Love, Like or Angry in Times of COVID-19? Analysing News Brands’ Audience Engagement on Facebook Amidst a Pandemic

Abstract

As an integral part of their online strategies and business models, news outlets diffuse their online content on social media platforms such as Facebook to increase traffic. They thereby succumb to the contingencies and constraints of third platforms infamous for their sudden changes in algorithms. In this article, we assess engagement patterns of 140,359 Facebook posts of 17 Belgian news brands between March 2020 and 2021. We map out differences in audience engagement of news outlets’ Facebook posts related and unrelated to the COVID-19 pandemic and differences between mainstream and alternative news outlets. We find that COVID-19-related posts generate more engagement and more so for mainstream media than for alternative media outlets.

1. Introduction

The rise of social media platforms has dramatically disrupted our relationship with media content in general and news content specifically. The 2021 Digital News Report was the first to rank social media over search engines and direct traffic as primary gateways to online news, with a much more pronounced difference among its respondents under the age of 35 (Newman et al. 2021). Around the world, private and public service media alike invest time and resources in building and maintaining active presences across the wide array of social media platforms currently available. Structural factors affecting all media corporations lead to similarities in business models and strategies, with social media platforms frequently perceived as ‘frenemies’, so both as friends as enemies (Sehl et al. 2021).

While being highly dependent on Facebook for online traffic and audience engagement (Hendrickx et al. 2023; Opgenhaffen and Hendrickx 2023), using it as an additional content-diffusion platform comes with a hefty price. News titles necessarily succumb to the contingencies and constraints of third platforms infamous for their sudden changes in algorithms and preferred content types (Nieborg and Poell 2018). This also means that news titles, dependent on social media, adopt distinct social media logics in diffusing news content to attract wide audiences (van Dijck and Poell 2013). For instance, their Facebook posts share many of the characteristics and lay-out of those posted by individuals and all other types of organisations or groups and effectively compete with them in order to increase engagement.

News media used to be delivered to people in easily discernible packages consumed habitually (Steensen and Westlund 2020). In the contemporary digital media era, they have to vie for ‘eyeballs’ or attention, particularly on social media amidst personalised timelines and different posts and media content originating from a host of different actors (Nixon 2020). Citizens still have some agency in relating to social media posts that even can reinforce certain pre-programmed steps of the platform (such as a like that gives the message more visibility) or change them (such as Facebook that had to change its policy around clickbait headlines because users started to distrust them and clicked less on them) (van Dijck and Poell 2013). For Facebook specifically, these include sharing or commenting on a post, or liking it through a collection of seven options, visualised in Figure 1 below. The traditional ‘Like’ button was compounded with other reaction types in 2016, which “allows users to provide emotive feedback, and allows for the collection of ever-more granular data from those users” (Sturm Wilkerson et al. 2021, p. 1043). It is noteworthy that the ‘Care’ like option was introduced only in April 2020 at the height of the first series of lockdowns following the COVID-19 pandemic, as it was explicitly intended to make “people feel a bit more connected with their friends and family during the pandemic” (Lyles 2020).

In this paper, we assess Facebook users’ engagement of a vast set of news posts (n = 140,359) published by 17 different Belgian news outlets between 1 March 2020 and 2021. We differentiate the different reaction types and set these off against different types of news topics (COVID-19 vs. non-COVID-19 news) and different types of news outlets (mainstream vs. alternative (see for instance Hiaeshutter-Rice and Weeks 2021). This multifaceted approach enables us to contribute to scholarship by offering empirical evidence on how people perceive and engage with online news from an array of different brands in unprecedented times of crisis. Our study allows us to shine light on how Flemish mainstream and alternative news media reported on COVID-19 on Facebook in the first year of the pandemic and how audiences engaged differently with the outlets’ Facebook posts.

2. Literature Review

2.1. News Brands and Audience Engagement: It’s Complicated

Engagement plays a crucial role in the conceptual Audience Dimensions framework of Philip Napoli. He argues that attentiveness, exposure, and loyalty are the key drivers of engagement, which through the degree of appreciation and emotion has the power to alter audiences’ behaviour (Napoli 2011). The specific relationship between audience engagement and news brands has been discussed among scholars for decades, though it has re-emerged as an interesting and intricate topic of debate in the light of social media platforms as new diffusion and consumption platforms (Ørmen 2016). He argued to better understand “how people attend to information about issues of public concern, become aware of the intricacies of these issues, and address each other about such issues” (Ørmen 2016, p. 18). Nelson later built on this definition and distinguished reception-oriented and production-oriented audience engagement, in which the former is more appropriate to our study as it “refers to the ways that audiences attend to the news” (Nelson 2021, p. 2357). Zayani (2021, p. 25) held the view that digitality and audience engagement are “the hallmark” of the current journalistic field. He operationalised this in his study of the Al Jazeera Network’s digital initiative AJ+ by collecting data of social media posts’ reactions, comments and shares as the key drivers of audience engagement on social media. Zayani’s approach ties in with Ørmen’s definition of audience engagement and news, as well as with our own conceptual and methodological frameworks as we too seek to operationalise audience engagement with news through an overt social media lens. As news consumption becomes increasingly steered by social media platforms (Newman et al. 2021) and the pandemic has been argued to only fuel this (Quandt and Wahl-Jorgensen 2021), we venture that it is only appropriate that research designs adapt to changing news consumption patterns. To complement this, we echo Krumsvik (2018), who notes how user engagement has become increasingly important for news organisations in recent years as user metrics including numbers of followers and clicks help “to ensure media’s social role”.

