After the storm : the consequences of online petitions for corporate reputation

(1)

After the storm:

The Consequences of Online Petitions for

Corporate Reputation

Thesis Master of Science Business Administration International Management Track

Student: Philipp Gerstheimer Student No.: 11418400

Thesis Supervisor: Dr. Michelle Westermann-Behaylo Second Reviewer: Dr. Mashiho Mihalache

Submission Date: 26 January 2018 Academic Year: 2017 – 2018

(2)

2

Statement of originality

This document is written by Student Philipp Gerstheimer who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Index of tables and figures

Table 1 – Country distribution of the sample ... 27

Table 2 - Descriptive statistics excluding dummy variables ... 30

Table 3 – Descriptive statistics including dummy variables ... 30

Table 4 - Distribution of frequencies I ... 31

Table 5 - Distribution of frequencies II ... 31

Table 6 - Correlation matrix ... 34

Table 7 - Multicollinearity ... 35

Table 8 - ANOVA statistic ... 37

Table 9 - Regression coefficients ... 37

Table 10 - Model summary moderated regression ... 40

Table 11 - Interaction term key ... 40

Figure 1 - Strategic Cognition, Issue Type and Responsiveness ... 16

Figure 2 - Conceptual model ... 22

(5)

5

Abstract

For decades, corporate scandal literature has explored the characteristics of issues, life cycles, and how activists address their concerns towards firms and their impact on business performance. However, the rise of web 2.0 and social media has increased the speed of information diffusion around the globe and improved the accessibility of information for everyone. Yet, the impact of this drastic increase in transparency and consumer criticism on the trajectories and outcomes of corporate scandals has received little academic attention. By employing a quantitative empirical case study on the effect of e-petitions on corporate reputation, the purpose of this research is to explore the missing link between the two fields. Based on stakeholder concern theory, it is hypothesised that a more substantial firm response to an e-petition leads to greater benefits in online consumer sentiment. The used sample consists of 134 e-petitions filed on the website change.org against a Global Fortune 500 firm. The cases are categorised along the three levels of respective corporate responses: absent, symbolic and substantive reactions. Corporate reputation was measured by scraping Twitter posts and using textual analytic software to determine their sentiment. The study found significant statistical evidence for an increase in online consumer sentiment and thus corporate reputation for symbolic corporate responses only. The findings suggest that social media listening is an important dimension of corporate scandal management. Moreover, the results indicate that substantive responses should be avoided if symbolic responses provide a reasonable alternative to diminish the intensity of the issue.

Keywords: e-petitions, corporate scandals, corporate reputation, crisis management, crisis response, issue life cycles, social media, stakeholder issues

(6)

6

1 Introduction

“It is not a crisis that does damage,

it is the response to a crisis that does damage to a company” (Galloway, 2017)

“I’d like my life back” said former CEO of British Petroleum (BP) Tony Hayward while giving an interview in response to the 2010 oil spill disaster in the Gulf of Mexico (Reed, 2012). The tragedy was a consequence of the Deepwater Horizon drilling rig blowout, resulting in 11 casualties and severe, long-lasting damage to the marine ecosystem. Hayward’s unfortunate choice of words in combination with his decision to take a day off to go sailing at the height of the crisis are two of the most outstanding symbols of the BP PR disaster. As a result of their stakeholders’ discontent with the firm’s crisis management, BP’s share value dropped by more than 50% and its corporate image sunk to an all-time low (Muralidharan, Dillistone, & Shin, 2011).

More recently in 2014, the fashion make Zara, a brand of the Spanish garment giant Inditex, caused a scandal when selling a shirt for children. The shirt was striped blue and white, with a prominent yellow star appliquéd on the front. Within hours of being released for sale, stakeholders expressed their deep discontent about the product through various social networks, as it reminded many of the uniform Jewish forced labourers had to wear inside German concentration camps. To channel the consumers’ discontent and increase the pressure on Zara to react, an e-petition against the firm was filed on the petition hosting website change.org (Edelman, 2014). Zara released polyglot reaction statements on Twitter the same day. The company apologised and emphasised the shirt was supposed to be reminiscent of a Western sheriff’s outfit. Zara also promised to remove the product from all retail outlets. In spite of the offense caused, Zara’s immediate and resolute reaction received a positive echo in the media (Cresci, 2014; Grindberg, 2014). Comparing one corporate crisis with another is difficult, especially in terms of responding to the discontent of the masses. But at the same time, what can be observed is how a firms’ response to a social issue can influence stakeholder sentiment, particularly in the fast-paced social media age.

It is insightful however, to explore the Zara case in more detail. This is because the firm is known to make (online) word-of-mouth marketing strategies a high priority (Royo‐ Vela & Casamassima, 2011). In this regard, Z. Chen & Berger, (2013) provided evidence that moderate controversy, such as that surrounding the Zara shirt, drives conversation and word-of-mouth (WoM). For two days Inditex’s share price experienced a small drop, before recovering and

(7)

7 then increasing steadily in the subsequent days. Hence, at least from the perspective of the financial markets, no sustainable damage to the firm could be reported.

Approaching corporate crises from a more theoretical perspective, business literature has examined the corporate processes of issue management in the pre-social-media era and provided frameworks such as issue life-cycle models which support management to react in a damage minimising manner (Litrico & David, 2016; Mahon & Waddock, 1992). However, the disruptive power of information and communication technologies (ICT) have decreased the timeframe available for companies to respond without losing control (Kaplan & Haenlein, 2010). Moreover, the shift towards the digital landscape has also affected the relative importance of different channels used for corporate communication and interaction between the firm and the consumer. Research has shown there is a benefit for companies in employing social media communication strategies which are built around personalised customer responses (Hewett, Rand, Rust, & van Heerde, 2016). However, this type of research has been conducted in the field of marketing, which is mostly aimed at refining advertising and corporate image enhancement strategies. Hence, the findings are predominantly related to proactive behaviour as opposed to the reactive nature of crisis response.

On the contrary to the developments in the marketing field, management scholars have continued to examine and improve corporate defence mechanisms allowing firms to minimise the imminent threat of traditional media moderated stakeholder sanctions (i.e. An, Gower, & Cho, 2011; McDonnell, 2015; McDonnell, King, & Soule, 2015). Moreover, research was done on the cognitive perception of issues by management and potential opportunities or threats stemming from these interpretations (Bundy, Shropshire, & Buchholtz, 2013). However, the developments modern ICT have scarcely been incorporated into the literature.

