Who’s on the same page? An analysis of Singapore’s political Facebook scene

Introduction

Scrutiny by tech giants such as Twitter and Facebook on pages that exhibit inauthentic social media behaviour has increased ever since the 2014 United States presidential elections. On 6 July 2020, Facebook stated that it had yet again taken down a pro-PAP page that violated its policies.

Why is this a concern? Since 2019, 62% of Singaporeans surveyed by the Reuters Institute consumed news via social media. Almost half (48%) specified that they did so on Facebook. Aside from the traditional players, such as the mainstream media, Facebook has enabled even small set-ups to propagate news and views with ease. Who runs these pages? How do they behave, and how are they different from the mainstream media? What impact do they have on our public discourse?

Some of these questions can be answered with reference to the facts. This article collates, cleans, and analyses around 34,000 Facebook posts posted from 12 such pages before 11 June 2020 that provide active commentary of current affairs to better understand:

how such pages tend to be operated
the nature of the discourse by such pages across different policy settings
the interaction (if any) between these pages

The 12 pages were selected as they can be considered non-mainstream outlets who reach their audiences primarily through social media. Some of these sites - The Independent Singapore and The Online Citizen - do have standalone sites but nonetheless interact with many of the sites studied on Facebook. These 12 pages are non-exhaustive and findings will be updated in line with studies of other pages.

Note: Fabrications About the PAP has since been taken down by Facebook on June 28. Facebook cited page behaviour in determining that the site had violated its community standards.

The main findings are:

Not all pages run the same. The timing of posts on these pages differ on two key metrics - (1) activity outside of office hours and (2) extent to which posts are scheduled
Beyond good and evil? The pages can broadly be divided into two distinct camps that are either exclusively pro or anti-Government divide when discussing domestic politics. Yet this categorisation does not hold up as well when analysing discussion of foreign affairs.
Echo chambers or sparring grounds? These 12 pages are aware of each other’s presence - especially those with larger follower bases. Some sites function exclusively as content generators while others serve to broadcast content that align with the page’s political sympathies.

An overview of the pages studied

The 12 pages differ on several metrics - number of likes, the date which they first became active, the number of posts to date as well as the location of their administrators. A summary of this can be found in the table below:

1. NOT ALL PAGES ARE RUN THE SAME

We can analyse the timing of each page’s posts over a day and within each hour. This reveals four distinct modes of operations.

In the chart above, post frequencies have been normalised (peak activity = 1) to provide a better visual comparison.

The first group which includes pages such as factcheckersg and SG Matters appear to operate exclusively during working hours while also scheduling posts throughout the day.

Posts tend to begin at 8am and stop by 8pm. On certain pages such as factcheckersg, no posts are observed before 7am and after 10pm. From a social media optimisation point of view, this may seem odd, as activity is usually higher during commuting, lunch, and after dinner hours to drive reach and reactions. It could suggest that this category of pages does not aim to maximise reach.

A sharp dip in post frequency during lunch hour - between 11am and 1pm - is also observed across sites in this category.

2 key dimensions of difference amongst pages

At the same time, these pages post at specific minutes in each hour. The regularity of posts on these pages suggests that these pages may be manned by personnel dedicated to running the page. This is because the timing posts is an added activity that page administrators have to factor in in order to structure their pipeline of posts. In the case of SG Matters, the persistent presence of posts beyond the start, mid and end-point of the hour may indicate the presence of spare capacity that monitors and responds to ad-hoc policy issues that emerge.

This similarity in operations is striking especially given the large difference in their post volumes and following.

The second cluster of pages consisting of factuallysg and Global Times Singapore tend to operate primarily during office hours but appear to have no discernable pattern in terms of post over the hour.

Like the first cluster of sites, these pages share the same features of low post volume during lunch and out-of-office hours. However, where they differ most distinctly is the post volume over the course of the hour.

The lack of a discernible post pattern suggests that the pages, while also possibly professionally managed, serve a different function from the pages identified in the first cluster. The randomness of post times over the hour could indicate their more ‘responsive’, just-in-time function of reacting to emerging political issues, or a lack of a pipeline of ready-made posts.

The third cluster consists of pages that operate beyond office hours and actively schedule posts.

These four pages - Everyday SG, fabrications about the PAP, The Online Citizen and Wake Up Singapore - continue to post sporadically at odd hours especially during times when social media is most active. It is interesting that all four pages command a significantly larger following than the pages that were in clusters 1 and 2. These pages also appear to have deliberate posting strategies and professionalisation, as evidenced by the presence of distinct posting times over the hour. This most closely mimics an optimised social media operation one might expect of a large business or organisation. Taken together, it suggests that the focus of these pages is to focus on catching eyeballs with the intent of influencing opinions to perspectives shared on the pages.