Chen and Pain (2021) have also looked at the interplay between audience engagement and social media. Based on a different survey of 588 Americans who have used Facebook to get news, they distinguished two types of social media news engagement: content interaction (interacting with content through commenting, sharing, quoting, or posting feelings about it) and exposure engagement (users’ measured attention for and positive emotions towards news on social media). The same survey was used for a different study which found that Facebook users are less likely to engage in news that people consider as controversial or makes them angry or anxious (Chen 2020, pp. 848–50). In our study, we solely focus on content interaction exposure through an assessment of over 140,000 Facebook news posts and users’ varying degrees of interacting with them.

Our study period covers the first year of the COVID-19 pandemic, the reporting of which has become a prime example of disaster communication (Perreault and Perreault 2021). It also relates to what Vasterman (2005) calls ‘media hype’, where a key event (e.g., in our case, the first death by COVID-19 or first vaccine administered) is followed by similar events, generating more and more media attention. News about COVID-19 often also had a scientific angle; just think of the many statistics about the number of infections or the scientific publications that (whether after going through a peer review) were used as a source to prove the effectiveness of the vaccines. Typical of this type of reporting is that a growing interest from the public leads to an increased media coverage that then ends up in a so-called hype pipeline (Caulfield and Condit 2012). Under the pressure of various actors and factors, scientific news can become more and more sensational, generating additional engagement, and thus, making it extra interesting to be covered, resulting in even more pressure. By comparing news about the pandemic with other news, we aim to effectively understand if and how the public exhibits a different form of engagement towards these two types of news stories. We venture that news related to the pandemic is more prone to making people feel more anxious, sad, or angry, exposes them to other types of news content, and makes them engage with it accordingly. In this regard, we seek to use our study to answer how people experience news in these times, which was one of the questions raised by Lewis (2020) in his commentary on journalism research during and beyond COVID-19. To strengthen our contribution, we deliberately offset news content that is and is not related to the pandemic to further shine light on how Facebook users engaged differently with these distinct types of news in the pandemic’s first year.

2.2. News Brands and Facebook: In an Open Relationship

The detrimental consequences of Facebook as a news diffusion platform and a genuine news source have been frequently discussed in scholarship from a host of different disciplines. Just like traditional newspapers or news outlets’ websites, social media platforms too ‘can provide individuals with information on pertinent political and social affairs that can engender further expression and participation’ (Chan 2016, p. 436). But from its inception, a key difference in online news consumption, be it via search engines, news aggregators or later social media platforms, is that citizens tend to come across news content unintentionally and incidentally without explicitly wanting to be informed (Matthes et al. 2020). On social media, news-related posts appear at ostensibly random places between personal status updates of Facebook friends in algorithmically personalised timelines. This means that social media platforms can indeed attract previously unserved audiences to news brands, but the latter also risks a dilution of their reputation, as brand awareness and loyalty matter to the business of journalism (Chen and Pain 2021, p. 367). They employ a range of tactics to maximise audience engagement with their Facebook news posts. These include carefully selecting the post time in what may appear as a linear television scheduling structure including a designated ‘prime time’ slot, choosing which news topics to post about, and using message vividness and interactivity, by using pictures and/or videos, using hyperlinks or tagging organisations or people (Guo and Sun 2020).

Based on a qualitative study of four Norwegian news brands, Kalsnes and Larsson (2018) inferred that ‘softer’ news topics are more prone to being shared on Facebook and Twitter than ‘harder’ news articles. Traditional news values are a suitable starting point for discussing and predicting which types of articles are being shared on platforms such as Facebook, though not all to the same extent. Various studies, which have looked at news content shared on Facebook from a host of countries, have shown that news presented with an emotional, negative tone are more prone to being shared (or liked or commented on) than news packaged more positively (examples include the works of Al-Rawi 2020; Robertson et al. 2023; Salgado and Bobba 2019; Sturm Wilkerson et al. 2021; Trilling et al. 2017). This ties in with the later finding of Lamot (2021) that mainstream news brands tend to ‘soften’ their news offering on Facebook as opposed to that of their own websites. The different types of reactions attributed to posts (e.g., ‘Like’, ‘Love’, ‘Wow’, ‘Angry’, …) have been linked to the tone or sentiment displayed in the post, as well as the salience or prominence attributed by the individual to the given issue or topic. Both the ‘Love’ and ‘Angry’ likes are explicitly linked to sentiment, while salience could only be linked to ‘Angry’ (Eberl et al. 2020).

