Read the latest campus information on coronavirus (COVID-19) here →

New Conflict Lab walks MBA students through the toughest workplace conversations

student in Conflict Lab roleplaying with instructor
Pearly Khare, MBA 23, role plays with his ‘boss’ and course instructor Bree Jenkins, MBA 19 during a class session. Photo: Brittany Hosea-Small.

Pearly Khare, MBA 23, was in a difficult spot. His ‘boss’ was confronting him about taking off early for vacation, leaving his colleagues “in the dust.” “I definitely understand how that impacted the team,” he said, adding that he gave her and his team advance notice. Then he apologized. 

Afterwards, MBA students who had watched the interaction discussed Khare’s apology to Bree Jenkins, MBA 19, who played the role of his boss.

 “If we apologize, and we’re not even sure of what we did or we are not genuinely sorry for what we did, it can be another form of conflict avoidance,” says Jenkins, co-instructor of the new Berkeley Haas MBA pilot course Difficult Conversations: Conflict Lab, where students roleplay tricky situations that are dreaded at work. “We should ask ourselves if it’s just because we want to move past the discomfort.”

From delivering a poor performance review to providing a critical work project assessment to firing an employee, things often got “spicy” during the 10-week session, says co-instructor Francesca LeBaron, MBA 19. But the class isn’t about right or wrong or about debating morality. “It’s about maintaining connection, even when we disagree with the person,” LeBaron said. “What is your objective? Is it to make this person feel heard, to problem solve, or to share your own needs? And how effective were you at achieving that objective?” 

Jenkins and LeBaron kicked off the semester with a speed conflict session (similar to speed dating) where students role-played a back-to-back series of conflicts to get a sense of the discomfort they would experience in the class. The exercise helped students to assess if this style of experiential learning was right for them. 

two students talking during class about having difficult conversations.
The class asks students to address the hardest parts of receiving difficult feedback. Photo: Brittany Hosea Small.

Ten undergraduate UC Berkeley students and a group of Berkeley Haas alumni—ranging from PWC partners to a Google exec to an NYU professor—also joined the class to play roles that would put students in the hot seat. 

In one session, alumna Kelly Deutermann, MBA 17, confronted Mridul Agarwal, MBA 23, about why he wanted to get off a project. Deutermann aggressively questioned Agarwal. “Do you want to be promoted? Do you want to be taken seriously? This is your chance.” When Agarwal explained that “it might not be the best project for me at this time,” Deutermann responded with, “This project needs to happen. Do you just not want to work hard to do it?” In this role play, Agarwal had to balance his own bandwidth and need for support with Deutermann’s demands for project management. 

After the difficult talk, Agarwal took a deep breath, and the two of them laughed and shook their heads. 

Friends coming up with solutions

Jenkins and LeBaron met in their first year at Haas. They were in the same cohort and found they shared a lot in common: They were both Consortium Fellows, student instructors for the Leadership Communications course, and board members for the Haas Center for Equity, Gender & Leadership (EGAL). After graduation, LeBaron went to work as an executive coach and mediator for startups at UC Berkeley’s accelerator SkyDeck; Jenkins runs leadership training courses as a senior leadership development associate at Pixar Animation Studios.

 “I noticed themes and trends with what we were doing at work,” Jenkins said. “There was conflict avoidance and harm from conflict that’s not dealt with effectively. We talked to friends in other organizations and we realized quickly that everyone is dealing with workplace conflict.” 

For example, LeBaron had recently coached startup founder and former Haas classmate Fahed Essa on how to fire someone. “Fahed is brilliant—has three masters degrees and has started three companies,” she said. “If he is still struggling with this, I bet many people are. I want Haasies to have this skill set that balances being compassionate with being honest and clear.”

After discussing the problem, Jenkins and LeBaron did what they were known for doing at Haas: they came up with a solution. With sponsorship from the Center for Social Sector Leadership at Haas (CSSL), they designed a syllabus for a pilot course completely devoted to managing difficult conversations. The class enrolled 32 MBA students, with a waitlist. 

To track their progress throughout the class, students provide one another with feedback, write papers addressing their own conflict styles, and identify conflicts in the media and how they can be improved using lessons from the course framework. “It’s really important that the students find ways to continue to practice this work after the class is complete,” Jenkins said. “They should have a clear understanding of where they are in their conflict journey and what they want to do to continue to grow.”

