How Have Corporate Codes of Ethics Responded to an Era of Increased Scrutiny?

Abstract

Over the past decade, corporate scandals have proliferated. These scandals, along with the emergence of the #MeToo movement and Environmental, Social, and Corporate Governance (ESG) mandates, have increased the scrutiny of corporations’ ethics culture. How have companies responded in terms of the language appearing in their public ethics documents? We compare the Code of Ethics in 2008 versus 2019 for a sample of S&P 500 firms. For the vast majority of firms, their Code of Ethics lengthened, with the average 2019 code having 29% more words (about 1760 words) than the 2008 average. The language of the codes has also changed. Words such as bribery, corruption, sustainability, speak up, bullying, slavery, and human rights all saw significantly higher usage in the later period. We review possible reasons for the dramatic changes and suggest what questions remain about the motivations behind them. Whether the changes we observe are primarily intrinsically motivated or simply market responses to public pressures is yet to be determined. What is clear from our findings is that society seems to be entering a new age of increasingly moral—or, at least, moralized—corporate governance.

Introduction

The last decade has been a challenging one in terms of corporate scandals and in matters of corporate responsibility. Some scandals have been environmental (BP’s massive oil spill in 2010), some related to substantial data breaches of customer information (Target in 2013), and some related to the overuse of financial incentives (Wells Fargo in 2016). In addition, a number of CEOs have resigned after the revelation of alleged inappropriate relations with employees or contractors: Intel (Brian Krzanich), CBS (Leslie Moonves), Wynn Resorts (Steve Wynn), McDonald’s (Steve Easterbrook), and Hewlett-Packard (Mark Hurd). Consistent with these high-profile resignations, in 2019, Strategy& found that for the first time in the history of their annual survey more CEOs were dismissed due to ethical issues rather than lagging firm performance or internal board struggles.¹

These scandals and the evolving thinking about corporate social responsibility (CSR) may have prompted a shift away from Milton Friedman’s claim that the only social responsibility of business is to increase its profits (Friedman, 1970). This shift is reflected in the 2019 “Statement on the Purpose of a Corporation” issued by the Business Roundtable and signed by 181 corporate CEOs (Business Roundtable, 2019). While asserting that “the free-market system is the best means of generating good jobs, a strong and sustainable economy, innovation, a healthy environment and economic opportunity for all,” nevertheless the 2019 “Statement,” in its own words, “overturned a 22-year-old policy statement that defined a corporation’s principal purpose as maximizing shareholder return” by “declaring that companies should serve not only their shareholders, but also deliver value to their customers, invest in employees, deal fairly with suppliers and support the communities in which they operate.”

The “Statement” implores companies to “create value for all of their stakeholders,” not just their shareholders. Although doing good for all stakeholders seems an obvious social objective for firms, whether the CEOs’ statement is altruistic or self-serving is still subject to much debate, and some would argue that Friedman’s simple assertion is a “least worst” solution.² Regardless of what the optimal social contract design should be, the CEOs in their declaration made clear that the domain of corporate ethics was expanding and that the systematic contemplation of a firm’s ethical posture was no longer merely a check-the-box regulatory artifact.

The question of how to incentivize firms to include social good in their operations has evolved from the simple label of “socially responsible investing” and coalesced in the past decade into the nonfinancial collection of factors labeled environmental, social, and governance (ESG). We examine ethics code changes over the 2008–2019 sample period to test whether or not managers are attempting to address this broader mandate. In particular, we use changes in corporate codes of ethics as a measurable outcome from this cultural shift. This allows us to ask the following research question: In an era where corporate misdeeds are more likely to become public and ESG concerns are increasing, have Standard & Poor’s 500 Index (S&P) managers changed the language contained in public documents related to business ethics?

A code of ethics allows a company to highlight its standards and aspirations for employee business conduct, as well as its commitment to environmental and other societal issues. As noted by Stevens (1994, p. 64), codes of ethics “are managerial tools for shaping change. They often demand from employees higher standards of behavior than required by law. A code may be part of a personnel policies manual, which many courts interpret as a legal contract between employee and employer, or it may be a separately issued document which stands alone.” Mathews (1987) proposes that codes of ethics allow companies to demonstrate a commitment to social responsibility and to promote an anticriminal and self-regulatory corporate culture.

To examine changes in language contained in public documents related to business ethics, we compare codes of ethics for 347 identical S&P 500 firms in 2008 and 2019 and document a number of notable trends. First, there has been a dramatic increase in average length. In 2008, the average number of words in the code of ethics was 6054, compared to 7821 for the average code in 2019. Thus, the average code has increased in length by 29% (almost 1770 words). Since the ethical landscape has become more complex, the lengthening of the code should perhaps be expected. Interestingly, some firms bucked the trend and dramatically shortened their code (3M cut over 97% of the words in their code).

