Better safe than sorry:
How Frames align in long-term Crisis Situations
exemplified by the Diesel Emissions Scandal
Eva Marie Busse 12287334
Master Thesis
University of Amsterdam Graduate School of Communication Master’s programme Communication Science
dhr. dr. Piet Verhoeven
Abstract
This study investigates the frame-building processes of the two domains public relations and news media in the diesel emissions scandal by answering the leading question, if the frames used in order to address the scandal by involved German
carmakers and the German media align. The aim is to research the occurrence of frame alignment in a longitudinal approach in order to determine if alignment can evolve over time and can therefore be persistent for long periods. A manual, quantitative content analysis with 168 articles published by the biggest German newspapers and 69 press releases from five German carmakers was performed. Results confirm the drawn solution of previous research, that frame alignment occurs during crisis situations mainly due to collective sensemaking. A new insight is found that frame alignment also emerges in order not to accelerate crises further, which can particularly be the case when organizations resist the media. The outcome is presented that frame alignment can be persistent in severe crises, since organizations then adapt to the steady framing of the media to avoid reputational risks. Also confirmed was the outcome that in long-term crisis situations, recurring alignment is possible. Frame alignment is identified as a concept that is necessary to acquire better understanding of communication dynamics in far-reaching crisis situations.
Introduction
“Nobody believes in diesel anymore.” These words were spoken by Ferdinand Dudenhöffer, Professor of Automotive Economics at the University of Duisburg-Essen in July 2017 (Spiegel Online, 2017, para. 8) and epitomize the misconduct of a whole industry. Diesel, a technology that was long believed to be superior to its gasoline opponent not only in costs and durability but also in terms of pollution (Jung & Park, 2016), is in a crisis along with the companies that brought it there: Carmakers.
Especially German manufacturers are involved in one of the biggest crises of the automobile industry to date (Fasse, 2019). A crisis can be defined as an event that poses a threat to the organizational reputation, effects the interaction with stakeholders, and generates negative results (Coombs, 2010). German carmakers are facing exactly these reputational threats and negative consequences, making the diesel emissions scandal a recent example of an organizational crisis. The scandal, also referred to as Dieselgate, initially began in the U.S. but mainly developed in Germany, the only country where it triggered a national debate (Presse- und Informationsamt der Bundesregierung, 2018) explaining why this country is in the focus of interest. The scandal began on 18 September 2015 when the Environmental Protection Agency (EPA) accused the German automobile manufacturer Volkswagen Auto Group (VW) of emissions rigging. The EPA detected so-called defeat devices in VW’s diesel engines that could lower the nitrogen oxides emissions (NOx) during laboratory tests (EPA, 2019) by up to 40 percent (Reuters, 2015). After the allegations in the U.S., falsifications were also discovered being practiced in Europe (Cünnen, 2015). The publicizing of the scandal had negative effects for VW: The stock price fell, the CEO had to resign (Jung, Chilton & Valero, 2017), thousands of cars had to be recalled, and a fine of millions of dollars had to be paid (Jung & Park, 2016; Ott, 2018). Besides the financial consequences, the trust of thousands of costumers had been lost and the reputation suffered damage (Jung et al., 2017). The head of VW’s Management Board Hans Dieter Poetsch described the situation as an “existence-threatening crisis for the company” (Schwartz & Cremer, 2015, para. 2).
focus of the media in 2015 (Balser, Fromm & Ott, 2016; Hank & Meck, 2016; Jahn, 2015), but denied being involved in the scandal (Glies & Rebstock, 2015; Hofsommer, 2015; Howe, Brock & Sackmann, 2015). In Germany, official investigations against other manufacturers did not begin until 2017, when new accusations of emissions rigging and concealment led to searches by public prosecutors. In March 2017, the headquarters of Audi AG (Campbell, 2017) and in May the bureaus of Daimler AG were searched. In June allegations against Porsche AG arouse. The last carmaker to make the news in Germany was Bayerische Motorenwerke Group (BMW) after the NGO Deutsche Umwelthilfe (“Environmental Action Germany”) accused the company of also using illegal defeat devices (Amelang & Wehrmann, 2018). All companies besides BMW became officially part of the manipulation, nevertheless every accused carmaker had to recall thousands of its cars for far too high NOx (Esch, 2018) or incorrect software (Hägler, 2018). The scandal had a considerable impact on one of the most important industries in Germany, financially as well as image-related
(Georgievski & Al Qudah, 2016).
If a crisis such as the diesel emissions scandal occurs, a reaction from the organization is advised due to the uncertainty (Stephens, Malone & Bailey, 2005) and negative emotions (Coombs & Holladay, 1996) the crisis produces. Communication is central for this reaction. Crisis communication can protect the reputation or repair reputational damage (Coombs, 2015). In times of mediatization (Ihlen & Pallas, 2014), companies should not take on an organization-based perception but should consider the stakeholder’s point of view. In line with the concept of framing (Entman, 1993), the perception of the public regarding an organizational crisis is highly influenced by the coverage of the media (Kim & Cameron, 2011). Framing is a useful paradigm for understanding strategic creation of communication (Hallahan, 1999) and sensemaking in crisis situations (Van der Meer, 2016). Framing presents certain bits of information to make them more noticeable (Entman, 1993). Frames can influence the public opinion and are used by the two domains public relations (PR) and media. For reputation management during a crisis, media coverage plays an essential role. Stakeholders will mainly learn about a
crisis from the media. The way the media frame the incident, therefore, has a great influence on the reputation and in most crisis situations the media will offer the frame that stakeholders will adopt. Reputational risks can occur, if the media reject the frame given by the organization. Thus, the organization either has to be the first to set the frame or needs to work within the frame given by the media (Coombs, 2007). During a crisis, frame alignment can therefore be expected. Alignment implies that the frames used by different actors become more linked and complementary (Snow, Rochford, Worden & Benford, 1986). This is also due to the uncertainty of the crisis that is attempted to be reduced and due to the aim to solve it (Van der Meer, 2016). The Dieselgate evokes the question of how the German carmakers and the German media reacted to the crisis. Of special interest is how the two domains framed this particular incident and if the phenomenon of frame alignment emerges. Since framing has the potential to shape the public opinion, it is important to gain insight in how
communication emerges between actors engaged in the same crisis such as the
Dieselgate and if the media have the power to set the frame. Furthermore, researching effects of frame alignment is necessary to acquire better understanding of
communication dynamics especially in far-reaching crisis situations. The following research question is proposed:
RQ: Do the frames used in order to address the diesel emissions scandal from
September 2015 until the end of 2018 by the involved German carmakers and the German media align?
