Breaking the Glass Ceiling

Hurwicz Distinguished Lecture 2017 MEA Conference

October 27, 2017

Marianne Bertrand Booth School of Business University of Chicago

Background

• Substantial gains for women over the last half century in many countries around the world: – Education – Labor force participation – Labor market earnings

• Large literature documenting reasons for these gains: – Innovations in contraception – Technological progress in home production activities – Better regulatory controls against discrimination – Labor demand shifts towards industries where female skills are disproportionately represented – Etc.

• Yet: – Women remain under-represented in high status/high income occupations – When represented in these occupations, women earn less than men

Example: C-Suite

Source: Catalyst Example: Economics

Source: CSWEP, 2017 Women's Labor Force Participation & Representation in the Upper Part of the Earnings Distribution Sample: All 25-64 Years Old Women Year is: 1970 1980 1990 2000 2010 In workforce 0.48 0.585 0.689 0.696 0.725 Share of working women with earnings at or above the xth percentile of the distribution of earnings among men working full time-full year, where xth percentile is: 50th percentile 0.051 0.071 0.127 0.18 0.194 80th percentile 0.012 0.014 0.028 0.044 0.05 90th percentile 0.006 0.007 0.0119 0.019 0.022 Share of women working full time-full year with earnings at or above the xth percentile of the distribution of earnings among men working full time- full year, where xth percentile is: 50th percentile 0.073 0.102 0.175 0.232 0.256 80th percentile 0.017 0.019 0.037 0.057 0.067 90th percentile 0.008 0.009 0.016 0.025 0.028 Women's Labor Force Participation & Representation in the Upper Part of the Earnings Distribution Sample: 25-64 Years-Old Women with a College Degree or More Year is: 1970 1980 1990 2000 2010 In workforce 0.611 0.734 0.817 0.807 0.811 Share of working women with earnings at or above the xth percentile of the distribution of earnings among men working full time-full year, where xth percentile is: 50th percentile 0.058 0.067 0.131 0.195 0.198 80th percentile 0.01 0.012 0.026 0.043 0.049 90th percentile 0.005 0.004 0.01 0.017 0.021 Share of women working full time-full year with earnings at or above the xth percentile of the distribution of earnings among men working full time- full year, where xth percentile is: 50th percentile 0.086 0.098 0.177 0.247 0.247 80th percentile 0.016 0.017 0.035 0.054 0.062 90th percentile 0.008 0.006 0.013 0.021 0.027 Source: Piketty, Saez and Zucman (2016) Gender pay gaps within occupation

Source: Goldin (2014) Women's Representation in the Upper Part of the Earnings Distribution Sample: 25-64 Years Old Women with a College Degree or More Year is: 1970 1980 1990 2000 2010

Share of women working full time-full year with earnings at or above the xth percentile of the distribution of earnings among men working full time-full year in the same occupation, where xth percentile is: 50th percentile 0.235 0.214 0.258 0.327 0.343 80th percentile 0.058 0.047 0.060 0.090 0.102 Why Should We Care? • Much of the popular discussion of the glass ceiling is typically framed as an issue of rights and fairness. – “Equal pay for equal work” – Current academic thinking on the glass ceiling, however, has focused on explanations that do not rely on employers treating women unfairly • E.g. there might be observed gender gaps in earnings even when employers practice “equal pay for equal work”

• An economy that is tapping into a limited pool (men) to find its leaders must be operating inside the efficiency frontier. – Hsieh et al. (2017): one-quarter of growth in US GDP between 1960 and 2010 can be explained by declining barriers to the entry of women and blacks in occupations where they were previously heavily underrepresented.

• Other efficiency-based related to how diversity in leadership roles might be productivity-enhancing. – Many organizations making the business case for “diversity and inclusion.” – (Economic research has fallen short so far of providing robust empirical demonstrations of the economic benefits of diversity.)

