What Works

Iris Bohnet

• Publisher: Belknap Press
• Year: 2016
• ISBN:
  9780674968592
• Topic: Business
• Subtopic: Organisational Behaviour
• Index: Business

Shortlisted for the Financial Times and McKinsey Business Book of the Year Award
A Financial Times Best Business Book of the Year
A Times Higher Education Book of the Week
Best Business Book of the Year, 800-CEO-READ Gender equality is a moral and a business imperative. But unconscious bias holds us back, and de-biasing people's minds has proven to be difficult and expensive. By de-biasing organizations instead of individuals, we can make smart changes that have big impacts. Presenting research-based solutions, Iris Bohnet hands us the tools we need to move the needle in classrooms and boardrooms, in hiring and promotion, benefiting businesses, governments, and the lives of millions."Bohnet assembles an impressive assortment of studies that demonstrate how organizations can achieve gender equity in practice… What Works is stuffed with good ideas, many equally simple to implement."
—Carol Tavris, Wall Street Journal "A practical guide for any employer seeking to offset the unconscious bias holding back women in organizations, from orchestras to internet companies."
—Andrew Hill, Financial Times

Part One

THE PROBLEM

1

Unconscious Bias Is Everywhere

Meet Howard Roizen, a venture capitalist, former entrepreneur, and proficient networker. A case study taught at many business schools describes how he became a power player in Silicon Valley. He co-founded a very successful tech company, then became an executive at Apple, subsequently turning his attention to venture capitalism. Most recently, he became a member of the boards of directors of several prestigious companies. He is a friend of Bill Gates and was close to Steve Jobs. He maintains one of the most extensive networks in Silicon Valley.

After studying the case, students were asked to evaluate Howard’s performance. They rated him as highly competent and effective. They also said that they liked him and would be willing to hire him or work with him.

But Howard does not actually exist. His real name is Heidi. He is a woman. And when studying the absolutely identical case with the protagonist being female, students find Heidi as competent and effective as Howard, but they no longer like or want to work with this successful entrepreneur and venture capitalist.

My friend Kathleen McGinn of Harvard Business School originally wrote the case study about Heidi Roizen in 2000 to highlight the steps taken by one successful entrepreneur to build and leverage personal and professional networks. A few years later, I encountered it again in a research seminar. Two professors had given half of their students McGinn’s original case, which accurately identified Heidi, and the other half the same case, but with “Howard” swapped in for Heidi. This allowed them to compare how students felt about Heidi and Howard.¹

Many business schools have since run the experiment, using it as a pedagogical tool to help their MBA students experience gender bias. Afterward, the students realize that the prototypical leader in their minds is male. Heidi does not look or act the part: she cannot be competent and likable at the same time. What is celebrated as entrepreneurship, self-confidence, and vision in a man is perceived as arrogance and self-promotion in a woman.

Women can’t win. If they conform to the feminine stereotype of nurture and care for others, they tend to be liked but not respected. Dozens of studies have now demonstrated that women face a trade-off between competence and likability. Women in stereotypically male domains encounter backlash at every juncture: when getting hired, compensated, and promoted. Psychologists believe that these negative reactions are due to a clash between our stereotypical perceptions of what women are or should be like (their gender roles), and the qualities we think are necessary to perform a typically male job. If women like Heidi demonstrate that they can do a “man’s job,” they no longer fit our mental model of the “ideal woman.” They violate norms, and people do not find norm violators appealing. Put differently, women who violate norms pay a social price.²

Most studies on the topic have been run with white men and women, mostly in the United States. We know little about whether the same dynamics apply to other races and ethnicities. What we do know is typically based on small samples, for the experiments have not been replicated widely. Thus we need to interpret them with care. While there exists no comprehensive study of whether the agency penalty established for white, female Americans also applies to African, Asian, Hispanic, and Native Americans, Robert Livingston and colleagues examined the question for African Americans. They found that black women are considered neither prototypical women nor prototypical blacks. Does this buffer them from some of the gender stereotypes white women experience? Indeed, experimental evidence showed that black women did not experience the same kind of backlash as white women when they expressed dominance rather than communality. In contrast, dominant African American men were penalized while white American men were not. If African American men are perceived as nonthreatening, they benefit. Physical features expressing warmth and deference have been shown to be an advantage for black male CEOs, but these same attributes hurt white male CEOs in the United States.

