#nzvotes: The dynamics of campaign communication on Facebook

During the 2020 general election campaign, the New Zealand Social Media Study (NZSMS) has now harvested more than 3,000 Facebook posts from the Labour Party, National Party, New Zealand First, the Green Party, ACT, the Māori Party, The Opportunities Party (TOP), Advance New Zealand (which also includes the New Zealand Public Party), the New Conservatives, and their leaders.

The data at this point presents a revealing picture of the use of Facebook in the campaign. The project’s findings have been widely reported and some of the political parties have responded to the team’s research. In line with the project’s findings that Advance New Zealand was spreading misinformation around COVID-19, Facebook, having conducted its own investigation, took the unusual step of ‘deplatforming’ the party just two days before the election. In justification, a spokesperson for Facebook said it “does not allow anyone to share misinformation about COVID-19 that could lead to imminent physical harm”. Something the NZSMS team had warned about early in the campaign.

In this final post, NZSMS research leaders Professor Jack Vowles and Dr Mona Krewel from Te Herenga Waka—Victoria University of Wellington’s Political Science and International Relations programme analyse the information collected throughout the last four weeks of the campaign and show how the parties’ social media unfolded.

Did social media campaigns become more positive or more negative over time? Was the spread of false information a single event for parties or did they continuously post fake news and half-truths? What political topics were trending over the course of the campaign?

As noted, Professor Vowles’ and Dr Krewel’s research has been widely reported and observed by the political parties, which raises the intriguing prospect that the reports of their research could have affected some parties’ subsequent posting strategies. Those using negative strategies could have become less negative, those promoting fake news and half-truths could have become less prone to do so. On the other hand, changes in strategy over the campaign could also simply reflect internally driven choices.

Who has been the most active campaigner on Facebook in the 2020 election?

“We have added together all activity by the political parties and their leaders. The biggest users of Facebook were, as expected, the Labour Party and New Zealand’s social media powerhouse Jacinda Ardern,” says Dr Krewel.

“The party coming in second is a bit of a surprise: the Māori Party. The main source of this is a large number of posts from party co-leader John Tamihere, who, of all the leaders, posted the most. Next comes National, as expected, but followed very closely by Advance New Zealand, whose leader data came only from Billy Te Kahika; whereas co-leader Jamie-Lee Ross did not use Facebook during the campaign. With two co-leaders active for the Green Party, they come next in the ranking order, followed by the New Conservatives and TOP. The two least active parties on Facebook were New Zealand First and ACT.

“The major parties, of course, have the deepest pockets and can invest a lot in their social media communication to woo voters online. However, the example of the Māori Party shows social media is also a great opportunity for the minor parties to make up for a smaller budget when they have a very engaged user of social media in their ranks.”

And what about the overall tone of the campaign?

“In early posts, we already explained the reasons why some parties ‘go negative’ and others ‘go positive’,” says Professor Vowles. “Government parties are more likely to choose a positive tone in their campaign, opposition parties have good reasons to be negative and attack, as they need the media attention.

“It is therefore surprising that, until the very end of the campaign, Jacinda Ardern’s ratio of positive to negative posts was lower than that of National leader Judith Collins. However, this is largely explained by Jacinda Ardern’s high rate of neutrality of tone early in the campaign. She started into the campaign by mostly campaigning in a ‘Prime Ministerial incumbent mode’. But roughly a week before the election, she finally geared up into campaigning mode, as was apparent in the final leadership debate. Jacinda Ardern’s positivity ratio is consistently tracking upward until it exceeds that of Judith Collins. Meanwhile, Judith Collins’s posts go so negative in the last week that, at one point, her positive and negative statements go into balance.”

“Focusing on the parties, meanwhile, Labour’s posts remain strongly positive throughout the campaign, while National’s ratio starts positive but becomes less so as the campaign progresses, following Judith Collins into almost negative territory by the last few days. Informed by repeated polling showing they were making no ground, the National Party and its leader were becoming increasingly desperate and tried to attack wherever they saw a chance.”

“The Green Party stayed in positive territory throughout,” says Professor Vowles. “The Māori Party is consistently positive at a slightly lower level, trending down only slightly as the campaign goes on. The New Conservatives are less positive, again, although becoming slightly more positive over the campaign. This could either have strategic reasons, as they wanted to ensure voters make their tick for them and not only vote against their political opponents, or be a reaction to being publicly called out for negativity. We cannot know for sure.”

Two parties started negative but became more positive as the campaign went on: ACT and TOP.

“ACT’s tone shows most variation, indicating either close attention to campaign impact over the campaign or, alternatively, perhaps a more random or uncoordinated line of attack,” says Professor Vowles. “In late September, ACT’s posts were 40 percent more negative than positive, but bounced back to 40 percent more positive only a few days later.”

“The trend for New Zealand First starts positive, as one would expect from a Government party,” says Dr Krewel, “but trends downward. Over the last week, New Zealand First ends up with much the same ‘tone ratio’ as that of ACT: at about 30–40 percent positive compared with 70 percent positive for the Green Party and between 50 and 60 percent for Labour.

“New Zealand First’s declining positivity was apparent more generally. Commentators have already identified this behaviour as a possible reason for New Zealand First’s failure to make more than marginal headway as the campaign went on. As a Government party, they should have concentrated more on their achievements in office.”

But what about the strongest form of dirty campaigning: misinformation?

“Compared with the total number of posts under scrutiny (more than 3,000), half-truths (fewer than 75 posts) and ‘fake news’ (less than 20 posts) were relatively absent in the 2020 New Zealand election and were mostly concentrated in the Facebook pages of Advance New Zealand, the New Conservatives and their respective leaderships,” says Professor Vowles.

