2.1 The tragic incident and initial managerial acknowledgement

The tragic incident led to an intensive debate on the role of insurance firms and industry insiders, clients, and the general public started to reflect on the current practices of the insurance industry. Following the incident, CEO Andrew Witty of UnitedHealth Group – the parent organization that UnitedHealthcare is governed by – gave an originally private speech to the company saying that the company is a positive exemplar in the American healthcare system and will protect their business practices that ‘guard against unsafe or unnecessary care’ [5]. However, Witty outwardly acknowledges that many Americans frustrated with the perceivably inadequate healthcare system in place, and that there is definitely progress to be made. In his first public statement since the shooting, titled “The Health Care System Is Flawed. Let’s Fix It.”, he wrote that the healthcare insurance system is a ‘flawed patchwork built over decades’ and that it is ‘clear we have not reached a perfect system yet’ [6]. He later states that while they are partly responsible for some controversial coverage decisions, it is truly the purpose of their organization to ‘build a better healthcare system that works for everyone’ with the help of the government and community.

While there had been ongoing discussions on current practices in the insurance industry and political debate on how to restructure the policies and regulations, after a brief stable period, stock prices continued to decline for most insurance industry players even after two weeks of the unwarranted event of a CEO assassination. We focus on this aspect in the following sub-section.

2.2 Stock market movement following the CEO assassination

The stock market reactions to the tragic event of CEO assassination evolved quite strangely, as the market initially mainly focused on the recent disclosures and financial health of the company and largely ignored the implications of CEO killing and potential public outrage. The day that Brian Thompson passed away itself, it was a normal day for UnitedHealth Group Inc. in terms of their stock prices. The stock as almost completely flat, opening at $611.02 and closing at $610.79 [2]. This was a peculiarly stable performance, despite the assassination occurring a few hours prior to the stock market opening that morning; the stock value was even up by around a point for most of the day. It appears that the financial market was more focused on the financial aspects rather than the unfortunate event of an insurance giant’s CEO assassination. On December 3^rd, UnitedHealth Group (NYSE: UNH) issued financial guidance ahead of its annual Investor Conference which was scheduled on December 4, beginning at 8:00 a.m. ET [14]. In that press release, the group projected 2025 sales of between $450 billion and $455 billion, which beats analysts’ consensus estimates of $431 billion. The financial market was content with UnitedHealth Groups’ future prospect and largely ignored the news of CEO assassination.

Meanwhile, the fatal shooting prompted a huge outpour of emotion in the general population [15]. Americans expressed their frustration, helplessness, and hostility towards insurance companies, some sharing personal anecdotes of their interactions with inscrutable mega-corporations. The incident unleashed a crashing wave of negativity not only onto UnitedHealthcare, but towards the insurance – and more specifically, the health insurance – industry as a whole. On social media, many posts criticized the strong profitability of health insurance companies in light of individuals’ reported challenges with coverage and reimbursement.

Due to this public outcry, UnitedHealth Group started experiencing difficulties the day after the murder [16]. Its share price declined sharply and continues to fall, losing more than $110 billion in market value by December 17^th. The Dow Jones Industrial Average fell from 45,014.04 to 43,449.90 points during the same period, with a large portion (804 points) of the loss largely due to UnitedHealth Group and thus the assassination’s online reception. Furthermore, other health insurance companies are also experiencing similar losses; CVS Health Corporation decreased by 25%, Cigna Healthcare dropped by 20%, and Humana fell by 19%. All of these companies saw a monetary loss by the billions.

2.3 Problems in insurance industry

2.4 Negative public sentiment: social media outcry and stock market reactions

In general, literature on news media and social media show that media content can sway investor sentiments and affect stock returns. Tetlock [21] find that negative news media coverage can robustly predict downward pressure on market prices. The finding suggests that media coverage serves as a proxy for investor sentiment. Tetlock et al. [22] posit that news media could contain certain soft information which is not captured by other publicly available information. Similarly, Dougal et al. [23] also find that financial journalists have the potential to influence investor behaviors.

A number of studies have also examined the relation between social media contents and stock market (e.g., [24–26]) and generally report that a higher proportion of negative social media posts/comments about a company leads to lower stock returns [27].

The assassination of UnitedHealthcare CEO on December 4^th led to a huge public outrage on social media platforms mainly criticizing the predatory practices and opportunistic behavior of insurance companies. Traditional news media also covered the story extensively and focused on the flawed practices in the insurance industry. Especially, the aftermath of Brian Thompson’s assassination saw an eruption of reactions across social media platforms, including TikTok, Instagram, X, Reddit, and Facebook; this was further concentrated and reported by news outlets and independent bloggers [8]. This in turn triggered a drop in UnitedHealthcare’s stock value.

For the first few hours after the incident and general reporting across news outlets, stock remained stable; however, the situation could not be more different the following days. The overarching corporation, UnitedHealth, lost 5.2% of its value, decreasing further as there was an outburst of vicious comments on social media, prompting the publication of multiple negative articles on UnitedHealthcare’s corporate practices [2]. Even as the company tried to salvage its image through executive announcements and appeals made with the help of their employees, it became evident that the damage was already done [28]. Social media saw a range of responses, from some users expressing sympathy for the CEO (primarily on professional networks such as LinkedIn), to others justifying the intent of the shooter [29,30]. Some posts were so vicious that X was compelled to remove them swiftly. In addition to such polarized perspectives, another significant portion of discourse surrounded users irately recounting their own experience with the healthcare insurance system.

Consequently, such negative sentiments shared through social media platforms and highlighted in traditional news media led to a meltdown of the insurance companies’ stock prices.

2.5 Relevant theories

2.6 Hypothesis development

As we have discussed in the earlier sections, following the assassination of the UnitedHealthcare CEO, insurance companies attracted negative publicity for profit-oriented business practices and insufficient attention to their clients’ needs. Post-assassination, social media platforms’ public outcry and conventional news media (e.g., newspapers) coverage analysis suggest that the public outrage at insurance companies is primarily attributed to ‘corporate profit and glamour’ and ‘corporate leaders’ privilege and personality traits’ – which could subsequently affect stock returns. The views are also in line with the ‘social banditry theory and investor sentiment’ – which implies that public outrage is directed more towards the riches and prominent entities and persons – and the ‘stakeholder theory’, which propagates that focusing narrowly on shareholder interests and corporate insider benefits could cost the shareholder value eventually.

