Silvia Majó-Vázquez, Jun Zhao and Rasmus Kleis Nielsen

Key findings

In this RISJ Factsheet, we analyse a sample of 2.96 million news-related tweets from a larger dataset of 43.5 million tweets collected during the 2017 French Presidential elections to examine the role of digital-born and legacy news media in online political discussions in France.

We find that:

• Legacy media, most notably newspapers and broadcasters, figure very prominently in the political discussion on Twitter. Legacy media generated more than seven times as much activity and engagement as digital-born news media during the election. 88.43% of news-related tweets collected either originated with or included explicit reference to legacy media, compared to 11.56% that originated with or included explicit reference to digital-born news media.

• A high number of followers and frequent tweeting does not automatically translate into high levels of engagement from audiences. Some prominent newspapers (Le Monde and Liberation), broadcasters (TF1 and BFM TV), and digital-born news media (Mediapart and Brut) see high levels of engagement, but some newspapers and digital-born news media have much lower levels of engagement than their general audience reach, follower count, and their frequent tweeting would lead one to expect.

• Attention and engagement is very unevenly distributed. A few media organisations like Liberation, TF1, and Mediapart are mentioned by others on average hundreds or even more than a thousand times for every time they tweet. In contrast, many news organisations, especially local and regional newspapers and smaller digital-born news media, find the level of engagement they receive does not correspond to the number of tweets they send, suggesting limited engagement. Interestingly, many of these organisations have a high ratio of following to followers – but despite thus engaging with many other accounts on Twitter see limited engagement in return.

Our findings suggest that news media organisations – especially a few prominent newspapers, broadcasters, and digital-born news media – figure very prominently in political discussions on Twitter, but also that a number of organisations with considerable audience reach, Twitter follower counts, and who tweet very frequently see much more limited engagement, and that local and regional newspapers as well as smaller digital-born news media see little engagement on Twitter.

General Overview

Digital-born and legacy media are competing to control the most central positions in the flow of online news. How this competition, which is even greater during major political events, unfolds on Twitter is still largely unknown. The 2017 French Presidential election offers a useful lens to investigate the role of digital-born and legacy media in one of the most important countries in Europe, and one that is different from the English-language markets most frequently analysed, both in terms of its political system and its media system.

The purpose of this RISJ Factsheet is to present an overview of 1) the allocation of news audience attention on Twitter during the French elections and 2) the potential influence of digital-born and legacy media in this platform of news distribution.

The French context is different from that found in many other markets in important ways. First, French digital-born news media are more prominent than in other European countries. Also legacy media, especially newspapers, have a historically weaker role here than in, for example, Germany or the UK (Nicholls, Shabbir, & Nielsen, 2016). Print newspapers have lower reach than in much of Northern Europe and although television remains the most important and widely used source of news in France, the overall audience is aging and eroding (Newman, Fletcher, Levy, & Nielsen, 2016). In contrast, the French increasingly turn to social media for news. 40% say they get news via social media, especially Facebook. 8% use Twitter for news, and the figures for Twitter, and for social media more widely, is higher for younger media users.

While still only used by a small minority as a source of news, Twitter has a broader political role that makes it important to understand political discussion there. In addition to serving as a platform for discussion and news distribution, political campaigns seek to use it to mobilise support and shape political discussions, and journalists and news media in turn pay attention to Twitter as they cover politics (Chadwick, 2013; Conway, Kenski, & Wang, 2015; Murthy, 2015; Quinlan et al., 2017; Stromer-Galley, 2014). We therefore want to analyse political discussion on Twitter during the French Presidential elections and specifically focus on the role of legacy and digital-born news media organisations in political discussions on Twitter.

Data Collection

Our dataset contains 2.96 million tweets collected between 2 April and 8 May 2017. We use a supervised third-party software for Twitter data collection, named Kalium, which calls the Streaming API. As Morstatter, Pfeffer, Liu, & Carley (2013) note the essential drawback of the Twitter Streaming API is the lack of information concerning what and how much data one gets. Yet we do know that it allows us to retrieve at most a 1% sample of all the data matching a set of pre-determined parameters (see Morstatter et al., 2013 for a systematic comparison between the Streaming API and the Twitter Firehose data).