During their final class, Jenkins and LeBaron took on a role-play with each other. Jenkins played a manager criticizing an employee for botching a critical client presentation. “I expected more of you,” Jenkins said. “I’m hearing that my actions didn’t meet your expectations. Can you tell me more about what that looked like for you?” LeBaron said. After more back and forth, they drilled down to the core issue: Jenkins was frustrated and disappointed because she wanted to appear competent in front of the client. The two decided to review all future presentations together before going to a client.

LeBaron asked the class to consider what Jenkins felt. “I don’t know if I made typos, but in her mind I made those mistakes,” she said. Her objective, she said, was to better understand her boss’ experience and unmet needs. “I can still hold my experience as true for me, while being curious about understanding her experience,” LeBaron said. 

Student gives feedback during conflict lab
Students practice giving and receiving feedback after role-playing a difficult conversation. Photo: Brittany Hosea-Small.

Working past fear through practice

After the 10-week class ended, students who identified themselves as conflict-avoidant at the start of Conflict Lab said they were starting to work past it.

Daryl Pugh, MBA 23, an executive recruiter before he came to Haas, said he’s learning to be “comfortable with discomfort” and was already using what he learned in class to help a friend through the difficulty of laying off employees. “I tried to talk to her through having that conversation and processing other people’s feelings, understanding what was happening and her interpretation of what was happening. We had a couple of sessions.”

What Pugh said he found most surprising over the weeks was understanding how inaccurately he can interpret the actions of others. “We need to focus on not ascribing emotion to people that could be just wrong,” he said. “That’s how we are trained our whole lives, even in social settings, is to interpret other people’s feelings. The only way to know how a person is feeling is to ask. This class taught me how to get others to express their feelings, then I can move past my observations and interpretations to a new level of understanding.”

Mariam Al-Rayes, MBA 23, said the course provided a set of tools that she plans to use at work and beyond. “I wish we’d learned this earlier in life,” she said. “The role playing was so useful—like when alumni talked to us as our managers. It was realistic and we applied what we learned in class first-hand.” 

On the journey to create a new class of conflict-embracing leaders, LeBaron and Jenkins are well on their way—and plan to offer the class again in Fall 2023.


A threat from a common enemy may no longer unite polarized Americans, study suggests

A scene from the U.S. Capitol on Jan. 6, 2021, just before rioters stormed the building. (AP Photo/Jose Luis Magana, File)

During World War II, Americans came together. They ate less meat and planted victory gardens. They lowered thermostats and rationed their gasoline. Republican, Democrat—it mattered little: Against a common enemy, American civilians were willing to sacrifice on behalf of American interests.

That was 80 years ago, when the political climate was less rife with partisan animosity. In 1960, 10% of parents said they would be uncomfortable if their child married someone from the opposing political party. By 2010, that figure was 33%.

“Intuitively it makes sense that common enemies unite people. It’s a folk theory that goes back to ancient Sanskrit writings,” said Douglas Guilbeault, an assistant professor at Berkeley Haas. “Given the state of polarization today, the question is whether we can get Republicans and Democrats working together in the face of a common threat.”

In new research published in the journal Nature Scientific Reports, Guilbeault and six coauthors found the opposite to be true. The series of experiments found that exposing partisans to information about a common enemy instilled in Republicans a deeper distrust of Democrats than they started out with. The same was not true of Democrats in the study.

The experiments

The researchers recruited about 1,700 Republicans and Democrats between October 2019 and January 2020 to take part in a survey. Participants were sorted into three groups, and each read a different article from Reuters: one with a patriotic tilt about Fourth of July celebrations around the U.S.; another—chosen to evoke a “common enemy”—about how Russia, Iran, and China were conspiring against the U.S.; and the third a neutral article on early human drawings discovered in South Africa.

In the second stage of the experiment, participants were told they could earn additional money based on the accuracy of their response to the question: “What percentage of immigrants between 2011 and 2015 were college educated?” After they gave their answers, participants were given an answer supposedly generated by a member of the opposing political party. (In fact, it was generated by a bot programmed to give a “guess” that differed from the participant’s by roughly 50 percentage points.) Participants were then given the chance to revise their guesses.

