University of Southern California Dissertations and Theses

Essays in applied microeconomics

(USC Thesis Other)

Essays in applied microeconomics

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ESSAYS IN APPLIED MICROECONOMICS
by
Monira Al Rakhis
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
August 2023
Copyright 2023 Monira Al Rakhis
Acknowledgments
I would like to thank my main advisor, Vittorio Bassi, for his invaluable feedback, continuous
support, and mentorship during the course of my PhD degree. I am also grateful for Jeff Weaver’s
clearguidanceandtechnicalsupportonmyprojects. Addiitonally, Iwouldliketoexpressgratitude
for the advice and feedback of Jeffrey Nugent, Nicolas Duquette, Simone Schaner, Daniel Bennett,
Thomas Chaney, Marianne Andries, Robert Metcalfe, Simon Quach, and Augustin Bergeron. I am
honored to have received funding from JPAL’s Jobs and Opportunity Initiative during my PhD
studies and I would like to thank the team at JPAL for their continued assistance throughout the
funding process.
For this project to come to fruition, it took an amazing team of women in Saudi Arabia and I
am forever grateful for each and every one of them. Many thanks to Jawaher AlSudairy, Miznah
AlOmair, Chaza Abou Daher, Munira AlSharif, and Razan Al Khalis for the time, effort, and
diligent work throughout my partnership with AlNahda. I would also like to thank the anonymous
youngladiesofthefocusgroupswhoprovidedinvaluableinsightsontheSaudifemalelabormarket.
It seemed at first a challenge to run an experiment in a country that I have never been to but this
experience has been nothing but amazing and I am grateful for work relationships that have turned
into friendships. Additionally, I would like to thank the deans and professors at Princess Nourah
University and King Saud University for facilitating the execution of my project. Lastly, I would
like to thank Abdullah AlGhamdi and Sara AlAmasi at Education for Employment (EFE) for their
assistance and cooperation.
MysuccessduringthisPhDprogramwouldnothavebeenpossiblewithoutmyfriendsAmyMahler
and Fatou (Fakinette) Thioune. I am blessed with the best of friends who were there to celebrate
it all, the good and the bad. From (extremely) long walks and coffee breaks with Amy to carpools
ii
and cooking breaks with Fatou, I take pride in knowing that although the past few years were
challenging, we also laughed the most and got to discover and enjoy LA the best way we could. My
appreciation also goes out to Rachel Lee, Jung Hyuk Lee, Tao Chen, Karim Fajury, Yi-Ju Hung,
Juan Espinosa Rajat Kochhar, Ruozi Song, and Francesco Gabrielle for great times spent together,
whether in KAP or out and about in LA. I would like to thank Andreas Aristidou, for not only
being a supportive friend (and ”Hand”), but also an upper classmate with all the advice on how
to successfully navigate a PhD program - from the core exams to submitting the dissertation.
No words can express my gratitude for my family. I am simply lucky to have been raised and loved
by incredibly supportive parents, Abdullah and Wafaa, who taught me all the values that helped
me power through this journey. I’m grateful for my siblings: Anwar, Nasser, Hussain, Dalal, and
Anfal for their belief in me, as well as my niece and nephew: Sheikha and Aziz, for being my cutest
supporters. Lastly, I am immensely grateful for my husband Bader and his undeniable support,
patience, and encouragement throughout my studies and beyond.
iii
Table of Contents
Acknowledgments ............................................................................. ii
List of Tables.................................................................................. vii
List of Figures................................................................................. viii
Abstract....................................................................................... ix
1 Social Norms and Women’s Job Preferences in Saudi Arabia............................... 1
Chapter One: Introduction ....................................................................... 1
Chapter One: Context and Qualitative Interviews ............................................. 4
Chapter One: Partnerships and Survey Executions ............................................ 7
Al Nahda Society and the Absher Survey of Men’s Perceptions......................... 7
Information Experiment at PNU and KSU................................................ 7
Chapter One: Eliciting Job Preferences: An Incentivized Choice Experiment ............... 8
Chapter One: Experimental Design.............................................................. 9
Overview ..................................................................................... 9
Treatment.................................................................................... 10
Outcome Variables .......................................................................... 11
Chapter One: Empirical Strategy................................................................ 12
Chapter One: Results............................................................................. 13
Chapter One: Conclusion......................................................................... 15
Chapter One: Figures............................................................................. 16
Chapter One: Tables.............................................................................. 19
iv
2 Eliciting Women’s Job Valuations:
An Incentivized Choice Experiment........................................................ 28
Chapter Two: Introduction....................................................................... 28
Chapter Two: Experimental Design ............................................................. 31
Discrete Choice Experiment ................................................................ 32
Chapter Two: Data ............................................................................... 33
Part-time Work.............................................................................. 33
Remote Work ................................................................................ 34
Segregated Workplace....................................................................... 34
Paid Maternity Leave ....................................................................... 34
Free Child Care Services.................................................................... 35
Chapter Two: Results............................................................................. 35
Elicited Valuations .......................................................................... 35
Heterogeneity in Valuations ................................................................ 36
Counterfactuals.............................................................................. 37
Chapter Two: Conclusion......................................................................... 38
Chapter Two: Figures............................................................................. 39
Chapter Two: Tables.............................................................................. 42
3 Evolving Gender Dynamics and Female Labor Outcomes: Evaluating the Female Driving
Policy in Saudi Arabia..................................................................... 45
Chapter Three: Introduction ..................................................................... 45
Chapter Three: Background...................................................................... 48
Chapter Three: Data.............................................................................. 50
Chapter Three: Empirical Strategy.............................................................. 51
Chapter Three: Results........................................................................... 54
Female Employment Outcomes............................................................. 54
Triple Difference Estimation................................................................ 55
Threats to Identification: Parallel Trends Assumption Test ............................. 55
Chapter Three: Conclusion....................................................................... 56
v
Chapter Three: Figures........................................................................... 59
Chapter Three: Tables............................................................................ 68
References..................................................................................... 72
vi
List of Tables
1.1 Treatments................................................................................... 19
1.2 Outcome Variables .......................................................................... 20
1.3 Balance Table................................................................................ 21
1.4 Aspiration Outcomes: Extensive Margin ................................................. 22
1.5 Aspirations By Underestimating Perceptions (Median) ................................. 23
1.6 Aspiration Outcomes: Intensive Margin .................................................. 24
1.7 Aspirations By Underestimating Perceptions (Median) ................................. 25
1.8 Elicited Valuation Outcomes ............................................................... 26
1.9 Valuations By Underestimating Perceptions (Median) .................................. 27
2.1 Descriptive Statistics ....................................................................... 42
2.2 Mean Valuations by College ............................................................... 43
2.3 Determinantes of Job Choices: Extensive vs Intensive Margin .......................... 44
3.1 Opening of Female Driving Schools in Saudi Arabia ..................................... 68
3.2 Difference-in-Difference Estimation ....................................................... 69
3.3 Difference-in-Difference Estimation - Males .............................................. 70
3.4 Parallel Trends Assumption Test .......................................................... 71
vii
List of Figures
1.1 Timeline...................................................................................... 16
1.2 Guessing Incorrectly......................................................................... 17
1.3 Guessing Correctly ......................................................................... 18
2.1 Part-time Valuation ......................................................................... 39
2.2 Remote Valuation ........................................................................... 39
2.3 Segregated Workplace Valuation ........................................................... 40
2.4 Maternity Leave Valuation.................................................................. 40
2.5 Free Child Care Valuation .................................................................. 41
3.1 Number of Driving Schools as of 2019 - Q4 ............................................... 59
3.2 Percentage of Saudi Females Working ..................................................... 60
3.3 Total Saudi Females Working............................................................... 61
3.4 Total Saudi Males Working................................................................. 62
3.5 Difference-in-Difference Model (1) ......................................................... 63
3.6 Difference-in-Difference Model (1) by Groups of Treatment ............................. 64
3.7 Difference-in-Difference Model (2) ......................................................... 65
3.8 Difference-in-Difference Model (2) by Groups of Treatment ............................. 66
3.9 Parallel Trends Assumption Test........................................................... 67
viii
Abstract
This dissertation contributes to our understanding of low female labor force participation rates in
developing countries - specifically the Middle East and North Africa (MENA) region. In the three
chapters, I specifically study female labor force participation in Saudi Arabia and the barriers that
women may face when making labor market decisions. In Chapter 1
1
, I investigate how marriage
market norms are misperceived by young women through a randomized controlled trial and I study
howcorrectingthesemisperceptionsaffectstheiremploymentaspirationsandbehaviors. InChapter
2, I employ an incentivized choice experiment to elicit young women’s valuations for different non-
wage job characteristics to calculate compensating wage differentials that may imrove matching
in the labor market. In Chapter 3, I exploit the staggered adoption of the female driving policy
in Saudi Arabia as a natural experiment to understand the impact of women’s independence and
work accessibility on their employment outcomes.
1
Chapters 1 and 2 were funded by Abdul Latif Jameel Poverty Action Lab (JPAL)’s Jobs and Opportunity
Initiative and received Internal Review Board (IRB) approval from University of Southern California (UP-21-00976-
AM001), Princess Nourah University (22-0062E), and King Saud University (KSU-HE-23-102).
ix
Chapter 1
Social Norms and Women’s Job Preferences in Saudi Arabia
Chapter One: Introduction
For some women in the developing world, the decision to join the labor force comes with the high
cost of making difficult household decisions about marriage, fertility, and child care. For those who
come from more conservative societies, this decision is further challenged by social norms, family
expectations, and cultural influence that may discourage even the most motivated of them. This
is especially the case for countries in the Middle East and North Africa (MENA) region. Based on
the Global Gender Gap report of the World Economic Forum (2022), the MENA countries rank
among the lowest in the Global Gender Gap (GGG) Index as well as the Economic Participation
and Opportunity (EPO) sub-index. One of the largest countries with the lowest female labor force
participation (FLFP) rates in MENA is the Kingdom of Saudi Arabia. The country’s GGG index
ranks 127 out of 146 countries and its EPO sub-index ranks 128 out of 156. Although Saudi
Arabia has initiated several policies over the past decade, creating a more friendly environment
for the female worker in the job market and easing several restrictions such as the male guardian’s
approval to work and not allowing women to drive, the FLFP rate still remains low in the country
at 32.3% for the female population aged 15 years and above in 2021
1
. This raises several questions
about women’s preferences and the restrictions they face that still keep them from working despite
the available opportunities and resources. With Saudi Arabia being one of the largest and most
conservative countries in the region, understanding these womens’ preferences and challenges gives
a solid foundation to build new policies and improve job opportunities for women not only within
1
General Authority for Statistics (2021) - Saudi Arabia
1
the country but for other countries in MENA and beyond.
Thispaper’smainobjectiveistounderstandtherolethatsocialnormsplayinyoungwomen’slabor
market decisions - especially marriage market norms and how working women are perceived by so-
ciety. Are young women misinformed about working women’s marriage prospects? Does correcting
their beliefs influence their employment aspirations and choices? This project gives an opportunity
to understand what employment choices a woman believes are socially appropriate to still preserve
her reputation as a desirable candidate for marriage. To answer these questions, I first survey
2
a nationally representative sample of Saudi men about their preferences when it comes to their
wives/future wives working. Then I run a randomized controlled trial on a sample 500
3
female col-
lege seniors at two public universities in Riyadh, Saudi Arabia: Princess Nourah University (PNU)
and King Saud University (KSU). This survey first starts by asking the young women about what
they believe young Saudi men desire in a future wife when it comes to employment choices. These
beliefsarethencorrected(orconfirmed)byactualstatisticsfromthemen’ssurvey. Afterrandomly
providingthisnewinformationtotheseyoungwomen, thesurveythenasksfurtherquestionsabout
their aspirations and attitudes towards certain job characteristics in order to observe the effect of
providing this information. In addition to changes in attitudes, their behaviors are then studied by
collecting their data through Education for Employment, a job matching agency in Riyadh that I
have partnered with to get access to job vacancies in the private sector. Through EFE, I observe
behavior on job applications and employment in order to see how salient these social norms are
and how easily they can be changed. I also administer follow-up surveys around eight months after
each cohort’s graduation to ask questions about their employment status as well as marital status.
The second objective of this project is to understand Saudi women’s job preferences and precisely
measure how much they value specific job characteristics. This was executed in an incentivized
choice setting where survey participants are given several pairs of job descriptions to choose from
and are given the option to be considered for the jobs the prefer. This is made possible through
the collaboration with EFE. The desired outcome of this part of the survey is, through the choices
they make, to be able to generate a profile for each applicant with their ideal job characteristics.
2
This survey is administered through Absher, a Saudi government services platform and was executed through
my partnership with AlNahda Society.