Roadmap for rest of talk

1. Why are women struggling to break the glass ceiling? What does the most recent research say? – Revisiting education (briefly) – Gender differences in psychological attributes – Women’s greater demand for flexibility/work-family balance • Children • Gender roles and gender identity norms • Why still so important today?

2. What role can corporate and public policy play, if any, in accelerating convergence at the top?

Education

• Source: Goldin, Katz and Kuziemko (2006) Percent with at least a college degree by age 30

.25 .35 .3 .4 1950 Share Share of menwomen and with at leasta degree, college by birth cohort 1960 Women birth birth cohort 1970 Men 1980 1990 But…

• While it is true that women have been more likely to complete a college degree than men since the late 1950s birth cohorts…:

– Convergence in share of women obtaining professional degrees or PhDs (which correspond to highest paying occupations) has been slower in the making

– Convergence in fields of study (in terms of expected labor market earnings) has also been slower

– No “reversal of the gender gap” yet as on these dimensions

.05 .15 .25 .1 .2 1950 Expected gender gap in log 90th pctile earnings gap gender log pctile 90th in Expected gap earnings gender log mean in Expected based based degree-fieldon combinationhighest of degreecompleted 1960 Men's vs women'spotentialMen's vs earnings birth birth cohort 1970 Expected gender gap in log 80th pctile earnings gap gender log pctile 80th in Expected 1980 1990 Overall

• Women’s over-taking of men in terms of number of years of education masks important remaining differences by gender in educational choices and how they map into expected earnings.

• Women’s decision to stay away from particular educational tracks might be in part an informed decision in light of the constraints and challenges women expect in the jobs that are associated with those educational tracks.

• But additional research needed to understand the full set of considerations that go into the educational choices today’s young women make.

Gender differences in psychological attributes

• One of the most active areas of research

• Flurry of laboratory studies over the last 15 years or so have documented robust gender differences in a set of psychological attributes

• Some of these psychological attributes may have direct relevance in explaining educational and labor market choices, as well as labor market outcomes, especially at the top of income distribution

• In particular: – Women are more risk averse – Women perform more poorly in competitive environments and shy away from such competitive environments – Women negotiate less/women do not ask – Women lack in self-confidence (while men tend to be overly confident)

• Nature or nurture?

Gender and risk aversion

Source: Dohmen et al (2011) Gender and risk aversion Source: Dohmen et al (2011)

Source: Dohmen et al (2011) Gender and competition

Source: Gneezy et al (2003)

Women are less willing to compete

Source: Niederle and Vesterlund (2007)

Relevance outside of the lab

• Wave of very recent work aiming to test relevance outside the lab.

• In particular, explaining gender gap in:

– Educational choices (Buser et al, 2014) – Job entry decisions (Flory et al, 2016) – Labor market earnings (Ruben et al, 2016)

Buser et al (2014)

• Study sample: Dutch secondary school students who are enrolled in the pre- university track.

• End of 9th grade: students have to choose one of four study tracks: – the science-oriented profile Nature & Technology (NT) – the health-oriented profile Nature & Health (NH) – the social science-oriented profile Economics & Society (ES) – the humanities-oriented profile Culture & Society (CS).

• Ranking of perceived prestige/challenge of each track: NT>NH>ES>CS

• Students perform tournament entry task in the lab (competitiveness measure)

• Correlate study track choice to competiveness measure

Buser et al (2014)

• Findings:

– 40 percent of boys vs. 17 percent of girls choose NT track (most prestigious)

– 15 percent of girls vs. 8 percent of boys choose CS track (least prestigious)

– All of this despite the fact that girls are as good at math as boys, and actually have a higher GPA.

– Differences in lab-based measure of competitiveness explain up to 23 percent of the gender gap in educational track choices.

