“Shock events: The impact of news media and communication strategies on listed companies’ share price”

Academics studied the theory of a company’s communication when it is involved into a crisis but they were less concerned about the impact of the communication on a listed company’s share price, especially when it resulted from a shock event. There is a lack of information about the role played by news media. The aim of this paper is to investigate if in cases of shock events (i) a company’s response strategy has a different effect on shareholders, observing the effect on share prices, and (ii) how the news media can affect the value change. Using the event study methodology, the Cumulative Abnormal Return of companies’ share prices involved in shock events was calculated. Statistics show a best effect of an accommodative response than a defensive strategy in cases of scandals and product recalls. There is no valuable impact of company communication in cases of incidents. With news media variable, the results show a worsening effect with bad news and a mitigating effect with good news. It was proved that the impact of a response strategy is surpassed by news media. When there is absolute certainty of guilt for a given situation, it is more convenient for management to apologize, and when there is no certainty, there was no substantial difference, because in the mind of an investor the focus shifts to the event itself. The news media has been shown to have a huge impact on investor perception, even more so than a company’s best response strategy. shareholders. the of had any effect on the corresponding the best type of to the shock the more appropriate communication strategy to each a approach to the individual felt by a


INTRODUCTION
The analyses to discover whether there is a link between the market share price of a company involved in an event crisis and the communication used to make public that event for listed companies, is one of the topics in the finance crisis communication theory.
Several academic studies have attempted to estimate the trend of a company's share price during a crisis and the consequent reactions from shareholders. Marcus and Goodman (1991) and Ferretti et al. (2015) tried to establish whether the type of declaration had any effect on the corresponding share price. Coombs (2006) regards the best type of response to propose when a specific crisis occurs, where the first step consists in clustering the shock event types and then to link the more appropriate communication strategy to each cluster. Jim et al. (2010) apply a psychological approach to understand the individual emotion felt by a generic shareholder.
Another important topic has emerged concurrent with the study carried out about Crisis Communication: the role of news media. Relevant studies (Kim & Cameron, 2011;An et al., 2011) have shown how in-The event study methodology has been detected as being the most suitable instrument for this type of study, as it allows considering multiple variables and identifying a time window for carrying out the analysis.
Indeed, in such cases most of studies aiming to find variations in share prices carry out a short-term analysis as in this way it is easier to manage to exclude those external factors that may invalidate results. For such reasons, the test shall be carried out over a total ten-day time frame: two days before and seven days after the date of the event.

LITERATURE REVIEW, THEORETICAL FRAMEWORK AND HYPOTHESES
In relation to the purpose, firstly it`s necessary to define the shock event, based on the concept of crisis. Various researchers (Weiner & Kahn, 1972) began to specifically address the concept of crisis, identifying three key elements that may lead a company to the crisis: high threat, limited time for decision making, and surprise element.
The analysis by Pearson and Claire (1998) is related to these three points, since the crisis is defined as "a low probability but high impact event which threatens the company's profitability and is characterized by ambiguity on the cause, the effects and the means with which it will be solved, as on the decisions which must be made rapidly." The main points of a crisis may be also used to consider the key characteristics of a shock event which, however, only turns out to be one part of the crisis process and it stands out for its speed of execution, which always allows one to identify the day on which it has occurred.
The literature has carried out important research on the subject, but the main study is Timothy Coombs' (1995Coombs' ( -2010 Situational Crisis Communication Theory (SCCT). It is a theory developed as a search guide concerning the best types of responses that a company intends to propose when a specific crisis occurs; for the purposes of this paper, it can be useful to understand whether it is connected with a company's financial recovery following the event.
To make any type of analysis, it is necessary to classify shock events; the basic principle offered here has been that investors attribute responsibilities to companies, whereby reputation becomes a crucial element for the purposes of the valuation of an investment.
Hence, shock events fall into the following different categories: a) Incidents: Unpredictable events for which investors are uncertain as regards a company's responsibility and in which the company itself may also turn out to be a victim of the crisis. b) Scandals: This type of a shock event is mainly attributable to a company. The behavior behind this type of shock event is often the result of shortcomings and slyness on the part of a particular entity. c) Product recall: This typology includes all cases in which a company has put on sale a product, which has turned out to be substandard or faulty and it has been obliged to recall it due to actual or potential damages.
