World Cup 2014 in the Twitter World: A big data analysis of sentiments in U.S. sports fans’ tweets

Highlights

• We used a big data approach to analyze U.S. soccer fans’ sentiments in their tweets.

• Tweets were used to express joy and anticipation and to express emotions.

• U.S. fans’ fear and anger were common and reflected U.S. team’s goals or losses.

• Emotions in two non-U.S. games were unclear.

• This paper provides support for disposition theory of sport and sentiment analysis.

Abstract

The present project collected real-time tweets from U.S. soccer fans during five 2014 FIFA World Cup games (three games between the U.S. team and another opponent and two games between other teams) using Twitter search API. We used sentiment analysis to examine U.S. soccer fans’ emotional responses in their tweets, particularly, the emotional changes after goals (either own or the opponent’s). We found that during the matches that the U.S. team played, fear and anger were the most common negative emotions and in general, increased when the opponent team scored and decreased when the U.S. team scored. Anticipation and joy were also generally consistent with the goal results and the associated circumstances during the games. Furthermore, we found that during the matches between other teams, U.S. tweets showed more joy and anticipation than negative emotions (e.g., anger and fear) and that the patterns in response to goal or loss were unclear. This project revealed that sports fans use Twitter for emotional purposes and that the big data approach to analyze sports fans’ sentiment showed results generally consistent with the predictions of the disposition theory when the fanship was clear and showed good predictive validity.

Introduction

On July 8, 2014, during the FIFA World Cup semi-final game at the Estadio Mineirao in Belo Horizonte, the host Brazil played the Germans. The German brilliance transformed into multiple goals and scored seven times and won the game by a wide margin of 7-1. The faces of the Brazilian fans in the stadium were captured by the camera and broadcast worldwide. Their faces turned from anticipation early in the game to surprise to sadness and to desperation after the many German goals. In contrast to the Germans’ joy and celebration, the sobbing, mourning, and desperation of the many Brazilian fans, young and old, men and women, are probably still vivid among the many who watched the game.

Admittedly, sports fans’ emotions change during games. Their faces and their body movements probably are natural ways to express their emotions. With the wide adoption of mobile devices and the easy access to the Internet, can sports fans’ emotions manifest in their writing on social media and through their interaction with others through the Internet and mobile devices? Based on Raney’s (2006) summary of sports fans’ motivations to consume mediated games, recent research in social media and electronic communication indicates that sports fans are motivated to use social media for a variety of purposes, including emotional release (e.g., Wang, 2013, Wang, 2015). Such research, although informative, is often based on a survey method and provides only correlational data related to motivations and intentions. Although significant, the indirect relationship between using the social media for emotion release and intention was around .10 in Wang (2013), indicating that motivations may not translate into intentions or behavior. Such research has yet to be confirmed by an analysis of sports fans’ behaviors, manifested by their tweets or messages on social media.

To that end, the present analysis examines sports fans’ emotions and behaviors on the social media through a big data approach to examine viewers’ responses to sports programming.

At the methodological level, we propose that instead of using a lab experiment, we can measure viewers’ emotions and responses through social media or using social media messages as media users’ responses to television shows or mediated sports programs. If results of our analysis are in line with our expectation of how fans would react based in disposition theory, it indicates that the social media are used for game-related emotions and that the big data analysis, as a fairly new text analysis method, is valid based on criterion validity. At the theoretical level, we examine whether the disposition theory of sports spectatorship can be used to examine cyber behaviors; that is, whether the emotional reactions to the games, manifested in tweets, change as a function of the game.

The present study focuses on the tweets with a location stamp from the United States, whereas the games took place in Brazil. Essentially, this means that the authors of the tweets viewed mediated games or gained information through various mediated sources instead of attending games in a stadium. Mediated sports are one form of entertainment and a source of enjoyment for sports fans. Because many people, including sports spectators, are probably chronically understimulated, they seek content to “psych themselves up” (Gantz, 1981). Raney (2006) called this “eustress” and stated that sports games provide a way for sports fans to experience emotional arousal or release emotions. Emotional responses can be both positive and negative.

In general, disposition theory states that sports enjoyment is a function of the content and one’s disposition (Bryant et al., 1981, Zillmann et al., 1989). This theory defines the viewers’ or participants’ feelings and attachments toward the team as affective disposition. In sports, one’s disposition toward a team is called fanship. Affective disposition determines one’s reactions toward and enjoyment of the sports events. The affective disposition toward a team ranges from intense liking to intense disliking. More specifically, the theory predicts that participants’ enjoyment of the event or games increases when their team or liked characters are linked to positive outcomes and their disliked teams experience negative outcomes, and their enjoyment of the events or games decreases when their team or liked characters experience negative events. If viewers are indifferent to a sports team, they are probably not excited or angered by that team’s win or loss.

More specifically, Raney’s (2006) stated that viewers view mediated sports as a means to be stimulated and to release emotions. First, sports games, as an entertainment source, may bring the spectators joy and happiness; the enjoyment or happiness was the highest when one’s liked team won and when one’s disliked team was defeated. Further, Zillmann et al. (1989) reviewed several studies and found that respondents’ enjoyment of the games changed over the course of game and participants rated the game more enjoyment when their team scored and rated the game less enjoyable when the opponent team scored. The latter may lead to anger and sadness among the spectators. Second, as a source to release eustress, the exciting, suspenseful, and possibly violent nature can get fans aroused, excited, and be thrilled by the victory (e.g., Bernhardt, Dabbs, Fielden, & Lutter, 1998; see also Raney, 2006 for a review). That is, sports fans’ emotional reactions during the game will vary as a function of the content and their own disposition.

