Tuesday, November 20, 2012

Attitudes To Attributes (Not Great News For The Greens)

Advance summary

1. Recent polling on voter views of the importance of a list of issues, and which parties are best trusted on those issues, shows that many voters regard a range of parties as having policy strengths in different areas, rather than assuming their preferred party is always right.

2. Party trust scores, when weighted by the perceived importance of issues, produce surprisingly accurate predictions of party vote share over the last two and a half years.

3. The Greens' current mediocre polling position, especially compared to before the 2010 election, is probably connected with the party neither "owning" key issues as strongly as it used to, nor being able to convince voters that those issues are important. 


This week's Essential Report contained federal polling on which out of a list of fifteen issues voters considered to be the three most important, and which of these issues voters thought were best handled by which party.  The exact form of the questions is Which are the three most important issues in deciding how you would vote at a Federal election? and Which party would you trust most to handle the following issues?

The more or less constant list of election issues polled by Essential dates from 2010.  Potential election issues come and go and in my view some of the issues listed will not be big vote-getters in 2013, while others that are not on the list will supplant them.  Especially, it's very likely that perceptions of attitudes to women will play a major role in the 2013 election, at least if both the current party leaders remain.

Although the list of issues is a little dated (and probably not meant to be too comprehensive in the first place), Essential's attribute polling is especially useful because it is three-cornered, and therefore can help obtain further insight into the current question of what is going on with the Green vote, and the age-old (and more complex than it may seem) question of why some people vote Green in the first place, while most don't.  A view I have found a lot of evidence for in the last few years is that at least a third of the Green vote at any level of politics seems to be a soft vote against the major parties - a vote that goes to the Greens by default and is easily picked off by left-wing and centrist independents. In Leadership Transition, Polling and the Greens, I examined attempts to connect a recent slight decline in Green support with the leadership change from Bob Brown to Christine Milne, and found that those attempts were inconclusive.

Looking at the latest attributes polling, what struck me immediately was the extent to which voters generally will pick across the political spectrum in deciding which party they trust most on an issue.  But this applies not only to major party voters, but to Green voters as well.

As the table as supplied by Essential contains a relatively high don't know rate, I decided it would be a useful exercise to take its list of issues, and redistribute both the "don't know" responses to the party trust question, and the Independent/Other responses to the voting intention question.  The idea was to see what could be determined about how the relative belief that each party was most trustworthy on given issues compared to that party's relative share of primary votes.  And the inevitable colour-coded spreadsheet looks like this:

Relative distribution of trust between parties (modified from Essential Research, 19 Nov 2012)
The Modified Primary row represents the proportion of all support for one of the three parties obtained by each party (ie with others/independents redistributed, an action that may slightly favour some parties over others.)  The Importance column represents the percentage of voters who placed the issue in their top three (the issues have been sorted by this figure).  For each issue, the figure for each party represents the proportion of voters who most trusted that party - out of those voters who named any party.

The following colour coding is used: pale green where the party's relative trust score exceeds its modified primary vote by at least 4 points, dark green where the party's relative trust score is more than double its modified primary vote. Orange where the party's relative trust score is at least 4 points below its modified primary vote score, red where it is less than half its modified primary vote score. Most trusted party on each issue shown in bold.

So, to the surprise of probably nobody, Labor is viewed strongly compared to its vote share on IR, education and local industry, and weakly on environment, asylum seekers, water supply, climate change and population growth.  The Coalition is viewed strongly compared to its vote share on security and war on terror, the economy, population growth, interest rates and (perhaps surprisingly) the vague criterion of "political leadership".  However the Coalition scrubs up poorly on the environment and climate change, and to a lesser degree water supply, IR and education.

The Greens differ from the major parties in that there are issues on which they are rated the best by several times, or by a small fraction of, their modified primary vote.  Few surprises here with the environment, climate change, water supply and asylum seekers out in front, poor ratings on education, Australian jobs, taxation, IR and security.  The Greens are rated particularly badly, even relative to their vote, on controlling interest rates and economic management.

These figures refute a common view of core Greens supporters as brainwashed locked-in diehards who believe their party has the answers about everything.  Clearly most Greens supporters in this sample either believe someone else has a better handle on the economy, or else are at least not convinced their party knows what it is doing in the area (or perhaps, the respondents just don't feel confident enough about economic issues to venture an opinion.)  The suggestion is that many Greens supporters are either open to ideas from other parties on many issues, or else have never thought some issues through, and are not prepared to assert "my party right or wrong" without so doing.

At the same time, the idea that voters refuse to vote Green because they don't believe in Green views on key Green political issues isn't supported by the evidence either.  Huge numbers of major party voters prefer the Greens' position on the environment and climate change, but this does not translate to votes.  The most obvious reason it does not translate is that the issues on which the Greens perform most strongly are issues that are considered important by low to modest proportions of the electorate at the moment. There is probably also a lot of inconsistency caused by ranking party performance on issues one issue at a time.  Some voters might say the Greens were best equipped to handle the environment, and the major parties best equipped to handle economic issues, without considering the interplay between the two and "costing" the damage each party caused on other issues into their views about its trustworthiness on a given issue.