Ferrer-Conill et al. (2021) analysed over a million posts from 482 different Scandinavian news brands. They sought to find differences in audience engagement between countries and type of media ownership, predominantly centred around commercial and state-owned outlets. The authors’ analysis spawned two interesting findings contrary to their hypotheses. First, the country with the lowest number of news posts, Denmark, had by far the highest number of average engagements in all three distinguished types (liking, sharing or commenting) per post, signifying that quality still trumps quantity. Second, news posts from public service media outperformed those of their private competitors in all three types of Facebook audience engagement. Their study is akin to the one at hand, yet we wish to expand their methodological framework in several ways. We also take into account the various like options as described earlier (Sturm Wilkerson et al. 2021) and operationalise these to more specifically pinpoint alterations in audience engagement with and attitudes towards news brands, including differences between mainstream and alternative news outlets, which have been defined to “represent a proclaimed and/or (self-) perceived corrective, opposing the overall tendency of public discourse emanating from what is perceived as the dominant mainstream media in a given system” and are gaining traction in media markets across the world (Holt et al. 2019, p. 862). According to the same authors, alternative news media set themselves apart by different content types and styles, and publishing routines and relationships with the societal system. Thereby, they too engage in “system-preserving tendencies” to, as overtly non-conformist minorities, maintain distance from their mainstream counterparts, including frequently criticising them and using this critique as foundation for their reporting and/or very existence (ibid., p. 864).

Scholarship has indicated that alternative, ideological news sources thrive on Facebook in terms of audience engagement, more so than posts of mainstream news outlets. Social media platforms were found to contribute to amplifying non-mainstream voices and opinions, notably politically conservative ones (Hiaeshutter-Rice and Weeks 2021; Larsson 2019). We seek to further explore the mainstream/alternative news brand conundrum in our own study of Belgian Facebook news brands’ posts and their degrees of audience engagement.

Due to the relative newness of the pandemic in terms of academic research, very few studies have appeared studying its relationship with audience engagement on Facebook at the time of submitting the paper at hand. We consider this as one of our main contributions to the state of the art. Nonetheless, we highlight one example of preceding work relevant to our study. In a non-peer-reviewed article, Oliveira et al. (2021) assessed 61,532 news headlines as posted on Facebook by Portugal’s biggest mainstream news brand between January and December 2020. The share of COVID-19-related posts fluctuated throughout the year, as did the different types of likes attributed by audiences to posts. The ‘Angry’ and ‘Sad’ reactions were more predominant in the weeks where local authorities (re)instated a state of emergency and lockdowns, whereas in general the ‘Angry’, ‘Sad’ and ‘Wow’ likes increased along with the rise of pandemic-related Facebook news headlines. We use these findings to hypothesise similar evolutions in our own data analysis by differentiating four quarters that are each representative for a different period in how the pandemic was dealt with and reported on. This will be further outlined in our methodology section. In this study, we distinguish the three forms of Facebook engagement (different types of reactions, shares and comments) from each other as they already represent different aspects of engagement. A like reaction, for example, is given the quickest, since it requires less commitment and less involvement compared to a comment or a share (e.g., Kim and Yang 2017). Reactions can be seen as “one-click communication” while posting a comment is an example of “composed communication” (Burke and Kraut 2016), which requires more effort. A share differs from the other two types of engagement in that the post ends up on your own timeline and is also visible to your network, shaping your own profile and identity (Kim and Yang 2017).

In summary, the literature review finds that the relationship between news brands and audience engagements has changed profoundly due to digitisation and social media platforms, and this was only exacerbated following the outbreak of the COVID-19 pandemic. At the same time, scholarship has argued that said platforms, with a predominant focus on Facebook, amplify both soft and negative news as well as conservative opinions. As we seek to gauge how news related to the pandemic was engaged with compared with other news and to what extent there are clear differences between Facebook posts from mainstream and so-called alternative news outlets, this leads us to the following two research questions:

• What is the difference in the number of (different types of) likes, shares and comments between Facebook posts about COVID-19 versus other topics?
• What is the difference in the number of (different types of) likes, shares and comments between Facebook posts of mainstream versus alternative news outlets?