Hewett et al. (2016) set out to increase the understanding of the interplay between social media and corporate communication by demonstrating that the effectiveness of a message through traditional media can be supported by a thorough social media communication strategy. The finding suggests that a corporate social media issue response could also benefit crisis management. However, there has been little exploration of the impact of firms’ responses to social media mediated crises. Since the digital environment has become increasingly essential for brand communications, this research will investigate the impact of corporate responses on online consumer sentiment in reaction to social media mediated petitions, so called e-petitions.

On a theoretical level, this research makes a contribution to the literature by examining to what extent international corporations can use the new media environment to diminish the threat potential of social media petitions, or even benefit from petitions. In this regard, Bundy

(8)

8 et al. (2013) asked for a better understanding of the relationship between issue salience and corporate responsiveness. Issue salience describes the degree of resonance and prioritisation a stakeholder issue experiences among corporate management. However, Bundy et al. (2013) excluded the dimensions of contemporary transparent and immediate consumer-firm interactions through social-media channels. Incorporating modern dimensions of ICT suggests a more thorough understanding of issues transported and moderated by social media vehicles for managements. Moreover, appropriate crisis response might yield more than sole damage minimisation and thus could lead in certain scenarios to an overall increase of corporate brand value. Hence, the research question is: “To what extent do e-petitions affect corporate business

performance?”.

The paper is organised as follows: in the first part, the literature on the effects of imminent issues on corporations is reviewed. This section is followed by an examination of social media petitions to give firm responses context and elaborate on the missing link shaped by the disruptive power of ICT. The second part will provide an analytical framework for the research and offer theoretical evidence for the threatening, but also beneficial characteristics of e-petitions for corporations. The fourth chapter familiarises the reader with the underlying research methods. This chapter includes the coding of the sample and the composition of the subsample, which is necessary to analyse the effects of firm actions in response to e-petitions on corporate brand value. In the fifth part, the research hypothesis is tested and statistical evidence for its support is provided. The paper closes with a discussion on the results, the theoretical and practical implications of the findings, and suggestions on areas for future research.

(9)

9

2 Literature review

2.1 Corporate reputation

Measuring the effect of a crises on corporate performance is a multidimensional task. While stock price changes require effort to monitor, tracking corporate reputation is far more multifaceted (Fombrun, Ponzi, & Newburry, 2015). This complexity results in numerous definitions for corporate reputation by different business disciplines. Business strategy defines it as “(…) intangible assets that are difficult for rivals to imitate, acquire, or substitute, and so create mobility barriers that provide their owners with a sustained competitive advantage” (Fombrun, Gardberg, & Sever, 2000, p. 243) . In contrast, marketing understands it as “(…) the corporate associations that individuals establish with the company name (Fombrun et al., 2000, p. 243). Synergising the various associations towards corporate reputation, Fombrun et al. (2000) developed a tool to empirically measure the reputation of a brand. The authors categorised corporate reputation along the dimensions of emotional appeal, products and services, vision and leadership, workplace environment, social and environmental responsibility and financial performance. These different dimensions denote the multifaceted characteristic of corporate reputation. Therefore, a thorough determination requires both quantitative and qualitative metrics beyond simple financial numbers (Basdeo, Smith, Grimm, Rindova, & Derfus, 2006; Rindova & Martins, 2012).

2.1.1 Traditional communication channels

According to Barney (1991), a firm’s reputation can be a source for sustainable competitive advantage. However, building a positive reputation to attract stakeholders requires long-term commitment and careful management (Ang & Wight, 2009), especially as negative events have the ability to damage that reputation (Dowling, 2004). A trigger will cause a change in corporate reputation, which in turn will make stakeholders react emotionally towards a stimulus that alters their perception of a firm in a positive or negative manner (Fombrun et al., 2015). In the context of this research, online stakeholder petitions are issues addressed towards firms, which have the potential to elicit a reputational change. Therefore, any given issue needs a vehicle in order to forward the concern towards the management of a firm.

Traditional vehicles for stakeholder concerns range from stakeholder influencing strategies (Frooman, 1999), to non-government-organisation (NGO) moderated activist tactics (Rehbein, Waddock, & Graves, 2004). The first can cover so-called usage strategies, in which a firm attaches a conditional string to the purchase of a resource; for example, the transition of a supplier towards a more ethical meat producer in the mid-term. NGO tactics include

(10)

10 campaigns such as boycotts, vendettas or letter-writing campaigns. On this subject, many of those ‘on-the-spot’ actions can be organised by a small group of individuals and include protest marches or leafleting. Nevertheless, these protests generally require thorough planning and physical commitment (Van Aelst & Walgrave, 2001). A further form of stakeholder influencing strategies are shareholder resolutions. Since the 1980s, and particularly during the early 1990s, the rise and increasing influence of shareholder resolutions in corporate governance was recognised by several scholars (i.e. Davis & Thompson, 1994; Goranova & Ryan, 2014). As Goranova et al. (2014) note, shareholder activism spans from the expression of discontent (David, Bloom, & Hillman, 2007; David, Hitt, & Gimeno, 2001) to the causing of actual strategic change in corporate management (Westphal & Bednar, 2008).

Lee & Lounsbury (2011) look at the topic from the stakeholder’s perspective. They describe three social movement mechanisms that could be applied to modify the effect of a resolution in the activists’ favour. Firstly, disruption to routine can gain the management’s attention by increasing the transaction costs of operational routines without initially winning proxy votes (Luders, 2006). Nonetheless, activists commonly seek to provide convincing arguments for the overall benefit of the organisation (M.-D. P. Lee & Lounsbury, 2011). Secondly, framing of the issue is used to present the issue to the management in a more rational and logical manner, as a means of eliminating toxic, and/or emotionally-driven arguments. Lastly, social movements try to increase third party support via the mobilization of relevant

constituents. As their own decision-making power is limited, this mechanism is often used to

increase public pressure on the management, forcing it to act. It incorporates critique, denunciation, or boycott invocations of influencers (M.-D. P. Lee & Lounsbury, 2011; Luders, 2006).

A shareholder can express his or her concern by informing the board directly, filing the issue in the annual statements of publicly traded companies or by creating media attention. Activist groups often utilize both options to increase pressure on the firm’s management (den Hond & de Bakker, 2007). Concerns arise once shareholders notice a shortcoming of the corporation in relation to its duties. Therefore, shareholder resolutions frequently seek to influence the balance between the profitability and the legitimacy of a firm, in favour of the latter (Uysal & Tsetsura, 2015). In the literature, shareholder resolutions are often related to models of issue life cycles, which typically cover four different stages (Rehbein et al., 2004; Wartick & Mahon, 1994). Tombari (1984) provides an early example of an issue life cycle model. The first stage of his model encompasses a corporate issue, starting with the heightened interest of the corporation’s stakeholders, in regards a societal concern. The growing awareness

(11)

11 leads to the second crisis stage, which manifests as increased attention from the public. During the third stage, the so-called trial period, a firm induces a solution to the concern. Following the intense activity surrounding the concern comes a period of relaxation, from which three possible scenarios can evolve into the fourth stage. Public awareness may return to an apathetic state, a level at which no further issue is noticed. However, the concern can also remain salient, characterised by a medium intensity level or additionally intensified in the case where the solution offered by the firm does not calm the concerns of the stakeholders (Mahon & Waddock, 1992).

