You want to be boss. You probably won’t be good at it.

Good managers are hard to find. Most companies pick managers based on personality traits, age or experience — and according to a recent National Bureau of Economic Research paper, they may be doing it wrong.

Co-authored by David Deming, Isabelle and Scott Black Professor of Political Economy at Harvard Kennedy School, the study concludes that companies are better off when they select managers based on two measures highly predictive of leadership skills.

The Gazette talked to Deming about the study’s findings. This interview has been edited for length and clarity.

What are the qualities that make a good manager, and why is it so hard to find them?

Being a good manager requires many different qualities that often don’t exist in the same person. First is the ability to relate well to others, to create what Amy Edmondson and others have called psychological safety, meaning the ability to make people feel stable and secure in their role so they are comfortable with critical feedback. That’s a key component of being a good manager. Communication skills are also essential. As a manager, you should know that there’s not one good way to deliver feedback to your workers because the words you use and the way you frame your statements also matter.

At the same time, you must also be analytically minded and open to different ways of doing things and be able to take a step back and reassess whether your team or organization is working as well as it could be. Overall, being a good manager requires both interpersonal skills and analytical skills. You also need to have a strategic vision — which is something that our study does not capture. Managers must have a sense of what their organization is trying to accomplish. Any one of those skills is hard to find. Having all three, and knowing when to use them, is even more difficult.

One of the paper’s most surprising findings is that people who self-nominate to be managers perform worse than those randomly assigned. Why is that?

In the study, we randomly assign the role of manager. That was half of the experiment. In the other half, we asked people which role they wanted, and we assigned the role of manager to the people with the greatest preferences for being in charge.

We found that people with the greatest preference for being in charge are, on average, worse than randomly assigned managers. It’s hard to know exactly why because there are a lot of factors in play, but we show evidence in the paper that they are overconfident in their own capabilities, and they think they understand other people better than they do. We all know people like that.

This was a surprising finding. And it’s important, because interest in leadership plays a big role in how companies pick managers. Companies have their own hiring and employee evaluation policies of course — they don’t pick managers randomly like we did — but it’s surely true that preference for leadership plays a big part in who gets promoted to management. For example, we find that men are much more likely to prefer being in charge, but they aren’t any more effective than women in the role of manager.

The main lesson I take from this finding is that there’s a big difference between preferences and skills; just because you want to be a manager doesn’t mean you’re going to be good at it. Organizations that take more scientific or analytical approaches to identifying good managers are going to come out ahead.

What are the best predictors for selecting a good manager, according to your paper?

It has nothing to do with how a person looks, how they speak, or what their preferences or personality traits are. None of those things are predictive. There are only two things that are: One is IQ as measured by the Raven’s Progressive Matrices test, which measures general and fluid intelligence, spatial reasoning, problem-solving, etc. But the one that’s more interesting to me is a measure of what we call economic-decision-making skill, or the ability to allocate resources effectively, that my co-authors and I created in a different paper. We use that very same measure in this experiment, and we found that it is highly predictive of being a good manager.

Why do you think these two tests predict being a good manager, but other traits like age, experience, personality, or gender do not?

If you want to predict who’s going to be going to be good at a specific performance task, in this case, managing a team to solve a problem, the best predictors are most closely related to what you’re asking someone to do. What matters is the ability to make decisions about the allocation of resources under time constraints; how to organize and motivate the members of your team to produce the most output. The lesson for me is that it’s a crutch to use personality traits and preferences to predict performance because they’re not that closely related to the performance you’re interested in.

We see this pattern elsewhere. There’s a huge research literature on figuring out who’s going to be a good teacher in the classroom, and study after study finds that characteristics such as age, gender, education, SAT scores, college major don’t do a very good job of predicting who’s going to be a good teacher. Yet if I put you in the classroom for a little bit of time and I see how much you improve student learning, that is a very good predictor, because it’s very closely related to the thing you ask people to do. If you want to know who’s going to be a good manager, make them manage. Don’t just rely on personality characteristics, or whether they raise their hand to say, “I want to do it.”

Why is it important to have good managers?

At the broadest level, it’s important to have good management because companies, universities, and other organizations face such an open-ended strategic landscape. They must tackle a variety of issues, such as where they should direct their attention, what are the most important things to focus on, and how to deploy resources toward solving certain problems. If you look at major corporations, they tend to be conglomerates that have many different divisions that do many different things. Google, just to give one example, in the beginning had a core product: a search engine. But now Google is Alphabet, and it still does search, but it also does venture investing, autonomous driving, drug discovery, and many other things.

If you zoom down to the micro level, a manager who leads a team of three or four employees faces the same sort of problems: What should I focus on? Who’s going to do what? How do I give people feedback? What are each person’s strengths and weaknesses? To be an effective manager, you must think about how to assign workers to roles to achieve the greatest success, and you must know how to communicate with a person to help them improve. The skill of being a good manager is probably underappreciated. Good managers are not necessarily the most vocal leaders; sometimes they’re quiet but effective, like diamonds in the rough.

The paper you and your co-authors wrote came up with a novel method to identify good managers. Can you explain?

It’s a hard problem to solve, because part of what makes a good manager is the people they’re supervising. If you give a manager a team of workers who aren’t very capable, that team is going to do a poor job, and if the workers are all-stars, they will make the manager look good regardless. In other words, when a team succeeds, we don’t know how much credit or blame to assign to the manager compared to other members of the team.

To solve that problem, we bring a bunch of people into a controlled lab setting, and we assign them a group task that they must do together. We randomly assign the role of manager to one of the three people on the team, and we ask them to lead their group in the task, and we see how well they do. Then we randomly assign each manager again to another group of workers. Each time, as a manager, you’re getting a different set of people, so we have a way to account for the quality of the workers you’re getting. And since we’re assigning workers, we can also identify who’s a good worker because we can see their performance with different managers.

What do you think the paper’s main contributions are to the literature of leadership and management in general?

I think the paper’s main contribution is to open the door to the idea that we can be scientific and analytical about selecting managers and that management is not a squishy thing that we can never get our arms around. We can measure management skill, and measuring it well unlocks huge productivity gains for organizations and for people. We’re doing this experiment in a lab; it’s not a real-world setting, but we are in talks with several folks to do this in the field. I do think it would work because we’re asking people to manage and we’re measuring their performance, and we’re showing you that there’s a repeatable predictive quality to this. Our contribution is to outline a very simple methodology for measuring who’s a good manager, and to say to people that they can use it. Figure it out in your own organization, and you will unlock big productivity gains.