Below, we discuss both aspects in more details, and present relevant hypotheses. We also performed a complementary social media analysis using Reddit comments (presented in Section 3), which reinforces the rationale behind our hypotheses.

3.1. Text data collection and BERTopic model implementation

We collected our comments data from Reddit (using Reddit API), comprising 69,726 original entries from posts with the query “Brian Thompson,” “Luigi Mangione,” and “UnitedHealthcare” between December 18 and 23, 2024. The posts we collected also cover three subreddit groups — “HealthInsurance,” “BrianThompsonMurder,” and “Luigi_Mangione” — capturing focused community discussions related to the incident. The data covers a substantial time range from December 4, 2024, to December 23, 2024, ensuring contemporary perspectives on the event. This broad scope provides a comprehensive basis for analyzing public opinion and thematic evolution over time.

Before performing the topic modeling analysis, the collected Reddit comments were subjected to a range of preprocessing techniques designed to improve the quality and suitability of the data for analysis [64]. These preprocessing methods were implemented to ensure the data was properly formatted for analysis and to enable BERTopic to accurately identify and interpret the semantic structures and meanings embedded in the text [63]. For data cleaning, the textual data was processed to remove emoticons, digits, punctuation, hyperlinks, redundant words, non-ASCII characters, and stopwords, as recommended by Xu et al. [65]. The Python scikit-learn library includes an English stopwords list, which comprises common words in English that carry minimal semantic or informational significance. These stopwords are usually removed in natural language processing tasks to improve the efficiency and precision of text analysis. Moreover, all uppercase letters in the dataset were converted to lowercase. To address potential redundancy due to overlapping data sources, duplicate texts were also filtered out. As a result, a total of 59,644 entries were retained for analysis.

We then apply the BERTopic model, which utilizes advanced language models to understand words within their specific contexts [64,66]. The capability of BERTopic allows it to identify intricate relationships among words, leading to enhanced accuracy in extracting topics from textual data. When applied to tasks like document clustering and topic assignment, BERTopic has shown greater effectiveness compared to traditional approaches [67]. Traditional topic models such as LDA and NMF often face challenges when applied to short and noisy texts, whereas BERTopic demonstrates superior performance in handling such data effectively [64]. Research has consistently shown that BERTopic surpasses LDA and NMF in terms of producing coherent and well-separated topics [64,67]. This suggests that BERTopic generates topics that are clearer and more logically structured. Additionally, BERTopic offers remarkable flexibility, making it suitable for analyzing diverse datasets, accommodating varying document lengths, and supporting multiple languages [68,69].

The BERTopic methodology, as outlined by Grootendorst [62], consists of three primary steps: generating numerical representations of documents through embedding, grouping documents into clusters, and applying the c-TF-IDF technique to analyze the clustered data. We use Sentence-BERT (SBERT) to transform preprocessed reddit comments into 384-dimensional embeddings, capturing semantic similarity [70]. To address the sensitivity of clustering algorithms like HDBSCAN to high-dimensional data [71], we apply Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction. UMAP effectively preserves the structural properties of data, similar to PCA and t-SNE [72]. When implementing HDBSCAN, we specified a minimum cluster size of 1,000 comments. As a result, 9 meaningful topics were identified, but 39,549 comments were classified as outliers. To reduce these outliers, we employed an approach based on cosine similarity, which was also utilized by Menezes et al. [63]. Using this method, all outliers were successfully reassigned to the 9 clusters.

3.2. Topic modeling

The shooting incident involving the health insurance CEO has sparked diverse discussions across social and political domains. Using BERTopic, we identified nine distinct topics that encapsulate the major themes of public discourse as shown in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Representative words and theme of each topic. This table presents the representative words and themes of the 9 topics clustered by the BERTopic model. Hypothesis relevance highlights how these topics provide support and context to our hypothesis development.

https://doi.org/10.1371/journal.pone.0334399.t001

Public outrage toward the health insurance industry is directly targeted at its profit-seeking focus on maximizing shareholder returns. This sentiment is captured in Topic 1 which emphasizes systemic issues such as high “profit,” “cost,” and “premium,” alongside calls for “universal healthcare.” Many interpret the shooter’s actions as a response to the perceived exploitation within a model that prioritizes shareholder profits over patient care. For example, comments like “All the patients being denied healthcare for the sake of shareholder return are people too”,“Health Insurance profit is blood money”, and “Does scaring greedy CEOs count as terrorism?” highlight the public’s anger toward an industry widely criticized for valuing profits over individuals’ health and well-being. Discussions emphasize the inequities of the current system, which calls for reform that prioritizes “nonprofit” or government-supported solutions. This topic highlights the tension between corporate profit focus and public health, using the shooter’s actions as a lens to critique systemic failures. The public outrage discussed in Topic 1 is particularly relevant to Hypothesis 1a, which posits that insurance firms with higher profits and revenue growth will face more negative cumulative abnormal returns (CARs). The negative sentiment toward profit-driven practices, as evidenced in this topic, aligns with the hypothesis that firms perceived as prioritizing financial gains over client welfare are subject to harsher market reactions.

It may be argued that social media analysis (e.g., analysis of Reddit comments) alone is insufficient to capture the full spectrum of sentiment influencing the market. Indeed, this is why we extend our analysis beyond social media and include Hypothesis 1b, which is rooted in literature and focuses on the role of traditional media coverage in shaping stock market reactions. Previous studies suggest that firms attracting greater media attention are often perceived as ‘glamorous’ [49] and are more likely to draw public scrutiny and anger during negative corporate events.

Topic 2 captures public sentiment surrounding alleged shooter Luigi Mangione’s actions, integrating keywords such as “support Luigi,” “hero,” and “shooter,” which indicate that certain social media users expressed support for him, alongside terms like “CEO,” “million,” and “Thompson,” which reflect heated debates over corporate leadership. This support underscores widespread dissatisfaction with the healthcare system, with many perceiving it as prioritizing corporate profit over patient care. Comments like “Brian Thompson made 10.2mil a year through UHC including stocks, comps, salary etc.” and “Most C.E.O.s are narcissist. You can only climb to the top if you lack empathy” illustrate public frustration with high executive compensation and narcissistic leadership traits. The keywords in this topic, blending support for Luigi Mangione with criticism of CEOs and their earnings, directly connect to Hypothesis 2, which examines how executive characteristics, such as high compensation and narcissism, influence stock market reactions. This topic encapsulates the intersection of public support, anger at corporate leadership, and broader systemic frustrations with the healthcare industry.