The literature discussed thus far shares a common trait in elucidating the lines of communication between the firm and its stakeholders at a certain point of time. However, brand communication is only discussed in the sphere of traditional media channels as this literature was published before the emergence of social-media or does not give attention to it. There is limited literature which explores the changes that can stem from this new form of issue vehicles.

2.1.2 Social media

The power shift between the firm and its stakeholders favours the stakeholders, owing to the disruptive power of social-networks, which has led to the establishment of new forms of issue vehicles, as described by Labrecque, vor dem Esche, Mathwick, Novak, & Hofacker (2013). One such example of a new vehicle is the e-petition hosted on a website like change.org. It is important to distinguish between the process of creating a petition, or seeding phase, and amplifying its strength by creating stakeholder awareness, leading to the breeding phase. Both seeding- and breeding mechanisms have the potential to add additional complexity to corporate crisis management due to the relevance of social-media for corporate communication (Berger & Milkman, 2012). An issue addressed in a e-petition can also be addressed in a shareholder resolution, as the two are not mutually exclusive. Beginning with emails and hyperlinks, message transportation systems have evolved to electronic message boards (Kahn & Kellner, 2004), blogs and today’s social networks (i.e. Facebook, RenRen) or micro-blogging social networks (i.e. Twitter, Weibo) (Burgess, 2015). Likewise, dedicated online petition platforms have emerged such as change.org, 38 Degrees or Avaaz.org (Chadwick & Dennis, 2017).

Fombrun et al., 2000 introduced one of the first multidimensional approaches to measure corporate standing, through which textual sentiment analysis of social media contents has been established as a novel form of determining reputation. The technique quantifies the prevailing notion towards a firm expressed by its stakeholders online (Peleja, Santos, & Magalhães, 2014). The micro-blogging service Twitter has emerged as the most fruitful source for user sentiment analyses (Ghiassi, Skinner, & Zimbra, 2013). In this regard, the overall

(12)

12 process of monitoring social media content is defined as social listening (Schweidel & Moe, 2014). Although the procedure is similar for most online venues (i.e. blogs, for a, social networks), there are differences in the general sentiment levels across them. For example, the sentiment expressed on a blog is generally more positive than the one expressed in an online forum. Therefore, Schweidel et al. (2014) suggest consistency in the venues listened to and a multi-platform approach in order to avoid misleading results.

2.1.3 E-petition characteristics

Although petitions generally represent discontent of stakeholders with a firm, an undertaking must evaluate different threat potentials for each individual concern, both due to limited resources and the nature and rationality of the request. In this regard, Eesley & Lenox (2006) provided evidence that a higher relative power of a stakeholder group increases the likeliness of a positive response of a corporation. Moreover, an increased legitimacy of the request and a higher evaluation of its urgency make firm reactions more probable (Mitchell, Agle, & Wood, 1997). Looking at corporate related e-petitions, little is known about their effectiveness and to what extent this effectiveness is moderated by the total number of supporters, what is sometimes characterised as slacktivism. This stakeholder behaviour can be defined as: “(…) a willingness to perform a relatively costless, token display of support for a social cause, with an accompanying lack of willingness to devote significant effort to enact meaningful change.” (Kristofferson, White, & Peloza, 2014, p. 1149). In contrast to the article of Kristofferson et al. (2014), Y.-H. Lee & Hsieh (2013) found that an individual who signs a petition is more likely to donate money for that petition’s related cause. This suggests a direct relationship between the number of backers and the possible impact of the petition on the firm’s reputation, as slacktivism seems to go beyond mere token support. However, the mutual relational ties between e-petitions and traditional mass media communication channels are yet to receive thorough investigation.

Since e-petitions are often related to the democratisation of societies and the empowerment of the individual, the literature has focussed predominantly on politically motivated e-petitions, overlooking e-petitions concerned with corporate issues. Whether there are differences in their characteristics, success rates and impacts has not been examined. Nevertheless, scholars could find a direct relation of success rates and the number of subscribers when analysing petitions published on the e-petition platform of the British government, Downing Street e-petitions (Escher, 2011; Navarria, 2010). However, a study by Wright (2012) intended to test the analytical framework on global democratic features (i.e. the direct vote system in Switzerland or Downing Street’s e-petitions) introduced by Smith (2009), suggests that Smith’s conclusions

(13)

13 may be partially biased. This is because the proposed framework does not control for duplicate petitions that get automatically rejected by the platform and the English government. Moreover, qualitative research suggests that the ecosystems of e-petition platforms are dominated by power-users who rely on their own vast social media networks to promote petitions they created themselves (Graham & Wright, 2014). As those individuals tend to be among the first to hand in a petition, they are less prone to suffer from rejection in return. The latter concern aligns with literature investigating the success of social media topics (Goel, Anderson, Hofman, & Watts, 2015). Online diffusion of information is predominantly caused by the large broadcast of one or a few single users and not achieved through broad structural virality. Hence, there is a premise to question whether results obtained from political e-petitions can be translated directly to corporation related e-petitions. Therefore, firms must be cautious when trying to understand why some petitions transpire to be more threatening than others.

2.2 The materiality of corporate response

In 1994, the conflict between the Zapatista Army of National Liberation and Mexico became one of the first internet empowered social movements. The Mexican government was pressurised by international investors to stop the war against the minority group. New technologies enabled the guerrillas to publish and share eye-witness information over the internet. By making it accessible to a large international audience, it enabled the guerrillas to shift sympathies in their favour and deter international capital (Martinez-Torres, 2001). Nowadays, advancements in ICT have led to the establishment of all kinds of social networks. From there, it was only a small step towards crowd-sourced activism. E-petitions are one form of this new, technology-backed stakeholder engagements. Although initially rooted in the political sphere, online petitions have also become a tool to address issues in the corporate landscape. A corporate issue is defined as “(…) a controversial inconsistency based on one or more expectational gaps (…) involving management perceptions of changing legitimacy and other stakeholder perceptions of changing cost/benefit positions (…)” (Wartick & Mahon, 1994, p. 309). Thus, corporate issues are an old phenomenon, yet their communication and impact have undergone significant changes during the recent decades.