Second, much of the language usage has changed over the last decade. To assist in documenting changes in language usage, we derive from extant word lists and the codes themselves eight categorical groupings of content words (Environmental, Human Rights, Corruption, Inappropriate Behavior, Trust and Transparency, Cyber, Social, and Governance). Consistent with our assertion of the broadening domain of corporate ethics, ‘Environmental,’ ‘Human Rights,’ and ‘Corruption’ word categories all experienced an increase in word count of over 90% between 2008 and 2019, and the ‘Cyber’ words made their first appearances.

Terms such as social media, transparency, slavery, sustainability, anticorruption, footprint, and trafficking appeared not at all or only infrequently in 2008 compared to significantly higher frequencies in the later period. As an example, travel companies and hotels did not focus attention on modern slavery or human trafficking in their 2008 codes. However, as investors and politicians have paid increasing attention to human trafficking, publicly traded travel companies changed their language. The Expedia Group (an online travel-related company) mentioned neither slavery nor trafficking in their 2008 code. In 2019, however, Expedia used the word slavery once and trafficking five times. Importantly, the recent codes also encouraged more active employee participation in ethical behavior. For example, the bigram (pair of consecutive words) speak up went from a count of 48 in 2008 to 766 in 2019.

Third, we document that some firms in the face of changing legal, harassment, and environmental norms, surprisingly decided not to change their code of ethics at all. For example, companies such as AT&T, New York Times, and Deere have effectively identical codes of ethics across the two periods. Lastly, we find evidence that some of our code of ethics word categories are linked to exogenous measures of firm ethical behavior.

Below, we provide a brief review of the literature on corporate codes of ethics (Sect. 2). We then provide our data and methods of analysis (Sect. 3). Finally, we summarize our findings, discuss some possible explanations for our findings, and indicate a limitation on what our data show (Sect. 4). We suggest that one possible motivation for the changes we document is simple window dressing rather than “principles-based” motivation. Our results related to exogenous measures of firm ethical behavior, however, suggest that this is not entirely the case. Thus, our study provides new intuition regarding the factors that may drive improvements in business ethics.

Literature on Codes of Ethics

Although some studies have used the language contained in annual reports to gauge ethical dimensions of the firm (Audi et al., 2016; Balvers et al., 2016), our study focuses on companies’ codes of ethics. Companies have been concerned about the ethical behavior of their employees for more than 100 years.³ One of the first large firms to have a formal code of business conduct was the retailer J. C. Penney. In 1913, the company’s founder believed in doing business by the standard of the Golden Rule, “Do unto others as you would have them do unto you” and drafted a set of principles for employees to follow.⁴ As mentioned by Stevens (1994), following the political scandal of Watergate and the passage of the Foreign Corrupt Practices Act of 1977 (outlawing the bribing of foreign officials to obtain business), US companies ramped up their usage of codes of ethics, at least in part to improve the public’s perception of them.

Several earlier studies documented the typical focus of the codes. For example, Cressey and Moore (1983, p. 53) find that many of the early business ethics documents gave more attention to “unethical conduct likely to decrease a firm’s profits than to similar conduct that might increase profits.” Using a sample of 281 US firms, Chatov (1980) documents that the five most commonly prohibited employee activities are extortion/kickbacks, conflict of interest, illegal political contributions, violation of laws, and the use of insider information. Sanderson and Varner (1984) reports that the codes for 39 of the top Fortune 100 companies were typically poorly written (i.e., long sentences and numerous relative clauses), reflecting the influences of the firm’s legal department.

Later publications typically analyzed outcomes of the codes. For example, using survey results for a sample of office equipment salespeople, Weeks and Nantel (1992) find that a well-communicated code of ethics might be beneficial for ensuring ethical behavior of the sales personnel. Similarly, McKinney and Moore (2008) document that employees with a written code of ethics are much less likely to find international bribery an acceptable practice. For a sample of four Chinese toy suppliers, Egels-Zandén (2014) finds that codes actually do improve workers’ rights. Just the presence of a code, according to Adams et al. (2001), seems to improve the perception of ethical behavior within an organization. Schwartz (2001), however, interviewed 57 managers, employees, and ethics officers at four major Canadian companies and finds evidence that corporate codes of ethics often do not, after all, affect employee and executive behavior.