“The extent to which someone used information from the other party to update their estimate gave us insight into cross-party cooperation,” Guilbeault said.

What they found was that only reading the “common-enemy” article about Russia, Iran, and China moved people’s guesses, and it appeared to increase animosity rather than bringing people closer. Specifically, Republicans who had read the article were less willing to use information provided by Democrats. The effect was stronger among those who described themselves as more conservative.

Real-world threats increase partisanship

The research project took an interesting turn when, on January 3, 2020, United States special forces in Iraq assassinated the influential Iranian general Qassim Suleimani. The news was saturated with the event and fear of war spiked domestically and abroad. That was midway through the researchers’ study and the events provided a natural experiment alongside the survey experiment—a real-time U.S. threat alongside the more abstract threat from a news article.

The researchers found that, after Suleimani’s assassination, Republican participants both identified much more strongly as American and were much less likely to cooperate with Democrats.

Asymmetric polarization

The finding that Republicans responded differently than Democrats may be explained by “asymmetric polarization,” Guilbeault said. A recent Pew Research Center survey found that Republicans were significantly more likely than Democrats to view the other party as un-American and a threat to the nation’s well-being (36% of Republicans versus 27% of Democrats).

The different parties’ views of what it means to be “American” may be what drove the different reactions, the researchers theorize. Even though a “common enemy” prompt might have pushed people to think of themselves as more “American” and moved them closer together, the presentation of an external threat may instead have further inflamed divisions.

“Because Democrats and Republicans appear to have very different definitions of what it means to be an American, then you can actually create more conflict by getting them to identify this way,” Guibeault said. “We find evidence consistent with that backfire effect here.”

(He acknowledged it’s also possible this particular threat resonated more with Republicans than Democrats, and results could vary with a different threat.)


The fact of contemporary polarization has been well documented in academia and the popular press. But what this polarization means for the health of the country remains uncertain. As Guilbeault and his colleagues demonstrated, one key implication is that partisan tensions, when strung high enough, can lead political rivals to see each other more as an external enemy than as a source of mutual strength.

This insight should have us on high alert going into the next election, which is slated to be one of the most polarizing yet, Guilbeault said.

“We saw it with COVID, where there was a common enemy and each party was simply pointing fingers at the other one,” he said. “Intensely polarized societies seem to create this backfire effect where, rather than bringing groups together, exposure to a common enemy makes them more likely to accuse each other of being on the enemy’s side.”


Read the paper:

An online experiment during the 2020 US–Iran crisis shows that exposure to common enemies can increase political polarization

By Eaman Jahani (UC Berkeley), Natalie Gallagher (Princeton University), Friedolin Merhout (University of Copenhagen), Nicolo Cavalli (Bocconi University), Douglas Guilbeault (UC Berkeley, Haas School of Business), Yan Leng (University of Texas at Austin, McCombs School of Business) & Christopher A. Bail (Duke University)

November, 2022

How salary benchmarking by employers affects workers’ pay

Young Asian business woman in blue suit receiving an envelope with a salary offer
Credit: iStock

A wave of pay transparency laws aimed at reducing inequities is giving millions of workers access for the first time to information on what co-workers make and what potential employers will offer.

Yet comparing salary information is nothing new for employers. While U.S. antitrust law prohibits employers from directly sharing salary information with each other, most mid-sized and large companies routinely use aggregated data from third parties to get a read on the going rates. 

The effects of this widespread practice, known as salary benchmarking, have never been systematically studied—until now. Following White House concerns that benchmarking may be used to suppress wages and benefits, a new study offers the first evidence on its impact on workers.

The conclusion: Benchmarking does not have a negative effect on pay for the average employee. While some salaries decrease and others increase after a company uses a benchmarking tool, salaries overall simply move closer to the benchmark. 

If there was a negative effect on salaries, it would be suggestive of anti-competitive effects,” said Associate Professor Ricardo Perez-Truglia, who authored the new National Bureau of Economic Research working paper with Zoe B. Cullen and Shengwu Li of Harvard University. “That’s not what we found. If anything, we see some small salary gains for low-skill occupations.”