The shift in communication patterns has resulted in a faster, more frequent and more impactful publication and transportation of corporate issues (Ingenhoff, 2013). From a corporate perspective, mass communication creates awareness among a firm’s stakeholders and can thus positively or negatively affect the reputation of the undertaking (Malmelin, 2007). Therefore, media reputation must be considered as organisational capital and, if used properly,

(14)

14 a resource for competitive advantage according to the resource-based view (Deephouse, 2000). The growing rate of recurrence of public issue awareness receives further significance as the influence of corporate reputation for corporate success continues to increase (Brammer & Pavelin, 2006). Hence, both the awareness and the impact of stakeholder concerns on corporate performance is growing. As a result, the task of corporate marketing has gone beyond battling for customers and total sales numbers. It has also moved to a general strategy that aims to increase attractiveness for investors (Helm, 2007). For example, transparent communication of sustainable corporate activities is one of the most significant marketing trends for the future (Achrol & Kotler, 2012). Therefore, active and reactive corporate messaging must become key functions of modern business management.

Despite recognising the change in communication behaviour, academic literature has mainly focused on defence mechanisms used by firms to protect themselves against social activism. Targeted corporations can respond to activists’ boycott calls by increasing their own pro-social claims (McDonnell & King, 2013). Moreover, those companies tend to respond to claims by installing social management devices to demonstrate commitment to social issues and deflect prevailing accusations made against them (McDonnell et al., 2015). A further protection mechanism of boycott-prone corporations is corporate-sponsored boycott. This method is used by denounced firms to disrepute their business rivals and to overshadow their own discretisation (McDonnell, 2015). Business literature has not yet fully explored the immediate outcomes of corporate responses to social activism. Therefore, evidence for the legitimacy of defensive tactics being the sole most appropriate approach is absent. As discussed, social media sentiment can be a volatile index to measure these outcome effects as consumers are able to express their emotions towards a firm in real-time.

Given the significant attention concerning the negative implications of stakeholder petitions, the question emerges whether social-media-amplified stakeholder concerns suggest a threat to corporations per se. Existing literature mainly describes the shareholder’s interest in maximising their share value, resulting in the agency problem (C. X. Chen, Lu, & Sougiannis, 2012; Jensen & Meckling, 1976). Consequently, numerous requests made by shareholders towards the management are driven by either short- or long-term profit maximisation (i.e. Eisenhardt, 1989; Ross, 1973). Nevertheless, these requests can also be related towards a social issue, which can cover various activities such as strategies towards labour improvements, emission reduction or donations of assets. In this context, Hillman & Keim (2001) found that social issue participation by a firm is negatively related to the maximisation of shareholder value. Examples of social issues in the context of their research are circumventing nuclear

(15)

15 energy or avoiding business relationships with countries accused of human rights violations. Thus, they are similar to issues articulated on e-petition platforms. However, their results conflicted with a meta-analysis on the relationship of corporate social performance and corporate financial performance by Orlitzky, Schmidt, & Rynes (2003), which did not include the study of Hillman et al. (2001). In contrast to the Hillman et al. article, Orlitzky et al. (2003) could find a general positive relationship between corporate social responsibility (CSR) and shareholder value creation. Furthermore, Orlitzky et al. (2003) showed that firms are not penalised by the market for showing a social commitment. As Hillman et al. (2001) suggested, different measures of financial performance, such as return on assets (ROA), return on equity (ROE) or Market-to-Book analysis, should be subject to further research. Therefore, the question whether corporate social commitment contributes to a firm’s performance in a positive or negative manner remains open for discussion.

2.2.3 Issue salience

Typically, the aim of any petition is to trigger a positive response of the accused firm to satisfy the expressed request (David et al., 2007). However, a response can also be negative, or not given at all. Likewise, shareholder and stakeholder resolutions demand for firm actions. Firm responses to issues can be characterised via three different levels; they can either be

substantive, symbolic or absent as depicted in Figure 1 (Bundy et al., 2013). This classification

describes the materiality of corporate response. A symbolic response is characterised by a verbal or written statement that does not place the issuer under any obligation. This type of response is commonly concerned with expressing regret or intentions for future action. A substantive response on the other hand, often takes place in form of a negotiation to achieve a consensus with the stakeholder. This can be achieved by removing the inconsistencies between both parties regarding corporate actions or with a distinct strategic change in response to the demand of the stakeholder. Alternatively, a firm can pursue accommodative or defensive tactics. These actions are concerned with embracing the idea behind a petition or actively reducing the threat of an issue. A decrease in threat level can be achieved by a firm through acknowledging its own misbehaviour, trying to mitigate the matter or de-emphasise the stake in it. If the response is

symbolic, firms can choose whether to take defensive or accommodative action (Bundy et al.,

(16)

16 Issue salience, or the

urgency and commitment with which a firm deals with a stakeholder issue, is an important moderator for stakeholder concerns addressed to firms, as it affects the cognitive interpretation of an issue by the corporate management (Bundy et al., 2013). One dimension of the proposed framework by Bundy et al. (2013), along with organisational identity, is the dimension of strategic frames

- that is the rationality of a corporation to consider an issue as salient when it offers the opportunity of a strategic goal (Polletta & Jasper, 2001). Therefore, the two most extreme outcomes of salience judgement by corporate management is either a true opportunity, if both the strategic frame and the organisational identity are entirely consistent with the issue, or a true threat, in case they are conflicting (Figure 1).

Hence, a salient issue stemming from a stakeholder petition can provide an opportunity for corporate management to increase shareholder value. According to Sarasvathy, Dew, Velamuri, & Vebkataraman (2010) there are three views of entrepreneurial opportunity. The

allocative view sees an opportunity as the “possibility of putting resources to good use to

achieve given ends” (Sarasvathy et al., 2010, p. 82). The discovery view considers an opportunity to be the “possibility of correcting errors in the system and creating new ways of achieving given ends” (Sarasvathy et al., 2010, p. 82). Lastly, from the creative view’s perspective, opportunities are possibilities “of creating new means as new ends” (Sarasvathy et al., 2010, p. 82). The discovery and creative view distinguish themselves from conventional definitions, such as the marketing opportunity, which considers opportunities to be areas where undertakings can leverage customer needs to increase their profitability (Kotler, 2012). Therefore, exploring the beneficial character of e-petitions against corporations provides a promising research premise. In this regard, the sole existence of an opportunity does not always imply that it should be seized upon, due to two different reasons (March, 1978). On the one hand, an individual or an entity can voluntarily resist to act. On the other hand, an opportunity, and/or threat, might not be recognised as such since managers can be bounded by their

Figure 1 - Adapted Strategic Cognition, Issue Type and Responsiveness Framework

(17)

17 rationality (DiMaggio & Powell, 1983; Kahneman, 2003). Hence, an unanswered question emerging from the theoretical framework of Bundy et al. (2013) is whether it can also be applied to online petitions. Furthermore, it is unclear whether the character of a corporate e-petition, that is an opportunity or a threat, is moderated by the increased speed and transitory nature of news transportation by social networks.