Our focus here is on news-related tweets, which we gathered 1) by collecting all tweets from a pre-defined list of Twitter handles and 2) by collecting additional tweets on the basis of relevant hashtags identified seperately.

Identifying media outlets and individual accounts

We started by defining a sample of French news media outlets based on their overall audience reach. Using comScore online metrics, we averaged the individual audience reach for French media outlets during the last three months starting from April 2017. We obtained the ranking of the most visited news sites in France with at least a percentage reach of 0.03%. To avoid leaving out important news sites, not in terms of their reach but instead of their brand awareness, we double-checked our list of news media with a group of five French experts including researchers and journalists (see the Acknowledgements). They helped us to identify two main types of additional media outlets: recently founded news sites that are not indexed in comScore yet (e.g. Brut, cFactuel, Brief.me or Explicite); or popular news sites that fall at the extremes of the ideological spectrum (e.g. Égalité & Récontiliation, Fdesouche or Dreuz Info). In total, we studied 27 digital-born outlets and 71 legacy media (see the Appendix for the full list). Finally, we incorporated the list of key actors by including the name of the eleven French presidential candidates and their political parties.

We gathered all tweets sent by our pre-selected set of media and political actors. We also collected all tweets including their name or their Twitter username. Tweets where their names were mentioned in URLs embedded in text were also gathered.

Identifying tags

Previous research has already pointed out the drawbacks of collecting tweets based on usernames as well as the complexity of tracking a multithreaded social event such as an election campaign (Jungherr, 2016).

To minimise the risk of missing other relevant tweets during the elections, we completed this primary data collection strategy with a second approach. We also included news-related tweets from other users by identifying relevant hashtags in the overall French Twitter conversation by tracking trending conversations twice per day. We used several online tools for this purpose. The most important hashtags in terms of the overall conversation were included in our Twitter crawler too.

Overall, we gathered 43.5 million tweets through this combination of approaches, which we then filtered following the criteria specified in Figure 3 (at the end of the Factsheet). Finally, we obtained a sample of 2.96 million tweets. 718,000 of them are original tweets and 2.24m are retweets.¹

News Content Activity

The first step in our analysis is to assess the distribution of news content volume throughout the electoral campaign cycle. Figure 1 shows the overall volume of tweets about the French Presidential elections by type of news media. The category legacy media includes national daily newspapers, monthly and weekly newspapers, regional and local daily newspapers, magazines, news agencies, radio, broadcasters and free national newspapers. Digital-born news media include a variety of news-related organisations launched online, most of them recently. The main findings of this longitudinal analysis are:

• Legacy media figure far more prominently in news-related Tweets during the French Presidential elections than digital-born news media do. 88.43% of the news-related content on Twitter either originate with or include explicit reference to legacy media. Only 11.56% of this content originate with or include explicit reference to digital-born news media.

• The volume of news-related Twitter activity is clearly driven by events and especially by the electoral cycle. Spikes in volume correspond to the most important stages of the electoral campaign i.e. electoral debates, the start of the campaign and the first and second polling days, as well as to events like the attack on the Champs Élysées.

• The first electoral debate, or “Le Grand Débat”, that took place on 4 April and featured eleven presidential candidates saw the highest amount of news-related activity registered in one day on Twitter. More than 126,000 messages around that date included legacy media content or mentioned this type of media outlet.

• Digital-born media content reached the highest volume on Twitter before the second and final polling day, but their output varies far less than that of legacy media and they account for much less activity overall.

• Of note, the jump in the overall number of tweets related to legacy media content reached its highest point only hours after both polling days. This contrasted with the subsequent 24 hours after the first round of the elections, where there was a substantial drop in legacy media news content. As tweets were only collected until 8 April, (the day after the second polling day) it is not clear whether this drop took place again on 9 April. However, we do observe an even greater jump in the immediate hours after the second round.

• Finally, the trend representing content by, or mentions to, digital-born outlets follows a more stable pattern with less spikes, whereas the line of legacy media shows deep falls and jumps before and after a salient electoral event.