The originality of this question lies in the exploration of an industry-wide scandal, that involves the whole German automobile industry, which has never been part of a crisis to this extent before, from a crisis communication perspective. It therefore allows to compare the actions of several companies and the news media in the same crisis in a
longitudinal study in order to highlight the occurrence of frame alignment in a new context. Being an ethical scandal that entails the betrayal of consumers and the German government and that ignores the global efforts to halt climate change (Li, Mcmurray, Xue, Liu & Sy,
2018), a social relevance for researching the Dieselgate is implied. Theoretical Background
Crisis Communication
All organizations can and most likely will experience a crisis at some point (Van der Meer, Verhoeven, Beentjes & Vliegenthart, 2017; Verhoeven, Tench, Zerfass, Moreno & Verčič, 2014). A crisis is “an incident that is unexpected, negative, and overwhelming” (Barton, 2001, p. 2), produces uncertainty (Stephens et al., 2005), and is “a threat to the organizational reputation” (Coombs, 2007, p. 163). Therefore, if a crisis occurs, an organization should react to it to reduce uncertainty, particularly because it is assumed that communication during and after a crisis is a key factor in assessing the long-term impact of it (Coombs, 2015). Additionally, with the emergence of new media that enable fast reporting to surfacing crisis situations (Holladay, 2010), crisis
communication is essential to all organizations to present the event from an organizational perspective (Verhoeven et al., 2014). Crisis communication can be defined “as the collection, processing, and dissemination of information required to address a crisis situation” (Coombs, 2010, p. 20) and can protect the reputation or repair reputational damage (Coombs, 2007). Reputation is a valuable asset that, when strong, was found to have a positive impact on financial performance, the ability to attract customers and employees, and to create competitive advantages (Abratt & Kley, 2012; Coombs, 2007). Reputation can shortly be defined as the “outcome of
interactions between stakeholders and the organisation over time” (Abratt & Kleyn, 2012, p. 1050) and lies in the perception of a company’s stakeholders. Stakeholders, thus, are crucial for the formation of reputation. Consequently, organizations need to acknowledge and know their stakeholders and their ability to negatively affect the performance of the organization, especially during and after a crisis (Van der Meer et al., 2017). The general public as a major stakeholder group mostly retrieve
organization-related information through “indirect experiences such as the news media” (Mason, 2014, p. 80). Hence, the news media are another important stakeholder that can have a direct impact on the reputation of the organization.
Consequently, media coverage is a substantive characteristic of reputation management (Carroll & McCombs, 2003; Coombs 2007). It accordingly becomes clear that
communicating with stakeholders during a crisis is inevitable (Van der Meer et al., 2017) and that especially the media as a stakeholder should be acknowledged (Carroll & McCombs, 2003; Coombs, 2007; Mason, 2014; Van der Meer et al., 2017).
Media Relations in Crisis Situations
As aforementioned, media coverage is important for reputation management during a crisis since most stakeholders receive information from the reports of news media (Holladay, 2010). Not only the organization, but also the media try to make sense of the incident and indicate how stakeholders should interpret the crisis
(Coombs, 2007). The media usually establish the final crisis interpretation because of their trustworthiness and gatekeeping function (Shoemaker & Vos, 2009). Gatekeeping theory illustrates the complex process through which events are covered by the mass media and explains how and why certain information gets media attention. It thus explains the power of the media to report on a crisis and to select which information is made available to the public (Shoemaker & Vos, 2009). Additionally, “organizational crises are likely to be covered by the media because they fit so well with news factors” (Van der Meer et al. 2017, p. 428). Consequently, the media can either put crises in the focus of social awareness or keep them off the public agenda. Since the public is mostly skeptical towards information provided by corporations (Dawkins, 2005), and media coverage has become a necessity to gain corporate legitimacy and reputation (Ihlen & Pallas, 2014), journalists have the ability to influence the public perception in times of a corporate crisis. Organizations, therefore, should be the first source to deliver information about the crisis before the media to reduce reputational damage (Coombs, 2015) and to have a voice in the crisis coverage (Holladay, 2010). By not reacting to the pace of the news media, companies enable the crisis situation to accelerate and lose the ability to influence the perception of stakeholders from an organizational point of view (Holladay, 2010; Van der Meer et al., 2017). The power of the media has also been studied in the theory of agenda-setting. The main
influences their salience on the public agenda” (Caroll & McCoombs, 2003, p. 37). Accordingly, the media shape the environment for institutions like corporations and are crucial for “building normative, regulative, and cognitive bases on which
corporations are evaluated” (Ihlen & Pallas, 2014, p. 7) with increasing significance. This process described as mediatization becomes apparent by the amount of attention and resources that is devoted to media relations by corporations (Ihlen & Pallas, 2014). This devotion is also explained by the impact the media can have on how the public thinks about an issue by presenting it in a certain tone (Kuttschreuter, Gutteling & de Hond, 2011). Media coverage can be positive, neutral, or negative about an issue (Klingeren, Boomgaarden, Vliegenthart & de Vreese, 2015) and even organizations can become subject of this evaluation (Deephouse, 2000). Since most stakeholders learn about a crisis in the news and adopt the interpretation of the media, tone has a significant effect on the public opinion (Nijkrake, Gosselt & Gutteling, 2015), demonstrating the powerful effects of news coverage in a crisis. Besides tone, media coverage is “characterized by the presence of certain news frames” (Nijkrake et al., 2014, p. 81). Framing is an approach that recognizes the influence of the media and further elaborates on the creation of communication.
Framing
In a crisis situation, a dynamic process of sensemaking and meaning construction occurs explained by the ambiguous nature of a crisis (Van der Meer, Verhoeven, Beentjes & Vliegenthart, 2014). The concept of framing is a potentially useful paradigm for
understanding such processes of communication and meaning construction (Hallahan, 1999). Frames structure the social world (Matthes, 2012) by providing context in which information is presented and processed (Hallahan, 1999). Framing consequently and “essentially involves selection and salience” (Entman, 1993, p. 52). To frame means selecting aspects of reality over others and making specific pieces of information “more noticeable, meaningful, or memorable to audiences” (Entman, 1993, p. 53). Thus, frames determine how most people understand the presented information. Framing helps
understanding how PR messages are strategically created, but frames are also commonly used by the media (Hallahan, 1999). “News framing is a vital process used by the news media to relate events in ways that have relevance and meaning for audiences” (Hallahan, 1999, p. 224) and goes beyond agenda-setting and gatekeeping theory by not only deciding which subjects are presented, but also how the mass audience perceives these subjects (Durrant, Wakefield, McLeod, Clegg-Smith & Chapmann, 2003). The news, therefore, have a vital say in how the public understands social reality, making news framing a “tool of power” (Hallahan, 1999, p. 223). Consequently, framing is about the question of which of the domains dominates how the world is seen (Hallahan, 1999). PR and media usually have domain-specific frame building processes driven by their professional identities, resulting in frame differentiation (Van der Meer, 2016).
Especially during highly uncertain events such as crises, framing has the power to shape the public opinion, because it diagnoses causes, makes moral judgements (Entman, 1993), and attributes responsibility (Coombs, 2007). As a result, frame negotiation between journalists and PR practitioners is a common practice (Hallahan, 1999) highlighting the interplay of the domains. Framing of a crisis is an important task of PR with the hope that frames are adopted by the media and, ultimately, by the public (Ihlen & Pallas, 2014). But how the media frame crises, generally has a great influence on the public perception and consequently on the reputation of companies. Quite often the media will offer the frame that most people will adopt. Reputational risk occurs if the media reject the frame given by the organization. Thus, PR practitioners need to work within the frame given by the media if necessary (Coombs, 2007). In the diesel emissions scandal it is assumed, that the media are more stable in their communication and set the frame, that the carmakers adopt over time. This is due to the long-term threat crisis situation, where reputational risks are high (Coombs, 2015) and the lost trust of the public in the
H1. The media change less in their use of frames during the scandal than the carmakers.
Situational Crisis Communication Theory (SCCT) as presented by Coombs and Holladay (Coombs & Holladay, 1996) identifies three crisis clusters that work as a framing approach: Stakeholders either frame the crisis as a victim, accidental, or preventable cluster, each with different amounts of attributed organizational
responsibility (see Table 1) (Coombs, 2007). SCCT is therefore based on attribution theory (Coombs & Holladay, 1996) and is concerned with the attributions made by stakeholders while searching for causes and making sense of the crisis (Coombs, 2007). In an organizational crisis, the amount of responsibility for the incident, that stakeholders attribute to the company, is especially relevant. The stronger the
attribution of responsibility to the company is, the stronger are the public’s feelings of anger, which result in a more negative view of the organizational image (Benoit, Table 1.
