It was not my intention to publish the full by-election study results until after the federal election, as I don’t want the information of why people voted the way they did to influence the general election activities, nor skew people’s responses to the rest of the surveys. However, given the LNP performance in Longman is being used as a reason to topple a Prime Minister, I have been urged by many to hand this information over. I’m posting it here rather than sending it privately to anyone in order to be completely transparent, and I won’t be releasing anything more than what is below.
The sample in Longman wasn’t sufficient from a quantitative perspective: only 56 of the less-than-engaged Longman voters could be enticed to participate. However, even with these small numbers, the study did accurately point to a high primary vote for Susan Lamb and a Labor win (as I commented to a number of media outlets in the week before the election, and on this blog). The results below are analysis of qualitative answers, so think of this as similar to the results of 6 focus groups, interviewed 3 times: Wave 1 in June, Wave 2 was 10 days before the by-election, and the Exit Poll wave on election day. There is a skew in the sample to men and those over 50, but there are participants supporting all the more prominent parties, all age groups, and they are nicely spread around the electorate with a slightly higher participation rate from Bribie Island.
REASON FOR VOTE
The dominant reason for voting one way or the other was personal characteristics of the candidate, with honesty and the ability to trust the individual being most frequently mentioned. Refugees, immigration and pensions were the only policy positions offered as a basis of vote decision. One respondent mentioned Malcolm Turnbull, negatively. Multiple references were made to Bill Shorten, but almost all were in the context of disliking the negative campaign, e.g. “the Kill Bill stuff was nonsense”.
Dislike of the negative campaigning was mentioned as a primary reason for vote decision by a small number of respondents, and featured frequently in other comments and the specific questions about what people thought of the campaigns. In addition, there were comments that the focus on the leaders (either positive or negative campaigning) was irrelevant or not what voters wanted to be talking about, local issues and “our real lives” being preferred topics for debate. The volume of phone calls and ads was seen as annoying and a turn-off. In all, none of the campaign contact, including GetUp! efforts in Longman, had a positive influence on vote decision other than to reinforce the choice of a small number of voters. The majority of respondents said the campaign contact had no influence on their decision, and in some cases, made them question their decision to support the candidate the campaign was promoting.
Most respondents reported seeing both news media and social media content about the by-election. However, it had little influence. Through all 3 waves, most respondents said they media exposure had no impact on their vote decision, and a minority said it reinforced their decision. Social media exposure caused some to question their vote decision, but not news media.
Conversations with other people did cause people to switch votes, and question their vote decision, as well as reinforcing decisions. Most influential conversations were with family members or close friends, and a number of those influential conversations took place on social media (particularly Facebook). Similar influence of personal conversations on social media were also noted in Braddon, and with a much higher level of influence in Mayo.
The method of analysis I am using to determine causality (the cause for each vote) is a very old method called Reason Analysis, which essentially involves reading all of an individual voter’s responses across all the waves as a single story, like a case study, and identifying those factors that caused the decision. There are plenty of academics who will say this isn’t demonstrating causality because you can’t rule out every other influence: noted, but you never can in complex social decisions like voting.
Overall, there were two main causes of vote decision in Longman identified using this method of causal analysis: loyal party supporters voted for their party; others voted for the candidates, mainly on their perception of which one was more honest/could be trusted, combined with the influence of personal contact and conversation. A significant number of respondents voted against the candidate they did not trust, rather than voting for anyone. More research is required to understand whether that assessment of trust/honesty was based on what they saw of candidates in the media, or if it was the results of social influence of others saying they trusted the candidate (or thought they were a liar), and honestly I’m not sure how you would do that research accurately in anything other than a controlled experimental setting.
There was a change in vote in the final week of the campaign of PHON voters shifting to the LNP. There was earlier movement (from Wave 1 to Wave 2) from Greens and don’t know to ALP, and both PHON and LNP also picked up some undecided voters. Two participants shifted twice, both to a minor party in Wave 2, and eventually voted for a major party on election day. Analysing the influences at these shift points is more illuminating than those who don’t shift, and all were doing so (at least in whole or part) on the basis of personal candidate factors: “she’s honest”, “the best candidate”, “they lied”, etc.
WHAT DOES THIS ALL MEAN
The poor showing in Longman was not about the Liberal National Party, or the Labor Party, or Malcolm Turnbull, or Bill Shorten. It was a decision about the candidates, their personal characteristics, and which campaign was annoying them the least. From my conversations with people in Longman with good knowledge of the electorate, that is pretty standard Longman behaviour. The kidney punch to the LNP can be put down to sympathy for Susan Lamb, the sophomore surge (the boost new members get in their second election), and Trevor Ruthenberg being a poor candidate.
Nothing about voter decision is simple, and there are significant limitations to this study. However, there is nothing in the data that I’m looking at that would indicate Longman is in any way indicative of what might happen in a federal election, nor is it in any way reflective on Government performance. No respondents mentioned Peter Dutton or Tony Abbott in the exit poll wave; there were a few mentions (mainly of Abbott) in earlier waves in reference to Government performance since the 2016 election or who they want to win the next election. However, they split pretty evenly as positive and negative, e.g. “Liberal National Coalition, but I’d prefer Abbott was the leader” or “LNP need to get rid of Peter Dutton and sort the refugee issue”. On balance, there was little indication a different leader of the party would shift their vote.