The issues canvassed can be broken down based on the figures above as follows:

Green-leaning issues (these are the issues on which the Greens perform the best and the Coalition the worst, relative to vote share): environment, climate change, water supply

Blue-leaning issues (the Coalition performs best and the Greens worst): economic management, interest rates, security, taxation (weakly), "political leadership"

Centralising issues (Labor performs better than either the Greens or Coalition, relative to vote share): IR, education, protecting Australian jobs, housing affordability (weakly), health (surprisingly weakly).

Polarising issues (the Coalition and Greens both outperform Labor): asylum seekers, managing population growth.

How To Predict Vote Share Without Even Trying?

Looking at the above, I noticed that the Greens' average trust score was well above their primary vote because of the influence of a few issues on which they are especially highly trusted.  So it idly occurred to me that multiplying each party's trust share on each issue by the perceived importance figure for that issue might produce some rough idea of how much positive contribution each issue made to the party's support.  For instance for the Greens, environment with its huge trust share figure and reasonable "importance" figure comes out as the most significant.  But something like health, on which the Greens are reasonably trusted for their vote share, still becomes quite significant too because of its high weighting.

Having thus weighted each party's level of support on the issue, it occurred to me that it might be interesting to add up all the weighted-by-issue-importance trust scores and see how the distribution of the totals by party compared to each party's vote share.  The answer is: surprisingly closely. The following is a worked example of how I did this. 

For the Nov 2012 poll, the Coalition's share of the sum of the weighted totals was 49.3, Labor's was 41.3 and the Greens' was 9.6.  This compares closely with the modified primary vote scores of 50, 39.1, 10.9.  

Fluke?  Apparently not. I tried the same trick for another six combinations of the same two ER questions in the past few years (all I could find, though there may be more).  Across seven readings for each party, the modelled primary vote shares by this very crude method were on average wrong by only 1.25 points.   The method on average under-predicted the Green vote by 1.4 points, over-predicted Labor by 1.8 and under-predicted the Coalition by 0.6, and explained 97% of variation in the Labor vote, 96% of variation in the Coalition vote and 73% in the Green vote.  (For the Greens, it tends to under-predict their vote when it is high.)

(And lest it sound like this is all just a post-hoc report of something that worked without testing against another data set, I did the proper thing by informing a witness of my suspicions halfway through and thereby making one part of my data set a test of the other part.)

So Does This Mean Voting Is Largely Driven By Issue Perceptions? 

Unfortunately, it is not that simple.  While it may look like voting for the three biggest parties is incredibly strongly driven by perceptions of how well they own particular issues combined with perceptions of how important those issues are, the ancient warning that correlation isn't causation needs to be sounded loud and clear on this one.  It's also possible that when voters change their opinions of parties for reasons not driven specifically by issues (such as leadership), they become likely to hold or express different views of either the party's performance on specific issues, or the importance of issues that are dear to that party.

The Green Vote and Key Issues

The most recent reading of the Green vote according to Essential was 10.  The last time Essential conducted attribute polling while the Greens were doing significantly better was in June 2011, when the Greens were polling 12 (because of Essential's use of two-week rolling averages, a two-point difference for the Greens is much more meaningful than for other pollsters).  They also conducted it in June 2010, when the Greens were polling 13.

Compared to June 2011, there has been very little change in voter trust levels for the Greens compared to the other parties on the set of issues listed.  As a share of the total trust for all three parties, the Greens' trust scores have not changed by more than 3 points in any instance.  Furthermore, the changes tend to cancel out (eg they are down slightly in trust on the environment, and up slightly in trust on climate change.)  What has happened, however, is that the percentage of voters who rate climate change as one of their top three issues out of the issue list has dropped from 15% in June 2011 (its highest level of the seven readings) to a miserable 9% now.  At the same time, voters have become more concerned especially about education, health and the economy - issues on which the Greens are rated reasonably in one case, poorly in a second, and disastrously in the third.

The comparison to June 2010 is rather different.  At that time the Greens enjoyed much higher relative trust scores on the environment (61 compared to 48.1 now) and climate change (50.7, now down to 42.1).  Perceptions of importance for issues the Greens perform well on were also slightly better then than now, but it is the Greens' reduced ability to "own" their own key issues since the 2010 election that is the most striking feature here.

As there was only one pre-2010-election attribute poll and Essential was a relatively new force in federal polling at that stage, I don't want to read too much into it.  But this is all consistent with my suspicion that the Greens are generally struggling both for firmer ownership of their key issues and to convince voters that "their" issues are important.  

No comments:

Post a Comment