3. Materials and Methods

3.1. The Flemish Media Market

Belgium has been classified in the democratic corporatist media system as well as the Western (liberal)-type (Brüggemann et al. 2014; Hallin and Mancini 2004). Like many Western and Northern European countries, it is marked by a tradition of quality journalism, a strong public service broadcaster and recent waves of increased media market consolidation. Belgium is also a multilingual country with various political layers and governments at the axes of sociocultural and linguistic barriers. Flanders, Belgium’s Dutch-speaking and affluent northern region of just under seven million inhabitants, boasts its own government with distinct media regulatory powers as well as its own media market and public service broadcaster (Picone and Donders 2020). Like in several other media markets, alternative news outlets have emerged in the last few years (Buyens and Van Aelst 2021). As shared ownership and consumption is low across the various Belgian media markets, we solely focus on Dutch-speaking news brands intended to inform Flemish citizens. We collected data from the 17 biggest Flemish news titles, consisting of 8 traditional daily newspaper and magazine titles, 1 private and 1 public television news broadcaster and 7 online-only or so-called alternative news titles, of which 3 are distinctly left-wing oriented and 2 are skewed to the right-wing. The news titles were selected due to their dominance in the Flemish media landscape in terms of circulation figures, viewer ratings, website visitor numbers and social media traffic following prior analyses carried out by the authors (Hendrickx et al. 2023).

3.2. Assembling Our Data Set

We used Crowdtangle for data collection, a Facebook-owned tool that tracks interactions on public content from Facebook pages and groups and has been used by publishers, content creators, and fact-checkers, and also more and more by academics who use the tool to access Facebook data of, for example, news pages, including the number of likes, shares and comments a news post generated. There are some drawbacks to the tool, such as the fact that it is owned by Facebook itself and forms a kind of black box that barely gives us as researchers insight into how accurate the retrieved data is, but nonetheless, Crowdtangle is considered a useful tool for conducting academic research on Facebook. We collected Facebook data from a series of legacy and alternative Flemish news outlets between 1 March 2020, which is seen as the beginning of the COVID-19 pandemic in Belgium, and 1 March 2021, when the vaccination campaign had begun, thus covering the whole first year of the pandemic. Specifically, we used this tool to collect all Facebook posts from the 17 most well-known news media in Flanders (Dutch-speaking part of Belgium) for one year (see Table 1 for list of brands) and the number of interactions (i.e., the likes, shares and comments) these news posts generated.

Table 1. News brands and classification.

News Brand	Classification
Apache	Alternative, left-wing
Business AM	Alternative, right-wing
De Morgen	Mainstream, newspaper
De Standaard	Mainstream, newspaper
De Tijd	Mainstream, newspaper
De Wereld Morgen	Alternative, left-wing
Doorbraak	Alternative, right-wing
Gazet van Antwerpen	Mainstream, newspaper
Het Belang van Limburg	Mainstream, newspaper
Het Laatste Nieuws	Mainstream, newspaper
Het Nieuwsblad	Mainstream, newspaper
Knack	Mainstream, magazine
MO*	Alternative, left-wing
Newsmonkey	Alternative, right-wing
PAL NWS	Alternative, right-wing
VRT	Mainstream, public television and radio
VTM	Mainstream, private television and radio

We distinguished the Facebook posts of news outlets dealing with and not dealing with COVID-19 using an automated dictionary approach with a self-created topic list (see

Table 2. Keywords in Dutch and translated to English. Note that the terms overlegcomité and veiligheidsraad are both used interchangeably for the meetings of various Belgian governments to decide on relaxing or tightening existing COVID-19 measures.

News Brand	Classification
1.5 m	1.5 m
Afstand	distance
Besmettingen	infections
bubbel	bubble
corona	-
coronapandemie	COVID-19 pandemic
coronavirus	-
COVID	-
COVID-19	-
ICU	-
intensive zorg	intensive care
intentieve zorgen	intensive care
lockdown	-
maatregelen	measures
mondkapje	face mask
mondmasker	face mask
opnames	hospitalisations
Overlegcomité	conciliation committee
overlijdens	deaths
pandemie	pandemic
SARS	-
vaccin	vaccine
vaccinatie	vaccination
variant	-
Veiligheidsraad	safety council
versoepelingen	relaxations
verstrengingen	hardening, escalations
virologen	virologists
viroloog	virologist
virus	-
ziekenhuisopnames	hospitalisations

for the full list).

Table 1. News brands and classification.

