Forget Myers-Briggs: To Build a Great Team, Focus on ‘Factor C’
By Cass R. Sunstein and Reid Hastie
In sports, some people are famous for “making other players better.” Magic Johnson, the great basketball player and winner of five National Basketball Association (NBA) championships, was not merely a terrific scorer, passer, and rebounder; he also transformed his teammates—some of them ordinary players—into stars. Early in his career, Michael Jordan was known to be great, maybe even the greatest of the great, but his teams just didn’t win. People wondered whether he could ever win a championship, because he “wasn’t a team player.”
In business, some people are thought to be like the young Michael Jordan—individual superstars who, apart from their own skills, don’t add much to team efforts. But there are others, like Magic Johnson, who are widely thought to make others better. Is it possible to say something about what kind of person does that? Not something impressionistic, intuitive, and anecdotal, but something that is actually based on evidence? Intriguing answers are starting to emerge, and they involve something called Factor C.
Social scientists have uncovered a statistical factor that reflects how people do on a large number of cognitive tasks; this factor is sometimes referred to as general intelligence. An obvious conclusion is that groups should seek people who have something like general intelligence. Indeed, that seems to be true. With respect to various measures of cognitive ability, there is a consistent finding across studies of many different small groups: groups with a higher average IQ tend to perform better.
Interestingly, studies of actual work groups find a somewhat looser relationship than do studies of laboratory groups. But the evidence unequivocally shows that groups with smarter members perform at higher levels. This is not exactly a remarkable observation. But it is a good point to keep in mind when you’re putting together a group of people.
There’s an important qualification to this finding, and to identify it, we have to return to basketball. In 2010, the Miami Heat basketball team created a kind of dream team, with three genuine superstars: LeBron James, Dwayne Wade, and Chris Bosh. James and Wade rank among the best basketball players of all time, and Bosh has been an All-Star. When the team was initially assembled, Sunstein had the good fortune to meet the Boston Celtics’ Hall-of-Famer Bill Russell, who is the greatest winner, and the greatest team player, in the history of professional sports. In his thirteen years in the NBA, Russell won the championship eleven times—an astounding championship rate of 85 percent. By the way, Russell was an anxious leader. He was famous for vomiting before big games, and when he was in the bathroom (vomiting), his teammates knew that they could expect to win.
Excited and a bit terrified (what can one say to Bill Russell?), Sunstein asked Russell whether the Miami Heat would win the championship that year. Sunstein was confident that Russell would say yes; who could possibly beat a team with James, Wade, and Bosh? Instead Russell offered a definite “No!” When asked for an explanation, Russell gave a quiet but firm response: “One ball.” And in fact, the Heat lost in the finals that year to the Dallas Mavericks, a far less skilled team—but a team.
Which brings us back to Michael Jordan. By the time he retired, Jordan had become one of the greatest all-time winners, besting Johnson (but not Russell) with six championships to his credit. What happened?
Here’s a clue. Late in Game 5 of the 1991 NBA Finals at the Forum in Los Angeles, the Chicago Bulls were ahead 3 to 1 in the series, and they were clinging to a fragile lead. Though double-teamed, Jordan was still shooting a lot—and missing. During a crucial timeout, Phil Jackson, the Bulls’ coach, looked Jordan right in the eyes and said, “Michael, who’s open?” Michael didn’t answer. Jackson asked again. “Michael, who’s open?” Jordan responded: “Pax.”
“Pax” was John Paxson, an unheralded guard who was a deadly shooter, at least when no one was covering him. Jordan got the ball to Paxson, who nailed a series of open shots. The Bulls claimed the first of their NBA championships. And by the way, the Miami Heat eventually became a team as well, winning two championships after its initial defeat, arguably because of improved teamwork.
Enough basketball. We have referred to Nancy-Ann DeParle and Jeff Zients, the anxious leaders in the Obama administration. They’re very different, but they know there’s just one ball, and they know who’s open. Although they’re worriers, they don’t just admire problems; they see around the bend and hunt for solutions. DeParle and Zients know the importance of assembling teams, not simply on the basis of skill, but also on the basis of teamwork and the capacity to add what the group most needs.