Crisis clusters identified in SCCT
Victim cluster: Weak attribution of responsibility for the crisis to the organizations. Organization is (also) seen as the victim.
Accidental cluster: Minimal attribution of responsibility to the organization. The actions leading to the crisis were unintentional.
Preventable cluster: Very strong attribution of responsibility to the organization. The organization knowingly took inappropriate actions that lead to the crisis.
Note. Adapted from “Protecting organization reputations during a crisis: the development and
application of situational crisis communication theory”, by Coombs, W.T., 2007, Corporate Reputation
Review, 10(3), pp. 163-176.
1997). “Communication [then] can be used in attempts to influence a person’s attributions, or the subsequent feelings attached” (Coombs & Holladay, 1996). Both the organization and media decide on a cluster to frame the crisis and consequently attribute responsibility. Crucial is which of the frames becomes the most dominant and will presumably be adopted by the public. To shape the public opinion from an
organizational perspective, SCCT presents recommended crisis response strategies for each crisis type. A non-fitting strategy can have serious consequences: “An
organization that falsely denies responsibility for offensive actions risks substantially damaged credibility if the truth emerges” (Benoit, 1997, p. 184).
Semetko and Valkenburg (2000) defined the most common frames used by European newspapers. These five frames are labelled conflict, human interest, economic consequence, morality, and responsibility frame (see Table 2). An and Gower (2009) investigated the usage frequency of these frames in event of a crisis. Attribution of
responsibility andeconomicframes were identified as the most common frames applied by the news media in a crisis. Due to the uncertainty, it is not surprising that the media as well Table 2.
Five news frames
Conflict frame: This frame emphasizes conflict between individuals, groups, or institutions as a means of capturing audience interest.
Human interest frame: This frame brings a human face or an emotional angle to the presentation of an event, issue, or problem.
Economic consequences frame: This frame reports an event, problem, or issue in terms of the consequences it will have economically on an individual, group, institution, region, or country.
Morality frame: This frame puts the event, problem, or issue in the context of religious tenets or moral prescriptions. Because of the professional norm of objectivity, journalists often make reference to moral frames indirectly—through quotation or inference, for instance—by having someone else raise the question.
Responsibility frame: This frame presents an issue or problem in such a way as to attribute responsibility for its cause or solution to either the government or to an individual or group.
Note. Adapted from “Framing European politics: a content analysis of press and television news”, by
Semetko, H. A., Valkenburg, P. M., 2000, Journal of Communication, 50(2), 93-109.
as the general public attribute responsibility and search for the cause of the event as already mentioned by Coombs (2007). Two other types of frames have been found to impact attributions of responsibility in crisis situations: Episodic and thematic frames (Semetko & Valkenburg, 2000). These differ in the way a crisis situation is presented. The thematic frame places a subject into a broader context, while the episodic frame offers more specific cases or examples to present an issue. Episodic frames are believed to be more emotionally engaging and facilitate a greater persuasive appeal as compared to the human interest frame (Gross, 2008; Semetko& Valkenburg, 2000).
Frame Alignment
The advent of a corporate crisis disrupts routines and the domain-specific processes of PR and media and “creates a need of meaning coherence” (Van der Meer, 2016, p. 21). Collective sensemaking emerges (Van der Meer et al., 2014) that produces “crisis-specific frames and frame alignment” (p. 753). Frame alignment implies that the frames used by different actors become more linked and complementary (Snow et al., 1986). In a crisis situation “frames function to organize experience and guide action” (Snow et al., 1986, p. 464) and to “make sense of the situation at play, limit negative outcomes, and in the end solve the crisis” (Van der Meer, 2016, p. 9). Frames align momentarily to “provide coherence [and] reduce uncertainty” (p. 15). Frame alignment can mostly be found in the early development of a crisis, where searching for information and attributing responsibility dominates the dynamics in both domains, organization and media (An & Gower, 2009; Coombs, 2007; Van der Meer, 2016). Evolvement of a crisis necessarily means that frame alignment is only temporal and cannot be taken for granted, because it is always “subject to reassessment and renegotiation” (Snow et al., 1986, p. 476) by the specific domains. But frame alignment might also occur as the result of pressure from the media: Organizations are forced to adopt the proposed frames by the media at some point, in order to prevent reputational risks (Coombs, 2007), especially in case of long-term threat crisis situations (Coombs, 2015). In this case, frame alignment could evolve over time and could exist for long periods. In order to find out, whether frame alignment also occurs in this way, the Dieselgate is investigated due to its extent. Because of the long timespan and diverse actors involved in the same crisis, the diesel emissions scandal is appropriate to close the research gap and enrich the frame alignment theory.
In the diesel emissions scandal, two separate cases of frame alignment are investigated: First, the alignment of VW’s communication with that of the German media and second, the alignment between the media and the other carmakers Audi,
Porsche, Daimler, and BMW. This is explained by the development of the scandal that initially only concerned VW and only over time evolved into an industry-wide crisis. The Dieselgate, therefore, is structured into three phases: the beginning of the VW scandal, the development of the VW scandal, and the evolvement into an industry-wide scandal. For frame alignment between VW and the media, it is expected that alignment occurs in the first phase of the scandal, where both domains seek for information, attribute responsibility, and collectively make sense. The expected frame alignment is shown in Figure 1. After the first phase, domain-specific processes are foreseen where frames will de-align again. Stated formally:
H2. Frame alignment between Volkswagen and the German media will be higher in phase 1 than in phase 2 and 3.
For the other carmakers it is estimated that frames in the domain itself align based on domain-specific processes. Frame alignment between the carmakers and the media is then for once presumed to occur in the first phase of the scandal due to the dynamics of collective sensemaking. After de-alignment in the second phase, it
emerges again in the last phase because of new allegations and evidence that leads to a re-emergence of the crisis (see Figure 2). The following hypothesis is formulated:
H3. Frame alignment between the manufacturers Audi, Porsche, Daimler, and BMW and the German media is higher in phase 1 and 3 than in phase 2.
It is furthermore suspected that consistent frame categories are adopted in the beginning of the scandal, but that the content of the frames is not necessarily identical, especially in regard to the question of responsibility. At this point, the domains possibly attribute responsibility differently based on their domain-specific identity. However, in the last phase, alignment of attribution of responsibility is expected to occur, in order to solve the crisis:
H4. The alignment of the attribution of responsibility between the carmaker Audi, Porsche, Daimler, and BMW and the German media is higher in the last period than in the first two phases of the scandal.
Methodology Research Aim
The interest of this study lies in the investigation of the Dieselgate in Germany. Since the scandal mainly developed in this country, the focus of interest is justified. Of particular relevance is the crisis communication of the involved German carmakers and
Figure 1. Expected frame alignment between VW and the German media
how the German media reacted to the scandal. The aim is to shed light on how the manufacturers and various newspapers framed the crisis. Furthermore, the occurrence of frame alignment between the companies and the media is explored in a long-term crisis situation, which allows a longitudinal approach to alignment. This enables researching if fame alignment can be persistent over time and thus allows the extension of the frame alignment theory.
