Has Australian politics become more polarised? A response to Lachlan Harris and Andrew Charlton

Is the “fundamental operating model of Australian politics breaking down?” That’s the dramatic question that Lachlan Harris and Andrew Charlton asked in their recent joint piece in the SMH, and answered with a emphatic ‘yes’. Among other things, Harris and Charlton claim that Australian politics has become more ideologically polarised over time, with both voters and candidates drifting away from the political centre. That’s plausible  – polarisation is a much-discussed phenomenon in the US – but I’m unconvinced by the data they put forward to support this claim. In this post I’ll consider one part of their argument,  that major party candidates have become more ideologically extreme over time.

Exhibit A in Harris and Charlton’s case that politics has become more polarised is this chart, which uses data from the Australian Candidate Study to suggest that fewer major party candidates are ‘moderate’ than used to be the case:
On the face of it, this looks pretty convincing. More than one-in-three major party candidates used to be a ‘moderate’; now it’s down to around one-in-ten. But I have three issues with this:

  1. The trend appears less stark if all the election years covered by the data are included;
  2. The definition of ‘moderate’ is questionable, and the trend appears much less stark if a more intuitive definition is used; and
  3. Declining response rates for the ACS mean we should be cautious about inferring anything about the political views of candidates based on responses to the survey.

First, the inclusion of other years in the survey. Harris & Charlton included only 1996, 2007, and the two most recent elections in their chart. Looking at those four elections, the downward trend looks monotonic – every election has seen a smaller proportion of moderates than the one before. But if we include other elections covered by the Australian Candidate Study, the picture is less clear. The trend still appears to be downwards, but 1996 looks much more like an outlier. The 2010 election featured about the same proportion of moderates as the 1993 or 2001 elections, and the 2013 figure was only a little bit lower. It’s true that the proportion of moderates was appreciably down in 2016, but reading too much into that result in the context of a declining response rate would be dangerous (we’ll come back to this).

My second quibble is with the definition of ‘moderate’ that Harris and Charlton use. They make the following claim:

In 1996 more than one in three Australian politicians (37 per cent) rated themselves as “moderate” – that is, centre-left Liberal and centre-right Labor politicians. This share has shrunk dramatically. At the most recent federal election in 2016 only one in 10 politicians described themselves as moderate.

But the Australian Candidate Study doesn’t ask politicians whether they describe themselves as ‘moderate’. Instead, it asks political candidates to place their own views on a left-right spectrum that runs from 0 to 10. Harris & Charlton have devised their own way of classifying candidates as ‘moderate’ or not based on this 0-10 scale. Harris & Charlton classify ALP candidates as ‘moderate’ only if they rate a 5 or higher on the left-right spectrum; members of the Liberal and National Parties qualify as ‘moderate’ only if they’re a 5 or lower.

But does this really accord with what people understand the term ‘moderate’ to mean in a political context? If an ALP member was very right-wing – a 10 on the 0-10 scale – then they’d qualify as a ‘moderate’ by the Harris & Charlton definition. Similarly, if an LNP member was on the far left, then he or she would qualify as ‘moderate’ in the Harris & Charlton classification. A person whose self-described ideology is far from the political centre qualifies as moderate, so long as their ideology is contrary to that of the bulk of their own party. But a centre-left ALP candidate, or a centre-right LNP candidate, wouldn’t qualify as moderate. To me, this classification doesn’t identify ‘moderates’ so much as it does ‘mavericks’ – those candidates whose views are at odds with their party.

In my view, a political moderate is someone whose views are close to the political centre. So, on the 0-10 scale, moderates would be people who rate a 4, 5, or 6 from either party. ALP candidates who are a ‘4’ qualify as moderate in my definition, as do LNP candidates who place themselves at 6. The chart below shows what the trend looks like if we use this definition of ‘moderate’.

Before 2016, you can make out a slight downward trend if you squint. But, even with 2016 included, the decline in ‘moderates’ is much less striking using this definition than the Harris & Charlton definition. Under the Harris & Charlton definition, the 2016 election featured only a third as many major party moderates as the 1996 election. But if we define moderates to mean ‘candidates in the middle of the political spectrum’, as in the graph above, the picture is much less alarming.

We can set aside the quibbling over the definition of ‘moderate’ and have a look at the full distribution of candidate respondents’ political views, as in the graph below. In every single election from 2001 to 2016 (inclusive), the median ALP respondent to the Australian Candidate Study placed themselves as a ‘3’ on the 0-10 left-right spectrum; the median Liberal or National Party candidate described themselves as a ‘7’. There no change in the middle of the distribution of either party.

Harris and Charlton are of the view that the parties are moving away from the median voter:

The political implications of increasing voter polarisation are profound: the fundamental operating model of Australian politics is breaking down. That operating model was based on the “median voter principle”, which assumes that the electorate is a continuum from conservatives on one side to progressives on the other. The combination of compulsory voting and a two-party system meant that the strategy to win the most votes in Australian politics was always to position your party as close to the median voter as possible.

But, in my view, the ACS data don’t really support this idea that the parties are moving away from the centre. At every election since 1996, the median ALP candidate has been a ‘3’ on the 0-10 ideological spectrum; since 2001 the median Coalition candidate has been a ‘7’.

Finally, I think we need to apply a few more grains of salt to the data. The Australian Candidate Study received 99 responses from major party (ALP + Coalition) candidates in the 2016 election. Of those, only 74 provided a response to the question about their position on the left-right scale. 14 Liberal candidates answered that question, along with 4 National Party candidates. That’s a small sample from which to draw inferences about the whole population of candidates. In fact, the codebook provided by the survey’s organisers contains a disclaimer to say that the responses to questions “cannot be interpreted as summary statistics of the population of interest.” We can’t assume that the handful of major party candidates who answered the survey, particularly from the Coalition side, are representative of the full range of major party candidates. The decline in ‘centrist’ LNP candidates in 2016, seen in the graph above, might reflect a general decline in moderate representatives, or it might reflect the declining response rate to the survey, particularly among Liberal and National candidates.

Harris & Charlton make a lot of important observations in their piece. They show that people are less satisfied with democracy, and voters’ trust in politicians has declined. These trends are important and troubling. But I find their claims about the ideological polarisation of the major parties to be unconvincing.

Thanks to Dr Shaun Ratcliff for commenting on a draft of this blog post.