During the last two decades there has been a rapid development in the research on Crisis Communication; after Coombs there were Covariation-Based Causal Attributions (Schwarz, 2008) and the ICM model (Jin, et al., 2007(Jin, et al., -2012. All the authors started from a psychological theory, which is used to study the different behavior that most people have in these types of situations, what is known as Attribution Theory (AT) by Bernard Weiner (1985Weiner ( , 2006Weiner ( , 2014. The Attribution Theory wants to identify factors that influence the opinion of any one person and how these factors will modify behavior and attitudes. Jolly and Mowen (1985) analyzed the effects on consumers of the recall of the product from the market, and they proved that a socially responsible action by a company led to public acceptance by bodies that had previously forced the company to recall the product, thus implicitly increasing the consumers' trust. Stockmyer (1996) examined the behavior resulting from product tampering in terms of variations in consumers' purchase intentions, Jorgensen (1996) studied the relationship between the response given by a company following an incident and consumer behavior, and finally, Bradford and Garnett (1995) tried to study the cases of unethical behavior verifying the relationship that was created between the crisis situation, the company's response and the effects caused by the correlation between the two elements. However, companies are more reluctant to take the blame given the possibility of thereby worsening their legal responsibility (Benoit, 1997).
Unlike previous studies, the SCCT tries to connect the potential damage to a company's image, which a crisis may provoke with the organizations' responsibility for what has happened.
Furthermore, Coombs deals with the main sources of information, that is, with the media; indeed, the news one can find on the daily papers identify the guidelines to be followed in order to direct the type of crisis in the right group. Hence, a crucial issue for the management may be to go and publicly report their version of the facts, at odds with the circumstances reported by the newspapers (Coombs, 2007).
In 2008, A. Schwarz tries to pursue what Coombs had begun, using as a hypothesis the covariation principle outlined by Kelley (1973) at the beginning of the 70s.
The covariation principle has highlighted that any isolated analysis of the effect of information on the attribution process may be misleading. Basically, relative to SCCT, a further strategy, prior to carrying out what is known as market communication, is that of influencing the knowledge acquisition of information. In other words, having relationships with influential newspapers or media, or special relationships with the most important stakeholders. All this would take place before the company's response concerning the crisis. Although this topic is strictly interesting, it was not possible for this study to collect this type of information.
At a primary level, it concerns the involvement in a situation of any one person, with reference to the relevance and to the consistency of the objectives. On the other hand, congruence ensures that an event is viewed positively; on the contrary, the objective's incongruence provokes a negative emotion. Finally, the involved parties shall be taken into consideration about their involvement in contribution to and responsibility in the event (Lazarus, 1991). At the secondary level, it concerns the different behavior that the involved parties may engage in when they must face the crisis.
From their numerous checks of ICM, authors found that the most dominant emotional state for any one stakeholder in the primary level in all crisis categories is that of anxiety.
The subsequent emotions felt by the publics in crises involving hostile takeovers, accidents and natural disasters were variations of sadness, anger, and fright, while the subsequent emotions felt by the publics involving CEO retirement, rumor and psychopathic acts were fright and anger (Pang et al., 2009).
Moreover, it is shown that an attitude of low involvement by a company may generate anger although it is often a subsequent reaction to the above-mentioned anxiety.
In highlighting the most accepted theories in the context of Crisis Communication, the role of media (daily papers, website, etc.) is clear, since they play a crucial part in the entire process. There have been many contributions in the literature on the subject like works by Kim and Cameron (2011) and An et al. (2011), based on what is known as the Framing Theory 1 .
