Content uploaded by Anat Ben-David
Author content
All content in this area was uploaded by Anat Ben-David on Mar 28, 2017
Content may be subject to copyright.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
393
Platform Inequality: Gender in the Gig-Economy
Arianne Renan Barzilay
& Anat Ben-David†
Americans are making extra money renting out a spare room,
designing websites, selling products they design themselves at home,
or even driving their own car. This ‘on demand’ or so-called ‘gig
economy’ is creating exciting opportunities and unleashing
innovation but it’s also raising hard questions about workplace
protections and what a good job will look like in the future.
–Hillary Rodham Clinton1
ABSTRACT
Laboring in the new economy has recently drawn tremendous social, legal,
and political debate. The changes created by platform-facilitated labor are
considered fundamental challenges to the future of work and are generating
contestation regarding the proper classification of laborers as employees or
independent contractors. Yet, despite this growing debate, attention to gender
dimensions of such laboring is currently lacking. This Article considers the
gendered promises and challenges that are associated with platform-facilitated
labor, and provides an innovative empirical analysis of gender discrepancies in
such labor; it conducts a case study of platform-facilitated labor using
Assistant Professor, University of Haifa Faculty of Law.
† Senior Lecturer, Department of Sociology, Political Science and Communication, the
Open University of Israel.
We are grateful to Einat Albin, Naomi Cahn, June Carbone, Jessica Clarke, Efrat
Daskal, Guy Davidov, Yossi Dahan, Debbie Dinner, Eldar Haber, Dafna Hacker, Yoram
Kalman, Laura Kessler, Shelly Kreiczer-Levy, Lilach Lurie, Faina Milman-Sivan, Sagit
Mor, Guy Mundlak, Orna Rabinovich-Einy, Amnon Riechman, Noya Rimalt, Betsy
Rosenblatt, Sharon Shakargy, Adam Shinar, and Oren Soffer for helpful suggestions
and valuable feedback, and to Niva Elkin-Koren for her generous support of this
research project. Thanks to Adam Amram for programming and data analysis
assistance and to Ido Porat and Ofer Toledano for research assistance. Thanks to
Benjamin Heller, Beata Safari, Christopher Mazza and the editors of the Seton Hall Law
Review for wonderful editorial assistance. This research was supported by the I-CORE
Program of the Planning and Budgeting Committee and The Israel Science
Foundation (1716/12).
1 Christina Reynolds, Reality Check: Hillary Clinton and the Sharing Economy,
HILLARYCLINTON.COM,
https://www.hillaryclinton.com/briefing/updates/2015/07/16/reality-check-
sharing-economy (last visited Dec. 17, 2016).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
394 SETON HALL LAW REVIEW [Vol. 47:393
computational methods that capture some of the gendered interactions hosted by
a digital platform. These empirical findings demonstrate that although women
work for more hours on the platform, women’s average hourly rates are
significantly lower than men’s, averaging about 2/3 (two-thirds) of men’s rates.
Such gaps in hourly rates persist even after controlling for feedback score,
experience, occupational category, hours of work, and educational attainment.
These findings suggest we are witnessing the remaking of women into devalued
workers. They point to the new ways in which sex inequality is occurring in
platform-facilitated labor. They suggest that we are beholding a third generation
of sex inequality, termed “Discrimination 3.0,” in which discrimination is no
longer merely a function of formal barriers or even implicit biases. The Article
sketches Equality-by-Design (EbD) as a possible direction for future redress,
through the enlisting of platform technology to enhance gender parity. In sum,
this Article provides an empirical base and analysis for understanding the new
ways sex inequality is taking hold in platform-facilitated labor.
INTRODUCTION ............................................................................ 394
I. OPPORTUNITIES AND CHALLENGES OF BALKANIZED LABOR .... 399
II. EMPIRICAL FINDINGS ............................................................... 403
A. Method ......................................................................... 405
B. Findings ........................................................................ 408
C. Discussion of Findings ................................................. 420
PART III. THE INEPTITUDE OF CURRENT LEGAL NORMS ............. 422
A. Employee Status ........................................................... 423
B. Antidiscrimination ....................................................... 423
PART IV. FROM DISCRIMINATION 3.0 TOWARDS EQUALITY-BY-DESIGN
........................................................................................... 427
A. A Third Generation of Discrimination ....................... 427
B. Towards A Platform for Equality? ............................... 429
CONCLUSION................................................................................ 431
INTRODUCTION
Flexibilization, globalization and privatization have presented
challenges for employment law for some time now.2 Sociologists and
legal scholars have documented and critiqued the precarious nature
2 See generally Katherine V.W. Stone, Flexibilization, Globalization and Privatization:
Three Challenges to Labour Rights in Our Time, 44 OSGOODE HALL L.J. 77, 77 (2006)
(noting that Flexibilization “refers to the changing work practices by which firms no
longer use internal labour markets or implicitly promise employees lifetime job
security, but rather seek flexible employment relations that permit them to increase
or diminish their workforce, and reassign and redeploy employees with ease”).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 395
and gendered implications of the work these forces have created.3 We
know that precarious work—work that departs from the model of a
full-time, year-round employment relationship with a single
employer—has historically been conducted mostly by women.4 In the
past few years, however, precarious work has expanded in magnitude,
scope, and trendiness for both men and women, professionals and
non-professionals.5 With the rise of the “sharing” economy, new
companies are using technology to initiate connections between
workers offering ad hoc labor and third parties in need of tasks
performed.6 The “sharing” or “gig” economy is generating widespread
conversation among academics, lawyers, and policy makers,7 even
3 Precarious work, characterized by low wages and the absence of job security, is
associated with part time employment, temp work, on-call work, home working, and
telecommuting. KATHERINE V.W. STONE, FROM WIDGETS TO DIGITS: EMPLOYMENT
REGULATION FOR THE CHANGING WORKPLACE 69–86 (2004); Judy Fudge & Rosemary
Owens, Precarious Work, Women, and the New Economy: The Challenge to Legal Norms, in
PRECARIOUS WORK, WOMEN, AND THE NEW ECONOMY: THE CHALLENGES TO LEGAL NORMS
3, 8, 12–13 (Judy Fudge & Rosemary Owens eds., 2006) (noting that this work is
performed largely by women). See generally ERIN HATTON, THE TEMP ECONOMY: FROM
KELLY GIRLS TO PERMATEMPS IN POSTWAR AMERICA (2011) (illustrating how the temp
industry transformed work in America). For the effects of economic inequality on
families, see generally JUNE CARBONE & NAOMI R. CAHN, MARRIAGE MARKETS: HOW
INEQUALITY IS REMAKING THE AMERICAN FAMILY (2014).
4 Fudge & Owens, supra note 3, at 8 (noting that flexible forms of labor, casual
labor, contract labor and outsourcing are associated primarily with women). For the
historical development of women’s labor, see generally ALICE KESSLER-HARRIS, OUT TO
WORK: A HISTORY OF WAGE EARNING WOMEN IN THE UNITED STATES 30, 36 (2003) (noting
part time and alternating jobs as historically common for women wage earners).
5 See Tamara Kneese et al., Understanding Fair Labor Practices in a Networked Age
(Data & Soc’y Research Inst. Working Paper, 2014), http://www.datasociety.net/pubs
/fow/FairLabor.pdf (indicating an increase in part-time, independent, contract,
freelance modes of labor and noting the “coolness” of individual risk); Orly Lobel, The
Gig Economy and the Future of Employment and Labor Law, U.S.F. L. REV. 2 (forthcoming),
[hereinafter Lobel, The Gig Economy], https://papers.ssrn.com/sol3/papers.cfm?
abstract_id=2848456 (discussing how new digital platform companies are disrupting
established markets).
6 See Matthew W. Finkin, Beclouded Work, Beclouded Workers in Historical Perspective,
37 COMP. LAB. L. & POL’Y J. 603 (2016) (describing the gig economy, and placing it in
historical perspective). See also Megan Carboni, A New Class of Worker for The Sharing
Economy, 22 RICH. J.L. & TECH. 1 (2016). See generally Kneese et al., supra note 5 (noting
that technology is central to the sharing economy flexible work patterns that enable
task-work and that websites connect individuals to customers who want specific tasks
performed).
7 See Thomas E. Perez, Sec’y of Labor, Remarks at the Department of Labor
Future of Work Symposium (Dec. 10, 2015), https://www.dol.gov/newsroom/
speech/20151210. See also Kneese et al., supra note 5; Benjamin Sachs, Uber: Employee
Status and “Flexibility”, ON LABOR (Sept. 25, 2015),
https://onlabor.org/2015/09/25/uber-employee-status-and-flexibility; Noah Zatz, Is
Uber Wagging the Dog With Its Moonlighting Drivers?, ON LABOR (Feb. 1, 2016),
https://onlabor.org/2016/02/01/is-uber-wagging-the-dog-with-its-moonlighting-
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
396 SETON HALL LAW REVIEW [Vol. 47:393
permeating the recent presidential race,8 and is likely to growingly
preoccupy law and policy in coming years. While there is no clear
definition of this economy, for our purposes, the Article will
characterize it by the disaggregation of consumption and the
segmentation of production via online platforms.9 Indeed,
fragmented, loose, informal, task-based forms of labor have been
amplified worldwide. We now see micro-labor on a macro scale—so
much so that some have claimed we are witnessing a “paradigmatic
shift” in the way we work.10
“Sharing” economy companies, at first benignly dubbed “peer to
peer” and marketed as fresh, innovative, and “collaborative,”11 have
mushroomed in popularity and scale. The “sharing” economy now
provides a wide and ever-broadening range of services, from driving,
to running errands, to professional tasks.12 While “sharing” economy
firms vary somewhat in the amount of control they exert over laborers,
their specific discursive terminology of laborers,13 the platforms they
provide, and the fees they collect, they all: (a) use the Internet; and
(b) endorse the ability of laborers to earn money, often from the
vicinity of one’s home and, importantly, on one’s own schedule. These
companies are increasingly criticized for ushering in with full force an
“on-demand,” “on-call,” “gig”–based economy, and for selling us uber-
capitalism under the guise of sharing rhetoric.14
drivers.
8 See Reynolds, supra note 1.
9 See Daniel E. Rauch & David Schleicher, Like Uber, But for Local Government Policy:
The Future of Local Regulation of the “Sharing Economy” 8 (George Mason Univ. L. & Econ.