2.3 Problem Statement

Fournier & Avery (2011) point out that firms had to realise that the web 2.0, and the beginning of the social media age, was designed for people and not for brands. It was intended to link and exchange information, making crowd activism and e-petitions possible (McNutt, 2014; Patrikarakos, 2017). By giving individuals the ability to connect and organise themselves, the modern internet has created a power swing towards the side of the consumers, resulting in increased public criticism and transparency (Kwak, Lee, Park, & Moon, 2010). As individuals leave their opinions all over the web, online user sentiment has become parallel to traditional consumer sentiment an important determinant for brand equity value (Keller, 2001; Keller & Lehmann, 2003; Orban, Nagy, Kjarval, & Sanchez, 2014). In response to the existing gaps in the literature, two research questions have been developed for this report, and they are as follows:

RQ1: “To what extent do social-media-moderated firm responses to e-petitions affect corporate business performance?”

RQ2: “What could be the benefits for international corporations from the new media environment in the event they are exposed to a social media petition?“

(18)

18

3 Analytical framework

Management literature has invested a substantial amount of resources in exploring the effects of crises on firm performance. Consequently, response models and best practices have been developed, which are thought to help an organisation to minimise the impact of an issue and diminish its threat potential (Bundy et al., 2013; Seeger, 2006). Likewise, a separate stream of academia has explored the intrusion of social media into the corporate communication environment (Laurell & Sandström, 2017). The web 2.0 has empowered stakeholders and thus supported consumer transparency and public criticism. Firms have come to realise that there is no alternative to relinquishing control and adapting their communication strategies (Fournier & Avery, 2011). Although modern issue vehicles have substantially increased the speed and spread of information, little is known about their degree of efficiency. Especially in the corporate landscape, contemporary research is often related to stakeholder issue theories from the 1980s and 1990s.

Further ambiguity arises as international management literature mostly relates the effect of stakeholder issues to threat prevention and avoidance for corporate management. This conflicts with corporate social responsibility theory, which has found evidence for positive spillovers for firms, if resolutions are addressed accurately (Orlitzky et al., 2003). Thus, perceived corporate threats might only be one side of a double-sided coin. The literature falls short in providing theoretical frameworks that allow for the categorisation of online petitions and/or highlight patterns to predict the threat or opportunity potential of social media transported issues. In this regard, Bundy et al. (2013) emphasised the need for a better understanding of the relationship between issue salience and corporate responsiveness. Proffitt & Spicer (2006) suggested incorporating the inception and mobilisation process of a petition into the research agenda, as it might influence the contemporary understanding of issue salience. However, scant literature has explored the connection between these two schools of thought. The importance of this missing link is further emphasised as new media receives increasing priority in modern corporate communication strategies (Dijkmans, Kerkhof, & Beukeboom, 2015; Hewett et al., 2016).

The anxiousness of corporations regarding the threat potential of social issues is not without a reason. Direct competitors that are able to offer a close substitute to what the accused firm markets, can profit from positive market share spillovers (Piazza & Jourdan, 2017). This is for example the case, when the petitioned undertaking fails to demonstrate remedial actions to the public, such as an official apology or changes in management (Connelly, Ketchen,

(19)

19 Gangloff, & Shook, 2016; Piazza & Jourdan, 2017). However, some examples of firm actions in response to petitions indicate that during these times, there might be even a benefit for corporations. In 2012, the online petition platform change.org hosted a petition against the American toy manufacturer Mattel. The petition asked for a bold Barbie®, designed to be a better companion for children suffering from chemotherapy (Bingham, 2012). In another case, Hasbro was faced with a petition on change.org demanding a gender-neutral version of its children’s toy the Easy-Bake Oven®_{. The company responded to the 45,000-vote backed}

initiative by offering additional versions of their product in blue and black (Driscoll, 2012). Hence, given both companies are for-profit organisations and the fact that the two products are still available, suggests that the firms are experiencing a benefit from marketing those products.

Whether these substantive responses were for-profit (Hasbro) or non-profit (Mattel), it seems likely that both contributed to an increase in the emotional brand value and thus the reputation of the two firms. In this regard He, Zhu, Gouran, & Kolo (2016) found that social responsibility is positively associated with the consumer-brand relationship. Their research provided evidence that cause-related marketing allowed firms to improve their moral evaluation by consumers. Moreover, it also positively moderates the intention to purchase various offers from the brand in question. Thus, the two companies most likely had the opportunity to leverage the e-petitions by piggybacking the viral nature of the campaigns and their wide stakeholder audiences, as the petition platform served as a strong broadcaster to stakeholders informing them about the cause-related marketing effort made (De Bruyn & Lilien, 2008). Another non-product related example was Greenpeace’s quest for a name for a humpback whale in 2007. The naming was regarded as a symbol to raise awareness against the Japanese government to hunt down 50 humpback whales in the South Pacific Ocean. In an open online poll, the name

Mister Splashy Pants was suggested and heavily favoured. Greenpeace hesitated to accept the

name initially yet bowed to the pressure of several social communities (i.e. reddit.com, Facebook). Today, Greenpeace’s late acceptance is seen as a successful management move, as the global awareness caused by the odd name served the success of the campaign in the long-term (Ohanian, 2009; Springer, 2015).

(20)

20 The previous examples can all be categorised as substantive responses to social media moderated petitions. Their outcomes suggest a certain benefit for business performance, either in terms of revenue, perceived corporate reputation or both. As it remains unclear how to quantify this benefit and whether there is even a different outcome between symbolic and substantive responses, the first hypothesis is formulated as:

Hypothesis 1: The more substantial the materiality of a corporate response to an e-petition, the greater the benefit for a firm in terms of its corporate reputation.

While there is academic evidence for a direct positive relationship between the number of supporters and the success of a political e-petition (Escher, 2011; Navarria, 2010; A. Smith, Fischer, & Yongjian, 2012; G. Smith, 2009) other literature cautions against translating these findings from the theoretical to the practical sphere without further control (Graham & Wright, 2014; Wright, 2012). Moreover, a study on the effectiveness of e-petitions published on change.org could not find statistical significance for a positive subscriber audience – petition

success relationship (Elnoshokaty, Deng, & Kwak, 2016). However, as this study did not filter

for a certain type of petition and instead scraped the platform through its application programming interface (API), further research on this relationship is required.