Helin and Sandström (2007) argue that one factor accounting for ineffectiveness of corporate codes of ethics results simply from employees’ lack of awareness of the content of the code. Davidson and Stevens (2013), in an experimental setting, use Bicchieri’s (2006) model of social norm activation to predict that a code of ethics will improve both manager behavior and investor confidence to the extent that it activates social norms that control opportunistic behavior. They also predict that a certification requirement will increase that activation.⁵ Consistent with their predictions, they find that a code of ethics improves manager behavior and investor confidence only when the code incorporates a public certification choice by the manager. The certification choice, therefore, made both the manager and the investor more aware of the code.

Forster et al. (2009) examine how similar the codes of ethics are between companies by using a regular expression method to parse the documents into individual sentences. Over 2.5 billion sentence comparisons between the codes were conducted. The authors report that the average number of exact sentence matches in S&P 500 companies’ codes of ethics is 36.8. They also document examples of boilerplate language across the sample and find several cases where the code of ethics for different firms are effectively identical. That is, some companies merely copy verbatim another firm’s code. Using a sample of 66 firms across five different industries, Holder-Webb and Cohen (2012) also find excessive overlap in code language. In terms of trends, Sharbatoghlie et al. (2013) document the increased convergence of US and global codes of ethics between 2006 and 2009.

Finally, we note also that the effects of informal moral norms versus explicit moral codes has long been a subject of philosophical investigation. As Blay et al. (2018) show, Adam Smith’s Theory of Moral Sentiments (1982 [1759]) claimed that society acted as a “mirror” that reflected others’ perceptions of one’s own conduct. Smith argued that our desire for “mutual sympathy of sentiments” led us to adapt our behavior and judgment to the expectations of others, even in the absence of an explicit moral code (Otteson, 2002). Kant (1981 [1785]), who read and was influenced by Smith, argued that it was solely one’s own internal commitment to a universalizable moral rule—a “categorical imperative”—that could give one’s actions moral worth. This raises the question of not just behavior but also motivation and intent when evaluating conduct, and presents a difficulty in assessing the purpose and effectiveness of corporate codes of ethics, a difficulty to which we return in the final section.

Data and Methods

S&P 500 Code of Ethics Data

In response to prior ethical lapses by a number of major corporations, including Enron and WorldCom, the US Government passed the Sarbanes–Oxley Act of 2002 (SOX). Although Section 406 of SOX requires only that public firms disclose whether they have adopted a code of ethics for management, the major US exchanges have strengthened that mandate by requiring codes of business conduct and ethics for all listed companies. Thus, publicly traded US companies have a Code of Business Conduct and Ethics that applies to all their employees.⁶ This exchange requirement for the code simplifies the task of obtaining the codes compared to the mail surveys used in earlier studies to build their sample (Murphy, 2005). Throughout our study, we use the terms “code of ethics,” “code of business conduct and ethics,” and “code” interchangeably.

We start with the sample of S&P 500 firms, an index of large, successful, publicly traded companies, included in the analysis of Forster et al. (2009) obtained from https://sraf.nd.edu/data/ethics. Of their sample of S&P 500 firms from December 2007, 347 firms continued to exist as of December of 2019. Firms not existing in 2019 were typically purchased by other publicly traded firms or went bankrupt. In December of 2019, we obtained codes of business conduct and ethics for the remaining sample of 347 S&P 500 firms. Typically, the codes are posted on the Corporate Governance section of the firm’s Investor Relations website.

Generally, a corporate ethics officer writes codes of ethics, often with the assistance or oversight of the legal or human resources department, or both. Although a firm’s board would bear ultimate responsibility for both the code’s content and enforcement, typically a board does not get directly involved with implementation unless there is a major scandal or one that involves a high-ranking executive. Nevertheless, the codes themselves claim authority over all of the firms’ operations, and any increase in length or content may suggest growing and changing concerns regarding the ethical culture of firms and the ethical behavior of employees.

MSCI’s ESG and Ethisphere Institute’s “Most Ethical Firm” Data

To link codes of ethics language to an outcome, we incorporate ESG ratings and Most Ethical Firm data into our analysis. ESG ratings measure how the firm treats its workers, the natural environment, and society. We obtain the ESG rating from MSCI’s website (https://www.msci.com/esg-ratings). MSCI’s ESG ratings are generated by analyzing 37 key issues (e.g., carbon emissions, labor management, tax transparency, chemical safety, and business ethics).⁷ MSCI grades companies on a AAA to CCC scale. For example, in 2020, 3M, Microsoft, and Kellogg received AAA ratings while General Motors and Wells Fargo had CCC ratings. We convert MSCI’s letter scale into a numeric scale (e.g., AAA is 7; AA is 6; … and CCC is 1). Because MSCI tends to focus their rating on larger firms, our S&P sample has 271 firms with available ratings. Smaller S&P firms such as Macy’s, Mattel, Harley Davidson, and Nordstrom lack ESG ratings.