Flory and List (2016)

• Field experiment: randomize job seekers who expressed initial interest in a job position into one of six different compensation regime treatments:

– T1/T2: piece-rate • T1: fixed wage of $15 an hour; job is done in team • T2: Fixed wage of $15 an hour; job is done alone – T3/T4: tournament (T4 >> T3) • T3: worker will be matched with another person also hired for the position; base salary is $13.50 an hour ; bonus is equivalent to $3 an hour if outperform matched worker. • T4: worker will be matched with another person also hired for the position; base salary is $12 an hour ; bonus is equivalent to $6 an hour if outperform matched worker. – T5: team-based tournament • T5: places two new hires into a team; compensation contingent on the group’s relative performance with respect to another similarly composed team; base salary is $12; bonus is equivalent to $6 an hour if outperform matched team. – T6: uncertainty • T6: base salary for the position is $13.50 per hour, plus a bonus that translates to $3 per hour if their assistance happens to contribute to journal article publications (uncertainty condition).

• Two types of jobs: general vs. sports-related

Source: Flory and List (2016) Source: Flory and List (2016) (Related to lab-based findings by Shurchkov 2012 and Grosse and Riener 2010: math vs. verbal tasks) Reuben, Sapienza and Zingales (2016)

• Measure psychological attributes (including willingness to compete) among MBA graduates (surveys and lab experiments)

• Subject pool: 286 male and 123 female students from Booth – Homogenous set of men and women that have all self-selected into MBA education (and are above the admission threshold) – Men and women in this program likely not representative of their respective gender group in terms of psychological attributes (because of the self-selection)

• Correlate competitiveness with salary at exit

• Most relevant study population wrt gender gaps at the top of the income distribution

Source: Reuben et al (2016) Source: Reuben et al (2016) Determinants of earnings

Source: Reuben et al (2016) Source: Blau and Kahn (2016) Overall

• 15 years later, field-based demonstrations that gender differences in psychological traits matter for education, job choices and earnings

• However: – Magnitudes so far remain limited wrt explaining gender gap in earnings

– Domain specificity (“male job” or “male task”) of attitudes highlight fragility/context-specificity of the psychological traits being measured • Nature vs nurture – Psychological attributes as reflection of socially created norms and identities • Also suggests value of “soft” policies that would reframe education and occupational choices to make them less threatening to women – Cf literature on stereotype threat and negation of stereotype threat

Women’s greater demand for flexibility

• Many of the higher-paying jobs have long hours and inflexible schedules

• Many of the financially more rewarding careers require continuous labor force attachment in order to stay on the “fast track,” which makes it difficult to combine those careers with job interruptions

• Because women remain the dominant providers of child care (as well as other forms of non-market work) this inflexibility is particularly detrimental to them.

Male and female mean and median annual salaries ($2006) by years since graduation (Chicago Booth MBA data)

Source: Bertrand, Goldin and Katz (2010)

Labor Supply by Gender and Years since Graduation

Number of Years since Graduation 0 1 2 3 4 5 6 7 8 9 ≥ 10 Share not working at all in current year Female 0.054 0.012 0.017 0.027 0.032 0.050 0.067 0.084 0.089 0.129 0.166 Male 0.028 0.005 0.002 0.003 0.007 0.004 0.008 0.008 0.006 0.011 0.010

Share with any no work spell (until given year) Female 0.064 0.088 0.116 0.143 0.161 0.193 0.229 0.259 0.287 0.319 0.405 Male 0.032 0.040 0.052 0.064 0.071 0.077 0.081 0.082 0.090 0.095 0.101

Cumulative years not working Female 0 0.050 0.077 0.118 0.157 0.215 0.282 0.366 0.426 0.569 1.052 Male 0 0.026 0.036 0.045 0.057 0.060 0.069 0.075 0.084 0.098 0.120

Mean Weekly hours worked for the employed Female 59.1 58.8 57.1 56.2 55.3 54.8 54.7 53.7 52.9 51.5 49.3 Male 60.9 60.7 60.2 59.5 59.1 58.6 57.9 57.6 57.6 57.5 56.7