Recent research found that the use of social media can be used by emotional reasons as well. However, the research in this area is still limited. Our extensive literature review showed that using a survey method, Wang (2013) analyzed the motivations why sports fans used social media while watching mediated games and found that motivations related to using social media to release emotions had a significant, but weak or moderate relationship with sports fans’ game enjoyment and their intentions to use social media during mediated games. As a cross-sectional study, Wang’s study did not measure sports fan’s actual behavior (i.e., authoring tweets) and as such did not provide evidence whether the sports fans indeed used the social media for emotional release purposes. To build on previous research, it is important to provide evidence for one’s actual use of the social media.

One way to examine sports fans’ emotional behaviors on social media is to use the traditional content analysis method which relies on human coders to identify the emotions that manifest in the tweets. In the age of a vast number of social media messages and posts, the traditional method seems to be limited because it usually deals with a small number of sampled messages. In the present research, we used a “big data” approach to analyze the sentiments of the sports fans.

Sentiment analysis analyzes people’s emotions, attitudes, or opinions toward various products or issues (Liu, 2012). Sentiment analysis and related research has increased considerably in the past decade. Existing sentiment analysis approaches are either based on linguistic resources or on machine learning. Sentiment analysis based on linguistic resources is centered on predetermined lists of positive and negative words and is more commonly used than machine learning (Taboada, Brooke, Tofiloski, Voll, & Stede, 2011). By counting how many times a word appears, this approach recognizes words with positive polarity (expressing a favorable sentiment toward an object), negative polarity (expressing an unfavorable sentiment toward an object), and no polarity (neutral). It also detects words with specific emotions or moods such as joy, sadness, and anger. As such, sentiment analysis analyzes the words directly and avoids the traditional, costly, and time-consuming content analysis. Sentiment analysis has been applied to a variety of field such as business, education and politics (Ceron et al., 2014, Pang and Lee, 2008, Tumasjan et al., 2010).

Twitter is a widely used social media platform and is a popular second screen. One distinctive feature of Twitter is that users can update contents instantaneously and frequently. Ji and Raney, 2014, Wang, 2013 stated that users use tweets to share their real-time reactions and emotions. Thus, tweets can reflect users’ thoughts in a real-time fashion and are a source of users’ real-time activities and social media use. Tweets can be used to explain, detect, or predict various phenomena (Kalampokis, 2013). A growing number of scholars have used Twitter to examine social media sentiments (Liu, 2012, Pang and Lee, 2008, Yu et al., 2013) or users’ responses. In a recent study, Ji and Raney (2014) examined the role that morality played during the consumption of TV entertainment. Ji and Raney examined the tweets that viewers tweeted during and after viewing the Season 3 finale of the British drama Downton Abbey. Ji and Raney found that viewers tweeted more tweets related to the young protagonist Matthew who died during the finale (a major event) compared to other characters and that more morality-related tweets were found for those characters who were involved in moral events than those who were not. The results showed that tweets were consistent with the contents and events in the TV drama and can be used as a way to examine viewers’ reactions.

The present analysis adopted a “natural experiment” approach to examine how U.S. sports fans reacted during the games when the U.S. National Soccer team competed in the FIFA World Cup 2014. We then used the sentiment analysis to examine the emotions manifested in the tweets (i.e., fans’ reactions). The analysis was limited to the English language tweets with a location stamp originated from the United States instead of using all tweets in English language because of the following reasons: Previous literature review (e.g., Raney, 2006) indicates that fans’ emotional reactions are determined by their affective disposition (fanship) and the content of the event. We assume that tweets in English from the United States would primarily root for the U.S. team. On the other hand, we are not certain about the teams that the authors of other English tweets were associated with. It should be acknowledged that not all English tweets with a U.S.-based location stamp were tweeted by a fan of the U.S. Soccer Team. The tweets could be potentially tweeted by for example, Germans or British, living in the United States. However, the number of these foreign dwellers is small compared to the number of U.S. fans. That is, we acknowledge the potential bias that foreign dwellers may bring into our data, but we choose to use U.S. location stamps as a proxy for U.S. fanship.

Second, we previously reviewed that sports fans’ emotional reactions and enjoyment change according to the win and loss of their own team (Bryant et al., 1981, Raney, 2006). As such, U.S. sports fans’ affective disposition toward the U.S. soccer team will lead them to experience various emotions and enjoyment that are results of the win or loss of goals of the U.S. team. Fans would be excited and experience positive emotions (e.g., joy and excitement) when their own team scored and would experience negative emotions (e.g., anger, disappointment) when the opponent team scored. The tweets during and shortly after the games can be correlated with the win or loss of the U.S. team because we previously reviewed evidence that tweets can be indicative of viewers’ reactions (Ji & Raney, 2006). On the other hand, the disposition theory states that if the sports spectators are indifferent, they are less likely to experience anger or joy. For example, the U.S. fans would less likely to be invested on the game between two unrelated teams, for example, a game between France and Nigeria on June 30, 2014. Thus, we hypothesized the following:

H1

Tweets with a U.S. location stamp (a) would show negative emotions (e.g., fear, anger) when the U.S. soccer team’s conceded a goal and (b) would show positive emotions (e.g., joy, hope) when the U.S. soccer team scored a goal.

H2

Tweets with a U.S. location stamp will remain “indifferent” to the win or loss of other teams.