The researchers used aggregated data from the nation’s largest payroll processing firm to see how much employers paid new hires in hundreds of job categories before and after they used the payroll firm’s salary benchmarking tool. They found that employers paid new hires much closer to the median wage after searching the market rates for those job titles.

As a result, some new employees earned more and some earned less than they would have otherwise. “For the most part, they sort of cancel each other out,” said Perez-Truglia.

Direct sharing prohibited by law

Although U.S. law prohibits employers from directly sharing information about their employees’ compensation with each other, they can access aggregated salary data from third parties such as management consultants and payroll processors. As part of the study, the authors surveyed human resource professionals who set pay at medium and large companies and found that 87.6% use salary benchmarks, usually from multiple sources.

In July 2021, Biden issued an executive order encouraging the attorney general and Federal Trade Commission to consider revising federal guidance that lets neutral third parties share compensation information with employers—but not employees—without triggering antitrust scrutiny. “This may be used to collaborate to suppress wages and benefits,” the White House said in a fact sheet.

At that time, there was no research on the impact of benchmarking. Perez-Truglia said regulators may be responding to studies and news reports suggesting that industry consolidation is giving employers too much market power when it comes to setting salaries. 

The new study, which began in 2019, looked at starting pay offered to new hires at 586 firms that gained access to the benchmarking tool between January 2017 and March 2020. The online tool is easily searchable by job title and is based on real, aggregated and anonymized payroll records of many millions of employees.

The researchers divided the new hires into two groups: nearly 5,300 new hires where the company had searched the benchmarking tool before hiring, and a “control group” of 40,000 hires in the same companies that had not been searched. They compared pay for the two groups in the five quarters before and five quarters after the company first gained access to the salary data.

As a second control group, they looked at salaries offered to about 157,000 people hired during the same time periods in comparable roles at similar companies that never gained access to the more precise salary tool.

‘Compression toward the benchmark’

They found that on average, both high and low salaries moved closer to the benchmark. Among “searched” positions, the absolute difference between the new hire’s starting salary and the median salary for that position dropped from about 20 percentage points before the firm gained access to the tool to about 15 percentage points after. This drop is “large in magnitude, corresponding to a 25% decline,” the authors wrote. 

To illustrate this “compression toward the benchmark,” suppose the median pay for a bank teller is $40,000. Without that information, one bank might pay $30,000 and another $50,000. “Once they see the benchmark information, they are much more likely to pay new hires around the $40,000 benchmark,” Perez-Truglia said.

The researchers added that “the effects on salary compression coincide precisely with the timing of access to the benchmark: The compression was stable in the quarters before the firm gained access to the tool, dropped sharply in the quarter after the firm gained access, and remained stable at the lower level afterwards.”

For searched positions, compression around the median wage was much stronger among low-skill positions, which generally don’t require more than a high school degree, than high-skill ones. Dispersion around the benchmark dropped by 40% for low-skill jobs versus 14.6% for high-skill ones. 

“This finding is largely consistent with the anecdotal accounts in interviews with compensation managers, according to which low-skill positions are treated as commodities and thus should be paid the market rate,” the authors wrote.

They did a similar analysis looking at average salary levels and found that benchmarking “does not have a negative effect on the average salary. If anything, the effect on the average salary is positive, but small in magnitude and statistically insignificant.”

Specifically, they found the average starting salary for all searched positions was either 0.2% lower or 1.7% higher than the two control groups, respectively.

There were some statistically significant gains for low-skilled employees: Their average pay increased by 5% or 6.7% depending on the control group. The study also found that the small boost in salary for low-skilled employees increased their retention. More precisely, there was an increase of 6.7 percentage points in the probability that those employees were still working at the company one year later. 

“This evidence suggests that firms may be using salary benchmarking to raise some salaries in an effort to improve, among other things, the retention of their employees,” the paper says.

For high-skill positions, starting pay went down by 2.9% or 1.6% depending on the control group, “but those effects must be taken with a grain of salt, because they are statistically insignificant,”  Perez-Truglia said.  

“Typically low-skilled employees are less likely to negotiate than high-skilled employees. They are often made take-it-or-leave-it offers,”  Perez-Truglia said.  Benchmarking could provide “more equality for those workers.”