The last cluster of pages - Democracy and Human Rights in Singapore, Fabrications By the PAP, State News Singapore and The Independent Singapore - are defined by posts outside office hours and irregular posting times across the hour. The goal of such pages, while still focussed on gaining traction for their posts, have different strategies to do so. These pages, with the possible exception of The Independent Singapore given its scale, exhibits greater evidence of community-run operations. Posters may not be available during office hours, and may not be deliberately prioritising posts.

It is interesting to note that the posting activity of Fabrications By the PAP witnesses a sudden, unique uptick in activity nearing midnight.

What is clear, however, is the growing importance of social media in structuring, responding to and influencing debates on current affairs. Page activity has increased exponentially across almost all of the 12 pages analysed, perhaps due to the impending general election. However, the increase in post activity has been sharpest in pages that are most politically engaged - they are namely Democracy and Human Rights in Singapore, SG Matters, Wake Up Singapore, Fabrications Against the PAP, The Online Citizen and The Independent SG.

2. BEYOND GOOD AND EVIL?

This analysis suggests that there is some truth that pages can be broken down into “pro” or “anti” government categories. However, such a categorisation may also be overly reductive. To analyse this, periods where more than 5 of the 12 pages are engaged in discussion about the same issue for more than 7 days were selected. The events selected are as follows:

I. Domestic issues: 2016 Bukit Batok By-election
II. Foreign affair issues: Terrex Incident

I. 2016 Bukit Batok By-election

## Selecting by post_volumeSelecting by post_volume

The 2016 Bukit Batok By-election witnessed a period of sustained activity between 7 of the 12 pages. The period of interaction began in earnest with the resignation of then-Member of Parliament David Ong from the People’s Action Party (PAP) on March 12 and ended with the release of the by-election results on May 7. Other significant personalities involved in the election are Murali Pillai, the PAP’s candidate in the by-election and Chee Soon Juan, candidate for the Singapore Democratic Party (SDP).

A sentiment analysis of these three personalities will be useful in understanding the political dispositions of Facebook pages. The text was pre-processed and then analysed for the sentiments using the NRC dictionary in R. One key limitation is that the dictionary will not be able to capture emotional language that is not in the English language. Nonetheless, a manual run through of the posts (by a native Singlish speaker) indicates that this was still a useful barometer of overall emotional sentiment towards the candidate on a page.

Across all pages, there was a consistent appeal to raw emotions when reaching out to their audiences. In many instances, the use of emotions clearly mapped onto a clear “either for or against” dynamic. For instance, fear was detected as the predominant emotion when The Independent Singapore was discussing the PAP’s candidate while anger featured prominently as an emotion when analysing Fabrications against the PAP’s coverage of the Opposition candidate, Chee Soon Juan.

Shades of grey when discussing foreign affairs

However, the simple “black and white” dynamic is challenged when analysing posts by the 12 pages when dealing with matters of foreign affairs. On November 23rd 2016, Hong Kong authorities announced that they had impounded nine Terrex vehicles and other military equipment that were bound for Singapore after a military exercise in Taiwan. In analysing posts related to the Terrex Incident, keyword search was used. Here, it can be noted how the divide that structures debates on domestic issues does not hold as well. Take pages typically considered “anti”. Although pages such as The Online Citizen and Fabrications by the PAP were focused more on the government’s handling of the Incident, pages such as Wake Up Singapore expressed anger at the seizure and demanded a return of the vehicles.

3. ECHO CHAMBERS OR SPARRING GROUNDS?

Amongst the 12 pages analysed, those with greater post volumes tend to perform the function of creating original content with the exception of Wake Up Singapore. Wake Up Singapore appears to function more as a propagator of content that has been produced by another page, specifically The Online Citizen and The Independent Singapore. A future study could further investigate the purpose of having a content aggregator - is the page seeking to pursue a particular agenda or does it seek to provide insight into the reposted content?

Amongst the pages that create original content, there are four pairs of interactions that have higher than average mentions:

Fabrications Against the PAP-The Online Citizen
SG Matters-The Independent Singapore
SG Matters-The Online Citizen
The Online Citizen-The Independent Singapore

These mentions often address claims made in the posts or articles by either one of the pages in the pair. Clearly, they do not operate in silos and are keenly aware of and may even be using each other’s large audience following.

Direct mentions of each other’s sites between the top 6 pages with the most posts on Facebook

Does this relationship exist when we analyse the interaction between pages with more posts and those with less? Not really. Pages with larger follower bases make little to no mention of the smaller-scale pages. In fact, pages with smaller follower bases tend to almost exclusively interact with articles posted by The Online Citizen.

Direct mentions between the top two pages and the 4 smallest sites (excluding Global Times given its focus on foreign affairs)

This is an ever-evolving study. Any updated analysis would include the general election period in its entirety. Should you identify any pages that you think should be included, do leave a comment below, and findings will be updated accordingly (when the author has some free time!). The compiled data - in xlsx, sorry people who may not have Excel - as well as code used to scrape and analyse the posts can be found in this GitHub repository.