News Brand	Classification
Apache	Alternative, left-wing
Business AM	Alternative, right-wing
De Morgen	Mainstream, newspaper
De Standaard	Mainstream, newspaper
De Tijd	Mainstream, newspaper
De Wereld Morgen	Alternative, left-wing
Doorbraak	Alternative, right-wing
Gazet van Antwerpen	Mainstream, newspaper
Het Belang van Limburg	Mainstream, newspaper
Het Laatste Nieuws	Mainstream, newspaper
Het Nieuwsblad	Mainstream, newspaper
Knack	Mainstream, magazine
MO*	Alternative, left-wing
Newsmonkey	Alternative, right-wing
PAL NWS	Alternative, right-wing
VRT	Mainstream, public television and radio
VTM	Mainstream, private television and radio

We distinguished the Facebook posts of news outlets dealing with and not dealing with COVID-19 using an automated dictionary approach with a self-created topic list (see Table 2 for the full list).

To assemble this topic list, we browsed articles and made a list of all frequently used or specific COVID-19 words used by journalists. After considering which words would be used regularly and mostly in COVID-19-related news coverage, we ended up with a list of 31 keywords. Using a dictionary-based approach, Facebook messages containing any of these words were retained. Both keywords and the message text, article title and first sentences of the article visible in the Facebook post, were transformed to lowercase to avoid mismatches due to capital letters. The biggest drawback of the approach we used is that we inevitably selected Facebook posts that were not really about COVID-19 (false positives) and missed Facebook posts that were about COVID-19 (false negatives) but do not mention any of the words in our list. After the automated topic coding was completed, we took a random sample of 100 Facebook posts and manually coded them. In 86% of the posts the topic coding was accurate1, which is rather high for topic coding. This high accuracy can be perhaps explained by the fact that the COVID-19 crisis introduced, on the one hand, a completely new vocabulary of medical-related words and, on the other hand, many existing words such as ‘pandemic’ and ‘distance’, which were almost exclusively used in relation to COVID-19. A potential limitation of our approach that should be mentioned is that we only looked at the Facebook caption and title of the article when classifying articles as COVID-19-related or not. Although this might increase the risk of missing COVID-19-related articles, this decision was made because we believe that many people make a comment based on the title or Facebook caption without reading the article first. By only looking at the caption and article title, we believe that we mimic most closely the way that people read and react to Facebook posts.

To better pinpoint changes over time, as the pandemic gradually lost its novelty and became a fixture in everyone’s lives, we divided our findings over four smaller periods of three months each. These roughly overlap with the pandemic’s dominance in public life and debate from a Belgian perspective, although we find that it is also applicable and appropriate for international affairs, though still predominantly Western-oriented research. March to May 2020 marked the period of the harshest lockdowns and restrictions, with eases and relaxations between June and August. Between September and November, infections and restrictions alike rose again. December to February 2021, finally, saw a continuation of most restrictions, including the holiday period and the advent of countries’ vaccination drives.

4. Results

4.1. Descriptive Findings

Looking at the macro level of our dataset, we collected a total of 140,359 Facebook posts of 17 Belgian (Flemish) news brands published between 1 March 2020 and 2021. Exactly 50,350 of those, or 35.87% of our total number, were found to be related to COVID-19. This means that the pandemic, perhaps unexpectedly, did not completely dominate the Facebook news cycle. Before answering our research questions, we delve deeper into the data and denote differences at the level of the outlet (mainstream versus alternative outlets) and those of periods throughout the year of study.

Unsurprisingly and in line with previous research (e.g., Hiaeshutter-Rice and Weeks 2021), a lower share of pandemic-related reporting was found for alternative outlets (29.65%) compared to mainstream outlets (37.14%). When we zoom further in on the individual outlets, two alternative right-wing news outlets (PAL NWS, 21.4%; Doorbraak, 21.2%) posted, by quite a bit, the lowest share of COVID-19-related Facebook posts. The highest shares were found at the public service broadcaster (VRT, 40.7%), private television broadcaster VTM (also 40.7%), a left-wing alternative outlet (De Wereld Morgen, 41.0%), quality newspaper De Morgen (40.3%) and the business-oriented quality newspaper De Tijd (40.4%). Overall, most mainstream outlets posted a comparable share of pandemic-related news, while there are larger discrepancies between alternative outlets (especially between right-wing and left-wing).

We divided our studied period into four three-month-periods and after comparing our data in the different periods, it becomes apparent that the first period of analysis (March–May 2020) had the highest share of COVID-related Facebook news posts (49.68%), with lower shares in the next two periods (30.66 and 29.82%, respectively). The fourth period, December 2020 to March 2021, saw an uptick of 33.43%, explained by the launch of the vaccination campaign and continued closures of shops, bars, and restaurants. Initially, Facebook posts about the pandemic took up nearly half of the total posted articles. After the first few months, this share already dropped significantly and stabilized around 30%. For all four periods, mainstream news outlets again outperformed alternative titles percentagewise. In the first period, alternative outlets devoted 45.82% of their posts to COVID-19-related news, which is quite close to the percentage of mainstream outlets (50.49). After the first period, alternative news outlets allocated, respectively, 25.31, 22.27 and 27.53% of their post to COVID-19 related articles, while this percentage remained higher for mainstream outlets (31.71, 31.45, 43.78%).