There is also a growing body of literature on pay transparency, but much of it focuses on giving employees more access to information, usually as a way to reduce gender and racial pay gaps. A new law in California requires all private employers with 15 or more employees to provide pay ranges in all job postings starting Jan. 1. Colorado, Washington, and New York City have passed similar laws.Although employees can’t access the same kind of “super-sophisticated” pay data that employers can use, Perez-Truglia said, they can use online sources such as Glassdoor, LinkedIn, and to get a read on competitive salaries. Companies are also consulting these sites, he added, which generally rely on employees reporting their pay at current or past employers

Study: Stereotypes of middle-aged women as less ‘nice’ can hold them back at work

Even as they achieve more power and capability on the job, middle-aged women can be held back by a perceived lack of “niceness,” new research finds.

A professional woman in a suit sits in a conference room and looks directly at the camera
Credit: iStock/Getty

Berkeley, Calif.—As a popular tenured professor at UC Berkeley’s Haas School of Business, Jennifer Chatman was used to teaching at the top of her game. But as she entered her 40s and gained even more expertise, she noticed something strange: Her student class evaluations started getting worse. 

“If anything, my teaching was getting even better, but students were harder on me,” she says. What’s more, when she spoke to her middle-aged female colleagues, she found that they were experiencing similar declines, while the men around them were not.

“If anything, my teaching was getting even better, but students were harder on me.” —Jennifer Chatman

Chatman went on to win Haas’ top award for teaching excellence and now leads the Berkeley Haas faculty as Associate Dean of Academic Affairs, but her latest research offers empirical support for her experience: An analysis revealed that evaluations of male professors remain consistent over time, while women experience a quick decline from their initial peak in their 30s and hit rock bottom around age 47.

The analysis is one part of a new research paper published this week in the journal Organizational Behavior and Human Decision Processes and co-authored with Professor Laura Kray and others.  The overall conclusion: Both men and women are perceived as more capable or effective as they get older, but only women are seen as less warm as they age—causing them to be judged more harshly.

Chatman notes that  at a time when women are only just beginning to approach parity in business schools and still make up only 6.4% of S&P 500 CEOs, the implications can be deadly to career ambition. “Middle age is a make-or-break time, when people are being groomed and considered for the top jobs,” says Chatman, the Paul J. Cortese Distinguished Professor of Management and co-director of the Berkeley Culture Center. “We have to look beyond the pipeline to see what’s actually happening in terms of the experiences women are having throughout their careers.”

“Middle age is a make-or-break time, when people are being groomed and considered for the top jobs…We have to look beyond the pipeline to see what’s actually happening in terms of the experiences women are having throughout their careers.” —Jennifer Chatman

Perceptions of “warmth” and “agency” are two fundamental measures that social science researchers have shown are critical to judging those around us. “The first thing we notice about someone is whether they are warm or cold,” explains Kray, who is the Ned and Carol Spieker Chair in Leadership and faculty director of the Center for Equity, Gender, and Leadership at Berkeley Haas. “It tells you something about whether they have good or bad intentions towards you. ‘Agency’ addresses the question of how capable we perceive them to be in achieving those intentions.”

Past research has established that, in general, women are stereotyped to be warmer than men, while men are perceived as having greater agency—or being more capable and assertive. This is a legacy of historical divisions in which women were in charge of child-rearing while men hunted or worked. “The stereotypes have outlived their utility,” Chatman said, adding that friction can emerge when women run counter to those stereotypes by achieving a position of greater agency at work. 

Studies have also shown that perceptions of both warmth and agency generally increase with age. However, no scholars have previously looked at both gender and age together to show how perceptions of men and women may differ. In a series of studies, Chatman and Kray set out to do just that, along with Haas doctoral researcher Sonya Mishra; Haas graduate Daron Sharps, PhD 19, now at Pinterest; and Professor Michael North of New York University. 

“Steve” vs. “Sue”

In an initial study, the researchers presented participants with a headshot of a hypothetical supervisor at a tech company—either a man, “Steve Wilson”, or a woman, “Sue Miller.” They were then given identical information about either Steve’s or Sue’s career and asked to rate them on adjectives such as “forceful” or “gentle” in middle age compared to when they were younger.