Research Paper Series, No. 15-01, 2015), http://papers.ssrn.com/sol3/papers.cfm?
abstract_id=2549919 (explaining the disaggregation of consumption). For general
fissured trends of employment, see also DAVID WEIL, THE FISSURED WORKPLACE: WHY
WORK BECAME SO BAD FOR SO MANY AND WHAT CAN BE DONE TO IMPROVE IT (2014).
10 See Orly Lobel, The Law of the Platform, 101 MINN. L. REV. 87 (2016). See also Mary
L. Gray, Your Job Is About to Get ‘Taskified’, L.A. TIMES (Jan. 8, 2016, 6:52 PM),
http://www.latimes.com/opinion/op-ed/la-oe-0110-digital-turk-work-20160110-
story.html (instead of hiring employees, firms can now post tasks on the web thus
fragmenting jobs; such “online piecework” represents a “radical shift in how we define
employment itself”).
11 RACHEL BOTSMAN & ROO RODGERS, WHAT’S MINE IS YOURS: THE RISE OF
COLLABORATIVE CONSUMPTION xiv–xv (2010).
12 Jeremias Prassl & Martin Risaktt, Uber, Taskrabbit, and Co.: Platforms as Employers?
Rethinking the Legal Analysis of Crowdwork, 37 COMP. LAB. L. & POL’Y J. 619, 622 (2016).
13 For example, Uber calls its laborers drivers, “partners,” and “independent
contractors.” See Partners, UBER.COM, https://partners.uber.com/join (last visited July
23, 2016). Taskrabbit calls its laborers “taskers.” See T
ASKRABBIT,
https://www.taskrabbit.com (last visited July 23, 2016).
14 See, e.g., TOM SLEE, WHAT’S YOURS IS MINE: AGAINST THE SHARING ECONOMY
(2015).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 397
Technological advances and changes in social and economic
organization often present moments of opportunity and challenge.15
Newly technologized contingent work practices are already creating
tremendous social change.16 Increasingly, such work practices are
remapping the frontier between home and work, public and private,
employment and contracting.17 They are also testing the boundaries
of legal responsibility.18 Employment law scholarship has begun to pay
attention to this phenomenon, focusing primarily on whether taskers
should be classified as employees or independent contractors.19 Still,
little is known about the gender dimensions of platform-facilitated
labor.20 We should start filling in this void and thinking through the
gendered implications and legal ramifications of laboring in the new,
“sharing” economy. This Article begins that process.
Part I considers the gendered promises and challenges associated
with platform-facilitated labor. On its face, platform-facilitated labor
has potential to enhance gender equality because laborers may
sometimes enjoy a degree of anonymity and inclusiveness when
offering work via platform, and a substantial degree of flexibility which
15 See YOCHAI BENKLER, THE WEALTH OF NETWORKS: HOW SOCIAL PRODUCTION
TRANSFORMS MARKETS AND FREEDOM 2 (2006).
16 Lobel, The Gig Economy, supra note 5, at 3.
17 See id. at 3–7; Naomi Schoenbaum, Gender and the Sharing Economy, FORDHAM
URBAN L.J. 1, 5 (forthcoming), https://papers.ssrn.com/sol3/papers.cfm?abstract
_id=2865710.
18 See, e.g., Order-Denying-Plaintiffs-Motion-for-Preliminary Approval of Settlement,
UNITED STATES DISTRICT CT., http://www.cand.uscourts.gov/EMC/OConnorvUber
Technologies (last visited Jan. 2, 2017) (Uber argues that because it sets minimal
controls over drivers’ hours, they are not Uber employees, thus testing the boundaries
of its legal responsibilities).
19 See, e.g., Keith Cunningham-Parmeter, From Amazon to Uber: Defining Employment
in the Modern Economy, 96 B.U. L. REV. 1637 (2016); Valerio De Stefano, The Rise of the
“Just-in-Time Workforce”: On-Demand Work, Crowd Work and Labor Protection in the “Gig-
Economy”, 37 COMP. LAB. L. & POL’Y J. 471, 471 (2016); Veena Dubal, Wage Slave or
Entrepreneur?: Contesting the Dualism of Legal Worker Identities, 105 CAL. L. REV.
(forthcoming 2017), http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2796728;
Benjamin Means & Joseph A. Seiner, Navigating the Uber Economy, 49 U.C. DAVIS L. REV.
1511, 1511 (2016); Brishen Rogers, Employment Rights in the Platform Economy: Getting
Back to Basics, 10 HARV. L. & POL’Y REV. 479, 479 (2016). Brishen Rogers, The Social Costs
of Uber, 82 U. CHI. L. REV. DIALOGUE 85 (2015). See also Guy Davidov, The Status of Uber
Drivers, ON LABOR (May 17, 2016), https://onlabor.org/2016/05/17/guest-post-the-
status-of-uber-drivers-part-1-some-preliminary-questions; Sachs, supra note 7
(considering whether flexibility enjoyed by workers can determine employee or
independent contractor status).
20 See Schoenbaum, supra note 17. Schoenbaum has claimed that the
pervasiveness of intimacy in services such as those provided through Uber and Airbnb
may prime sex stereotypes. On race and the sharing economy, see generally Nancy
Leong, New Economy, Old Biases, 100 MINN. L. REV. 2153, 2153 (2016).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
398 SETON HALL LAW REVIEW [Vol. 47:393
may be helpful for those with gendered familial responsibilities.21 On
the other hand, such work may actually be hindering laborers, since it
requires far less investment in workers, and offers fewer opportunities
for workers to establish close relationships with work-providers.
Moreover, at present such work provides a paucity of benefits for
laborers, and a dearth of protections against discrimination.
While the promises and challenges posed to gender equality by
this form of labor are considerable, there is an acute shortage of data
analyzing online platform work from a gender perspective. Part II
empirically examines how women are doing in this growing economy.
Through an empirical case-study focusing on workers on one global
platform (“the Platform”), it examines gender discrepancies in
platform-facilitated online labor. The Article employs a computational
approach that automatically extracts profile data from the Platform’s
Application Programming Interface (API).22 Rather than relying on
answers or data reported by users through surveys, extracting data
directly from the Platform’s API enables us to capture a snapshot of
the actual user profiles that are in use on the Platform. The
application of computational, unobtrusive methods is tailored to
capture the unique digital aspects of the “gig” economy by providing a
snapshot of the actual digital interactions that the Platform hosts, as
they are shaped by its technological affordances, and as they are made
available through the Platform’s API. The study analyzes over 4,600
online taskers’ requested rates, occupations, and work-hours. Using
statistical analysis, its findings illustrate a dramatic gender gap in the
hourly rate requested by men and women who are seeking work
through the studied platform. The findings show women’s average
hourly requested rates are 37% lower than men’s. Such gaps in hourly
requested rates persist even after controlling for feedback score,
experience, occupational category, hours of work, and educational
attainment. Surprisingly, among the different occupational categories
available on the Platform, the most pervasive gender gaps were found
with regard to those offering legal services.
21 Lack of anonymity has been suggested to prime sex stereotypes, see
Schoenbaum, supra note 17. Flexibility may prove helpful for those with caring
responsibilities. JOAN C. WILLIAMS, UNBENDING GENDER: WHY FAMILY AND WORK
CONFLICT AND WHAT TO DO ABOUT IT 84–86(2000). Family care still tends to be
gendered. See Naomi R. Cahn, The Power of Caretaking, 12 YALE J.L. & FEMINISM 177, 188
(2000).
22 Permission to extract the data was obtained through the Platform. See E-mail
from API Support Team, to Adam Amram (Mar. 3, 2016, 9:25 AM) (on file with
authors). For more on the data collection and for the benefits and limitations of this
kind of methodological approach, see infra Part II.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 399
Given these dramatic empirical findings, Part III begins to
consider the legal implications for working in platform-facilitated
online labor. It argues that to realize the promises of platform-
facilitated labor for gender equality we must: (a) make such labor a
sustainable work option for those heavily involved; and (b) mitigate
the perils faced disproportionately by women. It posits the ineptitude
of current legal norms to do both. Part IV suggests that we are
witnessing a third generation of sex-inequality, which we term
“Discrimination 3.0” in which sex discrimination may be occurring on
platforms, and in which platforms may (likely unconsciously) be
harboring gender inequality. It illuminates challenges that platform
inequality poses for antidiscrimination law more generally, and calls
for contemplating new mechanisms for promoting work equality. It
suggests that we could use platform technology itself to promote
Equality-By-Design (EbD) as a mechanism towards enhancing gender
parity in platform-facilitated labor.
I. OPPORTUNITIES AND CHALLENGES OF BALKANIZED LABOR
The “sharing” economy has been celebrated as a job creator, a
liberating option for those unable to attain stable employment, and as
providing freedom and flexibility.23 By some estimates, more than 22%
of U.S. adults (approximately 45 million people) have already offered
their labor and services in the “sharing” economy,24 with numbers likely
to grow.25 This Part outlines promises and pitfalls associated with this
23 See Natasha Singer, In the Sharing Economy, Workers Find Both Freedom and
Uncertainty, N.Y. TIMES (Aug. 16, 2014), http://www.nytimes.com/2014/08/17/
technology/in-the-sharing-economy-workers-find-both-freedom-and-uncertainty.
html. See also Paul Merrion & Fareeha Ali, Making Inroads: Women Cabbies on the Rise,
CHI. BUS. (Sept. 27, 2014), http://www.chicagobusiness.com/article/20140927/ISSU
E01/309279976/making-inroads-women-cabbies-on-the-rise.
24 Katy Steinmetz, Exclusive: See How Big the Gig Economy Really Is, TIME (Jan. 6,
2016), http://time.com/4169532/sharing-economy-poll. But other estimates found
that only 4 percent of the adult population had ever participated in the online
platform economy. Paychecks, Paydays, and the Online Platform Economy: Big Data on
Income Volatility, JPMORGAN CHASE & CO. 8–9, 21 (Feb. 2016),
https://www.jpmorganchase.com/corporate/institute/document/jpmc-institute-
volatility-2-report.pdf. Other research suggests workers on online-platforms comprise
a small but rapidly growing share of the economy. Lawrence F. Katz & Alan B. Krueger,
The Rise and Nature of Alternative Work Arrangements in the United States, 1995-2015,
SCHOLARS AT HARVARD (2016), http://scholar.harvard.edu/files/lkatz/files/katz_
krueger_cws_v3.pdf. See also The Online Platform Economy: What is the growth trajectory?,
JPMORGAN CHASE & CO. (Feb. 18, 2016), https://www.jpmorganchase.com/
corporate/institute/insight-online-platform-econ-growth-trajectory.htm.