4.2. COVID-19-Related News and Other News

This section answers our first research question. It discusses total interactions and reactions to Facebook posts on COVID-19 and other news topics, respectively.

4.2.1. Interactions

Across the 140,359 Facebook posts, we counted a grand total of 69,115,166 interactions. For the 50,350 pandemic-related posts specifically, the total amounted to 26,755,740 (or 38.71%). Keeping in mind that 35.87% of all posts were COVID-19-related, audiences overall engaged slightly more with COVID-19-related news than with other types of news.

Table 3. Different forms of interaction for COVID- vs. non-COVID-related news.

	COVID-19 News (N = 50,350)	Other News (N = 90,008)	Total (N = 140,358)
Likes	16,777,237 (58.72%)	29,665,112 (65.56%)	46,442,349
Shares	2,125,185 (7.44%)	2,596,790 (5.74%)	4,721,975
Comments	9,671,624 (33.85%)	12,988,567 (28.70%)	22,660,191

zooms closer in on the different forms of interaction: likes (total of all like options), shares and comments. COVID-19-related Facebook posts (58.72%) were compared to other less-liked news (65.56%). We did, however, see that COVID-19 news was, in comparison to other news, shared (respectively 7.44% and 5.74%) and commented on (respectively 33.85% and 28.70%) more often.

Additionally, when divided per quarter, our results revealed that the numbers of likes and shares of COVID-19 news posts gradually declined throughout the four periods, while comments progressively increased in volume (see Figure 2). We did not assess the content of Facebook users’ comments and are, hence, unable to make finite judgments on what explains the rise in commenting as a means of exerting audience agency. Our conjecture is that rather than liking or sharing posts, citizens progressively adopted their online agency and became more articulate and empowered to voice their personal opinions about unpopular measures such as lockdowns, the cancellation of events and concerts and the closures of schools and shops. During the fourth period, the rollout of the vaccination campaign began, which was a hot topic in public debate, which may also explain the increase in the number of comments.

4.2.2. Reactions

Since 2016, the traditional ‘like’ button was compounded with other reactions to give Facebook users the opportunity to provide more specific emotive feedback. Users can still ‘like’ posts but can also express that they have positive (‘Love’, ‘Care’, ‘Haha’, ‘Wow’) or negative (‘Sad’, ‘Angry’) feelings towards certain posts.

Table 4. Reactions to Facebook news posts.

	COVID-19 News	Other News	Total
Like	7,863,417 (46.87%)	13,487,028 (45.46%)	21,350,445
Love	1,270,007 (7.57%)	2,942,502 (9.92%)	4,212,509
Wow	1,206,053 (7.19%)	1,902,662 (6.41%)	3,108,715
Haha	1,840,187 (10.97%)	2,749,473 (9.27%)	4,589,660
Sad	2,202,706 (13.13%)	3,316,069 (11.18%)	5,518,775
Angry	1,927,734 (11.49%)	3,635,140 (12.25%)	5,562,874
Care	467,133 (2.78%)	1,632,238 (5.50%)	2,099,371

displays the different reaction options for COVID-19 and other news. The traditional ‘like’ option was still by far the most used reaction and made up almost half of all the reactions on both COVID-19 (46.87%) and other news (45.46%). Surprisingly, the ‘Love’ and ‘Care’ button were the options that were less used for COVID-19 news, which went against our own assumption that pandemic-related reporting would lead to ‘harder’ reactions. Although the ‘Care’ button was also less used on other news, the button was especially introduced in April 2020 to accommodate a sense of online connectedness, but was also not used more for COVID-19 posts. ‘Wow’ and ‘Sad’ were used slightly more often for COVID-19 news, while ‘Angry’ was slightly used more for other news. The ‘Haha’ button was used more often for COVID-19 news than for other news. This is a bit puzzling considering that intuitively we would not consider COVID-19 news very humorous or funny.

When further zooming in on news posts discussing COVID-19 and differentiating our dataset per quarter, a few intriguing tendencies surface, as schematised in Figure 3. The average use of ‘Love’, ‘Wow’ and ‘Sad’ reactions moderately declined throughout our four three-month periods. The ‘Care’ option shows an initial increasing trend in period one, two and three but in the last studied period, the use went down again. A similar trend can be seen for the ‘Angry’ button, which gained quite some relevance among Facebook users engaging with Flemish pandemic-related news, but in period four the average use slightly drops. The most interesting trend is the ‘Haha’ option, which shows a steep incline in our studied period. While already frequently used in the first 6 months, the ‘Haha’ option in the fourth period emerged as the most often-used reaction. The general ‘Like’ option is not included in the figure below as it scores considerably higher than the other options. Here too, though, a general decline in the average number of likes per post becomes apparent throughout our four periods, with a drop from 160 to 120—still twice as much as the second most popular reaction type in the fourth and final period, which was the ‘Haha’ reaction.