Method
The research question is tested by means of a manual, quantitative content analysis of
Phase of the scandal 1 2 3 VW’s domain-specific framing Media’s domain-specific framing VW’s domain-specific framing VW’s domain-specific framing Media’s domain-specific framing Media’s domain-specific framing Frame Alignment Frame De-Alignment No Frame Alignment
Figure 2. Expected frame alignment between the carmakers and the German media
German newspaper articles and press releases of involved German automotive manufacturers from September 2015 to December 2018. This time period is chosen, because it entails the first accusations against VW (thus the beginning of the crisis) and the development into an industry-wide scandal. The sample is divided into three phases: Phase 1 encompasses the beginning of the scandal, which starts with the exposal of VW’s incorrect emissions values in September 2015 and ends in December 2015. The second phase is from January 2016 until February 2017, in which the scandal develops. The last phase begins in March 2017, when the VW scandal turns into an industry-wide scandal in Germany and ends in
December 2018 (Amelang & Wehrmann, 2018). Therefore, the first and the last phase are considered as most important.
A manual content analysis is done, since an automated analysis does not necessarily do “justice to the complexity and depth of language” (Boumans, 2016, p. 27). With an
1 2 3 Media’s domain-specific framing Media’s domain-specific framing Phase of the scandal Frame Alignment Frame De-Alignment Media’s domain-specific framing Carmakers domain-specific framing Carmakers’ domain-specific framing Frame Alignment Carmakers domain-specific framing
automated content analysis, a non-biased demonstration of the extent to which the same topics are covered is given and an unaltered extraction of dominant topics is possible. But the automated analysis does not “take the valence of the texts into account” (p. 27). A manual option instead can extract the evaluation of the topics and the coder has the ability to compare the context in which language is used by the different actors. Additionally, due to the rather small sample, an automated analysis does not seem to be necessary (Boumans, 2016).
Sample
Press Releases. All press releases of the manufacturers from September 2015 to December 2018 thematizing the scandal and published online are selected. As long as models of a specific automotive brand are affected and the brand itself has its own corporate communication department, it is seen individually, apart from the Group it belongs to. This results in five companies of interest in Germany: Volkswagen Auto Group (VW), Audi AG (Audi), Daimler AG (Daimler), Porsche AG (Porsche), and Bayerische Motorenwerke Group (BMW) (MDR, 2018). Due to the small sample of statements given by the
manufacturers, the whole population (sampling unit) of press releases is considered. The unit of analysis results in an overall amount of 69 press releases. These are divided among the manufacturers as follows: VW published 36 press releases, Audi fourteen, Porsche seven, Daimler seven, and BMW five press releases.
Newspaper articles. Since the Dieselgate in Germany is researched, German newspapers are chosen for the analysis of the media coverage. The largest newspapers by circulation are particularly of interest, because they report nationwide and are relevant in terms of the formation of opinions among German citizens. Unfortunately, due to missing access to the largest newspaper BILD DEUTSCHLAND, only the second to fifth biggest newspapers are investigated. These are Süddeutsche Zeitung, Frankfurter Allgemeine Zeitung, Die Welt, and Handelsblatt (Statista, 2019). Additionally, only the online version
of Die Welt, WELT ONLINE, is accessible with the database at hand (NexisUni). To search for relevant articles, keywords such as “Diesel”, “Abgas”, “Abgasskandal” are chosen. To be evaluated, articles need to refer to the scandal, which should become apparent from the headings and/or sub-headings, or by scanning the article. Furthermore, they need to be in German, ideally refer to the manufacturers of interest, and be published in the given time frame. The initial search resulted in a sample unit of 1037 articles. Constructed week sampling, which is a type of stratified sampling, is chosen due to the examination of a large population, which is enabled by the electronic database use and the investigation of daily newspapers in a longitudinal study. The aim of constructed week sampling “in quantitative newspaper content analysis is to maximize sampling efficiency or to obtain the most accurate estimate of the population by sampling the least number of newspaper issues” (Luke, Caburnay & Cohen, 2011, p. 87). Constructed weeks are better suited than random samples in the case of traditional, printed daily newspapers (Hester & Dougall, 2007), as these differ greatly in their content depending on the day of the week. Weekend editions in particular allow for more extensive reporting. In order to take the different editions of daily newspapers into account and since this study mainly analyzes traditional newspapers, constructed week sampling is considered reasonable. Based on the findings of Luke et al. (2011), a stratified sample grounding on a total of six constructed weeks is conducted. Based on the importance of the phases, the sample yields the following: two constructed weeks for the first phase, one constructed week for the second, and three constructed weeks for the last phase. This way, the highlights of the crisis are sufficiently covered. Therefore, 42 days of coverage are investigated for each of the four selected newspapers, resulting in a unit of analysis of 168 articles. Consequently, the whole sample comprises 237 sample units from which 10% (n = 24) are double-coded to assure intercoder reliability by testing
Krippendorff’s Alpha (KALPHA) and Cohen’s kappa (KAPPA). Both methods are
and KALPHA furthermore takes the scale of measurement into account (Hayes &
Krippendorff, 2007). In order to ensure a consistent understanding of the literature between the coders, a coder training is conducted.
Operationalization of Frames and Frame Alignment
Since the aim of this study is to research how the carmakers and the media framed the crisis and if there is frame alignment, measurements for frames and for the concept of alignment as well as an operationalization for the term need to be created.
Operationalization of frames. Frames in this study are constructed via several frameworks: The crisis cluster based on SCCT (Coombs and Holladay, 1996), the thematic focus based on the five news frames (Semetko & Valkenburg, 2000), and the distinction between episodic and thematic frames, resulting in ten frame categories. Furthermore, the evaluation of the scandal and the question of responsibility is considered. To operationalize these variables, a codebook is developed.
Codebook. The codebook contains basic information about the newspaper articles
and press releases such as date, source, and title, as well as all items needed to answer the research question. This includes for once the ordinal variable Evaluation of the scandal with the values negative, neutral, positive, and none, as well as the nominal item Responsibility for the scandal. This variable contains among others all carmakers as categories as well as governments of different countries, the Automobile industry, diesel drivers, or the option none or others (for detailed description see Appendix A). Furthermore, the codebook
contains binary questions to reveal the use of frames which are explained in the following. It additionally includes coder instructions and examples to ensure reliability.
Frame Measurements. In order to measure the extent to which certain frames appear
in the context unit (article or press release), a set of questions is created to which the coder must answer yes (1) or no (0). The answers result from analyzing paragraphs, sentences, and words performing as units of coding. The majority of these questions were taken from the findings of Semetko and Valkenburg, who empirically tested the most ideal questions to
recognize underlying frame concepts in newspapers. They identified 20 questions that measure the five news frames presented in the theory. The questions are constructed in a certain manner and, if necessary, are slightly adjusted to the particular case of this study: Does the article or press release suggest that the automobile industry is responsible for the scandal? (responsibility frame), Is there a mention of financial losses or gains now or in the future? (economic frame), Does the article or press release reflect disagreement between parties/individuals/groups? (conflict frame), etc. Additional questions for the frames not presented by Semetko and Valkenburg are constructed independently (see codebook in Appendix A). To measure each frame, a minimum of two questions, but ideally more, is considered. With this, a coding system is structured that reveals if and to what extent different frames are used by the actors (Semetko & Valkenburg, 2000).
Pre-test. To verify the codebook empirically, a pre-test was carried out. This pre-test
showed that the codebook was largely designed effectively and successfully identified the frame constructs in the texts. Merely the questions defined by Semetko and Valkenburg regarding the responsibility frame were found insufficient for the investigation of this study, as they only showed the existence of the frame if responsibility was actually attributed to an actor. However, during the pre-test, it became apparent that responsibility is often a topic, but is not specifically attributed. For example, manufacturers try to absolve themselves from responsibility or newspapers are looking for responsibility without initially having an
answer. According to the questions adopted, there consequently would be no responsibility frame present. Both coders, however, clearly noticed the presence of the frame. Therefore, a further question was added to the codebook, which is exclusively aimed at the thematic existence of responsibility. According to this definition, the General responsibility frame is also present, if no actor is specifically held responsible, but the topic is addressed in general. After the pre-test, the codebook in total contains – besides basic information – 30 questions
to determine the presence of the now eleven defined frames and the variables Evaluation of the scandal and Responsibility for the scandal.