Figure 1. Timeline of overall number of tweets by media type

Most Active Media

A more in-depth analysis looks at the individual activity on Twitter of the 98 media outlets under study. Table 1 includes the ranking of the top 20 most active news media as measured by the number of original tweets posted during the French campaign. At this level, we find that:

• Le Figaro was the most active news media outlet overall.

• Le Huffington Post is the only digital-born outlet amongst the top 20 most active news media.

• Legacy media led the posting activity on Twitter by a great distance. National daily newspapers, broadcasters and radio are the most active categories.

• Amongst the three least active media users in our sample we find Le Canard Enchainé. This is the weekly newspaper that first reported on the alleged financial scandals affecting François Fillon’s wife, known for its political satire (and its off-hand approach to digital media).

• In this category of least active we also find two local daily newspapers, L’informateur d’Eu et L’Éclaireur and Le Républicain 47.

Table 1. Ranking of the 20 top most active media outlets on Twitter

Type	Brand	TW username	Tweets
National Daily Newpaper	Le Figaro	Le_Figaro	3198
TV	BFM TV	BFMTV	3004
TV	TF1/LCI	LCI	2183
National Daily Newpaper	Les Echos	LesEchos	2025
Radio	RMC	RMCinfo	1889
Radio	Europe 1	Europe1	1883
National Daily Newpaper	Le Monde	lemondefr	1675
Weekly Newspaper	L’Obs	lobs	1589
Weekly Newspaper	Le JDD	leJDD	1518
Free National Newspaper	20 Minutes	20Minutes	1366
TV	LCP	LCP	1242
Regional Newspaper	Ouest France	OuestFrance	1222
Magazine	Paris Match	ParisMatch	1195
Radio	RTL	RTLFrance	1155
News Agency	Agence France-Presse	afpfr	986
National Daily Newpaper	Liberation	libe	914
Weekly Newspaper	L’Express	LEXPRESS	876
TV	Franceinfo	franceinfo	871
Digital-born	Le Huffington Posts	LeHuffPost	866
Weekly Newspaper	Le Point	LePoint	863

Media Influence

So far, we have documented the overall importance of news media content during the French Presidential elections. The next step is to examine whether the prominence of legacy media and their high levels of Twitter activity are translated into greater levels of influence within the overall flow of news. To address this question we analyse the relation between the amount of news content produced and the number of mentions and replies received by each media outlet. Drawing on previous research on network role identification on Twitter (Gonzalez-Bailon, Borge-Holthoefer, & Moreno, 2013), we estimate the overall potential influence of each news outlet by relating their level of audience engagement to the relationship they establish with their followers. The basic idea here is that influence can be mapped along two dimensions. First, the ratio between how often an account tweets (messages sent) and how often it is mentioned or receives replies by others (messages received) and, second, the ratio between how many accounts it follows, and how many followers it has. In line with this, we first look at the ranking of most retweeted tweets² (Table 2) and find that:

• At the top of the ranking there is a digital-born outlet. A satirical video from Brut was posted at the end of the electoral campaign and one day before the final polling day. It obtained almost 40,000 retweets and over 25,000 likes.

• At the aggregate level, tweets from prominent television broadcasters dominated the ranking of most retweeted messages.

• All top ten most retweeted tweets contained some kind of audiovisual content, which speaks to the importance of compelling videos and pictures to boost the popularity of a news message on Twitter.

Finally, we look at the allocation of news audience attention by media brands and their relationship with their public (Figure 2) along the two dimensions of messages sent/received and following/followers, and find that:

• The level of Twitter activity does not seem to always explain the overall influence of news media outlets as measured by number of mentions and replies received by news content.

• The weekly newspaper Le Canard Enchainé reached the highest level of attention on Twitter during the French Presidential elections, although it was amongst the least active news media in this platform. This result highlights that offline news content also drives Twitter audience attention.

• Similarly, several digital-borns exhibit a higher level of audience attention in comparison to their posting activity. This is the case of Brut (number of posts=84), Mediapart (n=435) and Explicité (n=55) amongst others.