“Jacinda Ardern expressed no half-truth and Judith Collins two. As a party, Labour also did not post any half-truths, but the National Party was responsible for nine. ACT leader David Seymour posted one half-truth, his party two. New Zealand First leader Winston Peters posted also only one half-truth, his party none. The Green Party like Labour had none.

“Half-truths became less prevalent as the campaign went on, which again could be an effect of the parties and their leaders knowing a lot of fact-checkers, including our project, kept a sharp eye on them.”

Fake news was only found in the posts of Advance New Zealand (9), Billy Te Kahika (4), the New Conservatives (2), their leader, Leighton Baker (1), Judith Collins (1), and the National Party (2), says Professor Vowles. “Fake news posts also fell as the campaign went on, with a slight recovery in the last week, primarily from Billy Te Kahika.”

What political topics were trending over the course of the campaign?

“Of the slightly over 3,000 Facebook posts coded, more than 2,500 contained policy or issue content,” says Dr Krewel. “That is a lot. It means today’s campaigns are wrongly accused of being unpolitical. Modern campaign tools such as social media are not necessarily superficial and about only personalities rather than issues. However, that there is a certain number of non-political posts also shows social media is a medium that requires political leaders to be more informal and personal. But most posts are still political.

“As reported in earlier posts, the economy remained the major issue, the number of mentions dropping somewhat mid-campaign. Social and labour market issues were the second-most-mentioned issue, remaining consistent over most of the campaign, rising slightly at the end. Environmental and energy issues came next, followed by health (not including COVID-19) and transport, but none of these showed much variation over the campaign.”

And finally, did the campaigns do a good job in mobilising voters to turn out?

According to Professor Vowles, the pattern of voting returns has changed significantly over the past two or three New Zealand general elections.

“Advance voting has been increasingly facilitated and encouraged, changing the incentives shaping parties’ mobilisation efforts over the campaign. No longer is there a concentration on ‘getting out the vote’ on election day. Efforts are spread over the period of advance voting.

“However, not all parties’ Facebook posts designed to encourage going out to vote fit this pattern. Labour, National and the Māori Party show the clearest ‘step-up’ behaviour, although National’s efforts stepped up earlier and Labour only reached its peak in the final week. ACT stepped up its mobilisation efforts at the beginning of the advanced voting period but then fell back, subsequently redoubling its efforts over the last two weeks.

“The Green Party also stepped up when advanced voting started and steadily intensified its efforts until election day, although had also put in a very strong effort at the very beginning of the campaign, as did Advance New Zealand. Indeed, Advance New Zealand appears to have put almost all its efforts into mobilisation after the advent of advanced voting.”

But, says Dr Krewel, “it is New Zealand First that stands out in its apparent failure to transmit mobilisation messages during the period likely to have been most effective. New Zealand First seems to have intensified its efforts in the lead-up to advance voting but then failed to maintain the momentum, putting in the least effort of all the parties even just before election day. But as New Zealand First was a relatively low user of Facebook, this failure to concentrate its mobilisation efforts on the platform effectively may have mattered little in the end.”

CrowdTangle, a public insights tool owned and operated by Facebook, has been used to collect the data on which this commentary is based. This sample has then been coded by five human coders on the basis of CamforS/DigiWorld’s codebook.

Listen to the podcast

Pro-Vice-Chancellor of the Wellington Faculties of Humanities and Social Science and Education Professor Jennifer Windsor talks to Professor Jack Vowles and Dr Mona Krewel about the New Zealand Social Media Study.

Footnotes

  1. A post could include both negative and positive statements. Based on the valence and strength of the used statements, pictures or emotions, the post could be of a negative or positive but also balanced nature. This tendency could be derived from the overall impression of the statements, pictures and emotions included in the post. The decisive factor for coding was the impression about the valence of the statements, pictures and emotions an average reader would get after looking at the post.

    A post contained negative campaigning when it aimed at critically presenting the political opponent. This involved all forms of attack on the political opponent (party, politician, coalition, institution). Negative campaigning criticises socially relevant topics, uses stereotypical traits, highlights shortcomings as well as criticises and attacks qualities and behaviour of parties, politicians and related issues. Moreover, exaggerations and negative emotions such as fear, envy, blame and anger have also been considered negative campaigning.

    A post contained positive campaigning when it included positive statements, pictures and emotions of a supporting, encouraging, affirmative, beneficial or assertive nature and presented the advantages of a party’s own candidate, their goals and competencies.

    The graph shows the positive less the negative news. A five-day moving average has been used to smooth out short-term fluctuations and highlight the trend of the campaign period.

  2. A post contained fake news when it was completely or for the most part made up and intentionally and verifiably false to mislead voters. The usual disagreements and accusations between political actors were not coded as fake news here. If a coder assumed a post could include fake news, but was not fully sure, they were asked to do some fact-checking and visit news websites of reliable sources to see if something had already been identified as fake news. In case of doubt, coders were asked to take a conservative approach and code the absence of fake news. Therefore, the graphs presented here under- rather than overestimate the extent of fake news in the campaigns.

    When a post did not classify as fake news, coders were additionally asked if it contained half-truths—eg. things that were not completely accurate. Something coded as fake news automatically was coded as half-truths too. However, not all half-truths also qualified as fake news.

  3. A post contained a call for mobilisation when it encouraged the user to actively engage with and support the party in different ways and ultimately aimed at initiating/generating political participation. We distinguished between calls for mobilisation online and offline. Mobilisation online included all actions that take place online, like sharing a post or participating in an online petition. Mobilisation offline included all actions that take place offline, like going to cast a ballot or knock on doors.

    A five-day moving average has been used to smooth out short-term fluctuations and highlight the trend of the campaign period.