Kim and Cameron's approach (2011) tries to mix the Framing and Appraisal theory. In addition, they consider the concept of depth-of-processing, which, briefly, is used to understand what kind of decision-making process follows the occurrence of an emotion. Indeed, Lerner and Tiedens (2006) have demonstrated that a specific emotion influences the depth of thought; since anger is a more "certain" emotional state, it leads individuals towards heuristic processing, as opposed to sadness, considered more "uncertain", leading to systematic processing. Kim and Cameron, in agreement with Nabi (2007), try and demonstrate Lerner and Tiedens' contribution, find out whether heuristic and systematic processing actually takes place, also when opinion and assessment are influenced by external agents, such as new media. The results of the analysis confirmed the hypothesis that the persons who experienced anger induced by a crisis reported in newspaper articles were more likely to make sudden decisions without even considering the other possible aspects of the matter. On the other hand, it has symmetrically emerged that when there was sadness, readers were more inclined to obtain more information on the event and to read the news more carefully, postponing the decisional process. An et al. (2011), in a different manner compared to Kim and Cameron, study framing theory following the Iyengar and Simon (1993) model that identifies two different typologies, the Episodic News Frame and the Thematic News Frame respectively. The first type is characterized by the description of an event in terms of personal experience, with a focus on social and emotional issues. On the oth- 1 Framing theory has been developed starting from 1974, when Erving Goffman published the "Frame analysis: An essay on the organization of experience", a study in which the sociologist stated the existence of interpretative schemes with which the human being schematizes all the information he communicates and which is communicated to him. However, Erving is only a pioneer of the theory that reached its peak in 1993, when Robert M. Entman (1993) published an article entitled Framing: Towards clarification of a fractured paradigm, in which he defines the concept of frame as an interaction between selection and salience, and then continues his analysis in 2006 with his article Framing Bias: Media in the Distribution of Power. 2 Anxiety is not present because since it is common to all three categories, it is possible to omit it. er hand, the second is characterized by a description of the event in abstract and impersonal terms. Therefore, the news media may trigger people to blame the organization and have a negative attitude, impression, and image toward the organization through their use of thematic frames . However, news media have a critical role in shaping public opinion about who is responsible for causing or solving key social problems (Iyengar & Kinder, 1987). Although An et al.'s (2011) contribution is related to the purpose of this study, it was not possible for this study to go through this type of analysis.
Summarizing, based on Situational Crisis Communication Theory, the following division of shock events has been agreed: 1. Incidents; 2. Scandals; 3. Product recall.
The division is in accordance with Ferretti, Profumo and Tutore (2015) although the three authors, as had previously been done by Marcus and Goodman (1991), besides considering the disasters/catastrophes, only limit themselves to observing the event objectively and to cataloguing it according to the division they have identified and do not consider other factors such as the emotional aspect experienced by shareholders up to the consideration of the news media's role.
According to ICM model, emotions can be associated with each of the event classes as indicated in the diagram below 2 : Two main categories of possible communication strategies adopted to companies in response to a shock event can be then identified: • Defensive strategy: A company shall distance itself from what has occurred, it will try to minimize the effects of what has happened.
• Accommodative strategy: A public apology, the willingness to remedy to what has occurred by compensation, referring to the company's historical qualities.
In agreement with previous research, these two types of answers can be associated to different cases of shock depending on the previous emotional states in the following way: About news media, according to Kim and Cameron's contribution, they can be categorized as follows: • Bad news: The news that harshly criticize a company blaming it openly for what happened that lead anger to a potential shareholder.
• Good news: The news that are not too critical of a company, or which are generally quite impartial, with the possible shift of attention on other parties that can lead sadness to a potential shareholder.
In this framework, the aim of this paper is to empirically test theoretical predictions previously discussed. In more detail, it can be considered as a two-level purpose: 1. At first, predictions from attribution theory, situational crisis communication theory and ICM model are tested looking at investors' behavior depending on the response adopted by companies in different types of shock events. The results will give more consistence to previously reported results in literature to which they can be compared.
In the case of incidents, investors shall manifest various emotional states, and this process will lead to assessing a defensive response and an accommodative response in the same manner.
In the case of scandals, investors will feel a state of anger and they will prefer an accommodative response to a defensive one.
H 3 : In the case of product recalls, investors will feel a state of anger and they will prefer an accommodative response to a defensive one.
H 4 : News media influence investors' choices, improving (good news) or worsening (bad news) the effects of company communication.

METHODOLOGY
The event study methodology has been deemed the most appropriate to test the hypotheses. Based on this theory, the identified communications will constitute what is known as the event. According to McWilliams and Siegel (1997), the event window will always start two days before the declaration, and will always end seven days after, to have a ten-day time window 4 . Subsequently, the estimation window will be selected over which the so-called "normal" returns necessary to estimate the abnormal return will be calculated. The estimation window will be 200 days before the event.
Returns on shares and of the market are computed as in (1) where R t = Return at time t; P t = Price at time t; P t-1 = Price at time t -1.
With linear regression in (2), the Market Modes is estimated to get "normal" returns.
where Ȓ it = "Normal" return; R mt = Return of the relevant market index.
Once normal returns are estimated, it will be subtracted from the actual one, thereby deriving the abnormal return as in (3).