The Ethisphere Institute (EI) creates an annual list of the world’s most ethical firms.⁸ EI’s rating system contains “more than 200 multiple-choice and text questions that capture a company’s performance in an objective, consistent, and standardized way.” EI places the largest weight on the nominee’s Ethics & Compliance program. In 2020, EI selected 131 global firms for their honoree list. A number of S&P 500 firms in our sample made the 2020 list, including 3M, AT&T, Eli Lilly, IBM, PepsiCo, and GM.

Summary Statistics

The sample summary statistics for our 347 S&P 500 firms are reported in Table 1. The average number of words in the code of ethics was 6054 in 2008, compared to a mean of 7821 in 2019. This increase of over 1760 words represents a 29% jump in the average word count. Within each period, there is wide variation in code length. For example, the 5th percentile in 2019 contained 1554 words while the 95th percentile had 14,036 (almost than 10 times higher). In 2019, the bottom three S&P 500 companies in code of ethics word count are Starbucks (320), Boeing (344), and 3M (444). The three S&P companies with the most words are two financial companies and one health insurance firm: T. Rowe Price (37,100), Moody’s (22,342), and Humana (19,212).

Table 1 Code of Ethics summary statistics for S&P 500 firms with codes in both 2008 and 2019

In contrast to the general trend of increasing length in the sample, 3M, Pitney Bowes, Starbucks, and Eli Lilly all reduced their word counts by more than 90%. The reasons for these extraordinary reductions in this subsample appear to fall into two categories, neither of which suggests a reduced emphasis on ethical conduct by the firms. For example, 3M’s recent code claims that it is impossible for the company to list every possible ethical situation that their employees could face. Instead, 3M has their employees answer three questions when faced with an ethical circumstance. Similarly, Pitney Bowes introduces their shortened 2019 code by stating that “No code or policy can anticipate every situation that may arise,” and then focuses on six guiding principles. Thus, some companies acknowledge the increasing complexity of the ethical landscape and address this by building the code on fundamental principles that are expected to cover both the anticipated and unanticipated ethical issues that might arise. Alternatively, both Starbucks and Eli Lilly shortened their primary ethics code as it has become a core statement of values that references either a more comprehensive discussion of the topic (e.g., Eli Lilly’s Red Book) or an expansive collection of resources (e.g., Starbucks’ “livingourvalues.starbucks.com” website).

Conversely, State Street, Biogen, Snap-on, and Marsh & McLennan all increased their code length by more than 2000%. State Street, a large financial services and bank holding company, went from a code with 299 words in 2008 to one with 14,128 words in 2019 (55 pages in length). Their latest code has separate sections on “When to Speak Up,” “How to Speak Up,” “Gambling at Work,” “Anti-Tying Policy,” and “The Media.”

Also summarized in Table 1 are our exogenous measures of firms’ ethical behavior. For the 271 S&P firms with an ESG rating, Table 1 reports that the average and median firm has an ESG rating of approximately 4 (i.e., a grade of “BBB”). The ESG ratings are available only for years 2016 and 2020. Lastly, we create a Most Ethical Firm Dummy variable. Firms on the 2020 most ethical firm list are assigned a value of 1 for the variable; otherwise, 0. Ten percent of our sample made the EI’s list of most ethical firms in 2020.

Spearman Correlations of the Codes of Ethics

As reported in Table 1, the average Spearman correlations between ethics codes in 2008 and 2019—based on comparative word counts—are relatively low. The mean Spearman correlation is 35.4% while the median firm correlation is 29.7%. This suggests that the majority of firms have changed their code of ethics over time.

Table 2 reports the 10 lowest (Panel A) and highest (Panel B) Spearman correlations between each firm’s 2008 code of ethics and their code in 2019. The table also notes the number of words in each code of ethics. Generally, firms with vastly different code of ethics word counts report lower correlations while firms with effectively identical word counts have high code correlations. For example, Tyson Foods has the lowest code correlation (− 21%) due in part to its significant change in word counts; 1204 in 2008 compared to 2218 in 2019, which was attributable to a total rewrite of their 2008 document. The firm with the second highest inter-period correlation (0.99) is PACCAR Inc., a manufacturer of commercial trucks. Of the few changes occurring in PACCAR's code (most are simply a matter of adding headings), the 2019 version adds, in the section on harassment, “gender identity or expression” to the list of protected groups.