Good players mesh. This observation may be a cliché, but it’s nevertheless important to keep in mind: personalities—not merely abilities—matter. In this regard, most people’s intuition, at least in Western cultures, would lead us to two conclusions. First, some personality types are far better than others (Bill Russell and Magic Johnson, as opposed to the early Michael Jordan). Second, successful groups have a good mixture of personality types.
Those observations are not exactly wrong, but they do not take us very far toward specifying the right mixture of personality types for a particular task. What kinds of personalities would be best able to figure out an investment strategy for next year? To decide how to market a cell phone? To conduct a negotiation to acquire a company? To invent a new pharmaceutical drug or execute an emergency evacuation? To figure out how to get a start-up off the ground? To plan a litigation strategy for a copyright case?
A Popular, Bad Answer
To answer such questions, by far the most popular choice in American companies has been the Myers-Briggs Type Indicator, a test that separates individuals into sixteen basic “types” based on their self-reports of life preferences. A lot of companies take the test seriously.
Before we explain why pigeonholing people into personality types is not a good idea in business, let’s look at the general disadvantages of using personality assessments to predict behavior. Unlike members of more collectivist cultures (think Asia), Westerners seem obsessed with people’s personality traits as a means of understanding and predicting future conduct. If we want to know what employee Smith will do under conditions of time pressure, we tend to ask ourselves, What kind of person is Smith? That isn’t a crazy question, but in practice, even the most valid personality tests have only modest predictive power. Nonetheless, many people quite confidently rely on impressions of personality to explain the actions of their friends and colleagues. Their confidence is often misplaced.
The problem is that we exaggerate how consistent people are across both situations and time. If someone is lazy around the house, we might conclude that he is also likely to be lazy at work or in the gym. If someone was easily distracted in class, he will also be easily distracted at work. But in many cases, this kind of cross-situational behavioral consistency does not exist. Most of us are a lot less consistent in how we behave in different situations than we think we are. This goes in spades for our beliefs about the consistency of other people. Relying on these mostly false beliefs about behavioral consistency and hence predictability, we are overconfident about how accurately we understand and can predict others.
Social psychologists call this tendency to rely excessively on personality descriptors, and to be overconfident in our capacities to predict behavior, the fundamental attribution error. Attribution refers to everyday explanations that claim that other people’s behavior is a product of a trait (e.g., extroversion, laziness) possessed by the actor. If even weakly valid traits, like extroversion and conscientiousness, have low predictive value, an unscientific trait categorization like the Myers-Briggs score is going to be totally useless for purposes of prediction or for designing a team or another collective.
Consider, for example, this question from the Myers-Briggs test: “Would you rather work under a boss (a) who is good-natured, but often inconsistent or (b) who is sharp-tongued, but always logical?” Your answer to this question is aimed to assign you to a “feeling” or a “thinking” category. Or consider this question: “Is it true (a) that facts ‘speak for themselves’ or (b) that facts ‘illustrate principles’?” Your answer indicates whether you belong in the “perception” or “judging” category. Or this: “Are you the kind of person (a) who is external and communicative and likes to express yourself or (b) who is internal and reticent and keeps to yourself?” Your answer types you as extroverted or introverted. The sixteen Myers-Briggs types are defined by four dimensions: extroversion-introversion, sensing-intuition, thinking-feeling, and judging-perception.
Careful studies have shown the Myers-Briggs test does not accurately predict behavior. (Photo: GordSpence/Flickr, used under a Creative Commons license)
By some reports, almost 90 percent of the major US companies use the Myers-Briggs test for employee selection, placement, or counseling. But careful studies show that the test does not predict behavior of any kind with much validity. One problem is that the test has what statisticians call low test-retest reliability. If you retake the test after a one-month gap, there’s a 50 percent chance that you will fall into a different personality category from your original category. With such low reliability, the test is unlikely to have much predictive validity in companies that are deciding whether to assign an employee to a team for a three-month project.