Furthermore, a portion of the public has also focused on the patient group. In topic 3, keywords such as “provider,” “patient,” “hospital,” “billed” and “denial” suggest a focus on patient experiences and frustrations with healthcare services. The shooter’s actions have prompted many to share personal stories of denied claims, excessive medical bills, and inadequate access to care. This topic connects individual struggles with broader systemic issues, making the Luigi Mangione case a symbol of collective frustration. This topic is particularly relevant to Hypothesis 3, which posits that the shooting incident will have a more severe impact on the cumulative abnormal returns (CARs) of insurance companies in the ‘Hospital & Medical Service Plans’ industry segment. This segment includes firms such as UnitedHealthcare, Elevance Health, and their peers, which are the companies people typically turn to when dealing with medical expenses. Because of their direct role in managing healthcare payments and their visibility in patients’ financial and medical lives, these firms are uniquely positioned at the center of public dissatisfaction with healthcare costs and accessibility. As a result, companies in this segment are likely to experience the most significant market impacts, as they bear the brunt of the public’s anger and criticism following events like this one.

Similarly, the healthcare system itself comes under scrutiny in topic 4, as seen through words like “medicaid,” “income,” “deductible,” and “premium.” Public frustration with the complexity and financial burden of health insurance is evident. The shooter’s actions have further amplified calls for reform, with many tying their personal struggles with insurance companies to the broader systemic issues highlighted by the incident. This topic is also relevant to Hypothesis 3, which posits that the effect of the shooting incident on cumulative abnormal returns (CARs) will be more severe for firms in the ‘Hospital & Medical Service Plans’ industry segment. For industries tied to medical expenses, particularly those addressing regular healthcare needs such as doctor visits, hospital stays, and surgeries — the stock price declines are expected to be more pronounced, as these sectors face heightened public anger and criticism.

The Luigi Mangione case has also ignited broader discussions about societal and political tensions. Topic 5 explores the judicial proceedings surrounding the alleged shooter, focusing on terms like “jury,” “trial,” “attorney,” and “degree murder.” Many comments discuss jury nullification under this topic, suggesting it as a means to acquit the perpetrator despite legal guilt, reflecting widespread frustration with the healthcare system. For example, one commenter stated, “We need to spread the word about jury nullification - you can’t vote to find him not guilty,” while comparing Luigi’s actions to those against larger societal harms, emphasizing their anger toward healthcare companies. The legal debates emphasize the balance between justice for the crime and understanding the societal context of the shooter’s actions. Other topics, such as Topic 6, focus on the details of the crime. Keywords, such as “backpack,” “mask,” “bus,” and “dna”, point to discussions about the logistics and evidence of the crime. These details provide the foundational narrative for public discourse, focusing on how the crime was committed and the immediate response by law enforcement. In addition, in topic 7, terms like “violence,” “revolution,” “protest,” and references to leaders such as “biden” and “trump” indicate that the incident has become a flashpoint for broader frustrations with systemic inequality and government inaction. While some see the incident as symptomatic of deeper systemic problems, others use it as a platform for ideological debates, reflecting the polarized and contentious nature of modern political discourse. The public’s focus on terms like “revolution” and “protest” underscores the widespread desire for significant change.

In addition, Topic 8 focuses on debates about “free speech,” “twitter,” and “youtube,” reflecting discussions on platform policies, content moderation, and censorship. This topic reflects skepticism toward media platforms and disagreement with social media censorship. Topic 9 clusters informal and humorous commentary, including terms like ‘lmao,’ ‘joke,’ and ‘omg.’ These reactions provide limited substantive insight into the broader discussion of the incident, reflecting casual engagement rather than meaningful analysis.

3.3 Sentiment analysis on reddit comments

In this section, we use a fine-tuned DistilBERT model to analyze the sentiment of comments on the Reddit platform, aiming to assess the overall public sentiment toward the related topics. DistilBERT is a compact and efficient version of BERT, which is fine-tuned to classify English text into two categories: “Positive” and “Negative” with high accuracy [73]. This approach allows us to uncover the nuanced sentiments expressed in the posts while benefiting from the model’s speed and computational efficiency.

Fig 1 shows the proportions of negative and positive sentiment in the Reddit comments sample. Of the 59,644 comments, 75.41% expressed negative sentiment and reflected significant public anger and dissatisfaction triggered by the event. Notably, 41.97% of the positive sentiment comments appeared in Topic 9 (Section 4.1.2), which captures casual internet interactions such as “LOL” or “Thank you.” Negative sentiment dominated all other topics and highlighted widespread public frustration with the insurance companies. The prevalence of negative sentiment throughout all topics suggests significant investor unease and potential adverse effects on the market perception of the insurance industry. To further investigate this, we test our hypotheses in the following sections using the cumulative abnormal returns (CARs) of insurance companies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Sentiment distribution: A pie chart analysis.

https://doi.org/10.1371/journal.pone.0334399.g001

4.1. Stock sample, variable and methodology

4.1.1.. Sample collection.

This study focuses on U.S. listed insurance companies’ stock performance surrounding the shooting incident that occurred on December 4th. We collect firms’ stock price data from the Compustat-CRSP Merged Database provided by Wharton Research Data Services (WRDS). We then identify insurance companies based on their Standard Industrial Classification (SIC) codes, as reported in their disseminated EDGAR filings which specify the type of their business. Following widely used Fama-French 48 category industry classification, we consider firms with SIC codes ranging from 6300 to 6411 as part of the insurance industry, yielding an initial sample of 150 firms. To calculate abnormal returns, we include only firms with stock price data available around December 4th, resulting in a final sample of 134 observations for our abnormal return analysis. We acknowledge that many studies focusing on insurance firms exclude the firms with the SIC 6411 code, which pertains to “Insurance Agents, Brokers & Service” [74,75]. However, our initial results indicate that firms in this category were also significantly affected by public sentiment (with a highly significant and negative abnormal return around this event; results are not reported here). Therefore, in the context of this study we have included this sub-category in our sample.

