Stakeholders versus Firm Communication in Social Media: The Case of Twitter and Corporate Social Responsibility Information

Abstract

Building on legitimacy theory and prior work on stakeholder management, we study firm Corporate Social Responsibility (CSR) communication in social media. In particular, we analyze the content of over a million microblogs on Twitter relating to CSR in the banking industry. We focus on key issues considered by banks in their CSR reports, which we classify into Core or Supplementary depending on their connection with core business activities. We find that the use of Twitter to communicate CSR information in social media suggests that significant differences exist between the information interests of companies and stakeholders. Outside stakeholders focus on Core CSR issues, whilst firm insiders are relatively more likely to communicate Supplementary CSR issues. Firm insiders’ information dissemination appears biased towards favorable information, and consistent with a legitimacy-based use of social media. Event studies conducted on dates with significant exogenous CSR news confirm the findings of ‘parallel’ talking, and no resemblance in the CSR issues communicated by firms and stakeholders in social media.

Keywords:

• Corporate Social Responsibility (CSR)
• Stakeholder Management
• Social Media
• Banking Industry

JEL classifications:

• G21
• L86
• M14
• M21
• M49

Acknowledgements

We acknowledge helpful comments and suggestions from Hervé Stolowy (editor) and two anonymous reviewers, as well as from Leandro Canibano, Helena Isidro, Giovanna Michelon, and seminar participants at the Universitá di Padova, University of Exeter, Cass Business School, Universidad Carlos III de Madrid, University of Bristol, IE University, University of Bangor, UC3M-BS Institute of Financial Big Data, the 2017 EAA Annual Conference and the 2017 AAA Annual Conference.

Supplemental Data and Research Materials

Supplemental data for this article can be accessed on the Taylor & Francis website, doi:10.1080/09638180.2019.1708428

•  Appendix A – Criteria description. Examples of keywords, hashtags and users

•  Appendix B – Data gathering procedure

•  Appendix C – Core versus Supplementary CSR (baseline model) and DiffTalk results

•  Appendix D – CSR news in the banking industry

•  Appendix E – Illustrative case study ‘Kutxabank: Opportunistic disclosure and silence’

ORCID

Pablo Gómez-Carrasco http://orcid.org/0000-0001-6481-3340

Encarna Guillamón-Saorín http://orcid.org/0000-0001-5023-9579

Beatriz García Osma http://orcid.org/0000-0002-6600-8342

Notes

1 Social media enhance the flow of reviews, complaints, recommendations and comments addressed to inform about a product, brand or company, and even to influence the behaviour or attitudes of users and consumers (Jeacle & Carter, 2011; Jansen et al., 2009).

2 Prior research analyses communication with stakeholders primarily through the study of firms’ websites (Gomez & Chalmeta, 2011; Capriotti & Moreno, 2007; Unerman & Bennett, 2004).

3 Similar to the work of Lee et al. (2015), who analyse whether firms use social media to react to negative events related to their products, we delve into social media use in reaction to crises, focusing on reputational ones.

4 We follow, in applying these labels of ‘Core’ and ‘Supplementary’ CSR, the work of Gomez-Carrasco et al. (2016), which links Carroll’s (1991) ‘Pyramid of CSR’ theory with the strategic versus non-strategic CSR debate (e.g., Burke & Logsdon, 1996). Using these labels helps to simplify our discussion.

5 According to the Observatory of Ethical and Solidarity Finance Association, between 2008 and 2012, the assets held in these entities multiplied by 4.5, reaching almost one billion euro. The disappearance of savings banks and their social welfare projects left a gap in the market that other types of banking institutions seek to occupy.

6 These are the stakeholders to whom to pay attention. In practice, however, managers may not pay attention to them, as they ‘may not perceive the stakeholder field correctly’ (Mitchell et al., 1997, p. 871).

7 Stakeholder management involves managing expectations (Mitchell et al., 1997), leading to a communication in which both ends share information, learn, and revise their expectations (Manetti, 2011), ultimately leading to the creation of a network of mutual responsibility (Manetti & Bellucci, 2016; Unerman & Bennett, 2004).

8 For example, Friedman and Miles (2006) argues that, during a crisis event, public and media perceptions are driven by emotions enhanced by the feeling of risks, loss, tragedy and corporate deception.

9 For example, the petition website Change.org has more than 100 million unique users in 196 countries. This is a platform to start or support a petition, creating social pressure. An example of their power is the petition of a customer of Bank of America, requesting the removal of a $5 monthly fee in debit cards. This petition had over 300,000 supporters in a month, likely influencing the company to eliminate it (‘Tell Bank of America: No $5 Debit Card Fees’ http://www.change.org/petitions/tell-bank-of-america-no-5-debit-card-fees, accessed October 2019).