3
The target sample size of this study is 1500 college seniors and data collection is still ongoing.
2
This helps elicit monetary values they place for specific job qualities - such as valuing flexible work
hours or segregated work spaces. This provides a better understanding of what the young female
job seeker in this society is looking for in a job in order to better treat the current labor market
frictions. This section of the project is explained and analyzed in detail in Chapter Two.
With the sample of 500 female seniors, I document several key results. First, 90% of young women
underestimate how acceptable men find it for their wives to work. The median guessed estimation
is 20 percentage points below the actual percentage. Second, I find that once women’s perceptions
are corrected, their work aspirations and valuations of different job characteristics change. As for
the treatment effects, I study these effect on three sets of outcomes: extensive and intensive margin
outcomes, as well as elicited valuations of job characteristics. I do not find an overall treatment
effect on extensive margin outcomes (likelihood of applying for a job, reservation wage, etc). In
the heterogeneity analysis however, I do find that on the extensive margin, overestimators report a
27% higher reservation wage as a result of the treatment. On the intensive margin, the treatment
has a 5% statistically significant effect on the likelihood of apply for an office job as opposed to a
remotejob. Moreover, underestimatorsreporta6%, 9%, and8.7%higherlikelihoodofapplyingfor
full-time, office, and nonsegregated jobs after correcting their beliefs. As for elicited valuations, the
treatment has no effect on women’s valuations of different job characteristics. Under-estimators,
however, value segregated workplaces by 1108 SARs less as a result of the treatment.
This project relates to a vast and growing literature on social norms and female labor supply
choices and the role of information in reshaping these attitudes. The role that cultural norms
play in explaining gender inequality in developing countries is discussed in Jayachandran (2015)
and Jayachandran (2021), with examples of policies that could help remove such barriers. Goldin
(2014) and Fern´ andez and Fogli (2009) also show how culture and social norms impact female
labor force participation. Dhar et al. (2019) observe intergenerational transmission of norms about
gender roles in India and Dhar et al. (2018) explore the effect of an information experiment on
gender attitudes in adolescents. Coffman et al. (2017) find that providing social information has
an effect on high-stakes decisions and that these effects persist over time. Specific to Saudi Arabia,
several experiments have been implemented to study this issue. Aloud et al. (2020) and Bursztyn
et al. (2020) have shown that social norms and information gaps about working women are the
3
main reason why women still choose to stay unemployed. Miller et al. (2019) use data on firms in
Saudi Arabia to show that firms find it costly to integrate both men and women if the number of
female workers in their organization is below a certain threshold.
Icontributetothisliteratureintwoways. First, theeffectofinformationonsocialnormsisstudied
from a different angle. Previous papers have shown that women’s decisions could be influenced by
learningaboutotherpeoples’socialacceptanceofemploymentchoices-suchastheirpeersandhow
this affects their own employment decisions. This project focuses on marriage market returns to
labor supply and how updating women’s knowledge about what men find acceptable in a potential
wife may influence their employment decisions. This sheds light on how peoples’ perspectives of a
”good reputation” change as the country undergoes social change and how providing information
aboutthesechangescanaffectwomen’schoices. Second,mostoftheresearchincountrieslikeSaudi
Arabia has studied the demand side of the labor market when it comes female labor participation,
believing that the root of the problem may be stemming from the work environments and the
unequal opportunities between genders. This paper aims at studying the labor supply side by
understanding the preferences of women in these societies and whether their perceptions of social
norms impacts their valuations of different job characteristics.
Chapter One: Context and Qualitative Interviews
To begin this project, focus groups with several Saudi female job-seekers were arranged to better
understandthecontextandasksomedetailedquestionsabouttheirbeliefsandperceptionstowards
women working in the Saudi labor market. Young university graduates were sourced and requested
to join the focus group sessions
4
. These participants come from different cities, backgrounds, uni-
versities - with some coming from conservative families and others more progressive. Interestingly,
the request to arrange focus groups of around 4-5 young women to discuss this topic was politely
declined by almost all of the participants. This confirms the social stigma around this topic since
young women prefer not to share their opinions with people within their society.
4
Each participant was rewarded with 55 Saudi Riyals worth of gift cards to Saudi Arabian online shopping
platforms.
4
Thequestionsaskedintheseinterviewsincludedpersonaldemographicquestionssuchasage,family,
and marital status, questions about education and level of experience, job market preferences and
what the ideal job would look like, and finally their opinions about the constraints holding women
back from working. The potential job-seekers
5
were all recent university graduates between 23-24
years of age coming from the Eastern Province (Al-Sharqiya), Al-Dammam, and Riyadh.
To understand the role that family plays on these young women’s decisions and the possibility
of intergenerational norm transmission, some questions were asked about parents’ education, em-
ployment, religiosity, and level of involvement in their job market choices. From their answers, it
seemslikethosewhocomefrommorereligiousandconservativehouseholdsrefertotheirfathersfor
advice and approval when it comes to such decisions. It also happens that these women attended
all-female universities and their mothers work in all female jobs, mostly in the education sector.
According to one of the participants, the reason for her fathers’ involvement is that ”he is more
aware of the different industries and the environments that will be socially acceptable for her to
work in”. When asked about the risks of working in environments that may be male-dominated
or not socially acceptable, the repeated answer is about harming their reputation as young single
women in society, their marriage prospects, and also committing blasphemy by not complying with
the laws of religion. As for the participants from more progressive households, they usually discuss
with their friends and family members who are currently in the market for advice while the parents
are generally supportive of their final decisions.
When asked with the question ”what do you think is the main restriction for women when joining
thelaborforce?”,theanswersagainvariedbyindividuals’backgroundsaswellasthecitiestheylive
in. Those from progressive families and living in urban cities believe there really is not much of a
social restriction nowadays and the problem may be that there aren’t enough jobs to meet the new
level of demand. However, most of the respondents stated that even with the fundamental changes
happening over the past decade, women are still not used to talking to strange men or being with
themforlongperiodsoftime. Ifthejobsdonotprovidetheprivacyandcomfortthatwomenprefer
and have been accustomed to, they would rather not work until the ”perfect job” presents itself.
5
Job-seekers in this context does not necessarily mean that they are actively looking for a job but are interested
in applying/working within the next 6-12 months.
5
An important feature of that perfect job is to have a separate floor or area for women only, where
they can talk freely and even remove their hejab
6
if they wish. These answers give rise to several
important questions on women’s preferences and the need to meet them. If women are willing to
workandareactivelyseekingforjobs, andjobopportunitiesarebecomingincreasinglyavailablefor
young educated women, then understanding women’s preferences more deeply will prove valuable
for employers and policymakers that would help improve women’s participation.
Marriage is the prominent issue when it comes to job market decisions, as most of the participants
claimed. It is not only important to keep traditional wife roles in mind when applying for a job
(evenifstillsingle),butitisalsoimportanttofindajobthatmaintainsayoungwomen’sreputation
and desirability to improve marriage prospects. If such job opportunities are not available, then
it is safer to not work at all. For example, one of the respondents explained that ”if a man is to
choose a wife, he will definitely prefer that she does not work in an environment with other men
- even if she is okay with it”. Another respondent explains that ”men would also prefer that the
wife works in the public sector where the hours are shorter and more flexible and the pay is higher
than the private sector. In that case, women will look for any job, not necessarily one that would
match her degree or qualifications, but at least she guarantees that she signals her preparedness for
marriage and can fulfil her duties as a wife.” Interestingly, regardless of the respondents’ income
status, when asked if they expect to contribute to household expenses when they do get married,
they all agreed that even if they have a job and it pays well, it is the man’s duty to provide for
the family. It is therefore concluded that the rising day-to-day expenses in Saudi Arabia and the
possibility of sharing costs with the husband in the future is not a factor to consider when applying
(or not applying) for a job.
The last round of questions bridges the different conclusions thus far by asking, why do you think
menhavethesepreferencesforafuturewifeandhowdoyouknowthatthisisthecase? Theanswer
wasunanimousthatthisiswhattheyhavebeenraisedtobelieveandthisiswhatthey’veseenwith
their parents, older sisters and older cousins. This confirms that these norms have been transferred
to these women from their parents or family members of an older generation, and have never been
questioned even after the severe changes society has gone through over the last decade. This raises
6
Hejab: head covering worn by Muslim women when around strange men
6
the main research question of this paper, are women misinformed about their marriage prospects if
they chooose to work? Does correcting their beliefs change their attitudes and behaviors towards
employment? Will they value nonwage job characetristics differently?
Chapter One: Partnerships and Survey Executions
Al Nahda Society and the Absher Survey of Men’s Perceptions
This project is executed in partnership with Al Nahda Society in Riyadh, Saudi Arabia. Al Nahda
is a non-profit organization that runs projects and programs aimed at empowering women socially
and economically. In partnership with AlNahda, we administered a national survey of 2300 men’s
preferences and opinions regarding their wives/future wives working. This survey was distributed
through “Absher”, an online government services platform, which is used by everyone in Saudi
Arabia to complete different government-related tasks such as passport renewal and government
paperwork. The survey data was weighted (by age) so that the sample is representative of the
underlying population. We find that on average, 76% of men at least agree with the different
statements shown to them. Table 1.1 shows the statements that were shown to these men.
Information Experiment at PNU and KSU
The main experiment was conducted through a survey of senior undergraduate females at the
PrincessNourahBintAbdulrahmanUniversity(PNU)andKingSaudUniversity(KSU)inRiyadh,
SaudiArabia. Thistargetedsampleisidealforthepurposeofthisexperimentbecausethesewomen
are surveyed during their final semester at the university, making them potential job-seekers in the
next following months. The target sample is 1500 participants from five different colleges: College
of Business Administration, College of Computer Sciences, College of Law, College of Languages,
and College of Social Work. These specific colleges were selected due to the availability of job
opportunities that we were able to provide to them during the survey. The sample size is chosen
based on power calculations for two treatment arms and several standardized minimum detectable
7
effects (MDE) found in the literature for the different desired outcomes. As for the execution, the
universities send out the survey link to the targeted sample via email, explaining the scope of the
study (Saudi women’s social and economic role in the country) and the rewards they would be
receiving for participating in the study. Each participants receives a 25-SAR (7 USD) Amazon gift
card for completing the survey as well as an opportunity to win one of four 1500-SAR (400 USD)
cash prizes at the end of the study.
As previously mentioned, this research project was conducted in partnership with Education for
Employment (EFE) in Riyadh, Saudi Arabia. EFE is a nonprofit organization which provides
training for youth job seekers as well as match them with job opportunities across the MENA
region. By linking the survey participants to EFE at the end of the survey, data is collected on
thesejob-seekers’jobsearchbehavior,availableoptionsatthetimeofsearch,andfinaljobdecisions.
This also facilitates follow-up surveys after the experiment is fully executed. The follwoing sections
explain in detail the two main objectives of this research project and how the survey will embed
the two experiments to answer the relevant research questions.
Chapter One: Eliciting Job Preferences: An Incentivized Choice
Experiment
It is clear that even with the structural changes happening in Saudi Arabia over the recent years,
women still believe in somewhat traditional gender roles in the household and perhaps the current
available jobs do not cater to these preferences. By placing a value on specific job characteristics,
it can be very useful in providing jobs for women as the Saudi female labor market expands. The
way to elicit these job seekers’ preferences is to provide pairs of different available job vacancies
and their characteristics and extract their willingness to pay for different qualities of a job. This
includes full-time vs part-time jobs, segregated workplace, office vs remote work, paid maternity
leave, and child-care support. This idea follows the work of Mas and Pallais (2017), Alfnes and
Rickersten(2013)andBelotetal.(2018). Byprovidingalistofdifferentpairsofjobsandobserving
their choices, a profile is created for each respondent with all their preferred job characteristics.
Moreover, this method is used to place a value on specific characteristics. For example: if two jobs
8
are presented with the exact same characteristics and job description but one provides a women-
only work environment and pays 3000 Saudi Riyals less than, then it is concluded that women
”would pay” 3000 SARs a month to be able to work in a female only environment. The same
logic applies to placing a value on other characteristics such as on-site childcare or flexible working
hours. ChapterTwoexplainsdelvesintothisincentivizedchoiceexperimentandprovidesadetailed
analysis. For this chapter, the elicited valuations of are used as a set of outcome variables.
Chapter One: Experimental Design
Overview
The results from the men’s Absher survey were used in the information treatment arms of the RCT
on female students at PNU and KSU, where the survey is distributed to the students by email from
theuniversity.
7
EFEthenprovidestheprivatesectorjobpostingstoshowtothestudentsinthejob
preference elicitation section of the experiment. EFE also collects data on students who apply for
these jobs and records their employment outcomes. Lastly, Al Nahda Society administers follow-up
surveys to learn about the students’ socioeconomic status after graduation. The pre-analysis plan
for this RCT was pre-registered in the AEA registry
8
.
Thisstudyconsistsoftwosurveyrounds–themainsurveyiswherebaselineinformationiscollected,
the information treatment is randomly given to the participants, and the first set of outcomes is
collected. This survey is administered starting the summer semester of 2022 until spring semester
2024 on senior graduating students in 5 colleges (College of Business Administration, College of
Law, College of Computer Sciences and Information Technology, College of Languages, and College
of Social Work) at PNU and KSU. Finally, around 8 months after each cohorts’ responses are
collected, the follow-up survey is administered by phone to collect the remaining outcome variables
of the experiment. Figure 1.1 shows the planned timeline and the key features of two surveys.
7
At PNU, this survey was posted on each students’ Blackboard account and participation was mandatory.
8
AlRakhis, Monira. 2023. ”Social Norms and Women’s Job Preferences in Saudi Arabia.” AEA RCT Registry.
February 25. https://doi.org/10.1257/rct.9403-2.0
9
Treatment
A target sample of 1500
9
senior undergraduate students are randomly assigned into control and
treatment groups. Randomization is assigned at the individual level and is not stratified. The con-
trol group has 500 students, treatment groups T1 and T2 have 500 students each. At baseline, the