We collect executive compensation data from the Execucomp database available on WRDS. Since the database primarily includes S&P 1500 firms, we supplement missing values with executive compensation data extracted from DEF 14A filings using the SEC Filings API (https://sec-api.io/). For the market stock return, we obtain the indexes of S&P 500 from Yahoo finance. Besides the stock data, we obtain the firms’ accounting variables as our control variables from Compustat database. We then merge stock returns with firms that have available compensation and firm characteristic data, resulting in a final regression sample of 116 observations. We winsorize continuous variables at 1% level to address for outliers. Table 2 presents the summary statistics for the variables used in our analysis. Appendix 1 (supplementary information file: S2 Appendix) provides detailed definitions of these variables.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary statistics. This table shows the sample descriptive statistics for the key variables used in the paper. There are 116 observations in the dataset. We report the sample size (N), mean, standard deviation (Std), 25th, 50th (Median) and 75th percentile statistics for the dataset. Detailed variable descriptions can be found in Appendix 1 (supplementary information file: S2 Appendix).

https://doi.org/10.1371/journal.pone.0334399.t002

4.1.2. Main variable: cumulative abnormal returns.

We use the standard event-study methodology of [76] to analyze the impact of the shooting incident on the stock prices for the insurance firms. To analyze the stock market’s response, we construct Cumulative Abnormal Returns (CARs) as our main variable, which is a widely used measure of stock performance in event studies within the finance literature [77,78]. We compute the abnormal return around the incident day using a three-day event window [77]. Since the shooting incident is an accidental event and should not have any impact on the stock prices before the shooting date, we adopt an event window starting from exact the shooting date to two days after the shooting date. The abnormal return for firm i on day t is:

where represents the raw return of the common stock of firm i on day t, and is the expected return on stock i on day t. We use the market model to calculate the expected return on stock i:

where is the daily market return, and and are calculated for each firm by regressing the firm’s returns on the market returns during the time period of 200 trading days to 30 trading days before the shooting date. Though the literature usually uses daily value weighted return from CRSP as market returns, but the data for 2024 is not available on CRSP when we conduct our analysis. Thus, we adopt the S&P 500 download from Yahoo Finance as market return proxy. The S&P 500 is a value-weighted portfolio of 500 of the largest U.S. stocks and is widely recognized as a benchmark proxy for the overall market [79]. Over the interval of three days around the shooting day, beginning with day 0 and ending with +2, the cumulative abnormal return, (simply referred to as the CAR) is:

Our first objective is to test if the CARs for insurance companies are significantly different from zero. We perform the following hypothesis test in Section 4.2:

where is the cumulative abnormal return. We use standard t-test statistics to determine the statistical significance of the results.

4.1.3. Regression models.

To examine our hypothesis proposed in Section 2.6, we then use multiple regression models to examine the impact of firm and CEO characteristics on insurance companies’ CAR. For Hypothesis 1a and 1b, our regression models are defined as:

where is the cumulative abnormal return for firm i within the three-day trading window around December 4th. is the explanatory variable for corporate profit and growth. Specifically, we employ two measures to capture this: and . measures the firm’s profitability, calculated as the ratio of Income Before Extraordinary Items (ib) to Total Assets (at). is a binary indicator that equals 1 if the firm’s average revenue growth rate over the past three years exceeds the sample median, and 0 otherwise. is measured by the total number of news articles mentioning the firm in the past one or three years in the Factiva database. We also control other firm characteristics in the model, including firm size, book-to-market ratio, and firm age. Appendix 1 (supplementary information file: S2 Appendix) provides detailed definitions of all variables used in the analysis.

We are primarily interested in the sign of , which represents the impact of the explanatory variables on firms’ cumulative abnormal returns (CAR). Similarly, when testing alternative Hypothesis 2, we substitute the explanatory variables with executives’ compensation and their personal traits:

where represents the executive pay for firm i, we use three different proxies to measure the compensation level: the CEO’s individual compensation, the total compensation for all executives reported in the firm’s filings, and the average compensation level of executives. measures CEO narcissism and is calculated as the natural logarithm of one plus the ratio of first-person singular pronouns to all first-person pronouns used by the CEO during the Q&A section of earnings calls [80,81]. Finally, the regression model for testing Hypothesis 3 is:

where is a dummy variable that equals 1 if the firm belongs to the same industry segment as UnitedHealthcare, identified by SIC code 6324. Since our sample comprises cross-sectional data with unique firms observed in the same year, we do not include firm or year-fixed effects. However, to enhance the robustness of our results, we incorporate additional control variables into the regression, despite the relatively small sample size. We also report robust t-statistics to ensure the reliability of our empirical findings.

One of the distinct advantages of our empirical set-up is that it is based on a truly exogenous event (CEO shooting) – this is a rare incidence in the North American corporate world, and no one was anticipating such an unfortunate event. This virtually eliminates the possibility of a ‘reverse causality’ in our empirical setup, which is a major concern in corporate finance empirical studies.

4.2. Economic effects of the shooting on insurance companies

Given the public anger and negative sentiment identified in Section 3, we examine whether the incident has a significant economic effect on the stock prices of insurance companies. To assess this impact, we analyze the cumulative abnormal returns (CARs) across various event windows, ranging from a three-day window (0, + 2) to an eleven-day window (0, + 10) following the shooting date [78]. Fig 2 illustrates the cumulative abnormal returns (CARs) for all insurance companies from ‘day 0’ (the shooting incident) to ‘day 10’. The CARs show a sharp decline in the first two days, reflecting an immediate negative market reaction. On day 3, a temporary rebound occurs, likely due to opportunistic investors attempting to buy the dip, given the substantial profit levels of these companies. However, this recovery is short-lived, as the negative sentiment and broader impacts of the shooting incident persist, causing CARs to continue declining over the subsequent days. The overall trend highlights the sustained adverse effects of the event on the insurance sector’s stock performance.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Cumulative abnormal return of insurance companies following the december 4th shooting incident.

