8 Take-Home Points from the Malta Election 2017 Big Data Project
Steering away from traditional polling methods, big data analysis provided an innovative perspective to understanding social sentiment during Malta’s election run-up. Now that the results are out, what did ICON and Minely’s big data project teach us about online sentiment analysis? Did it reflect [or predict] the actual results?
1. Average Sentiment is NOT Enough
The first and most important finding from this big data initiative is that it is crucial to understand that focusing on ‘average sentiment’ alone is simply not enough. While the sentiment may provide insights into the effectiveness of the campaign narrative and its ability to win ‘hearts and minds’, it does not correlate directly to voter intention. Indeed, the analysis of the interactions and activity of unique active users can give much more accurate results. ‘Unique’ is the key word here.
The first graph below demonstrates Active Users on social channels attributed to PL and PN. This visual shows quite a large discrepancy between the active users for Partit Laburista (PL) and Partit Nazzjonalista (PN) which discrepancy was maintained through most of the campaign period, and ultimately in the final results.
When reviewing this data in terms of Active Users in percentage terms and over a grouped month, this second chart immediately reveals a significant PL user base. Users who routinely and equally like content from both PN and PL are considered ‘unknowns’ (marked in black in the graph below). After eliminating such ‘unknowns’, data shows PL at 53.8% and PN at 46.2%. This is rather close to the eventual election result which placed PL at 55.04% and the PN at 43.68%.
A quick glance at the daily average sentiment score (graph below), on the other hand, paints a different picture. Here it can be seen that PN’s average sentiment score was ahead of PL’s, especially towards the end of the campaign. Whilst this may have potentially been the case, positive general engagement with social media posts should not be prioritised above (or confused for) vote-winning.
2. How Influential is Social Media in Malta?
It is important to understand the influence that social media and digital platforms have on Maltese society and, more specifically, on voters during an electoral campaign. Although digital is increasingly becoming part of day to day life, the whole spectrum of media channels must be kept in mind.
This is especially so during an electoral campaign where many different methods of communication are used. These multi-channel strategies cover a range of digital and offline campaigns such as email marketing, print advertising, direct mail, television advertising, mass meetings and home visits.
Digital does indeed play a large role, however, it raises the question; is the public still resorting to offline communication as the credible resource to educate and inform themselves about electoral campaigns?
An interesting study carried out by the Malta Independent on Sunday 4th June 2017, stated that 77% of survey respondents followed the election campaign on television whilst 53% followed on social media. Furthermore, respondents who leant towards the PL were more likely to follow the campaign on television whilst those more inclined to vote for the PN were more likely to follow the campaign online. This discovery has significant demographic ramifications and consequences.
Furthermore, respondents who leant towards the PL were more likely to follow the campaign on television whilst those more inclined to vote for the PN were more likely to follow the campaign online. This discovery has significant demographic ramifications and consequences.
3. What About Your Inbox?
Our analysis did not cover email communication whatsoever. It’s worth noting that this remains a strong tool in Malta and may also have served as a major communication and activism channel during the campaign.
A deeper analysis would be required to determine whether different sender names, sending times and formats stimulated higher open rates and ultimately more successful conversions.
4. What Works For One…
What works well for one political party may not necessarily work well for another. For example, a short personal video of Simon Busuttil summarising what he announced during a mass meeting was highly effective and received very high engagement on social media by his supporters.
On the other hand, Joseph Muscat’s longer videos of his speeches at mass meetings fared much better than his short ones. This may reflect the different follower personas on the candidates’ social media channels which will lead us to our next point.
5. One Size Does NOT Fit All
Today’s marketing and online communications have moved well away from the one-size-fits-all approach. In fact, audience targeting and personalisation now play a significant role in modern marketing strategies. Whether it’s targeting people via their location, their previous online behaviour and so on, today’s sophisticated campaigns analyse the individual user in minute detail. Using this method in an electoral campaign would enable political parties to create tailor-made content for particular demographics.
By providing the right information at the right time, and taking voters’ needs and preferences into account, they could potentially influence undecided voters. The PN and PL campaigns seemed to have used only basic targeting (age, gender and locality) to reach their audiences. Whilst this represents the traditional way of understanding social stratification, it may be less accurate in today’s complex society. We believe that behaviour-based groupings may be accurate. Therefore, two completely different demographics such as pensioners and youths may have the same liberal preferences and should thus experience the same ads and content.
We believe that behaviour-based groupings may be accurate. Therefore, two completely different demographics such as pensioners and youths may have the same liberal preferences and should thus experience the same ads and content.
This also applies to online advertising, such as Google PPC and remarketing. The two parties seemed to have focused more of the media maxim of repetition to stimulate top-of-mind recollection, rather than using the highly specific targeting methods offered by digital tools.
SEO tactics were also given secondary importance in both campaigns. The parties seemed to rely on their pre-existing popularity and renowned names to drive traffic to their sites. However, both parties missed an opportunity to position themselves well on Search Engine Results Pages (SERPs) for issue-based keywords. The mainstream media successfully claimed that space instead.
6. Little Campaign Integration
Although both parties were very much present online, there was little campaign integration. Instead, the channels were used in silos. For example, a great emphasis was placed on social media activity, however very few of the Facebook posts linked to the parties’ websites or apps, or encouraged users to submit their email address for updates. Directing traffic to their websites wasn’t as high a priority as getting the message across. This is commonplace during shorter campaigns which put pressure on campaign leaders to bet on merely the main action(s) necessary to win.
7. Tracking Codes
While it was not possible to detect all the tracking codes embedded in the source code of both party sites, it’s likely that there has been a degree of user-tracking to provide both parties with analytical insights. It would be interesting to deliver insights into whether this tracking proved useful in message-formulation.
8. Will Social Media Replace Traditional Media?
As Facebook, or any social media platform for that matter, does not create any content itself; it will never be able to replace independent and investigative journalism which produces good quality, verified content. Social media is, however, a channel of information aggregation and a medium in which to effectively distribute content. Its use as a channel for consumption has proven that the mediums are changing and that users will obtain information in a way that is most convenient to them.
Although both parties were highly active online, the political campaigns did not present anything specifically novel in terms of digital campaigning.
As mentioned above, this may be because the Maltese population seems to still interact principally with offline data sources (see here), or perhaps because the campaign was so short it left its digital protagonists breathless. Nevertheless, we have seen some good traction in terms of the employment of Facebook-Live as a method of political communication, and the digital democracy model (‘ideat’ platform) employed by the PN to attract crowd-sourced involvement. There is tremendous power in opening citizen access to politics which, as we have seen in other industries, may have a transformative approach.
Social Media is not just a range of channels for a campaign to amplify existing communication. It provides invaluable data that can influence the formulation and execution of the campaign strategy. Currently, these channels are being used to repeat a rehashed version of the central message, however, a more sophisticated campaign allows for the generation of psychometric variables to correctly profile and target users with the right message at the right time. By doing so, all social media channels become, not only a broadcasting medium but a listening tool as well, thus shaping and expanding the campaign’s strategy.
By combining such data points from online users, political parties will be able to predict with some accuracy what type of message would be the most effective for each person. This, in turn, should move away from broadcasting one national campaign and instead will lead to the creation of multiple sub-campaigns tailored around the expertise and intelligence sourced from public social data.
Would you like to find out how ICON can use sentiment analysis, big data and machine learning to assist your business?