Table 2 Lowest and highest 10 Code of Ethics Spearman correlations between 2008 and 2019 for a sample of 347 S&P 500 Firms

AT&T, one of the 20 largest firms by market capitalization in 2008, changed its code of ethics by only 10 words (correlation of 98%) between 2008 and 2019. In 2019, some of the changes include AT&T adding the phrase “or other corporate codes or policies” and “through AT&T’s Hotline found on AT&T’s website,” and dropping the phrase “that are made in good faith” when describing retaliation against employees who report potential violations. One potential explanation of this is that AT&T also has a Code of Business Conduct, which is not part of their Investor Relations web page and appears to be more of an internally facing document.

Code of Ethics Word Subcategories

As noted in Loughran and McDonald (2016, p. 1208), the natural language processing literature “has long emphasized the importance of developing categorization procedures in the context of the problem being studied (e.g., Berelson, 1952).” Similarly, Loughran and McDonald (2011) show that generic word categories work poorly when applied to the specific content of corporate filings and develop word lists based on their prior information about business terminology.⁹ Simply put, word meanings vary substantially with context. At the same time, self-generated lists seem subjective and susceptible to selection bias.

To our knowledge, there is no comprehensive lexicon of business ethics terms. We attempt to create a balanced solution by using, where possible, existing lists supplemented by our assessment of the 23,421 unique words used in all of the sample codes. From the 23,421 words in our sample documents and the words/categories previously identified in other research, we create 8 groups of words totaling 570 content words. The eight categories are: Environmental, Human Rights, Corruption, Inappropriate Behavior, Trust and Transparency, Cyber, Social, and Governance. Because the primary focus of our study is descriptive and not inferential, we believe the benefits of the corpus-specific lexicons outweigh the potential pitfalls of self-generated lists.

Baier et al. (2020) (BBK) created an ESG list (containing no bigrams) of 482 words from usage in annual reports and proxy statements of the 25 largest market capitalization companies in the S&P 100 Index. BBK break their list of 482 ESG words into three main categories—ESG. For our Environmental words, we used the Environmental category of BBK with nine words/phrases (climate change, conservation, environmentally, footprint, global warming, pollutant, recycled, sustainable, and sustainably) added to their core list.

Working from BBK’s Social word list, we created our Human Rights word list. More than half of our Human Rights words overlap with the BBK Social list. To the original BBK social list, we added seven words (discriminatory, diverse, human rights, slave, slavery, tolerance, and trafficking). For completeness, we also created a separate BBK Social word list containing BBK Social words not included in one of our other subcategories. Examples of BBK Social words that are not on our Human Rights list include charity, childbirth, drug, warranty, and teacher.

We create our Corruption list by extending a subset of words from the BBK Governance list. From the BBK Governance list, we include seven words (bribery, corrupt, corruption, crimes, embezzlement, misconduct, and whistleblower) on our Corruption word list. We extend this to a total of 31 words, although notably some of those are simply inflections of the BBK words. In addition, we created a separate list labeled BBK Governance, which includes words on the BBK Governance list not contained in any of our other subcategories. Examples of BBK Governance words include IRS, tenure, salary, ballot, pension, and grandchildren.

A November 9, 2017 CNN article (“The words we use to describe sexual harassment”) assisted us in creating the Inappropriate Behavior list.¹⁰ The Trust and Transparency word list contains the 21 trust words of Audi et al. (2016) with the five additions of speak up, transparent, transparently, truthful, and truthfully.

Words relating to computers and the internet have more recently entered daily language. Our last category is Cyber words (cyber, cybersecurity, malware, and social media), which captures the changing language contained in codes relating to the internet. None of the Cyber words appeared in our sample’s code of ethics in 2008.

Our analysis also counts nine bigrams (pairs of consecutive words) and one trigram (three consecutive words) in the tabulation of our word subcategories. Specifically, we include “climate change” and “global warming” in the Environmental category; “human rights” in the Human Rights category; “gender harassment,” “quid pro quo,” “sexual assault,” “sexual harassment,” and “unwanted advances” in the Inappropriate Behavior category; “speak up” in the Trust and Transparency category; and “social media” in the Cyber category.

Language Changes in the Code

Table 3 reports language changes in the codes of ethics between the two time periods for a select subset of our ethics subcategories.¹¹ Panel A reports that the aggregate count of Environmental words increased 93% (going from a count of 2264 in 2008 to 4360 words in 2019). A number of Environmental words saw dramatic increases in usage between the two periods. For example, carbon, climate change, footprint, and sustainability all experienced word count increases of at least 1400% from 2008 to 2019. However, not all Environmental words saw spikes in usage. Surprisingly, the bigram global warming had an aggregate count of only one in all the codes in 2019 while the word pollutant never appeared in 2019 compared to three times in 2008. The word environment was not included on our Environmental word list because it is typically used in the context of the “work environment” and not the “natural” environment.