https://doi.org/10.1371/journal.pone.0334399.g002

Table 3 presents the cumulative abnormal returns (CARs) for three sample sets. Panel A reports the CARs across various event windows for all insurance companies in the sample. The CARs for the six evaluated event windows are significantly negative at the 1% level, indicating a strong negative stock market reaction to the shooting incident among insurance firms. The CARs range from −1.658% (over the (0, + 2) window) to −7.299% (over the (0, + 10) window). Since news and reports about the incident began emerging on the second day, which could lead to public awareness later, we also examine windows starting from day + 1 (one day after the shooting), and the CARs remain significantly negative.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Impact of December 4th shooting incident – univariate CAR analysis. This table presents the cumulative abnormal returns (CARs) for a number of windows surrounding the day of the shooting of the CEO of UnitedHealthcare on December 4th. Panel A presents CARs for all types of insurance firms, Panel B reports CARs for “hospital & medical service plans” insurance companies, while Panel C presents CARs for all other types of insurance companies. The classification of insurance companies are those with standard industry classification (SIC) code between 6300 and 6411. “Hospital & medical service plans” insurance companies, classified under SIC code 6324, provide prepaid or contract-based health services, including hospital, medical, and other healthcare services, to members or subscribers in exchange for specified subscription charges, with coverage often facilitated through prearranged agreements with healthcare providers. UnitedHealthcare and its competitors, including Elevance Health, Cigna Group, Humana Inc, and Centene Corp, all fall under this category. ***, **, * stand for statistical significance at the 1%, 5%, and 10% level, respectively.

https://doi.org/10.1371/journal.pone.0334399.t003

We further analyze insurance firms that have the closest relationship with UnitedHealthcare by focusing on those within the same industry, identified by sharing the same SIC code. Panel B of Table 3 reports the CARs for insurance companies with a SIC code of 6324, which includes UnitedHealth and many of its competitors. We find all CARs are significantly negative in this industry, the CARs range from −4.216% (over the (0, + 2) window) to −11.158% (over the (0, + 10) window). Panel C of Table 3 reports the CARs for all other insurance firms not included in Panel B. While the CARs are quantitatively smaller, they remain significantly negative.

Overall, the shooting incident represents a significant economic event, leading to substantial stock price declines for all insurance firms, with particularly severe impacts on the category that includes UnitedHealth and its competitors.

4.3. Corporate profit and cumulative abnormal returns following the shooting

From this section, we apply the multiple regression model introduced in Section 4.1.3 to examine which firm characteristics contribute to more negative CARs among insurance firms. Our analysis begins with testing Hypothesis 1a. As we found in Section 4.1, the public expressed significant anger and negativity on online platforms toward insurance companies, specifically criticizing their overemphasis on profit and revenue growth, which has generated widespread dissatisfaction. For example, we observed comments such as, “You don’t have to be insured to know how greedy the insurance companies have been,” and “No one should be profiting millions from healthcare, healthcare services, and medications… all this is critical to all Americans.” If customers and investors share these concerns, we would expect firms with higher profits and revenue growth compared to their peers to experience more negative CARs, leading to a negative coefficient for the explanatory variable.

We begin by regressing CARs on the firms’ profit and high revenue growth to test the hypothesis that higher profits and revenue growth lead to more negative abnormal returns. Table 4 presents the regression results across four specifications. The coefficient on Profit is negative and statistically significant at the 5% level in columns [1] and [3], with values of −0.045 and −0.047, respectively. This indicates that higher profitability is associated with more negative CARs, supporting the hypothesis that public and investor dissatisfaction with profit-driven motives impacts stock performance negatively. The coefficient on High_Rev_Growth is consistently negative and statistically significant at the 1% level in column [2] (−0.018) and at the 5% level in column [4] (−0.017). This finding suggests that firms with higher revenue growth experience greater stock price declines, potentially reflecting public frustration with aggressive growth strategies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Relationship between corporate profit, glamour, and cumulative abnormal returns (CARs) following the fatal shooting of the CEO. This table presents the relationship between the firms’ profit and revenue growth with the cumulative abnormal returns (CARs) after the day of the shooting of the CEO of UnitedHealthcare. The dependent variable is the cumulative abnormal return (CAR) from day 0 to day 2. In column (1) and column (3), CAR is regressed on firm profit, while in column (2) and column (4), CAR is regressed on a dummy variable indicating high revenue growth. The variable High_Rev_Growth is a binary indicator that equals 1 if the firm’s average revenue growth rate over the past three years exceeds the sample median, and 0 otherwise. Several firm accounting and stock characteristics are controlled for in the regressions, including firm size, book-to-market ratio, and firm age. Definitions of variables are provided in Appendix 1 (supplementary information file: S2 Appendix). Robust t-statistics are in parentheses. Asterisks *, ** and *** indicate the significance levels of 10%, 5% and 1%, respectively.

https://doi.org/10.1371/journal.pone.0334399.t004

The economic significance of the regression results is particularly large. A one standard deviation increases in Profit or High_Rev_Growth, the CARs of insurance companies decrease by 37.9% or decreases by 53.2% relative to the sample average CAR, respectively (The standard deviation of Profit and High_Rev_Growth is 0.15 and 0.502. For regressions in columns (1) and (2), coefficients of -0.045 and -0.018 reflect changes in CARs of -39.7% ( ) and -53.2% ( ), respectively. As reported in Table 1, CAR has a mean value of −0.017).

We further divide the sample based on whether firms have positive or negative profitability, motivated by the observation in Table 2 that the median profit is positive and the possibility that the market may interpret profit levels differently depending on firms’ financial condition. The results show that the coefficient on profit is significantly negative only among firms with positive profitability, while it is statistically insignificant among loss-making firms. This suggests that public anger and investor backlash was concentrated on firms that were profitable before the incident, reinforcing the interpretation that visible profit-taking prior to the event became a focal point of social criticism and market reaction.

The results provide strong evidence that, following the shooting, higher profitability and revenue growth in insurance firms significantly amplify negative stock price reactions. This finding aligns with the hypothesis that public and investor dissatisfaction with profit-driven behaviors is reflected in market outcomes.

4.4. Media attention and cumulative abnormal returns following the shooting

In this section, we examine Hypothesis 1b, which posits that insurance firms with greater media attention experience more negative impacts on their CARs. To test this, we replace the explanatory variable in the multivariable regression model with the measure based on the number of media reports over the past one and three years, respectively. This approach follows prior studies that use media coverage frequency as a proxy for firms’ media exposure [82,83]. If the negative sentiment observed in Section 4 is reflected in the CARs, we would expect a negative coefficient for media coverage.

