Many people who have worked in teams can name someone whose presence made the whole group work better. Meetings ran more smoothly, tasks landed with the right person, and the team got more done than its members might have managed apart. Economists at Harvard set out to test whether that impression holds up under controlled conditions, and whether such people can be identified.

Their answer, published in the journal Econometrica in 2021, was that they can. Some participants consistently lifted the performance of whichever team they were placed in. What distinguished them was not intelligence, personality or education, but how well they scored on a test of recognizing what other people are thinking and feeling.

Testing teamwork like a laboratory experiment

The study, by Ben Weidmann and David Deming of the Harvard Skills Lab, was designed to get around a basic problem with judging teamwork in real workplaces. In an office, it is hard to separate one person’s contribution from that of the colleagues around them, the manager above them or the luck of the project they were handed.

So the researchers built teams at random. According to a 2023 account in Harvard Magazine, participants first completed tasks on their own, so their individual abilities could be measured. They were then placed in groups of three to work on similar tasks together, including recognizing shapes, memorizing words and solving math puzzles. The groups were shuffled repeatedly, so each person worked with several different sets of teammates.

With individual scores in hand, the researchers could predict how well each team should do. Then they looked for people whose teams kept beating that prediction regardless of who else was in them. As the paper’s abstract puts it, “Some people consistently cause their team to exceed its predicted performance.” The authors called these people team players.

What set the team players apart

The researchers then compared team players with everyone else. They found no difference in IQ, personality, education or gender. The one measure that did stand out was the Reading the Mind in the Eyes Test, a standard assessment in which people look at photographs of eyes alone and choose the word that best describes what the person is thinking or feeling.

Team players scored higher on that test. When the authors combined it with the team player measure into a single social skills factor, they found that social skills improved team performance about as much as IQ did.

How team players helped is less certain. The authors reported suggestive evidence that they raised their teammates’ effort: teams that included one tended to use more of the time they were given, which in turn was linked to better results. Another possibility the researchers raised is that team players help make sure each task goes to the person best suited to it.

The authors are open about the study’s limits. It was relatively small, run in a controlled setting with set puzzles rather than real jobs, and Weidmann described the eyes test as “very first-generation tech.” The experiment shows that the effect is consistent and measurable in the lab. It does not show how large it is in a typical workplace.

A pattern seen in other groups

The Harvard findings echo earlier work on what makes groups effective. In 2010, a team led by Woolley, publishing in Science, randomly assigned 699 people to 192 groups that tackled a range of problems together.

According to a summary by the Harvard Kennedy School’s Gender Action Portal, groups that did well on one kind of task tended to do well on others, suggesting a general “collective intelligence.” That group-level ability was only moderately linked to the intelligence of individual members.

What mattered more was how the group worked. Groups did better when members scored higher on the same eyes test of social sensitivity, and when conversation was shared rather than dominated by a few people. Groups with a higher proportion of women also scored higher, which the researchers linked to women’s higher average scores on social sensitivity. Group cohesion, satisfaction and motivation did not predict performance.

Why employers have been paying more for social skills

The lab results fit a broader shift in the job market that Deming documented in an earlier study. Writing in the Quarterly Journal of Economics in 2017, he analyzed US employment data from 1980 to 2012. According to the paper’s abstract, jobs requiring high levels of social interaction grew by nearly 12 percentage points as a share of US employment over that period, while jobs that were math-intensive but less social, including many in science and technology, shrank by about 3.3 points.

The pay gap moved in the same direction. Deming found that the labor market return to social skills was much greater in the 2000s than in the mid-1980s and 1990s. As the Harvard Gazette reported, jobs requiring both strong social and strong math skills saw the biggest gains, with employment up 7.2 percentage points and wages up 26%.

Deming’s explanation is that computers have taken over a growing range of routine tasks, including routine thinking work, but remain poor at dealing with people. “Social skills reduce the cost of coordinating with others,” he told the Gazette. In his model, people who coordinate well can trade tasks with colleagues and specialize in what they do best, which makes the whole group more productive.

What one large company found

Google reached a similar conclusion from inside its own workforce. Its internal study of team effectiveness, known as Project Aristotle, examined what distinguished its most effective teams. As InfoQ reported in 2016, the company concluded that how team members interacted mattered more than who was on the team.

Of five factors it identified, researcher Julia Rozovsky described psychological safety, the sense that it is safe to take risks and speak up in front of colleagues, as “most foundational and most important.” The others were dependability, clear structure, meaningful work and a sense of impact. Project Aristotle was a company study rather than peer-reviewed research, and at least one consultant has questioned how much it added to existing work on teams, but its emphasis on interaction rather than individual talent matches the academic findings.

What the research can and cannot say

None of these studies offers a simple checklist. The Harvard experiment identified team players with a single test that one of its authors describes as early technology, and it used puzzles rather than real projects. The collective intelligence work and the labor market data show patterns across many people, not what any one person should do. None of the studies reported here tested whether these skills can be taught.

What the studies do suggest is consistent. Across controlled experiments, group studies and decades of employment data, the ability to read other people and coordinate with them shows up again and again as something that makes teams work better and that employers increasingly pay for. Raw intelligence still matters. On this evidence, it is not the only thing that does.

Photo: Theo Decker / Pexels