Table 3 Selected word changes between 2008 and 2019

The Human Rights word list, as reported in Panel B, had a 92% increase in usage between the time periods. The three highest words in percentage increases are slavery, human rights, and trafficking. This suggests that the topic of modern slavery and trafficking has gained importance in US codes of business conduct. Also of note, the word transgender occurred 0 times in 2008 compared to 16 times in the later period.

Some countries have a historical culture of bribery and corruption when dealing with local government officials. Two main US trading partners, Mexico and China, historically score poorly on worldwide corruption rankings. As an example, in 2019, Transparency International in their Corruption Perception Index (CPI) ranked Mexico #130 and China #80 out of 198 countries.¹² Increasingly, US companies are educating their employees about what constitutes bribery in the code of conduct. Consistent with this assertion, Panel C of Table 3 reports that the count of our Corruption word list increased 95%. Words such as corruption, anticorruption, bribery, extortion, and whistleblower all saw significantly higher occurrences in the later period.

As noted earlier, inappropriate behavior led to the recent dismissals of a number of prominent CEOs. We would thus expect that the language of US codes of ethics would see a dramatic increase of Inappropriate Behavior words. Yet surprisingly, the count of Inappropriate Behavior words saw only a modest increase of 23% in word usage, less than the 29% increase in average word counts. In fact, words such as groping, pornographic, quid pro quo, and sexual all experienced declines in usage. The count of sexual harassment usage actually fell 30% (326 to 228) between 2008 and 2019.

The count of Trust and Transparency words between 2008 and 2019 increased by 56%. Of all our lists, the Trust and Transparency words had the highest raw word counts in both 2008 and 2019. The words transparently, transparent, trusted, and transparency all experienced large count increases. Although businesses may not have a moral obligation to be transparent or fair, these terms also seem to reflect changing moral concerns. Relatedly, the bigram speak up, referring to an employee’s obligation to stand up for what is right, jumped from 48 occurrences in 2008 to a count of 766 in 2019. Of all the words on the Trust and Transparency list, only virtue experienced a decline in counts between 2008 and 2019. Since the typical firm code focuses on the moral behavior of its employees, words dealing with trust and integrity should and do frequently appear.

The last panel of Table 3 reports the pattern for Cyber words. This list contains four internet-related terms (cyber, cybersecurity, malware, and social media). The aggregate count of Cyber words went from 0 in 2008 to 1504 in 2019, serving as a clear example of the changing domain of ethics codes. Like all languages, English evolves and continuously adds new words or phrases. For example, the bigram “social media” never appeared in any S&P 500 code of ethics in 2008, versus appearing 1331 times during 2019. PepsiCo uses “social media” four separate times in their 2019 code, even having a section on “What are some examples of social media use that violates our policies?”

Table 4 reports the summary statistics for each of the subcategory word lists. The average S&P firm in our sample had 6.52 Environmental words in 2008, compared to 12.56 in the same firm’s code in 2019. As noted in the table, the change in Environmental words is − 6 at the 5th percentile, while the 95th percentile is + 24. Some firms added significantly to their discussion of the environment following BP’s oil spill in 2010. For example, Noble Corporation, an offshore oil drilling company, went from one Environmental word in 2008 to one of the top counts in 2019 (57). In 2019, Noble Corporation even added a section in their code on environmental stewardship.

Table 4 Summary statistics for various subcategory words in both 2008 and 2019

Human Rights words also saw an increase in usage between 2008 and 2019. The average change in Human Rights words was slightly more than nine. As an outlier, Cisco Systems had a relatively low human rights count of nine in 2008, while in 2019 their human rights count was 69 (the highest count). CH Robinson Worldwide, an international shipping and transportation company, saw the largest increase in Corruption words, growing from two in 2008 to 88 in 2019 (ranking #1).

The smallest average percentage increase of the subcategory word lists was for Inappropriate Behavior words. The average count increased only from 15.32 to 18.79. However, a number of firms significantly increased their text devoted to inappropriate behavior by managers. One of the top movers was CBS, which had only two Inappropriate Behavior words in their 2008 code compared to 93 in 2019. As noted in the introduction, Leslie Moonves, the long-time chairman and CBS CEO, was dismissed in December of 2018 following numerous allegations of abuse and sexual harassment.

