INTRODUCTION
News organizations are often hailed as gatekeepers of democracy, responsible for maintaining public trust and staying impartial in their quest to deliver truth and information. However, an expanding body of evidence suggests that news reporting is becoming more polarized, influenced by political ideologies (Aggarwal et al., 2020; Flamino et al., 2023; Goldman et al., 2023). Political bias in the media, especially when amplified on social media platforms like Facebook and X (formerly known as Twitter), significantly influences the public’s attitudes and behaviors, as seen in the 2016 presidential election (Flamino et al., 2023). Furthermore, politically polarized news sources tend to introduce increased negativity into their content to align with their audience’s preferences (Bellovary et al., 2021). This media polarization undermines our democratic processes, fuels misinformation, and erodes public trust.
Politically polarized news coverage of financial crises can also yield far-reaching consequences. While polarized reporting of financial news can influence trading behaviors and disrupt markets (Goldman et al., 2023), the impact of biased and negative reporting of financial crises can be even more profound. As illustrated in Mohiuddin et al.'s (2016) study, excessive polarization and a negative tone in coverage by both Democratic and Republican-leaning editorials during global financial crises can breed distrust and frustration among the general public and hinder future societal progress. Despite the importance of this issue, there is no clear understanding regarding the presence and extent of political polarization in news media’s reporting of financial crises.
In this study, we aim to investigate political polarization in the recent financial coverage of the significant collapse of Silicon Valley Bank (SVB). For this purpose, we examined the sentiments expressed on Twitter/X related to the SVB crisis by news organizations of different political leanings using a sentiment analysis approach. Sentiment analysis is a natural language processing technique that determines whether the tone of the data is positive, negative, or neutral, where a higher negative or positive sentiment score denotes stronger emotional content. By analyzing the sentiments expressed in news articles, tweets, and other forms of media content, researchers have gained insights into biases and political polarization within the news media. Various studies have shown that politically polarized news coverage that appears to be highly subjective or biased (Aggarwal et al., 2020) or strongly emotional (Bellovary et al., 2021) tends to have high sentiment scores (showing strong positivity or negativity). We hypothesized that news organizations with political leanings (left or right bias rating) will have more emotional content, hence higher sentiment scores than news organizations with neutral ratings while reporting financial crises on their social media handles. Social media platform Twitter/X was selected for this study as it has proven to wield substantial influence, at times surpassing traditional media, in connecting and engaging users with finance-related news (Milas et al., 2021).
MATERIALS AND METHODS
We identified the political orientations of news organizations by referencing the AllSides Media Bias Ratings (2019). We categorized a total of 35 news outlets into five distinct groups, each comprising seven members: far-left, left-leaning, center, right-leaning, and far-right.
Sentiment Analysis
We collected tweets related to the Silicon Valley Bank crisis that had been posted by these news organizations from March 1, 2023, to April 4, 2023, through data from the Kaggle platform. To gather these tweets, the Kaggle dataset utilized Snscrape to extract data from the public Twitter/X APIs, employing the search query ‘silicon valley bank.’ In total, we amassed 279,000 tweets during this specific time frame.
To gauge the sentiment and tone of the news coverage surrounding the crisis, we conducted sentiment analysis on the dataset using Python. This analysis entailed the utilization of the NLTK (Natural Language Toolkit) library and the VADER (Valence Aware Dictionary and Sentiment Reasoner) lexicon tool. VADER, developed by Hutto and Gilbert in 2014, is a lexicon and rule-based sentiment analysis tool designed to capture sentiment expressions within social media text. VADER assigns polarity scores (PS) on a scale that ranges from -1 (indicating strong negativity) to +1 (indicating strong positivity). This scale is further categorized into three groups: positive, negative, and neutral. Negative PS values correspond to scores ranging from -1 to -0.33, while positive PS values are associated with scores ranging from 0.33 to 1. Scores that fall between -0.33 and 0.33 are classified as neutral PS values.
Word Clouds
A word cloud provides a visual representation of text, where words that appear most frequently are displayed in a larger and more prominent format. To capture the most common themes, we created word clouds using the Python Matplotlib and word-cloud libraries. Before analysis, we removed stop words, symbols, numbers, and other unnecessary terms and converted all characters to lowercase. We employed the Collections and regular expressions libraries to count the most frequently used words and their respective frequencies.
RESULTS
Sentiment Analysis
Mean scores: Since the subject of coverage revolved around a financial crisis, the majority of tweets shared by the selected news organizations resulted in negative sentiment scores. In other words, the overall tone of the posts leaned toward negativity. Figure 1 illustrates the average sentiment scores for five news outlet groups, categorized based on their political orientation (Far right, Far left, Center, Right- and Left-Leaning). We conducted a one-way ANOVA to examine the influence of political orientation on sentiment scores. The analysis revealed that there was no statistically significant difference in sentiment scores among the five groups categorized by their political leanings (single-factor ANOVA, F (4, 30) = 1.47, p = 0.2362). Despite the absence of statistically significant variations between the groups, it’s worth noting that, on average, the far-right exhibited the most negative scores among the five political leanings, while the center displayed the least negativity.