25 See Molly Cohen & Arun Sundararajan, Self-Regulation and Innovation in the Peer-
to-Peer Sharing Economy, 82 U. CHI. L. REV. DIALOGUE 116, 116 (2015).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
400 SETON HALL LAW REVIEW [Vol. 47:393
new form of labor for women’s economic equality and security.
On the one hand, the new online, free-access “sharing” economy
form of laboring shows great promise for enhancing women’s
economic equality through participation in the online workforce for
two main reasons. The first is that, at least in some cases, laborers enjoy
a greater degree of anonymity and potential inclusiveness when
offering services online, which could offset bias, barriers, and
discrimination still faced by women in the general workforce. This may
be especially true if anonymity and gender-blindness are preserved in
online platforms, which is often not the case; but it may potentially also
be the case if anonymity is not preserved, given the gender
discrepancies in pay, promotion, and opportunities the workplace has
long exerted.26 Given that some “sharing” economy work is based
online rather than in-person, and is horizontal rather than hierarchal,
women may also find it easier to negotiate for equal pay. After all, the
income generated by the same online task should not be affected by
the laborer’s gender.
The second reason for optimism is that in most cases, laborers in
the “sharing” economy enjoy a substantial degree of flexibility in
setting their work schedules.27 That flexibility is especially important
26 For the persistent existence of the wage gap, see generally Fact Sheet: The Wage
Gap Is Stagnant in the Last Decade, NAT’L WOMEN’S L. CTR. (Sept. 2013),
http://www.nwlc.org/sites/default/files/pdfs/wage_gap_is_stagnant_2013_2.pdf.
The National Women’s Law Center data is only on full–time earners. The wage gap is
even more severe for the many women who are relegated to part-time, temporary,
contingent work. See Labor Force Statistics from the Current Population Survey, UNITED
STATES DEP’T OF LABOR, http://www.bls.gov/cps/aa2014/cpsaat37.htm (last visited
Jan. 2, 2017) (revealing that women’s median wages for full-time, year-round work
were 82% of their male counterparts’); BUREAU LABOR STATISTICS, U.S. DEP’T OF LAB.,
HIGHLIGHTS OF WOMEN’S EARNINGS IN 2008, at 1–2 (2009),
http://www.bls.gov/opub/reports/womens-
earnings/archive/womensearnings_2008.pdf (showing occupational segregation and
generally lower earnings for women than men); On Pay Gap, Millennial Women Near
Parity—For Now, PEW RESEARCH CTR. (Dec. 11, 2013),
http://www.pewsocialtrends.org/2013/12/11/on-pay-gap-millennial-women-near-
parity-for-now (showing young women are making progress and starting their working
lives earning nearly the same as young men). See also D
EBORAH L. RHODE, WHAT
WOMEN WANT: AN AGENDA FOR THE WOMEN’S MOVEMENT 7, 25–38 (2014) (discussing a
persistent gender gap in leadership); Christianne Corbett & Catherine Hill, Graduating
to a Pay Gap: The Earnings of Women and Men One Year after College Graduation, AM. ASS’N
OF UNIV. WOMEN (Oct. 2012), http://www.aauw.org/files/2013/02/graduating-to-a-
pay-gap-the-earnings-of-women-and-men-one-year-after-college-graduation.pdf
(reporting that women earn less than men already one year after graduation, across
different occupations). See generally MARIA CHARLES & DAVID GRUSKY, OCCUPATIONAL
GHETTOS: THE WORLDWIDE SEGREGATION OF WOMEN AND MEN (2004) (reporting that
men and women still work in significantly segregated occupations).
27 Carboni, supra note 6; Drive with Uber: Earn money on your schedule, UBER,
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 401
for working caregivers, who are still predominately women.28 Indeed,
it seems like “sharing” economy companies are aiming to attract
women using precisely these rationales: these companies market
themselves as empowering women by providing them with the
flexibility they need to balance work with family and other gendered
responsibilities.29 Work and family both carry great significance in
most people’s lives, so providing shorter work hours and flexible
schedules for both men and women may potentially prove beneficial
in the search for work-family balance.30 On its face, then, the
mushrooming of the “sharing” economy might be seen as a positive
force for women’s empowerment and equality. After all, the idea is
that in the “sharing” economy a person is the designer of her labor.
With no boss to tell her when to work, or which assignments to take
on, she is the architect of her work life. Additionally, the flexibility of
the “sharing” economy offers to both men and women the promise of
gainful employment alongside family-care, potentially even changing
the normalized gendered roles of caretaking and breadwinning.31
However, the view of cyberspace as an ideal realm where all can
participate equally, free from historical, social, and physical restraints
has already been critiqued as utopian.32 The picture does indeed look
https://get.uber.com/drive/ (last visited Dec. 17, 2016).
28 See Kathryn Abrams, Gender Discrimination and the Transformation of Workplace
Norms, 42 VAND. L. REV. 1183, 1195, 1235 (1989); Arianne Renan Barzilay, Parenting
Title VII: Rethinking the History of the Sex Discrimination Prohibition, 28 YALE J.L. & FEMINISM
55, 100 (2016).
29 For example, Uber has stated:
[F]reedom is helping (literally) drive another wave of women’s
empowerment: the opportunity to fit work around life, rather than the
other way around. Around 20 million Americans work fewer hours than
they would like for “non-economic reasons,” according to the Bureau of
Labor Statistics. These include personal commitments, in particular
child care, that can make full-time jobs so difficult. . . . It’s one of the
reasons Uber last year announced a commitment to get one million
women drivers using our app by 2020. Because driving a car isn’t just a
way to get to work—it can be the work. For women around the world,
Uber offers something unique: work on demand, whenever you want it.
Drivers can make money on their own terms and set their own schedules.
Blaire Mattson, This International Women’s Day, Women Take the Wheel, UBER NEWSROOM
(Mar. 7, 2016), https://newsroom.uber.com/driven-women.
30 See Arianne Renan Barzilay, Back to the Future: Introducing Constructive Feminism
for the Twenty-First Century—A New Paradigm for the Family and Medical Leave Act, 6 HARV.
L. & POL’Y REV. 407, 432–35 (2012).
31 See JOAN C. WILLIAMS, RESHAPING THE WORK-FAMILY DEBATE, WHY MEN AND CLASS
MATTER 2 (2010) (discussing how workplace norms pressure men into breadwinning
roles and women out of them). For the gendered roles of caretaking and
breadwinning, see Cahn, supra note 21, at 188, 191, 200–01, 214.
32 See Mary Anne Franks, Unwilling Avatars: Idealism and Discrimination in Cyberspace,
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
402 SETON HALL LAW REVIEW [Vol. 47:393
far more complicated, in the context of platform-facilitated, on-
demand labor. Companies like Uber treat drivers as contractors rather
than employers, thereby avoiding worker protections such as overtime,
minimum wage, family leave, and unemployment insurance.33 Of
course, freelancing, temp work, and telecommuting have been around
for a long time, but the “sharing” economy’s rapid growth in recent
years has given rise to a growing online service economy that applies
the contract-worker model across various sectors. This trend has a
dramatic influence on the workforce and the organization of
employment, as the volume, ease, and scope of online precarious labor
is increasing.34 According to some evaluations, by 2020, 40% of
American workers will be working as independent contractors,35 likely
making platform-facilitated labor even more popular. These so-called
contracting, freelancing, or “tasking” work models require far less
investment in workers, offer fewer opportunities for workers to
establish relationships with employers, and provide fewer benefits and
a paucity of protections against discrimination than do long-term or
full-time employment models.36 Additionally, the sheer number of
online taskers competing for a given task may encourage the lowering
of bidding rates.37 Along with information gaps about the actual work
a given task entails, the pressure to lower one’s price may generate
exploitative work practices.38
20 COLUM. J. GENDER & L. 224, 225 (2011).
33 Deepa Das Acevedo, Regulating Employment Relationships in The Sharing Economy,
20 EMP. RTS. & EMP. POL’Y J. 1, 2 (2016) (noting that sharing economy work often
entails no benefits). See Finkin, supra note 6, at 611, 615 (explaining that when workers
are not considered “employees” the purchaser of their labor need not bear benefits
such as minimum wage, or family leave); see also Carboni, supra note 6; Henry Ross,
Ridesharing’s House of Cards: O’Connor v. Uber Technologies, Inc. and The Viability of
Uber’s Labor Model in Washington, 90 WASH. L. REV. 1431, 1431 (2015).
34 See Lobel, supra note 10, at 1; The Online Platform Economy: What is the growth
trajectory?, JPMORGAN CHASE & CO. (Feb. 2016), https://www.jpmorganchase.com/
corporate/institute/insight-online-platform-econ-growth-trajectory.htm.
35 Joanna Penn & John Wihbey, Uber, Airbnb and Consequences of the Sharing Economy:
Research Roundup, JOURNALIST’S RESOURCE, http://journalistsresource.org/studies/
economics/business/airbnb-lyft-uber-bike-share-sharing-economy-research-
roundup#sthash.XMg2yvqU.dpuf (last updated June 3, 2016). See also Cunningham-
Parmeter, supra note 19, at 4 (citing INTUIT, INTUIT 2020 REPORT: TWENTY TRENDS THAT
WILL SHAPE THE NEXT DECADE 20 (2010)).
36 See Vicki Schultz, Feminism and Workplace Flexibility, 42 CONN. L. REV. 1203 (2010);
Michelle A. Travis, Equality in the Virtual Workplace, 24 BERKELEY J. EMP. & LAB. L. 283
(2003).
37 See Finkin, supra note 6, at 617 (explaining how global competition may erode
wages).