These findings are in contrast to models of double jeopardy, assuming that people with multiple “subordinate identities,” for example, African American women, are subject to more prejudice than those with only one—African American men or white women. Identities appear to be not only additive but intersecting in ways that current research is starting to uncover. Erica Hall and collaborators suggest that a person’s gender profile is composed of the “genders” of a person’s sex and race and that we should take this gender profile into account to better understand gendered perceptions of occupational “fit.”³

Gender stereotypes appear to generalize to some degree across cultures. For example, many societies differentiate warmth from competence when judging social groups; high-status groups are stereotypically perceived as competent but lacking in warmth, and the male stereotype is associated with the cultural ideal. The pervasiveness of these stereotypes has real effects on how people are evaluated.⁴

Again and again, the pattern that has emerged from experiments assessing women who are performing stereotypically male jobs—an aircraft company’s assistant vice president for sales, for example—looks like this:

• When performance is observable, successful women are rated as less likable than men.
• When performance is ambiguous, successful women are rated as less competent than men.

In the latter case, when evaluators cannot easily measure quality, they fill in the blanks with stereotypes. In one recent study, Katy Milkman, Modupe Akinola, and Dolly Chugh sent thousands of professors at academic institutions across the United States an email from a phantom student requesting a ten-minute meeting the following week to learn more about a doctoral program the professor was involved in. The name of the student, however, varied: some were self-evidently male, others female; each sounded either white, African American, Hispanic, Indian, or Chinese. Almost 70 percent of the professors responded, and most agreed to meet with the student. However, they were significantly less likely to respond to a student who was not a white male than to a white male student. The bias was most pronounced in the field of business administration, with 87 percent of white men but only about 62 percent of all the women and students of color combined receiving a response. The professor’s own demographic characteristics generally did not matter—Hispanic female professors were as likely to favor white male students as white male professors were (the one exception being Chinese students who requested a meeting with a Chinese professor). Did the professors, likely unconsciously, perceive the white men as more competent or more deserving of their attention?⁵

Another field experiment is illuminating. A man and a woman applied for a laboratory manager position at a university. They were otherwise identical and had the same qualifications. Science faculty rated the male candidate as significantly more competent than the female candidate and were more likely to hire him. The faculty’s pre-existing bias against women affected their evaluation. In a further exploration of gender biases associated with STEM fields (science, technology, engineering, and mathematics), researchers had people hire candidates to perform a specific job: a well-defined arithmetic task on which both genders perform equally well. When the evaluators knew nothing except the candidates’ gender, men were about twice as likely to be hired as women. The bias was barely affected, however, when the candidates were allowed to provide information on their qualifications. In keeping with other findings, when the male candidates did so they tended to boast, and when the female candidates did so they tended to under-report how good they were, behaviors evaluators did not take into account. Only information about how well the candidates had done on the task in an earlier round helped reduce the bias. But even this precise information was unable to eliminate it completely.⁶

How about men? What if men are evaluated for a counterstereotypical job, such as the assistant vice president of Human Resources? While this question has been studied less, it appears as if men in counterstereotypical roles experience some of the same bias-informed dynamics as women do—with one important exception: their likability is not affected. Male human resources managers may well be evaluated as less competent, but holding a stereotypically female position does not make them less likable. Women, thus, are in a double bind that men are not. They are perceived as either likable or competent but not both.

Biases about whether or not a person fits matter. Being disliked, in addition to being extremely unpleasant, can injure, even derail your career. Mothers might be even more affected than women without children. Much evidence suggests that stereotypical perceptions of warmth work against mothers in the labor market. Unlikable individuals have been shown to receive worse performance ratings and be deemed less worthy of salary increases or promotions than their more likable counterparts. This appears to be true for both men and women. But while colleagues have lots of reasons to dislike someone, from dishonesty to arrogance, “it is only women, not men, for whom a unique propensity toward dislike is created by success in nontraditional work.” This quote, from Madeline Heilman, one of the leading researchers in this field, can be rephrased more bluntly. Because of our biases, we tend to react to successful women much like we react to dishonest men: we do not like them and do not want to work with them.⁷

Numerous additional field experiments have been conducted in which male and female candidates, otherwise equally qualified, have applied for the same jobs, and again and again bias has been found to influence outcomes. Whether applying to be waiters or waitresses in the United States, or accountants, engineers, computer programmers, or secretaries in the United Kingdom, or financial analysts in France, sex-based discrimination has been influential. A review of the evidence concludes that both women and men tend to be discriminated against in jobs that are associated with and dominated by the opposite sex. Men were discriminated against when seeking jobs as secretaries and women discriminated against when seeking jobs as engineers.⁸