True to former studies, the participants rated both individuals higher on characteristics of agency as they got older. However, even with identical descriptions and such little information by which to judge, the participants rated Sue lower on characteristics related to warmth as she aged, while Steve’s ratings didn’t change. “It’s just stunning,” Chatman says. “These stereotypes are so hard-wired and deeply entrenched that they come out even when absolutely identical information is provided about a man and a woman.” 

In a second study, the researchers asked nearly 500 professionals in executive leadership classes to ask real-life colleagues to perform an assessment measuring them on attributes including assertiveness and agreeableness. Interestingly, women received the same ratings on warmth regardless of their age; however, middle-aged men in the class were rated higher on warmth than were younger men. 

“In these circumstances, women were not perceived as less warm in an absolute sense, but they’re still being perceived as less warm compared to men,” says Kray. “So anytime they are being considered in juxtaposition to men at that age group, they may be at a disadvantage.”

Evaluations of university professors

In the final study, Chatman and Kray went back to the original source of the research to analyze a large dataset of university professor evaluations, allowing them to literally compare a person’s performance to their younger selves to see how it changed with age. (The researchers did their best to control for factors such as whether professors had children or took on extra non-teaching work as they rose through the ranks.)

Sure enough, they found that male professors’ evaluations remained consistent over time. Meanwhile, evaluations for female professors quickly declined from their initial peak in their 30s, hitting a low point around age 47. After that, they steadily increased again, achieving parity with men by their early 60s. “At that point, there are different stereotypes of women, and they may benefit from being seen as more grandmotherly,” says Kray. 

This is an especially dramatic finding because teaching is a skill that should improve with age and experience, Chatman points out. “It did for men, but for women, they were evaluated as worse teachers compared to their own earlier performance, a finding that strongly implicates stereotypes rather than actual performance” says Chatman.

To drill down further into the reasons behind the decline, the researchers also analyzed the actual comments written by students using linguistic software that identified hundreds of adjectives. They found that words such as “caring,” “nice,” and “helpful,” declined for women along with their scores. “When women were getting their lowest teaching ratings, there was an uptick in complaints about their personality,” Kray says.

All of these results confirm the researchers’ suspicions that even as women gain more power and capability as they gain experience, they are dinged for not fulfilling stereotypical prescriptions for “niceness.” 

“There seems to be something about the very nature of career progression that seems to lead people to perceive women as less warm and therefore less likable as their agency increases,” says Chatman. In part, these findings offer validation to women, explaining why they might experience backlash and stagnation in middle age just as their careers are on the rise.

Awareness is essential

The researchers caution, however, against taking away the idea that women should strive to be warmer or less capable—a struggle every woman who has sat through a meeting with overconfident men already knows. “I would hate for the message to be that women need to be more careful about how they present themselves,” Chatman says, “because these findings already point to the fact that women have a narrower band of acceptable behavior.”

“As women move into positions of evaluating others, they should not be afraid to speak up about double-standards and be change agents from within committees charged with evaluating others’ work.” —Laura Kray

Instead, the researchers hope that the results can help create awareness that bias may affect how women are considered for promotions, differently from how men are considered. “We need to create systems and standardization for how we discuss and evaluate candidates,” Kray says, “and either exclude feedback on personality, or make sure it is considered equally for men.” 

As women do rise through the ranks, she adds, their own firsthand knowledge of these persistent stereotypes can help them educate the men around them to make decisions based on merit and ability, rather than stereotypes about perceived warmth. “As women move into positions of evaluating others,” Kray says, “they should not be afraid to speak up about double-standards and be change agents from within committees charged with evaluating others’ work.” 


The paper:

Agentic but not warm: Age-gender interactions and the consequences of stereotype incongruity perceptions for middle-aged professional women
By Jennifer A.Chatman, Daron Sharps, Sonya Mishra, Laura J. Kray, and Michael S. North
Organizational Behavior and Human Decision Processes, November 2022

Press contact:

Laura Counts, Berkeley Haas media relations:, 510-205-9570

New study undermines the theory that depressed people are just more realistic

A young businessman stands at a window with his hand on his forehead
Credit: Adobe Stock

Are depressed people simply more realistic in judging how much they control their lives, while others view the world through rose-colored lenses, living under the illusion that they have more control than they do?

