So whats the deal with preferences?

Australian politics has become progressively more interesting, with the percentage of primary vote being cast for one of the two major parties steadily going down over the last three federal elections. So the upcoming Western Australian state election in March and four others coming up in 2018 will all be interesting to watch if the last federal election is anything to go by.

The Attack of the Minors

The 2016 federal election in Western Australia was not as interesting as it was in the rest of the country. There were fewer minor parties running and both One Nation and Nick Xenophon Team were not present, as they were in the eastern states. This won’t be the case in the state election, with about 16 registered parties fielding candidates, including One Nation.

Most of these probably won’t really have much of an impact. They won’t have the same media attention as the larger political parties and generally they cater to very specific groups – unlike One Nation, which is likely to have an impact, with some polls showing One Nation projected to receive 13% of the primary vote.

One Nation and the 2016 Federal Election

One Nation had a number of candidates in a number of electorates in Queensland at the last federal election, and they attracted a significant percentage of the primary vote. This can probably be attributed as much to the ongoing drop in interest in the major parties as the high visibility One Nation attracts from the media.

The level of interest One Nation attracted echoes a similar phenomenon seen in the 2013 Federal Election in Queensland with the Palmer United Party, which fielded a large number of candidates within the state and received a respectable percentage of the primary vote.

Where do the votes come from?

Australia uses preferential voting for almost all elections. In federal elections for the House of Representatives, the preferences are fully distributed, to produce a two party preferred (TPP) statistic for each district. Most of the time this ends up splitting the votes of minor parties between the ALP and the Coalition, a pairing of the Nationals and the Liberal Party (or the Liberal National Party if you live in Queensland), except in those rare cases when a representative from a minor party or an independent wins a seat or gets enough votes to come second.

The TPP statistics are published by the Australian Electoral Commission (AEC) and are available through the commission’s website. These statistics represent a mix of the voters’ actual preferences, how the minor parties encourage their supporters to vote via how to vote cards, and which two parties have the highest count of votes.

In Australia the Labor Party (ALP) is seen as left-leaning and the Coalition is seen as right, though these assumptions probably don’t stand up under close examination. Consequently, it has been generally assumed that the ALP has been leaking primary votes to the Greens, as the ALP has historically been left of the Coalition and there are few alternatives on that end of the Australian political spectrum. This assumption is well-supported by the two candidate preference flows.

Based on the statistics from the AEC, the Greens’ preferences flow very strongly to the ALP across the majority of electorates, with the Liberal Party only receiving a much smaller percentage of the preferences. There are very few parties where there is such a one-sided flow of two-party-preferred preferences as is seen with the Greens and the ALP. The preference flows from the Greens to the ALP also support the assumption that a lot of the voters that the Greens have gained have been lost by the ALP, or at the very least from groups who would have had the potential to become ALP voters.

So who is One Nation absorbing first preferences from of the two major parties? It is easy to assume that those who voted One Nation in the last election would put the Liberal Party ahead of the ALP on the ballot. One Nation does appear to be on the conservative side of politics and a lot of the reporting on One Nation and preferences in the upcoming Western Australian election seems to be framed by this belief.

However, this assumption is not very well supported by the data from the last federal election. The two-party-preferred preference flows from One Nation to the ALP and the Liberal Party were fairly equal in 2016, though in 2010 and 2013 there was a bias towards the Liberal Party.

Grouping the Minors

A correspondence analysis was performed to explore how the parties relate to each other using the preference flow data from the AEC website. This technique uses frequencies to determine associations within and between the rows (minor parties) and columns (major parties their preferences were allocated to) of a cross table.

The data from the 2016 federal election was processed prior to analysis and each row had the party which received the primary vote, the party that their preferences would have exhausted on, and the mean of each electorate’s total votes transferred for the pairing. Parties that received fewer than 10,000 votes overall were removed from the ‘from’ data set. The only parties used for the ‘to’ data set were those four with the most candidates across the country.

A three dimension model was chosen as it explained at least 80% of the variance seen in the data. Each of the dimensions generated explains the relationships within the table based on the data from the columns (and rows) and each specific one accounts for a certain proportion of the variance. For example, the Greens (To_GRN) contribute most to the variation explained by the first dimension, with Labor (To_ALP) and the National Party (To_NP) contributing the most to the second. The National Party (To_NP) and the Liberal Party (To_LP) contribute the most to the third.

In the plots below the columns for preferences allocated to One Nation (From_ON), Katter’s Australia Party (From_KAP), Independents (From_IND), Nick Xenophon Team (From_XEN) and the Country Liberal Party (From_CLP) from the Northern Territory were suppressed. Their positions relative to the other parties are displayed, but they are not used in calculating the placement of the points.

These plots illustrate the relationship between many of the more well supported parties and where the preferences of those who voted for them ended up. Based on how those who vote for the minor parties place their preferences you can also see how they relate to each other, for example Rise Up Australia (From_RUA) and Family First (From_FFP) voters both have very similar voting patterns when it comes to the order they place the major parties, or their willingness to follow a how to vote card, which reflects where the minor party sees itself relative to the other parties.

On dimension one, the distance between One Nation to Labor and Liberal Party is minimal. One the second dimension One Nation is closer to the Liberal Party while on the third it is closer to Labor. When compared to the location of the Greens on both of these plots, where it remains close to Labor on all three dimensions, the location of One Nation suggests that those who were voting for One Nation in the last federal election did not have a clear preference between the two major parties. The Nick Xenophon Team occupies a similar position relative to Labor and the Liberal Party, except it does skew closer to the other parties.

Based on the preference data available from the AEC, the assumption that One Nation voters are being drawn predominantly from those who would otherwise vote for the more conservative of the two major parties is not supported. At least based on the data used here, it would be fair to assume that both of the major parties are probably losing first preference votes to both One Nation and the Nick Xenophon Team.

With parties like One Nation and Nick Xenophon Team likely to be fielding more candidates over time, and the electorate’s willingness to cast their vote for an alternative when one is available, the drop in the major parties’ primary vote is highly likely to continue.