students are asked about their demographic characteristics, education and experience, ideal spouse
characteristics, and their work plans/aspirations after graduation. As mentioned, a representative
survey of Saudi Arabian men was previously conducted to collect men’s marriage preferences and
their attitudes towards their wives/future wives working. All students (treatment and control) are
shown the statements that were included in the men’s survey and are asked to guess what percent-
age of young, educated Saudi Arabian men agreed with these statements. To incentivize this step
and receive their most accurate perceptions, each participant enters a draw to win one of four 1500
SAR cash prizes for completing the survey and guessing these questions correctly increases their
chances of winning that cash prize. Each statement is shown individually and guesses are collected
before showing the next statement.
In the treatment groups, after collecting students’ guesses for all statements shown to them, the
statements are shown again with the actual percentages of men who agreed with the statements.
A figure is displayed showing how far they are from the correct answer for each statement. If they
scored within 5 percentage points of the correct answer, a statement is displayed telling them their
answer was correct and that now they have gained one more chance to win the cash prize. Figures
1.2 and 1.3 show different examples of what treated participants will see after responding to the
questions.
The difference between T1 and T2 is the type of statements that are shown. In T1, the statements
areaboutmen’sacceptabilityoftheirwivesworkingandwhetherworkisperceivedtoaffectwomen’s
roles as wives and mothers. In T2, in addition to the statements that were provided in T1, I
show statements about men’s acceptability of their wives working in specific (unconventional) work
environments such as non-segregated workplaces and leaving the house for work. Table 1.1 shows
9
This version of the paper displays results from 500 respondents as data collection is still ongoing.
10
thestatementsshowntoeachgroupandTable1.3showsthattreatmentandcontrolgroupcovariates
are balanced, which confirms that the randomization process of the treatment was successful.
Outcome Variables
As previously mentioned, there are different sets of outcomes that correspond to different parts
of the study. The first set of outcomes, collected in the first round, is related to aspirations and
preferences. In the career aspirations questions, I ask about their likelihoods of applying for a
job and their reservation wage (extensive margin outcomes) and their likelihood for applying for
different job types (intensive margin outcomes). Then, in an incentivized choice setting, I provide
pairs of real job vacancies to choose from (based on their qualifications) to elicit their valuations
for the following non-wage characteristics: segregated workplace, remote work, part-time
work, option to extend paid maternity leave, and option for free childcare close to
the workplace. I incentivize this by informing them that one of their choices will be selected at
random and our partner organization (EFE) will send them the full job description of that job and
can assist with applying for the job. I observe if the information provided in the treatment affects
the preferences and thus the valuations of the treatment groups relative to the control.
The second set of outcomes is related to behaviors and career-related choices. At the end of the
survey,ImentionthattheorganizationthatIpartneredwith(EFE)cansendthememailswhenever
they receive new job vacancies, and I observe whether they choose to sign up for these emails over
a subscription to newsletters/offers from a popular e-commerce website in Saudi Arabia. Then, I
tell the respondents that we are providing them with 5 more tickets, which they can allocate to the
1500 SAR lottery or to getting the full job description of a second job posting of their choice
(in addition to the first one already randomly selected). Lastly, to understand the salience of the
treatment effect, I observe their employment choices, which can be made through EFE or through
other employment opportunities. For jobs through EFE, the organization collects data on their
behavior throughout the study period. This data includes whether they apply for a job, attend the
interview, accept the offer, and if they applied/attended any of the job training programs that the
organization provides. For other employment choices, we administer a follow-up survey around 8
months after each cohorts’ graduation to ask whether they applied/searched for jobs, are employed
11
(if so, what are the characteristics of the job), and we also ask about their marriage outcomes (if
married, what are their spouse characteristics and do they have children).
10
Table 1.2 displays the
outcome variables and categorizes them into primary and secondary outcomes.
For this version of the paper, since I still have not reached the targeted sample size, I pool the two
treatments (T1 and T2) and do the analysis on treatment versus control to increase power. Also,
since the survey is still ongoing, I have yet to report results on actual behavior changes from EFE
and the follow-up surveys. The results so far will show aspiration and preference outcomes on the
extensive and intensive margin for a sample of 500 college seniors.
Chapter One: Empirical Strategy
Themainpartoftheanalysisistoestimatetheimpactoftheinformationtreatmentontheoutcome
variables. The main specification is as follows:
Y
isc
=α 0
+α i
Treat
i
+X
i
′
β +γ s
+θ c
+ϵ isc
(1.1)
where Y
i
sc is the outcome variable for respondent i, Treat
i
is an indicator variable equal to 1 if
respondent i is treated, and X
i
is a vector of control variables. Estimated results are reported with
control variables selected by Double Lasso. The control variables include personal demographics
and background (age, family income, parents’ education) and educational attainment (type of high
school, major, GPA, work experience). I first estimate the results without fixed effects. However,
since students belong to different colleges and graduate at different times, I include college γ s
and
cohort fixed effects as well θ c
– which is also used to increase statistical power at this stage of the
analysis.
I estimate the effects of each treatment arm separately (control vs T1, control vs T2) to evaluate
whether the type of information we reveal creates a differential effect across treatment groups. We
10
Outcomevariableswithlimitedvariation(wheremorethan95%ofobservationshavethesamevalue)areomitted
from the analysis.
12
then pool T1 and T2 (control vs T) to estimate the overall effect of being treated with information
on men’s preferences. The results of this version of the paper only show the pooled treatment
effects since I am still underpowered to evaluate the treatment arms separately.
Heterogeneous Effects
Since I assume that the treatment effect comes from the correction of beliefs by providing informa-
tion on men’s perceptions, I evaluate the differential impact of the treatment by the misperception,
measured as the difference between students’ prior beliefs about men’s perceptions and the actual
perceptions. The assumption is that the treatment would have a higher effect on women who
believe men are not supportive of women working (and so for whom the misperception is larger
at baseline). I calculate this misperception by a dummy variable that is equal to 1 if the average
guesses for all 8 statements is below the median guess of all female respondents (-21%) and equal
to 0 if it is above the median. I employ this strategy to preserve power, since 90% of women under-
estimated these percentages and the number of overestimators is not large enough if the dummy
variable threshold is at 0. The specification in this case is as follows:
Y
isc
=α 0
+α 1
Treat
i
+α 2
Cov
i
+α 3
Treat
i
xCov
i
+X
i
′
β +γ s
+θ c
+ϵ isc
(1.2)
where Y
isc
is the outcome variable for respondent i, Treat
i
is an indicator variable equal to 1 if
respondent i is treated, Cov
i
is the heterogeneity covariate of interest - in this case it is a dummy
variableformisperception,Treat
i
xCov
i
isthevariableofinterestindicatingtheinteractionbetween
treatment and the covariate, and X
i
is a vector of control variables. As with the main specification,
this is measured using college and cohort fixed effects and controls are selected using Double Lasso.
Chapter One: Results
Tables 1.4 and 1.5 display the results of the pooled treatment on the extensive margin aspiration
outcomes: likelihood of applying for a job, reservation wage, and the number of lottery tickets
13
they would allocate to get a second job posting from EFE. From the results of Table 1.4, I cannot
conclude that the treatment had a statistically significant effect on this set of outcomes. However,
in Table 1.5, I find in Column (2) that this treatment had a statistically significant effect on
overestimators. More precisely, those who overestimated the number of men who agreed with the
shown statements – ie. were under the misperception that men are more accepting than they
actually are, report a higher reservation wage than their counterparts, and this wage is 27% higher
than the mean reservation wage of the control group. This confirms that women do internalize the
norms around marriage when making labor market decisions and when their beliefs are corrected,
they require a higher reservation wage to offset the social cost.
For the intensive margin outcomes, Tables 1.6 and 1.7 display the treatment effect on women’s
likelihoods of applying for different types of jobs. Column (2) of Table 1.6 shows that there is an
overall treatment effect on the likelihood of applying for an office job (as opposed to a remote job).
In other words, young women who are provided with this information treatment are 5.5% more
likely to apply for a office job than those who did not receive the treatment. Table 1.7 shows the
results of underestimators and the differential impact this treatment has had on their aspirations.
Columns (1), (2), and (3), show that treated women who underestimated are more likely to apply
for a full-time, office, and non-segregated job by 6%, 9% and 8.7% respectively. This is in line
with my prior hypothesis. The assumption here is that once young women learn that, contrary to
their priors, men actually prefer their wives to work, they will find that it is now more socially
acceptable to be a working wife. This leads to increased likelihoods of applying for job types that
were otherwise thought to be frowned upon such as: full-time job, an office job (away from the
home), and a non-segregated job where they may work and interact with strange men.
In terms of elicited valuations, I do not find an effect in the main regression specification in Table
1.8. However, inColumn(3)ofTable1.9, Ifindthatwomenwhounderestimatedthenormactually
value segregated workplaces by 1108 SARs less - which is consistent with the result from Table 1.7.
This result is in line with the previous results, showing that working in a segregated workplace is
not as valuable now that they have new information about men finding this more acceptable.
14
Chapter One: Conclusion
Low female labor force participation rates in the MENA region remain an issue of concern in
development research. By studying Saudi Arabian women’s preferences and choices, important
conclusions can be made about how women in this region of the world prefer to work, how mar-
riage market norms affect their decisions, and what can be done to improve their labor market
outcomes. In this paper, I find that women internalize the social cost of work and take into account
society’s (especially prospective husbands’) perceptions when making labor supply decisions. It is
particularly interesting that women who underestimate men’s acceptability of their wives working
in different job types change their job preferences after receiving the information treatment and are
now more likely to apply to jobs that otherwise seemed unacceptable. These results are important
for employers as such revelations can help improve labor market frictions. They are particularly
important for policymakers as well since the spread of information is a cost-effective method of
informing young women about society’s support of their career choices and labor decisions. Data
collection is still in progress, however, and results may change as I reach the targeted sample size.
Employment data from EFE as well as follow-up surveys are ongoing to analyze respondents’ job
choices after graduation in order to complete the last set of outcomes for this study.
15
Chapter One: Figures
Figure 1.1: Timeline
16
Figure 1.2: Guessing Incorrectly
Note: This figure is an example of what a respondent sees when she incorrectly guesses the percentage of men who
agreed with this statement: ”When I decide to marry, I prefer to marry a woman who is employed”.
The red arrow shows the distance between her guess (30%) and the actual percentage (83%).
17
Figure 1.3: Guessing Correctly
Note: This figure is an example of what a respondent sees when she correctly guesses (or guesses within 5 pp) the
percentage of men who agreed with this statement: ”When I decide to marry, I prefer to marry a woman who is
employed”. The red arrow shows how close her guess (80%) is to the actual percentage (83%), and the note informs
her that she guessed correctly and increased her chances of winning the cash prize.
18
Chapter One: Tables
Table 1.1: Treatments
Treatments
Control Group
What percentage of single, educated, high-income men do you think agreed with the following statement?
- When I decide to get married, I would prefer to marry a woman who is employed
- I think a woman is capable of being a good wife and mother while still having a successful career
- Women who have promising careers and who are financially independent are more desirable for marriage
than women who choose not to work
- My parents will be happy and supportive if my wife/future wife has a successful career
- It is acceptable if my wife/future wife has a full-time job
- If my wife or future wife wants to drive a car to go to work, I will encourage her
- It is acceptable for my wife to work in a place that may require her to go to the office every day
- It is acceptable if my wife/future wife works in a job that requires her to work or interact with other men
Treatment Group 1
Only statements directly related to work affecting women’s roles as wives/mothers are shown
What percentage of single, educated, high-income men do you think agreed with the following statement?
- When I decide to get married, I would prefer to marry a woman who is employed
- I think a woman is capable of being a good wife and mother while still having a successful career
- Women who have promising careers and who are financially independent are more desirable for marriage
than women who choose not to work
- My parents will be happy and supportive if my wife/future wife has a successful career
Treatment Group 2
In addition to statements from T1, statements about acceptability of women working in specific
work environments are included
What percentage of single, educated, high-income men do you think agreed with the following statement?
- When I decide to get married, I would prefer to marry a woman who is employed
- I think a woman is capable of being a good wife and mother while still having a successful career
- Women who have promising careers and who are financially independent are more desirable for marriage
than women who choose not to work
- My parents will be happy and supportive if my wife/future wife has a successful career
- It is acceptable if my wife/future wife has a full-time job
- If my wife or future wife wants to drive a car to go to work, I will encourage her
- It is acceptable for my wife to work in a place that may require her to go to the office every day
- It is acceptable if my wife/future wife works in a job that requires her to work or interact with other men
19
Table 1.2: Outcome Variables
Outcome Variables
1. Career Aspirations
Primary Outcomes
- Likelihood of applying for a job
Secondary Outcomes
- Index for applying for a job with more than 10 weeks paid maternity leave, a job with option for free childcare close to workplace
(T1 outcome)
- Index for: applying for a job in non-segregated workplace, full-time job, office job only (no remote)
(T2 outcome)
- Likelihood of applying for job at environment-friendly company (placebo)
2. Elicited Job Preferences
Primary Outcomes
- Students’ elicited valuation for each of the 5 non-wage characteristics is used as an outcome (segregated workplace, remote work,
part-time work, option for more paid maternity leave, free childcare close to work), used as values and ordered Probit
3. Career-related Choices
Primary Outcomes
- Lottery allocation: variable that measures how many “tickets” out of 5 does the student allocate to getting assistance from
EFE in applying for a second job of their choice
- EFE data collection: collects data on student choices and behaviors: what jobs they apply to, do they attend the interview,
do they get offers, and whether they accept the offer (between baseline survey and follow-up survey)
- Employment outcomes from follow-up survey: searched for jobs (binary yes/no), number of jobs applied for,
employed (binary yes/no), if employed (sector, industry, wage, non-wage characteristics)
Secondary Outcomes