There are many unknowns in relation to the underlying patterns that cause petitions to attract user attention and increase participation rates. Taking the article of Wright (2012) as a basis, it can be assumed that the more supporters a petition has acquired, the higher is the likeliness that key influencers are inside this pool of stakeholders. A key influencer is a person who can broadcast his opinions to many people, for example through a vast personal social-network (Kietzmann, Hermkens, McCarthy, & Silvestre, 2011). Therefore, the moderating role of participation rates are subject to further investigation as they might be a misleading metric towards the perceived threat potential of a petition (Chadwick, 2008; Wright, 2012).

Since literature has not fully explored the moderating effect of participation rates on the success of online petitions, this paper seeks to improve the categorisation of its salience for management. Therefore, the second research hypothesis is formulated as following:

Hypothesis 2: A higher participation rate in online petitions positively moderates the relationship between materiality of corporate response and corporate reputation.

(21)

21 It is unsurprising that the likeliness of a campaign going viral increases in correlation with the amount of global attention it receives. In this regard, Grégoire, Salle, & Tripp (2015) identified six different types of social media complaining situations. Although generally focussed on the complaints of individual customers, it may serve well as a foundation for the life-cycle analysis of e-petitions. After an initial service failure, customers voice their issues through social-media networks, such as Twitter or Facebook. If the concern is not addressed adequately, the customer starts ‘bad-mouthing’ about the issue through other networks that rely on a word-of-mouth structure (i.e. Pinterest, Instagram, Flicker). Moreover, he or she will proceed to the tattling stage involving third-party services, such as blogs or petitions websites (Grégoire, Tripp, & Legoux, 2009). At this point, the accused firm is at the crossroad to resolve the issue or let the complaint become viral. If the issue is at the threshold of a viral publication, the initial service failure has proceeded to the spite stage. The authors recommend transparent communication regarding the steps the firm intends to take to resolve the matter paired with reasonable compensation. In the case that the firm fails to prevent a viral break-out, the issue will ultimately advance to the feeding the vultures stage. From there, the firm has lost control and its competitors will take advantage. However, it is important to note that the article by Grégoire et al. (2009) automatically treats a viral campaign as a threat. Thus, although a firm may resolve a conflict favourably, the initial damage still requires resolution first (Joireman, Grégoire, Devezer, & Tripp, 2013).

In the event of a corporate crisis, two effects play a significant moderating role. Firstly, social media allows for the inception of a parallel paracrisis, which is a publicly observable crisis for any individual with non-restricted internet access (Coombs & Holladay, 2012). Secondly, the mass media - social media relationship receives significant importance. As social media news is picked-up by traditional mass-media, the fourth estate transports the issue to the offline world, resulting in additional public attention and increased institutional interest (Pang, Nasrath Begam Binte Abul Hassan, & Aaron Chee Yang Chong, 2014). The article concludes that conventional-media coverage stems from factors like celebrity involvement, human interest in conflicts, and the novelty of the issue or massive customer dissatisfaction.

(22)

22 As the impact of the traditional media – social media’s relationship on online petitions is still relatively unknown in academia, the question arises to what extent online user sentiment is isolated from traditional media coverage for stakeholder resolution scenarios, which leads to a third research question:

Hypothesis 3: A higher proportion of online word of mouth directly concerned with the petition positively moderates the relationship between materiality of corporate response and corporate reputation.

3.1 Conceptual model

The hypothesised relationships between the discussed theories are visualised in a conceptual model, depicted in Figure 2. A higher materiality of a corporate response to an online petition is expected to have a positive effect on the firm’s reputation. This relationship is positively moderated by a higher participation rate for a petition and negatively moderated by a higher proportion of online word-of-mouth directly concerned with the respective petition.

(23)

23

4 Research design

4.1 Sample and data collection

This paper employs a self-coded dataset, which has been developed in cooperation with a total of six Master’s students from the University of Amsterdam in 2017. Thereby, the English version of the social-petition website change.org was scraped regarding any petition against one of the Global Fortune 500 firms between the years 2007 to 2017 that sought to initiate change on a social level. Thus, petitions addressing product related issues were not included in the data set. Although there are multiple other e-petition websites in operation, change.org has more than 100,000,000 users, and recognises itself as the biggest campaign hosting website (Change.org, 2017). In total 2,466 petitions could be coded.

Stemming from this data set, a sub-sample was drawn and included any petition that was labelled as ‘being successful’. This labelling was done by the respective petition filer and published on the petition page hosted by change.org. The new sub-sample contained 233 cases, a success rate of around 9.5% for all coded petitions. Out of these successful petitions, 44 were categorised as having evoked a symbolic firm response, 161 led to a substantive response and 33 did not trigger a corporate reaction at all. As this sub-sample was created to cover cases of symbolic and substantive material responses of a firm, the 33 cases without a reported corporate response were deleted. Such cases can be explained by missing information that was not provided, or incorrect labelling by the petition filer. In order to achieve more equality among the remaining category sizes of the sub-sample (symbolic and substantive) and to ensure variety, the 161 cases from the substantive category were randomly sampled down by the divisor of 3.5, leading to integer 46 cases.

A third category was added to the sub-sample to control for the effects of a successful petition on user sentiment in comparison to the effects of requests that remained unanswered. Therefore, 46 unsuccessful petitions with the most subscribers were added to the sub-sample. As a result, the sub-sample size consists of n=134 cases, which are distributed into: [1] 46 cases without a firm response, [2] 42 cases with a symbolic firm response and [3] 46 cases with a substantive firm repose. The original amount of 44 symbolic cases had to be reduced by two, as no social media listening was possible for these due to an insufficient number of Tweets.

To measure the impact of an e-petition on the corporate reputation of a targeted firm, the echo in the social media landscape was captured and analysed. As Hewett et al. (2016) argue, Twitter data is a fitting proxy for social media online word-of-mouth, that is “any positive or negative statement made by potential, actual, or former customers about a product

(24)

24 or company, which is made available to a multitude of people and institutions via the Internet” (Hennig-Thurau, Gwinner, Walsh, & Gremler, 2004, p. 39). One of the reasons for Twitter’s suitability is its focus on the information and communication exchange between individuals and corporations (Kwak et al., 2010). As such, 80% of all Twitter posts contain a brand name (Orban et al., 2014). Secondly, Twitter broadcasts, the user’s content publicly by default (Hughes, Rowe, Batey, & Lee, 2012). Thirdly, although Twitter’s API settings do not permit scraping of historic content that has been published longer than seven days ago without further financial commitment, the website’s html architecture allows running programmed scripts to scrape the desired content with moderate technical expertise. To fetch the Twitter data, published content is collected in a mining or scraping phase, in which all necessary content is gathered. The data is then analysed by textual analytic and textual processing software to determine the sentiment of each stakeholder’s contribution.