• The most active accounts, even when these are associated with very prominent news media, do not necessarily see particularly high levels of engagement on Twitter. The broadcasters BFM TV and TF1 have high general audience reach across offline and online media, high follower counts, and are very active on Twitter, and see high levels of engagement as measured by messages sent/received, but this is less so in the case of, for example, the newspapers Le Figaro and Les Echos. In contrast, news media like the newspaper Liberation and the digital-born news media site Mediapart seem far more influential on Twitter judging by overall engagement than their general online audience reach or their Twitter following alone would suggest.

• The ratio of messages received and sent is negatively related to the ratio of following over followers. In other words, news media outlets that achieve higher levels of attention do not necessarily follow large numbers of other accounts. Particularly interesting here is how attention and engagement is very unevenly distributed, and how our data documents that many news organisations, especially local and regional newspapers and smaller digital-born news media, tweet almost as often as other Twitter users engage with their tweets, suggesting limited engagement. Interestingly, many of these organisations have a high ratio of following to followers – but despite thus engaging with many other accounts on Twitter see limited engagement in return.

Figure 2. Distribution of media outlets according to network position and conversation engagement³

Summary

This document provides a descriptive analysis of the role of digital-born and legacy media in the flow of news content on Twitter during the French Presidential elections in 2017. We have brought evidence to understand better the re-allocation of roles between digital-born and legacy media looking at Twitter during a major political event. We have found that legacy media dominate the overall production of news content on Twitter. However, once we look at the audience engagement we have seen that Twitter activity is not always associated with the same levels of influence as measured by audience engagement. Our analysis has helped identify both the limited number of very influential, mostly legacy media, brands that figure very prominently in online political discussions in France, those that seem more influential on Twitter than their audience reach and following count alone would suggest, some who are punching below their weight, and a significant number of smaller legacy and digital-born news media organisations that see very little engagement on Twitter.

Figure 3. Sampling and filtering process

References

Chadwick, A. (2013). The Hybrid Media System: Politics and Power. Oxford: Oxford University Press.

Conway, B. A., Kenski, K., & Wang, D. (2015). ‘The Rise of Twitter in the Political Campaign: Searching for Intermedia Agenda-Setting Effects in the Presidential Primary’. Journal of Computer-Mediated Communication, 20(4), 363–380.

Gonzalez-Bailon, S., Borge-Holthoefer, J., & Moreno, Y. (2013). ‘Broadcasters and Hidden Influentials in Online Protest Diffusion’. American Behavioral Scientist, 57(7), 943–965. http://doi.org/10.1177/0002764213479371

Jungherr, A. (2016). ‘Twitter Use in Election Campaigns: A Systematic Literature Review’. Journal of Information Technology & Politics, 13(1), 72–91.

Morstatter, F., Pfeffer, J., Liu, H., & Carley, K. M. (2013). ‘Is the Sample Good Enough? Comparing Data from Twitter’s Streaming API with Twitter’s Firehose’. arXiv Preprint arXiv:1306.5204.

Murthy, D. (2015). ‘Twitter and Elections: Are Tweets, Predictive, Reactive, or a Form of Buzz?’ Information, Communication & Society, 18(7), 816–831.

Newman, N., Fletcher, F., Levy, D. A., & Nielsen, R. K. (2016). Digital News Report 2016. Oxford: Reuters Insitute for the Study of Journalism.

Nicholls, T., Shabbir, N., & Nielsen, R. K. (2016). Digital-Born News Media in Europe. Oxford: Reuters Insitute for the Study of Journalism.

Quinlan, S., Gummer, T., Roßmann, J., & Wolf, C. (2017). “Show Me the Money and the Party!” – Variation in Facebook and Twitter Adoption by Politicians. Information, Communication & Society, 1–19.

Stromer-Galley, J. (2014). Presidential Campaigning in the Internet Age. Oxford: Oxford University Press.

Acknowledgements

The authors would like to thank Leonie Rivere for her valuable work as a research assistant. They also gratefully acknowledge the work of Eurecat’s Digital Humanities Research Group, Andreas Kaltenbrunner, Pablo Aragón and Matteo Manca, who helped us to collect and analyse the data of this report and accommodate all our requirements. Lastly, we thank the following researchers and journalists’ contribution in helping us to design our sample of French news media: Julia Cage, assistant professor at Sciences Po Paris, Department of Economics; Arnaud Mercier, professor of communication at the University of Paris 2; Alex Vicente, culture correspondent based in France for El País; Joan Carles Peris, France correspondent, Catalan Public Television, TV3; Mariona Vivar Mompel, freelance journalist.