Table 5 reports the results. In columns [1] and [3], the independent variable is media attention over the past one year, measured as the number of articles related to the firm in the Factiva database. The coefficient for media attention is negative and statistically significant at the 1% level in column [1] (−0.020) and at the 1% level in column [3] (−0.019), suggesting that firms with higher media attention in the past year experience more negative CARs. In columns [2] and [4], the independent variable is media attention over the past three years, measured in the same way. The coefficient is also negative, with statistical significance at the 10% level in both column [2] (−0.010) and column [4] (−0.009). This indicates that extended media attention over a longer period is similarly associated with more negative stock price reactions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Relationship between media attention and cumulative abnormal returns (CARs) following the fatal shooting of the CEO. This table presents the relationship between the median attention with the cumulative abnormal returns (CARs) after the day of the shooting of the CEO of UnitedHealthcare. The dependent variable is the cumulative abnormal return (CAR) from day 0 to day 2. In column (1) and column (3), CAR is regressed on the median attention in the past one year, while in column (2) and column (4), CAR is regressed on the median attention in the past three years. The median attention is measured as the number of articles (in thousands) related to the firm in the Factiva database in the past one or three years, respectively. Several firm accounting and stock characteristics are controlled for in the regressions, including firm size, book-to-market ratio, and firm age. Definitions of variables are provided in Appendix 1 (supplementary information file: S2 Appendix). Robust t-statistics are in parentheses. Asterisks *, ** and *** indicate the significance levels of 10%, 5% and 1%, respectively.

https://doi.org/10.1371/journal.pone.0334399.t005

Overall, the results support Hypothesis 1b, which posits that glamour firms with greater media attention experience a more negative impact on cumulative abnormal returns (CARs). The significant negative coefficients for media attention in both the one-year and three-year periods indicate that firms with higher public visibility and scrutiny face stronger adverse stock price reactions following the incident.

4.5. Executive compensation and cumulative abnormal returns following the shooting

In this section, we test Hypothesis 2a, which posits that firms with higher CEO compensation will experience a more negative impact on cumulative abnormal returns (CARs). To evaluate this hypothesis, we include three executive compensation measures, all transformed using their natural logarithms, as explanatory variables in the multiple regression model to analyze their relationship with CARs following the incident. These measures are [1] the logarithm of CEO compensation, [2] the logarithm of the total compensation of all executives reported by the firm, and [3] the logarithm of the average compensation of executives within the firm. Together, these measures capture both the compensation of the top executive (CEO) and the overall level of executive compensation observed by investors, providing a comprehensive and scaled view of how executive pay influences market reactions.

Table 6 reports the results. Columns [1] and [4] show that CEO compensation is significantly negatively associated with CARs, with coefficients of −0.015 and −0.013, respectively, both significant at the 5% level. This suggests that firms with higher CEO pay experienced more negative stock price reactions. Columns [2] and [5] examine the average compensation of executives. The results indicate a negative relationship with CARs, with coefficients of −0.022 (significant at 1%) and −0.021 (significant at 5%), highlighting that firms with higher average executive compensation faced stronger negative abnormal returns. Columns [3] and [6] analyze the total executive compensation. The coefficients, −0.019 and −0.018, are both significant at the 5% level, providing further evidence that higher total executive compensation exacerbates negative stock price reactions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Relationship between Executive Compensation and Cumulative Abnormal Returns (CARs) Following the Fatal Shooting of the CEO. This table presents the relationship between the executive compensation with the cumulative abnormal returns (CARs) after the day of the shooting of the CEO of UnitedHealthcare. The dependent variable is the cumulative abnormal return (CAR) from day 0 to day 2. In column (1) and column (4), the dependent variable CAR is regressed on the natural logarithm of the CEO’s total compensation. In column (2) and column (5), CAR is regressed on the natural logarithm of the average compensation of all executives reported by the firm. In column (3) and column (6), the independent variable is the natural logarithm of the total compensation value of all executives reported by the firm. Several firm accounting and stock characteristics are controlled for in the regressions, including firm size, book-to-market ratio, and firm age. Definitions of variables are provided in Appendix 1 (supplementary information file: S2 Appendix). Robust t-statistics are in parentheses. Asterisks *, ** and *** indicate the significance levels of 10%, 5% and 1%, respectively.

https://doi.org/10.1371/journal.pone.0334399.t006

Overall, the results support Hypothesis 2a, indicating that firms with higher CEO and executive compensation are more negatively impacted in terms of CARs following the incident. These findings suggest that public and investor scrutiny of executive pay plays a critical role in shaping market reactions to such events.

4.6. CEO narcissism and Cumulative Abnormal Returns following the shooting

In this section, we test Hypothesis 2b, which posits that firms with higher CEO narcissism will experience more negative effects on cumulative abnormal returns (CARs). To test Hypothesis 2b, we construct a measure of CEO narcissism measured as the natural logarithm of one plus the ratio of first-person singular pronouns to all first-person pronouns used by the CEO during the Q&A section of earnings calls [80,81]. This measure is incorporated as an explanatory variable in our multivariable regression model, alongside controls for firm characteristics, including size, age, and book-to-market ratio.

Table 7 presents the results of regressions analyzing the relationship between CEO narcissism and cumulative abnormal returns (CARs) for insurance firms following the shooting of UnitedHealthcare’s CEO. In column [1], CEO narcissism is measured as the natural logarithm of one plus the ratio of first-person singular pronouns to all first-person pronouns used by the CEO during the Q&A section of earnings calls. The coefficient for this variable is positive (0.026) but not statistically significant, indicating no strong evidence that this measure of CEO narcissism significantly influences CARs. Column [2] uses the original ratio of first-person singular pronouns as the independent variable. The coefficient (0.004) is again positive but lacks statistical significance, suggesting that this alternative measure of narcissism also does not have a meaningful impact on CARs. Similarly, we do not find significant results on CEO narcissism or first-person pronouns in columns [3] and [4]. As a robustness test, we use an alternative measure of CEO narcissism – calculated as the ratio between the total compensation of the CEO and that of the next-highest-paid executive in the firm [73]. This measure also leads to insignificant results.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Relationship between CEO Narcissism and Cumulative Abnormal Returns (CARs) Following the Fatal Shooting of the CEO. This table presents the relationship between the CEO Narcissism and the cumulative abnormal returns (CARs) after the day of the shooting of the CEO of UnitedHealthcare. In columns (1) and (3), the independent variable is CEO narcissism, measured as the natural logarithm of one plus the ratio of first-person singular pronouns to all first-person pronouns used by the CEO during the Q&A section of earnings calls. In column (2) and (4), the independent variable is the original ratio of the first-person singular pronouns used by the CEO during the earnings calls. Several firm accounting and stock characteristics are controlled for in the regressions, including firm size, book-to-market ratio, and firm age. Definitions of variables are provided in Appendix 1 (supplementary information file: S2 Appendix). Robust t-statistics are in parentheses. Asterisks *, ** and *** indicate the significance levels of 10%, 5% and 1%, respectively.