Source: Bertrand, Goldin and Katz (2010)

Gender Wage Gap by Number of Years since MBA Graduation Number of Years since MBA Receipt 0 2 4 7 9 ≥ 10 1. With no controls -0.089 -0.213 -0.274 -0.331 -0.376 -0.565 [0.020]* [0.032]* [0.043]* [0.062]* [0.079]* [0.045]*

With controls: 2. Pre-MBA -0.08 -0.172 -0.221 -0.271 -0.32 -0.479 characteristics [0.021]* [0.033]* [0.044]* [0.065]* [0.084]* [0.045]*

3. Add MBA -0.054 -0.129 -0.166 -0.2 -0.257 -0.446 performance [0.021]* [0.032]* [0.042]* [0.063]* [0.082]* [0.044]*

4. Add labor market -0.053 -0.118 -0.147 -0.141 -0.181 -0.312 exp. [0.021]§ [0.031]* [0.042]* [0.063]§ [0.082]§ [0.044]*

5. Add weekly hours -0.036 -0.069 -0.079 -0.054 -0.047 -0.098 worked [0.020] [0.030]§ [0.041] [0.060] [0.078] [0.042]§

Source: Bertrand, Goldin and Katz (2010)

Gender pay gaps within occupation

Source: Goldin (2014)

Non-linear pay and gender pay gap within occupation

Source: Goldin (2014)

Job features and the gender pay gap

Source: Goldin (2014) Women are willing to pay for flexibility

• Recent studies estimating willingness to pay (WTP) for various job attributes (e.g. trading off $ for longer hours worked or availability of part-time option):

– Mas and Pallais (2017): • Applicants to call center jobs randomize into various menus of job attributes • Women, particularly those with young children, have higher WTP for work from home and to avoid employer scheduling discretion

– Wiswall and Zafar (2017): • NYU undergraduates’ hypothetical choices for jobs with different menus of job attributes – More relevant population • Women on average have a higher WTP for jobs with greater work flexibility and job stability, and men have a higher WTP for jobs with higher earnings growth. • Gender differences in preferences explain at least a quarter of the early-career gender wage gap.

Why is inflexible work particularly difficult for women?

• First-order explanation: children

Gender Gap in Labor Supply: The Role of Children

(controls include Pre-MBA characteristics, MBA performance, cohort*year fixed effects)

Dependent Variable Not working Actual post-MBA Log (weekly hours experience worked)

Female 0.084 -0.286 -0.089

[0.009]* [0.039]* [0.013]*

Female with child 0.20 -0.66 -0.238

[0.024]* [0.094]* [0.031]*

Female without child 0.034 -0.126 -0.033

[0.007]* [0.031]* [0.012]*

Source: Bertrand, Goldin and Katz (2010) Estimating child penalties

• Recent wave of papers using rich administrative (register) panel data to compute child penalties: – Angelov et al (2016): Sweden – Kleven and Landais (2016): Denmark

• Approach: document within-couple change in earnings after parenthood

• Identifying assumption is that the decision of when to enter parenthood is not induced by unobservable information of a changed direction of the income and wage trajectory of one of the spouses. – Event-study

• Note: – Overall population – Scandinavia -> longer parental leave; weaker gender norms.

Sweden

Source: Angelov et al (2016)

Denmark

Source: Kleven and Landais (2016)

Denmark

Source: Kleven and Landais (2016)

Denmark Denmark Puzzle?

• Why are women, and especially the more educated ones, today still paying such a disproportionate price in the labor market for carrying a couple’s children?