38 Brad Stone, My Life as a Taskrabbit: A Short Career in the Distributed Workforce,
BLOOMBERG (Sept. 13, 2012), http://www.bloomberg.com/news/articles/2012-09-
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 403
One can argue that these work practices are actually preventing
economic sustainability and equality for women, and that the “sharing”
economy, far from helping laborers overcome the work-family conflict,
may be worsening it by reducing the human subject to a mere
commodity. Because detached and fragmented labor places the ideal
of stable employment and self-sufficiency beyond the reach of many
laborers, thereby requiring them to work for more hours to make ends
meet, it may create numerous risks for workers and families.39 This
form of labor may carry additional specific, gendered risks for
caregivers. For example, arranging, scheduling, and providing
childcare when one is an on-call worker makes juggling work and
family even more difficult to sustain.40 Furthermore, some have
claimed that the “sharing” economy heightens the salience of sex
(because of the intimacy associated with some transactions and the
accessibility to photographs and names online), which primes sex-
stereotypes, often considered harmful for gender equality.41
II. EMPIRICAL FINDINGS
Given the considerable theoretical benefits and challenges
illustrated above, it is especially important to begin to empirically
examine how women are actually faring in the online “sharing”
economy. Women comprise a substantial share of “sharing” economy
laborers.42 Women still do the lion’s share of familial caregiving, while
most lucrative jobs are constructed for workers free from such
responsibilities.43 The attraction of flexible schedules, combined with
women’s second-class status in the workplace generally,44 may make
women especially susceptible to the lure of fragmented tasking
services.45 But to what degree are these new forms of work reorganizing
13/my-life-as-a-taskrabbit. For a more optimistic assessment of working in the sharing
economy, see Lobel, supra note 10.
39 See Carboni, supra note 6.
40 See Miriam A. Cherry, A Taxonomy of Virtual Work, 45 GA. L. REV. 951 (2011). See
Angela P. Harris, Theorizing Class, Gender, and the Law: Three Approaches, 72 L. &
CONTEMP. PROBS. 37, 44–51 (2009) (stating that the gender divide is fundamental to
economic production).
41 Schoenbaum, supra note 17.
42 Katz & Krueger, supra note 24, at 11–12 (observing a “notable rise in the share
of workers in alternative work arrangements that are women”). See also Steinmetz,
supra note 24.
43 See WILLIAMS, supra note 21, at 14–19.
44 Abrams, supra note 28, at 1191, 1196 (noting that women have been
disadvantaged as workers by the fact that central features of the workplace have been
constructed by and for men). See supra note 26.
45 See Laura T. Kessler, The Attachment Gap: Employment Discrimination Law, Women’s
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
404 SETON HALL LAW REVIEW [Vol. 47:393
around the same gendered lines? Are the new platforms enhancing
women’s opportunities in important and substantial ways, or are they
merely replicating “old” economy gender inequality?
Social science data has long pointed to the persistent existence of
a wage and leadership gap and of occupational segregation in the
workplace.46 The Internet has been critiqued as a place in which sexual
harassment has dramatic discriminatory effect on women, and in
which women have had to silence their potentially rewarding online
presences due to cyber harassment.47 Some attention has recently been
paid to the “sharing” economy’s possible effects on underprivileged
groups like racial minorities,48 and new research has shown that blacks
are discriminated against on Airbnb.49 Some have found that customer
satisfaction ratings (important tools for users on various platforms)
overwhelmingly favored men over women.50 Women have been found
to receive less money than men when selling the same merchandise on
eBay.51 But how is gender playing out in the unregulated online
“sharing” economy labor context? Do women still earn less than men,
even online? Are the age-old maladies confronted by women in the
“old” economy morphed through technology in the new one?
There is an acute shortage of data analyzing online platform work
from a gender perspective. We investigate a global online platform
that connects work-seekers of various occupations with online tasks to
be performed. We use data from the studied platform as a case study
through which to examine gendered dimensions of work in the “gig”
economy. On the studied platform, people can register either as work-
seekers or as potential work-providers. Work-seekers create profiles in
which they provide information about the services they perform, their
Cultural Caregiving and the Limits of Economic and Liberal Legal Theory, 34 U. MICH. J.L.
REFORM 371 (2001) (arguing that a lack of parental leave policies creates an
“attachment gap” for women in the workforce).
46 See supra note 26 and accompanying text.
47 Franks, supra note 32.
48 Nancy Leong, The Sharing Economy Has a Race Problem, SALON (Nov. 2, 2014),
http://www.salon.com/2014/11/02/the_sharing_economy_has_a_race_problem.
49 Benjamin G. Edelman & Michael Luca, Digital Discrimination: The Case of
Airbnb.com (Harvard Bus. Sch., Working Paper No. 14-054, 2014),
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2377353; Benjamin Edelman,
Michal Luca & Dan Svirsky, Racial Discrimination in the Sharing Economy: Evidence From a
Field Experiment (Harvard Bus. Sch., Working Paper No. 16-069, 2016),
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2701902.
50 Larry Kim, Gender Bias in Online Marketing: Data Shows Women Are Undervalued by
21%, WORDSTREAM (Apr. 3, 2015), http://www.wordstream.com/blog/ws/2014/05/
13/gender-bias.
51 Tamar Kricheli-Katz & Tali Regev, How Many Cents on the Dollar? Women and Men
in Product Markets, 2 SCI. ADVANCES 1 (2016).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 405
skills, and their requested “hourly rate.”52 There is no box to check on
the profile page for the gender of work-seekers, but names are
required and photographs are commonly used as profile pictures.53
Potential work-providers post jobs to which work-seekers apply. After
the potential work-provider reviews the profiles of those who have
applied, they may contact those who seem the most fitting to complete
the online task.54 After an interview, which usually takes place outside
of the Platform (such as via email or phone), the work-provider sends
an offer to the work-seeker and when the work-seeker accepts the
terms, a contract is signed.55 After the task is completed, the Platform
transfers pay via an escrow account. The work-provider then rates the
performance of the work-seeker, which appears on her profile as a
“feedback score.”56 The Platform collects its fee as a percentage of
every transaction.57
A. Method
Traditionally, studies on the gender pay gap primarily rely on data
obtained from surveys, coupled with demographic data.58 By contrast,
in this study we undertake a computational approach to measuring
gendered dimensions of working on the Platform by extracting profile
data from its Application Programming Interface (API). Rather than
relying on surveys or questionnaires that are based on answers or data
reported by users, extracting the data from the Platform’s API enables
us to capture a snapshot of the actual user profiles that are in use on
the Platform. Since we aim to examine the gendered dimensions of
working via an online platform, and because the “gig” economy is
operating through platforms, the application of computational,
unobtrusive methods is tailored to capture the unique digital aspects
of the “gig” economy, by providing a snapshot of the actual digital
interactions that the Platform hosts, as they are shaped by its
technological affordances,59 and as they are made available through
52 This information appears on the Platform’s website (link on file with authors).
53 Id.
54 Id.
55 Id.
56 Id.
57 Id.
58 See supra note 26.
59 The term “technological affordances” relates to the ways with which technology
shapes sociability. It examines the ways humans (users), perceive objects as
possibilities for potential actions and act upon them. The term is widely used by
scholars of social media platforms. See, e.g., Ester Weltevrede & Erik Borra, Platform
Affordances and Data Practices: The Value of Dispute on Wikipedia, 3 BIG DATA & SOC’Y 1
(2016).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
406 SETON HALL LAW REVIEW [Vol. 47:393
the Platform’s API.60
As a first step, we collected data. After gaining permission from
the Platform to access their API, we built a python script to
automatically extract a sample of users’ profile data according to the
following parameters: 1) profiles from the U.S.; 2) profiles of private
people (as opposed to agencies that provide services through its
workers); and 3) profiles that were active on the Platform between
June 2015 and March 2016. In total, we retrieved 24,000 user profiles
within these dimensions.
Subsequently, we set to obtain a pool of gender identifiable
profiles. While the Platform does not provide a specified field for
gender in the user profile, the gender of a user is known to people or
companies who provide work opportunities either through the user’s
name, or through their profile picture. Therefore, in order to
automatically determine the gender of the extracted profiles, we used
two web services:
a. Genderize.io, a web service aimed at identifying gender based
on English names.61
b. Angus.ai, a web service aimed at identifying gender based on
photos.62
To validate the automated gender identification of the extracted
profiles, we only selected profiles in which both services indicated an
80% or more certainty about the gender, and additionally, we
compared the findings of each service with the other, and only took
profiles that both services identified as the same gender. After
comparing both services, we remained with profiles whose average
gender accuracy was 98% for name identification and 96% for photos.
After this, 4,669 profiles remained.
From the extracted profile data, we further selected fields for
analysis according to fields articulated by the Platform’s API: 1)
Occupational Category (“Accounting & Consulting,” “Admin
Support,” “Customer Service,” “Data Science & Analytics,” “Design &
Creative,” “Engineering & Architecture,” “IT & Networking,” “Legal,”
“Sales & Marketing,” “Translation,” “Web, Mobile & Software
Development,” “Writing”); 2) “Hourly Rate” – the hourly pay rate
determined by the work-seeker; 3) “Feedback Score” – a score on a
scale of one to five given to the user by those who utilized the user’s
60 See Christine Hine, Internet Research and Unobtrusive Methods, 61 SOC. RES. UPDATE
1 (2011) (presenting the benefits of such an approach).
61 See Determine the Gender of a First Name, GENDERIZE.IO, https://genderize.io/ (last
visited Dec. 12, 2016).
62 See ANGUS.AI, https://www.angus.ai/ (last visited Sept. 21, 2016).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 407
labor after completing a task through the Platform: this field shows the
average feedback score a user received from previous assignments; and
4) “Total Hours” – the accumulative number of hours a user worked
through the Platform. Note that in our dataset the total number of
hours does not include tasks billed with a fixed price, but only those
with an hourly rate.
To further enrich the analysis, we computed two additional fields
based on the data extracted from the API. The first is years of
experience. From the profiles’ self-description field, we automatically
extracted the description of previous work, and calculated years of
experience as a subtraction of the first year from the last year worked
in each previous job. We added the number of years worked in each
previous work to calculate an estimate of the user’s work experience in
years. Parallel jobs conducted in the same year were excluded from
the calculation of this field, so as not to skew the data.
The second computed field is the level of education attainment.
The education field returned by the Platform’s API contains all
degrees mentioned by a given profile. To determine a profile’s level
of education, we manually selected the highest degree mentioned, and
kept it in a separate field we called “degree.” Subsequently, we
clustered the different degrees into the following categories of level of
education attainment: high school, associate degrees, bachelor
degrees (undergraduate), master’s degrees (graduate), and doctorate
degrees (including J.D.). Since the extraction of the highest degree
was performed manually, the field of level of education attainment was
computed only on the occupational category “Legal.”