About the authors

Sílvia Majó-Vázquez is Research Fellow at the Reuters Institute for the Study of Journalism at the University of Oxford

Jun Zhao is Senior Research Fellow at the Department of Computer Science at the University of Oxford

Rasmus Kleis Nielsen is Director of Research at the Reuters Institute for the Study of Journalism at the University of Oxford

Appendix

Table A1. Reach and Twitter statistics by French media outlets

Type	Brand	TW username	Online Audience Reach	Followers
National Daily Newpaper	Le Figaro	le_Figaro	18.53	2,636,827
National Daily Newpaper	Le Monde	lemondefr	19.72	7,101,100
National Daily Newpaper	Les Echos	lesEchos	5.64	984,067
National Daily Newpaper	Liberation	libe	4.70	2,505,670
National Daily Newpaper	La Tribune	laTribune	1.96	179,558
National Daily Newpaper	La Croix	laCroix	2.25	40,617
Free National Newspaper	20 Minutes	20Minutes	10.03	2,246,457
Free National Newspaper	Cnews Matin	CNEWSMatin	0.64	14,109
Regional Newspaper	Le Parisien	le_Parisien	12.30	1,856,493
Regional Newspaper	Ouest France	ouestFrance	7.69	492,288
Regional Newspaper	La Depeche du Midi	ladepechedumidi	4.09	69,608
Regional Newspaper	Sud Ouest	sudouest	3.35	383,636
News Agency	Agence France-Presse	afpfr	0.09	2,634,211
Digital-born	Le Huffington Posts	leHuffPost	7.40	945,085
Digital-born	Slate	slatefr	2.44	658,114
Digital-born	Mediapart	mediapart	1.94	2,017,606
Digital-born	Atlantico	atlantico_fr	1.54	69,312
Digital-born	Arrêt sur Images	arretsurimages	0.14	487,942
Digital-born	Acrimed	acrimed_info	0.09	42,494
Digital-born	NonFiction	nonfiction_fr	0.03	6,112
Digital-born	Fakir	fakir_	n/a	31,228
Digital-born	cFactuel	cFactuel	n/a	3,539
Digital-born	Brut	brutofficiel	n/a	41,899
Digital-born	Rue89	rue89	n/a	1,432,801
Digital-born	Les Jours	lesjoursfr	n/a	26,829
Digital-born	Bondy Blog	leBondyBlog	0.07	30,985
Digital-born	Égalité & Récontiliation	eetR_National	0.51	30,139
Digital-born	Fdesouche	F_Desouche	0.59	24,443
Digital-born	Wikistrike	wikistrikeW	0.15	6,104
Weekly Newspaper	L’Obs	lobs	10.66	1,190,600
Weekly Newspaper	L’Express	LEXPRESS	10.24	1,193,875
Weekly Newspaper	Le Point	lePoint	5.19	641,771
TV	BFM TV	BFMTV	10.50	2,158,449
TV	TF1/LCI	LCI	5.13	94,871
TV	France 24	FRANCE24	1.37	2,770,705
TV	Franceinfo	franceinfo	13.10	1,045,639
Radio	RTL	RTLFrance	5.45	692,539
Radio	Europe 1	europe1	3.45	1,248,844
Radio	Franceinter	franceinter	2.35	1,199,497
Radio	RFI	RFI	1.35	1,447,527
Regional Newspaper	La Voix du Nord	lavoixdunord	2.33	293,432
Regional Newspaper	Le Midi Libre	Midilibre	n/a	128,142
Regional Newspaper	Le Progrès	Le_Progres	1.48	131,000
Regional Newspaper	La Provence	laprovence	1.25	174,744
Regional Newspaper	L’Est republicain	lestrepublicain	1.27	80,946
National Daily Newpapers	France Soir	france_soir	1.07	18,238
Local Daily Newspaper	Le Républicain Lorrain	lerepu	0.79	14,318
Local Daily Newspaper	La Montagne	lamontagne_fr	0.98	81,366
