Herald Insights interactive shows how population characteristics affect the probability of voting for one party over the other.

Places with higher proportion of Pacific voters are most likely to vote for Labour Party over any other party. Whereas, places with high proportion of European voters are most likely to vote against the Labour Party.

The model behind the interactive combines the 2014 election results with the census data.

Readers can explore the comparisons between different parties.

Read more - Labour party's game plan to beat National in next year's election

The way to read the interactive is to consider proportions of individuals.

Some starting from different charts are below -

  • Labour is less likely to get votes in places with high proportion of full-time employees compared to National.
  • Best predictors for Green voters are places with high proportion of people being self-employed and atheists.
  • Best predictors for National voters are places with high proportion of people from European ethnicity.
  • In a Labour versus Greens contest, places with higher proportion of educated voters are more likely to vote for Greens.

A comparison between different parties allows readers to see the shift in voting patterns for different population characteristics.

However, it is not possible to infer voting patterns for individuals using this analysis.

The model uses meshblock data around voting places. The results seen on the visualisation show which factors result in higher votes for a particular party compared to others.

In case of the horizontal line chart, the model singles out the impact of individual characteristics.

Even considering the social patterns, such as higher proportion of European ethnicity, it is useful to consider that people whether it is the ethnicity or living in a less diverse neighbourhood that has the impact on voting patterns.

The scatterplots which allow readers to explore meshblocks simply show the relationship between two variables. It does not remove the effect of different characteristics and should be read as complementary.

The interactive is based on analysis by Peter Ellis. It includes data from the 2013 census and the 2014 election results, combined by Mr Ellis in nzelect R package.