Intercoder reliability. The results of the intercoder reliability tests for each variable
are presented in Table 3. All variables except the Thematic and Episodic frame have a KALPHA and KAPPA of .80 or above, which is considered as the norm for a good reliability test (Hayes & Krippendorff, 2007). The two other variables are still considered reliable and are used for the analysis. The low intercoder reliability is explained by the fact that both frames occurred simultaneously in some of the units of analysis, and the coders disagreed as to which one dominated.
Table 3.
Intercoder reliability
Variable KALPHA KAPPA
Evaluation .88 .88 Victim cluster 1.00 1.00 Accidental cluster 1.00 1.00 Preventable cluster .84 .84 Thematic frame .59 .59 Episodic frame .59 .59 Conflict frame 1.00 1.00
Human interest frame .86 .86
Economic frame .84 .84
Responsibility frame .86 .86
Morality frame .82 .82
General Responsibility frame 1.00 1.00
Responsibility for the scandal 1.00 1.00
Note. n = 48
Operationalization of frame alignment. Since frame alignment is a rarely researched topic, there is no uniform definition or methodological operationalization of the term. Therefore, alignment in this study is operationalized in regard of the obtained data. Alignment occurs, if actors during a given time period adopt the same or similar frames. By performing a manual content analysis, the occurrence of equal frames and equal content can be uncovered. Frame alignment becomes apparent with answering
the aforementioned questions, to measure the extent to which a frame category appears and is then further identified with two constructs: Correlations and means.
Correlations. Based on previous research, frame alignment is particularly tested
by correlations, as these display the reciprocal relationships between the domains (Van der Meer et al., 2014). Alignment is determined by the strength of the correlations between the used frames in the different phases. To test correlations, the binary questions covering the frames need to be transformed into metric data. It is not reasonable to simply average all questions that cover one frame for each unit of
analysis, since some variables (such as the Victim cluster or the Thematic and Episodic frames) are still constant and therefore nominal after the transformation. Additionally, the sample sizes of the carmakers and the media are unequally distributed with a significant smaller sample size on part of the manufacturers. Since statistical significance in correlations is substantially influenced by sample size, the aim is to standardize and maximize the sample size for all domains. For this reason, frame alignment is operationalized as general alignment that is examined over all frames together. Considered are thus all thirty questions individually for each phase and not as part of the frame category they belong to. A new data set of thirty variables is created by determining the means for each question over all units of analysis for every phase. This way, the new metric data set meets all requirements needed for testing
correlations and determines, if general alignment is present.
Means. In order to find out which of the frames are useddifferently by the domains, means are compared. For this, t-tests for independent samples are considered. Again, a metric data set is required that demands a reasonable operationalization of the term alignment. Since t-tests are more robust with regard to constant variables and unequal as well as small sample sizes, frame alignment for this analysis is investigated using the eleven defined categories. For this, the means of all questions covering one frame are averaged for every single unit of analysis divided by phase. This way, the sample sizes of the domains do not change. Differences in framing are hereafter determined on the level of the individual frame categories. Thus, if means differ
not be present in the particular phase. This new data set is as well used to determine frequencies. Together with the results of the correlations as well as frequencies, the amount of alignment between the domains in each phase can be analyzed.In this study, a significance level of 5% is assumed (p < .05). The Kolgomorov-Smirnov test showed no normal distribution for both new data sets, p < .05. Subsequently, three analyses have been conducted: To test hypotheses 1, 2, and 3, Spearman’s rank correlations and t-tests for independent samples were performed divided by phase. To test hypothesis 4, the nominal variables Responsibility for the scandal and Source were investigated for each phase separately using cross tabulations and Fisher’s exact test explained by the unequal sample sizes.
Results Frequencies and t-test for independent samples
All results of the descriptive analysis are shown in Table 4. In the following, the most important findings of frequencies are presented together with those of the t-test for independent samples.
Phase 1. Results revealed, that in the first phase of the scandal the media (n = 54) evaluated the crisis as highly negative (100%) and framed it as a Preventable cluster (M = 1.00, SD = .00). In the investigated articles in this phase words such as faked, manipulated, and cheated were commonly used, which explain the evaluation. The cluster and evaluation are in line with the most dominant frames in this phase, which are the General responsibility frame (hereafter referred to as GRF) (M = 1.00, SD = .00), the Responsibility frame (M = .61, SD = .23), and the Economic frame (M = .37, SD = .49). In this phase, the media attributes deliberate responsibility for the scandal in particular to VW and talks about economic consequences.
VW (n = 15), too, assessed the scandal as mainly negative (73,3%), or didn’t evaluate it (26,6%) and also framed it as preventable (M = .78, SD = .42) showing their regret. The GRF (M = .80, SD = .41) and the Responsibility frame (M = .65, SD = .37) are as well most dominantly used by VW. The results of the t-test also show the approximation of frames used by the two domains: In the first phase, there are only