- EFE Subscription: Binary (0/1) variable for whether the student chooses to sign up for offers and discount codes from “Noon.com”,
a popular E-commerce website in Saudi Arabia (0) or EFE newsletters that provide information on new job opportunities (1)
- Marriage outcomes from follow-up survey: married (binary yes/no), if yes: characteristics of husband, number of children
20
Table 1.3: Balance Table
Variable Control Group Treatment Group Difference
College of Business Admin. 0.249 0.299 0.050
(0.433) (0.458) (0.227)
College of Computer Sciences 0.401 0.394 -0.007
(0.492) (0.489) (0.869)
College of Languages 0.186 0.147 -0.040
(0.391) (0.354) (0.239)
College of Law 0.090 0.089 -0.001
(0.288) (0.285) (0.960)
College of Social Work 0.073 0.072 -0.002
(0.262) (0.259) (0.947)
GPA 4.497 4.560 0.063
(0.621) (0.517) (0.222)
Driver’s License 0.311 0.295 -0.016
(0.464) (0.457) (0.707)
Work Experience 0.119 0.133 0.014
(0.325) (0.340) (0.661)
Public School 0.695 0.688 -0.007
(0.462) (0.464) (0.869)
Job Excitement 9.239 9.275 0.036
(1.519) (1.465) (0.794)
Family Support for Job 9.301 9.370 0.069
(1.514) (1.594) (0.636)
Family Income 20,546.912 18,020.994 -2,525.918+
(15,170.916) (13,299.140) (0.076)
Mother’s Education 0.370 0.396 0.026
(0.484) (0.490) (0.569)
Mother Worked 0.197 0.180 -0.017
(0.399) (0.385) (0.693)
Married 0.098 0.073 -0.026
(0.299) (0.260) (0.316)
Observations 177 348 525
21
Table 1.4: Aspiration Outcomes: Extensive Margin
(1) (2) (3)
Apply for Job Reservation Wage EFE Lottery Tickets
Treatment 0.096 887.907 0.123
(0.182) (853.654) (0.169)
Y Mean 8.89 8328.28 2.11
Obs. 503 300 309
Standard errors in parentheses
Results shown with selected controls and college fixed effects
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
22
Table 1.5: Aspirations By Underestimating Perceptions (Median)
(1) (2) (3)
Apply for Job Reservation Wage EFE Lottery Tickets
Treatment -0.107 2230.928
∗ -0.064
(0.222) (1322.453) (0.239)
Treatment× Underestimate 0.410 -2638.377
∗ 0.344
(0.367) (1370.370) (0.337)
Y Mean 8.89 8328.28 2.11
F-Test 0.29 0.68 0.24
Obs. 503 300 309
Standard errors in parentheses
Results shown with selected controls and college fixed effects
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
23
Table 1.6: Aspiration Outcomes: Intensive Margin
(1) (2) (3) (4) (5)
Full-time Office Non-Seg. Mat. Leave Child Care
Treatment 0.218 0.481
∗∗∗ 0.162 0.048 -0.191
(0.179) (0.163) (0.257) (0.169) (0.237)
Y Mean 8.76 8.93 7.33 8.94 8.46
Obs. 503 503 503 503 503
Standard errors in parentheses
Results shown with selected controls and college fixed effects
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
24
Table 1.7: Aspirations By Underestimating Perceptions (Median)
(1) (2) (3) (4) (5)
Full-time Office Non-Seg. Mat. Leave Child Care
Treatment -0.117 0.125 -0.339 -0.286 -0.194
(0.253) (0.227) (0.335) (0.216) (0.336)
Treatment× Underestimate 0.647
∗ 0.684
∗∗ 0.983
∗ 0.649
∗ 0.013
(0.353) (0.317) (0.506) (0.335) (0.486)
Y Mean 8.76 8.93 7.33 8.94 8.46
F-Test 0.03 0.00 0.09 0.16 0.60
Obs. 503 503 503 503 503
Standard errors in parentheses
Results shown with selected controls and college fixed effects
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
25
Table 1.8: Elicited Valuation Outcomes
(1) (2) (3) (4) (5)
Part-time Work Remote Work Segregated Maternity Leave Child Care
Treatment 122.056 21.041 -478.901 347.589 73.517
(356.366) (284.691) (321.328) (238.608) (705.841)
Y Mean 1359.38 -507.58 1203.82 39.94 -3216.27
Obs. 466 485 463 495 488
Standard errors in parentheses
Results shown with selected controls and college fixed effects
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
26
Table 1.9: Valuations By Underestimating Perceptions (Median)
(1) (2) (3) (4) (5)
Part-time Work Remote Work Segregated Maternity Leave Child Care
Treatment 430.955 -464.956 208.535 80.784 146.111
(527.330) (433.513) (442.855) (306.262) (1034.253)
Treatment× U.E. -590.880 915.950 -1315.954
∗∗ 518.161 -90.932
(714.487) (599.460) (654.485) (482.595) (1424.401)
Y Mean 1359.38 -507.58 1203.82 39.94 -3216.27
F-Test 0.74 0.26 0.02 0.10 0.95
Obs. 466 485 463 495 488
Standard errors in parentheses
Results shown with selected controls and college fixed effects
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
27
Chapter 2
Eliciting Women’s Job Valuations:
An Incentivized Choice Experiment
Chapter Two: Introduction
Non-wage job attributes have played a central role in providing work-life balance for employees in
the private sector. However, we have yet to find how non-wage job attributes affect job-seeker’s
employment choices when presented with different alternatives. Understanding how job-seekers
value different work arrangements may provide insight on how to efficiently match job vacancies
with job-seekers. This is especially important for economists and policymakers alike when studying
emerging labor markets - such as the female labor markets in developing countries. Female labor
force participation rates remain low in developing countries. Some of these countries are keen on
improving the matching process in the female labor market to boost female employment. However,
in such contexts, the female demographic is one that has not been extensively studied in the labor
marketduetoitslowrepresentation. Itisplausiblethatpolicymakersarepromotingandregulating
different work arrangements that do not meet women’s desired job characteristics, which intensifies
labor market frictions and leads to inefficient outcomes.
One of the developing worlds’ lowest female labor force participation rates is in the Kingdom of
Saudi Arabia, with rates as low as 30% General Authority for Statistics (2021). With the recent
progressive policies promoting women’s social and economic representation, the country is on a
trajectory to support women’s political and economic representation . However, with the majority
ofemployeesbeingmeninthepast,littleisknownaboutwhatcharacteristicswomenwouldpreferin
28
ajobandwhatwouldmotivatethemtowork. Author-Yearshowthatthereasonbehindlowfemale
employment in Saudi Arabia is that employers find it too costly to provide a segregated workplace
to accommodate female workers - assuming that women would only work if the workplace is fully
segregated. However, this may not be the case as women value segregated workplaces differently
and we can consider compensating wage differentials to bridge this gap. This applies to other non-
wage job attributes that policymakers and employers may have overlooked due to limit available
data and research on women’s job preferences.
In this paper, I elicit young women’s valuations for five different job characteristics that are most
relevanttofemaleemploymentinthiscontext: part-timework, remotework, segregatedworkplace,
paid maternity leave, and free child care services. I survey 500 college seniors in Saudi Arabia as
part of a larger study on social norms and women’s job preferences in the country. I employ an
incentivized discrete choice experiment whereby I display pairs of real job vacancies that match
their qualifications and ask them to select their preferred job. The two jobs displayed are similar in
allaspectsexceptforwageandthenon-wageattributethatIaminterestedinvaluing. Iincentivize
1
truthful responses by telling them that one of their choices will be selected at random and we will
provide them with the full job description and assist them in the application process.
Since the sample I study is college seniors who were invited to take a survey on the social and
economic role of Saudi women (and not specifically job-seekers), I am able to reduce selection
bias and elicit estimates from a sample of women that is more representative of the young female
population. I find that young women value part-time work them most, followed by segregated
workplaces and maternity leave. Interestingly, they place a negative value on remote work and
free child care services. These results contradict popular belief that women would prefer to have
the convenience of working from home as well as having free child care services. Such results are
important for employers since resources may be reallocated in order to provide job opportunities
that better match women’s preferences. I also explore how women’s individual characteristics may
impact these valuations, especially when it comes to their perceptions of social norms and what job
is acceptable for a woman to be working in. I find that perceptions of the norms around women
1
Participation in this study is also incentivized by providing Amazon gift cards and entering a draw for a cash
prize upon completion.
29
working affect their decision to enter the workforce while individual characteristics such as family
background affect their valuations for non-wage job attributes.
This paper contributes to two strands of literature. First, it is closely related to Mas and Pallais
(2017), WiswallandZafar(2017), andBelotetal.(2018)wherethemaingoalistoelicitjobseekers’
valuations for different job attributes through discrete choice experiments. Mas and Pallais (2017)
estimate willingness to pay distributions for different work arrangements in a discrete choice exper-
iment using and find that workers do not prefer flexible work arrangements and are willing to pay
20% of their wage to avoid unpredictable work schedules and and 8% for the option to work from
home. Wiswall and Zafar (2017) use a stated preference approach by providing undergarduate
students with hypothetical jobs to elicit their valuations for a wide range of job characteristics.
Feld et al. (2022) experiment with different methods to elicit jobseekers valuations and find that
discrete choice experiments are most effective, which is the the approach used in this paper. My
main contribution to this body of literature is that the this experiment is incentivized using real
job vacancies that closely match the respondents’ valuations. This overcomes critique of hypothet-
ical settings as well as respondents’ inattention when job vacancies seem out of reach. Second,
the respondents are part of a larger study and were not invited to join the survey for job-seeking
purposes. This minimizes selection of job-seekers that may bias these valuations in the direction of
a more ”eager-to-work” portion of the population.
Second, I contribute to the literature on low female labor force participation rates in developing
countries by showing how elicited job valuations can improve job market frictions and boost female
employment. Multiple studies, including Jayachandran (2015) and Goldin (2014) have shown that
cultural norms play a major role in explaining gender inequality in developing countries. Specific
to Saudi Arabia, Aloud et al. (2020) and Bursztyn et al. (2020) have shown that information
gaps around social norms have played a major role in limiting female employment in the country.
Miller et al. (2019) find that due to the social norms of gender segregation, firms find it costly to
integrate women and provide segregated workplaces. I contribute to this literature by providing
young women’s elicited valuations for different job characteristics and provide interesting results on
what women in such cultures actually value. Using these valuations, we can improve job matching
efforts of policymakers and employers and provide insight on compensating wage differentials that
30
may attract young female workers into the workforce.
Chapter Two: Experimental Design
This survey experiment is part of a larger study that I conducted in 2022 to learn about marriage-
relatedsocialnormsinSaudiArabiaandhowtheyaffectwomen’slaborsupplydecisions. Isampled
500 female seniors at Princess Nourah University and King Saud University in Riyadh, Saudi
Arabia. The students were invited to join the study through an email announcement that was
sent out by their respective universities. The email announcement stated that the survey was
about the social and economic role of women in Saudi Arabia and we were interested in young
women’s opinions on the matter. They each received a link to the survey and completed it online.
To incentivize participation, the announcement stated that they will each receive a 25-SAR gift
card from Amazon Saudi Arabia and they will all enter a draw to win one of four 1500-SAR cash
prizes upon completion. At the beginning of the survey, I asked the respondents to share their
full names and contact information in order to reach out to them for a follow-up survey after their
graduation. The participants were not informed that the survey had a section about their labor
supply preferences nor that we will provide them with job opportunities until they reached that
section at the end of the survey. This is to minimize selection of those who were interested in
applying for a job at baseline.
To elicit the respondents’ valuations with the most accuracy, I further incentivized truthful re-
sponses by providing them with real job postings that matched their qualifications in a discrete
choice setting following the work of Amanda Pallais. I partnered with Education for Employment
(EFE) in Riyadh, a job matching agency that provided real private sector job opportunities at the
time of the study. The sample of students was selected based on the available jobs and therefore,
notallseniorstudentswereinvitedtoparticipate. SeniorstudentsfromCollegeofBusinessAdmin-
istration, College of Computer Sciences and Information Systems, College of Languages, College of
Law, and College of Social Work were the ones who were invited to participate due to the nature
of the job opportunities at the time. Since I know the majors and minimum Grade Point Averages
(GPA) that the employers are looking for, matching job postings would appear to each respondent
31
based on her specific major and GPA. The job postings would state that this employer is looking
to employ someone of the respondent’s qualification and shows the monthly wage in Saudi Riyals
and working hours, as well as the varying characteristic that we are interested in valuating. The
name of the company/employer and the details of the job description are not displayed to minimize
the effect that job titles/positions would have on their choices. Before showing them the list of job
postings, I tell them that one of their job choices will be selected at random and we will provide
them with its full job description (name of company, position, location) and assist them in the ap-
plication process through EFE (preparing resumes and applications, training for interviews, etc).
This is to capture their revealed preferences and incentivize attention during the survey.
Discrete Choice Experiment
I aimed at eliciting the respondents’ valuations for five different non-wage attributes: part-time
work, remote work, segregated workplace paid maternity leave, and free child care services. For
each attribute, each respondent was shown two job descriptions at a time and was asked to select
her preferred choice. The respondents were told the two job descriptions were similar in all aspects
except the monthly wage and the non-wage attribute. We displayed only two job descriptions at a
time: the baseline job and the alternative. Whenever the alternative job is not selected, we display
the next pair of jobs where the wage of the alternative job is slightly higher than the previous one.
The difference between the baseline wage and the alternative wage that the respondent eventually
selects is the elicited valuation of that non-wage attribute. For example, to elicited valuation of a
non-segregated workplace, the two job descriptions show a baseline job that is segregated and pays
8,000 SARs and an alternative job that is non-segregated that pays 7,000 SARs. If the baseline
job is selected, I increase the wage of the non-segregated job in the next pair of choices until the
alternative is selected. Suppose the respondent selects the non-segregated job at 13,000 SARs,
this means that the respondent dislikes non-segregated workplaces and would require a 5,000 SAR
compensating differential to select that job over a segregated one. If the respondent selects the
alternative job from the beginning, then she values that work arrangement the most and is willing
to pay 1,000 SARs to work in a non-segregated environment. For coding purposes, we cap this
valuation at an amount higher than the maximum valuation of each attribute. Similarly, if she
32
chooses the baseline job throughout the entire section and never switches, then she dislikes the
alternative the most and we cap this valuation at an amount lower than the least valuation of each
attribute. The job choices and increments differ based on the availability of jobs at the time the
respondentistakingthesurveyandherqualifications. Thissectionproducestheelicitedvaluations
for each non-wage attribute, which range from negative to positive values.
Chapter Two: Data
The discrete choice experiment described previously produces the key variables that I am inter-
ested in studying. Most job opportunities for college graduates offer similar wages and workplace
amenities. However,somemajorsareknowntogetofferedhighersalariesthanothers. Forexample,
positionsforbusiness, computerscience, andlawmajors(BCLmajors)offerhigherstartingsalaries
than positions for languages and social work majors (LSW majors). The increments I use to raise
the wage of the alternative job posting is also different depending on the major and availability of
jobs. After eliciting each respondents’ valuations for the five different non-wage characteristics, I
average out these valuations to understand the overall sentiment of young women in this context
towards these job arrangements. This section describes the non-wage job attributes in detail