Finally, we used descriptive statistics and regression models to
analyze the data. Specifically, we computed the differences in the
average hourly rate of women and men across all occupational
categories, and used a standard t-test to compare mean differences by
gender within occupational category. After confirming that there are
significant interactions between occupational categories and the
hourly rate, we subsequently conducted a two-way analysis of variance
(“ANOVA”) with the log of the hourly rate in order to assess the effect
of gender on the hourly pay in each occupational category. We then
repeated the model, each time testing for possible interactions with
the following variables: “feedback score,” “years of experience,” and
“total hours” worked on the Platform. Finally, we used a linear
regression model on the log of the hourly rate to compute the ratio in
the hourly rate of women and men in each occupational category, as
well as for all categories taken together.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
408 SETON HALL LAW REVIEW [Vol. 47:393
B. Findings
Our findings show a dramatic gender gap in platform-facilitated
online work, on the Platform. They show that although the overall
number of male and female profiles in the dataset is equally
distributed (N = 2321 women, 2348 men), and so is the average
feedback score of male and female profiles (3.21 for women, 3.17 for
men), women have worked a larger total number of hours (N =
773,666) than men (N = 611,912). However, on average, women’s
hourly rate is 37% lower than men’s: the overall average hourly rate
for women is $28.20 per hour, compared to an average hourly rate of
$45.07 for men. It should be noted that the proportion between
female and male profiles varies across the different occupational
categories (see Figure 1), and so does the gap in the hourly rate of men
and women in each category (see Figure 2). Half of the occupational
categories (N=6) are populated by more women than men, namely,
Customer Service (77% women), Admin & Support (73% women),
Legal (70% women), Translation (65% women), Writing (65%
women), and Sales & Marketing (62% women). Three categories have
more male profiles: Engineering & Architecture (83% men), IT &
Networking (82% men), and Data Science & Analytics (73% men). In
the remaining categories (N = 3) male and female profiles are more or
less equally distributed: Accounting & Consulting (54% women), Web,
Mobile & Software Development (49% women), and Design &
Creative (44.1% women).
Figure 1. The number of male and female profiles in each occupational category.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 409
Figure 2. Average hourly rate by occupational category and gender.
Table 1 shows a breakdown of the average hourly rate per
occupational category. Since the distribution of the average hourly
rate was not normal, we conducted a t-test to compare differences on
the log of the average hourly rate of men and women. We report that
the hourly rate gap exists in all categories, albeit with significant
differences between categories. For example, “Legal” stands out as the
occupational category where the average hourly rate of women makes
only 37% of men’s 100%. At the other extreme, women’s profiles in
the “Design & Creative” category have an average hourly rate that is
almost equal to that of men (95%). A significant hourly rate gap is
reported for categories with a majority of female profiles. In the
categories “Accounting & Consulting” and “Customer Service,” for
example, the average hourly rate of women is only 62% and 64%,
respectively, of the average hourly rate of men. Although in the
categories “Translation” and “Writing,” where there is also a majority
of female profiles, the reported average hourly rate gap is narrower
(79% and 83%, respectively), the differences between the average log
of the hourly rate of men and women in these categories have not been
found statistically significant (see Table 1). In categories with a
majority of men’s profiles, the hourly rate gap varies from 65% in “IT
& Networking” to 80% in “Data Science & Analytics.”
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
410 SETON HALL LAW REVIEW [Vol. 47:393
Table 1. T-test on Log of Hourly rate
Hourly rate
T
-
test
on the log
of Hourly
rate
Category Gender N Mean Std
Female
compared
to Male
Significance
A
ccounting &
Consulting
female 204 36.52 46.05
-38%
male 186 58.96 55.34
Admin Support female 278 23.86 14.99 -32%
male 84 35.03 25.79
Customer Service female 273 17.26 10.03 -36%
male 101 26.97 27.55
Data Science &
Analytics
female 121 36.65 40.33
-20%
male 318 45.83 34.66
Design & Creative female 162 33.50 24.87
-
5%
male 205 35.13 23.53
Engineering &
Architecture
female 60 26.45 15.69
-36%
male 289 41.31 26.08
IT & Networking female 64 32.72 24.60 -35%
male 281 50.01 40.61
Legal female 302 28.88 31.86 -63%
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 411
male 129 77.93 71.04
Sales & Marketing female 213 32.50 22.44 -32%
male 195 48.10 36.69
Translation female 292 22.57 13.78
-
17%
male 155 27.03 22.06
Web, Mobile &
Software Dev
female 98 35.41 19.45
-25%
male 270 47.38 37.67
Writing female 254 29.13 21.15
-
21%
male 135 37.10 35.27
ALL female 2321 28.20 26.01 -37%
male 2348 45.07 39.65
P<.05 P<.01 P<.001
In order to verify that the differences in the hourly rate of
women and men are not a result of other factors, several alternative
explanations were tested. Specifically, we attempted to rule out that
the occupational category itself, the years of experience worked
outside of the Platform, the feedback score or the total hours worked
on the Platform are the variables that account for most of the
differences found in the hourly rate of women and men. A two-way
analysis of variance (“ANOVA”) model shows that the interaction
between gender and the occupational category has a significant effect
(F = 40.19, p < 0.001, R-square = 0.165). See Table 2.
Table 2: Two-way ANOVA on the log of the hourly rate
Source DF Sum of
Squares
Mean
Square
F
Value
Pr > F
Model 23 400.133947 17.397128 40.19 <.0001
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
412 SETON HALL LAW REVIEW [Vol. 47:393
Source DF Sum of
Squares
Mean
Square
F
Value
Pr > F
Error 4645 2010.480699 0.432827
Corrected
Total
4668 2410.614646
R
-
Square Coeff Var Root MSE log hourly rate mean
0.165988 19.79749 0.657896 3.323126
Source DF Type III SS Mean
Square
F
Value
Pr > F
Category 11 165.7583917 15.0689447 34.82 <.0001
Gender 1 94.0194458 94.0194458 217.22 <.0001
Category*Gender 11 41.2556977 3.7505180 8.67 <.0001
Furthermore, gender remained a significant predictor of the
hourly rate when each of the other independent variants (years of
experience, feedback score, hours of work) was added to the model.
For example, there is a significant interaction between years of
experience and the log of the hourly rate. However when the variant
years of experience was added to the model together with gender and
occupational category, the interaction between gender and the log of
the hourly rate remained significant (F = 39.20, p < 0.001, R-square =
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 413
0.19). A two-way multivariate analysis of variance (“MANOVA”) model,
comparing gender, occupational category, feedback score, years of
experience and the total hours worked on the Platform resulted in a
significant multivariate effect (F = 19.64, p < 0.001, R-square = 0.18).63
See Table 3. This confirms that the alternative explanations to the
differences found in the hourly rate of women and men do not rule
out the significant effect of gender.
Table 3. Two-way MANOVA on the log of the hourly rate.
Source DF Sum of
Squares
Mean
Square
F
Value
Pr > F
Model 26 187.924453 7.227864 19.64 <.0001
Error 2249 827.657393 0.368011
Corrected
Total
2275 1015.581846
R
-
Square Coeff Var Root MSE log hourly rate mean
0.185041 17.97336 0.606639 3.375214
63 Note that the variant “Total hours” was computed on values that are larger than
zero, therefore only relating to people who have already completed hourly work on
the Platform (N = 2275).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
414 SETON HALL LAW REVIEW [Vol. 47:393
Source DF Type III SS Mean
Square
F
Value
Pr > F
Category 11 39.98418673 3.63492607 9.88 <.0001
Gender 1 61.79873027 61.79873027 167.93 <.0001
Category*Gender 11 13.83327945 1.25757086 3.42 0.0001
Feedback Score 1 5.00322945 5.00322945 13.60 0.0002
Experience 1 12.77920923 12.77920923 34.73 <.0001
Total Hours 1 1.32609520 1.32609520 3.60 0.0578
Since we found significant interaction between the
occupational category and the hourly rate, and after confirming that
the profiles in each category are unique, we further conducted
separate regression analyses on the effect of gender on the hourly rate
for each of the twelve occupational categories, as well as for all
categories taken together. As can be seen in Table 4, with the
exception of the categories “Design & Creative,” “Translation” and
“Writing,” we found a significant effect of gender on the hourly rate.
We subsequently converted the difference in the log values of the
hourly rate of women and men into a ratio of women’s hourly rate out
of men’s. We can therefore confirm that in nine out of the twelve
occupational categories there are statistically significant gaps in the
hourly rate of women compared to men. For example, in the
“Customer Service” category, women’s hourly rate is 73% of what their
male peers requested. In the “Legal” category, women’s hourly rate is
as low as 44% of the hourly rate asked by men. Finally, we report that
after controlling for the separate effect of gender on the hourly rate in
each of the occupational categories, when all occupational categories
are taken together, women’s hourly rates are only 66% of what men
request. This figure corresponds with the 37% gap reported for the
differences in the average hourly pay for men and women at the outset
of the findings section.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 415
Table 4. Regression analysis of gender on the log of the hourly rate
Category N
Regression
coefficient
of female vs.
male
(on log of
Hourly rate) Significance
Female/
male
ratio (on
Hourly
rate)
A
ccounting &
Consulting
390 -0.4752 0.62
Admin Support 362 -0.3122 0.73
Customer Service 374 -0.2893 0.75
Data Science &
Analytics
439 -0.2826 0.75
Design &
Creative
367 -0.0516 NS 0.95
Engineering &
Architecture
349 -0.4107 0.66
IT & Networking 345 -0.3947 0.67
Legal 431 -0.8261
0.44
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
416 SETON HALL LAW REVIEW [Vol. 47:393
Sales &
Marketing
408 -0.3436 0.71
Translation 447
-
0.1086 NS 0.90
W
eb, Mobile &
Software
Development
368 -0.2112 0.81
W
riting 389
-
0.1378 NS 0.87
ALL
466
9
-0.4109 0.66
We have seen that the category “Legal” somewhat deviates from
the other categories under examination, since it displays the largest
gap between the average hourly rate of men and women, but our t-test
on the log of the average hourly rate of men and women was not
statistically significant. In order to further understand the outstanding
hourly rate gap found in the “Legal” category, we examined its internal
distribution. As Figure 3 shows, the hourly rate distribution of women
in this category is highly skewed, with the vast majority of profiles
offering the lower end of the hourly rate scale.