In the category of Trust and Transparency word counts, Cisco Systems had the largest raw increase. The firm went from a count of 67 Trust and Transparency words in 2008 to 267 in 2019. Conversely, Pitney Bowes had the biggest drop in Trust and Transparency word count (decreasing from 116 words in 2008 to only 7 in 2019). The changes for both companies relate, in part, to corresponding large changes in their overall word count.

Both the BBK Governance and BBK Social categories experienced increased average word counts between 2008 and 2019. CVS Health Corp, for example, added 374 additional BBK Social words from 2008 to 2019. At the other extreme, an outlier is alcohol company Brown Forman, which dropped 650 BBK Governance words as its code of ethics overall word count declined from 21,164 words in 2008 to 4559 words in 2019.

Change in ESG Rating

We have ESG ratings at two points in time, 2016 and 2020, for a sample of 271 S&P firms. Thus, we examine if there is any association between changes in our word category counts and changes to the firm’s ESG rating. Target Corp. had the biggest positive change in ESG rating (a rating of CCC in 2016 compared to a rating of A in 2020). Thus, Target Corp. had a change in ESG of + 4 (going from an ESG score of 1 to 5). Conversely, Boeing, Chubb, Cardinal Health, Marriott International, and Southwest Airlines all had an ESG rating change of − 2. The variation in time series of ESG ratings allows an analysis on what language in the code of ethics is linked with changes in the externally obtained rating.

Table 5 reports the regression results with the Change in ESG Rating as the dependent variable. Since the variable of interest is the change in rating, the majority of the independent variables also focus on changes over time. Our independent variables are Change in Environmental Words, Change in Human Rights Words, Change in Corruption Words, Change in Inappropriate Behavior Words, Change in Trust and Transparency, Change in Cyber Words, Change in BBK Governance Words, Change in BBK Social Words, Change in Market Value of Equity, Spearman Code Correlation, and the natural logarithm of the number of code of ethics words in 2019. For the sample, the firm with largest increase in market value of equity (stock price multiplied by number of shares outstanding) during our sample period is Amazon. In the table, the t-statistics are in parenthesis.

Table 5 Regression with change in ESG ratings as the dependent variable

In the regression, only one of our variables is statistically significant at conventional levels in explaining the change in the ESG rating. We find that the Change in Inappropriate Behavior words is significantly associated with fluctuations in the ESG rating (at the 5% level). The positive coefficient on the Changes in Inappropriate Behavior variable means that higher counts of inappropriate behavior words are associated with significant improvements in the firm’s ESG rating. It appears that MSCI views firms more positively that lay out exactly what constitutes inappropriate behavior by managers/employees. Given that CEO inappropriate behavior is a leading cause of employment termination, this seems a reasonable focus for the rating agency.

Most Ethical Firm Dummy

As our last empirical test, we examine the relation between being classified as one of the world’s most ethical firms and levels of our selected word categories. In the Table 6 logit regression, the dependent variable is the Most Ethical Firm Dummy (set to 1 if the firm is on the 2020 EI annual list of the world’s most ethical firms, otherwise 0). The percentage of Environmental, Human Rights, Corruption, Inappropriate Behavior, Trust and Transparency, Cyber, BBK Governance, and BBK Social words in the 2019 code of ethics are independent variables. Also included as independent variables are the natural logarithm of market value of equity in 2019 and the Spearman correlation between the 2008 and 2019 codes.

Table 6 Logit regression with most ethical firm dummy as the dependent variable

In the logit regression, the coefficient on Trust and Transparency has a value of 60.92 and is statistically significant at the 5% level (z-statistic of 2.28). Of the 10 independent variables, it is the only variable with a significant coefficient. The positive coefficient implies that higher frequencies of Trust and Transparency words in the firm’s 2019 code of ethics are associated with a higher probability of being identified as a most ethical firm according to the EI. As noted earlier, EI places the highest weight in their rating on the nominee’s Ethics and Compliance program. Of all the language contained in the code of ethics, it appears that the EI rating is most affected by trust words such as ethics, ethical, integrity, trust, and transparency.

Because successful ethical firms have assumed integrity and trust values into their organizations through their codes of ethics, the linkage between higher relative counts of Trust and Transparency words in their code and being externally classified as a most ethical firm is expected. This relation is consistent with the intuition that these changes in firms’ codes of ethics are being implemented and are playing an important role in building corporate culture. Transparency encourages accountability in firms, and the fact that this category also includes other related moral behavior terms (Panel E) further substantiates the intuition. As PepsiCo (a 2020 EI most ethical firm) notes, “Our Code is at the center of everything we do. It reinforces our core values, and is the foundation of our strategic vision.”