ˆ. ˆˆ
The AR will then be grouped cumulatively, the CAR (Cumulated Abnormal Return), for three different intervals over the event window. They will be calculated as follows 5 : Depending on the type of examined case studies, statistical tests will be carried out in order to compare two sample data sets (CARn defensive response vs. CARn accommodative response) so as to verify if the populations from which they have been extracted may be considered significantly different. For this purpose, as a first step, descriptive statistics of each sample are calculated, including the Shapiro-Wilk test (and D'Agostino- Pearson) to verify the normality of distributions, as well 5 The CAR are the total abnormal returns up to the indicated days, the first, the fourth and the seventh after the declaration, respectively.
as the Levene test for the equality of variances. Should the normality test be positive (normal distribution for both), a T-Test comparing averages, with equal or different variances depending on the result of the Levene test, will be used to decide whether the working hypothesis must be accepted or not. Should the result of the normality test be negative, the comparison will be carried out using Mann-Whitney's non-parametric test.
A first analysis is carried out on the two samples (only CAR7) represented by all the defensive response cases vis-à-vis all the accommodative response cases with no distinction of shock type.
As regards H 1 formulated for the case of incidents, namely that a company's response should not influence the results, the two-tailed test will be used.
For hypotheses H 2 and H 3 , in which one expects the accommodative response to give better results than the defensive one, the one-tailed test will be used. The analysis of the impact of the news is only carried out on CAR7. The comparison between more than two samples has been carried out according to the same logic. Firstly, descriptive statistics of each sample are calculated, the Shapiro-Wilk/D'Agostino-Pearson tests are carried out to verify the normality and the Levene test to verify the equality of variances. In case of normality, an ANOVA test will allow determining whether the four groups are similar to one another or not, and should the result be that they are significantly different, the Tuckey HSD test will be used to compare the group pairs.
In case of non-normality, the first comparison will be carried out using the non-parametric Kruskal-Wallis test, and should the groups turn out to be significantly different from one another, the following non-parametric tests will be carried out on the pairs: Nemenyi, Mann-Whitney, Dunn, Schlaich-Hamerle, Conover.

RESULTS AND DISCUSSION
The sample has been selected by identifying communications of companies listed on a regulated market that are involved in sensational cases of incidents, product recalls and scandals. On this basis, 39 case studies have been selected, divided into three types of events (15 incidents, 11 scandals, and 13 product recalls) starting from the beginning of the 2000s.
The sources of the announcements have been the companies' official websites, Bloomberg's and Thomson Reuters Eikon's databases and official press releases, respectively; the dataset is coherent with the studies carried out in the literature. As regards the impact of news media, the articles present on the most popular newspapers in the world 6 have been selected, using the LexisNexis database, where it has been possible to indicate the exact period over which the viewed news should be dated, for the search. Share prices have been obtained from Bloomberg's and Thomson Reuters Eikon's databases.
As a first look to the reaction of investors to the two different possible managerial responses in case of shock events, defensive or accommodative, the data presented reports on the descriptive statistics of the two CAR7's samples of all cases, together with statistical tests. In the following, results for single cases are reported and discussed.
Concerning incidents, hypothesis H 1 assumes that the two populations from which samples are drawn are not significantly different.
As The results are reported in Table 3. As expected, for all CARs, the defensive sample mean is consistently lower than accommodative. Only for CAR1, the difference is not statistically significant, while for CAR4 and CAR7 it is with a very low p-value.
Following the results, hypothesis H 2 can be considered as verified. According to Marcus and Goodman (1991), in case of scandals, investors prefer an accommodative response with respect to a defensive one.
Results are consistent with Coombs (2010), especially due to his links with attribution theory that demonstrated that scandals could be associated with the internal locus and with the concept of controllability, key elements for the attribution of responsibility to a company.  For product recalls, hypothesis H 3 also assumes an investor's preference for an accommodative response. Results are reported in Table 4. Once again, for all three CARs, sample means are consistently different. The difference is statistically significant for CAR1, not for CAR4 and the t-test's p-value for CAR7, which is only slightly higher than 5%. At a 95% confidence level the null hypothesis must be accepted, but it can be rejected at 94%.
Given this small difference, the tendency is to consider the hypothesis verified and these results support the economic theories based on the attribution theory (Jolly & Mowen 1985). These theories state that in case of product recall, if a firm has a socially responsible behavior, consumers will regain, or even increase, their confidence. Similarly, Coombs (2010) showed that in such cases, a reconstructive strategy would have increased the probability for the firm to regain reputation.
Otherwise, the results presented here contrast with Marcus and Goodman (1991). They found no significant differences between managerial responses in case of dangerous products. It should be noted that Marcus and Goodman (1991) only considered five cases associated to important lawsuits or even resulting in victims among consumers.