B
ARZILAY
&
BEN
-D
AVID
(D
O
N
OT
D
ELETE
) 2/16/2017
1:55
PM
2017] PLATFORM INEQUALITY 417
Figure 3. Distribution of hourly rate of male and female profiles in the “Legal
Category.”
One possible explanation for this huge gap is the fact that, in
the studied platform, the “Legal” category hosts a variety of tasks and
skills, from paralegals to attorney-at-law. Therefore, we further
examined the ties between gender, years of experience, and the level
of education in this category. As Figure 4 shows, the hourly rate gap is
apparent in all ranges of work experience. When the hourly rate gap
between women and men is compared against the level of education
in the “Legal” category, the findings are even more striking: while at
the associate degree level the average hourly rate of women is 106%
that of men, at the undergraduate level there is already a gap of 65%
in favor of men, and at the graduate level (which includes LL.M.,
M.B.A., etc.) the average hourly rate of women makes only 34% of the
average hourly rate of men. At the doctorate level (J.D.) women J.D.’s
hourly rates are 55% to men’s 100% (see Figure 5). A regression
model on the effect of gender, years of experience, the level of
education on the log of the hourly rate in the legal category confirms
that the effect of gender remains significant even given other factors
that explain the variance in the hourly rate (F=31.87, p<0.001, R-
square = 0.28). See Table 5.
B
ARZILAY
&
BEN
-D
AVID
(D
O
N
OT
D
ELETE
) 2/16/2017
1:55
PM
418 SETON HALL LAW REVIEW [Vol. 47:393
Figure 4. Legal category: Average hourly rate by years of experience and gender.
Figure 5. Legal category: Average hourly rate by level of education and gender.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 419
Table 5. “Legal” category: Regression model of the effect of gender,
years of experience and level of education on the log of the hourly rate
Dependent Variable: log hourly rate
Number of Observations Read 431
Number of Observations Used 239
Number of Observations with Missing Values 192
A
nalysis of Variance
Source DF Sum of
Squares
Mean
Square
F Value Pr > F
Model 3 43.34248 14.44749 31.87 <.0001
Error 235 106.53534 0.45334
Corrected Total 238 149.87782
Root MSE 0.67331 R
-
Square 0.2892
Dependent Mean 3.33536
A
dj R
-
Sq 0.2801
Coeff Var 20.18691
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
420 SETON HALL LAW REVIEW [Vol. 47:393
Parameter Estimates
V
ariable DF Parameter
Estimate
Standard
Error
t
V
alue Pr > |t|
Intercept 1 2.55237 0.20678 12.34 <.0001
Gender (Female) 1
-
0.45163 0.10090
-
4.48 <.0001
Degree 1 0.26824 0.04262 6.29 <.0001
Experience 10.01159 0.00559 2.07 0.0393
C. Discussion of Findings
Our findings show that, on average, women’s hourly rates are
significantly lower than men’s when considering the same tasks,
despite similar levels of educational attainment, feedback score, and
length of experience.
Two questions derive from this data. The first concerns
women’s undervaluation of themselves. This could be caused either
by women’s increased need for money (which may explain why they
work more on the Platform), or because the online hourly rates are
operating in the shadow of the offline market in which women often
earn less than men. Should women simply post higher hourly rates?
Literature on women and negotiations has identified a significant
difference between men and women in their propensity to negotiate
for wages and an a-priori lowering of salary expectations among women
to avoid negotiation.64 Recently women have been publically urged to
“lean-in”65 by demanding higher pay, for example. However, new
research suggests that posting higher hourly rates may not always be
beneficial for women. This research shows that when women are not
64 See generally LINDA BABCOCK & SARA LASCHEVER, WHY WOMEN DON’T ASK: THE
HIGH COST OF AVOIDING NEGOTIATION—AND POSITIVE STRATEGIES FOR CHANGE (2007).
65 SHERYL SANDBERG, LEAN IN: WOMEN, WORK, AND THE WILL TO LEAD (2013).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 421
sure what is expected in negotiations, or when those expectations are
murky, asking for more money does not always work as well for women
as it does for men, and may actually hurt women financially.66
A second question is whether there are actual differences in
pay. The answer to this question relates to our dependence on the
data provided by the Platform’s API. Since the actual payment
received by users per hour is not available for extraction from the
Platform’s API, we are unable to systematically analyze actual payment
received by users per hour for similar tasks and occupations, compared
to their hourly rate. Despite this limitation, it is most likely that gender
differences in pay remain due to the Platform’s role in shaping the
interaction between work seekers and work providers: work seekers
apply to jobs posted by work providers, after which the latter can review
the applications and contact applicants for interviews, based on their
profiles which include their hourly rates.67 At this point, some
negotiation may go on between users and those in need of tasks
performed, however the negotiation itself is not a structured element
of the Platform’s interface and may take place using another medium,
such as email, chat, or telephone.68 There may be some negotiation
regarding timeframe and expediency, and a move to a fixed rather
than hourly rate. Such fixed rate will likely be derivative or connected
to the posted hourly rate. While it is a possibility that during
negotiation men’s hourly rates will go down, and women’s rates will
increase, thus making the actual pay more equitable, it seems more
likely that some gender discrepancy remains, especially in categories
in which the difference in hourly rates is huge, such as the “Legal”
category (women’s average hourly rate is 63% lower than men’s), or
large such as “Accounting & Consulting” (women’s average hourly rate
is 38% lower than men’s), making it increasingly plausible that
ultimately women are paid less for the same tasks. The gaps we found
resonate with findings from studies on the pay gap in the general
economy, further suggesting that the “gig” economy is not an anomaly
to this pattern.69
66 See Christine Exley, Muriel Niederle & Lise Vesterlund, New Research: Women Who
Don’t Negotiate Might Have Good Reason, HARV. BUS. REV. (Apr. 12, 2016),
https://hbr.org/2016/04/women-who-dont-negotiate-their-salaries-might-have-a-
good-reason; Christine L. Exley, Muriel Niederle & Lise Vesterlund, Knowing When to
Ask: The Cost of Leaning-in (Harvard Bus. Sch., Working Paper No. 16–115, 2016),
http://www.hbs.edu/faculty/Publication%20Files/16-115_a6680006-be03-44be-ab6f-
3625d3f21c00.pdf.
67 See supra note 52.
68 Id.
69 See supra note 26.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
422 SETON HALL LAW REVIEW [Vol. 47:393
While our computational approach to measuring differences
between men and women in the “gig” economy through the data
provided by the API of the studied platform limits our ability to
measure actual pay, it allows for large data computation, and an actual
snapshot of differences in the average hourly rates of men and women
on the Platform. Furthermore, it uniquely captures some of the
marketplace interactions that take place ahead of pay negotiations or
actual pay in platform-facilitated labor. While survey data relies on
answers to questions reported by men and women, our data set
captures the shaping of new dynamics of self-evaluation and self-
presentation in the “gig” economy. Through these interactions, our
findings suggest that we are witnessing the remaking of women into
devalued laborers.
PART III. THE INEPTITUDE OF CURRENT LEGAL NORMS
Advances in technology are powerful tools that have influenced
both industry and gender relations.70 As online platforms enhance
widespread changes in labor, the law must adapt to the changing
workplace. What can be done to ensure that women realize the
promises of work in the “sharing” economy?
Law can provide at least two tools for enhancing gender
equality in this context. First, online, platform-based labor should
become an option that provides sustainable work for those heavily
involved. This requires some labor protections like minimum wage,
pro rata health insurance, and family leave, at least for those working
a significant number of hours on platforms. This first tool will be
discussed briefly in infra Part III.A as others have dealt with the
question which it implicates: when workers should be considered
employees. The second tool, addressed in infra Part III.B, focuses on
the inability of antidiscrimination law to deal with women’s platform-
labor subordination. As our case study illustrates, there are dramatic
gender disparities in hourly rates in platform-facilitated labor. While
the case study does not purport to examine all platform-facilitated
online labor, it should raise serious concerns about women’s equality
in this context, and about platforms’ potential role in cultivating
gender equality. Efforts to mitigate the perils faced disproportionately
by women might include updating antidiscrimination laws to diminish
gendered pay gaps or imagining entirely new forms of regulation.
70 Nancy B. Schess, Then and Now: How Technology Has Changed the Workplace, 30
HOFSTRA LAB. & EMP. L.J. 435 (2013).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 423
A. Employee Status
The debate about labor protections directly implicates the
question of whether taskers are employees or independent
contractors. The law’s answer to this question may vary according to
the amount of control the given platform possesses over its taskers.71 If
taskers are considered employees of a certain company, they become
eligible for protective labor laws such as minimum wage and
overtime.72 Online platforms, however, are fiercely resisting efforts to
classify taskers as employees.73 Platforms insist that their workers are
nothing more than independent contractors.74 Uber, for example,
exerts considerable control over its drivers by setting rates and making
termination decisions,75 yet considers them independent contractors.76
For workers on other platforms which do not set laborers’ rates, it
might be more difficult to argue for employee status. Yet even for such
workers, considering all laborers on a given platform as business
entities and micro-entrepreneurs ignores both the gatekeeping role
such platforms have undertaken, and the surplus they attain for every
transaction. Some have argued that we should view taskers’
relationships to platforms as an intermediate status between traditional
employment and independent contracting (proposing a third status of
dependent contractors or independent workers).77 This approach
would afford workers some labor protections, but not all.78
B. Antidiscrimination
Employment status is also critical to the second component,
which focuses on gender in particular. The protections of
antidiscrimination laws such as Title VII are triggered only in an
71 See supra note 19.
72 See Sachs, supra note 7; Miriam A. Cherry, Beyond Misclassification: The Digital
Transformation of Work, 37 COMP. LAB. L. & POL’Y J. 577, 578, 581–82 (2016). See generally
Cunningham-Parmeter, supra note 19.
73 See supra note 19.
74 This is clearly evident from platforms’ terminology of laborers, see supra note
13, and see also the litigation in O’Connor v. Uber Techs., Inc., 58 F. Supp. 3d 989
(N.D. Cal. 2014); and Cotter v. Lyft, Inc., No. 13-CV-04065, 2016 WL 1394236 (N.D.