While the evidence is still thin, there are some early signs that we are starting to see a change in this trend. For a 2013 segment of CNN’s Anderson Cooper 360°, the Heidi-Howard study was repeated at New York University’s Stern School of Business. The students still rated the successful female leader as less trustworthy than her male counterpart, but they no longer liked her less. Indeed, they reported being more willing to work for her than for the successful man. In 2015, a scientific study reporting a reversal of this trend for entry-level jobs in academia appeared in the Proceedings of the National Academy of Sciences. Wendy Williams and Stephen Ceci found in five hiring experiments—in which faculty evaluated profiles of hypothetical male and female candidates applying for jobs as assistant professors in biology, engineering, economics, and psychology—a substantial pro-female bias in all disciplines but economics. Are we starting to harvest the fruits of all the work that has been done to equalize the playing field in science, technology, engineering, and math—at least at the entry level?

As I write this book, it is too early to tell. Maybe faculty in biology, engineering, and psychology are overcompensating for past inequities in an effort to move the needle a bit more toward gender equality in hiring. Perhaps STEM fields are catching up with other areas where we have seen an increase in gender diversity at the entry level. Why this is happening in some fields but not in my own, economics, remains unclear.⁹

Another field where gender diversity at the entry level has dramatically increased is law. Analyzing panel data from more than 6,000 lawyers employed by one of the largest law firms in the world from 2003 to 2011, Ina Ganguli, Ricardo Hausmann, and Martina Viarengo found that an almost equal number of male and female lawyers entered at the associate level. However, this did not translate into closing the gender gap at the top, where only 23 percent of partners were female. Instead, the gender gap at the most highly ranked position, partner, was strongly related to a gender gap in promotions. In their book Through the Labyrinth, Alice Eagly and Linda Carli discuss how gender stereotypes constrain women’s access to leadership roles. In particular, biases affected evaluations of women vying for the very top positions, or what is commonly known as the glass ceiling effect. In contrast to the entry level, there is no closure of the gender gap at the top in sight.¹⁰

Ganguli and colleagues’ research shows why. The lawyers in their study worked for one firm in thirty-three different offices located in twenty-three countries on four continents. The three researchers had access to an almost unprecedented amount of individual-level data, including wages, bonuses, performance appraisals, educational background, employment status, career trajectory and leaves, as well as demographic characteristics. Promotion gaps persisted after controlling for all of these variables, including the fact that men and women left the firm at equal rates. But the degree of promotion gaps varied across countries—even though, at least in theory, the firm was guided by the same set of policies and practices. Female lawyers found it hardest to climb up the career ladder in countries where stereotypical thinking about gender roles was most pronounced, such as in the Russian Federation, Singapore, or Thailand, based on both survey data from the World Values Survey and the World Economic Forum Global Gender Gap Index as well as data concerning each country’s gender gap in political representation. The playing field for female lawyers was more even in countries such as Belgium, the Netherlands, or Sweden.¹¹

Another study examined the performance evaluation bias toward highly successful women in a number of contexts, including commanding officers in the US military. It turns out the evaluating officers gave female subordinates whose pay grades were close to their own lower performance scores than they gave to male subordinates. The authors identify this phenomenon as gender hierarchy threat. Female (but not male) subordinates whose objective performance was strong were punished by male (but not female) evaluators for violating gender norms.¹²

Let’s take stock: the gender gap in leadership is real; its relationship to the gender gap in promotions is real; a connection exists between the promotion gap and the extent of stereotypical attitudes. These dynamics have been demonstrated in various contexts and countries, but too little is known to determine to what extent they generalize from whites to all other demographic groups. The stereotypes about “leadership fit”—or lack thereof—are hardly based on evidence. There simply are not enough women in positions of leadership to draw reliable inferences. Interestingly enough, when employees who prefer male leaders in theory are exposed to female leaders, they do not give them lower ratings, a large 2011 survey finds. The bias against female upper-level managers is in our heads—or, to quote from one of my favorite textbooks on organizational economics by two Stanford economists, Paul Milgrom and John Roberts: “even if the beliefs are completely groundless, no disconfirming evidence ever is generated because women never get a chance to prove the beliefs are wrong. Thus, the baseless beliefs survive, and with them, the unjustified discrimination.”¹³

Our minds are not well equipped to deal with what is commonly known as survivor bias. We constantly make inferences based on biased samples. The archetypal example is a study of World War II bombers. With the hope of making them safer, the planes were examined for weaknesses after they...