Is Your Company’s Code of Conduct Encouraging … Misconduct?

“These codes of conduct sometimes look like they’ve been copied and pasted,” says Maryam Kouchaki, an associate professor of management and organizations at Kellogg.

But there are some subtle differences in the language used to communicate these similar points. Some codes use warm, communal words such as “we”—for example, “We always put customers first.” Others convey the same points with more formal, detached language: “Employees always put customers first.”

In a recent study, Kouchaki and her colleagues explored whether these subtle differences might influence people’s willingness to comply with a code of conduct. To the researchers’ surprise, they found that communal language seems to be linked to more rule breaking. This may be because people believe they’re unlikely to be punished harshly. “You think, ‘Even if I’m caught, they are going to be kind, tolerant, forgiving,’” Kouchaki says.

That doesn’t necessarily mean that companies should strike all communal language from codes of conduct. But they should emphasize enforcement as well, she says. Firms need to balance making people feel like part of the company while also “making sure you communicate that bad behavior isn’t tolerated here.”

Cookie-Cutter Codes of Conduct

Kouchaki, along with Francesca Gino at Harvard Business School and Yuval Feldman at Bar-Ilan University, became interested in codes of conduct because they are the primary way that companies try to regulate workers’ actions.

“The ultimate goal is to curb unethical behavior,” Kouchaki says. “We want to see how successful these codes of conduct are in fulfilling those values.”

So the team looked for differences across codes. Research assistants read codes of conduct from Fortune 100 companies and recorded factors such as length, format, and section topics. The assistants found that “there were a lot of similarities,” Kouchaki says. They weren’t going to get very far by trying to identify differences in content.

But the research assistants noticed variations in language. About half the codes used personal language such as “we” and “us,” while the other half used impersonal language such as “employees,” “members,” or the company’s name.

The team decided to test whether one type of language more effectively deterred bad behavior. Their initial intuition was that warmer codes with communal words would work better. The communal language would make employees feel more loyal, they reasoned, and less likely to do something unethical that would harm the company’s reputation.

Studying the Effects of Different Codes of Conduct

The researchers recruited 120 people on Amazon Mechanical Turk for an online experiment. Half the group read a communal code of conduct, which contained statements such as “We emphasize integrity, fairness, and respect” and “We value integrity in all aspects of our work.” The other half read the same code, except that “we” was replaced with more formal terms such as “staff.”

Next, participants were given 10 math puzzles to solve. They didn’t have to show their answers, only state whether they had solved the puzzle or not, which let participants claim they had solved it even if they hadn’t.

Five of the puzzles were actually impossible to solve. So if a participant reported solving one of them, the researchers knew that person had cheated.

Contrary to the team’s expectations, people who read the communal code were more likely to cheat. They reported solving an average of 2.83 unsolvable puzzles, while those who read the formal code reported solving 2.06.

The researchers were surprised and wondered if the result was a fluke. So in another experiment, they hired 151 freelancers on Upwork, an online freelancing platform, to perform data entry tasks for a few weeks. One-third of participants read the communal code of conduct and one-third the formal code; the remaining workers did not receive a code.

As part of their assignment, the freelancers were asked to type the text in captchas (images containing letters and numbers). If they finished more than 35 captchas in three minutes, they would earn an extra $2. However, the workers were told that their work could not be checked, so they would self-report the number of captchas completed. In reality, the researchers could check whether this report was accurate.

During the first week, 40% of participants who read the communal code of conduct inflated their performance, compared to only 14% of those in the formal code group and 25% of those who didn’t read a code. The amount of cheating dropped in the second week, but the overall trend remained the same.

And in a similar lab experiment with 134 students, the communal-code group cheated nearly three times as much on a problem-solving task as the formal-code group.

“After two or three studies, we saw that the effect is robust,” Kouchaki says.

Do Certain Codes Imply Forgiveness?

Why would a warmly worded code of conduct encourage bad behavior? For a long time, “we were stuck,” Kouchaki says.

Maybe communal language made people feel more loyal to the group, but that loyalty actually caused shadier behavior. A committed employee might feel she was helping the firm by earning higher profits, even if such gains were obtained unethically. “It’s easier to justify,” Kouchaki says.

Or perhaps warm and communal language made people care more about the group and thus feel more pressure to perform, again motivating them to boost results at all costs.

But when Kouchaki’s team ran experiments to test these hypotheses, the results didn’t bear them out. People who read a communal code of conduct didn’t seem to feel more loyal or experience more performance pressure than those who read a formal code.

So the researchers considered a third explanation. Maybe when the language is communal, “the group is being seen as forgiving and tolerant,” Kouchaki says.

To test that hypothesis, the team asked 95 individuals at a university to pilot-test surveys for a research lab. After reading one of the two codes of conduct, participants rated the lab on a variety of traits. The people who read the communal code rated the research lab higher, on average, on warm traits such as forgiving, kind, and sociable. But the language didn’t affect how participants rated the lab for non-warm traits such as competence or fairness.

In a similar online study, participants also rated how harshly they expected to be punished for making a mistake. The group that read the formal code of conduct rated the punishment severity higher than the communal code group. Thus, it appeared that communal language may have given people the impression that transgressions would be more easily forgiven.

Finally, the team wanted to test whether communal language was linked to acts of corporate misbehavior in the real world. So the researchers examined codes of conduct from 188 S&P 500 manufacturing firms. They also looked up news articles about bad behavior at those companies, such as employees committing fraud or breaking environmental laws, and found 873 such violations from 1990 to 2012. Employees at companies with “we” language were more likely to act unethically.

Setting Clear Ethical Expectations

The study shows “how challenging and complex it is to try to guide people to be good people,” Kouchaki says. “It’s not just the presence of the code of conduct that matters. It’s your culture.”

Does this mean that all companies should write their codes of conduct in formal language? Not necessarily, she says. Previous research by Kouchaki and other coauthors has shown that writing in a more formal, business tone is linked to a greater focus on competitive, business-oriented values, which could itself lead to more immoral behavior.

But if a firm uses warm and communal language, they should also make it clear that violations will be punished—and follow through with those penalties, she says.

“What helps is to be very clear about expectations and enforcing those,” she says. “You don’t want to just look warm. You want to look warm and moral. That’s the best combination.”