Discussion and Conclusions

Our analysis demonstrates that corporate codes of ethics have seen dramatic changes in the period from 2008 to 2019. The average length of ethics codes increased from 6054 words in 2008 to 7821 words in 2019, an average increase of 29%. Although a handful of firms defied the trend and shortened their codes, most increased by a relatively large amount. Terms such as social media, slavery, sustainability, footprint, and trafficking appeared not at all or only infrequently in 2008 but appeared with significantly higher frequencies in 2019. Consistent with the intuition that codes of ethics play an important role in building and maintaining corporate culture, we document a positive linkage between Trust and Transparency words (e.g., ethics, respect, and trust) and being selected as a most ethical firm by the EI. We also find that changes in Inappropriate Behavior words are positively linked with changes in ESG ratings.

There are several possible explanations for the growing length of corporate codes of ethics. They could be responses to: increasing immorality or corruption among corporate executives or other employees; a growing sense that consumers, customers, or investors care about moral values; growing worries about legal liability or negative reputational effects from bad public relations stories; or a combination of these concerns. However, all of these possibilities might also ultimately reflect a standard corporate motivation to increase profits. If any of these matters could jeopardize corporate profitability, then it would be simply good business practice to attend to them.

A different explanation would be a growing realization of proper moral principles, or an increasing concern among board members and executives of implementing proper moral principles. A test of this latter possibility—what we might call a “principles-based” motivation—would be whether firms adopt and implement them even if they lead to net losses.

A longstanding moral tradition holds that genuine moral principles involve sacrifice; if one gets rewarded for, or benefits from, one’s moral principles, or if they come at no cost to one, then one accordingly gets little moral credit for them. As Kant (1981 [1785]) argued, true moral virtue is displayed when we follow a moral rule simply because it is the moral rule, regardless of the consequences. The Kantian argument raises this question: Would firms increase, or even maintain, their commitment to ethical principles, and to their codes of ethics, if doing so led to negative financial consequences—to, for example, demonstrable net loss of profitability?

Many defenses of CSR include the claim that engaging in CSR will ultimately pay off in greater profitability, due to positive public relations, positive reputational effects, or increased likelihood of new and returning customers, all leading to increased sales (see El Ghoul et al., 2011; Attig et al., 2013; Malik, 2015; Stoian & Gilman, 2017). Their arguments tend to be: adopting CSR policies is the right thing to do, and doing so will benefit the firm in the long run.

Yet, the addition of the latter claim raises the possibility that it is the motivation driving the change, while the former is a welcome coincidence. If so, then the increased attention to corporate codes of ethics we have observed is susceptible to being described as virtue signaling, moral grandstanding (Tosi & Warmke, 2020), or following and responding to changes in cultural mores, rather than a reflection of sincere or principled commitment. The wide variation we found in the length of codes might also suggest that companies are experimenting with ways to match or reflect rapidly changing cultural mores.

It is perhaps not an accident that corporate codes of ethics have expanded during a time when moral values have enjoyed increased emphasis among the public, and have expanded as the public has increased its express concern for moral values. This suspicion is reinforced by the fact that the particular directions the expanded codes of ethics have taken—for example, sustainability, which saw a 2275% increase in usage—coincide with increasing concerns among the public about the same issues.

Corporations have dedicated considerable time and resources to revising and expanding their codes of ethics. These expansions must, then, be responding to and reflecting a significant concern, and this concern has strengthened substantially during the period we review. Whether this concern is a result of companies attempting to compete and succeed in markets increasingly characterized by newly changing consumer sentiment or a result of sincere, and new, commitment to moral principle is not obvious from our findings.

To give full moral credit to companies for newly enhanced moral commitments, we would need to know what companies’ true motivations are for revising and expanding their codes of ethics. We would also need to know what actual changes in corporate behavior the new codes entail: Are they merely attempts to satisfy changing consumer perception and sentiment, or potentially changing legal liability or regulatory expectations? Or, do they reflect actual behavioral change in the companies?

The linkages we found between Trust and Transparency words and most ethical firm designations, and between Inappropriate Behavior words and changes in ESG ratings, suggest that changing firms’ codes of ethics in these ways might improve their ethical cultures. Moreover, the test suggested above—whether companies would adhere to their new codes if doing so led to net losses—might constitute one way to discern their motivating concerns and intentions. Addressing those questions is the subject of future research, though looking at written corporate codes of ethics will not by itself be dispositive. What is clear from our findings here, however, is that we seem to be entering a new age of increasingly moral—or, at least, moralized—corporate governance.