Research: How Risky Behavior Spreads

The challenge in some organizations is not enough risk-taking: Employees are too cautious and not willing to try new things even when they’d be beneficial, on average, for the organization. In other organizations, the problem is excessive risk-taking: Risky behavior spreads through an organization until something goes wrong. From poor financial decision-making to unethical behavior, excessive risk-taking can sink a company.

So how does risk-taking spread through an organization? The extreme uncertainty surrounding Covid-19 provided a unique environment for studying that question. With the onset of the pandemic, people around the world simultaneously questioned which behaviors were appropriate in order to reduce both individual and societal risk of exposure to the virus. That allowed us to test how canonical theories of learning work in tandem to spread risk-taking.

In a study of people’s behavior post-lockdown and pre-vaccinewe document a phenomenon we call “risk creep.” Thisrefers to a growing tolerance of risky behavior, which may potentially be caused by near missesorevents that could have resulted in a negative outcome but, by chance, did not.

Our research clarifies two key channels through which risk-taking spreads: from social learning and from experiential learning, or trial-and-error. Companies need to understand both of them and how they potentially interact to encourage or discourage risky behaviors. The more managers understand what guides employees’ behavior, the better they can predict it.  Ultimately, this can help them anticipate downstream consequences in order to preemptively communicate with employees and calibrate risk more appropriately.

Theories of Risk-Taking

Decades of work on social norms shows that people are often influenced by observing what others do. These observations help people understand which behaviors are common and which are likely to gain them social rewards or punishment. They are often considered sufficient for learning new behaviors. Researchers refer to inferences based on observing others as “social learning.”

As any manager knows, employees respond less to what they are told is appropriate behavior and more to what they see others doing in the workplace. In strong cultures, these two go hand in hand, reinforcing each other. Southwest Airlines, for example, instructs its flight attendants to risk having fun, but new flight attendants really learn how to behave by observing their colleagues going off script with safety announcements or playing practical jokes. In observing others, they learn the appropriate level of risk for trying something new.

But what happens when there are no obvious cues from the social environment? This occurs in situations where culture is weak or during times of intense change, so that there is little to no information to help people decide what is socially acceptable behavior. Here, people likely rely on their own experiential trial-and-error learning. People may “test the waters,” taking a modest amount of risk and then assess the result — an assessment that is guided by emotions more than by rational calculation.

How does this work? If someone takes a risky action one week, do we expect them to do the same thing next week?

The answer lies in how dangerous the result of the risky action feels. Imagine being distracted by a text while driving and accidentally swerving into another lane. Once you catch your breath, you are likely to put the phone down for at least a few minutes. Alternatively, if people engage in risky behavior without serious consequences, they may develop a sense of safety and become less careful with their behavior. Imagine you responded to the text staying squarely in your lane. You may feel a bit more emboldened to continue texting. We call this latter phenomenon “risk creep.”

The academic literature on the psychology of decision making has explored both when people become more risk averse and more risk tolerant (see these 2012 and 2016 papers).  It has also examined how social learning or experiential trial and error could account for these outcomes.  Yet, these are studied separately rather than in the same context. Our study of Covid-19 behavior helps us measure whether risk aversion or risk tolerance wins out, accounting for both possible mechanisms of social and experiential learning.

“Risk creep” during Covid-19

In a five-month longitudinal field study following lockdown and prior to vaccines, we tracked what people did when they left their homes. We collected eight surveys from 304 students who had recently returned to campus and the surrounding neighborhood to take classes remotely. They took a baseline survey and seven weekly follow-up “pulse” surveys, which included a subset of questions from that baseline survey. The seven pulse surveys allowed us to track changes in behavior and perceptions over time. In all surveys, participants reported how many times they left their home to take part in any of six categories of activities.

We categorized the activities into 1) non-discretionary activities, necessary for day-to-day living (going outside the home for food, to run errands, or school activities) and 2) more discretionary activities, which were relatively less essential for day-to-day living, and were foregone by many during lockdown (going outside the home to exercise, gathering with others in small social groups, attending large events). To examine social learning, we asked participants how many people they saw taking part in those same activities the prior week. To examine experiential learning, we measured people’s perceptions of the riskiness of their own behavior the week before.

We found that people’s level of non-discretionary activities (errands for things such as groceries or a drug store, school study groups) was unchanged during the time period. However, people who saw others engaging in discretionary activities outside of the home (exercise, social gatherings, and large events) did more of these same activities the following week, evidence of a creeping tolerance of risk associated with social learning.

Likewise, people who said they took part in riskier public activities one week gradually engaged in more subsequent discretionary activities the following week. Again, people show a creeping tolerance of risk from the inconsequential outcomes of their own experimentation.

The results from our study suggest that even when social learning is strong and influences behavior, it does not crowd out experiential learning. This may especially be the case when social learning is disrupted due to random events (forgetting a mask and thus coming across a new decision one has not yet had to make). Thus, there is always a need for vigilance against excessive risk-taking.

Implications for Companies

The lesson for companies is, in a nutshell: Beware of close calls. If someone does something risky, whether intentionally or not, and things turn out alright, they’ll be more likely to do it again. If someone sets an insecure password and nothing happens, their intuitive brain is “learning” that it’s OK. If someone accidentally overbills a client but no one notices, they’re more likely to do it again. If someone makes a risky trade and it pans out, they’ll take more risk next time.

In effect, “cutting corners” — even if it’s accidental to begin with — will lead to more corners cut going forward. When things work out, we tend to ignore or discount our good luck, and so the behavior or process no longer feels as risky to us.

This pattern is most dangerous when the risk is relatively low, because of how it combines with social learning, as the pandemic perfectly illustrated. Even in 2020, a person who forgot a mask and so went maskless on an errand was still fairly unlikely to get the virus. The “risk creep” effect then leads them to be more likely to go maskless the next time. Then social learning amplifies the effect, as others see the maskless person and incorporate that into what they think of as socially acceptable. A bit of good luck sets off a chain reaction that ends in more risky behavior.