To work around Twitters restrictive API infrastructure, a self-programmed Python script was employed to collect historic Tweets addressed to the petitioned firm. The program was designed to scrape any public message addressed directly at the respective firm, which is indicated in a Tweet by an “@” followed by the corporate account name (i.e. @cocacola). The scraped time frame covered a six-day period, consisting of the two days before the day a petition was labelled successful and three days after it was resolved. In rare cases, scraping Tweets directly addressed to a firm was not possible. This was the case because at the time of the success note’s publication, the undertaking in question did not have a corporate Twitter account. This occurred particularly for older petitions, owing to the lesser importance of social media at the time. In these cases, the search was amended to a hashtag in combination with the firm name (i.e. #JPMorgan). Further complexity occurred as it was difficult to determine a period for social media listening to the online chatter regarding an unsuccessful petition. This is because change.org does not provide information concerning completion dates to those stakeholder demands. To circumvent these limitations, the average duration between the publication of a successful petition and its respective corporate response was calculated at 143 days. Hence, the Tweets between the 141th and 146th day after an unsuccessful petition was published were scraped. An overview of the search tags that were used is displayed in Appendix A.

(25)

25

4.2 Variables

4.2.1 Dependent variable – Corporate reputation

The dependent variable [DV] of the analysis is operationalised as the percent change in user sentiment in the selected social media environment. Sentiment can be measured by analysing a text fragment, which means looking for keywords that indicate a positive, negative or neutral notion of a message or passage (Lai & To, 2015). An analysis can either be done manually or by using lexical concept mapping software. For this research, the latter of the two was conducted by employing the tool MonkeyLearn®_{, which is based on a machine learning}

algorithm continuously improving the quality of the analysis by developing and training an own text analysis module for each research project. In this regard, the sentiment data had a consistency of 84.21% between the Tweets classified as positive, negative or neutral by MonkeyLearn® and a manual sorting. The consistency was tested by drawing 100 random Tweets and coding these along the same three sentiment levels. The discrepancy between the automatically coded values and the real values can be explained to a large extent by the rhetorical use of irony and sarcasm, as both make it difficult for an algorithm to determine the true sentiment of a text.

Hence, each of the Tweets in the 134 Twitter samples was coded with a notion. To calculate a sentiment score, a positive output was assigned with the value “1”, a neutral with the value “.5” and a negative output was given the value “0”. The percental change of each Twitter set was computed with the two sentiment scores that could be derived from each sample, resulting from the prior-to-the-success and post-the-success timeframe (in the case of the non-successful samples the assumed peak of social media echo). Since the research is interested in the changes in user sentiment, the method is appropriate as it is not context dependent on the general sentiment level of a firm. The method further explains why two symbolic cases had to be removed from the sample, since neither yielded enough Tweets to compute an a priori or a post score.

4.2.2 Independent variable – Materiality of corporate response

The independent variable [IV] used in this study is the materiality of a firm’s response to a petition. The variable was coded within the initial data set of the change.org petitions into three categories. Using this approach meant that a petition could remain unanswered and thus unsuccessful [1] or trigger a firm’s response. This reaction could either be of a [2] symbolic or substantive nature [3]. The categorical variable was coded into two separate dummy variables, which both have unanswered petitions level as their reference group. Thus, the first dummy variable is symbolic petitions and the second substantive petitions.

(26)

26 An example of a symbolic response is the reaction of McDonald’s South Africa to a petition published on March 3, 2016. The fast-food giant was asked to commit to the sourcing of cage-free eggs for its products. The petition attracted 18,195 subscribers and inspired the fast-food producer to respond to the demand. The South African subsidiary released a statement in which it promised to transition to cage-free egg sourcing by 2025 (Guse, 2016). This reaction was coded as symbolic, because at the time of the release besides the unbinding written promise no monetary or physical commitment was made. An example for a substantive response can be found in the case of Mr Leo Bretholz, who petitioned the French railroad service SNCF to admit its historic guilt for collaborating with the German Nazi regime. The petition created on January 27, 2014 attracted 165,000 supporters and was labelled successful on December 11, 2014. The French government signed a binding agreement to provide a $60 million compensation fund, which also represented the government’s admission to be held accountable for its complicity in World War II crimes (Shaver, 2016).

4.2.3 Moderating variables – Petition subscribers & Online word-of-mouth

This research incorporated the number of subscribers to a petition as the first moderating variable of the analysis. The website change.org allows every user with a registered email address to back a petition with his or her account. However, accounts do not need an identity verification, and thus can be anonymous.

The second moderating variable is the percentage of Tweets caused by the change.org petition. In other words, this is the proportion of online word-of-mouth (eWoM) of all Twitter content in the analysed period that can be directly related to a petition. Therefore, the Twitter data sets were scanned for keywords that could also be found in the respective title and summary of the request. As an example, Delta Airlines was petitioned to prohibit the transport of hunting trophies. The airline responded to the request by banning trophies from its aircrafts with immediate effect at the time the statement was released (Delta Air Lines, Inc., 2015). During the scraped period, a total of 1,417 tweets were publicly addressed to Delta. Out of theses, 10.66% (n =151), contained the word “troph*” or “hunting”. In this regard, the “*” search-term allows to spot different words that start with troph-, such as trophy or trophies.

4.2.4 Control variables

The analysis incorporates four control variables to monitor the effect of country internet regulation and industry characteristics. The Freedom on the Net Index by Freedom House is used as a first variable. The U.S. elections in 2016 and the analysis of the presidential campaigns’ effectiveness of the candidates have increased public awareness of the threat of social media manipulations. The probability and practicality of digital mass persuasion has

(27)

27 added a new, more complex layer to the fully democratic web, as it is sometimes falsely and romantically transfigured (Matz, Kosinski, Nave, & Stillwell, 2017). Moreover, the nesting of digital mass persuasion strategies makes it significantly more difficult to disclose these as opposed to traditional web-site or IP-address blockades, which can be used by governments to restrict information to domestic users. The second control variable is the degree of individualism of a society in which the petition was published. Therefore, the Individualism-Collectivism dimension from Hofstede’s cultural framework was taken. The variable is numeric and has a score between 0 and 100. Both Freedom on the Net and Individualism control for international differences across the data set. An overview of the distribution of countries where the petitions in the sample set were filed in and their respective scores for the two control variables is displayed in Table 1.