• True that biological mechanics of having children have remained mainly unchanged

• But several forces seem at first glance to have been operating towards weakening this disproportionate price…

• In particular: – Amount of non-market work – Gender role attitudes

Amount of non-market work

• Technological change and cheaper outsourcing options have reduced need for “double-shift” (labor market work + non-market work) • Especially relevant for the more educated/those with highest earnings potentials

• But important countervailing force has been growing amount of time spent on parenting, especially among the more educated • Particularly relevant in the US • Guryan et al 2008 ; Ramey and Ramey 2009

Gender role attitudes/gender identity

– On the one hand: • Gender role attitudes appear weaker today than in the past, as one would expect if these attitudes are endogenously responding to market changes/new educational landscape

– On the other hand: • (Among historically tracked gender role attitudes) slowest to converge are views regarding conflicts between working mothers and well-being of their children.

• Self-reports (because of demand effects) may mask “true” attitudes

• Some “dormant” gender identity norms may only start biting when women’s position in the labor market improves. – Ex: “Men should earn more than their wives”

• Also, what about other manifestations of gender norms: – “Women should not compete” – “Women should not take too much risk” » Cf domain-specificity; nature vs. nurture

Gender role attitudes

Sample: Women, college educated or more Approve of Mother married Women as working Pre-K kids don't Wife should women suited for doesn't hurt suffer if mother help husbands Birth cohort: working politics as men children works career first 1943 or before 0.88 0.75 0.77 0.60 0.27 1944-1957 0.92 0.88 0.84 0.68 0.09 1958-1973 0.92 0.85 0.83 0.73 0.09 1974 or after 1.00 0.82 0.83 0.78 0.07

Sample: Men, college educated or more

1943 or before 0.84 0.73 0.58 0.39 0.27 1944-1957 0.91 0.80 0.65 0.50 0.14 1958-1973 0.91 0.82 0.67 0.59 0.13 1974 or after 0.90 0.78 0.73 0.73 0.20

Source: GSS, 1972-2014 “Men should earn more than their wives”

Distribution of relative earnings across couples US administrative data

Source: Bertrand, Kamenica, Pan (2015) “Men should earn more than their wives”

• Other results in Bertrand, Kamenica and Pan (2015):

– Marriage rate lower in marriage markets where potential wife likely to earn more than potential husband – Women more likely to be out of the workforce/work fewer hours if their potential earnings exceed that of their husbands. – Couples where the gender identity norm is violated (e.g. wife earns more than husband) associated with: • More marital unhappiness; more divorce • Higher gender gap in housework

Burzstyn et al (2016)

• Questionnaire on job preferences and personality traits filled by newly admitted students to an elite MBA program.

• High stake environment: answers to be used by career services for internship placement.

• Randomly-selected students thought their answers would be shared with classmates (public condition).

• Compare answers for single vs. non-single (women and men) in public vs. private condition. Source: Burzstyn et al (2017) Source: Burzstyn et al (2017) Source: Burzstyn et al (2017) Puzzle?

• Why are women, and especially the more educated ones, today still paying such a disproportionate price in the labor market for carrying a couple’s children?

• Even with declining need for non-market work and weakening gender role norms, other forces might have been pushing in the other directions:

• In particular: – Technological change and job attributes: • How has the labor market valuation of inflexible work attributes changed over time? – Correlation between occupational earnings and elasticity of annual income with respect to hours worked – How has it changed over time?

– Change in marriage rate and spousal characteristics

Occupation-Specific Elasticity of Earnings to Hours Worked Panel A Sample of occupations: All Year Mean Median 1980 0.26 0.25 1990 0.45 0.44 2000 0.56 0.54 2010 0.57 0.63 2015 0.59 0.63 Panel B Sample of Occupations: 100 Highest Paying Occupations Year Mean Median 1980 0.32 0.28 1990 0.54 0.52 2000 0.67 0.69 2010 0.64 0.64 2015 0.66 0.64 Panel C Sample of Occupations: 50 Highest Paying Occupations Year Mean Median 1980 0.31 0.27 1990 0.61 0.66 2000 0.72 0.71 2010 0.67 0.64 2015 0.69 0.72 Panel D Sample of Occupations: 10 Highest Paying Occupations Year Mean Median 1980 0.24 0.41 1990 0.57 0.66 2000 0.62 0.54 2010 0.52 0.78 2015 0.55 0.78 Families of Employed College Educated Women Family Charistic: Married Never Married With Child