Ta b le 4 . Fr a me s di vi ded by dom ai n and phas e and ra n ke d b y fre q u en cy Me d ia ( n = 1 6 8 ) Ca rm ak er s (n = 3 3 ) VW ( n = 3 6 ) Ph a se 1 G R F (M = 1 .0 0 , SD = .0 0 ) Th em at ic f ra m e (M = 1 .0 0 , SD = .0 0 ) G R F (M = .8 0 , SD = .4 1 ) Pr ev en ta b le c lu st er ( M = 1 .0 0 , SD = .0 0 ) G R F (M = .8 3 , SD = .4 1 ) Pr ev en ta b le c lu st er ( M = .7 8 , SD = .4 2 ) Th em at ic f ra m e (M = . 6 9 , SD = .4 7 ) Vi ct im c lu st er ( M = .5 0 , SD = .5 5 ) Th em at ic f ra m e (M = .6 7 , SD = .4 9 ) Re sp o n si b il it y f ra m e (M = .6 1 , SD = .2 3 H IF (M = .2 9 , SD = .4 0 ) Re sp o n si b il it y f ra m e (M = .6 5 , SD = .3 7 ) Ec o n o m ic f ra m e (M = .3 7 , SD = .4 9 ) Pr ev en ta b le c lu st er ( M = .2 8 , SD = .4 8 ) H IF (M = .4 7 , SD = .3 8 ) Ep is o d ic f ra m e (M = . 3 3 , SD = . 4 8 ) Mo ra li ty fra m e (M = .2 2 , SD = .3 4 ) Ep is o d ic f ra m e (M = .3 3 , SD = .4 9 ) Co n fl ic t fr am e (M = .2 0 , SD = .4 1 ) Re sp o n si b il it y f ra m e (M = .1 9 , SD = .3 3 ) Mo ra li ty f ra m e (M = .0 7 , SD = .2 6 ) H IF (M = .1 9 , SD = 0 .3 9 ) Co n fl ic t fra m e (M = .1 7 , SD = .4 1 ) Co n fl ic t fr am e (M = .3 1 , SD = .2 6 ) Mo ra li ty f ra m e (M = .0 4 , SD = .1 5 ) Ec o n o m ic f ra m e (M = .1 1 , SD = .2 7 ) Ec o n o m ic f ra m e (M = .0 7 , SD = .2 6 ) Vi ct im c lu st er ( M = .0 0 , SD = .0 0 ) Ep is o d ic f ra m e (M = .0 0 , SD = .0 0 ) Vi ct im c lu st er ( M = .0 0 , SD = .0 0 ) Ac ci d en ta l cl u st er ( M = .0 0 , SD = .0 0 ) Ac ci d en ta l cl u st er ( M = .0 0 , SD = .0 0 ) Ac ci d en ta l cl u st er ( M = .0 0 , SD = .0 0 ) Ph a se 2 G R F (M = 1 .0 0 , SD = .0 0 ) G R F (M = .8 0 , SD = .4 5 ) G R F (M = .7 1 , SD = .4 7 ) Pr ev en ta b le c lu st er ( M = 1 .0 0 , SD = .0 0 ) Vi ct im c lu st er ( M = .8 0 , SD = .4 5 ) Pr ev en ta b le c lu st er ( M = .6 0 , SD = .5 1 ) Ep is o d ic f am e (M = .6 6 , SD = .4 8 ) Th em at ic f ra m e (M = .8 0 , SD = .4 5 ) Ep is o d ic f ra m e (M = .6 0 , SD = .5 1 ) Re sp o n si b il it y f ra m e (M = .5 9 , SD = .3 0 ) Co n fl ic t fr am e (M = .6 0 , SD = .5 5 ) Re sp o n si b il it y f ra m e (M = .5 4 , SD = .4 1 ) Ec o n o m ic f ra m e (M = .40, SD = .5 0 ) Ep is o d ic f ra m e (M = .2 0 , SD = .4 5 ) Ec o n o m ic f ra m e (M = .4 0 , SD = .5 1 ) Co n fl ic t fr am e (M = .3 8 , SD = .4 9 ) H IF (M = .1 5 , SD = .3 4 ) Th em at ic f ra m e (M = .4 0 , SD = .5 1 ) Th em at ic f ra m e (M = .3 8 , SD = .4 8 ) Re sp o n si b il it y f ra m e (M = .0 7 , SD = .1 5 ) H IF (M = .1 3 , SD = .3 0 ) H IF (M = .2 4 , SD = .3 9 ) Pr ev en ta b le c lu st er ( M = .0 7 , SD = .1 5 ) Co n fl ic t (M = .0 9 , SD = .2 9 ) Mo ra li ty fra m e (M = .2 0 , SD = .1 2 ) Ec o n o m ic f ra m e (M = .0 0 , SD = .0 0 ) Mo ra li ty ( M = .0 5 , SD = .1 7 ) Vi ct im c lu st er ( M = .0 0 , SD = .0 0 ) Mo ra li ty f ra m e (M = .0 0 , SD = .0 0 ) Vi ct im c lu st er ( M = .0 0 , SD = .0 0 ) Ac ci d en ta l cl u st er ( M = .0 0 , SD = .0 0 ) Ac ci d en ta l cl u st er ( M = .0 0 , SD = .0 0 ) Ac ci d en ta l cl u st er ( M = .0 0 , SD = .0 0 ) Ph a se 3 G R F (M = .95 , SD = .21 ) G R F (M = .7 3 , SD = .4 6 ) G R F (M = .67, SD = .5 2 ) Pr ev en ta b le cl us ter ( M = .84 , SD = .37 ) Th em at ic f ra m e (M = .6 4 , SD = .4 9 ) Ec o n o m ic f ra m e (M = .6 7 , SD = .5 2 ) Ep is o d ic f am e (M = .7 6, SD = .4 3 ) Ep is o d ic f ra m e (M = .3 6 , SD = .4 9 ) Pr ev en ta b le c lu st er ( M = .5 0 , SD =. 5 0 ) Re sp o n si b il it y f ra m e (M = .48 , SD = .38 ) Pr ev en ta b le c lu st er ( M = .3 5 , SD = .4 7 ) Th em at ic f ra m e (M = .5 0 , SD = .5 0 ) Co n fl ic t fra m e (M = .4 7 , SD = .5 0 ) Re sp o n si b il it y f ra m e (M = .3 3 , SD = .4 0 ) Ep is o d ic f ra m e (M = .5 0 , SD =. 5 0 ) Ec o n o m ic fra m e (M = .45 , SD = .4 0 ) Ec o n o m ic f ra m e (M = .2 3 , SD = .4 3 ) H IF (M = .3 8 , SD = .4 1 ) Th em at ic f ra m e (M = .24 , SD = .43 ) Vi ct im c lu st er ( M = .2 3 , SD = .4 3 ) Re sp o n si b il it y f ra m e (M = .3 3 , SD = .3 7 ) H IF (M = .2 3 , SD = .40 ) Co n fl ic t fr am e (M = .1 4 , SD = .3 5 ) Mo ra li ty f ra m e (M = .3 3 , SD = .3 7 ) Mo ra li ty f ra m e (M = .05 , SD = .17 ) H IF (M = .0 4 , SD = .0 9 ) Co n fl ic t fr am e (M = .0 0 , SD = .0 0 ) Vi ct im c lu st er ( M = .01 , SD = .05 ) Mo ra li ty f ra m e (M = .0 3 , SD = .1 4 ) Vi ct im c lu st er ( M = .0 0 , SD = .0 0 ) Ac ci d en ta l cl u st er ( M = .0 0 , SD = .0 0 ) Ac ci d en ta l fr am e (M = .0 0 , SD = .0 0 ) Ac ci d en ta l cl u st er ( M = .0 0 , SD = .0 0 )
differences between the Human interest frame (hereafter HIF), t(69) = -2.43, p = .018, d = .28, the Economic frame, t(30.438) = 2.31, p = .028, d = .39, and the Morality frame, t(15.241) = -2.59, p = .020, d = .53. Therefore, the majority of frames does not differ. At this point, it should be noted that all variables used for testing were not normally distributed, Kolmogorov-Smirnov test, p < .05.
The carmakers Audi, Porsche, Daimler, and BMW (in the following only referred to as carmakers) (n = 6) did not evaluate the scandal at all (66,6%), or assessed it negatively (33,3%). In contrast to the other two actors, they dominantly framed the crisis in the first phase as a Victim cluster (M = .50, SD = .55), followed by the Preventable cluster (M = .28, SD = .48) indicating their attempt to prove their unguiltyness. The other carmakers also mainly used the GRF (M = .83, SD = .41), but only used the Responsibility frame as the fourth most frequent frame (M = .19, SD = .33), which suggests that the manufacturers are addressing responsibility, but do not consider themselves responsible. If the results of the t-test are considered, it becomes clear that the carmakers differ slightly from the media in their use of frames: Significant differences are found between the use of crisis cluster, where the media mainly apply the Preventable cluster (M = 1.00, SD = .00), t(5.000) = 3.99, p = .010, d = .87, and the carmakers the Victim cluster (M = .50, SD = .55), (t(60) = -7.36, p < .001, d = .69. Furthermore, a different utilization of frames was found between the Episodic, t(55.000)= 5.10, p < .001, d = .57, and Thematic frame, t(55.000) = 5.10, p < .001, d = .55, and the use of the Responsibility frame, t(60) = 4.26, p < .001, d = .48.
Phase 2. In the second phase, the media (n = 29) again only evaluated the scandal as negative (100%) and preventable (M = 1.00, SD = .00), and predominantly used the GRF (M = 1.00, SD = .00), Responsibility frame (M = .59, SD = .30) as well as the Economic frame (M = .40, SD = .50). The media have remained consistent in their communication.
