Mood Map: How FB Users In Singapore React To News On Social Media

FB has been in the news for all the reasons lately, and there are growing questions about how the platform influences public opinion, particularly on politics and contentious issues.

There’s scope for a full research project on this topic, if you can get your hands on enough data from FB. What’s publicly available won’t tell us the full picture. But it can perhaps give us a glimpse of how Singaporeans react and interact with news on FB, the dominant social media outlet here.

Charts in this post are based on 149 weeks (week of Dec 27 2015 to week of Oct 28, 2018) of FB interactions on the pages of CNA, ST and TODAY — the three dominant English news outlets in Singapore. They also have the biggest “FB fan base” among media outlets in Singapore. While I could include non-English and non-mainstream news outlets, their fan bases on FB are too small to affect the overall trend.

For context: CNA leads the field with about 2.7 million “Likes”, while ST and TODAY have 1.2 million and 751,000 “Likes” respectively. The number of “Likes” on a page doesn’t directly translate to the number of people you reach on FB, but it is one of many factors involved. A more detailed discussion is beyond the scope of this post.

Several caveats are in order before we delve into the charts:

• I sourced the data from CrowdTangle, an analytics service owned by FB. CrowdTangle only shows you a limited set of data from public FB pages published by news outlets, personalities etc. In other words, it merely helps a user gather public FB data more efficiently. I have no access to the broader set of data from personal FB accounts.
• I’ve excluded comments and shares, as they are more complex behaviour which require more data and ninja data skills to analyse(neither which I have at this point as a data noob). For instance, how do you detect and quantify something like sarcasm and snark, which dominate much of social media commentary?
• The CrowdTangle data is available only in bulk aggregate form. There is no breakdown in terms of age, gender, or geography, for instances. This is problematic as you can’t say that the publicly available data fully represents interaction from users in Singapore — the FB pages of CNA, ST and TODAY attract a considerable regional audience. So, say, if a post has 1,000 “Likes”, you wouldn’t be able to tell how many of those came from users in Singapore, or if the bulk came from readers elsewhere(unless you work in FB’s data shop). My working assumption is that any skewing of the data by non-Singaporean users gets evened-out over time (in this case a two-year period) as these three FB pages primarily target a local audience.
• There’s no reliable way (publicly at least) to measure passive behaviour on FB, say, if I merely read a post by CNA, ST or TODAY and don’t interact directly with the post. Does that mean I’ve not interacted with that post? Clearly not. But I don’t know how such data is defined or captured at FB.

Despite the severe limitations, the available data still shed some interesting insights into user behaviour in Singapore towards news. Just bear in mind that the charts here only show one aspect of how FB users in Singapore behave towards news overall.

CHART #1: The dominance of “Likes”

There’s a widely-held assumption that many people take to social media to rant or express their disapproval of the news of the day. But Chart 1 seems to dispel that notion.

The blue line for “Total Interactions” charts the total weekly number of interactions on the three FB pages — the sum of “likes”, shares, comments, as well as those who clicked to express specific responses like “angry”, “wow”, “haha”, “sad”, or “love”.

The orange line just below the blue one charts the weekly number of “likes” specifically. That mess of green, red, purple, brown and pink lines below the orange line represents the weekly numbers for those who chose to register a specific emotion(more on this later).

It is amply clear that “liking” a post is the dominant and most generic behaviour for news-related FB posts — several orders higher than other possible forms of behaviour such as sharing, comments or registering anger or sadness (see the breakdown at the end of this post).

Or put another way, those who opt to express a specific emotion when interacting with a news-related FB post are in the minority.

And among those who chose to express a specific emotion, what do you think the dominant reaction was? The answer might surprise you (see Chart #2 below).

The weeks with the highest spikes in FB interaction are self-explanatory and not the least bit surprising, as huge news events took place during those periods.

But it is interesting to note how sharp that spike in total interaction was that week of Aug 07 when swimmer Joseph Schooling won Singapore’s first Olympics gold. That level — just over 2 million in total interactions — has not been breached since. Other big news events that interest Singaporean FB users seem to hit just around the 1-million mark in total interactions.

What does this mean for newsrooms? Simply put, your managers and your teams have to learn to anticipate the spikes in interaction, and “ride” those social media surges when they are on the upswing.

It is pointless trying to pour more resources into breaking coverage when interactions are far off their peak and coming down sharply.

CHART #2: Sad!

To better visualise the patterns of those who chose to express a specific emotion when interacting with news, I’ve removed the data for total interaction and “likes”.

This chart surprised me the most. I did not think that “sadness” would be the dominant emotion — by far — among FB users who opted to register a specific response outside of the generic “like”.

The “saddest” week for FB users was an extraordinary one — barely 24 hours after Prime Minister Lee Hsien Loong suffered a brief “fainting spell” that unnerved the country on Aug 21 2016, the country’s sixth President SR Nathan died.

Meanwhile, the outpouring of online grief for Inuka, the Singapore-born polar bear, was stunning to see as well. Who knew?

Chart #3: Loving it

Charts #3 and #4 illustrate how tricky it can get when interpreting social media data, due to the ways responses on FB are designed.

The three main spikes among FB users who “loved” a particular FB news-related post closely mirror the three spikes in total interaction and “likes” seen in Chart #1.

This begs the question: Is there a real qualitative difference between “like” and “love” for ordinary FB users? If so, when does one decide between one and the other? My sense is that there is no appreciable difference for most users.

CHART #4: Lol

Humour is infamously hard to parse and decipher from a technological perspective. It is interesting to see that more users chose “haha” as an option than “angry” over the last two years when interacting with news on the FB pages of CNA, ST and TODAY. But it is far from clear what it is that FB users find particularly funny or amusing.

There were many conflicting views on Halimah Yacob becoming Singapore’s first female President, that’s for sure. But I’m not sure “haha” is the response that comes to mind first…..Ditto Dr Mahathir Mohamad returning to power with a historic election win. Huge news, but why would one react with “haha”?

Maybe this is just a prank by FB’s product team and the joke’s on those of us who take this option literally or seriously.

CHART #5: Triggered

Anger, as a specific response to news on the three FB pages, registered the second lowest set of interaction spikes overall.

This is a good place to trot out that classic line about “absence of evidence is not evidence of absence”: the FB data does not show the actual level of societal anger, or lack of, among people who had read or reacted to a particular piece of news.

The chart merely gives a sense of the proportion of people who chose to react in this specific way given the options on FB.

That said, it is interesting to see that more FB users were “triggered”

by news of Halimah Yacob becoming the President than news of the verdict on the appeal of the City Harvest Church case.

FINAL THOUGHTS:

This is my second (and still rudimentary) attempt at applying some new data skills. As they say, you learn best with real data and problems that you are trying to solve.

While basic, I think this approach can be applied to FB news pages across different regions to get a sense of how users in those areas interact with news-related posts. I don’t know if this can be improved to include a “predictive” element, but let’s see what future lessons in data science hold.

Finally, a breakdown of some key data nuggets used in this exercise:

• Time Frame: Week of Dec 27 2015 to week of Oct 28, 2018 (149 weeks)
• Total number of Interactions: 86,874,680
• Likes: 45,399,049
• Shares: 21,395,145
• Comments: 5,872,141
• Sad: 3,544,478
• Wow: 3,354,195
• Haha: 3,001,952
• Angry: 2,347,336
• Love: 1,947,013
• FB pages tracked: CNA, ST, TODAY

Errors are all mine, naturally.