That’s the general idea behind “depressive realism,” a theory that has held sway in science and popular culture for more than four decades.

The problem is, it’s just not true, new research finds.

“It’s an idea that exerts enough appeal that lots of people seem to believe it, but the evidence just isn’t there to sustain it,” says Professor Don Moore, the Lorraine Tyson Mitchell Chair in Leadership and Communication at UC Berkeley’s Haas School of Business and co-author of the study, in press at the journal Collabra:Psychology. “The good news is you don’t have to be depressed to understand how much control you have.”

Depressive realism

The concept of depressive realism stems from a 1979 study of college students examining whether they could predict how much control they had over whether a light turned green when they pushed a button. The original research concluded that the depressed students were better at identifying when they had no control over the lights, while those who weren’t depressed tended to overestimate their level of control.

Moore and his colleagues set out to try to replicate those findings as part of a broader effort to restore trust in scientific research, much of which is woven into the fabric of the scientific community and wider culture. Researchers are revisiting bedrock studies to shore up the most basic of scientific principles: Can the research—and its conclusions—be replicated?

Why test the theory of depressive realism in particular? Its decades-long infusion into science, culture, and even potential mental health treatment policy makes it important, Moore says. The original study, for instance, was cited more than 2,000 times in subsequent studies or research, according to Google Scholar.

“At the top of the list of reasons why we ought to revisit this particular article is its widespread acceptance in both the scholarly and popular literature,” says Moore, who studies overconfidence, confidence, and decision-making. “That means a lot of people are building theories or policies premised on this effect being true. If it’s not, it’s really important to establish that.”

Replicating the original study

Moore co-authored the study with Amelia Dev, BA 17, who served as lead author and is now a University of Miami doctoral student; UC Berkeley psychology professor Sheri Johnson; and former undergraduate student researcher Karin Garrett, BA 21.

The authors studied two groups of participants, whom they screened for depression via a questionnaire. The first group of 248 participants came from Amazon’s Mechanical Turk, an online service that provides paid survey-takers and study participants from a range of backgrounds, in this case all over 18 years old. The second group was made up of 134 college students who participated in return for college credit.

The researchers added or used more modern and robust measurements for the study. For example, they added a mechanism to measure bias, and experimentally varied the amount of control participants actually had.

Participants performed a task similar to that in the 1979 study. In 40 rounds, each chose whether to press a button, after which a lightbulb or a black box appeared. Each was told to figure out whether pushing (or not pushing) the button impacted whether the light came on. After the rounds, each reported how much control they had over the light.

Both the online groups and college student groups were split into three experimental conditions. Each condition experienced different relationships between the button and the light during the 40 rounds. The participants in the first two conditions had no actual control over the light’s appearance, yet saw it illuminate one-quarter or three-quarters of the time, respectively. Participants in the third condition had some control, seeing the light three-quarters of the time after pushing the button.

The researchers were unable to replicate the original study’s results. In fact, people in the online group with a higher level of depression overestimated their control—a direct contradiction to the original study. That finding may be driven by anxiety rather than depression, the researchers note, an observation Moore says merits further study.

In the college student group, depression levels had little impact on their view of their control, the authors found.

Researchers also tested for overconfidence. Study participants were asked to estimate their scores on an intelligence test. Depression had no impact there, either.

Results undermine the theory

The results, Moore says, undermined his belief in depressive realism.

“The study does not suggest that there are benefits to being depressed, so no one should seek depression as a cure to their cognitive biases,” Moore says.

Imagine, for example, a manager hiring someone who is depressed because they believe—based on the original study—that the person is less likely to be overconfident and will have better judgment. That would be a mistake, Moore says.

While depression may not improve judgment, the issue of how to accurately gauge our level of control in various situations has broader implications throughout life, Moore says.

“We live with a great deal of uncertainty about how much control we have—over our careers, our health, our body weight, our friendships, or our happiness,” says Moore. “What actions can we take that really matter? If we want to make good choices in life, it’s very helpful to know what we control and what we don’t.”

Read the full study:

Sadder ≠ Wiser: Depressive Realism is not Robust to Replication
By Amelia Dev, Don A. Moore, Sheri Johnson, Karin Garrett
Collabra:Psychology (in press)