VW (n = 15) almost equally evaluated the crisis as negative (33,3%), neutral (26,6%), and not at all (40%) indicating a difference compared to the beginning. Regardlessly, only the Preventable cluster (M = .80, SD = .45) was used again. VW also used the GRF (M = .71, SD = .47), Responsibility frame (M = .54, SD =. 41), and
Economic frame (M = .40, SD = .51) most often. A common trend between the two domains VW and media can be observed, which is also confirmed by the t-test: Both domains only differed in the use of the Preventable cluster, t(12.000) = 3.21, p = .008, d = .68, the Conflict frame, t(37.030) = 2.36, p = .023, d = .36, and the HIF, t(12.000) = 2.31, p = .040, d = .55.
The carmakers (n = 5) withheld from evaluating the scandal in this phase and increasingly framed the crisis as a Victim cluster (M = .80, SD = .45) and rarely ever as preventable (M = .07, SD = .15). They primarily used the GRF (M = .80, SD = .45) and the Conflict frame (M = .60, SD = .55), but again infrequently applied the
Responsibility frame (M = .07, SD = .15). The data suggests that the carmakers
consider themselves as victims of news reporting which is supported with expressions such as other manufacturers hurt the image of the whole industry. A significant
difference in using the Responsibility frame compared to the media was also confirmed by the t-test, t(31) = 5.35, p < .001, d = .69. Further differences are again evident in the use of the crisis cluster (Victim cluster, t(4.000) = -4.00, p = .016, d = .89, Preventable cluster, t(4.000) = 14.02, p < .001, d = .99), and the Episodic, t(31) = 2.09, p = .045, d = .35 and Thematic frame, t(31) = -2.09, p = .045, d = .35. Additionally, a significant difference was found in the utilization of the Economic frame, t(27.000) = 4.18, p < .001, d = .63. Hence, more frames differ than compared to the first phase. The effect size Cohen's d also points to strong differences in this phase with values mainly above .50.
Phase 3. Consistent with their previous evaluation, the media (n = 85) evaluated the scandal as negative (91,76%) or not at all (8,24%) in the last phase. The
Preventable cluster is also still predominant (M = .84, SD = .37). Surprisingly, the Victim cluster (M = .01, SD = .05) is used for the first time in this phase, which can be explained by the wrong accusations against BMW, which were withdrawn after
prosecution. Also consistent are the frames used by the media: The GRF (M = .95, SD = .21) and Responsibility frame (M = .48, SD = .38) are again the most frequently used ones, followed by the Conflict frame (M = .47, SD = .50) and the Economic frame (M = .45, SD = .40). The increased use of the Conflict frame in the last phase results
from comprehensive newspaper coverage on charges and lawsuits.
VW (n = 6) still has a balanced evaluation of the crisis (negative 33,3%, neutral 16,6%, positive 33,3%, and neutral 16,6%), but is the first to rate the scandal positively and the only domain to do so, indicating a development in its communication. The dominant frames of VW are the GRF (M = .67, SD = .52), the Economic frame
(M = .67, SD = .52), and the HIF (M = .38, SD = .41). The Responsibility frame on the other hand is used less frequently than before (M = .33, SD = .43). When comparing the means of both domains - media and VW - in the last phase, only one significant difference can be found between the use of the Conflict frame, t(83.000) = 8.70, p < .001, d = .69.
The carmakers (n = 22) still rarely evaluated the scandal (81,82%) or assessed it negatively (18,18%). The data indicates a change in the framing of the crisis cluster: In the third phase, the carmakers predominantly used the Preventable cluster (M = .35, SD = .47) over the Victim cluster (M = .23, SD = .47). This is in line with the dominant frames. The GRF (M = .73, SD = .46) as well as the Responsibility frame (M =. 33, SD = .40) are the most frequently used frames in this phase indicating the carmakers’ acceptance of responsibility. The Economic frame (M = .23, SD =. 43) is the third most used frame which can be attributed to the increased presence of reports on fines and damage payments. The trend in the carmakers’ framing is seen in Table 4. In contrast to the previous phases, the results of the t-test - especially the effect size Cohen's d – point to less strong differences between the framing of the media and the carmakers. Other than in the second phase, almost all differences were found to have a weak effect strength of under .50 indicating a weaker contradiction in the domains’ communication than before. Although a total of eight frames differ in use (see Table 5), Cohen's d points to small differences at the end of the scandal. A development is particularly evident in the use of the Responsibility frame, where the weakest difference is found in the last (d = .22) and the strongest in the second phase (d = .69). None of the domains ever used the accidental cluster. Moreover, looking at frequencies, responsibility is most often attributed or thematized by all domains and economic consequences are discussed. By examining the results of the Spearman's rank correlations frame
alignment can be determined and the hypotheses can be validated. Table 5.
T-tests for independent samples between framing of the media and the carmakers Phase 1 Phase 2 Phase 3 t d t d t d Victim cluster -7.36* .69 -4.00* .89 -2.49* .48 Preventable cluster 3.99* .87 14.02* .99 5.50* .48 Thematic frame -4.90* .55 -2.09* .35 -4.89* .43 Episodic frame 5.10* .57 2.09* .35 4.40* .37 Conflict frame .36 .05 -.85 .15 3.66* .47 HIF -.63 .08 .45 .07 6.33* .57 Economic frame 2.18 59 4.18* .63 -.47 .05 Responsibility frame 4.26* .48 5.35* .69 2.33* .22 Morality frame -1.32 .51 .42 .08 .42 .04 GIF 1.00 .41 1.00 .45 2.59* .47 Note. *p < 0.5, VW excluded.
Spearman’s rank correlations
Hypothesis 1. To test hypothesis 1 and to determine if the media change less in their use of frames than the carmakers, Spearman’s rank correlations were conducted within the domains. Therefore, the communication within the domain media was compared between phase 1 and 2, between phase 2 and 3, as well as between phase 1 and 3. The process was also carried out with the other two actors, carmakers and VW. The analysis showed that the strongest, statistically significant correlations of frames between phases were found in the media’s communication, indicating little changes in frames. The data also confirmed that the media was the first domain to report on the crisis. VW’s framing also changed only slightly, with strong correlations between each of the phases. Only the carmaker’s framing indicates disagreement and significant change over time. Correlations between the second and the third phase revealed a weak, statistically not significant relationship, rs = .28, p = .138. A detailed description of this analysis is seen in Table 6. From the table it can be concluded that the used frames by the carmakers have changed from the second to the last phase. If frequencies are considered, the conclusion is drawn, that the frames by the carmakers converge with those of the media in the last phase. In the end, the GRF, the Responsibility
frame, and the Economic frame were among others the most frequent used frames by all domains (see Table 4) acknowledging responsibility for the crisis. The media, therefore, change less in their use of frames, supporting hypothesis 1.
Table 6.
Spearman’s rank correlations within the domains
Phase 1 – 2 Phase 2 – 3 Phase 1 – 3
Media .93* 1.00* .93*
VW .88* .85* .77*
Carmakers .57* .28 .48*
Note. * p < .05
Hypothesis 2. In order to examine frame alignment between the domains and to test hypothesis 2 and 3, Spearman’s rank correlations were performed between the domains. Results revealed that in all three phases there was a strong, positive and statistically significant correlation between the frames used by the media and VW, indicating frame alignment throughout the phases. The highest correlation (and therefore frame alignment) was found in phase 2, rs = .90, p < .001, followed by the first phase, rs = .86, p < .001. In the last phase, the weakest Spearman’s rho was examined, rs = .70, p < .001. Hypotheses 2, that suggested that frame alignment
between VW and the German media will be higher in the first than in the other phases, consequently needs to be rejected. Frame alignment is lowest in the last phase, but - opposing the expectations - highest in the second phase of the scandal. It is also, contrary to the initial assumption, consistent over the scandal.