(28)

28 The third variable controls for the year a petition was published. The contemporary debate whether Facebook is just a platform or a media house and thus must regulate the content inside its network, indicates the growing importance of social media (Galloway, 2017). The variable consists of eight categories, which represent the years 2010 until 2017. Although change.org has been in operation since 2007, no petitions published earlier than 2010 were included in the data set. The overall user sentiment of a firm, which was measured throughout the six-day period, was chosen as the fourth control variable. Owing to the methods used to determine the sentiment, the variable is numeric and can have a value between 1 and 0. An overview of the sentiment scores for the analysed firms can be found in Appendix A.

(29)

29

5 Results

5.1 Data cleaning and controlling

To proceed with the statistical analysis, the finalised subsample was imported into IBM SPSS. Initially, the data was checked for missing or falsely entered variables and controlled for correlational matters by running descriptive statistics. Table 2 presents the descriptive statistics of all variables used in the model prior the dummy variable coding. Table 3 displays all variables excluding the IV Materiality of Response, the numerical moderating variable Number

of Subscribers and the control variable Year of the Petition’s Publication as these were

transformed into their respective dummy variables. In terms of missing variables, no abnormalities can be reported except for the variables Freedom on the Net and Degree of

Individualism. In both cases, retrieving data was not possible due to missing information on

Nepal’s and Palestine’s internet freedom and Armenia’s and Palestine’s level of individualism. Therefore, the valid amount of cases had to be reduced to 131 cases in total.

The average change in Twitter user sentiment is 8.26 and has a standard deviation of 33.30. The maximum positive sentiment increase can be reported with 181.25%, while the most extreme decrease is -87.22%. The mean number of petition subscribers in this sample is 84,818.21. Moreover, the petition size levels are non-normally distributed, since very large (n=47; 35.07%) and very small (n=31; 23.13%) petitions are the most and second most common levels as depicted in Table 4. Regarding Twitter user sentiment, a median of 0.43 can be derived from the data set, while on average 2.8% of all Tweets analysed could be related to the petition directly. The respective maximum value of all sentiment scores is 0.73, whereas the minimum is 0.12. Both control variables Freedom on the Net and Degree of Individualism show high skewness to the right and the left respectively. The skewness is caused by the dominance of the United States as the origin country for the petitions in the sample (n=99; 73.88%). The maximum value for internet freedom in the sample is 87, the minimum value is 15 (U.S. =21). The maximum value for Hofstede’s individualism dimension in the sample is 91, the minimum value is 18 (U.S. =91).

The 46 cases of unanswered petitions showed an average user sentiment of 42.80% during the observed period. In contrast. symbolic cases had an average of 41.75%, which is 2.44% lower. The average of the substantive cases is 45.01%. Thus, the difference to absent responses is a surplus of 5.17%. Consequently, symbolic cases have a 7.80% lower average than substantive cases. The discussed frequencies are displayed in Table 5.

(30)

30 Table 2 - Descriptive statistics excluding dummy variables

Table 3 – Descriptive statistics including dummy variables

(31)

31 Table 4 - Distribution of frequencies I

(32)

32

5.2 Correlation and multicollinearity

To control the selected sample for correlation issues, a pairwise correlation check of the variables, which are selected and coded for the linear regression analysis, was conducted. Therefore, the Kendall’s Tau measure was employed. For the given variables, this correlation coefficient is the best fitting alternative as it evaluates the relationship between ranks. In contrast, the Pearson correlation coefficient evaluates the relationship between continuous variables, which is conflicting with the dummy variable setup of this research (Field, 2013). The correlation matrix is displayed in Table 6.

No significant correlations between the dependent variable and any other variables used for the analysis can be reported. In contrast, four significant correlations can be found between the symbolic responses dummy and other variables. The first is between the substantive responses dummy (coefficient -.489; p =.000). Further correlations for this variable are reported for the medium petition dummy (coefficient .351; p =.000), the large petition dummy (coefficient -.277; p =.001) and the very large petition dummy (coefficient -.294; p =.001). The substantive response dummy correlates significantly with the small petition dummy variable (coefficient .192; p =.027), the very large petition dummy variable (coefficient -.202; p =.020),

Freedom on the Net (coefficient -.171; p =.041) and the dummy variables for the years 2013

(coefficient .183; p =.035), 2014 (coefficient .218; p =.012), 2016 (coefficient -.173; p =.046) and 2017 (coefficient -.178; p =.040).

Regarding correlations between the dummies of the moderating variable petition size and the control variables - leaving the publication years aside, there is a significant relationship between the small petition dummy variable and Freedom on the Net (coefficient -.174; p =.037). The second moderator variable Proportion of Tweets correlates with very large petitions (coefficient .155; p =.042). Moreover, a significant correlation between medium sized petitions and the degree of individualism prevails (coefficient -.222; p =.007). Another significant relationship exists between the Freedom on the Net and the degree of individualism (coefficient -.393; p =.000). This suggests that countries with highly individualistic societies generally tend to be less restrictive with internet regulations for their citizens. Furthermore, there are some additional significant correlations among dummy variables. All significant correlations are flagged and can be retrieved from Table 6.

The highest correlation among the analysed variables in this stems from the relationship between Year 2015 Dummy and the Year 2016 Dummy (coefficient -.543; p =.005). The second highest can be found between Year 2012 Dummy and Year 2017 Dummy (coefficient -.515; p =.013) Therefore, two of the coefficient exceeds the thresholds for mediocre correlational

(33)

33 strength (> |.5|) or severe correlational strength (> |.8|). Consequently, the control variables Year

of the Petition had to be excluded from the regression analysis. To control whether one of the

explanatory variables is linearly associated with the dependent variable of the model, a check on multicollinearity was conducted. Therefore, the Variance Inflation Factor [VIF] and the Tolerance stemming from the regression model are checked. The results are displayed in Table 7. No abnormalities and issues with multicollinearity are to be reported among all variables except the dummies for the years 2012, 2014, 2015, 2016 and 2017. All other explanatory variables exceed the threshold for Tolerance (> .20) and have a low VIF score indicating no collinearity (< 5.00). Therefore, the decision to exclude the control variable Year of the Petition receives additional support. All other analysed variables are strongly related to each other, allowing to proceed with the selected model components towards a further regression analysis. The analysis further confirms the results from the correlation analysis detecting significant correlations between the variables that in most cases are not too severe.

(34)

34 Table 6 - Correlation matrix

(35)

35 Table 7 - Multicollinearity