2
.
Part-time Work
The baseline job for a fresh college graduate in this context is usually a full-time job. I display
a baseline full-time job that offers 8000 SARS for BCL majors and 6000 SARs for LSW majors.
The alternative part-time job starts with 10000 SARs for both majors and goes down in different
increments depending on the major until it reaches 2000 SARs.
2
The survey was first administered at PNU, after which some changes were made and more jobs were added
before administering it at KSU.
33
Remote Work
As with the full-time job attribute, working in an office is generally a baseline job characteristic for
fresh graduates. I display a baseline full-time, in-office job that offers 8000 SARs for BCL majors
and5000SARSforLSWmajors. Thealternativejobisaremoteworkopportunitythatstartswith
15000 SARs for both major categories. This job goes down in different increments until it reaches
2000 SARs.
Segregated Workplace
Generally for these non-wage attributes, I elicit the valuation of the varying/alternative work
arrangement because I assume it is the desired job attribute that women in this context would
prefer. However, for segregated vs nonsegregated workplaces, even though segregated workplace is
the baseline characteristic of all jobs, it is actually the non-wage attribute that I am interested in
valuing and not the alternative non-segregated arrangement. In this section I display a full-time,
in-office, segregated (women only) job that offers 8000 SARs for BCL majors and 5000 SARS for
LSWmajors. Thealternativejobisanon-segregatedworkopportunitythatstartswith7000SARs
for BCL majors and 4000 SARs for LSW majors. This job goes up in different increments until it
reaches 15000 SARs for both major categories.
Paid Maternity Leave
As per Article 151 of the Saudi Labor Law, every working woman is entitled to 10 weeks of paid
maternityleaveatatime. Toelicitwomen’svaluationsfortheoptiontoextendthispaidmaternity
leave,Iprovidepairsofjobswherethebaselineis10weeksandthealternativeisajobthatprovides
theoptiontoextendthisperiodbeyond10weeks. Idisplayabaseline10-weekpaidmaternityleave
job that offers 8000 SARs for BCL majors and 6000 SARS for LSW majors. The alternative job
with the option to extend starts with 15000 SARs for both major categories. This job goes down
in different increments until it reaches 3000 SARs for both major categories.
34
Free Child Care Services
The Saudi Labor Law support women with children by providing free child care services through
the Qurrah Program. Women working in the private sector with a salary below 8,000 SARs are
eligibleforthisservice. Iprovidepairsofjobswherethebaselinejobisonethatmakesthemeligible
for this service, and an alternative that does not provide free child care. As with the segregated
workplace valuation, the assumption here is that women would value child care and therefore I
elicit the valuations of a job with free child care services. The baseline job with free child care
offers 8000 SARs for BCL majors and 6000 SARS for LSW majors. The alternative job with no
child care with 8500 and 8100 SARs for BCL and LSW majors respectively. This job goes up in
different increments until it reaches 11500 SARs for BCL majors and 8750 for LSW majors.
Chapter Two: Results
In this section, I report the elicited valuations of each of the five non-wage characteristics described
inthedatasection. Ithenexplorehowthesevaluationsvarybytherespondents’educationalattain-
ment, general perception of social norms, and personal demographics such as mother’s education -
which may impact their flexibility towards non-conventional job attributes.
Elicited Valuations
Table 2.3 displays the valuations of each of the different non-wage job attributes. They are cate-
gorized by college and Column (6) shows the average valuation of the total sample of respondents.
The main survey that I used to elicit these valuations also asked about the respondents’ reservation
wage and likelihood for applying for a job after graduation. This table shows that on average,
the reservation wage of college seniors is around 8700 SARs, with Law and Social Work majors
requiring slightly higher wages to consider accepting a job offer. Interestingly, of the five non-wage
attributes, young women value part-time work the most, where they are willing to accept a job
that pays around 1380 SARs lower in order to have the flexibility of a part-time job. This accounts
35
for 15.9% of the average reservation wage. The second most valued attribute is having a segregated
workplace. Young women value segregated workplaces by around 921 SARs, which accounts for
10.5% of the average reservation wage. As for paid maternity leave, the respondents’ reported a
positive valuation for the option to extend paid maternity leave beyond 10 weeks, although this
valuation is low relative to the first two attributes at around 241 SARs or 2.7% of the average
reservation wage. The last remaining job attributes, remote work and free child care, were valued
negatively. Remote work was valued at -504 SARs, meaning that these respondents are willing to
accept a lower wage for an office job to avoid working from home. The most negative valuation was
giventofreechildcareservicesat-3142SARs. Thisresultissurprisingsinceitisgenerallyexpected
that women would prefer free child care services. There are three possible explanations here. First,
since the child care service is part of the Qurrah Program, these jobs could be perceived as less
prestigious and women may therefore prefer to work in a job that does not make them eligible for
this service. Second, multi-family homes are very common in Saudi Arabia and most women live
with several family members that could help with child care while they are at work, which makes
this service unnecessary. Lastly, in-house nannies are also very common in this context and most
families are able to hire an assistant to babysit the child while the mother is at work. Therefore,
while the idea of free child care seems like an attractive job attribute, these women may not find it
necessary. Figures 2.1, 2.2, 2.3, 2.4, and 2.5 display the distributions of these valuations.
Heterogeneity in Valuations
I investigate whether personal characteristics have differential impacts on job choices. Job choices
are categorized into two margins: extensive and intensive. On the extensive margin, I explore
the effects on likelihood to apply for a job after graduation and on reservation wage. On the
intensive margin, I explore the effects on the elicited valuations explained in the previous section.
The personal characteristics are chosen to understand how women’s family backgrounds, average
perception of social norms (question asked in the main study about how women perceive men’s
norms around their wives working), and educational attainment (GPA, type of high school) can
affect their flexibility around different work arrangements. Summary statistics are displayed in
Table 2.1 and results are summarized in Table 2.3.
36
In Columns (1) and (2), I find that women’s average perceptions (AP) of the norms around women
working is a significant factor affecting employment decisions on the extensive margin, along with
GPA. The results are mixed on the intensive margin. Having a mother with at least a college
education, having a driver’s license, going to a private high school, and having a higher perception
of social norms (AP) are all signs of a more progressive respondent. I find that these attributes
are negatively correlated with the valuations of a segregated workplace (Column (5)) and some are
positively correlated with the valuation of paid maternity leave (Column (6).
Counterfactuals
The second purpose of this paper is to try to understand the Saudi labor market frictions when
it comes to female employment. Female labor force participation remains low and most of the
literature aims at understanding the labor supply side. However, it is plausible that young women
are eager to work if certain job attributes were available, or if the compensating differentials were
better when their desired attributes are not provided. For example, it is costly to provide free
child care services for women, and I find that women actually negatively value this job attribute.
Resourcescouldbeallocatedelsewheretoprovideworkingconditionsthatwomenpreferinorderto
reduce these labor market frictions and raise female employment. The next step is to use historical
firm-level data to understand that nature of the jobs that were available to women and do some
counterfactual exercises to study how female employment could have been different had these firms
provided the desired attributes and/or the higher wages to compensate for the lack thereof. For
example, suppose a current employer allocates resources to providing full-time jobs to women at
a certain monthly wage. Given this paper’s findings, we know that women value part-time work
and are willing to accept a lower wage for the flexibility and convenience that a part-time job
provides. Such changes in the labor demand side can help meet women’s needs, increase their
employment, and reduce the market’s frictions. Another example is those companies who provide
(or are planning to provide) free child care services. Given my findings, this employer is better off
allocating resources to provide segregated workplaces for women rather than free child care since
this feature is the second most valued among young women and may also help employers cut costs
since providing a segregated workplace is a one-time fixed cost as opposed to child care services.
37
This exercise will quantify the cost=effectiveness of such changes and help determine how much
more female employment we can expect after implementing them.
Chapter Two: Conclusion
This paper employs an incentivized discrete choice experiment to elicit women’s valuations of dif-
ferent non-wage attributes in Saudi Arabia. I find that women value part-time work the most out
of the different characteristics studies and value free child care services the least. Perceptions of the
norms around women working affect the decision to apply for a job, whereas individual character-
istics affect valuations of different job characteristics. These results will be used in counterfactual
exercises to better understand the labor market frictions and how resources can be reallocated to
improve female employment. The findings of this paper are especially important for policymakers
as its insights show that active labor market policies are needed to bridge the current gap be-
tween female labor supply and demand. Incentivized elicitation is a powerful tool that may be
used in many other contexts. The strategies and methodologies employed in this paper may be
used and extended to understand other labor market challenges such as youth unemployment and
automation.
38
Chapter Two: Figures
Figure 2.1: Part-time Valuation
Figure 2.2: Remote Valuation
39
Figure 2.3: Segregated Workplace Valuation
Figure 2.4: Maternity Leave Valuation
40
Figure 2.5: Free Child Care Valuation
41
Chapter Two: Tables
Table 2.1: Descriptive Statistics
Mean SD Min Max N
AP -23.63 15.74 -68 24 513
GPA 4.54 0.55 2 5 525
Driver’s License 0.30 0.46 0 1 523
Public School 0.69 0.46 0 1 523
Family Income 18896.49 14010.31 1800 90000 427
Mother Worked 0.19 0.39 0 1 366
Mother’s Education 0.39 0.49 0 1 514
42
Table 2.2: Mean Valuations by College
(1) (2) (3) (4) (5) (6)
CBA CCS Languages Law CSW Total
Reservation Wage 6848.10 8758.43 8969.70 11352.94 9972.22 8741.78
(2780.83) (3784.02) (3472.84) (19536.88) (8670.10) (7802.64)
Part-time Work 627.36 2457.92 644.58 875.00 -52.63 1380.55
(4251.08) (3097.42) (3365.47) (4596.67) (4261.38) (3785.57)
Remote Work -870.00 51.98 -463.86 -1295.45 -1427.63 -504.07
(3627.65) (2166.23) (2973.24) (4047.42) (3991.08) (3104.75)
Segregated 1465.65 612.07 810.98 613.64 1052.63 921.11
(3375.36) (3592.35) (3448.60) (2788.81) (4458.20) (3522.84)
Maternity Leave -151.85 371.29 590.36 17.05 440.79 241.04
(2797.60) (2045.08) (1648.90) (3439.00) (2424.14) (2396.13)
Child Care -3503.88 -1592.04 -4268.67 -5652.27 -4750.00 -3142.42
(8867.65) (6554.41) (7885.68) (9042.06) (6986.23) (7800.47)
43
Table 2.3: Determinantes of Job Choices: Extensive vs Intensive Margin
(1) (2) (3) (4) (5) (6) (7)
Apply for Job Reservation Wage Part-time Work Remote Work Segregated Maternity Leave Child Care
AP 0.022
∗∗∗ 11.160 -4.646 -17.085
∗ 12.068 5.415 28.900
(0.006) (23.441) (10.764) (9.689) (11.568) (7.653) (21.379)
GPA -0.018 1141.704
∗∗ -517.884
∗ -175.045 390.106 -172.792 52.590
(0.153) (539.458) (280.661) (238.442) (254.300) (161.295) (675.009)
Driver’s License 0.158 -301.911 186.196 286.934 -830.578
∗∗ 323.056
∗∗ -2444.336
∗∗∗ (0.200) (680.425) (346.504) (253.569) (324.380) (161.361) (715.936)
Public School 0.307 477.561 57.248 287.285 768.985
∗∗ -68.037 2350.155
∗∗∗ (0.220) (726.783) (349.613) (290.001) (361.662) (190.410) (750.445)
Mother’s Education 0.073 -63.502 515.959 1042.529
∗∗∗ -107.267 593.333
∗∗∗ 584.606
(0.194) (1065.491) (327.290) (249.881) (322.332) (204.782) (727.802)
Y Mean 8.926 8765.000 1394.850 -491.753 926.566 239.394 -3202.869
Obs. 503.000 300.000 466.000 485.000 463.000 495.000 488.000
Standard errors in parentheses
Results shown with college fixed effects
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
44
Chapter 3
Evolving Gender Dynamics and Female Labor Outcomes: Evaluat-
ing the Female Driving Policy in Saudi Arabia
Chapter Three: Introduction
Gender inequality in developing countries has been an important issue for decades. As of 2022,
developing countries have accounted for most of the Global Gender Gap (GGG) measure (World
Economic Forum, 2022). Of the developing countries, the Middle East and North Africa (MENA)
ranks last, along with South Asia, making it a particularly important region for research. Studies
thus far have shown that inequality in education, nutrition, and health access have been major
contributors to the persistence of this gap. Evidence has shown that gender inequality in human
capital investments leads to inequality in social and economic outcomes, creating a cycle that
women find difficult to deviate from. However, most countries in the MENA region rank fairly well
in terms of equality in education and health. Female educational attainment is relatively high and
yet labor force participation rates remain as low as 30%. It is therefore interesting to understand
why populations of healthy, well-educated women still rank poorly in terms of economic and social
participation.
This question highlights a slightly different branch of gender research, which stresses the role that
societal norms and cultures play in feeding gender gaps. In many of the MENA countries, religion
and culture have shaped the way that societies function – from big economic decisions to day-to-
day household choices. Part of this culture is the heavy dependence women have on their male
guardians. Men are the breadwinners and the decision-makers. To many women, finding a man
45
to marry at an early age is a form of financial and social security. It has become the norm, such
that being single in the late 20’s and early 30’s is frowned upon. Such ideologies have created a
great lack of autonomy amongst most women. This is especially the case in the Gulf countries,
where women are identified by the men in their lives and are considered their subordinates. For
example, it is considered disrespectful to address a woman by her first name and she must be
formally referred to as the daughter, wife, or mother of her male guardian. Such norms create an
acceptance of male dominance and marginalize women’s roles, which is hypothesized to perpetuate
the lack of economic participation we see today.
This study exploits a natural experiment in Saudi Arabia to study whether giving women one of
their most basic human rights will affect their labor market decisions. Specifically, it investigates
the policy of allowing women to drive that was implemented in June 2018. This policy is seen
as one of the biggest steps to introduce women in Saudi Arabia to a new norm where they can
commute freely without any dependence on a male – whether it be a male guardian or a hired
private driver. In a way, this also frees many women from the need to ask for permission to leave
the house. Many men have abused the power they have had by trying to control where the women
in their lives are allowed to go and when - simply because they were the ones to drive them. This
policy was introduced with a goal to raise the low female labor participation rates, which have
been stagnant in Saudi Arabia for decades. It is crucial to note that the effects of this policy on
employment outcomes should be considered through two different channels. On one hand, it is
likely that simply having easier access to transportation would motivate women to work. By this,
we would conclude that the main reason unemployment rates are low is because women found it
difficult to commute and therefore chose not to work. On the other hand, a woman’s decision to
work could be affected by cultural norms and the custom of guardianship. If that is the case, then
the policy could be an introduction to a new gender dynamic that gradually shifts’ women’s - and
men’s - perspectives, leading to a sense of female independence and autonomy that affects social
andeconomicdecisions. Wewouldexpectashiftinlaborparticipationdecisionsaswellasachange
in social decisions such as age of marriage, number of children, and decisions to file for a divorce.