4.3. Mainstream and Alternative News Outlets

Our second research question focuses on the difference in audience engagement between mainstream and alternative news outlets’ Facebook posts. Comparing the likes, shares and comments of COVID-19 related post for alternative and mainstream media outlets separately, we see that posts of alternative media outlets are slightly more liked (61.22%) and shared (9.93%) than posts of mainstream media (58.66% and 7.38%, respectively). The percentage of comments is slightly higher for mainstream media (33.96%) compared to alternative media Facebook posts (28.85%).

When we compare the use of the different like options between alternative and mainstream media outlet posts about the pandemic, we see some interesting differences. On the one hand, the ‘Angry’ reaction was used far more for alternative media posts (23.82%) than for mainstream media posts (11.21%). The same was true for the ‘Haha’ reaction (alternative: 19.49%, mainstream: 10.77%). The ‘Wow’ and ‘Sad’ reaction were, on the other hand, used more for mainstream media posts (7.27% and 13.33%, respectively) compared to alternative media outlets (3.79% and 4.57%). Within the different alternative media outlets, there was a clear distinction between left-wing and right-wing outlets when it came to the ‘Haha’ reaction. The ‘Haha’ reaction saw only very limited use for Facebook posts of left-wing outlets (Apache: 7.94%, De Wereld Morgen: 1.27% and MO*: 0.50%), while right-wing outlets generated this reaction more (Doorbraak: 14.20%, Business AM: 13.06%, PAL NWS: 20.66% and Newsmonkey: 10.98%). For the ‘Angry’ reaction, the dichotomy between left-wing and right-wing alternative outlets was less clear. Although the share of Angry reactions was high for PALNWS (36.10%) and Doorbraak (20.54%), the left-wing alternative outlet Apache (17.17%) also generated quite a lot of ‘Angry’ reactions. The other left-wing outlets (De Wereld Morgen: 10.51% and MO*: 8.43%) and the right-wing outlet Newsmonkey (4.42%) had a much smaller share of ‘Angry’ reactions.

In the next step, we performed a negative binomial regression to estimate how the different variables (news outlet and news topic) affected audience engagement (combination of shares, like options and comments). To achieve this, we recoded news outlets into a dichotomous variable for mainstream and alternative outlets, with alternative news outlets as the reference category. To account for the differences in reach between the different outlets, we included the number of followers of the Facebook page of the outlet when they posted the news article as the control variable. We also included the period (again differentiating between the four consecutive quarters in our year of study) to see if the beginning of the pandemic affected audience engagement differently. In a second model, we also included an interaction between the news outlet and news topic to assess if news related to COVID-19 had a different effect on user engagement for alternative news outlets compared to their mainstream counterparts. The output of the negative binomial regression is presented in

Table 5. Negative binomial regression with audience engagement as dependent variable.

	Model 1		Model 2 (Interaction)
	Incidence Rate Ratios	Conf. Int. (95%)	Incidence Rate Ratios	Conf. Int. (95%)
News topic (Non-COVID-19 = 0)	1.07 ***	[1.05–1.08]	0.67 ***	[0.64–0.70]
News outlet (Alternative = 0)	1.09 ***	[1.06–1.11]	0.94 ***	[0.92–0.97]
Period 2	0.87 ***	[0.85–0.89]	0.87 ***	[0.85–0.89]
Period 3	1.05 ***	[1.03–1.08]	1.05 ***	[1.03–1.08]
Period 4	0.99	[0.97–1.01]	1.00	[0.98–1.02]
Followers	1.00 ***	[1.00–1.00]	1.00 ***	[1.00–1.00]
News topic*News outlet			1.72. ***	[1.64–1.79]
Intercept	101.49 ***	[99.06–103.99]	113.67 ***	[110.70–116.74]
R squared Nagelkerke	0.431		0.435
N	140,356		140,356

*** p < 0.001.

below.

In our first model we found a significant difference between engagement of COVID-19-related and non-COVID-19-related news on Facebook, where pandemic related news generated more interaction than non-COVID-19-related news. While other variables remained constant in the model, COVID-19 news, compared to non-COVID-19 news, was expected to have a rate 1.07 times greater for news engagement. The period variable was also found to have a significant impact on engagement, except for period 4. We can see that overall, period 1 had more interactions than period 2 but less than period 3. This is in line with what we saw in the data presented above. As 2020 progressed, 2021 loomed and the pandemic remained endemic, or recurrent; there was also more user activity. In the descriptive analysis, we already found that the nature of the interaction changed. The number of shares declined, while the number of comments increased over our studied period. Comments remained similar throughout the year.

