Assistant Professors Zoë Cullen and Katherine Coffman; image by John Ritter

Assistant Professors Katherine Coffman and Zoë Cullen came up as economists in largely male-dominated environments where, during their years of graduate studies and private-sector work, it was impossible to overlook the imbalanced gender dynamics around them. Now, as experimental economists, their research is helping define the complex factors that contribute to the workplace gender gap—and sparking ideas about how to mitigate it. Here, they talk about the nature of experimental research and the give and take between lab and fieldwork.

How do you settle on a particular research question to investigate?

Katherine Coffman: I’m looking for two things. The most important one is whether it has the potential to really make a difference for outcomes. It has to be a big and managerially relevant question, with first-order importance. The second part is whether I have a clever way of studying it. Can I design an experiment to isolate exactly what I think the causal mechanism is? It’s not just that A happens and then B happens, but can I actually show that A causes B? That actually is pretty limiting to some of the types of research questions you can tackle, at least initially.

What’s an example of that?

Zoë Cullen: Harassment is something that many of us who are focused on gender would like to understand better, but the idea of hiring someone to randomly harass other people in the workplace is unethical. It has made that topic elusive, for a lot of reasons, with the methods that we use.

Zoë, your study of schmoozing and the old boys’ club drew on big data sets. How do you use data to isolate the forces at play in a workplace setting?

ZC: When I’m looking for a way of causally identifying what channels are active, I’m also looking for, as Katie mentioned, a situation that is really meaningful to somebody’s life. I know that employees talk quite a bit about their managers and how that manager affects their career progress, for example, so I’m looking for ways in which employees don’t necessarily have control over that.

When employees don’t have control over something that matters to them, this usually points toward potential policy interventions and also naturally occurring interventions that we can study cleanly. By comparison, proactive choices by employees are sometimes harder to study because disentangling the type of person who makes the choice from the impact of the choice itself adds another challenge. In my data, I saw that employees who smoked would sometimes be assigned to managers who also smoked, and this led them to take more breaks together that they wouldn’t have otherwise taken. But the manager assignment was out of their control, and this meant that some part of the variation in breaks with the manager was also out of their control.

Because I’m building off of these administrative data sets that other people have collected over many years, it makes it possible for me to study questions around how interactions play out for years down the road. But in some ways the decisions about what data to collect, and how the environment is organized, weren’t made by me. Given those constraints, I have a lot of envy for experimentalists who work in a lab.

KC: There are a lot of fantastic complementarities between fieldwork and lab work. Zoë’s projects have all of this rich complexity from the field, where you can see beautiful things like long-term outcomes and high stakes. But I also think her job is harder, because she has to be really clever about what counts as random variation or quasi-random variation. In some ways, my work is at the other end of the spectrum, because a lot of my environments are really abstract. I have college students in a lab, I’ve designed the environment, I can perfectly randomly assign people to different conditions, and everything’s over in an hour. Zoë’s approach allows her to say the things that I can’t say, and vice versa.

Does having this complement of approaches help fill out a fuller picture?

ZC: It can be challenging to feel that way when our papers are sent to journal referees working in different fields that have different expectations for clean identification. But, yes, we fundamentally know that we’re in it together.

KC: I can give you an example. I wrote an experimental paper about men and women having different propensities to guess on standardized tests. Say you’re taking the SAT and you’re unsure about the answer. Do you guess or skip that question? You could look at field data and see that women skip more questions than men, but it would be hard to really know why. So I designed a lab experiment where I administered something that looks just like the SAT to college students. In the first part of the study, I looked at whether they chose to guess or skip questions.

In the second part of the study, which you could not do in the field, I administered the same test and forced people to answer all the questions. By doing so, I could see how people would have performed had they chosen to answer rather than skip. I can compare the results of a man and a woman who performed exactly the same when you force them to answer every question but the woman ends up leaving a few more blank in the first part and leaves points on the table. That’s a pretty nice way to demonstrate in the lab that this might be important to think about when we’re interpreting SAT scores.

Since then, I’ve been able to go back to the field, with that knowledge from the lab, and look at how policies that reduced skipping among test-takers—removing penalties for wrong answers—helped to reduce the gender gap in test scores.

What’s your outlook on the overall impact of research in this area?

ZC: For me, each paper’s feedback opens up so many new questions, and the whole process happens so organically, it feels like you’re having a conversation with the rest of the profession. My paper about schmoozing dynamics, for example, is one where people are pushing me to understand the reason why, when a female has a female manager, they don’t end up spending more time together the way men do with their male managers. I would like to figure out why that happens and the kinds of activities that women could be doing during the workday to have a similar effect. Each paper opens up new behavioral work that I hadn’t thought about previously and sets me on a different path.

KC: Broadly, I think you could see some of these results and still feel dispirited about the state of gender differences in the workplace, but there’s also a lot to be optimistic about. Just 20 years ago, the economics literature was in a mode of documenting gender differences, whether it was about discrimination or the different outcomes for men and women. But more recently we’ve gotten much more to a place where we’re looking at why, what are the mechanisms, and like Zoë says, what do we do about it. None of that work could have happened without the work that came before, but seeing effective interventions and things that are implementable is exciting—for us and for practitioners, too.

Katherine Coffman studies the sources of gender gaps in economically important contexts and designing and testing policies aimed at closing them. Her work focuses on the role of beliefs: How do stereotypes bias the beliefs that individuals hold about themselves (and others), and how do these biased beliefs shape decision-making? More about her paper on standardized tests can be found here.

Zoë Cullen studies the design of labor markets and the choices of employers and labor platforms that affect matters of public interest, such as pay transparency, pay inequality, and criminal-background screening. Her paper about schmoozing and the old boys’ club can be found here.

Post a Comment