Weekly Newspaper	Le JDD	leJDD	0.80	321,913
Regional Newspaper	L’Union	UnionArdennais	0.63	49,629
Type	Brand	TW username	Online Audience Reach	Followers
Local Daily Newspaper	DNA Dernières Nouvelles d’Alsace	dnatweets	0.51	113,668
Local Daily Newspaper	Le Courrier Picard	CourrierPicard	0.49	86,667
Local Daily Newspaper	Lyon Mag	lyonmag	0.25	15,200
Local Daily Newspaper	La Manche Libre	lamanchelibre	0.24	6,274
Local Daily Newspaper	La République de Seine-et-Marne	LaRep77	0.24	3,564
Local Daily Newspaper	Le Populaire	lepopulaire_fr	0.27	13,782
Digital-born	Le Petit Journal	lepetitjournal	0.30	9,281
Regional Newspaper	L’Écho Républicain	lecho_fr	0.26	5,216
Regional Newspaper	L’Yonne Républicaine	lyonne_fr	0.21	6,184
Regional Newspaper	Le Berry Républicain	leberry_fr	0.27	11,023
Regional Newspaper	Vosges Matin	VosgesMatin	0.24	6,726
Regional Newspaper	L’Est Eclair	lesteclair	0.17	8,833
Regional Newspaper	Nord Eclair	NordEclairWeb	0.24	17,324
Local Daily Newspaper	Le Phare Dunkerquois	pharedk	0.10	3,782
Weekly Newspaper	Le Canard Enchainé	canardenchaine	0.20	407,141
Regional Newspaper	Le Journal du Centre	lejdc_fr	0.18	6,657
Local Daily Newspaper	Le Républicain 47	LeRep47	0.06	1,026
Local Daily Newspaper	L’Eveil	leveil43	0.06	2,900
Local Daily Newspaper	L’informateur d’Eu et L’Éclaireur	InfoEclaireur	0.05	1,038
Regional Newspaper	L’Ardennais	UnionArdennais	0.11	49,629
Local Daily Newspaper	Montceau News	montceaunews	0.09	754
Digital-born	Infonormandie	infoNormandie	0.10	8,762
Regional Newspaper	La Gazette	lagazette95	0.15	3,176
		echoregional95	n/a	1,814
Digital-born	WeDemain	WeDemain	0.06	21,567
Regional Newspaper	Nord Littoral	Nordlitt	0.08	5,923
Digital-born	Mediacités	Mediacites	n/a	1,448
Digital-born	Explicite	expliciteJA	n/a	43,168
Digital-born	The Conversation	FR_Conversation	0.09	11,384
Digital-born	Brief.me	briefmenews	n/a	5,172
Digital-born	Les Moutons Enragés	Moutonenrages	0.21	4,390
Digital-born	Dreuz Info	Dreuz_1fo	0.68	12,509
Digital-born	Riposte Laïque	1RiposteLaique	0.40	9,520
Magazine	Paris Match	ParisMatch	2.84	1,116,503
Weekly Newspaper	Telerama	Telerama	3.40	334,759
Weekly Newspaper	Challenges	Challenges	1.88	196,801
Weekly Newspaper	Les Inrocks	lesinrocks	1.69	1,083,546
Radio	France Culture	franceculture	n/a	437,559
National Daily Newpaper	L’Humanité	humanite_fr	1.11	322,578
TV	CNews	CNEWS	0.19	981,305
National Daily Newpaper	L’Opinion	lopinion_fr	0.52	60,698
Weekly Newspaper	Courrier International	courrierinter	1.16	227,116
TV	LCP	LCP	0.24	370,838
TV	TF1	TF1	13.01	4,817,643
Radio	RMC	RMCinfo	n/a	125,278
Digital-born	Agora Vox	agoravox	0.74	13,379
Weekly Newspaper	La Vie	LaVieHebdo	0.27	22,140
Weekly Newspaper	Témoignage Chrétien	TChebdo	n/a	10,387
Monthly Newspaper	Capital	MagazineCapital	2.23	21,823

Source for online reach: comScore MMX Key Measures, % reach desktop only, average last 3 months, April 2017