In this paper, I specifically explore the effect of the female driving policy on female employment
levels. This policy has gotten attention from the media all over the world, with a great expectation
46
that female employment levels will rise as a result. It is therefore interesting to explore whether
the inability to drive was indeed holding back Saudi women from joining the labor force, or if
there still remains a deeper underlying barrier that keeps women from realizing their full potential.
Since women in different regions of the country got access to drivers’ licenses at different times, a
difference-in-difference design was used to conclude that there is still no evidence that the driving
policy has causally affected female employment. This result was unexpected given that this policy
was clearly intended to boost female employment levels. However, it is a result that is quite
important for policy. Females having the right to drive is definitely a non-negotiable human right
andthispolicyhashadmanypositiveeffectsonSaudiwomen. However,therestillremainsastigma
about working women and in order to motivate women to work, policymakers should address this
issue differently. This supports previous research about female labor participation in Saudi Arabia.
Aloud et al. (2020) and Bursztyn et al. (2020) have shown that social norms and information gaps
about working women are the main reason why women still choose to stay unemployed.
This paper relates to several strands of literature. First, it complements research that has explored
“identity economics”. Akerlof and Kranton (2000) show how gender identity and sense of self
impact economic outcomes. Jensen and Oster (2009) show how the exposure to urban lifestyles
through cable television change rural women’s autonomy and affects their behaviors in India. Bassi
and Rasul (2017) explore the effect of papal visits on fertility-related beliefs and behaviors. The
secondtopicisculturalandsocietalnormsanditseffectongenderinequality. Jayachandran(2015)
has given a literature discussion of the role that cultural views play in explaining gender inequality
in developing countries. Goldin (2014) and Bertrand (2011) show how social norms impact female
labor force participation rates. Aloud et al. (2020) and Bursztyn et al. (2020) show the importance
of social norms and information gaps in explaining female labor force participation rates in Saudi
Arabia. Lastly, this paper also relates to the literature on gender economics. Bandiera et al. (2018)
show how a policy intervention of adolescent women in Uganda leads to higher engagement in self-
employment, less teenage pregnancies, and delayed marriages and childbearing. Barbanchon et al.
(2018) study the tradeoff between commute and wage and find that women’s maximum accepted
commute to work is much smaller than men – reflecting the constraints women might face when
making labor decisions.
47
This study’s contribution to the literature is two-fold. First, considering the fact that the MENA
region is one of the most conservative regions in the world, it is very well understood that female
decisions are heavily influenced by culture and religion. In some cases, gender equality policies will
only achieve the desired economic goals if paired with major social awareness about women’s roles
in the economy. Evaluating the effects of this new driving policy carves the path that new policies
should be taking. With Saudi Arabia being the most conservative and having female labor force
participation rates as low as 30%, it makes it a great context to study and would shed light on the
common phenomenon that we see in the region as a whole. Second, it is generally difficult to find
a causal effect of a gender-related policy at a nation-wide level in this region. This is particularly
because such policies are usually applied to the entire nation, creating a setting for correlational
studiesbutrarelyforcausalinference. Thispolicy, however, hasauniquesetofcircumstancessince
its implementation begins gradually due to several regional constraints. Its staggered adoption
(which we will assume to be exogenous) allows for a difference-in-difference estimation that will
help find the causal effect of evolving gender norms on the outcomes of interest.
Chapter Three: Background
TheWorldEconomicForum(2022)ranksSaudiArabia127/146basedonanoverallindexofgender
inequality. As it is the case in many other MENA countries, the country paradoxically ranks fairly
wellineducationalattainmentandhealth/survivaloutcomes. However,economicparticipationand
political empowerment still rank poorly. On the social side, around 40% of women are married by
theageof25,asopposedtoonly12%ofmen. Interestingly,accordingtotheArabBarometerSurvey
of 2011
1
, on average, 75% of males support female labor force participation. The culture however
has created a gender dynamic that reinforces women’s dependence on men - which perpetuates
behavior such as early marriage, childbearing, and the choice not to work outside the home. This
raises a noteworthy point – are women substituting work with marriage and wage income with
spouse income? This could be their solution since, given the circumstances, it is easier to get
married and financially depend on the spouse than to find a job in a setting that does not support
1
Princeton (2011)
48
the social norms.
UpuntilSeptember2017,womenwerenotallowedtodriveforseveralculturalandreligiousreasons,
including “women on the street distract men and will cause accidents” and “women driving cars
will cause traffic and will make it difficult for men to get to work”. This has made the trip to work
(or anywhere else) impossible without permission, a man to drive, or in many cases an expensive
hired driver. The announcement of the driving policy has caught the attention of the media all
over the world. Before this, Saudi Arabia was the only country in the world that banned women
from driving. After the ban was lifted, Saudi women began to see the tip of a new era that will
at least institutionally give them their rights and hopefully shape the new norm within society in
the years to come. Businesswomen Council leader Deena Al Faris was interviewed by The National
(UAE) in an article entitled the (2018) when this policy was first announced and after asking her
about what reform opponents would do she said “Because if the leader is with us - they don’t have
a chance to say no”. She speaks on behalf of Saudi women and her expression draws strength from
the fact that the leader is now “with them” and society must accept it.
Moreover, I interviewed a Saudi female journalist (Lulwa Shalhoub (2020)) to better understand
the context and how women perceived this new policy. She has shared her experience of getting
her license after the ban was lifted and gave insight on how Saudi women have reacted to this
policy. When asked about how driving may impact women’s decisions to work, she explained that
”having the ability to drive themselves to work makes it easier to pursue their professional lives
without having to rely on a male relative and/or a private driver. Hiring a driver is very expensive
especially for women who are on low pay. Other options, like Uber are also expensive on the long
run and unsustainable”. Also, ”There’s no reliable public transportation system in Saudi and so
women have to rely on drivers or drive themselves now. I do think it gives a sense of independence
and control over their the structure of their workdays”. Indeed, Harvard Kennedy School’s Policy
Brief (2018) has shown survey results and around 68% of women in Riyadh, the capital city of
Saudi Arabia, hire a private driver. Given that the average salary of such drivers is around USD
1000, it is highly expensive to get to places if it is not with a male guardian.
The policy reform was announced in September 2017 and was ordered to take effect in June 2018.
49
Womenwho are interested ingettinga driver’slicense arerequired togo through formaltrainingin
all-femaledrivingschools. Suchtrainingtakesabout3weeks ofdaily writtenandphysicalpractice.
However, having these all-female schools requires funding, space, and capacity to train the female
population. Such resources are not available in all 13 administrative regions of Saudi Arabia. As
of today, only 7 of the 13 regions have driving schools. Figure 3.1 displays a map of the regions,
where darker regions have a higher number of driving schools. This creates a natural experiment
where women in treated regions get the chance to drive before others.
It is critical to understand why some regions had schools before others to rule out any possible
confounding factors that would bias the results. The argument here is basically the availability of
resources and is not related to our outcomes of interest. In figures 3.2, 3.3, and 3.4, I show pre-
trends of employment outcomes of all 13 regions from the fourth quarter of 2016 until the second
quarter of 2018, which is when the policy was enforced. In Figure 3.2, total Saudi females working
as a percentage of the total employed population is shown for each region during the time period
before the policy was enforced.
ThefractionofemployedpersonsineachregionwhoareSaudifemalesseemstofollowaverysimilar
trend in all of Saudi Arabia, which gives an indication that even more urban cities such as Riyadh,
the capital city, had stagnant female employment levels. Therefore, any effect on employment that
may be found from the opening of the driving schools could safely be attributed to this treatment
and not to region-specific characteristics that could have been driving female employment. Figures
3.3 and 3.4 show the same pattern. Total Saudi females and total Saudi males employed are each
graphed individually and having these parallel trends before the treatment supports the exogeneity
ofhavingthedrivingschools. Thisassumptionwillbetestedmoreformallyinthefollowingsections.
Chapter Three: Data
The data used in the empirical estimation is collected from Saudi Arabia’s General Authority for
Statistics (2021) (GAS). The GAS has provided data on Saudi female employment levels as well
as Saudi male employment levels. All variables are collected quarterly and at the regional level
50
from the fourth quarter of 2016 until the fourth quarter of 2019. Female labor force participation
rates would have been the ideal outcome variable for this study because I am interested in women
joining the labor force (starting to look for a job and not necessarily getting hired) since it may be
too soon to see an effect on actual employment. Seeking a job and joining the labor force, however,
is likely to be impacted much sooner. However, female population of working age is required in
order to calculate female labor force participation rates as a percentage of female population (ages
15-64) as calculated by the International Labour Organization. This quarterly, region-specific data
was not available at the time this study was conducted
2
.
Then, to gather exact information on the location, number, and opening dates of female driving
schools, Twitterwasusedtosearchfortweetsabout”OpeningofschoolXinregionY”onaspecific
date. The timing of these schools is critical as it is the treatment and the main variable of interest.
Theschoolswerethencontactedtoconfirmtheseopeningdates. Table3.1showstheexactopening
date (and from that the quarter is recorded) of the first driving school in each treatment region
3
.
Chapter Three: Empirical Strategy
To estimate the effect that the opening of a driving school has on Saudi female employment levels,
I exploit the staggered timing of the opening of these schools across the Kingdom’s administrative
regions. I expect that if this policy is effective, then regions with earlier adoption of the policy
should have higher female employment outcomes as a result. Since different regions adopt the
policy at a different time, this setting is fit for a difference-in-difference model with heterogeneous
treatments. Recentliteraturehasshownthattheconventionaldifference-in-differencewithtwo-way
fixed effects models may not be the optimal choice in settings with staggered treatment adoption
(Callaway and Sant’Anna, 2021; Borusyak et al., 2023; Goodman-Bacon, 2021; Sun and Abraham,
2021). This is mainly due to the fact that when treatments are not homogeneous, these regressions
may incorrectly treat ”already-treated” groups as ”untreated” which essentially makes an a treated
2
Data collection for the periods after 2019 has paused since Covid-19 and more recent data has been requested
from the GAS.
3
In the case of Riyadh, the actual opening date of the first school was February 2018 but women began training
in June when the policy was enacted and therefore the training day was used as treatment.
51
groupacontrolgroupforthatspecificperiodandleadstobiasedtreatmenteffects. Specifically, the
approach introduced by Roth et al. (2022) highlights that estimated coefficients from a standard
TWFE model in this case may have opposite signs due to ”negative weighting problems”. This
issue is addressed by distinguishing between groups that are never treated versus groups that are
not-yet treated and assigning the relevant weights.
The identifying variation comes from across-region differences before and after the opening of the
driving schools. Assuming exogeneity of the driving schools in each region, the estimation equation
is as follows:
y
jt
=βDrivingSchool
jt
+γ j
+ϕ t
+ϵ jt
(3.1)
where y
jt
is the employment outcome variable in region j at year-quarter t,
DrivingSchool
jt
is the treatment variable, which is the opening of the first driving school in region
j and year-quarter t, γ j
is a region fixed effect and ϕ t
is a time fixed effect. Because the number of
regions is small (13), the regions were clustered using the wild cluster bootstrap method following
Roodman et al. (2019). In this case, the estimator is a weighted average of different causal effects
suchasregionsthatadoptfirstandregionsthatadoptlateron. Theyproposeanimprovedvariance
estimator to the standard DID variance estimator. A revised version of this paper is forthcoming
with changes to the methodology to incorporate the recent advances in difference-in-differences
(DD) estimation with staggered adoption with strategies that tackle the limitations associated
with two-way fixed effects estimation when treatment effects are heterogeneous. Specifically, the
approach introduced by Roth et al. (2022) highlights that estimated coefficients from a standard
TWFE model in this case may have opposite signs due to ”negative weighting problems”. This
issue is addressed by distinguishing between groups that are never treated versus groups that are
not-yet treated and assigning the relevant weights.
To further support this, a ”triple difference (DDD) design” was initially used in order to explore
whether the regions that open driving schools have affected male employment levels as well. Then
52
we can find the differential change that this policy has on female employment levels. This more
preciselyisolatestheeffectthatthepolicyhasonfemalesonly,whilealsoaccountingforanychanges
that may happen to male employment levels. The outcome variable used here is Saudi employment
tobetterexplorethepolicythatishappeningatdifferenttimes,indifferentregions,andsupposedly
affecting male and female employment levels differently. The equation estimated was as follows:
y
gjt
=β 1
DrivingSchool
jt
+β 2
DrivingSchool
jt
× 1(Gender =F)+γ gj
+ϕ gt
+ϵ gjt
(3.2)
where y
g
jt
is the employment level in region j at year-quarter t for gender g, DrivingSchool
jt
remains the treatment variable that is affecting different regions at different quarters and is mea-
suring the effect of driving schools on employment levels, and DrivingSchool
jt
× 1(Gender = F)
is the differential increase for female employment over male employment. However, with the new
estimation strategies of difference-in-difference with staggered adoption, I am limited to estimating
two separate regressions for men and women and comparing the estimates to infer the differential
effect.
It is important to mention the outliers shown in figures 3.3 and 3.4. Riyadh, Makkah, and the
Eastern Province generally have much higher female and male employment levels than the rest of
the regions. These regions are the largest and are more urbanized than the rest and therefore this
difference is expected. For that reason, the natural logarithm of these employment levels will also
be used in the estimation to dampen the effect of these outliers and get more precise estimates.
As previously mentioned, the identifying assumption here is that regions that get a driving school
would not have otherwise changed differently than regions that do not. A possible argument would
be the fact that more progressive or urban regions are likely to start training women earlier, which
means at baseline could have different effects. So, there remains the concern that there could be
region-specific policies or events that would also be influencing the opening of the driving schools
andcausingfemaleemploymenttochange. Toformallytesttheparalleltrendsassumptiondiscussed
previously, the following equation is estimated
4
:
4
Estimated using wild cluster bootstrap method at the region level.
53
y
jt
=β 0
DrivingSchool
t− 2
jt
+β 1
DrivingSchool
t− 1
jt
+βDrivingSchool
jt
+γ j
+ϕ t
+ϵ jt
(3.3)
where DrivingSchool
t− 2
jt
and DrivingSchool
t− 1
jt
are dummy variables equal to one if the quarter is
2 (or 1) quarters before the actual opening of the driving school in region j. This acts as a placebo
test and if the opening of the driving schools in each region is truly exogenous, we expect β 0