Individual differences: It’s crucial to emphasize that a significant variation exists in the degree of negativity when reporting among news organizations sharing the same political leanings. In Figure 2, you can observe the sentiment scores of each news organization’s Twitter/X handle regarding SVB posts. The longer the bar lines, the more negative the tone of the news source. A cutoff (dotted line) that was precisely set one standard deviation out from the mean of center-leaning news organizations—the top16% of news sources with extreme negative sentiment—was used to depict the news outlets with the most negative posts on the SVB crisis. News organizations with scores between -0.308 and -0.55 indicate a significantly higher level of negativity and a significant deviation from the dataset’s average sentiment levels. Based on this cutoff, we identified 5 far-right, 2 far-left, 2 left-leanings, 1 right-leaning, and 1 center organization(s) that exhibited a more negative tone than the mean plus one standard deviation of the center outlets. Conversely, Newsweek (center) and the SF Chronicle (far-left) displayed a less negative tone than the mean plus standard deviation of the center news outlets. As depicted in Figure 2, 5 out of 7 far-right news organizations are situated at the far end of the sentiment negative score spectrum, signifying a greater use of a negative tone in their coverage of the financial crisis.
Word Clouds
While word clouds may not qualify as a quantitative analysis, they offer a valuable approach to inspecting and understanding the most frequently used words in posts, thereby capturing recurring themes in news outlets’ reporting. In Figure 3, we present the word clouds for right (red), left (blue), and center (gray) news outlets during the SVB crisis. To simplify, we amalgamated far-right and right-leaning data, as well as far-left and left-leaning data, to generate word clouds for right and left, respectively. Words that appear more frequently in a dataset are depicted as larger and, at times, bolder.
__left_(blue)_and_center_(gray)_-_leaning_news_organization_gen.png)
It’s worth noting that the terms “collapse” and “failure” emerged as the most commonly used words across all three political-leaning news outlets. To explore the emotional dimension further, we singled out emotionally charged words within the extracted tweets. In this context, right-leaning news outlets exhibited the highest frequency of negative words (489), followed by center (465), and left (347). Nevertheless, there was no substantial difference in word frequency across the political-leaning news outlets.
DISCUSSION
The recent financial crisis, notably the dramatic collapse of Silicon Valley Bank (SVB), captured the world’s attention and was extensively covered by mainstream newspapers and their social media outlets. X, in particular, played a pivotal role in the unfolding SVB crisis, as acknowledged by various studies (Bales et al., 2023; Khan & Anupam, 2023). Often referred to as history’s first Twitter-fueled bank run, these events underscored the immense power and influence of social media.
Our study aimed to investigate whether the sentiment expressed in news coverage via social media channels during the crisis was affected by the political affiliations of news organizations. Our findings reveal that, on average, there was no statistically significant difference in the sentiment conveyed by news organizations, irrespective of their political leanings. Even far-right and far-left-leaning news outlets, on average, exhibited a similar degree of negativity in their coverage. Similar results were reported in studies where sentiment analysis was conducted on mainstream non-financial news across news organizations with varying political orientations (Dalal et al., 2019; Flamino et al., 2023).
One possible explanation for this shared negative sentiment across different groups is the phenomenon that negative news tends to spread rapidly and elicit higher engagement. It’s a commonly employed strategy by news organizations, regardless of their political affiliations, to connect with and captivate their core audience. Given that news outlets strive to cater to the preferences and sensitivities of their viewers or readers, outlets representing extreme ideological positions, such as far-right and far-left, often incorporate a heightened level of negative emotionality in their content as a means to engage their audience (Bellovary et al., 2021; Mohiuddin et al., 2016). This common sentiment was also evident in the word cloud, where a prevailing theme shared among right, left, and center-leaning news organizations revolved around words like “collapse” and “failure.”
Furthermore, there might be a collective awareness among news organizations when addressing the financial crisis. It’s conceivable that certain events, like a financial crisis, can surpass ideological differences and evoke shared emotional reactions. This underscores the idea that, when a situation is perceived as a substantial challenge to deeply held beliefs, regardless of political orientation, it can generate a common response. On a positive note, the lack of a substantial overall difference in negativity may also suggest that media outlets are making efforts to deliver balanced and objective reporting.
However, even in the absence of a substantial overall difference, it’s crucial to acknowledge the significant individual variability in negativity scores within news organizations sharing the same political leanings. The variability may be due to the existence of a diverse range of political beliefs, perspectives, and reporting approaches within the media landscape. The higher prevalence of negativity in far-right-leaning news organizations may suggest that outlets with clear ideological stances tend to emphasize negativity that aligns with their beliefs, thereby catering to the preferences of their core audience. For example, even if there might be some shared perception when dealing with a financial crisis, news discussions can be influenced, to some extent, by whether a Republican-dominated or Democratic-dominated government is in power. During the SVB crisis, a Democratic-led government was in control, potentially leading to significantly different sentiments expressed by the far-right outlets compared to the center, aligning with their conservative audience’s preferences.
CONCLUSION
Our current study highlights that the negative tone in the coverage of the SVB crisis by various news organizations is not overtly influenced by political ideologies. However, it is notable that many news outlets tend to adopt an excessively negative tone when discussing financial crises. Given the pivotal role of the news media in shaping public attitudes and decisions, there’s a growing need for a commitment to fair and impartial reporting.
It’s important to acknowledge that sentiments can evolve over time, and the sentiment scores observed during the SVB crisis may not be representative of sentiments in different timeframes or regarding unrelated matters. Generalizing these findings to other financial contexts should be done cautiously. Further research is required to explore the factors contributing to the observed variability within each media group. This can provide valuable insights into media reporting dynamics during financial crises and other significant events.
AUTHOR’S NOTE
I have no conflict of interest to disclose.