For the purpose of this paper, the final concern is about the news media effect: how do bad news or good news affect investors' behavior? How do they interact with firms' communication? To answer these questions, all data (with a focus on CAR7s) have been categorized with respect to both communication response and news type. Four samples are shown: defensive response and bad news, defensive response and good news, accommodative response and bad news, accommodative response and good news. It is expected that bad news worsens the effect of communication responses while good news improves it.
At first, it is worth considering the four samples together. Results are reported in Table 5. Results are reported in Table 6. The comparison between groups shows that: 1. Def/bad vs def/good: CAR7 means are -26.13% and -7.68% against an overall mean of defensive group of -18.75% (Table 1). The difference between populations is statistically significant in all the five tests. It can be assessed that with no doubt in this case news affect investors' behavior confirming the hypothesis that bad news worsens the result while good news improves it.
2. Def/bad vs acc/bad: although means are different (defensive lower than accommodative), only one test out of five results significant, indicating that there is no strong evidence supporting that difference. Since in the overall test defensive and accommodative were significantly different, one can say that the effect of bad news appears to outclass the effect of communication.
3. Def/bad vs acc/good. The largest difference can be observed between sample means, and it is statistically significant in five cases out of five. News affect the results and contribute to spread out difference between defensive and accommodative already discussed (see Table 1).
4. Def/good vs acc/bad. CAR7 means are very close and no statistical significance supporting a difference between them has been found. Despite the lack of statistical significance, it should anyway be noted that the mean of the defensive case supported by good news is now greater than that of the accommodative associated with bad news, reminding that without considering news, the defensive response showed a statistically significant lower mean, one can say that the effect of news overrides the effect of communication.
5. Def/good vs acc/good. Although the defensive case shows a lower average CAR7 than accommodative, now they are quite close to each other, and there is no statistical evidence supporting a difference between populations. Good news cancels the effect of communication outclassing it.
6. Acc/bad vs acc/good. Sample means are different, and the difference is statistically significant in three out of five cases. This is what was expected and allows us to say that the effect of news is important.
These results fill a gap in the literature providing the first empirical evidence on news media effect and giving a first confirmation of frame and appraisal theories.

CONCLUSION
The purpose of this paper is twofold: (1) to find out the best company response strategy to minimize the impact on share price when the company is involved in a shock event, and (2) how news media can influence investors' emotions and their perception of company communications.
This study aimed to provide further empirical evidence supporting crisis communication theories and as a new contribution to provide an empirical validation of framing theory and news media effect on share prices, filling the literature gap.
As for the first level of the purpose of this study, the results show that, from an overall point of view, when a shock event occurs, an accommodative response is the best strategy for the company management. Going deeper into the subject and considering different cases of shock events, together with associated emotions, statistical evidence has been found that accommodative responses are better than defensive for scandal and product recall cases, while the communication strategy chosen in case of incident does not definitely affect investors' behavior. All the results are in accordance with theoretical predictions. About news media, the results show that groups with the same type of response prove to be different if they are divided based on news, thus confirming that the effect of the latter worsens (bad) and mitigates (good) the attribution of responsibility by investors. Conversely, compared to the same type of news (def/bad & acc/bad; def/good & acc/good), a statistical consistency that supports the difference in average returns between the groups could not be found; it can therefore be stated in such cases that the effect of the news cancels that of communications overriding their impact on investors' behavior. Finally, comparing the opposite cases (def/bad & acc/good; def/good & acc/bad), there is a further confirmation that in any case of shock event, an accommodative strategy will be more accepted by investors, but if negative opinion is expressed in the news, this ameliorative effect of communication may be overridden by the media's opinion.
Therefore, a company's management has a reference framework based on theoretical considerations empirically validated as a guide for choosing the best communication strategy following various types of shock events. The results suggest that a socially responsible behavior apologizing and struggling to compensate event consequences is better than a denial strategy refusing to assume responsibility. That could also improve reputation among media, which may be induced to be favorably disposed toward the company and less critical, minimizing the effect of the shock.
Future research may be aimed at finding out whether the way in which a company had addressed the market on previous occasions will influence the investor's opinion. In the literature, this component is known as intensifier (Coombs, 2004). Furthermore, the presence of another emotional state known as "schadenfreude" can be verified (Coombs & Holladay, 2005), analyzing competitors' market price behavior.
It will also be interesting to find out if the behavior of investor and the media can be influenced by the previously adopted strategies by the company's management in order to try to guide their opinions.