Panel A Year = 1980 Women's occupational earnings rank:

Below top 100 0.68 0.19 0.51 Top 100 0.61 0.23 0.42 Top 50 0.58 0.24 0.39 Top 10 0.54 0.30 0.34 Panel B Year = 1990 Women's occupational earnings rank: Below top 100 0.69 0.17 0.54 Top 100 0.64 0.21 0.46 Top 50 0.63 0.22 0.43 Top 10 0.62 0.23 0.43 Panel C Year = 2000 Women's occupational earnings rank: Below top 100 0.67 0.18 0.50 Top 100 0.65 0.20 0.47 Top 50 0.64 0.21 0.46 Top 10 0.64 0.21 0.46 Panel D Year = 2010 Women's occupational earnings rank: Below top 100 0.66 0.19 0.49 Top 100 0.65 0.19 0.47 Top 50 0.65 0.20 0.47 Top 10 0.68 0.19 0.49 Families of Employed College Educated Women (with College Educated Husbands) Husband's occupation Elasticity of earnings has higher elasticiy of Family Charistic: to hours worked in earnings to hours husband's occupation worked than wife's

Panel A Year = 1980 Women's occupational earnings rank:

Below top 100 0.24 0.72 Top 100 0.27 0.46 Top 50 0.28 0.51 Top 10 0.24 0.66 Panel B Year = 1990 Women's occupational earnings rank: Below top 100 0.43 0.74 Top 100 0.48 0.48 Top 50 0.50 0.40 Top 10 0.48 0.52 Families of Employed College Educated Women (with College Educated Husbands)

Husband's occupation Elasticity of earnings has higher elasticiy of Family Charistic: to hours worked in earnings to hours husband's occupation worked than wife's

Panel C Year = 2000

Women's occupational earnings rank: Below top 100 0.55 0.74 Top 100 0.60 0.50 Top 50 0.63 0.45 Top 10 0.60 0.59 Panel D Year = 2010

Women's occupational earnings rank: Below top 100 0.56 0.67 Top 100 0.59 0.52 Top 50 0.61 0.46 Top 10 0.55 0.62 Roadmap for rest of talk

1. Why are women struggling to break the glass ceiling? What does the most recent research say?

– Revisiting education (briefly) – Gender differences in psychological attributes – Women’s greater demand for flexibility/work-family balance – Gender roles and gender identity norms

2. What role can corporate and public policy play, if any, in accelerating convergence at the top?

Augmenting work-family amenities within the workplace • Firm-level HR and public policies aimed at making the workplace more family-friendly, such as: – Longer maternity leave – Part time work, shorter hours; flexibility during the workday – Working remotely • While offering such flexibility may achieve the objective of attracting/ retaining women in the workforce, it will not reduce the gender gap as long as this flexibility is negatively priced in the market. – A matter of job design • Some of these policies may backfire: • For example, longer leave raises costs for employers of hiring women of child-bearing age; may lead employers to not assign women to the most important jobs or clients; may also keep women out of the workforce for “too long” to ensure a re-entry on the fast-track.

Augmenting work-family amenities within the workplace • There might be a tradeoff between reducing whatever is left of the gender gap in labor participation and reducing the gender gap in earnings

• Blau and Kahn, 2013:

– Country-level panel evidence suggests that part (30 percent) of the US plateauing in labor force participation compared to other OECD countries can be accounted for by more aggressive work-family balance policies in non-US OECD

– But higher representation of women in high-paying managerial and professional occupations in US compared to non-US OECD

Gender-neutralizing childcare

• Encouraging more fathers to take up parental leave

– Sweden, Norway, parts of Canada’s experience with dedicated paternity leave into their parental leave policy (“daddy quotas” or “daddy months”) • Duvander and Johansson, 2012; Ekberg et al., 2013; Dahl et al., 2014; Patnaik, 2015

– Evaluation of these policies suggests that: • Fathers do take up their dedicated quota, but not more (so far at least). • Amount of child care provide by fathers go up, even after the child’s first year of life (Schober, 2014) • Effects on women’s labor market outcomes appear muted (so far at least).

