Local events work better regarding country branding than bigger ones. At such local events, the opinion leaders are all lovers, and the content of the emoticons is highly positive. Moreover, given that a lack of public administration in social networks has been detected, active involvement is recommended to promote the territorial brand as opinion leaders.

Second, the most appropriate opinion leaders (those oriented towards brand love and not hate) must be identified before the sporting event occurs. Choosing the proper social media opinion leader is not so simple when a country wants to improve its image using social networks. A critical problem is finding a fit between the product (country brand), the influencer-

Third, we recommend different use of social media for country branding depending on the type of event. (i) In mega sports events, social networks could be less recommended for country branding because opinion leaders tend to be haters on Twitter and Instagram. On Facebook and YouTube, everything can happen because there are as many lovers as haters. (ii) Only one social network is recommended for country branding in semi-massive sports events: Twitter. YouTube is not recommended because most opinion leaders on YouTube act like haters, and everything can happen on Facebook and Instagram. (iii) In local sports events, all social networks are recommended for country branding because opinion leaders tend to act as lovers, which means the tone of the message is positive. They are highly recommended on Twitter and Facebook.

Fourth, emojis represent a promising tool for evaluating emotional responses. Our results have shown a positive emotional influence on events hosted in Spain (Davis Cup tennis and Valencia Marathon in our case) and events held outside Spain (World Cup football in Qatar). It is recommended that country brand managers (or city brand managers) support all sporting events by sending good sportsmen and women as ambassadors of the territory they are trying to promote.

METHODOLOGY

Sample

The present study focused on three events of a different calibre: a football mega event (Qatar Football World Cup), a semi-massive tennis event (Davis Cup 2022) and a local marathon event (XLI Marathon Valencia Trinidad Alfonso 2022). Based on Kumar and Sachdeva (2021), four social networks were monitored for two months (1 November to 31 December 2022): Facebook, Instagram, Twitter and YouTube.

All the posts in these events talking about the country brand Spain in different parts of the world in the eleven social networks mentioned in Table 1 were retrieved to obtain data. For each post, the emojis used were counted. In addition, for each social network, we identified (i) the top influencers (those with the highest number of followers), (ii) the opinion leaders (those who published the highest number of posts about our topic), (iii) the lovers (those opinion leaders that published the highest number of posts transmitting a positive sentiment); and (iv) the haters (those opinion leaders that published the highest number of posts transmitting a negative sentiment).

Sentiment analysis was performed based on the social listening tool provided by Atribus, which allows you to analyze various elements of the text that indicate emotion: conversational tone, emoji, punctuation, context, etc. The text analysis platform employed allowed going beyond the black-and-white nature of positive/negative sentiment to identify a specific sentiment. A brand sentiment score or rating was obtained to classify an opinion leader as a lover or hater, expressed as a percentage of positivity or negativity shown in online conversations around a brand. Since an opinion leader’s conversations are rarely negative or positive, social listening tools indicate what percentage of negativity and positivity they show. Hence, usually, opinion leaders are both lovers and haters. Based on Aro, Suomi, and  Gyrd-Jones (2023, hate-love speech detection techniques from social media were applied using classical and deep learning classifiers. This article does not focus on the particular sentiment of each opinion leader, only the signs of their general feeling (positive-lover and/or negative-hater). We have mainly followed a supply perspective. From a demand approach, we analyzed the most used emojis in response to each post and the number of times it was shared, but not the number of likes.

The collection was conducted using the tool Atribus. As a result, we obtained a total of 1,716,800 posts (Tabl 1): 1,700,00 from the mega-event (Qatar Football World Cup), 13,200 from the semi-massive event (Davis Cup 2022) and 3,600 from the local event (XLI Marathon Valencia Trinidad Alfonso 2022). For this study, we focused only on four social networks. So, our final sample was composed of 1,642,319 posts:

• 1,637,058 from the mega-event (Qatar Football World Cup)
• 3,348 from the semi-massive event (Davis Cup 2022) and
• 1,913 from the local event (XLI Marathon Valencia Trinidad Alfonso 2022).

Table 1. Sample of posts analyzed in three different sports events

Source: based on Atribus

All data collected by the platform are public (i.e., social profiles publicly sharing opinions on any digital platform). Therefore, profiles with restricted access cannot be analyzed. The socio-demographic variables (such as age or gender) of the profiles that have issued some opinion are treated in an anonymized aggregate form; therefore, it is impossible to show this data in an individualized format. As in the study of Aro, Suomi and  Gyrd-Jones (2023), access to data was based on user permissions to comply with general data protection regulations and avoid legal penalties.

Statistical Tool

Following Casado-Molina, Rojas-de Gracia, Alarcón-Urbistondo and Romero-Charneco (2022), data was processed to conduct a descriptive study in order to learn more about how opinion leaders act to generate positive or negative communication reactions among users. Atribus ad hoc software was used to monitor the four social networks.

To determine whether a user is a lover or a hater, we analyze the number of opinions issued by each user to detect which users have been the most active for each digital platform. Once these opinions were analyzed, we determined negative, positive or neutral opinions through sentiment analysis. Sentiment detection in social listening data involves analyzing the language used on social networks and the Internet to determine the opinion or attitude expressed (Thelwall, Buckley and Paltoglou, 2012). To do this, we employ natural language processing (NLP) algorithms to analyze the sentiment or tonality of the message expressed in each post or comment.

We relied on Bayes’ theorem for the sentiment detection process to calculate the probability that a given mention had a positive, negative or neutral sentiment (Gamallo, 2014). As such, we assign weights or scores to keywords or semantic features most strongly associated with a specific sentiment. These weights can influence the calculation of conditional probabilities and, thus, the sentiment estimation. For example, highly emotional or emotional sentiment is more likely to be associated with a specific sentiment (Gamallo, 2014). The assignment of weights to each mention, depending on the semantics used, was performed using different approaches (i.e. word dictionaries with predefined sentiment scores, the training of machine learning models to learn the relative importance of other features or the manual adjustment of weights according to expert knowledge).

Key informers identification

Based on previous literature (Ouvrein, Pabian, Giles, Hudders and Backer, 2021), and as introduced in previous lines, top influencers were identified based on their popularity (number of followers). A leading influencer has given their opinion about the topic by posting one or an enormous number of publications.

On the contrary, opinion leaders are defined based on the number of posts published (Ouvrein, Pabian, Giles, Hudders and Backer, 2021). That is, an opinion leader is a person who has posted the highest number of posts about the topic (mentioning Spain during a sports event). Opinion leaders are, therefore, those who publish the most (regardless of the number of followers they have).

Lastly, brand lovers and haters are opinion leaders correlated with positive and negative sentiments (Ouvrein, Pabian, Giles, Hudders and Backer, 2021). In other words, those users who have published the most posts with for or against message tones when talking about an event mentioning the country Spain.

Emojis count

Following Vidal, Ares and Jaeger (2016), the frequency of use of each emoji was determined by counting the number of posts that contained it. Frequencies were calculated at the aggregate level for each sporting event.