https://doi.org/10.1371/journal.pone.0334399.t007

Overall, the results do not provide strong support for Hypothesis 2b, as neither measure of CEO narcissism demonstrates a significant relationship with CARs. This suggests that investor anger is not directed at the CEO’s subjective attitudes or personality traits but is instead more focused on executive compensation levels and firm profits. The findings highlight the tension between public welfare and the pursuit of shareholder value maximization, which appears to be a central driver of negative market reactions.

4.7. Impact on the targeted industry

In this section, we test Hypothesis 3, which states that the stock market reaction to the assassination of UnitedHealth Group’s CEO was more severe for insurance firms in the same industry segment. These firms are classified under SIC code 6324, which corresponds to “Hospital & Medical Service Plans.” If investors were particularly concerned about firms with similar business models and practices to UnitedHealth, such as Elevance Health, Cigna Group, Humana Inc., and Centene Corp, then companies in the same sub-industry should show more negative cumulative abnormal returns (CARs). To test this hypothesis, first we compare the CARs between SIC 6324 (treated group) with insurance firms outside this sub-industry. We further conduct a matching analysis comparing firms in SIC 6324 (treated group) with insurance firms outside this sub-industry but similar in size and book-to-market ratio. Following the approach of Fu et al. [84], we select control firms that fall within 50% to 150% of each treated firm’s size and book-to-market ratio. From these, we identify the two closest matches using minimum Euclidean distance. For one treated firm, we could not find a control firm as per the matching criteria stated above. Thus, the process yields a matched sample of 9 treated firms and 18 control firms.

Table 8 reports the results. Panel A shows unadjusted comparisons between the treated and control groups. Firms in SIC 6324 experience significantly more negative cumulative abnormal returns (CARs) in the short-term event windows. The difference in CAR (0, + 2) is −2.764% and in CAR (1, +  2) is −2.977%, both statistically significant. Panel B presents results after matching. The short-term effects remain significant. The matched treated firms have a CAR (0, + 2) of −5.529%, compared to −2.222% for the control firms, a difference of −3.308% (significant at the 5% level). For CAR (1, +  2), the difference is −3.147%, significant at the 1% level. In contrast, the differences for longer event windows such as CAR (0, + 5), CAR (0, + 10), CAR (1, +  5), and CAR (1, +  10) are not statistically significant. It appears that the differentiating stock market reactions between SIC 6324 sub-group firms and other insurance firms are primarily observed with a very short window.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Relationship between Insurance Segment and Cumulative Abnormal Returns (CARs) Following the Fatal Shooting of the CEO. This table presents the relationship between Insurance Segment and the cumulative abnormal returns (CARs) after the day of the shooting of the CEO of UnitedHealthcare. Panel A reports univariate differences in CARs between treated firms, defined as those in the ‘Hospital & Medical Service Plans’ insurance segment (SIC 6324), and control firms, which belong to other insurance industries. Panel B presents univariate comparisons where control firms are matched to treated firms based on firm size and book-to-market ratio. Panel C shows multivariate regression results where CARs are regressed on the dummy variable Hospital&Medical_Insurance, which equals 1 for firms classified under SIC 6324 and 0 otherwise. UnitedHealthcare and its competitors, including Elevance Health, Cigna Group, Humana Inc., and Centene Corp., are included in this category. The regression controls for firm characteristics such as size, book-to-market ratio, and firm age. Definitions of variables are provided in Appendix 1 (supplementary information file: S2 Appendix). Robust t-statistics are in parentheses. Asterisks *, ** and *** indicate the significance levels of 10%, 5% and 1%, respectively.

https://doi.org/10.1371/journal.pone.0334399.t008

Panel C of Table 8 shows multivariate regressions using the full sample of insurance firms. We regress short-term CARs on an indicator for SIC 6324 membership and control for firm size, book-to-market ratio, age, R&D intensity, leverage, and past stock returns. The SIC 6324 indicator remains negative and statistically significant for CAR (0, + 2) and CAR (1, +  2) windows. However, for other CAR windows, we do not observe any significant difference. These results indicate that firms in this segment experienced larger stock price declines during very short-term windows, even after adjusting for observable firm characteristics.

Overall, the evidence supports Hypothesis 3. The assassination triggered a sharper market response for firms in the Hospital & Medical Service Plans segment. This suggests that investors viewed these firms as sharing key similarities with UnitedHealth and expected them to face similar risks or public scrutiny.

4.8. Extending Reddit sentiment analysis to explain cross-firm variation in CARs

To further investigate whether firms associated with more negative social media sentiment experienced greater stock price declines, we construct firm-level measures of negativity exposure based on the Reddit discussions in Section 3. Specifically, we identify negative comments posted during the event window (0, + 2) that reference firms in our sample and count how many times each firm appears in comments. This procedure would yield a firm-level variable that captures the extent of public negativity exposure on social media, which allows us to test whether firms more frequently criticized in online discussions experienced larger stock price declines.

While this integration is conceptually intuitive, Reddit comments do not contain structured tags indicating which firm they refer to, and a direct name-based matching between comment text and company names can lead to mismatching. For instance, the word “citizen” in the comments may refer to U.S. citizens rather than the firm Citizen Inc. To address this challenge, we apply Named Entity Recognition (NER), a Natural Language Processing (NLP) technique that automatically detects and classifies entities mentioned in the text [85]. We then match the recognized organization entities to firm names in our sample based on exact string matches. This approach allows us to retain only those instances where a firm is explicitly identified as an organization, avoiding mismatches with common nouns that are spelled the same as company names [86]. While this approach limits the number of comments that can be linked to firms, it enhances the accuracy of our sentiment exposure variable.