and
β 1
to be insignificant.
Chapter Three: Results
Female Employment Outcomes
To explore the effect that this policy has had on Saudi female employment levels, equation (1) was
estimated and the results are shown in Table 3.5. As explained previously, this is difference-in-
difference with heterogeneous treatments estimation and is clustered at the region level using the
wild cluster bootstrap method. It is shown that the opening of the driving schools in the treatment
regions has no significant effect on Saudi female employment levels. Column (1) shows the effect
on the total number of female employment and column (2) shows the effect on the log of female
employment to account for any outliers from larger regions. Since the opening of a driving school
was the proxy for the enforcement of the driving policy, we cannot conclude that the ability of
Saudi women to drive independently in the treated regions has had a statistically significant effect
onfemaleemploymentlevels. Figure3.5showstheAverageTreatmentEffectontheTreated(ATT)
of Equation (1), which visually confirms that there is little if any change after the treatment. Since
this method allows me to group the regions based on their year-quarter of treatment, Figure 3.6
shows a detailed image of how different groups were impacted by the policy. Each plot in the figure
represents the effect the policy has had on the regions that opened schools in the same quarter.
For example, the regions that opened driving schools in the 3rd quarter of 2018 show no effect post
treatment until around the 5th quarter though these results get less precise with time. Looking at
54
the remaining plots, it is clear that if there is any change post treatment, these changes are less
precise and therefore I cannot conclude that these results are significant. Figures 3.7 and 3.8 show
the same results for the log of Saudi Females Employed (Model 2) and these results are robust to
using logarithms. Using the same example as Model (1), the regions that were treated in the 3rd
quarter of 2019 show little to no effect and the results lose precision towards the 5th quarter after
the policy.
Triple Difference Estimation
Tofurtherinvestigatetheeffectofthispolicy,thetripledifferenceequationwasestimated. However
in order to incorporate the recent changes to the DID estimation, I estimate the effect on males
individually to observe whether this policy has had an effect on males and the possibility of having
a differential effect on females over males. As with the female results, Table 3.3 shows that this
policy has no significant effect on males either.
Threats to Identification: Parallel Trends Assumption Test
A main concern previously discussed, which may be a threat to identification in this study, is the
fact that the opening of driving schools may not entirely be exogenous as some regions could be
more urbanized or liberal than others. If so, these regions could be at an advantage and adopt
this policy earlier than others. Table 3.4 displays the results of equation (3), which formally tests
the parallel trends assumption and shows that it is sufficient to make this assumption. Although
the treatment variable ”Driving School Opening” is not significant, it shows that had these driving
schools opened in earlier quarters (2 quarters or 1 quarter earlier), this would not have had an
effect on female employment levels either. This result rules out the fact that some regions who
adoptthispolicyearlierandopentheirschoolssoonermayhaveotherfactorsaffectingemployment
levels. Figure 3.9 displays this graphically, showing that the treatment variable being 2 quarters
and 1 quarter earlier than the actual quarter of opening are closer to zero than the actual treat-
ment variable ”Driving School Opening”. However, all variables remain statistically insignificant.
Another possible strategy is to run a regression to understand determinants of having a driving
55
school first such as: average income per household, median age, average education per household,
ratio of female-to-male population, and population density. If any of these variables are significant
in determining the opening of a driving school, they would possibly be used in a set of controls in
the main regression
5
.
Chapter Three: Conclusion
In this paper, I use the opening of female driving schools as a measure for the gradual imple-
mentation of the female driving policy in Saudi Arabia. I find that so far, women’s ability to
drive does not have a causal effect on female employment levels. These results are robust across
different empirical designs and different employment outcomes. Thus, contrary to popular belief,
females not having the right to drive may not be root cause of the low female employment levels in
Saudi Arabia. This policy change nevertheless is a historic one and is a major step towards gender
equality in the Kingdom.
This study, however, does have its limitations. First, the policy change was implemented in June
2018 and is therefore considered fairly recent. With cultural norms being persistent, this policy
could indeed have an effect on female employment in the long-run. This time period shortcoming
is also magnified with the COVID-19 lockdowns that have stopped the openings of new schools for
around 2 years and has also slowed employment overall. Therefore, more time is needed to make
these results more compelling. Second, the use of female employment levels as opposed to female
laborparticipationratesisalsoalimitation. Womenaremorelikelytojointhelaborforceandbegin
searchingforjobsthantoactuallybeemployedinsuchashortperiodoftime. Third,take-upofthis
driving option is very important for the analysis. As Shalhoub (2020) mentioned in the interview,
”many women chose not to change their lifestyle and preferred to continue to have drivers. Some
are afraid of driving in general as driving in Jeddah is quite dangerous for example and needs a
significant level of confidence and road awareness. Others just enjoy the comfort of being served.”
Therefore, the results of this paper could be further strengthened by collecting individual-level
data on drivers’ licenses and the number of hired private drivers to examine women’s reactions to
5
These variables were requested but not received in time for this study.
56
this driving option as well as substitutability between hired private drivers and personal driving.
Moreover, despite the rapid improvements we see in gender equality, the custom of guardianship is
still prevalent in the Kingdom. This may prevent women from working even with easier access to
transportation. The results therefore support the literature on gender inequality and culture, but
further research is needed to understand the role that underlying cultural values and beliefs play
in female work decisions.
To delve into the root cause of the low female labor participation rates, surveys are also needed
to examine peoples’ perceptions of female employment before and after this policy change. This
new policy could have a social learning role, where people get more accustomed to women being
independent and more economically active. If that is the case, then this learning curve may be
steep since these cultural norms tend to persist over time. There could be some changes in peoples’
perceptions and attitudes but it may still be too soon to see changes in decisions and behavior,
especially since the policy change is relatively recent. With more data collected over time, a main
concern in interpreting the results could be addressed, which is the mechanism by which having a
school is affecting these outcomes (if there is an effect). Will it be autonomy or will it simply be
the fact that women are able to drive to work now? If it is autonomy, is it simply from: living in
a region that started training women, a feeling that is developed after being able to drive, or is it
a result of getting a job or going to school? These questions would be better understood through
heterogeneous analysis. Looking at women who already work, already in college, already married,
etc and how this policy affects their perceptions and behavior will shed light on the exact channel
that this policy is working through.
These surveys could also be useful in understanding what other factors could further empower
women and motivate them to work. As explained by Shalhoub (2020), ”I think seeing more women
in leading positions in Saudi Arabia nowadays encourages motivates more women to join the work-
force. OnlyrecentlydidSaudiappointfemaleambassadorsforexample. Rolemodelsareimportant
to push women forward and the more these are promoted and celebrated in the media, the more it
will become normalized and encourage men to support their female members of the family to work.
Also, the economic factor plays a role, as it is no longer enough for the man in the family to be the
soleprovider. BothhavetocontributetothehouseholdgiventheincreasedVATratesandthehigh
57
living expenses nowadays”. Therefore, collecting data on peoples’ opinions on these issues will be
vital in understanding what policies could positively affect female autonomy and decision-making.
Lastly, the decisions and behaviors that are expected to change as a result of this policy need
not only be work-related. Given that this policy is seen to be positive, social outcomes are also
expected to change and therefore would be an interesting extension to this paper. Giving women
the opportunity to be more independent and self-sufficient might have an effect on the decision to
get married at an early age for example. This is especially the case since now there are alternative
options with less restrictions than before such as working and going to college. By the same logic,
women may tend to initiate divorce more than before once they feel like they have an outside
option with this empowering reform. Fertility rates are also expected to decline, which is in line
withtheliteratureonautonomyandfertilityrates. Andfinally,womenareexpectedtogotocollege
or pursue higher education more than before. However, individual-level data is again required in
order to explore these social behaviors and whether this policy is impacting those decisions.
58
Chapter Three: Figures
Figure 3.1: Number of Driving Schools as of 2019 - Q4
59
Figure 3.2: Percentage of Saudi Females Working
Source: General Authority for Statistics - Kingdom of Saudi Arabia.
60
Figure 3.3: Total Saudi Females Working
Source: General Authority for Statistics - Kingdom of Saudi Arabia.
61
Figure 3.4: Total Saudi Males Working
Source: General Authority for Statistics - Kingdom of Saudi Arabia.
62
Figure 3.5: Difference-in-Difference Model (1)
63
Figure 3.6: Difference-in-Difference Model (1) by Groups of Treatment
64
Figure 3.7: Difference-in-Difference Model (2)
65
Figure 3.8: Difference-in-Difference Model (2) by Groups of Treatment
66
Figure 3.9: Parallel Trends Assumption Test
67
Chapter Three: Tables
Table 3.1: Opening of Female Driving Schools in Saudi Arabia
Region Opening Date of 1st School Number of Driving Schools Today
Riyadh June 2018 5
Makkah January 7, 2019 2
Madinah August 1, 2018 2
Qassim March 18, 2019 1
Eastern Province May 8, 2018 4
Asir - 0
Tabuk June 24, 2018 1
Hail - 0
North Border - 0
Jazan May 2018 1
Najran - 0
Al-Baha - 0
Al-Jouf - 0
68
Table 3.2: Difference-in-Difference Estimation
(1) (2)
Total Saudi Females Employed Log TSFE
ATT 1910.528 -0.003
(1391.761) (0.017)
Observations 169 169
Standard errors in parentheses
Wild cluster bootstrap at region level
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
69
Table 3.3: Difference-in-Difference Estimation - Males
(1) (2)
Total Saudi Males Employed Log TSME
ATT 150.528 -0.009
(2007.954) (0.017)
Observations 169 169
Standard errors in parentheses
Wild cluster bootstrap at region level
∗ p< 0.10,
∗∗ p< 0.05,
∗∗∗ p< 0.01
70
Table 3.4: Parallel Trends Assumption Test
(1) (2)
Total Saudi Females Employed Log TSFE
Driving School Opening (t-2) 2969.5 0.00949
(0.326) (0.521)
Driving School Opening (t-1) 2797.7 0.00844
(0.353) (0.622)
Driving School Opening 5542.6 0.0163
(0.2563) (0.5305)
Observations 169 169
P-values in parentheses, wild cluster bootstrap at region level.
71
References
(2018). Arab youth survey 2018: Overwhelming majority of young saudis support more women’s
rights. The National - UAE.
Akerlof, G. A. and Kranton, R. E. (2000). Economics and Identity. The Quarterly Journal of
Economics, 115(3).
Alfnes, F. and Rickersten, K. (2013). Non-market Valuation: Experimental Methods.
Aloud, M. E., Al-Rashood, S., Ganguli, I., and Zafar, B. (2020). Information and Social Norms:
Experimental Evidence on the Labor Market Aspirations of Saudi Women. Technical Report
w26693, National Bureau of Economic Research, Cambridge, MA.
Angrist, J. D. and Pischke, J.-S. (2008). Mostly Harmless Econometrics: An Empiricist’s Com-
panion. Princeton University Press.
Ashraf, N., Berry, J., and Shapiro, J. M. (2010). Can Higher Prices Stimulate Product Use?
Evidence from a Field Experiment in Zambia. American Economic Review, 100(5):2383–2413.
Bandiera, O., Buehren, N., Burgess, R., Goldstein, M., Gulesci, S., Rasul, I., and Sulaiman, M.
(2018). Women’s Empowerment in Action. Other papers.
Barbanchon, T.L., Rathelot, R., andRoulet, A.(2018). GenderDifferencesinJobSearch: Trading
off Commute Against Wage. page 84.
Bassi, V. and Rasul, I. (2017). Persuasion: A Case Study of Papal Influences on Fertility-Related
Beliefs and Behavior. American Economic Journal: Applied Economics, 9(4).
Belot,M.,Kircher,P.,andMuller,P.(2018).HowWageAnnouncementsAffectJobSearch-AField
Experiment. SSRN Scholarly Paper ID 3338629, Social Science Research Network, Rochester,
NY.
Bertrand, M. (2011). New Perspectives on Gender. In Card, D. and Ashenfelter, O., editors,
Handbook of Labor Economics, volume 4, pages 1543–1590. Elsevier.
72
Borusyak, K., Jaravel, X., and Spiess, J. (2023). Revisiting Event Study Designs: Robust and
Efficient Estimation. arXiv:2108.12419 [econ].
Brief, H. K. P. (2018). Allowing Women to Drive in Saudi Arabia May Reduce Cost of Travel.
Harvard University.
Bursztyn, L., Gonz´ alez, A. L., and Yanagizawa-Drott, D. (2020). Misperceived Social Norms:
Women Working Outside the Home in Saudi Arabia. American Economic Review, 110(10):2997–
3029.
Callaway, B. and Sant’Anna, P. H. (2021). Difference-in-Differences with multiple time periods.
Journal of Econometrics, 225(2):200–230.
Coffman, L.C., Featherstone, C.R., andKessler, J.B.(2017). CanSocialInformationAffectWhat
Job You Choose and Keep? American Economic Journal: Applied Economics, 9(1):96–117.
Dhar, D., Jain, T., and Jayachandran, S. (2018). Reshaping Adolescents’ Gender Attitudes: Evi-
dence from a School-Based Experiment in India. page 88.
Dhar, D., Jain, T., and Jayachandran, S. (2019). Intergenerational Transmission of Gender Atti-
tudes: Evidence from India. The Journal of Development Studies, 55(12):2572–2592.
Feld, B., Nagy, A., and Osman, A. (2022). What do jobseekers want? Comparing methods to
estimate reservation wages and the value of job attributes. Journal of Development Economics,
159:102978.
Fern´ andez, R. and Fogli, A. (2009). Culture: An Empirical Investigation of Beliefs, Work, and
Fertility. American Economic Journal: Macroeconomics, 1(1):146–177.
General Authority for Statistics, S. A. (2021). Labor Force.
Giuliano, P. (2020). Gender and Culture. Working Paper 27725, National Bureau of Economic
Research. Series: Working Paper Series.
Goldin, C. (2014). A Grand Gender Convergence: Its Last Chapter. American Economic Review,
104(4):1091–1119.
Goodman-Bacon, A. (2021). Difference-in-differences with variation in treatment timing. Journal
of Econometrics, 225(2):254–277.
73
Grewenig, E., Lergetporer, P., and Werner, K. (2020). Gender Norms and Labor-Supply Ex-
pectations: Experimental Evidence from Adolescents. CESifo Working Paper 8611, Center for
Economic Studies and Ifo Institute (CESifo), Munich.
Jayachandran, S. (2015). The Roots of Gender Inequality in Developing Countries. Annual Review
of Economics, 7(1):63–88.
Jayachandran, S. (2021). Social Norms as a Barrier to Women’s Employment in Developing Coun-
tries. IMF Economic Review.
Jensen, R. and Oster, E. (2009). The Power of TV: Cable Television and Women’s Status in India.
The Quarterly Journal of Economics, 124(3):1057–1094. Publisher: Oxford Academic.
Mas, A. and Pallais, A. (2017). Valuing Alternative Work Arrangements. American Economic
Review, 107(12):3722–3759.
Miller, C., Peck, J., and Seflek, M. (2019). Integration Costs and Missing Women in Firms.
Technical Report w26271, National Bureau of Economic Research, Cambridge, MA.
Princeton (2011). Arab Barometer Wave II – Arab Barometer.
Roodman, D., Nielsen, M. , MacKinnon, J. G., and Webb, M. D. (2019). Fast and wild: Bootstrap
inference in Stata using boottest. The Stata Journal: Promoting communications on statistics
and Stata, 19(1):4–60.
Roth, J., Sant’Anna, P. H., Bilinski, A., and Poe, J. (2022). What’s trending in difference-in-
differences? a synthesis of the recent econometrics literature. arXiv.
Shalhoub, L. (2020). Saudi female driving policy.
Sun, L. and Abraham, S. (2021). Estimating dynamic treatment effects in event studies with
heterogeneous treatment effects. Journal of Econometrics, 225(2):175–199.
Wiswall, M. and Zafar, B. (2017). Preference for the Workplace, Investment in Human
Capital, and Gender*. The Quarterly Journal of Economics, 133(1):457–507. eprint:
https://academic.oup.com/qje/article-pdf/133/1/457/30636559/qjx035.pdf.
World Economic Forum (2022). Global Gender Gap Report. Technical report, World Economic
Forum.
74