One group of researchers, for example, studied the actual value of the Myers-Briggs test in teamwork applications and concluded that it “does not account for behavioral differences nor does it exactly clarify which characteristics of a particular function an individual may exhibit.” More generally, careful scientific reviews, including statistical meta-analyses, find little value in personality test scores, including scales that are more valid predictors of behavior than the Myers-Briggs. It follows that the Myers-Briggs is not likely to provide guidance for managers trying to staff a team to perform a novel task.
At the same time, and the weakness of the Myers-Briggs notwithstanding, team performance does seem to depend on whether people are sociable in the relevant sense (a point for Bill Russell). More technically, research suggests that measures of preferences for social working conditions have provided fruitful clues for a successful team. Groups composed of people who prefer to work in teams do end up working better. While the relationship between social preference and work success appears to be modest, it is real, and wise groups take it into account.
An Intriguing Finding
We need a lot more work on whether certain types are helpful to groups. But one especially thought-provoking development in recent research has been reported in studies from a group associated with the Center for Collective Intelligence at MIT. These researchers wondered if there could be some general method to assess the problem-solving capacity of a team across many types of intellectual and social problems. They conducted two large-scale tests of two- to five-member groups, solving problems such as brainstorming, answering IQ test questions, solving moral dilemmas, and even playing checkers.
Their central finding is that three individual-member measures combined into a useful measure of collective IQ, which the researchers called Factor C. First, the average of the members’ scores on a test of social perception (the Reading the Mind in the Eyes Test) predicted performance: the higher the average, the higher the group’s performance. This test was originally invented by the psychologist Simon Baron-Cohen (brother of the actor and comic Sacha Baron-Cohen) to diagnose autism in children. The person taking the test is shown a series of photos of just the eyes of another person and asked to judge what emotion the person in the photo is experiencing (e.g., playful, irritated, bored).
The second measure, unevenness of participation, or the tendency of a few members to dominate a discussion, was negatively related to group performance. Intriguingly, the third measure, the number of women on the team, positively predicted performance.
Perhaps the most striking conclusion is that the Factor C measure predicted team performance better than conventional measures of intelligence did. (Compare the 2010 Dallas Mavericks, consisting of team players, to the star-studded 2010 Miami Heat.) Average IQ and highest IQ were not correlated with team performance nearly as highly as Factor C. More technically, IQ measures were correlated in the +.10 to +.20 range; Factor C was correlated at the high level of +.50 (+1.00 is the highest possible correlation).
The other noteworthy result was the simple, direct relationship between the percentage of women members and performance. This correlation was not simply a diversity factor; rather, the more women, the better the performance. Other research supports this basic finding. But we should be pretty careful here, because the finding may be explained by the observation that women are also consistently better than men at a variety of social perception and social judgment tests (like the Reading the Mind in the Eyes Test used in these studies).
With respect to Factor C, Nancy-Ann DeParle and Jeff Zients are superstars. DeParle can size up a room, and a person, in an instant. She can take its emotional temperature, see behind the happy talk, and supply exactly what’s needed. And within Sunstein’s first month at the White House, a (female) friend of his said, with a combination of incredulity and admiration, that Jeff Zients “has the most emotional intelligence of any man I’ve ever met.” At the time, Zients was a newcomer to government, and he had a ton to learn. But he had already mastered something very much like Factor C.
To be sure, we should be careful to avoid extravagant conclusions here. At this stage, it is difficult to identify exactly what underlies the high correlation between Factor C and performance on the problem-solving tasks. What seems to be the most important is the capacity of the individual members to cooperate with one another and to coordinate their performance. Unfortunately, it is always difficult to define the true causal factor in correlations of this type.
Nonetheless, the finding does suggest an important direction for research on whether particular individuals are especially good (other things being equal) at contributing to group performance.
Take that possibility seriously and, by analogy with research on individual performance, see if a general, quantifiable measure of Factor C can be defined. Such a measure would be extremely valuable in allowing us to design groups that are more effective. It would also stimulate further research on the nature of the effective ingredient captured at least partly by Factor C.