Hypothesis 3. Correlations between the framing of the German media and the carmakers were only found to be statistically significant in the last phase, where a strong, positive relationship was found, rs = .75, p < .001. A weak, positive relationship was also confirmed in the first phase of the scandal, that was marginally significant, rs = .33, p = .077. In the second phase, no significant relationship was found, rs = .05, p = .778. With regard to frame alignment, frames between the two domains strongly converge in the last phase, which suggests that frame alignment is present. An
approximation of frames is also visible in the first phase, but it is not as strong as in the last. Therefore, frame alignment is also present in the beginning of the scandal, but less
distinctive. Only in the second phase, there is no alignment. Hypothesis 3, stating that frame alignment between the manufacturers Audi, Porsche, Daimler, and BMW and the German media is higher in phase 1 and 3 than in phase 2, can thus be confirmed. A comparison of all correlations can be found in Table 7.
Table 7.
Spearman’s rank correlations between the domains
Phase 1 Phase 2 Phase 3
Media – VW .86* .90* .70*
Media – Carmakers .33 .05 .75*
Note. *p < .05
Hypothesis 4. To answer hypothesis 4 and to determine, if alignment of the
attribution of responsibility between the carmakers Audi, Porsche, Daimler, and BMW and the German media is higher in the last period than in the first two phases of the scandal, the two-sided Fisher’s exact test and the measure of association Cramér's V were examined using cross tabulations between the nominal variables Responsibility for the scandal and Source.By using a filter, VW as a source was excluded from the investigation. Statistically significant relationships between the media’s and the carmakers’ attribution of responsibility were confirmed in all three phases. A positive and strong relationship was found in phase 1, p < .001, V = .72. The strongest positive relationship between both variables could be determined in phase 2, p < .001, V = .94, whereas results for the last phase only confirmed a moderate positive relationship, p < .001, V = .53. The results thus indicate that the strongest difference in attribution of responsibility is in the second, the smallest difference in the last phase. Cramér’s V furthermore illustrates a change in the last phase, where the media and carmakers approach each other, and the carmakers start
attributing responsibility to themselves. As shown in Table 8, both domains among others attribute responsibility to Audi, Porsche, or the whole automobile industry. Hypothesis 4 is thus confirmed. This result corresponds to that of the correlations. The whole
completeness, the information from VW is also included. Never considered responsible were BMW, the U.S. government, or diesel drivers.
Table 8.
Attribution of responsibility divided by domains and phases
Media (n = 168) Carmakers (n = 33) VW (n = 36) Phase 1 VW 86,6 16,6 81,3 Audi 13,8 16,6 0 Automobile Industry 1,6 0 0 German Government 1,6 0 0 None 1,6 66,6 18,7 Phase 2 VW 72,2 16,6 60,0 Audi 13,8 16,6 0 Automobile Industry 5,5 0 0 German Government 8,3 0 0 None 0 66,6 40,0 Phase 3 VW 45,4 0 50,0 Audi 18,5 31,8 0 Porsche 2,0 9,0 0 Daimler 13,4 0 0 Automobile Industry 11,3 4,5 0 German Government 1,0 0 0 None 14,4 54,5 50,0
Note. Figures in percent. Only categories that have been mentioned are included.
Conclusion and Discussion
The current study has been conducted to research framing of the diesel emissions scandal and to investigate the occurrence of frame alignment between the two domains public relations and media. This study hence contributes to the still under-researched topic of frame alignment. The aim of this study was to gain insight in how communication and framing emerges between actors engaged in the same long-term crisis. Particularly interesting in this study was the investigated crisis, which was a persistent scandal, that included several different actors. This way, the dynamics in the same domain as well as between different domains such as PR and media can be observed in a longitudinal approach. For this, the communication of the carmakers VW, Audi, Porsche, Daimler, and BMW was compared with that of the largest German
newspapers in circulation by answering the following research question: Do the frames used in order to address the diesel emissions scandal from September 2015 until the end of 2018 by the involved German carmakers and the German media align? The research question was approached with four hypotheses. Results confirmed three of them and showed that frame alignment was present between the carmakers and the media during the scandal, thereby answering the leading question. In order to examine how frame alignment was exactly manifested and which additional insights were gained, the hypotheses are discussed in the following and an overarching conclusion about the results is drawn.
Hypothesis 1 dealt with the change in the domain’s communication and expected the media to be more stable in their framing and therefore the initiator of frames. Since several theories such as agenda-setting, gatekeeping theory, or SCCT identify the media as one of the most important stakeholders of corporations, that can have a serious impact on the reputation (Carroll & McCombs, 2003; Coombs, 2007; Shoemaker & Vos, 2009), determining the initial source of the used frames is
important. The assumption of hypothesis 1 was confirmed and thus indicates that the media is probably the initiator of the frames rather than the manufacturers. It was also detected that the media was the first to report on the crisis. For VW, following the frames of the media can be explained by the attempt to reduce negative threats and, in line with SCCT, repair the reputational damage. VW admitted to its wrongdoings and worked within the frames and tone of the media, especially because there was simply evidence that spoke against the manufacturer. Hence, following Benoit (1997) and Coombs (2007), VW reacted according to theory. But, following SCCT, VW could have had more influence on the media’s and presumably other stakeholders’
evaluation, if it would have been open about all violations in the beginning. Instead, the investigated newspaper articles revealed that investigations increasingly discovered misconduct on the part of VW, which made it impossible to end the crisis at an early stage and instead intensified it. Nevertheless, the obtained data suggests that VW’s alignment with the media possibly helped VW to stay out of the main focus in the last phase, when the crisis re-emerged due to new evidence against other manufacturers.
However, due to the nature of news factors and the length of the scandal, a decrease in the negative evaluation of VW by the media was nevertheless expected (Kepplinger, 2008; Van der Meer et al., 2017). Audi, Porsche, Daimler, and BMW were inconsistent in their communication and are thus not believed to be the initiators of the aligning frames in the end of the scandal. A noticeably difference within the selected frames of the carmakers was especially found between the second and the last phase. This can be explained by the evidence against the manufacturers, the aim to solve the crisis, and by the pressure of the media that holds potential negative threats to the reputation, if the carmakers continue to contradict them. Not admitting to their malpractice from the beginning on, led the crisis to accelerate and made the manufacturers the continuous focus of speculation, which had a negative impact on their reputation. In line with SCCT (Benoit, 1997; Coombs, 2007), denying responsibility at first led to a harsher reaction of the media in the end of the scandal.
Hypotheses 2 assumed frame alignment between VW and the media. Based on previous research (Van der Meer, 2016; Van der Meer et al., 2014), frame alignment was expected to occur in the first phase of the scandal due to the dynamics of collective sensemaking. Afterwards, domain-specific framing was predicted, which results in de-alignment. The hypothesis could not be confirmed, since alignment was found in all three phases. This finding is therefore not consistent with the findings of previous studies. The outcome can possibly be explained by the fact that the severity of the crisis and the ongoing reporting on it did not allow VW to frame it
domain-specifically. It is important to notice at this point that the crisis was extensively covered by the media in Germany, which is linked to gatekeeping theory and agenda-setting. The German media put the crisis on the public agenda in all four years and raised social awareness of the scandal, while presenting the crisis in a mostly negative tone, which was most likely adopted by the general public. Therefore, VW continued to admit its fault and adapted to the general tone, especially to that of the media, in order to prevent further backlashes and not to further accelerate the crisis. VW’s mainly consistent communication over the years therefore can be explained by the