Cal. Mar. 11, 2015).
75 Cunningham-Parmeter, supra note 19, at 1, 11–13, 33–35.
76 Id.
77 Seth D. Harris & Alan B. Krueger, A Proposal For Modernizing Labor Laws for
Twenty-First-Century Work: The “Independent Worker”, HAMILTON PROJECT (Dec. 2015),
http://www.hamiltonproject.org/assets/files/modernizing_labor_laws_for_twenty_fi
rst_century_work_krueger_harris.pdf.
78 Id.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
424 SETON HALL LAW REVIEW [Vol. 47:393
employment relationship.79 Even if workers do eventually win the
battle over their employment status, antidiscrimination laws may
require updating for the new, online economy. Since online labor is
often performed under gender-blind policies,80 and without face-to-
face interaction, discriminatory treatment may be especially difficult to
prove.
If, as some have argued, taskers should be viewed as dependent
contractors or independent workers,81 antidiscrimination laws would
apply, but other worker protections like health insurance mandates
would not.82 Under this scenario, workers in this intermediate category
would be significantly less protected than if they were deemed
employees. Women might not be legally discriminated against, but all
workers would still lack important supports. The lack of family leave
may affect women more significantly than men.83 And, of course,
intentional discrimination would remain difficult to prove. Still, an
intermediate status may be better for workers than the status quo, in
which workers are currently treated as independent contractors.84
Certain basic labor protections are not part of this relationship by
operation of law, such as minimum wage and family leave, or even
antidiscrimination laws such as Title VII (though some have long
sought to amend antidiscrimination laws to cover independent
contractors).85
Yet even if online taskers are relegated to independent
contractor status, one could argue that platforms still cannot permit
discrimination to occur on their platforms. Scholars have recently
claimed that some “sharing” economy companies, especially those
offering services like housing or transportation, are “functional
substitutes” for traditional public accommodations and therefore that
federal public accommodation law should be updated to capture racial
discrimination in the “sharing” economy.86 Sex discrimination, on the
79 Pub. L. No. 88-352, tit. VII, 78 Stat. 241, 253–66 (1964) (codified as amended
at 42 U.S.C. §§ 2000e to 2000e-17 (2012)).
80 The Platform in our case study, for example, does not have a designated field
for gender on its profile “template” but does have a field for name and photo.
81 Harris & Krueger, supra note 77.
82 Id.
83 Women still perform the lion’s share of family-work. See Schoenbaum, supra
note 17, at 3.
84 Ross, supra note 33, at 1438–41.
85 Lewis L. Maltby & David C. Yamada, Beyond “Economic Realities”: The Case for
Amending Federal Employment Discrimination Laws to Include Independent Contractors, 38
B.C. L. REV 239 (1997).
86 Nancy Leong & Aaron Belzer, The New Public Accommodations, 105 GEO. L.J.
(forthcoming 2017), http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2687486
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 425
other hand, can be litigated under state public accommodation
legislation, such as the Unruh Civil Rights Act in California.87 That Act
bars sex discrimination by business establishments.88 If, as platforms
claim, their laborers are business partners/micro-entrepreneurs/
independent contractors, then one could argue that platforms are
business establishments that must treat clients (here, the taskers)
without discrimination.89 Following the common law doctrine
requiring places of public accommodation “to serve all customers on
reasonable terms without discrimination and . . . to provide the kind
of product or service reasonably to be expected from their economic
role,”90 the Unruh Act forbids business establishments from engaging
in sex discrimination in dealing with their clients.91
Similar laws can be found in other jurisdictions as well.92 Three
doctrinal questions arise from this type of law, though: first, are online
platforms business establishments under the Unruh Act? While most
public accommodation doctrine has been interpreted to refer to
physical accommodation,93 some precedents have extended Unruh’s
(draft from July 23, 2016).
87 CAL. CIV. CODE §§ 51–51.3 (West 2016).
88 Id. § 51, 51.5(a).
89 See also Noah Zatz, Beyond Misclassification: Gig Economy Discrimination Outside
Employment Law, ON LABOR (Jan. 19, 2016), https://onlabor.org/2016/01/19/beyond-
misclassification-gig-economy-discrimination-outside-employment-law.
90 In re Cox, 474 P.2d 992, 996 (Cal. 1970).
91 The Unruh Civil Rights Act, California Civil Code Section 51, provides, in
pertinent part:
(b) All persons within the jurisdiction of this state are free and equal,
and no matter what their sex, race, color, religion, ancestry, national
origin, disability, medical condition, genetic information, marital status,
sexual orientation, citizenship, primary language, or immigration status
are entitled to the full and equal accommodations, advantages, facilities,
privileges, or services in all business establishments of every kind
whatsoever.
CAL. CIV. CODE § 51 (West 2016).
92 See, e.g., Alberta Human Rights Act, R.S.A. 2000, c. A-25.5 (Can.); Treaty of
Amsterdam Amending the Treaty on European Union, the Treaties Establishing the
European Communities and Certain Related Acts art. XIII, Oct. 2, 1997, 1997 O.J. (C
340); Council Directive 2004/113, 2004 O.J. (L 373) 37 (EC) (implementing the
principle of equal treatment between men and women in the access to and supply of
goods and services); Isabelle Chopin & Eirini-Maria Gounari, Developing Anti-
Discrimination Law in Europe – The 27 EU Member States Compared, EUR. COMMISSION 101
(Nov. 2009), https://ec.europa.eu/migrant-integration/librarydoc/developing-anti-
discrimination-law-in-europe—-the-27-eu-member-states-compared.
93 “Physical accommodation” refers to accommodations such as inns, hotels,
restaurants, country clubs. Kelly Catherine Chapman, Gay Rights, the Bible, and Public
Accommodations: An Empirical Approach to Religious Exemptions for Holdout States, 100 GEO.
L.J. 1783, 1785–86 (2012). See, e.g., Koebke v. Bernardo Heights Country Club, 115
P.3d 1212 (Cal. 2005) (regarding country clubs).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
426 SETON HALL LAW REVIEW [Vol. 47:393
liability to websites when, for example, they denied access due to sexual
orientation.94
Second, are there legitimate business reasons (permissible
under the statute) to allow taskers to offer their services at substantially
different prices? From platforms’ perspective, for example, this might
generate more diverse profiles. Third, and most importantly, is the
Unruh definition of discrimination too narrow? Under Unruh, cases
have typically been concerned with assuring equal opportunity of
access (whether to rental property or entrance to restaurants, country
clubs, and other businesses).95 But in online labor, gender
discrimination does not seem to appear in the form of formal barriers
or overt exclusions:96 both men and women are free to participate as
online taskers, yet pay inequality seems to occur in platform-facilitated
labor. Under the Unruh Act, prohibiting this kind of disparity would
require an interpretation that ensures not just equal access to
platforms but prevents substantive discrimination from taking place
within platforms. Courts have noted that the Unruh Civil Rights Act
must be construed liberally.97 Still, some courts have required proof of
intentional discrimination,98 an extremely difficult burden to satisfy in
the online context.
94 See, e.g., Ashleigh Bergeron, Butler v. Adoption Media, LLC: Eradicating Sexual
Orientation Discrimination in Cyberspace, 17 L. & SEXUALITY 173, 179–80 (2008)
(examining Butler v. Adoption Media, LLC, 486 F. Supp. 2d 1022 (N.D. Cal. 2007),
which declared that a website was a “business establishment” and its policy of allowing
access only to heterosexual couples was not supported by legitimate business reasons).
95 See Alison Rothi, Changing Ideas About Changing Diapers, 25 WHITTIER L. REV. 927
(2004) (discussing the Unruh Act).
96 See, e.g., case study discussion supra Part II.
97 Munson v. Del Taco, Inc., 208 P.3d 623, 626 (Cal. 2009) (finding that liberal
construction is necessary to carry out the Act’s purpose to “create and preserve a
nondiscriminatory environment . . . by ‘banishing’ or ‘eradicating’ arbitrary, invidious
discrimination by such establishments. . . . [The] Act ‘serves as a preventive measure,
without which it is recognized that businesses might fall into discriminatory
practices’”) (citations omitted). The court holds that in the context of disability
discrimination, a plaintiff proceeding under Section 51(f) of the Unruh Act “may
obtain statutory damages on proof of an ADA access violation without the need to
demonstrate additionally that the discrimination was intentional.” Id. at 628.
98 See Timothy B. Liebaert, The Death of the Unruh Civil Rights Act: An Examination of
the Act After Harris v. Capital Growth Investors XIV and an Argument in Favor of
Liberalizing the Act, 29 W. ST. U. L. REV. 1 (2001); Note, The Andiscrimination Principle in
the Common Law, 102 HARV. L. REV. 1993 (1989). See, e.g., Greater L.A. Agency on
Deafness, Inc., v. Cable News Network, Inc., 742 F.3d 414, 425 (2014) (holding that
under the Unruh Act, plaintiffs must show intentional discrimination and affirmative
misconduct on the part of those violating the Act—much more than disparate impact
of a facially neutral policy); Koebke v. Bernardo Heights Country Club, 115 P.3d 1212
(Cal. 2005).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 427
Moreover, even if discrimination jurisprudence could be
interpreted to resolve the doctrinal difficulties noted above, bringing
claims under antidiscrimination law is not the only possible way, and
may possibly not even be the best way, to tackle women’s subordinate
position in online labor. Indeed, filing such cases may generate
publicity and have some “educational” effect on the market. But
discrimination claims are largely based on private rights of
enforcement which require that taskers’ legal consciousness be raised99
and, once raised, that they would be capable and willing to present,
and possibly litigate, claims. Since such claims are often filed
individually, and after the fact, and given the magnitude and scale of
online labor, their effect may be quite limited.
PART IV. FROM DISCRIMINATION 3.0 TOWARDS EQUALITY-BY-DESIGN
A. A Third Generation of Discrimination
The doctrinal difficulties regarding both federal anti-
discrimination law and state public accommodation law, expose the
ineptitude of current legal norms to address the devaluation of
women’s work in today’s on-call, platform-facilitated economy. Of
course, we are only beginning to understand the relationship between
gender and the “gig” economy, and so this Part provides a nascent
attempt to conceptualize gender inequality in this emerging context.
