Bozen – Advances in textual analysis allow researchers to extract useful financial information from the financial statements and the text of the accompanying reports. Can we “read between the lines” of the financial statements? Does textual information provide investors with additional measures to predict bankruptcies or fraud? Can computers help financial analysts read news articles or extract subtle cues from the earnings conference calls to better understand a firm’s financial health? Or, perhaps, the text of risk disclosures reveals the unintended messages of the management, such as their position on ethical issues or attitude towards corruption? These are all interesting questions potentially answered by the technology of textual analysis.
Language correlates with firm’s performance, future stock and fraudulent activities
The exponential increase in computing power in the late 1990s and early 2000s allowed researchers to use advanced methods to actively examine the impact of qualitative information on a firm’s accounting and financial metrics. The online availability of annual reports, conference call transcripts, and texts from social media provide ample material to apply the technology. The words used by managers to describe their operations and financial statements, and the language used by media to report on firms and markets are correlated with firm’s performance, future stock, and even fraudulent activities of management.
Textual analysis has become particularly popular in accounting and financial literature after a seminal paper by Loughran and McDonald 2011. Published in the high-ranking Journal of Finance, the authors interpreted the words commonly used in the annual reports of public companies to create positive and negative sentiment lists. For example, words achieve, attain, efficient, improve, profitable, or upturn are considered positive, while untrue, omit, weaknesses, and fraud are negative. The authors suggested that understanding the sentiment of managerial communication is important to interpret the financial position of the firms.
Another textual methodology to identify common themes in documents is called topic modeling. Using rigorous mathematical and statistical approaches, researchers can reduce millions of words to a handful of easily interpretable topics. This technique allows, for example, to differentiate paragraphs of an annual report discussing climate risk challenges from those talking about executive compensation. One of the widely-used topic modeling algorithms is based on a seminal research paper by Blei et al., 2003. In our paper, we use Blei et al.‘s approach to study companies’ corporate Codes of Conduct (CoC).
Companies’ corporate Codes of Conduct
A Code of Conduct is a company-specific document that describes employee responsibilities and guides them in making everyday work-related decisions, for example, approaching government officials or handling a work conflict. Corporate codes cover a broad range of financial and non-financial risks that can disrupt or threaten an organization‘s operations. Legal Research Network found that companies consider the following topics as top five in their codes: conflicts of interest, data privacy, cybersecurity, hostile work environment, and bribery and corruption.
The codes have become near-ubiquitous in the business world as literally every public firm is requested to maintain a CoC for its employees and senior officers. Investors place increasing importance on the corporate codes because they may influence decisions, behaviors, and financial outcomes. Firms increasingly emphasize the importance of the Codes and personal responsibility for violating them: The CEOs of Boeing (2005), Hewlett- Packard (2010), Best Buy (2012), Priceline (2016), Intel (2018), and McDonald‘s (2019) have left the firms for various violations of ethical standards (USA Today, 2018).
Study: Textual analysis of Codes of Conduct
In our research we focus specifically on corporate Codes of Conduct because they feature many topics and less legally scrutinized language (as opposed to, for example, the annual reports). It is nevertheless challenging to assess if the code reinforces the true commitments and risks or simply recites them as an attempt to reach minimal compliance with regulatory requirements, which researchers call “impression management”.
One of the topics typically included in the codes is anti-corruption. Corporations are increasingly required – and expected – to commit to having a corporate culture on anti-corruption and to publicly disclose their position and their efforts on such topics. But corruption remains an increased concern for regulators and investors. In our study we answer the following research question: is there a strong corporate communication on corruption, or does corporate “tone from the top” serve only an “impression management” role to appear compliant? Moreover, is this corporate communication consistent with the corruption risk profile?
We use a sample of Standard & Poor 500 companies that are required to maintain a corporate CoC and consider topics of CoC as evidence of anti-corruption language. We use the firm-adjusted World Bank Control of Corruption Index to measure firm-level corruption risk. We find a positive association between the anti-corruption topics and the firm-adjusted level of corruption risk. That is, the language used varies depending upon the corruption risk faced by the firms. This result suggests that CoC is not simply used for impression management: the firms are aware of the corruption risks and reflect them in their anti-corruption language.
Prior research also notes that corporate communication does not work in isolation in creating risk awareness but rather in conjugation with other risk management practices, such as external auditor procedures. To investigate this, we also examine if there is any interaction between the two main risk control instruments: anti-corruption communication and audit. We find that when corruption risk is high, external audit oversight becomes an essential factor in corporate anti-corruption communication. Our results confirm that one can indeed “read between the lines.” Corruption risk disclosures are informative to investors and textual analysis can provide insights into a firm’s risk management practices. Our study may open new avenues for auditors who may care about the information content of CoC when evaluating a firm‘s risk practices. However, communicating the rules and risks has its limitations. Anti-corruption communication needs to be supplemented by other control activities such as audit oversight, suggesting that a mix of traditional audit-based controls and appropriate narrative is required to achieve desired behaviours.
Olga Bogachek, Francesco Grossetti and Miles Gietzmann
THE AUTHORS Bogachek is an Assistant Professor of Accounting of the Faculty of Economics and Management at the Free University of Bozen-Bolzano. Her research focuses on textual analysis of corporate disclosure, investors’ short-termism, auditing and blockchain.
Grossetti is an Assistant Professor at the Department of Accounting of Bocconi University, Milan and a fellow at the Bocconi Institute of Data Science and Analytics (BIDSA).
Gietzmann is a Professor at the Department of Accounting of Bocconi University, Milan.
Info
Glossary
ample: reichlich, genügend Codes of Conduct: Verhaltenskodex, Verhaltensregeln‚ bribery: Bestechung ubiquitous: allgegenwärtig, omnipräsent impression management: Impression Management oder Selbstdarstellung sind bewusste oder unbewusste Inszenierungsstrategien zur Herstellung eines bestimmten Ansehens in der öffentlichen Meinung external auditor: externer Prüfer