10 As an example, Twitter permits posting URL links or direct content by uploading documents, such as letters in reply to stakeholders. See, for example, a reply of The White House (@WhiteHouse) sent out in response to concerns about Zika of an individual stakeholder (https://twitter.com/WhiteHouse/status/769930430917316608, accessed October 2019).

11 We consider as firm insiders the official Twitter accounts that companies use. We also add the personal accounts of managers and other staff members who declare their affiliation with the firms on their Twitter bio. Since this information is public and the affiliation is publicly declared, other users can easily identify managers and staff as corporate accounts. In Online Appendix A section ‘2) Criteria description: Twitter users,’ we describe all the types of Twitter accounts identified. The firm insiders are corporate accounts, managers and employees. Outside stakeholders are all others: public administrations, other companies, mass media, NGOs, civic associations, unions and a generic category of individual users.

12 A microblog it is a type of blog in which users can post small pieces of digital content like pictures, video or audio on the Internet. It differs from a blog in its smaller content. The most popular one is Twitter, and in our paper a microblog is a tweet. Microblogging is popular among users because of its portability and immediacy.

13 Principle 7: ‘Managers should acknowledge the potential conflicts between (a) their own role as corporate stakeholders; and (b) their legal and moral responsibilities for the interests of stakeholders, and should address such conflicts through open communication, appropriate reporting, and incentive systems, and, where necessary, third party review’ (Clarkson Center for Business Ethics, 2002, p. 260).

14 Since its inception in 2006, Twitter limited the length of posts to 140 characters, which applies for the sample period of our analysis. In November 2017, after testing one month for a small group of users, the company extended this limit to 280 characters.

15 Spain is amongst the countries with a higher Twitter penetration rate (Hawelka et al., 2014). About 5.3 million Spaniards used Twitter in 2012 according to Nielsen (2012). Also, according to Alexa.com, in 2014 Spain was the fifth country in the world by number of visits to Twitter with 4.1% of the total, only exceeded by US (28.5%), India (5.8%), UK (4.9%) and China (4.9%), all with significantly higher populations.

16 As acknowledged by Linde (2017, p. 13), the first (financial) crisis was not the one with the worst impact; it was the second (euro zone) crisis that plunged the economy into a deep recession, as he noted that ‘Banco de España estimated that the 2009 recession would have what the economists call a “V” shape, and not a “W” shape, without anticipating the strong impact of the second recession on many credit institutions’ solvency.’ See, also, the work of Otero-Iglesias, Royo, and Steinberg (2016) for details. Moreover, the most critical evidence of banks wrongdoing, i.e., court sentences in favour of bank customers started to be ratified by the Spanish Supreme Court in 2013.

17 Financial data are obtained from Bankscope. We consider the 2012 year-end figures, the latest available for all entities. We cover all the categories of credit institutions recognized by the Bank of Spain: commercial, savings, cooperative, and foreign banks branches. Due to the restructuring in the sector, savings banks are the former savings banks reconverted into commercial banks. We also include ‘ethical banking’ as an additional category.

18 The software is a self-constructed tool that assists in the manual classification of the relevant tweets as explained in Online Appendix B.

19 We follow Friedman and Miles (2006) approach, whereby stakeholders’ relations are considered from the viewpoint of the stakeholder (and not of the firm). We also consider intermediaries in the relationship between traditional stakeholders and firms. E.g. the media. This approach is appropriate given our focus on Twitter as there will be stakeholders who, although important, will not be considered in the current study either because they are not present on Twitter or because they cannot be identified in that specific role (i.e. shareholders).

20 Bankia’s case provides anecdotal evidence in support of the hypothesis that, in a controversial industry, the information disseminated through social media has a noticeable negative bias, enhancing reputational problems for these companies. Our main results are consistent if we exclude Bankia from the analyses.

21 There was no activity on Twitter for certain companies some days. Therefore, these observations were not included in the sample. This reinforces our finding on the importance of CSR communication on Twitter, with 74.66% of days being classified as having CSR communication. A data-point is therefore the number of tweets that are tweeted on a specific day about a specific bank. It is the bank-day number of tweets.

22 Outliers are indicative of 1) measurement error; or 2) variability in the distribution. We do not winsorize the data, as we collect our data manually and extreme values are not suspect of being spurious data points due to measurement error. Winsorizing would eliminate the data we are interested in, as we focus on those days of maximum activity and reactivity in Twitter, when stakeholders at large communicate CSR events and information. We make a number of transformations of our dependent variables (using natural logarithm of core and supplementary CSR, using the ratio of core and supplementary CSR tweets to total tweets, standardizing the number of tweets, using decile ranks) to ensure that our results are not driven by atypical points in the distribution. See Dyckman and Zeff (2019) for a detailed discussions of the pitfalls in windsorizing accounting data.