In the past, plaintiffs in discrimination suits could point to a specific
culprit, whether an individual or a policy, and the court would have to
answer a basic underlying question: “Who did the discriminating?” In
the first generation of sex discrimination claims, formal barriers played
a significant evidentiary role. In the second generation, the quest was
to identify underlying sex biases by a manager, company policy, or a
workplace culture, dynamic, or organizational practice that lead to
entrenching unequal access, exclusion, glass ceilings or harassment.100
But in today’s world of platform work, as demonstrated in our
case study, we are witnessing what may be a third generation of sex
inequality—Discrimination 3.0. Our focus and the questions we ask
must be adapted in two ways. First, the questions of access and
inclusion should be supplemented with questions of usability. Men
and women generally have equal access to the platform, yet the way
99 See Amy Blackstone et al., Legal Consciousness and Responses to Sexual Harassment,
43 LAW & SOC’Y REV. 631 (2009).
100 See Susan Sturm, Second Generation Employment Discrimination: A Structural
Approach, 101 COLUM. L. REV. 458, 465–74 (2001). Others have described the first
phase of gender discrimination as ending exclusion and the second as transforming
male centred norms, see Abrams, supra note 28, at 1186.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
428 SETON HALL LAW REVIEW [Vol. 47:393
they may be able to benefit from platform-facilitated labor seems to
vary. Second, the goal should increasingly shift from aiming to
determine who is doing the discrimination to answering how the
discrimination is being effectuated. Indeed, the how question has been
suggested by antidiscrimination scholars in other contexts,101 but it is
increasingly important in the online labor arena, where intentional
discrimination, and even biases towards an individual tasker, are often
virtually impossible to prove. We should therefore increasingly be
asking how inequality is being (re)produced and in which
institutionalized ways is discrimination enabled by platforms, and by
society? We should steadily be inquiring how discrimination is
facilitated and harbored by the practices of platform-facilitated on-call
labor. But also, consider the structural disadvantages with which
laborers come to use “gig” economy platforms to begin with.102
As for platforms, we know that they not only represent social
interactions, but that the way platforms are structured—their
architecture, meaning their technological affordances and code—
affects human behavior.103 Platform codes are invisible to most of us,
yet they operate as de facto law in the online arena.104 The way they
regulate is a function of their algorithmic design.105
In our case study, we revealed consistent gender discrepancies
regarding hourly rates. But women were not excluded from the
Platform, which could be considered prohibited intentional
discrimination, and the Platform may not have even portrayed them in
a biased or sex-stereotyped manner. In fact, on its face, its interface is
gender agnostic, in the sense that it does not structure gender as an
official field in the user’s profile. One may argue that the platform
101 Sturm, supra note 100, at 460–61. See also Martha Albertson Fineman, The
Vulnerable Subject: Anchoring Equality in the Human Condition, 20 YALE J.L. & FEMINISM 1
(2008) (arguing that we should move from focusing on individual discrimination
focused on identity and look at the ways social institutions distribute power and
resources).
102 For example, structural disadvantages include gendered pay gaps in the offline
labor market and lack of paid family leave. See supra note 26; see also Maxine Eichner,
Square Peg in a Round Hole: Parenting Policies and Liberal Theory, 59 OHIO ST. L.J. 133,
148–50 (claiming that U.S. parental leave policy fails to protect a sufficient concept of
parenting); Laura T. Kessler, Keeping Discrimination Theory Front and Center in the
Discourse over Work and Family Conflict, 34 PEPP. L. REV. 313, 314 (2007) (explaining
connections between gender bias and work/family conflict).
103 LAWRENCE LESSING, CODE VERSION 2.0 5–6, 24, 34 (2006); see also Joel R.
Reidenberg, Lex Informatica: The Formulation of Information Policy Rules Through
Technology, 76 TEX. L. REV. 553, 555 (1998); Lawrence Lessig, The Law of the Horse: What
Cyberlaw Might Teach, 113 HARV. L. REV. 501, 505–06 (1999).
104 LESSING, supra note 103, at 58.
105 Id.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 429
solely reflects age-old social dynamics. Yet the codes of a platform may
nonetheless condition some of the social interactions it hosts.106
Contemplating the ties between platforms’ policies, their
technological affordances and user behavior, along with opportunities
in play outside of the platform in the market and in the family, may
explain some of the discriminatory effect we observed.107 These
interactions, taken together with women’s structural disadvantages in
the labor market and the family more generally suggest that platform-
facilitated labor should be viewed cautiously as a liberating, equality-
enhancing work option for women.
B. Towards A Platform for Equality?
The aim of this last subsection is to briefly sketch a possible
direction for enhancing equality in platform-facilitated labor in the
future. If “sharing” economy companies are made aware of the
discrimination taking place on their platforms, they may be inclined to
consider ways to offset inequality. Changes within a platform’s design
may prove necessary to prevent today’s forms of sex discrimination at
work, and regulating platforms through law and technology may
advance women’s economic equality in online workplaces. Yet, it may
not be quite as straightforward as merely embedding the value of
equality into a technology and then expecting it to create a direct
positive effect on society.108 There may well be unintended
consequences to this move or indeterminate ramifications which we
106 On the role of the technological affordances of social media platforms in
shaping discrimination, see Anat Ben-David & Ariadna Matamoros Fernandez, Hate
Speech and Covert Discrimination on Social Media: Monitoring the Facebook Pages of Extreme-
Right Political Parties in Spain, 10 INT’L J. OF COMM. 1 (2016).
107 When further thinking about how discrimination may be facilitated through
platforms regarding the mediation of the work related social interactions and the
hourly rate gap, we should be thinking about platforms’ existing features and
affordances, but also consider those features that are absent from the interface.
Consider the following two examples. First, the negotiation process mentioned above.
In most cases, the negotiation on the terms of actual pay that incorporates a social
interaction between the work-seeker and the work-provider takes place outside of the
Platform. Such practice of posting hourly rates that are pending negotiation, along
with the distancing of the negotiations from the Platform, may contribute to the
preservation of age-old gender dynamics in the labor market. Second, the Platform’s
design to conduct feedback scores without substantive guidelines may also cause those
more in need of such work to a-priori lower rates to receive a better feedback score.
108 Mary Flanagan, Daniel C. Howe & Helen Nissenbaum, Embodying Values in
Technology: Theory and Practice, in INFORMATION TECHNOLOGY AND MORAL PHILOSOPHY
322–23 (Jeroen van den Hoven & John Weckert eds., 2008). See also Helen
Nissenbaum, From Preemption to Circumvention: If Technology Regulates, Why Do We Need
Regulation (and Vice Versa)?, 26 BERKELEY TECH. L.J. 1370 (2011) (“[I]f we have
technology, why do we need law?”).
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
430 SETON HALL LAW REVIEW [Vol. 47:393
cannot anticipate. However, given that technology is already
functioning as regulation, those who care about gender equality must
consider ways to support the de-facilitation of platform inequality.
Since platforms have the best access to large-scale information about
their workers’ hourly rates and pay, as well as control over
membership, shaping of profile categories, feedback scores, and
whether to allow for negotiations, platforms themselves could best be
suited to enact pre-emptive measures through their affordances and
codes to de-facilitate and counteract discrimination.109 “Sharing”
economy companies could promote within their architecture an
“Equality-by-Design” (EbD): the structuring of platforms in a manner
that is sensitive to prevailing forms of gender discrimination, in ways
that extend beyond merely omitting gender as a formal element of
platforms’ template for profiles or not portraying women in a biased
manner.
This goal of EbD could be affected through measures platform
designers, researchers and policy makers should start to imagine,
contemplate and discuss. These could include, for example, having
platforms inform the market of average hourly rates by providing
notices of average hourly rates for certain tasks to taskers, or publishing
reasonable hourly rates, or suggesting hourly rates to users. Platforms
may also reflect on the ways profiles are displayed and how the
application system might enable discrimination, or on providing
transparent guidelines for assessing users’ performance and,
importantly, on whether negotiations should be structured or
narrowed. Such changes, if put in place, could be quite easily
monitored by platforms’ technologies, and verified and evaluated for
their equality-enhancing capacities.110
In fact, such mechanisms and reflections could be self-imposed
and self-regulated by platforms.111 It is fair to assume that companies
109
See Katharine T. Bartlett & G. Mitu Gulati, Discrimination by Costumers (Duke L.
Sch. Pub. L. & Legal Theory Series, No. 2015-4, 2016), http://papers.ssrn.com/sol3/
papers.cfm?abstract_id=2540334. With regard to sexual harassment law, see Mary
Anne Franks, Sexual Harassment 2.0, 71 MD. L. REV. 655 (2012).
110 See Flanagan, Howe & Nissenbaum, supra note 108, at 344. For the dangers
associated with algorithmic regulation, see, for example, Maayan Perel & Niva Elkin-
Koren, Black Box Tinkering: Beyond Transparency in Algorithmic Enforcement, FLA. L. REV.
(forthcoming), http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2741513.
111 Some sharing economy companies have begun to prohibit posting materials
that bans users based on sexual orientation. Nick Duffy, Accommodation website Airbnb
removes listing that banned gay couples, PINKNEWS (Nov. 23, 2014, 11:52 AM),
http://www.pinknews.co.uk/2014/11/23/accomodation-website-airbnb-removes-
listing-that-banned-gay-couples. For an example of effective self-regulation in the
general economy, see Sturm, supra note 100, at 509–19.
BARZILAY & BEN-DAVID (DO NOT DELETE) 2/16/2017 1:55 PM
2017] PLATFORM INEQUALITY 431
aiming to attract women laborers would be well advised to consider
such mechanisms. However, imposition of legal liability might
ultimately be necessary to initiate some of these reforms.
CONCLUSION
This Article aims to ignite a conversation about equality in the
“sharing” or “gig” economy. It draws attention to the gender inequality
taking hold within digital platforms. It empirically shows that gender
inequality is re-configured and reproduced through platform-
facilitated labor, casting doubt on the liberating and equality-
enhancing promise platform-facilitated labor carries for women.
Looking ahead towards the legal and policy choices that are likely to
take place in the near future regarding labor in the “sharing”
economy, the realization that gender inequality is reproduced on
platforms must force us to begin to contemplate legal, social and
technological mechanisms to mitigate this phenomenon.