Beyond the institutional recommendations made here, wise groups should devote real attention to social abilities, including the capacities both to participate and to listen, in selecting personnel and in devising social norms. A person’s preference to work on teams, especially when the preference is linked to social skills, is a good predictor—as is the ability to read other people’s emotional states.
Face-to-Face versus Online
But when, exactly, will Factor C really matter? Some groups have a lot of members who operate independently. Some of these members have little emotional intelligence, and in the usual respects, you wouldn’t consider them team players. Maybe they’re a bit autistic. But they’re sensational at their individual jobs, and the groups are a lot better because of them. Most groups have members and even leaders like that. (Steve Jobs was not exactly famous for his emotional intelligence.) Why aren’t they a problem?
It is important to make a distinction that cuts across many types of collectives (teams, committees, electronic crowds, and so forth). Some team members act wholly or somewhat independently to complete the task. Other team members have to coordinate, perhaps even at the lowest levels of work, on that task. In terms of evaluating the importance of personality, the difference greatly matters.
Many tasks can be “crowdsourced,” in the sense that individuals can work on their contributions independently until the ultimate stage in the task requires integrating or selecting a contribution. This is true in many forecasting tasks (“take an average”), tournaments (everyone suggests his or her individual best solutions and anyone can see which solution is the best), elections (each voter seeks out some information, perhaps with social deliberation, and then submits a vote or conclusion independently). In other cases, ingenious aggregation devices (prediction markets) integrate individual, independent solutions into a collective answer.
By contrast, many collective endeavors require coordination from the beginning to the end. The most obvious example is the classic face-to-face, physically connected team that must combine interdependent pieces and adjust them repeatedly to make them fit together into a functioning whole. (By the way, there is a lot of interest these days in the idea of telecommuting, which can make work a lot better for a lot of people, especially if they have young children. And if coordination is not required, telecommuting should be just fine. But when face-to-face interaction is needed—and it often is—telecommuting comes with a real cost. For overall productivity, the record of telecommuting is mixed.)
Consider a team constructing a physical product—say, a bridge, a ship, or an entrée. The members must interact frequently and with quick, coordinated responses, or the product risks falling apart. Many less tangible “products” also require more coordination and real-time interaction. For example, the development of a litigating position within a law firm may well require continuing discussions, and if people are trying to create a new tablet or computer, it is probably best if they work with one another.
What is remarkable about the rise of electronic networks is how many tasks that seemed to require face-to-face, real-time collaboration can be performed with much less coordination over networks. The game-changing example is software coding, which seemed to demand face-to-face teamwork until landmark projects like the open-source operating system Linux were developed.
Before 1992, there were many efforts to create distributed software systems to support the collaborative development of computer programs. Many commercial programs were written online, with loosely synchronized team processes. But in 1991, Linus Torvalds took this method to the next level when he invited thousands of programmers to contribute to his operating system and shifted the coordination to a late-stage selection and editing process, instead of an early-stage guidance role. Obviously, Wikipedia is another example of a new method of collaborating, where identification occurs mostly in a wide-open, early stage and where coordination and integration only occur very late in the process in a selection and editing stage.
All this is to say that social skills and an appetite for working on teams may not be essential in every kind of successful collective, especially those that do not require a face-to-face, real-time process. Extreme independence may even be a plus. But if the team process requires coordination in early stages or throughout the process, then Factor C and other aspects of sociability will matter a great deal. There may be only one ball, but whether that’s a problem depends on the nature of the task that the group is asked to complete.
Cass R. Sunstein, the former administrator of the White House Office of Information and Regulatory Affairs, is the Robert Walmsley university professor at Harvard Law School and a Bloomberg View columnist.
Reid Hastie is the Ralph and Dorothy Keller Distinguished Service Professor of Behavioral Science at the University of Chicago’s Booth School of Business.
Reprinted by permission of Harvard Business Review Press. Excerpted fromWiser: Getting Beyond Groupthink to Make Groups Smarter by Cass R. Sunstein and Reid Hastie. Copyright 2015. All rights reserved.