Based on the mapping, we construct two measures of firm-level exposure to negative sentiment: [1] Negative_comments, which counts the number of negative Reddit comments in which a firm is mentioned, and [2] Negative_comments_dummy, an indicator equal to one if the firm is mentioned in at least one negative comment. We then regress the CARs on these measures to examine whether firms that are more frequently criticized on Reddit experience larger stock price declines. Table 9 reports the results. The coefficient on Negative_comments is negative and statistically significant at the 1% level in column [1] (–0.004), while Negative_comments_dummy yields a significant negative coefficient at the 1% level in column [2] (–0.037). The results remain robust after including additional controls, with significance levels at 10% and 5% in columns [3] and [4], respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 9. Relationship between Negative Reddit Sentiment and Cumulative Abnormal Returns (CARs) Following the Assassination of the UnitedHealth CEO. This table presents the relationship between firm-level exposure to negative Reddit sentiment and cumulative abnormal returns (CARs) following the assassination of the CEO of UnitedHealthcare. The dependent variable is the cumulative abnormal return from day 0 to day 2. In columns (1) and (3), CAR is regressed on Negative_comments, which represents the number of negative Reddit comments mentioning the firm. In columns (2) and (4), CAR is regressed on Negative_comments_dummy, an indicator equal to one if the firm is mentioned in at least one negative Reddit comment, and zero otherwise. Several firm accounting and stock characteristics are controlled for in the regressions, including firm size, book-to-market ratio, and firm age. Definitions of all variables are provided in Appendix 1 (supplementary information file: S2 Appendix). Robust t-statistics are reported in parentheses. Asterisks *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

https://doi.org/10.1371/journal.pone.0334399.t009

These findings suggest that investors reacted more severely to firms that were specifically named and criticized in online Reddit discussions. In line with our interpretation, social media narratives appear to channel public outrage in a targeted manner, amplifying stock price declines for firms perceived to be more blameworthy.

4.9. Placebo test using sudden CEO deaths in other industries

To further validate our interpretation of the assassination’s industry-wide impact through the lens of social bandit theory, we examine whether the sudden death of CEOs in other industries triggered similar market responses among peer firms. Specifically, we test the market reaction within the same industry following two high-profile CEO death events. These events also triggered media coverage and public attention but did not generate widespread societal backlash or industry-level moral criticism. This contrast helps highlight the distinctive nature of the health insurance case and underscores the role of public outrage and investor sentiment in shaping abnormal stock returns.

Panel A of Table 10 analyzes the market reasons in the transportation industry following the fatal shooting of Philip Trenary, former CEO of Pinnacle Airlines and a prominent civic figure in Memphis. Although the incident attracted media coverage due to Trenary’s community profile, it did not spark strong public discussion about the industry’s conduct or moral concerns. Event study results show a small and marginally significant positive CAR of 0.76% over the (0, + 2) window (significant at 10% level), with no significant effects in other windows.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 10. Sudden CEO Deaths in Other Industries – Univariate CAR Analysis. This table reports the cumulative abnormal returns (CARs) for industry peer firms surrounding the sudden deaths of CEOs in two different industries. Panel A presents CARs for transportation firms around the fatal shooting of Philip Trenary, the former CEO of Pinnacle Airlines, on September 27, 2018. Panel B presents CARs for electronic equipment firms around the death of Micron Technology’s active CEO, Steve Appleton, who died in a plane crash on February 3, 2012. Peer firms are defined by the same Fama-French 48 industry classification as the firm experiencing the CEO death. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

https://doi.org/10.1371/journal.pone.0334399.t010

Panel B investigates the electronic equipment industry in response to the accidental death of Steve Appleton, then-CEO of Micron Technology, in a 2012 plane crash. As an active CEO, Appleton’s death was widely reported. However, public reaction remained focused on the personal tragedy rather than corporate behavior. CARs for peer firms were positive and statistically significant in a variety of windows: 1.40% over the (0, + 5) window (significant at 1% level), and 1.80% over (0, + 10) (significant at 1% level).

These findings are not intended to suggest that the CEO deaths in other industries necessarily produce large industry-wide stock price effects. Rather, the absence of significant negative CARs in these cases reinforces our central argument. Both events lacked the strong public outrage, moral condemnation, and narrative spillover that are central to the social banditry mechanism. Without a wave of collective indignation directed at the industry, investor sentiment remained relatively unaffected. Unlike other high-profile CEO deaths, the assassination of the UnitedHealth CEO sparked rare collective outrage toward the industry, fueling investor pessimism and driving the significant negative spillovers we document.

4.10. Robustness tests using extended event windows

We examine the robustness of our findings by re-estimating the main regressions using extended event windows of (0, + 3) and (0, + 5). Across specifications in the untabulated results, the coefficients on high revenue growth, media attention, and compensation-related variables remain statistically significant and directionally consistent with those in the baseline (0, + 2) window. While the coefficients on CEO narcissism remain insignificant. Notably, for these significant variables, the magnitudes of the coefficients decrease in the (0, + 3) window but increase again in the (0, + 5) window, ultimately exceeding their baseline levels. This non-monotonic pattern aligns with the cumulative abnormal return trajectory in Fig 2, which shows a temporary rebound in the market around day 3 before continuing its downward trend.

In contrast, the coefficient on profitability remains negative but becomes statistically insignificant in the extended windows. A likely explanation is that high-profitability firms are perceived as fundamentally undervalued after initial price declines and are more likely to attract speculative contrarian traders, thereby offsetting the early sell-off [87,88]. As a result, the early negative effect of profitability becomes muted over the longer window due to countervailing upward price pressure. Moreover, expanding the event window increases the risk that unrelated news adds noise to abnormal return estimates [11]. Still, the coefficient on profitability continues to decline and becomes more negative in the (0, + 5) window compared with (0, + 3), suggesting that the market was torn between selling on negative sentiment pressure and buying on high profit fundamentals during the rebound, rather than ignoring profitability altogether in the longer windows.