When looking at RQ2, or the distinction between alternative and mainstream media outlets, we see that overall, the latter generated more audience interaction than the former. Mainstream media are expected to have a rate 1.09 times greater for news engagement compared to alternative media outlets.

When we zoom further in on the difference in engagement between news topics and news posts, we find a significant difference. In Figure 4, the interaction is plotted to make interpretation more straightforward. Here, we see that the audience engagement for non-COVID news was similar for the alternative and mainstream media Facebook posts. This is not the case for COVID-related news. Here, we found a difference between the alternative and mainstream media posts, where COVID-related news generated more user engagement for Facebook posts by mainstream media outlets.

5. Discussion and Conclusions

In this paper, we collected 140,359 Facebook posts by 17 different mainstream and alternative news outlets from Flanders (Belgium), all published between 1 March 2020 and 2021. We assessed the Facebook posts for their various interactions and sought to find out if we were able to denote differences along the lines of various media brands and of news coverage that was and was not related to the COVID-19 pandemic, which we consider as the paper’s most useful and novel contribution to existing scholarship. With 35.87% of our dataset constituting pandemic-related news posts, we found that COVID-19 only moderately dominated the news cycle to the extent one could have suggested, at least on Facebook. We were unable to confirm that this corresponded completely with the actual online news offer of published articles by news outlets’ websites for our vast sample, particularly as recent scholarship also applicable to the Flemish media market signalled instances of so-called ‘softening’ processes of news content on Facebook versus what is published at news outlets’ websites as online news articles (Lamot 2021). Overall, we found that Facebook users engaged more with posts related to the pandemic (RQ1). Although COVID-19-related posts were liked slightly less, they generated more shares and comments compared to other news. Looking at reactions, we also found differences in user engagement. The ‘Wow’ and ‘Sad’ reaction was, not surprisingly, used more for COVID-19-related posts. Remarkably, however, the “Haha” response was also used more with COVID-19 news. When zooming in further, we found that this is mostly explained by the fact that the ‘Haha’ response was used a lot with posts from right-wing alternative media outlets.

The fact that posts about the pandemic generated more interaction is in line with previous studies of engagement towards COVID-19 news. Important here is that pandemic-related posts were not just about hard news but also often used a human interest/soft news angle, where social media editors could pull out all the stops to provoke interaction, for example through status updates and headline changes. This gives us a double effect that boosts engagement: the hype of the topic plus in addition social media rhetoric.

Regarding the difference between types of news media, the analyses show that in terms of COVID-19-related news, mainstream accounts elicited more engagement than alternative accounts (RQ2). This could be explained by the tendency of people to turn to the better-known, mainstream media in their search for information and news during times of crisis. According to Media Systems Dependency Theory, mass media are seen as best to fulfil information needs, especially in times of uncertainty. It is, therefore, not surprising that traditional media brands such as TV stations and elite newspapers played a crucial role as news sources during the pandemic and experienced an increase in user numbers as public service media in many Western nations tend to be considered as the most trustworthy sources of information in times of crisis (Ali et al. 2020; Van Aelst et al. 2021).

In the descriptive analysis, we found several things to be more outspoken and extreme in the first quarter of our time frame, meaning March to May 2020. This period of course overlapped with the, thus, far-harshest period of the pandemic in Belgium and various other (Western) nations, with strict lockdowns and other restrictions. As previously shown by studies, the pandemic was new and unexpected to many, and this was of course also reflected in the production (Libert et al. 2021), diffusion and consumption of news content (Hendrickx 2021; Mellado et al. 2021). Interestingly, some forms of engagement were higher in the first period (shares), while other forms (liking and commenting) increased in the later periods. We gauge that the mainly similar trends in dealing with the pandemic, from a government and societal perspective, as well as the similarity of the Flemish media market with others in terms of mainstream and alternative news outlets represented, together enhance the transferability of our analyses’ findings, at least to other Western markets and nations.

A shortcoming to this paper is that our results are to a large extent descriptive, as we can only contextualise the findings of this paper to the extent of other relevant scholarship and wider political and societal trends. It was not possible for us to properly explain how and why citizens use their agency on a daily basis to choose if, when and how to engage with news outlets’ Facebook posts, which form just a fraction of the countless decisions they make on a daily basis to exert agency and (not) contribute to audience engagement (Picone et al. 2019). With audience engagement and user metrics only gaining relevance within journalism practice and studies alike, future research can pay more attention to the how and why questions of individual citizens and audiences engaging with online news via social media in their daily lives.