 | 10 |
We did (briefly) consider putting a form on the website that would allow the user to enter the prevalence of different preference orders - but soon realised that the sheer number of possible preference orders would make this a bigger job than most casual web surfers would care to take on. There are actually 13699 different ways to rank some or all of seven parties in order, and surprisingly many of these seem to occur in the voter population.