23 Untabulated tests show no differences in the way the three different types of internal accounts, – corporate, managers and employees, behave on Twitter. Managers and employees often just retweet the corporate account which led us to simplify the coding by classifying all these tweets under the category ‘corporate.’

24 Savings banks were created in Spain in the mid-nineteenth century with the main purpose of improving the living conditions of poorer urban classes through affordable credit conditions and promoting savings, to protect these people from usury networks.

25 For completeness, we also construct DiffTalk, the difference between TalkCore and TalkSupp, a proxy for the relative preference of stakeholders to mention Core versus Supplementary CSR issues. The higher the value of DiffTalk, the more focused the communication is on Core CSR issues. For example, if there is large following in Twitter for a particular Bank, it may rank high (score 10 or 9) in both TalkCore and TalkSupp, this would mean DiffTalk will be small. In contrast, if a Bank ranks high in Core CSR discussion and low on Supplementary CSR, DiffTalk will be positive and significant. The results obtained for DiffTalk are reported in Online Appendix C.

26 In most banks the sum of firm insiders and outside stakeholders is the total number of stakeholders. However, in some banks, there are also publicity/sponsoring stakeholders. We control for the presence of these stakeholders in our analyses. In our analyses, they are classified as ‘Twitter accounts aimed at the generation of purely promotional or advertising content, which sometimes could be even classified as spam’ (see, Online Appendix A, section 2.C). On average, these accounts represent 6.7% of users.

27 At the time of our analyses, financial information disclosures in the banking industry were regulated by IFRS/Spanish GAAP and Circular 4/2004 of the Bank of Spain. This regulation did not include any specific rules with respects to CSR disclosure or communication on social media. Non-financial information disclosure was not regulated in Spain until 2017, when the new Circular 4/2017 overruled the old one.

28 Our sample consists of commercial banks, entities linked to former savings banks, cooperative banks and entities classified as ‘ethical banks.’ Amongst them, the savings banks have the greatest social impact and are highly controversial, as many of them had to be bailed out by the government and were involved in the fraudulent sale of complex financial products that caused significant losses to savers.

29 We also study the association between Core and Supplementary CSR communication (as measured by the natural logarithm of the total number of Core CSR tweets (Core_CSR) and of Supplementary CSR tweets (Supplementary_CSR), respectively) and the presence of firm insiders and outside stakeholders. The results are consistent with those reported in Panel B of Table 4.

30 Twitter is characterized by rapid communication. Indeed, most communication takes place on the event date and the following day. As an example, an analysis of Twitter trends reveals that top 20 trends start on average 60 days before (median of 25 days). Perhaps more importantly, even these top trends only stay as such for an average of 35 min (median of 30 min). See: http://www.vuelio.com/uk/blog/how-many-tweets-make-a-trend/, accessed October 2019.

31 Although our events are selected to be exogenous to the users, potentially, mass media such as journalists have some say over the timing of publishing the news. Thus, we provide evidence separately for outside stakeholders excluding media outlets. Our results do not vary if we do not eliminate media-classified users.

32 Event_S, Event_I and Event_A are time dummies, they take the value of 1 on the day when there is an event, for all banks; 0 otherwise. These dummies capture potentially confounding effects associated with dates with great activity, if there is contagion across the industry, and for example, indignation with one bank spills over to other banks. Specific_Event is a bank-time variable. It takes the value of 1 for the bank and date affected; 0 otherwise.

33 See original release at: https://www.sec.gov/news/press-release/2018-219, and information on settlement with SEC at https://www.sec.gov/news/press-release/2018-226. Accessed October 2019.

34 Under Spanish legislation, lenders must pay back the mortgage in full: the bank may seize the mortgaged house, and with real estate prices going down during the crisis, this was usually not enough to cover the debt. This meant that families were often left without a house and with a sizable debt still to be repaid.

35 Bankia implemented aggressive strategies to raise capital, selling complex financial products to retail investors without providing enough information about their risks. The so-called preferential shares were sold particularly to the elderly, who believed they were buying safe, high-yield deposits. In fact, they were high-risk illiquid bonds not covered by the national depositors’ compensation scheme.

Additional information

Funding

The authors gratefully acknowledge funding from the Ministerio de Ciencia, Innovación y Universidades [ECO2016-77579], Comunidad de Madrid (H2015/HUM-3353), FEDER UNC315-EE-3636, the Catedra UAM-Auditores Madrid, and from UC3M-Twittiment.