Gender-neutralizing childcare

• While such policies go to the core of what is holding women’s back by trying to speed up the shift in gender norms, it is possible that these policies may have perverse effects in the interim.

• Example: “daddy months” in academia

• Antecol, Bedard and Stearns (2016): – Build a dataset on the universe of assistant professor hires at top-50 economics departments from 1985-2004 – Show that the adoption of gender-neutral tenure clock stopping policies substantially reduced female tenure rates while substantially increasing male tenure rates.

Source: Antecol et al (2016) Affirmative action

• The most significant public policy initiative in Europe aimed at increasing gender balance in top corporate jobs has been the introduction of gender quotas on corporate boards

• How are these quotas theoretically supposed to address the glass ceiling? – Mechanical effect (but limited by low numbers of directly affected) – Short-circuiting path dependence in executive appointments – Homophily – Role models

• Additional argument related to the value of diversity… – …which we empirically surprisingly know very little about

• These theoretical benefits are supposed to exceed the potential costs of such policies – Reinforcing stereotypes/self-fulfilling prophecies

Female representation on corporate boards Board quotas • Bertrand et al (2017) evaluate effects of the Norwegian reform (first of its kind) on women’s labor market outcomes – focus on women in business who are most likely to benefit from the reform

• Mixed findings (interim results): – On the one hand, the reform succeeded in reducing gender disparities on corporate boards: • Not just in terms of numbers (mechanical) but also in terms of observable qualifications – Beyond these direct gains for the narrow set of women on boards, not much evidence of spillovers to other women who could have indirectly benefitted: • In particular, female executives at firms with a greater share of women on board (due to the quotas) do not have improved career prospects

• Ultimately, these results are not that surprising in that quotas seem to be a very indirect (at best) response to what is holding women’s back

Engineering more diversity

• Corporate policies also often incorporate some elements of affirmative action, more or less formally:

– E.g. pro-active steps towards achieving minimal female representation on key committees

– Given women’s under-representation, this may imply more committee work for the average woman than the average man

– Example: academia: Misra et (2011) survey 350 faculty members at the University of Massachusetts Amherst in 2008–09. • Three-quarters of women associate professors, compared with half of their male counterparts, had played major service roles. • A third of the women had served as undergraduate directors, compared with 17 percent of the men. • (Women promoted to FT professor at lower rate than men)

Engineering more diversity

• While this may reflect women’s higher willingness to say yes to “non- promotable tasks” and employers’ willingness to exploit this higher willingness… – Vesterlund, Babcock and Weingart, 2017

• A more charitable interpretation is that employers’ primary goal is to achieve greater gender diversity with the hope of positive ripple-down effects for other women within the organization

• While I am not aware of any robust data to document the prevalence of such practices and their impact, the possibility of perverse effects in the interim period is clear: – E.g. Non-linear pay and gender pay gap within occupation

Concluding remarks

• What about labor market discrimination/sexism? – Not arguing, or believing, that it is irrelevant – But strongly arguing that other explanations are quantitatively very relevant – Absent direct testing, discrimination is the “residual”

• Many trends are moving in the “right direction” for women, some very quickly (such as the large and rising reverse gender gap in completed schooling) and some more slowly (such as the declining conservativeness of gender norms).

• It is possible though, as we hinted at above, that changes in the structure of work and job design over the last 40 years may not have not been as beneficial to women.

• How the next wave of technological change in the workplace (i.e. artificial intelligence) will change the structure of work is anyone’s guess.