Abstract (if available)

Abstract This dissertation contributes to our understanding of low female labor force participation rates in developing countries - specifically the Middle East and North Africa (MENA) region. In the three chapters, I specifically study female labor force participation in Saudi Arabia and the barriers that women may face when making labor market decisions. In Chapter 11, I investigate how marriage market norms are misperceived by young women through a randomized controlled trial and I study how correcting these misperceptions affects their employment aspirations and behaviors. In Chapter 2, I employ an incentivized choice experiment to elicit young women’s valuations for different non- wage job characteristics to calculate compensating wage differentials that may imrove matching in the labor market. In Chapter 3, I exploit the staggered adoption of the female driving policy in Saudi Arabia as a natural experiment to understand the impact of women’s independence and work accessibility on their employment outcomes.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on applied microeconomics

PDF

Three essays on the microeconometric analysis of the labor market

PDF

Essays in development economics

PDF

Essays in labor economics: demographic determinants of labor supply

PDF

Essays on family and labor economics

PDF

Essays on development economics and adolescent behavior

PDF

Essays on political economy and corruption

PDF

The impact of minimum wage on labor market dynamics in Germany

PDF

Essays in environmental economics

PDF

Three essays on human capital and family economics

PDF

Essays on wellbeing disparities in the United States and their social determinants

PDF

Essays in financial economics

PDF

Essays on environmental economics: education, employment, and experiments

PDF

Against the wind: labor force participation of women in Iran

PDF

Essays on the platform design and information structure in the digital economy

PDF

Three essays on health & aging

PDF

Three essays on macro and labor finance

PDF

Essays on education programs in Costa Rica

PDF

Sustaining open source software production: an empirical analysis through the lens of microeconomics

PDF

Essays on development and health economics: social media and education policy

Asset Metadata

Creator Al Rakhis, Monira (author)

Core Title Essays in applied microeconomics

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Degree Conferral Date 2023-08

Publication Date 02/08/2025

Defense Date 08/08/2023

Publisher University of Southern California. Libraries (digital)

Tag applied microeconomics,female labor force participation,labor markets,OAI-PMH Harvest

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Bassi, Vittorio (committee chair), Duquette, Nicolas (committee member), Nugent, Jeffrey (committee member), Weaver, Jeffrey (committee member)

Creator Email alrakhis@usc.edu,monira.alrakhis@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113296577

Unique identifier UC113296577

Identifier etd-AlRakhisMo-12207.pdf (filename)

Legacy Identifier etd-AlRakhisMo-12207

Document Type Dissertation

Rights Al Rakhis, Monira

Internet Media Type application/pdf

Type texts

Source 20230808-usctheses-batch-1080 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA

Repository Email cisadmin@lib.usc.edu